Extensive Screen for Bacterial Endosymbionts

FEMS Microbiology Ecology, 91, 2015, fiv047

doi: 10.1093/femsec/fiv047 Advance Access Publication Date: 27 April 2015 Research Article

RESEARCH ARTICLE Extensive screen for bacterial endosymbionts reveals taxon-specific distribution patterns among bees (Hymenoptera, Anthophila) Michael Gerth1,∗, Abiya Saeed2, Jennifer A. White2 and Christoph Bleidorn1,3

1Molecular Evolution and Systematics of Animals, Institute for Biology, University of Leipzig, Talstrasse 33, D-04103 Leipzig, Germany, 2Department of Entomology, S-225 Agricultural Science Center North, University of Kentucky, Lexington, KY 40546-0091, USA and 3German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Deutscher Platz 5d, D-04103 Leipzig, Germany ∗ Corresponding author: Molecular Evolution and Systematics of Animals, Institute for Biology, University of Leipzig, Talstrasse 33, D-04103 Leipzig, Germany. Tel: +49 341 9736743; E-mail: [email protected] One sentence summary: By sampling a large number of species of a local bee fauna, we find that host phylogeny predicts the absence and presence of bacterial endosymbionts best. Editor: Julie Olson

ABSTRACT

Bacterial endosymbionts play key roles in arthropod biology, ranging from beneficial mutualists to parasitic sex ratio manipulators. The number of described endosymbiotic bacterial taxa has accumulated continuously in recent years. While the understanding of arthropod–microbe interactions has advanced significantly, especially in model organisms, relatively little is known about symbiont distribution and effects in non-model organisms. As a first step to alleviate this gap in understanding, we performed an endosymbiont survey in bees (Anthophila), an ecologically and economically important group of hymenopterans. To this end, we sampled 170 bee species and screened by PCR for the presence of Wolbachia, Rickettsia, Arsenophonus and Cardinium. Detected strains were then further diagnosed by additional markers. Additionally, we tested if certain ecological traits, bee phylogeny or geographic origin of bees explain endosymbiont distribution. Our results indicate that supergroup A Wolbachia are very common in bees and that their distribution can be significantly correlated to both host ecology and phylogeny, although a distinction of these factors is not possible. Furthermore, bees from the same region (Old World or New World) are more likely to harbour identical Wolbachia strains than expected by chance. Other endosymbionts (Rickettsia, Arsenophonus) were less common, and specific to particular host taxa, suggesting that host phylogeny is a major predictor for endosymbiont distribution in bees.

Keywords: Wolbachia; Arsenophonus; Rickettsia; Cardinium; PCR screen

INTRODUCTION tion from parasitoids (Oliver et al. 2003; Xie et al. 2013), viruses (Hedges et al. 2008; Teixeira, Ferreira and Ashburner 2008)or Intracellular bacterial endosymbionts of arthropods may have toxins (Kikuchi et al. 2012). In addition to evidently beneficial profound impacts on many aspects of their hosts’ biology. effects, some bacteria are known as ‘reproductive parasites’ of Conspicuous examples include metabolic provisioning (Douglas arthropods. Typically, these bacteria alter their hosts’ reproduc- 1998; Akman et al. 2002; Hosokawa et al. 2010)aswellasprotec- tion, often to expedite their own spread. Strains of the genus

1 2 FEMS Microbiology Ecology, 2015, Vol. 91, No. 6

Wolbachia (α-Proteobacteria) are probably the most common re- et al. 2012, 2014). In the present study, we focused on four of productive manipulators in arthropods, infecting approximately the most common intracellular endosymbionts in insects: Wol- 40% of all terrestrial species (Zug and Hammerstein 2012). Phe- bachia, Rickettsia, Arsenophonus and Cardinium. We surveyed 170 notypes induced by Wolbachia include cytoplasmic incompatibil- species that make up about one third of the German bee fauna ities, feminization, male-killing and parthenogenesis (Werren, for the presence of these bacteria and further classified the Baldo and Clark 2008). Similar effects have been attributed to strains by sequencing multiple loci. Furthermore, multilocus se- other symbionts as well, e.g. Arsenophonus sp. (γ -Proteobacteria), quence typing (MLST) data from German and USbees were used Rickettsia sp. (α-Proteobacteria), Cardinium sp. (Bacteroidetes) and to assess geographical differentiation of Wolbachia strains. The Spiroplasma sp. (Mollicutes) (Gherna et al. 1991;Werrenet al. 1994; depth of sampling allows us to infer universal distribution pat- Williamson et al. 1999; von der Schulenburg et al. 2001; Weeks, terns of endosymbionts among bees and broadens our knowl- Marec and Breeuwer 2001; Zchori-Fein et al. 2001). Often, how- edge of endosymbiont prevalence in arthropods in general. ever, symbiont–host relationships are complex, and a categori- cal separation into mutualistic or parasitic lifestyles is not possible (Gill, Darby and Makepeace 2014; Zug and Hammerstein MATERIALS AND METHODS 2015). Insect collection and DNA extraction In arthropods, Wolbachia has been detected across many insect orders (Ros et al. 2009), in spiders (Baldo et al. 2008), Bees were collected between 2009 and 2013 from sites all over scorpions (Baldo et al. 2007) and in some crustacean species Germany. The sampling was designed to reflect the taxonomic (Cordaux et al. 2012). Previous Wolbachia screens revealed some diversity of bees in Germany and thus included species of all major lineages (termed ‘supergroups’ A and B) to be ubiquitously native families and most genera (Table S1, Supporting Infor- spread, with some evidence for host ecology and geographical mation). When available, multiple individuals per species were distance influencing strain distribution (Jeyaprakash and Hoy collected. Notably, this sampling design may result in underesti- 2000; Russell et al. 2009; Stahlhut et al. 2010). Other arthropod mations, especially for endosymbionts with generally low preva- endosymbionts seem to be more unevenly distributed: while lence, such as Arsenophonus, Rickettsia and Cardinium (Duron et al. generally much rarer than Wolbachia, some host taxa appear 2008; Russell et al. 2012). Furthermore, a meta-analysis con- to be ‘hotspots’ for the diversity of certain endosymbiotic lin- cluded that Wolbachia is usually distributed in ‘most-or-few’ in- eages, e.g. spiders for Cardinium (Duron et al. 2008), coccinellid dividuals of a species (Hilgenboecker et al. 2008), suggesting that beetles for Rickettsia (Weinert et al. 2009) or flies of the taxon our sampling of few individuals per species might have missed Hippoboscoidea for Arsenophonus (Novakov´ a,´ Hypsaˇ and Moran low prevalence of Wolbachia (or other endosymbiont) infections 2009). However, this could also reflect sampling biases in ex- in some species. Altogether, 330 individuals of 170 bee species isting endosymbiont screens. Often, these screens covered few were sampled and used in subsequent endosymbiont screens, species from a large taxonomic spectrum (Werren, Windsor and representing about one third of the German bee fauna (Table Guo 1995; Jeyaprakash and Hoy 2000; Duron et al. 2008)and S1, Supporting Information). A number of these samples had consequently, only few insect taxa can be considered to be ex- been used in prior studies investigating Wolbachia in bees: 75 haustively screened, e.g. ants (Russell 2012; Russell et al. 2012). species were screened for Wolbachia in Gerth, Geißler and Blei- Nevertheless, such data are needed as a first step in understand- dorn (2011) and also included in the present dataset. An ad- ing how and at what level bacterial endosymbionts interact with ditional 38 species were taken from Gerth, Rothe¨ and Bleidorn their hosts, especially in non-model arthropod taxa. (2013), while 57 species included in this study were so far not in- Here, we sampled a large proportion of a local bee fauna vestigated for the presence of Wolbachia. None of the specimens to subsequently perform an endosymbiont screen. Bees (An- had previously been screened for any symbionts other than Wol- thophila) are a group of aculeate Hymenoptera, with seven cur- bachia. We extracted DNA from muscle tissue dissected from bee rently recognized families (Michener 2007). Many aspects of bee thoraces by proteinase K digestion and subsequent chloroform biology are areas of intense research, such as their diversity (Car- extraction. It should be noted that screening muscle tissue in- dinal and Danforth 2013) and the evolution of sociality (Brady stead of reproductive organs may result in missing the presence et al. 2006; Cardinal and Danforth 2011). Bee ecology and evo- of endosymbionts. However, bacterial symbionts were shown to lution is tightly linked to that of flowering plants, as pollen be widely distributed in many host tissues (Saridaki and Bourtzis is the main source of protein for bees (Westrich 1989). Conse- 2010; Faria and Sucena 2013). Furthermore, our approach was quently, they act as important pollinators (Greenleaf and Kre- used to detect Wolbachia in bees before (Gerth, Geißler and Blei- men 2006; Garibaldi et al. 2013) and therefore impact human dorn 2011;Gerth,Rothe¨ and Bleidorn 2013) and did not seem wellbeing (Losey and Vaughan 2006). Bee species differ in their to impact the detection of the symbionts encountered in the pollen preferences, and while some species are generalists, oth- present screen. The success of the DNA extractions was tested ers need pollen of a single or a few closely related plant species by a PCR of the single-copy gene long-wavelength rhodopsin as de- for development (Praz, Muller¨ and Dorn 2008;Sedivy,Muller¨ and scribed in Danforth et al. (2004). The remains of the insects were Dorn 2011). In addition to pollen, bees require specific nesting retained in 95% ethanol as vouchers. substrates within a certain foraging range of their pollen source In addition, 13 US bee species were sampled between August (Zurbuchen et al. 2010). These ecological requirements, in addi- 2011 and May 2013 (Table S1, Supporting Information). US spec- tion to pronounced differences in phenology, limit the potential imens were collected free-hand, with nets, and with bee bowls for physical contact of different species as a prerequisite for lat- in and around Lexington, Kentucky. Individuals were placed in eral symbiont transfer. We therefore tested if bee ecology is also 95% ethanol and stored at −20◦C until identification and DNA reflected in their endosymbiont composition. extraction. Investigations of microbe–bee interactions have hitherto fo- Individual bees were surface sterilized using a 5% bleach so- cused mainly on the honeybee Apis mellifera, arguably the eco- lution (for 60 s), followed by three 95% ethanol rinses (60 s) and nomically most important bee species (Hamdi et al. 2011; Lan- finally a DI water rinse (60 s). DNA was extracted from the ab- gridge 2014), or on the microbiome of sweat bees (McFrederick domen of each specimen using DNeasy kits (Qiagen) following Gerth et al. 3

manufacturer’s instructions with a 3 h incubation time. Extrac- version 1.0 (Kuck¨ and Meusemann 2010). We determined the tion quality was evaluated by screening for positive DNA pres- best partitioning scheme (16S + first codon positions of protein- ence through CO1 (mitochondrial cytochrome oxidase subunit 1) coding genes, second codon positions, third codon positions) or EF1-α (elongation factor 1-alpha). and nucleotide substitution models (GTR + + I for all parti- tions) for all subsequent phylogenetic analyses with Partition- Endosymbiont PCR screens Finder version 1.1.1 (Lanfear et al. 2012). We used RAxML for a combined ML search and bootstrapping with 1000 pseudorepli- All German specimens were screened by PCRs for the presence cates. Bayesian inference of Rickettsia phylogeny was performed of four of the most common bacterial endosymbionts in in- with MrBayes version 3.2.2 (Ronquist and Huelsenbeck 2003). sects: Wolbachia, Rickettsia, Arsenophonus and Cardinium. Specific Four Metropolis-coupled Monte Carlo Markov Chains were run primers as well as appropriate positive and negative controls two times for 5 million generations with a sampling frequency were used in all PCRs (Table S2, Supporting Information). We of 100. After convergence of the two runs was diagnosed by a used ftsZ (cell division protein) primers from Baldo et al. (2006) standard deviation of split frequencies below 5%, we discarded to screen for the presence of Wolbachia. For the detection of Rick- the first 12 500 trees of the sample as burnin. A majority rulecon- ettsia, Arsenophonus and Cardinium, we performed nested PCRs. sensus tree was calculated from the remaining trees; posterior First, we targeted the entire 16S rRNA gene with universal bacte- probabilities were derived from the frequency of bipartitions. rial primers. In the second step, we employed nested 16S primers For Arsenophonus, 16S sequences from bee hosts were aligned specific for each of the endosymbionts, using the PCR product with sequences from NCBI GenBank (Fig. S2, Supporting Infor- of the first reaction as a template (Table S2, Supporting Infor- mation) with SSU-ALIGN. We included Arsenophonus strains iso- mation). Endosymbionts were then further characterized by se- lated from the bees A. mellifera and Megachila rotundata. For the quencing partial atpA (ATP synthase F1 alpha subunit) and coxA latter, we used 16S amplicon data from McFrederick, Mueller (cytochrome C oxidase subunit 1) for Rickettsia and partial groEL and James (2014). After downloading the reads from NCBI se- (GroEL protein) for Arsenophonus. All obtained PCR products were quence read archive (accession number SRR826752), trimming sequenced using the Sanger technique in both forward and re- was performed automatically with QTrim (Shrestha et al. 2014) verse directions by GATC Biotech AG (Konstanz, Germany). We using default settings. The reads were then mapped onto the only considered bacterial symbionts to be present in the sample sequence of their closest match (as determined in McFreder- when at least one of the PCR products could be sequenced suc- ick, Mueller and James 2014) in NCBI GenBank (DQ115536) us- cessfully. Accession numbers of Wolbachia ftsZ sequences can be ing segemehl 0.2 (Hoffmann et al. 2009). Finally, a majority rule found in Table S1 (Supporting Information). Sequences of other consensus sequence of all mapped reads was included into the endosymbionts were submitted to NCBI GenBank under acces- 16S alignment. Phylogenetic analyses were then performed with sion numbers KP183239–KP183262. RAxML as described above. Because phylogenetic estimates de- For German Hylaeus hosts and the 13 US bee species, five rived from 16S datasets are often unreliable in Arsenophonus and MLST loci for Wolbachia were sequenced for each Wolbachia strain related taxa (Novakov´ a,´ Hypsaˇ and Moran 2009), we additionally (see Table S2, Supporting Information for details on all em- sequenced groEL of Arsenophonus from both of the two clusters ployed primers and Table S4, Supporting Information for ac- of bee Arsenophonus recovered in the 16S analysis. Therefore, we cession numbers). American samples were sequenced by AGTC used MAFFT (L-INS-i) to align the obtained sequences with the (Lexington, KY, USA) or Beckman-Coulter (Danvers, MA, USA). groEL dataset of Duron et al. (2014) including 51 Arsenophonus like organisms (ALO) and 5 outgroup taxa. Next, we performed ML Phylogenetic analyses as described above as well as NeighborNet analyses using Split- sTree version 4.12.6 (Huson and Bryant 2006). All sequences were manually checked for quality and assembled to contigs with CLC Main Workbench 6.9 (CLC bio, ˚ Arhus, Denmark) or with Geneious 6.0.6 (BioMatters Ltd., Testing for predictors of Wolbachia infections NZ). Wolbachia ftsZ sequences from bee hosts were aligned and masked by using a template from the MLST website Most bee species are solitary, i.e. single females construct a nest (http://pubmlst.org/wolbachia/). To enable supergroup assign- and provide their offspring with pollen. Therefore, key ecologi- ment, we included two reference ftsZ alleles each from super- cal requirements for all bees are nesting substrates and pollen groups A, B and F. The best-fitting nucleotide model was deter- sources, both of which must meet certain criteria that differ mined by jModeltest version 2 (Darriba et al. 2012). We then con- among bee species (Zurbuchen and Muller¨ 2012). Conceivably, structed a maximum likelihood (ML) tree under the GTR + + bee species with similar ecological requirements are found in I model of nucleotide substitution with RAxML version 8 (Sta- similar habitats and are therefore more likely to exchange bac- matakis 2014). teria than bee species whose ecology precludes physical con- To classify Rickettsia strains from bee hosts, we used three loci tact. To test if bee ecology is correlated with endosymbiont com- (16S, atpA and coxA) that were recently employed to reconstruct position, we chose three well-documented ecological traits of evolutionary relationships of Rickettsia from arthropod and ver- bees (pollen preference, phenology, nesting substrate) and ex- tebrate hosts (Weinert et al. 2009). Consequently, we assembled amined their association with Wolbachia infections. Specifically, a dataset of these previously published sequences (66 strains) we tested if Wolbachia is more frequently encountered in (I) and Rickettsia sequences from bee hosts. The protein-coding bees specialized on a few pollen sources (oligolectic) or in bees genes coxA and atpA were translated to amino acids, aligned without marked pollen preference (polylectic); (II) spring species and then backtranslated with TranslatorX version 1.1 (Abascal, (phenological peak in March–April) or summer species (peak Zardoya and Telford 2010) and MAFFT version 7.1.64b (Katoh in May–August); and (III) ground-nesting bees or cavity-nesting et al. 2002) using the L-INS-i strategy. 16S sequences were aligned bees. For each of these tests, we only included species for which and masked with SSU-ALIGN version 0.1 (Nawrocki 2009). All the corresponding trait was unambiguously documented in the three loci were concatenated to a supermatrix with FASconCAT literature (Westrich 1989). 4 FEMS Microbiology Ecology, 2015, Vol. 91, No. 6

Additionally, we tested if Wolbachia is more common in bees serving a similarly high or higher level of shared strains within whose corresponding kleptoparasite (or host for kleptoparasitic regions by chance. This test was conducted using the full dataset bees) harbours Wolbachia. Kleptoparasites are bees that do not (N = 54 species), a reduced dataset using just members of the An- collect pollen but rather use nests and pollen of other bees to drenidae and Halictidae across the two regions (N = 27 species), provide for their own offspring. This lifestyle has evolved mul- and an even smaller dataset using only members of the Halicti- tiple times independently (Litman et al. 2013) and is found in dae (N = 15 species). For the sake of simplicity, we only used one a large proportion of bee species (Michener 2007). It has re- of the two Wolbachia strains within the doubly-infected Lasioglos- cently been hypothesized that Wolbachia may be transmitted sum trigeminum for this analysis. When the test was re-run using horizontally from bee larvae (or pollen provisions) to their klep- each of the two strains in L. trigeminum, it produced congruent toparasites, although this could not be established as a pre- results. dominant mechanism of Wolbachia transmission (Gerth, Rothe¨ and Bleidorn 2013). Information on the ecological traits and RESULTS on kleptoparasite–host associations were taken from Westrich (1989, and references cited therein) and is summarized in Ta- Wolbachia prevalence and distribution among bees ble S3 (Supporting Information). We used Fisher’s exact tests We detected Wolbachia in 107 of 170 screened bee species (63%). that were performed in the R environment (R Development Core For the species of which multiple individuals were tested, Wol- Team 2012) to assess the significance of the trait associations bachia was either present in all or none of the individuals (Ta- with Wolbachia infections. ble S1, Supporting Information). Although Wolbachia was by far the most commonly encountered endosymbiont in our study, Wolbachia Test for geographical isolation of strains from it was not evenly distributed among bee taxa (Fig. 1). Consid- bee hosts erable differences exist in the prevalence of Wolbachia in different families of bees, e.g. in Halictidae, 90% of the tested species To test if Wolbachia distribution among bees is influenced by (N = 29) harboured Wolbachia, whereas in Megachilidae, Wol- geography, we compiled a dataset of 58 MLST profiles from bachia was present in only 13% of the investigated species (N bees, comprising 46 already published Wolbachia MLST pro- = 31). Furthermore, similar differences were observed between files from German bees (Gerth, Rothe¨ and Bleidorn 2013), the genera within the family Apidae: Wolbachia was found in 6 newly sequenced Wolbachia strains from German Hylaeus 72% of the investigated Nomada species (N = 25), but only in two species and 8 Wolbachia strains from 13 US bee species se- out of eight and one out of five tested Bombus and Anthophora quenced for this study. Accession numbers and allele num- species, respectively (Fig. 1). Wolbachia prevalence also differed bers of all MLST loci employed are listed in Table S4 (Support- when grouping the investigated bees by ecological features. We ing Information). We used templates from the MLST database found it to be significantly more frequent in ground-nesting than (http://pubmlst.org/wolbachia) to align each of the five loci. in cavity-nesting bees and also more frequent in spring species Identical profiles were excluded from subsequent analyses, re- than in summer species (Fig. 2). Furthermore, bees whose sulting in 38 unique Wolbachia MLST profiles from bee hosts. We reconstructed a phylogeny of these strains with ClonalFrame version 1.2 (Didelot and Falush 2007). Using a Bayesian framework, ClonalFrame infers relationships between clonal organisms (such as bacteria) from MLST data, while incorporating potential events of recombination (Didelot and Falush 2007). Two independent runs with 1 million generations and a burnin of 250 000 generations each were performed. Convergence of runs was assessed with the implemented methods of Gelman and Rubin (1992) before the consensus tree was built out of all posterior trees. We then mapped the origin of the Wolbachia hosts (Germany or USA) onto the Wolbachia MLST phylogeny and tested if this trait (origin) is significantly associated with the phylogeny. To this end, we used the software BaTS (Bayesian Tip- association Significance testing) which implements three test statistics for assessing trait–tip associations, with null hypothe- ses always being random trait–tip associations (Parker, Rambaut and Pybus 2008). We used the posterior sample of ClonalFrame trees as input for BaTS. Because this phylogenetic approach did not take identical MLST profiles into account, we also used the Hunter–Gaston index to calculate the probability of two bees from the same region sharing identical Wolbachia strains (Hunter and Gaston 1988). The weighted average of this index across both regions was used as a test statistic, s. To estimate how likely this observed value would be if geography had no influence on Wol- bachia strain distribution, we used the Resampling Stats for Ex- Figure 1: (a) Cladogram showing bee family-level relationships based on recent phylogenetic analyses (Danforth et al. 2013). (b) Proportion of Wolbachia infected cel Add-in (http://www.resample.com) to randomly redistribute bee species per family (dark bars) and genus (light bars). Only taxa with at least Wolbachia the strains among the bee species across both regions five sampled individuals are displayed. Number of species tested is givenin and subsequently recalculated the weighted Hunter–Gaston in- parentheses. (c) Approximate proportion of German species of each bee family dex. We used 10 000 iterations to determine the probability of ob- screened in this study (grey). Species numbers are taken from Westrich (1989). Gerth et al. 5

Figure 2: Correlation of bee traits and Wolbachia prevalences. Bars represent the proportion of Wolbachia infected species in % for (a) species whose corresponding kleptoparasite or host bears Wolbachia (W+) and does not bear Wolbachia (W −); (b) ground-nesting and cavity-nesting species; (c) spring species and summer species; (d) oligolectic and polylectic bees. Significance levels were assessed with Fisher’s exact tests (see the section ‘materials and methods’ for details). corresponding kleptoparasite (or host, for kleptoparasitic bees) countered in any of the investigated bee species. Rickettsia was carries Wolbachia are significantly more likely to host Wolbachia detected in six species of two closely related halictid genera themselves (Fig. 2). Pollen preferences (oligolecty and polylecty) (Halictus and Lasioglossum)andinOsmia bicornis (Megachilidae). were not related to Wolbachia frequency. While it was present in at least 1/3 of individuals in infected Almost all of the Wolbachia strains from bees could be as- Halictidae, we found Rickettsia only in a single individual out of signed to supergroup A by phylogenetic analysis of ftsZ se- 12 sampled O. bicornis specimens. Phylogenetic analysis based quences (Fig. S1, Supporting Information). Supergroup B Wol- on three loci clustered halictid endosymbionts within the bel- bachia were only detected in a single group of bees, Hylaeus lii group, which includes Rickettsia of various arthropods (Fig. 4). (Fig. S1, Supporting Information). In this genus, we found super- Rickettsia from O. bicornis fell within the adalia group, together group A Wolbachia (five species), supergroup B Wolbachia (three with endosymbionts originally isolated from coccinellid beetles species) as well as seven uninfected species (Fig. 1). Addition- (Fig. 4). ally, a supergroup B allele (ftsZ) was detected in Hylaeus varie- Our screen further revealed Arsenophonus to be present in gatus. In US bees, we found another potentially recombinant the colletid species Colletes cunicularius (two out of three tested Wolbachia strain in Augochlorella aurata with three supergroup individuals), Colletes halophilus (2/6) and Colletes hederae (7/8). B alleles (gatB, fbpA, ftsZ) and two supergroup A alleles (coxA, Phylogenetic analysis of Arsenophonus 16S of Colletes hosts and hcpA). Finally, of 170 screened German bee species, only Osmia GenBank data resulted in a weakly supported topology (Fig. S3, caerulescens harboured supergroup F Wolbachia. This finding was Supporting Information). Most Arsenophonus strains from bees reported before (Gerth, Geißler and Bleidorn 2011;Gerthet al. were recovered in the same clade (hereafter Colletes clade I), to- 2014) and further confirmed here by sampling additional indi- gether with Arsenophonus isolated from honey bees (A. mellifera). viduals from Germany (Table S1, Supporting Information). The However, one strain was clustered with Hippobosca and Pyrrho- only O. caerulescens specimen sampled from the USA also carried coris endosymbionts (Colletes clade II). NeighborNet analysis of this strain. the groEL dataset placed Colletes clade I within Arsenophonus ssp. Phylogenetic reconstruction of selected Wolbachia strains (Fig. 5), and clade II within the outgroups Providencia, Proteus and from Germany and the USA based on MLST data revealed no ob- Photorhabdus. A very similar topology was also recovered by ML vious geographical patterns (Fig. 3). Furthermore, test statistics analysis of the same dataset (Fig. S4, Supporting Information). implemented in BaTS showed no significant impact of host origin on the Wolbachia phylogeny. However, with the exception to DISCUSSION the above-mentioned supergroup F strain, no Wolbachia strain was shared between Old World and New World bees (Fig. 3) By investigating a large proportion (about one third) of a local bee and some bees from the same region shared identical Wolbachia fauna, we found three bacterial endosymbionts to be distributed strains. Random redistribution of Wolbachia strains revealed that in taxon-specific patterns. Wolbachia was by far the most com- bees within a region were statistically more likely to share a Wol- monly detected endosymbiont. The here reported prevalence of bachia strain than would be expected by chance (Table 1). 63% is in line with statistical estimations of Wolbachia abun- dances among terrestrial arthropods in general (Hilgenboecker Other bacterial endosymbionts in bees et al. 2008; Zug and Hammerstein 2012)aswellaswithaprevious screen in bees on a smaller scale (75 investigated species, Gerth, Endosymbionts other than Wolbachia were only rarely identi- Geißler and Bleidorn 2011). However, arthropod Wolbachia infec- fied and restricted to few host lineages. Cardinium was not en- tion frequencies reported from empirical studies (Duron et al. 6 FEMS Microbiology Ecology, 2015, Vol. 91, No. 6

Figure 3: ClonalFrame genealogy of 58 Wolbachia strains from bees. Circles on internal nodes indicate ClonalFrame support values (black ≥ 95, grey ≥ 90). Symbols on tips correspond to single Wolbachia strains from bees, circles represent bee hosts from Germany and squares represent those from the USA. Colour of symbols indicates taxonomic association of the Wolbachia hosts: blue—Apidae, green—Andrenidae, yellow—Halictidae, red—Colletidae, purple—Melittidae. MLST profile numbers are given in Table S4 (Supporting Information).

Table 1: Hunter–Gaston index for the probability of two bees from the of Andrenidae and Halictidae are ground nesting and show a same region sharing identical Wolbachia strains (s) and the probabil- high Wolbachia infection rate whereas megachilid species are ity of observing this index under random distribution of Wolbachia usually cavity nesting and show very low Wolbachia prevalence. strains among bee hosts (P-value). Similarly, kleptoparasitic bee lineages are often closely related Group Observed sP-value to their host (Litman et al. 2013)—compatible with the observa- tion that parasitic lineages are often closely related to their hosts All species (N = 54) 0.1 <0.0001 (‘Emery’s rule’, Emery 1909). Additionally, although data on Wol- Andrenidae and Halictidae (N = 27) 0.099 0.0004 bachia absence/presence seem to suggest a potential pathway of Halictidae (N = 15) 0.243 0.0014 horizontal transmission between kleptoparasites and their bee hosts, this is not supported by MLST data (Gerth, Rothe¨ and Blei- dorn 2013). Therefore, it is problematic to link host ecology to Wolbachia infection rates based on the here presented data. 2008; Russell et al. 2012) are usually much lower, suggesting that On the contrary, there is some evidence for bee phylogeny Wolbachia prevalence in bees (especially in Halictidae and An- shaping endosymbiont distribution, i.e. Wolbachia abundance drenidae) is exceptionally high. varies greatly among bee families and genera (Fig. 1). While most In a previous study, Wolbachia strains from bee hosts were Wolbachia from bee hosts are supergroup A strains, we found su- investigated using MLST data in a phylogenetic framework in pergroup B Wolbachia in three species of Hylaeus (Fig. S1, Support- order to identify ecological factors that may facilitate horizon- ing Information). Furthermore, between supergroups recombi- tal transmission of Wolbachia strains (Gerth, Rothe¨ and Bleidorn nation was detected via MLST analysis for the Wolbachia strain 2013). Here, we instead asked if the presence (or the absence) of within H. variegatus. A previous analysis of Wolbachia strains Wolbachia in bees corresponds to bee ecology. Indeed, some eco- from bees did not detect a single supergroup B allele in 75 MLST logical factors correlate with Wolbachia frequencies in bees, such profiles (Gerth, Rothe¨ and Bleidorn 2013), suggesting that super- as phenology and nesting habitats (Fig. 2). The presence of Wol- group B in German bees is indeed limited to Hylaeus hosts. Other bachia in the corresponding kleptoparasite or host also seems studies have shown that at least on a higher taxonomic level (e.g. to increase the likelihood of encountering this bacterium in a insect orders), Wolbachia supergroups are host specific: super- bee species. Although these correlations show statistical sup- group F is the most common Wolbachia in termites (Lo and Evans port, they may be misleading as bee phylogeny and ecology are 2007), which is also true for supergroup A in Diptera (Stahlhut not independent of each other. For example, most of the species et al. 2010) and supergroup B in Lepidoptera (Russell et al. 2009). Gerth et al. 7

Figure 4: ML inference of Rickettsia strains from bees and other arthropod hosts, based on three genes from the dataset of Weinert et al. (2009). For ease of reading, the tree was pruned to exclude the outgroup Orienta and the Rickettsia groups hydra, tundra and torix (full tree in Fig. S2, Supporting Information). Numbers above nodes correspond to bootstrap support from 1000 pseudoreplicates and numbers below are posterior probabilities from the Bayesian analysis of the same dataset. Only values above or equal to 70% or 0.7 are shown. Terminal labels correspond to host names except for Rickettsia species. Rickettsia group names are used in accordance to Weinert et al. (2009).

A potential reason for this host specificity was recently demon- Raychoudhury et al. 2009;Gerth,Rothe¨ and Bleidorn 2013)and strated by Hughes et al. (2014). The authors found that the na- microbiome structure itself is also determined by host phy- tive microbiota of Anopheles gambiae interferes with the verti- logeny (Colman, Toolson and Takacs-Vesbach 2012;Yunet al. cal transmission of an artificially introduced Wolbachia strain. 2014), this may provide an explanation for the success of cer- When considering that Wolbachia strains are regularly transmit- tain Wolbachia strains in some host lineages, but not in others. ted horizontally, i.e. invade new host species (Baldo et al. 2008; However, the native microbiota of the bee species screened for 8 FEMS Microbiology Ecology, 2015, Vol. 91, No. 6

Figure 5: NeighborNet of Arsenophonus and related taxa based on groEL sequences. Arsenophonus from Colletes are highlighted in bold. Group names were used as employed by Duron et al. (2014). Labels correspond to host names, except for the ones marked by an asterisk. ML analysis resulted in a similar topology (Fig. S4, Supporting Information).

Wolbachia remains to be investigated. In summary, while some German and US bees is a supergroup F strain that was detected ecological traits correlate with Wolbachia presence, we think that in O. caerulescens from both continents. Osmia caerulescens was host phylogeny (on the level of families and genera), i.e. the sum introduced in the USA (Droege 2012), and therefore its Wolbachia of factors governed by host genetic background, seems to ex- strain was likely present before the geographical separation of plain Wolbachia distribution in bees best. O. caerulescens populations investigated here. Although susceptibility to Wolbachia infections is likely as- Similar to supergroup B Wolbachia, Rickettsia in bees are lim- sociated with bee phylogeny, this does not seem to be true for ited to a single lineage of bees (Halictidae, genera Halictus and Wolbachia strain phylogeny. As demonstrated recently, Wolbachia Lasioglossum) and cluster within the bellii group (Fig. 4). When strains found in closely related bees are not more similar ge- excluding Rickettsia from Lasioglossum semilucens, halictid Rick- netically than expected by chance (Gerth, Rothe¨ and Bleidorn ettsia form a monophyletic clade in our analysis. Other hosts 2013). The same conclusion can be drawn from the here pre- of the Rickettsia bellii group include a wide range of arthropods, sented MLST genealogy of Wolbachia strains (Fig. 3). However, we including vertebrate ectoparasites. Our findings thus confirm a found evidence for geographic isolation among Wolbachia strains general trend that was reported for Rickettsia and also for Wol- from bees. This separation is not reflected in a phylogenetic pat- bachia distribution (Weinert et al. 2009;Gerth,Rothe¨ and Blei- tern, but rather the high probability of two bees from the same dorn 2013): similar strains are found in closely related hosts, al- region sharing identical Wolbachia strains, which is very unlikely though horizontal transmission between distantly related hosts due to chance (Table 1). This indicates limited horizontal trans- occurs as well. To identify pathways of this horizontal transmis- mission of Wolbachia strains between the investigated regions. sion based on the here presented data would be speculative, es- Spatial separation of Wolbachia strains was demonstrated with a pecially as Halictidae are usually ground-nesting pollen general- large dataset previously (Russell et al. 2009), and may express ists like most bee species in Germany (Westrich 1989). Rickettsia a general pattern. The only Wolbachia strain shared between was also found in a single specimen of O. bicornis (Megachilidae) Gerth et al. 9

out of 12 individuals sampled from various populations. Since lated to host phylogeny and ecology. Additionally, there seem Rickettsia may be transmitted horizontally, e.g. via host plants or to be geographic barriers of horizontal Wolbachia transmission. host vertebrates (Perlman, Hunter and Zchori-Fein 2006; Caspi- As similar conclusions were drawn from a recent meta-analysis Fluger et al. 2011), its occurrence in O. bicornis maybetheresult of endosymbiont screens in arthropods (Russell et al. 2012), this of an individual horizontal transmission event or even contami- may reflect a general trend. Furthermore, the distribution of rare nation via gut content rather than evidence for a stable infection endosymbionts in bees (supergroup B Wolbachia, Rickettsia, Ar- in this species. senophonus) is restricted to specific host lineages. Further inves- The distribution of Arsenophonus among bees strikingly mir- tigations are needed to identify host specific traits that govern rors the pattern observed for Rickettsia.Wefoundittobepresent these distribution patterns. One causative factor might be the in- in just three species of a single genus (Colletes). All 16S sequences digenous microorganisms that are specific at least in some bee of Arsenophonus from Colletes were almost identical, except for species (McFrederick et al. 2013; McFrederick, Mueller and James the strain found in one specimen of C. halophilus (Fig. S3, Sup- 2014). Our study underlines the importance of detailed taxon- porting Information). The same distinct clusters were also recov- wide screens as the basis for exploring the processes shaping ered when analysing the groEL dataset (Figs 4 and S4, Support- endosymbiont distribution patterns. ing Information). Using 16S, both clusters could be assigned to Arsenophonus, whereas with groEL, the strain of C. halophilus was SUPPLEMENTARY DATA placed among the outgroups Proteus, Photorhabdus and Providen- cia. It is unclear which placement is more probable, as 16S was Supplementary data is available at FEMSEC online. shown not to be a reliable phylogenetic marker in Arsenophonus (Novakov´ a,´ Hypsaˇ and Moran 2009) and recombination among strains may also occur (Sorfova,´ Sker´ıkova´ and Hypsaˇ 2008). Ar- ACKNOWLEDGEMENTS senophonus was previously detected in honey bees (Babendreier We are grateful to the colleagues and students who helped in et al. 2007; Aizenberg-Gershtein, Izhaki and Halpern 2013; Corby- collecting bees: Detlef Bernhard, Robert Mayer, Anja Rudolph, Harris, Maes and Anderson 2014) and in the alfalfa leafcutter Stefan Schaffer & Ronny Wolf. We want to express our gratitude bee (Megachile rotundata, McFrederick, Mueller and James 2014). to Olivier Duron who kindly provided PCR positive controls. Mar- Based on 16S sequences, Arsenophonus from Colletes cladeIand inus Sommeijer and Robert Paxton kindly provided specimens from honey bees are very similar and distinct from Arsenophonus of C. halophilus. We thank three anonymous reviewers for com- of M. rotundata, suggesting at least two independent origins of ments on earlier versions of this article that helped in improving Arsenophonus in bees. Interestingly, two of the three infected Col- it. We acknowledge public authorities in Germany for permitting letes species are close relatives and one might speculate that the collection of protected species. We thank Martin Schlegel for the infection with Arsenophonus dates back to their common an- providing laboratory equipment and facilities, Allison Dehnel for cestor. However, the third species carrying Arsenophonus (C. cu- laboratory support, and the University of Leipzig for funding. nicularius) is not closely related and is also strongly separated in its phenology from the other two species (Kuhlmann et al. 2009). FUNDING Arsenophonus and related organisms (‘ALO’) have been isolated from various insect hosts (Duron et al. 2008;Novakov´ a,´ AS was supported by Kentucky’s National Science Foundation Hypsaˇ and Moran 2009), but seem to be especially diverse in the Experimental Program to Stimulate Competitive Research, grant ectoparasitic louse flies (Hippoboscidae) and bat flies (Streblidae 0814194. and Nycteribiidae) (Morse et al. 2013; Duron et al. 2014), in which Conflict of interest. obligate associations have evolved (Morse et al. 2012). The nature None declared. of the Arsenophonus symbiosis in Colletes cannot be resolved with absence/presence data alone. However, in an obligate symbiosis REFERENCES one would expect almost 100% infection frequencies, which is not the case in Colletes. Abascal F, Zardoya R, Telford MJ. TranslatorX: multiple align- Moreover, the impact of all of the here investigated endosym- ment of nucleotide sequences guided by amino acid trans- bionts on their bee hosts remains obscure. Although sex ratio lations. Nucleic Acids Res 2010;38:W7–W13. distorting effects have been described for Arsenophonus (Gherna Aizenberg-Gershtein Y, Izhaki I, Halpern M. Do honeybees shape et al. 1991), Rickettsia (Werren et al. 1994; von der Schulenburg the bacterial community composition in floral nectar? PLoS et al. 2001)andWolbachia (Werren, Baldo and Clark 2008), we One 2013;8:e67556. are not aware of any evidence supporting endosymbionts as the Akman L, Yamashita A, Watanabe H, et al. Genome sequence of cause for skewed sex ratios in bees. The investigation of other the endocellular obligate symbiont of tsetse flies, Wiggleswor- potential endosymbiont effects, such as metabolic provision- thia glossinidia. Nat Genet 2002;32:402–7. ing (Brownlie et al. 2009), virus protection (Teixeira, Ferreira and Babendreier D, Joller D, Romeis J, et al. Bacterial commu- Ashburner 2008) or cytoplasmic incompatibilities (Stouthamer, nity structures in honeybee intestines and their response Breeuwer and Hurst 1999), would require laboratory cultures and to two insecticidal proteins. FEMS Microbiol Ecol 2007;59: experiments, which are challenging for bees due to very long 600–10. generation times. Baldo L, Ayoub NA, Hayashi CY, et al. Insight into the routes of Wolbachia invasion: high levels of horizontal transfer in the spider genus Agelenopsis revealed by Wolbachia strain and mi- CONCLUSIONS tochondrial DNA diversity. Mol Ecol 2008;17:557–69. Baldo L, Dunning Hotopp JC, Jolley KA, et al. Multilocus sequence Our results show that the presence of seemingly universal en- typing system for the endosymbiont Wolbachia pipientis. Appl dosymbionts such as supergroup A Wolbachia in bees is corre- Environ Microb 2006;72:7098–110. 10 FEMS Microbiology Ecology, 2015, Vol. 91, No. 6

BaldoL,PrendiniL,CorthalsA,et al. Wolbachia are present in Gerth M, Rothe¨ J, Bleidorn C. Tracing horizontal Wolbachia move- Southern African scorpions and cluster with supergroup F. ments among bees (Anthophila): a combined approach using Curr Microbiol 2007;55:367–73. MLST data and host phylogeny. Mol Ecol 2013;22:6149–62. Brady SG, Sipes S, Pearson A, et al. Recent and simultaneous Gherna RL, Werren JH, Weisburg W, et al. Arsenophonus nasoniae origins of eusociality in halictid bees. PRSocB2006;273: gen. nov, sp. nov, the causative agent of the son-killer trait 1643–9. in the parasitic wasp Nasonia vitripennis. Int J Syst Bacteriol Brownlie JC, Cass BN, Riegler M, et al. Evidence for metabolic 1991;41:563–5. provisioning by a common invertebrate endosymbiont, Wol- Gill AC, Darby AC, Makepeace BL. Iron necessity: the secret of bachia pipientis, during periods of nutritional stress. PLoS Wolbachia’s success? PLoS Neglect Trop D 2014;8:e3224. Pathog 2009;5:e1000368. Greenleaf SS, Kremen C. Wild bees enhance honey bees’ pol- Cardinal S, Danforth BN. The antiquity and evolutionary history lination of hybrid sunflower. P Natl Acad Sci USA 2006;103: of social behavior in bees. PLoS One 2011;6:e21086. 13890–5. Cardinal S, Danforth BN. Bees diversified in the age of eudicots. Hamdi C, Balloi A., Essanaa J. et al. Gut microbiome dys- PRSocB2013;280:20122686. biosis and honeybee health. J Appl Entomol 2011;135: Caspi-Fluger A, Inbar M, Mozes-Daube N, et al. Horizontal trans- 524–33. mission of the insect symbiont Rickettsia is plant-mediated. Hedges LM, Brownlie JC, O’Neill SL, et al. Wolbachia and virus pro- PRSocB2011;279:1791–6. tection in insects. Science 2008;322:702. Colman DR, Toolson EC, Takacs-Vesbach CD. Do diet and tax- Hilgenboecker K, Hammerstein P, Schlattmann P, et al. How onomy influence insect gut bacterial communities? Mol Ecol many species are infected with Wolbachia?—a statistical 2012;21:5124–37. analysis of current data. FEMS Microbiol Lett 2008;281: Corby-Harris V, Maes P, Anderson KE. The bacterial communi- 215–20. ties associated with honey bee (Apis mellifera)foragers.PLoS Hoffmann S, Otto C, Kurtz S, et al. Fast mapping of short se- One 2014;9:e95056. quences with mismatches, insertions and deletions using in- Cordaux R, Pichon S, Hatira HB, et al. Widespread Wolbachia index structures. PLoS Comput Biol 2009;5:1–10. fection in terrestrial isopods and other crustaceans. Zookeys Hosokawa T, Koga R, Kikuchi Y, et al. Wolbachia as a bacteriocyte- 2012;176:123–31. associated nutritional mutualist. P Natl Acad Sci USA Danforth BN, Brady S, Sipes S, et al. Single-copy nuclear 2010;107:769–74. genes recover Cretaceous-age divergences in bees. Syst Biol Hughes GL, Dodson BL, Johnson RM, et al. Native micro- 2004;53:309–26. biome impedes vertical transmission of Wolbachia in Danforth BN, Cardinal S, Praz C, et al. The impact of molecular Anopheles mosquitoes. P Natl Acad Sci USA 2014;111: data on our understanding of bee phylogeny and evolution. 12498–503. Annu Rev Entomol 2013;58:57–78. Hunter PR, Gaston MA. Numerical index of the discriminatory Darriba D, Taboada GL, Doallo R, et al. jModelTest 2: more ability of typing systems: an application of Simpson’s index models, new heuristics and parallel computing. Nat Methods of diversity. J Clin Microbiol 1988;26:2465–6. 2012;9:772. Huson DH, Bryant D. Application of phylogenetic networks in Didelot X, Falush D. Inference of bacterial microevolution using evolutionary studies. Mol Biol Evol 2006;23:254–67. multilocus sequence data. Genetics 2007;175:1251–66. Jeyaprakash A, Hoy MA. Long PCR improves Wolbachia DNA am- Douglas A. Nutritional interactions in insect-microbial sym- plification: wsp sequences found in 76% of sixty-three arthro- bioses: aphids and their symbiotic bacteria Buchnera. Annu pod species. Insect Mol Biol 2000;9:393–05. Rev Entomol 1998;43:17–37. Katoh K, Misawa K, Kuma K, et al. MAFFT: a novel method for Droege S. Handy Bee Manual. Beltsville, MD: USGS Native Bee In- rapid multiple sequence alignment based on fast Fourier ventory and Monitoring Lab, 2012. transform. Nucleic Acids Res 2002;30:3059–66. Duron O, Bouchon D, Boutin S, et al. The diversity of reproductive Kikuchi Y, Hayatsu M, Hosokawa T, et al. Symbiont-mediated parasites among arthropods: Wolbachia do not walk alone. insecticide resistance. P Natl Acad Sci USA 2012;22: BMC Biol 2008;6:27. 8618–22. Duron O, Schneppat UE, Berthomieu A, et al. Origin, acquisition Kuck¨ P, Meusemann K. FASconCAT: Convenient handling of data and diversification of heritable bacterial endosymbionts in matrices. Mol Phylogenet Evol 2010;56:1115–8. louse flies and bat flies. Mol Ecol 2014;23:2105–17. Kuhlmann M, Almeida E, Laurenne N, et al. Molecular phylogeny Emery C. Uber¨ den Ursprung der dulotischen, parasitischen und and historical biogeography of the bee genus Colletes La- myrmekophilen Ameisen. Biol Cent 1909;29:352–62. treille, 1802 (Hymenoptera: Apiformes: Colletidae), based on Faria VG, Sucena E.´ Wolbachia in the Malpighian Tubules: mitochondrial COI and nuclear 28S sequence data. Insect Syst Evolutionary dead-end or adaptation? J Exp Zool Part B Evol 2009;40:291–318. 2013;320:195–9. Lanfear R, Calcott B, Ho SYW, et al. Partitionfinder: com- Garibaldi LA, Steffan-Dewenter I, Winfree R, et al. Wild pollina- bined selection of partitioning schemes and substitution tors enhance fruit set of crops regardless of honey bee abun- models for phylogenetic analyses. Mol Biol Evol 2012;29: dance. Science 2013;339:1608–11. 1695–701. Gelman A, Rubin DB. Inference from iterative simulation using Langridge G. Buzz off, that’s my bee! Nat Rev Microbiol multiple sequences. Stat Sci 1992;7:457–72. 2014;12:659. Gerth M, Gansauge M-T, Weigert A, et al. Phylogenomic analyses Litman JR, Praz CJ, Danforth BN, et al. Origins, evolution, and uncover origin and spread of the Wolbachia pandemic. Nat diversification of cleptoparasitic lineages in long-tongued Commun 2014;5:5117. bees. Evolution 2013;67:2982–98. Gerth M, Geißler A, Bleidorn C. Wolbachia infections in bees (An- Lo N, Evans TA. Phylogenetic diversity of the intracellular sym- thophila) and possible implications for DNA barcoding. Syst biont Wolbachia in termites. Mol Phylogenet Evol 2007;44: Biodivers 2011;9:319–27. 461–66. Gerth et al. 11

Losey JE, Vaughan M. The economic value of ecological services broad molecular surveys and a systematic review. PLoS One provided by insects. Bioscience 2006;56:311–23. 2012;7:e51027. McFrederick QS, Cannone JJ, Gutell RR, et al. Specificity be- Russell JA, Goldman-Huertas B, Moreau CS, et al. Special- tween lactobacilli and hymenopteran hosts is the excep- ization and geographic isolation among Wolbachia sym- tion rather than the rule. Appl Environ Microb 2013;79: bionts from ants and lycaenid butterflies. Evolution 2009;63: 1803–12. 624–40. McFrederick QS, Mueller UG, James RR. Interactions between Saridaki A, Bourtzis K. Wolbachia: more than just a bug in insect fungi and bacteria influence microbial community struc- genitals. Curr Opin Microbiol 2010;13:67–72. ture in the Megachile rotundata larval gut. PRSocB Sedivy C, Muller¨ A, Dorn S. Closely related pollen generalist bees 2014;281:20132653. differ in their ability to develop on the same pollen diet: ev- McFrederick QS, Wcislo WT, Hout MC, et al. Host species and idence for physiological adaptations to digest pollen. Funct developmental stage, but not host sociality, affects bacterial Ecol 2011;25:718–25. community structure in socially polymorphic bees. FEMS Mi- Shrestha RK, Lubinsky B, Bansode VB, et al. QTrim: a novel tool crobiol Ecol 2014;88:398–406. for the quality trimming of sequence reads generated us- McFrederick QS, Wcislo WT, Taylor DR, et al. Environment or ing the Roche/454 sequencing platform. BMC Bioinformatics kin: whence do bees obtain acidophilic bacteria? Mol Ecol 2014;15:33. 2012;21:1754–68. Sorfova´ P, Sker´ıkova´ A, Hypsaˇ V. An effect of 16S rRNA in- Michener CD. The Bees of the World, 2nd edn. Baltimore, Maryland: tercistronic variability on coevolutionary analysis in symbi- Johns Hopkins University Press, 2007. otic bacteria: molecular phylogeny of Arsenophonus triatomi- Morse SF, Bush SE, Patterson BD, et al. Evolution, multiple ac- narum. Syst Appl Microbiol 2008;31:88–100. quisition, and localization of endosymbionts in bat flies Stahlhut JK, Desjardins CA, Clark ME, et al. The mushroom habi- (Diptera: Hippoboscoidea: Streblidae and Nycteribiidae). Appl tatasanecologicalarenaforglobalexchangeofWolbachia. Environ Microb 2013;79:2952–61. Mol Ecol 2010;19:1940–52. Morse SF, Dick CW, Patterson BD, et al. Some like it hot— Stamatakis A. RAxML version 8: a tool for phylogenetic anal- Evolution and ecology of novel endosymbionts in bat flies of ysis and post-analysis of large phylogenies. Bioinformatics cave-roosting bats (Hippoboscoidea, Nycterophiliinae). Appl 2014;30:1312–3. Environ Microb 2012;78:8639–49. Stouthamer R, Breeuwer JAJ, Hurst GDD. Wolbachia pipientis:mi- Nawrocki EP. Structural RNA homology search and alignment crobial manipulator of arthropod reproduction. Annu Rev Mi- using covariance models. PhD Thesis, Washington University crobiol 1999;53:71–102. in Saint Louis, 2009. Teixeira L, Ferreira A, Ashburner M. The bacterial symbiont Wol- Novakov´ a´ E, Hypsaˇ V, Moran NA. Arsenophonus, an emerging bachia induces resistance to RNA viral infections in Drosophila clade of intracellular symbionts with a broad host distribu- melanogaster. PLoS Biol 2008;6:e1000002. tion. BMC Microbiol 2009;9:143. Von der Schulenburg JHG, Habig M, Sloggett JJ, et al. Incidence of Oliver KM, Russell JA, Moran NA, et al. Facultative bacterial sym- male-killing Rickettsia spp. (alpha-proteobacteria) in the ten- bionts in aphids confer resistance to parasitic wasps. P Natl spot ladybird beetle Adalia decempunctata L. (Coleoptera: Coc- Acad Sci USA 2003;100:1803–7. cinellidae). Appl Environ Microb 2001;67:270–7. Parker J, Rambaut A, Pybus OG. Correlating viral phenotypes Weeks AR, Marec F, Breeuwer JA. A mite species that consists with phylogeny: accounting for phylogenetic uncertainty. In- entirely of haploid females. Science 2001;292:2479–82. fect Genet Evol 2008;8:239–46. Weinert LA, Werren JH, Aebi A, et al. Evolution and diversity of Perlman SJ, Hunter MS, Zchori-Fein E. The emerging diversity of Rickettsia bacteria. BMC Biol 2009;7:6. Rickettsia. PRSocB2006;273:2097–106. Werren JH, Baldo L, Clark ME. Wolbachia: master manipulators of Praz CJ, Muller¨ A, Dorn S. Specialized bees fail to develop on non- invertebrate biology. Nat Rev Microbiol 2008;6:741–51. host pollen: Do plants chemically protect their pollen? Ecol- Werren JH, Hurst GDD, Zhang W, et al. Rickettsial relative asso- ogy 2008;89:795–804. ciated with male killing in the ladybird beetle (Adalia bipunc- R Development Core Team. R: A Language and Environment for Sta- tata). J Bacteriol 1994;176:388–94. tistical Computing. Vienna, Austria: R Foundation for Statisti- Werren JH, Windsor D, Guo LR. Distribution of Wolbachia among cal Computing, 2012. neotropical arthropods. PRSocB1995;262:197–204. Raychoudhury R, Baldo L, Oliveira DCSG, et al. Modes of acqui- Westrich P. Die Wildbienen Baden-Wurttembergs¨ I & II. Stuttgart: sition of Wolbachia: horizontal transfer, hybrid introgression, Ulmer, 1989. and codivergence in the Nasonia species complex. Evolution Williamson DL, Sakaguchi B, Hackett KJ, et al. Spiroplasma poul- 2009;63:165–83. sonii sp. nov., a new species associated with male-lethality in Ronquist F, Huelsenbeck JP. MrBayes 3: Bayesian phyloge- Drosophila willistoni, a neotropical species of fruit fly. Int J Syst netic inference under mixed models. Bioinformatics 2003;19: Bacteriol 1999;49:611–8. 1572–4. Xie J, Butler S, Sanchez G, et al. Male killing Spiroplasma protects Ros VID, Fleming VM, Feil EJ, et al. How diverse is the Drosophila melanogaster against two parasitoid wasps. Hered- genus Wolbachia? Multiple-gene sequencing reveals a pu- ity 2013;112:399–408. tatively new Wolbachia supergroup recovered from spider Yun J-H, Roh SW, Whon TW, et al. Insects gut bacterial di- mites (Acari: Tetranychidae). Appl Environ Microb 2009;75: versity determined by host environmental habitat, diet, 1036–43. developmental stage and phylogeny. Appl Environ Microb Russell JA. The ants (Hymenoptera: Formicidae) are unique and 2014;80:5254–64. enigmatic hosts of prevalent Wolbachia (Alphaproteobacte- Zchori-Fein E, Gottlieb Y, Kelly SE, et al. A newly discovered ria) symbionts. Myrmecological News 2012;16:7–23. bacterium associated with parthenogenesis and a change in Russell JA, Funaro CF, Giraldo YM, et al. A veritable menagerie host selection behavior in parasitoid wasps. P Natl Acad Sci of heritable bacteria from ants, butterflies, and beyond: USA 2001;98:12555–60. 12 FEMS Microbiology Ecology, 2015, Vol. 91, No. 6

Zug R, Hammerstein P. Still a host of hosts for Wolbachia:analy- Zurbuchen A, Landert L, Klaiber J, et al. Maximum forag- sis of recent data suggests that 40% of terrestrial arthropod ing ranges in solitary bees: only few individuals have the species are infected. PLoS One 2012;7:e38544. capability to cover long foraging distances. Biol Conserv Zug R, Hammerstein P. Bad guys turned nice? A critical assess- 2010;143:669–76. ment of Wolbachia mutualisms in arthropod hosts. Biol Rev Zurbuchen A, Muller¨ A. Wildbienenschutz–von der Wissenschaft zur 2015;90:89–111. Praxis, Bristol-Schriftenreihe 33. Bern: Haupt Verlag, 2012.