The Bacteriome of Bat Flies (Nycteribiidae) from the Malagasy Region: a Community Shaped by Host Ecology, Bacterial Transmission Mode, and Host-Vector Specificity

Abstract

The Nycteribiidae are obligate blood-sucking Diptera (Hippoboscoidea) flies that parasitize bats. Depending on species, these wingless flies exhibit either high specialism or generalism toward their hosts, which may in turn have important consequences in terms of their associated microbial community structure. Bats have been hypothesized to be reservoirs of numerous infectious agents, some of which have recently emerged in human populations. Thus, bat flies may be important in the epidemiology and transmission of some of these bat-borne infectious diseases, acting either directly as arthropod vectors or indirectly by shaping pathogen communities among bat populations. In addition, bat flies commonly have associations with heritable bacterial endosymbionts that inhabit insect cells and depend on maternal transmission through egg cytoplasm to ensure their transmission. Some of these heritable bacteria are likely obligate mutualists required to support bat fly development, but others are facultative symbionts with unknown effects. Here, we present bacterial community profiles that were obtained from seven bat fly species, representing five genera, parasitizing bats from the Malagasy region. The observed bacterial diversity includes Rickettsia, Wolbachia, and several Arsenophonus-like organisms, as well as other members of the Enterobacteriales and a widespread association of Bartonella bacteria from bat flies of all five genera. Using the well-described host specificity of these flies and data on community structure from selected bacterial taxa with either vertical or horizontal transmission, we show that host/vector specificity and transmission mode are important drivers of bacterial community structure.

INTRODUCTION

Bats are increasingly recognized as natural reservoirs of a large number of emerging infectious agents (1–4). It is thus implicit that vectors of bat-borne disease will play important roles in the epidemiology and dynamics of infectious agents that can eventually emerge in human populations. Further, bats are hosts to different ectoparasites, including mites, fleas, ticks, and bat flies (5, 6). Bat flies (Diptera: Hippoboscoidea) are obligate blood-sucking parasites that are classically divided into two families—the Streblidae and the Nycteribiidae (7). Together, the Hippoboscidae (louse or ked flies), Streblidae, and Nycteribiidae are referred to as the Pupipara sensu stricto due to their adenotrophic viviparity, where all larval developmental stages occur within the adult female's body and the larva are nourished by milk glands until they are ready to pupate. This particularity of the Pupipara sensu stricto is thought to promote vertical parasite transmission, thus influencing the epidemiological role of these vectors in disease transmission.

To date, studies of microorganisms associated with nycteribiids have mainly focused on two groups of bacteria—Bartonella spp. (8–11) and Arsenophonus-like organisms (referred to here as ALOs) (12–15). These bacterial genera offer contrasting model systems for investigating the biotic factors driving the structures of associated microbial communities. Bartonella species are parasitic intracellular bacteria that infect vertebrate erythrocytes, and many species are considered to be zoonotic (16) and are linked to disease in humans (17–20), including the recently identified bat-associated Bartonella mayotimonensis (21). Nycteribiids are known to host a wide variety of Bartonella spp. (10), and contrasting patterns of Bartonella-bat host associations have been described across Africa (8, 9, 22), South America (23–25), Europe (26), and Asia (27).

In contrast, ALOs are symbiotic organisms that are engaged in complex interactions with a variety of arthropod species (28–30). The group making up the ALOs is a monophyletic lineage that includes different genera, such as Arsenophonus, Aschnera, Riesia, and other unnamed groups (ALO-1 to ALO-3) (12–15). In bat flies, ALOs are primary and obligate endosymbionts that are vertically transmitted from infected mothers to offspring due to the presence of bacteriocytes in the milk glands of all nycteribiid species (12–15). It is thought that ALOs may play a role in the provisioning of B vitamins, which are deficient in vertebrate blood, the only food source for these flies (13). The ubiquity of ALOs in bat flies, in which infection is at fixation, corroborates the hypothesis of an obligate endosymbiont (12–15). Such patterns have been found in other exclusive blood-feeding species like bedbugs (31) and tsetse flies (32), two insect groups which rely on a single food source throughout their developmental cycle and harbor beneficial microbes that provide nutrients absent from their restricted diets.

Nycteribiid species show various levels of bat host specificity. For example, a recent study reported a lack of host specificity and genetic structure in Cyclopodia horsfieldi, found on several fruit bats of the genus Pteropus (Pteropodidae) in southeastern Asia (33). On islands in the western Indian Ocean, it has been observed that members of the genera Penicillidia and Nycteribia are relatively promiscuous with respect to their host associations, exchanging freely between several Miniopterus spp. (Miniopteridae) and Myotis goudoti (Vespertilionidae) (5). Interestingly, these insectivorous bat species are often found living in syntopy (in physical contact within day roosts), which likely facilitates vector sharing (34). The picture is rather different for fruit bat flies belonging to the genus Eucampsipoda, which have been observed to uniquely parasitize pteropodids of the genus Rousettus on continental Africa or on islands in the southwestern Indian Ocean (5) despite these bats often sharing caves as day roost sites with other bat species (35). Similarly, Eidolon fruit bats are infested exclusively by flies from the genus Cyclopodia, with Cyclopodia dubia and Cyclopodia greeffi specifically parasitizing Malagasy Eidolon dupreanum and African Eidolon helvum, respectively. Thus, bat fly communities in Madagascar, neighboring islands in the Comoros archipelago, and continental Africa are composed of species that either are specialists (e.g., Eucampsipoda madagascarensis, Eucampsipoda theodori, Eucampsipoda africana, Cyclopodia dubia, and Cyclopodia greeffi found strictly on Rousettus madagascariensis, Rousettus obliviosus, Rousettus aegyptiacus, Eidolon dupreanum, and Eidolon helvum, respectively) or parasitize a broader range of hosts, such as Penicillidia leptothrinax and Nycteribia stylidiopsis, which are found on several Malagasy Miniopterus spp. as well as on Myotis goudoti. Under these interesting biological circumstances, variations in the patterns of infectious agents that transfer between mammalian and arthropod hosts are likely determined by a number of factors, including the specificity of host-vector interactions, the nature of vector-microorganism interactions (ranging from strict parasitism to mutualism), the transmission mode, and the confounding factors of host ecology (Fig. 1).

FIG 1.

Host specificity between bats and bat flies (Nycteribiidae) in Madagascar and the nearby Comoros archipelago (5).

Here, in order to investigate the associations between bacteria and nycteribiids, we have obtained bacterial 16S gene pyrosequencing data from bat flies of the genera Eucampsipoda, Penicillidia, Nycteribia, Cyclopodia, and Basilia (Paracyclopodia), which were taken from bats sampled in the western Indian Ocean region (Madagascar and the Union of the Comoros). We identify a number of previously undescribed bacterial associations within the Nycteribiidae, compare the phylogenetic relationships of some of these novel bacterial taxa, and propose a model for how host-vector-parasite interactions may shape the bacterial communities hosted by these flies.

MATERIALS AND METHODS

Taxon sampling and identification.

Bats were captured using mist nets and harp traps, generally placed at cave entrances, or butterfly nets to obtain individuals from day roost sites in caves. This study was conducted in strict accordance with the terms of research permits issued by national authorities (Direction du Système des Aires Protégées, Direction environment et des Forêts, and Madagascar National Parks [Madagascar] and Centre National de Documentation et de Recherche Scientifique [Union of the Comoros]), following the laws of these countries, and the associated research permit numbers are listed in Acknowledgments. Upon capture, each individual cataloged bat was placed in a separate clean cloth bag until the collection of relevant biological data, and they were examined for ectoparasites. The ectoparasites were collected with forceps and stored in separate vials containing 70% ethanol. Morphological identification of the ectoparasites was carried out using published keys and descriptions (36), and host bat specimens are housed in the Field Museum of Natural History (FMNH) (Chicago, IL) and in the Département de Biologie Animale, Université d'Antananarivo (UADBA) (Antananarivo, Madagascar). A considerable portion of the nycteribiid specimens used here came from a previous study that evaluated the evolutionary origins of bat flies on the Comoro Islands and Madagascar (5). New samples of Cyclopodia dubia and Basilia (Paracyclopodia) sp. were collected on Madagascar from Eidolon dupreanum and Scotophilus spp., respectively, as part of this study.

Nucleic acid extraction and pyrosequencing design.

Nycteribiid flies were removed from ethanol storage and dried and were then crushed using 2- by 2-mm tungsten beads in a TissueLyser (Qiagen). Nucleic acids were extracted using a Qiagen EZ1 robot with the DNA tissue kit, according to the manufacturer's protocol and as previously published (5). Nucleic acids were pooled by species for 16S gene pyrosequencing as detailed in Table S1 in the supplemental material.

The 16S gene pyrosequencing was performed as previously described (37). Briefly, 16S V3 and V4 variable regions were amplified via PCR-specific primers targeting the upstream and downstream regions of the V3 to V4 segment; the 3′ end of forward (TACGGRAGGCAGCAG) and reverse (GGACTACCAGGGTATCTAAT) bacterium-specific primers were bound at the 5′ end by multiplex identifier (MID) tags, a GS FLX key, and GS FLX adapters. The quantity of each PCR product was then determined with PicoGreen, and all products were mixed together in equimolar concentrations before 454 GS FLX sequencing (Genoscreen).

All reads of <250 bp in length were discarded using the Geneious Pro software package (38). Remaining reads were analyzed using the SILVA online next-generation sequencing (NGS) tool (www.arb-silva.de/ngs). Raw sequence reads were aligned with a gap extension penalty of 2 and a gap penalty of 5. Reads were filtered based on the following quality criteria: minimum length, 250 bp; minimum quality score, 30; maximum percent ambiguities, 1%; minimum base pair score, 30; and maximum percent repetitive, 2%. Remaining reads were clustered into operational taxonomic units (OTUs) at a threshold sequence identity of 99%. OTUs were classified by BLAST score comparison against the SILVA rRNA database version 115, with a classification similarity threshold of 93%. OTUs were used exclusively for initial taxonomic interpretations, and all quantitative data were calculated from the numbers of original sequence reads. The statistical significance of differences in proportions of sequence data was calculated by chi-squared comparisons with a Yates correction.

Sequence data and genetic analyses.

Specific gene fragments were amplified using primer pairs detailed in Table 1. Sequence data were analyzed using the Geneious Pro software package (38). All final sequence data were deposited in GenBank, and the relevant accession numbers are listed in Table 2. Coding sequences were aligned in-frame using the Geneious Pro translation align tool and the standard ClustalW cost matrix. Noncoding sequences were aligned using MUSCLE (39). Network analyses were performed in SplitsTree4 (40) using the neighbor-net algorithm. Maximum likelihood phylogenies were generated through RAxML v8.0.0 (41) using the RAxML GUI (v1.3) with 1,000 bootstrap replicates and the GTR+Gamma substitution model as suggested by JModelTest2 (v.2.1.6) (42). Coding sequence alignments were partitioned by base position, and single or multiple outgroup sequences were specified prior to analysis. The best tree outputs from RAxML were used as topology inputs in the R software package, exploiting the ape (43) and vegan (44) packages in order to test evolutionary congruence using the ParaFit algorithm (45). Bacterial phylogenies used for ParaFit analysis contained only sequences obtained in this study and were compared against a host-nycteribiid phylogeny generated based on cytochrome oxidase I (COI) sequence data. COI sequences were taken from reference 5, and data from Cyclopodia dubia and Basilia (Paracyclopodia) sp. were provided by B. Ramasindrazana and S. M. Goodman (unpublished data).

TABLE 1.

Primers used in this study

Description	Forward primer	Reverse primer	Reference/source
16S gene pyrosequencing	TAC GGR AGG CAG CAG	GGA CTA CCA GGG TAT CTA AT	Genoscreen
16S gene, 1,350 bp	AGA GTT TGA TCM TGG CTC AG	TAC GGY TAC CTT GTT ACG ACT T	Unpublished data (P. Tortosa)
Bartonella gltA	GGG GAC CAG CTC ATG GTG G	AAT GCA AAA AGA ACA GTA AAC A	23
Bartonella 16S	AGA GTT TGA TCM TGG CTC AG	TAC GGY TAC CTT GTT ACG ACT T	15
Bartonella rpoB	CGC ATT GGC TTA CTT CGT ATG	GTA GAC TGA TTA GAA CGC TG	62
Enterobacteriales 16S	GGG TTG TAA AGT ACT TTC AGT CGT	CCT YTA TCT CTA AAG GMT TCG CTG GAT G	51
Rickettsia gltA	GGT TTT ATG TCT ACT GCT TCK TG	CAT TTC TTT CCA TTG TGC CAT C	63
Wolbachia wsp	GTC CAA TAR STG ATG ARG AAA C	CYG CAC CAA YAG YRC TRT AAA	49
Wolbachia hcpA	GAA ATA RCA GTT GCT GCA AA	GAA AGT YRA GCA AGY TCT G	48
Wolbachia fbpA	GCT GCT CCR CTT GGY WTG AT	CCR CCA GAR AAA AYY ACT ATT C	48
Wolbachia ftsZ	ATY ATG GAR CAT ATA AAR GAT AG	TCR AGY AAT GGA TTR GAT AT	48

TABLE 2.

GenBank accession numbers for sequences generated as part of this study^a

IDb no.	Species	Host species	Host ID no.	Country	Bartonella gltA	Enterobacteriales 16S	Rickettsia		Wolbachia
							gltA	16S	wsp	fbpA	ftsZ	hcpA
SC1	Basilia (Paracyclopodia) sp.	Scotophilus marovaza	FMNH 221393	Madagascar	KT751157	KT751105			KT751166	KT751134	KT751140	KT751163
SC2	Basilia (Paracyclopodia) sp.	Scotophilus marovaza	FMNH 221393	Madagascar		KT751104
SC3	Basilia (Paracyclopodia) sp.	Scotophilus marovaza	FMNH 221393	Madagascar		KT751103			KT751167	KT751135
SC4	Basilia (Paracyclopodia) sp.	Scotophilus marovaza	FMNH 221393	Madagascar		KT751102
1b	Basilia (Paracyclopodia) sp.	Scotophilus robustus	FMNH 209163	Madagascar	KT751147	KT751101
1	Cyclopodia dubia	Eidolon dupreanum	FMNH 221295	Madagascar	KT751146	KT751130
2	Cyclopodia dubia	Eidolon dupreanum	SMG 17826/FMNH 221296	Madagascar		KT751129
3	Cyclopodia dubia	Eidolon dupreanum	SMG 17827/FMNH 221297	Madagascar		KT751128
4	Cyclopodia dubia	Eidolon dupreanum	SMG 17828/FMNH 221298	Madagascar	KT751148	KT751127			KT751164	KT751132	KT751138	KT751161
5	Cyclopodia dubia	Eidolon dupreanum	SMG 17830/UADBA 32975	Madagascar	KT751149	KT751126
6	Cyclopodia dubia	Eidolon dupreanum	SMG 17831/UADBA 32976	Madagascar	KT751150	KT751125			KT751165	KT751133	KT751139	KT751162
7	Cyclopodia dubia	Eidolon dupreanum	SMG 17832/UADBA 32977	Madagascar	KT751151	KT751124
8	Cyclopodia dubia	Eidolon dupreanum	SMG 17833/UADBA 32978	Madagascar		KT751123
9	Cyclopodia dubia	Eidolon dupreanum	SMG 17834/UADBA 32979	Madagascar		KT751122
1c	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17698/FMNH 221366	Madagascar		KT751120
3b	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17776/UADBA 32967	Madagascar		KT751119
2c	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17777/UADBA 32968	Madagascar			KT751158
10	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17780/UADBA 32971	Madagascar		KT751118
6c	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17783/UADBA 32974	Madagascar			KT751159
5c	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17781/UADBA 32972	Madagascar				KT751131
14c	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 17931/UADBA 33647	Madagascar		KT751121
J50	Eucampsipoda madagascarensis	Rousettus madagascariensis	SMG 16259/FMNH 209105	Madagascar		KT751117
J13	Eucampsipoda theodori	Rousettus obliviosus	SMG16715/FMNH 220041	Union of the Comoros		KT751116
J14	Eucampsipoda theodori	Rousettus obliviosus	SMG 16715/FMNH 220041	Union of the Comoros		KT751115
J22	Eucampsipoda theodori	Rousettus obliviosus	SMG 16751/FMNH 220043	Union of the Comoros	KT751156
GR16	Nycteribia stylidiopsis	Miniopterus gleni	SMG 17841/UADBA 33030	Madagascar	KT751152	KT751108
GR17	Nycteribia stylidiopsis	Miniopterus gleni	SMG 17841/UADBA 33030	Madagascar		KT751107
GR15	Nycteribia stylidiopsis	Miniopterus gleni	SMG 17841/UADBA 33030	Madagascar		KT751109
GR12	Nycteribia stylidiopsis	Miniopterus gleni	SMG 17846/FMNH 221333	Madagascar		KT751110
GR6	Nycteribia stylidiopsis	Miniopterus griveaudi	SMG 17654/UADBA 32960	Madagascar
GR8	Nycteribia stylidiopsis	Miniopterus griveaudi	SMG 17682/UADBA 32962	Madagascar		KT751113
GR9	Nycteribia stylidiopsis	Miniopterus griveaudi	SMG 17689/FMNH 221413	Madagascar		KT751111, KT751112
2b	Nycteribia stylidiopsis	Miniopterus griveaudi	SMG 17759/FMNH 221343	Madagascar		KT751114
GR19	Nycteribia stylidiopsis	Miniopterus griveaudi	SMG 17761/UADBA 33004	Madagascar		KT751106
11b	Penicillidia leptothrinax	Miniopterus griveaudi	SMG17860/FMNH 221354	Madagascar		KT751097
11c	Penicillidia leptothrinax	Miniopterus aelleni	SMG 17922/FMNH 221440	Madagascar	KT751143
GR20	Penicillidia leptothrinax	Miniopterus griveaudi	SMG 17757/FMNH 221341	Madagascar	KT751153
GR18	Penicillidia leptothrinax	Miniopterus griveaudi	SMG 17772/UADBA 33015	Madagascar		KT751093, KT751094			KT751171
J35	Penicillidia leptothrinax	Miniopterus petersoni	SMG16806/FMNH 209186	Madagascar		KT751092
J38	Penicillidia leptothrinax	Miniopterus petersoni	SMG 16806/FMNH 209186	Madagascar			KT751160
12b	Penicillidia leptothrinax	Miniopterus sp.	SMG 17865/UADBA 33032	Madagascar		KT751096
12	Penicillidia leptothrinax	Miniopterus sp.	SMG 17866/UADBA 33033	Madagascar	KT751144
13	Penicillidia leptothrinax	Miniopterus cf. ambohitrensis	SMG17867/UADBA 33034	Madagascar		KT751095
13b	Penicillidia leptothrinax	Miniopterus sp.	SMG 17866/UADBA 33033	Madagascar	KT751145				KT751169	KT751136	KT751141
18b	Penicillidia leptothrinax	Miniopterus sp.	SMG 17884/FMNH 221429	Madagascar					KT751170	KT751137	KT751142
GR21	Penicillidia sp. (cf. fulvida)	Miniopterus griveaudi	SMG 17753/FMNH 221337	Madagascar	KT751154	KT751098, KT751099
9b	Penicillidia sp. (cf. fulvida)	Miniopterus sp.	SMG 17878/FMNH 221423	Madagascar		KT751100

^aGeographical, arthropod, and bat host information is also presented, along with laboratory and museum identifier codes for the host mammal.

^bID, identification.

Nucleotide sequence accession numbers.

All PCR-generated sequences in this study were submitted to GenBank under the accession numbers given in Table 2.

RESULTS

Bacterial community composition as revealed by 16S gene pyrosequencing.

Pyrosequencing data were obtained from seven independent samples that originated from seven nycteribiid species: Eucampsipoda madagascarensis, Eucampsipoda theodori, Penicillidia sp. cf. fulvida (a Penicillidia sp. that looks like Penicillidia fulvida), Penicillidia leptothrinax, Nycteribia stylidiopsis, Cyclopodia dubia, and Basilia (Paracyclopodia) sp. Samples contained pooled DNA that was extracted from different numbers of nycteribiid flies. A summary of sample composition, as well as the statistics of pyrosequencing data acquisition, including numbers of reads obtained and quality control results, is presented in Table S1 in the supplemental material.

We then analyzed bacterial diversity at the class level. The bacterial communities associated with all bat fly samples were dominated by three bacterial phyla, Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria, which represented 17%, 3%, and 78% of total sequences, respectively. The vast majority of Betaproteobacteria were observed in Eucampsipoda theodori and showed the closest similarity to uncultured bacteria of the family Neisseriaceae. The majority of Neisseriaceae are common commensal bacteria that are often associated with mammals, but Neisseria gonorrhoeae and Neisseria meningitides are the known causative agents of gonorrhea and meningococcal meningitis in humans, respectively.

Of the sequences identified as Alphaproteobacteria, 55% came from Wolbachia, 26% came from Bartonella, and 17% came from Rickettsia. Nearly all Gammaproteobacteria (>99%) came from the order Enterobacteriales, and for all bat fly species but one (Cyclopodia dubia), they were mostly distributed between two genera, Arsenophonus and Aschnera, which clearly belong to the ALO group (12–15). The case of Cyclopodia dubia is subtly different; on the basis of a BLAST search, most of the bacterial sequences from this bat fly species were found to have identity to Enterobacteriales members other than ALOs, such as Providencia, Erwinia, and Dickeya. Interestingly, the bacteria originating from Cyclopodia dubia may actually form a new Enterobacteriales group (see below).

Bacterial community data are summarized in Fig. S1 in the supplemental material.

We then further investigated bacterial taxa by sequencing additional loci. Taxa were selected because of their possible medical importance, such as the genera Rickettsia and Bartonella because of their relation to pathogenic strains causing rickettsioses and bartonellosis, respectively, or because they were common to multiple species of nycteribiid fly and represented a significant proportion of the total number of reads obtained by pyrosequencing, such as Wolbachia and Enterobacteriales (threshold retrospectively set at 2.5% of the total number of reads).

Rickettsia.

During pyrosequencing, 16S sequences closely related to Rickettsia were obtained from Eucampsipoda madagascarensis and Penicillidia leptothrinax. These sequences represented 5.5% and 15.3% of reads in Eucampsipoda madagascarensis and Penicillidia leptothrinax, respectively. The gltA locus of Rickettsia was amplified from Eucampsipoda madagascarensis and Penicillidia leptothrinax samples. Classification using BLASTx showed that the obtained gltA sequences showed little homology to known species of Rickettsia, with the strongest match showing 90% identity to gltA from Rickettsia raoultii, a bacterium identified in hard ticks from China (46). In addition, a 1,350-bp fragment of the 16S gene of Rickettsia was amplified from Eucampsipoda madagascarensis. Although the phylogeny of Rickettsia could only be poorly resolved based on these data (not shown), sequence similarity was strongest with Rickettsia gravesii, Rickettsia barbariae, and Rickettsia tarasevichiae, bacteria previously shown to be associated with hard ticks belonging to the genera Amblyomma and Ixodes.

Wolbachia.

Pyrosequencing of the 16S locus revealed that Wolbachia was present in all sampled fly species except Eucampsipoda madagascariensis and Eucampsipoda theodori. The percentage of reads matching Wolbachia ranged from <1% in Nycteribia stylidiopsis to nearly 40% in Basilia (Paracyclopodia) sp. PCR-based screening using the wsp system (47) was then used to test individual nycteribiid samples for the presence of Wolbachia. In accordance with the pyrosequencing results, Wolbachia was identified in Penicillidia leptothrinax (14/23; 61%), Penicillidia sp. (cf. fulvida) (1/1; 100%), Cyclopodia dubia (6/9; 67%), and Basilia (Paracyclopodia) sp. (5/5; 100%).

A selection of DNA fragments from the Wolbachia wsp, fbpA, ftsZ, and hcpA genes (48) was amplified from Cyclopodia dubia, Basilia (Paracyclopodia) sp., and Penicillidia leptothrinax. Sequence data were additionally obtained from the wsp gene of Wolbachia from Penicillidia sp. (cf. fulvida), showing 100% identity to those sequences obtained from Wolbachia infecting Penicillidia leptothrinax; however, attempts to amplify other loci were unsuccessful for these samples. As has been observed elsewhere (49, 50), network analysis of wsp gene data against reference sequences from the Wolbachia multilocus sequence typing (MLST) database (http://pubmlst.org/wolbachia/) suggested high levels of diversification and recombination between Wolbachia spp. from well-defined phylogenetic groups (see Fig. S2 in the supplemental material), and thus, the wsp gene data were abandoned for phylogenetic characterization of the sequences obtained from our bat fly samples.

Network and phylogenetic analyses of concatenated sequences (fbpA, ftsZ, and hcpA) suggested that all identified bat fly Wolbachia species belong to the F subgroup (Fig. 2). Sequence types from different nycteribiid species were grouped by host species and were paraphyletic, sharing identity with Wolbachia strains from multiple host types and geographical origins, which is suggestive of independent evolutionary histories and multiple introductions of Wolbachia into bat fly populations.

FIG 2.

Wolbachia network analysis and group F phylogeny. The presented analyses are based on concatenation of three gene loci (fbpA::ftsZ::hcpA); reference data were acquired from the Wolbachia MLST database (http://pubmlst.org/wolbachia/). The presented network structure was generated from aligned sequence data in SplitsTree4 using the neighbor-net algorithm. The phylogenetic tree was produced using RAxML, and bootstrap support from 1,000 replicates is indicated by dots of different sizes on each internal node. Sequences from this study are highlighted in red, orange, and blue, representing host origins from Cyclopodia dubia, Basilia (Paracyclopodia) sp., and Penicillidia leptothrinax, respectively. Asterisks indicate that the hcpA region was missing from the sequence data originating from P. leptothrinax. However, it should be noted that identical, but less well supported, topologies were generated when using only fbpA and ftsZ (data not shown).

Arsenophonus-like organisms.

The Enterobacteriales were by far the most predominant bacterial taxa identified in pyrosequencing analyses, with 73% of all sequencing data identified as either ALOs or as a new and unnamed Enterobacteriales group. Further 16S data were obtained from multiple samples of each nycteribiid species. Network analysis was performed using reference strains listed in Table S2 in the supplemental material. Nycteribiid endosymbionts belonging to the Enterobacteriales fell into three distinct groups (Fig. 3). Insectivorous bat-associated Nycteribia and Penicillidia possessed ALOs associated with the Aschnera subgroup, whereas fruit bat-associated Eucampsipoda and insectivorous bat-associated Basilia (Paracyclopodia) sp. possessed ALOs belonging to the Arsenophonus subgroup. Interestingly, network analysis suggested that the endosymbionts from Cyclopodia dubia form their own, new group within the Enterobacteriales family, which is closely related to the known ALO members (Cyclopodia group in Fig. 3).

FIG 3.

Enterobacteriales network diagram generated from 16S gene sequence data. The presented network structure was generated from aligned sequence data in SplitsTree4 using the neighbor-net algorithm. The Cyclopodia group contained only sequences from this study and refers to a newly identified taxon of the Enterobacteriales, all originating from C. dubia. Sequences used to generate this representation can be found in Tables S3 and S4 in the supplemental material. All sequences from bat flies in this study are indicated by colored circles, as identified in the key.

Bartonella.

Bartonella sequences were detected by pyrosequencing in all bat fly species. Eucampsipoda theodori, the only bat fly in this study from the Union of the Comoros and parasitizing Rousettus obliviosus, showed significantly higher proportions of sequences from Bartonella (P < 0.0001) than those of its sister species Eucampsipoda madagascarensis from Madagascar occurring on Rousettus madagascariensis. The percentage of reads derived from Bartonella ranged from <1% in Basilia (Paracyclopodia) sp., Penicillidia leptothrinax, and Nycteribia stylidiopsis to 17.4% in Eucampsipoda theodori.

The gltA locus was amplified from these nycteribiid samples. Only 15 sequences that demonstrated clear identity to unambiguously annotated Bartonella strains in BLAST searches and that were monophyletic (with Bartonella reference sequences in preliminary phylogenetic studies) were retained for analysis. Bartonella sequences were obtained from all bat fly species except Eucampsipoda madagascarensis. Additional Bartonella-specific amplifications based on the 16S and rpoB loci were also attempted but were largely unsuccessful.

The gltA sequences obtained from the nycteribiids of Madagascar and the Comoros archipelago were seen to cluster into five distinct groups (Fig. 4). Potential limitations of the gltA locus have been discussed elsewhere (10).

FIG 4.

Bartonella phylogeny based on the gltA locus, generated using 1,000 bootstrap replicates in RAxML. Dots on internal nodes represent bootstrap support of >0.75. Sequences from this study, and similar bat- and bat fly-associated sequences from other studies, are grouped into monophyletic groups by color. Bats drawn on leaves indicate sequences that originated from bats or bat fly specimens. The key at the bottom of the figure details bat fly and host bat origins of the associated sequences. Abbreviations used for bat flies are as follows: Cy., Cyclopodia; Eu., Eucampsipoda; P, Penicillidia; N, Nycteribia. Abbreviations used for bats are as follows: Ei., Eidolon; Mi., Miniopterus; My., Myotis; R., Rousettus; Sc., Scotophilus. Numbers in parentheses represent the numbers of identical sequences that were grouped into operational taxonomic units in RAxML.

Evolutionary congruence.

The global-fit method, ParaFit, was used to test the hypothesis of evolutionary congruence between nycteribiids and the bacteria detected in this study, by comparing phylogenetic structures to those obtained using COI data from bat flies. ALOs demonstrated a significant level of congruence (P < 0.001), whereas congruence was not significant for Bartonella (Table 3). As the topological structure may result from a difference in levels of polymorphism of the chosen markers (16S for ALOs and gltA for Bartonella), we compared their nucleotide diversities (pi value) using DnaSP v5 (http://www.ub.edu/dnasp/). This revealed that 16S was actually less polymorphic than gltA (pi, 0.056 and 0.117 for 16S and gltA, respectively), thus allowing us to reject the hypothesis of an artifactual topology of ALOs resulting from higher resolution of the 16S locus. It was not possible to test Wolbachia or Rickettsia lineages using this methodology due to the limited number of relevant bacterial taxa identified in nycteribiid hosts.

TABLE 3.

Evolutionary congruence between Nycteribiidae and associated bacteria^a

Parameter	Wolbachia	Arsenophonus-like organisms	Bartonella
ParaFit Global	NAb	0.087	0.060
P value	NA	0.001	0.172

^aCongruence testing for Wolbachia was not possible due to the presence of only three independent taxa.

^bNA, not available.

DISCUSSION

Here, we provide characterization of the bacterial microbiota associated with numerous species of Nycteribiidae bat flies occurring in Madagascar and the nearby Comoros archipelago and provide more detailed taxonomic information for some bacterial taxa of particular interest.

Bacterial communities associated with the nycteribiid bat fly genera Eucampsipoda, Penicillidia, Nycteribia, Cyclopodia, and Basilia (Paracyclopodia) were described by pyrosequencing of the 16S locus. Variation was found in community composition in different species of Nycteribiidae; however, these may be due to stochastic infection effects, differences in sample pooling between species, and/or limits in sequencing repeatability. It should also be noted that taxon sampling was subject to amplification bias due to the primers used to amplify the 16S locus (see Table S1 in the supplemental material). Although we cannot quantify the significance of these effects, our data suggest that the vast majority of bacterial diversity in all samples can be described by the presence of not more than two bacterial classes—the Alphaproteobacteria and the Gammaproteobacteria. While the ALOs (Gammaproteobacteria) have been previously described as obligate, primary symbionts in bat flies (12–15), the nature of the interactions between other bacteria and bat flies remains unknown. The overall homogeneity of the bacterial community structure across nycteribiid species, independent of bat host species, suggests that infection is not merely opportunistic, but selected and commonly occurring taxa likely form beneficial, positively selected interactions with their arthropod hosts. On the other hand, strongly significant differences between samples from Madagascar and the Comoros archipelago (Betaproteobacteria in Eucampsipoda theodori) suggest that environmental variables will also determine which interactions can be sustained.

Our results along with previous studies (12–15) corroborate the widespread presence of ALOs in bat flies while identifying a likely new genus of endosymbiont of the Enterobacteriales that infects the fruit bat (Eidolon dupreanum)-associated nycteribiid Cyclopodia dubia. While ALOs, and especially Arsenophonus, are globally common symbionts estimated to infect ca. 5% of insect species (28, 51), we further confirm that bat flies harbor the highest diversity of Enterobacteriales strains reported to date (Arsenophonus, Aschnera, and the Cyclopodia-associated group). We also found congruent patterns of codivergence between bat flies and ALOs, a result likely due to the association between members of the Nycteribiinae subfamily (which includes Penicillidia leptothrinax, Penicillidia sp. [cf. fulvida], and Nycteribia stylidiopsis) and Aschnera, which are known to have codiverged over a long evolutionary period (8, 26). However, the evolutionary history of these endosymbionts in bat flies is complex; all strains found here do not cluster within a specific bat fly clade but rather exhibit distinct evolutionary origins showing that they underwent repeated horizontal transfer between distantly related host species. This pattern is well illustrated by the presence, in some bat fly species, of Arsenophonus strains, which are closely related to strains from very diverse insect species, such as aphids and parasitoid wasps. Two distinct evolutionary strategies are thus acting on ALO endosymbiosis; Aschnera is highly specialized with regard to its hosts, with ancient acquisition followed by codiversification, while Arsenophonus is more generalist and acquired through recent horizontal transfers. The presence of a previously unidentified group of bacteria in Cyclopodia dubia testifies to the diverse nature of symbiotic interactions, and further work will be required to better understand how these findings modify our understanding of the evolutionary history of nycteribiid-endosymbiont interactions.

The genus Wolbachia was observed to infect Penicillidia spp., Cyclopodia dubia, and Basilia (Paracyclopodia) sp. This common reproductive manipulator has previously been identified in bat flies (13); however, it is of note that we observe Wolbachia in multiple nycteribiid species. Interestingly, all Wolbachia species that were identified in bat flies belonged to the F supergroup, an emerging supergroup of Wolbachia that has been associated with a broad spectrum of arthropod hosts, including the orders Scorpiones, Blattodea, Coleoptera, Hemiptera, Isoptera, Neuroptera, Orthoptera, Phthiraptera, Thysanoptera (52), and Diptera (the order containing the Nycteribiidae), which are more commonly associated with supergroups A and B (51, 53). Eucampsipoda did not harbor Wolbachia, and only a subset of bat flies infesting insectivorous bats tested positive for Wolbachia, showing that, unlike the ALOs, Wolbachia is not an obligate endosymbiont of nycteribiid flies. Although the number of samples that could be sequenced in greater detail was relatively limited, nycteribiid-Wolbachia sequences appeared to be species specific but paraphyletic through MLST analyses, suggesting multiple introductions of independent lineages to different nycteribiid species. However, a common lineage of Wolbachia was observed in Penicillidia leptothrinax and Penicillidia sp. (cf. fulvida). Facultative association between nycteribiid Wolbachia suggests that their interaction over evolutionary history is more recent than that of the ALOs despite the fact that fluorescence in situ hybridization-based physiological studies suggest a common mechanism of vertical transmission for all nycteribiid bacterial endosymbionts (13). This is in keeping with observations in other arthropod species, where Wolbachia strains are seen to have complex phylogenetic histories due to their dual capacity to form stable coexisting mixed populations within arthropod communities (54, 55) and to invade entire populations by manipulating host reproduction (56). While the potential of Wolbachia to manipulate the reproduction of bat flies is unknown, it is interesting to note that a recent study observed biased sex ratios in bat fly populations (Trichobius frequens) in Puerto Rico (57).

The data generated in this study also add to the knowledge of the increasing diversity of Bartonella subgroups that are associated with bat flies and their bat hosts. It is generally thought that Bartonella species demonstrate specificity to mammalian hosts; however, the exact role of vector-host specificity in establishing host-specific interactions is not fully understood (58). Here, the detection of Bartonella in all tested bat fly species adds credence to the theory that Bartonella and bat flies may form mutually beneficial, positively selected interactions, which may drive host specificity. However, the lack of evolutionary congruence between Bartonella and the Nycteribiidae suggests that any selective benefit is insufficient to generate stable symbiosis despite their likely ability to be transferred vertically from mother to pupa. In the case of the insectivorous bat-infesting Nycteribiidae of Madagascar and the Comoros, transmission through blood-feeding is likely to play a direct role in the horizontal transfer of Bartonella, as multiple nycteribiid flies (i.e., Nycteribia spp. and Penicillidia spp.) feed on the same bat hosts (5). This was seen to be the case here as, for example, a single clade in our phylogenetic analysis (colored orange in Fig. 4) contained Bartonella variants from Nycteribia and Penicillidia as well as from five different bat species. However, the same phylogenetic analysis also suggested exchange between Cyclopodia dubia and Basilia (Paracyclopodia) sp. (group colored blue in Fig. 4), taxa which are not known to interact directly or with the same bat host species. This suggests that other direct or indirect mechanisms may promote exchange between hosts and thus drive the intraspecific diversity of Bartonella observed within the Nycteribiidae. This is in contrast with other members of the Pupipara sensu stricto; ked flies, blood-sucking vectors of ungulates, are thought to form strict interactions with Bartonella, which have resulted in host-specific associations (10). The difference between these two situations is likely not related to vector host specificity, as members of the family Hippoboscidae are highly adapted to their respective hosts (59). Instead, we can imagine that the ecology of different bat hosts may promote the diversity of Bartonella within the Nycteribiidae, where host populations that more frequently come into contact with other hosts undergo more frequent horizontal exchange. This may help explain global Bartonella diversity as well as the previously reported regional differences in host specificity between Bartonella and bats (see the introduction).

Overall, a relatively low diversity of bacteria was seen to be associated with the studied bat flies from Madagascar and the Comoros, with the principal infecting taxa being the Enterobacteriales, Bartonella, and Wolbachia. Various patterns of intraspecific genetic diversity were observed between these models, which we explain by differences in the nature of the bacterium-nycteribiid endosymbiotic relationship, transmission mode, host-vector specificity, and variations in the associated bat host ecology (Fig. 5). The transmission modes of these three bacterial taxa vary; while all three are thought to be transferred vertically from mother to pupa during adenotrophic viviparity, only Bartonella may be transferred in the blood of the bat hosts. Bartonella species, where the highest levels of intraspecific genetic diversity were observed, thus have the opportunity for direct horizontal transfer due to bat fly host promiscuity and may also be transferred by bat-host interactions that are independent of these ectoparasites, such as with other blood-feeding vectors (e.g., Streblidae, mites, or fleas) or habitat overlap with other hosts. The frequency of horizontal transfer of Bartonella is thus likely to be influenced by ecological factors, such as the breeding seasonality of bats as well as their migration patterns, geographical distribution, and colony population size and structure. In contrast, the Enterobacteriales form strict, primary endosymbiotic relationships with their nycteribiid hosts, but evidence of horizontal transfer between hosts that infest the same bat species may be observed, suggesting that proximity likely drives the frequency of horizontal endosymbiont exchange.

FIG 5.

Factors affecting speciation patterns and monospecific microorganism diversity for different host-associated microorganisms.

In conclusion, while the global bacterial community structure associated with the Nycteribiidae is likely primarily driven by the establishment of mutually beneficial relationships between microorganism and host, the origins and evolution of these associations are complex. Additionally, host-associated microorganisms are known to interact (60), and bacterial community structure in insects may be influenced positively by cooperation or negatively by competition or exclusion and in turn may affect the biology of their hosts (reviewed in references 61 and 12). Bartonella species appear to be the only bacteria with known pathogenic potential that form strict relationships with the tested Nycteribiidae bat flies, suggesting that these arthropods may be a true reservoir of Bartonella infection. The presence of other likely pathogenic bacteria such as Rickettsia is only anecdotal and may even remain detrimental to the health of these arthropods. However, these unique vectors remain fascinating from an epidemiological point of view due to the diverse nature of the interactions that they form, especially in tropical settings.

Funding Statement

European Commission (EC) provided funding to David Arthur Wilkinson, Colette Cordonin, Yann Gomard, Beza Ramasindrazana, and Pablo Tortosa under the grant RunEmerge.