Diversity and evolutionary patterns of bacterial gut associates of corbiculate bees

17
Diversity and evolutionary patterns of bacterial gut associates of corbiculate bees HAUKE KOCH,* DHARAM P. ABROL, JILIAN LI § and PAUL SCHMID-HEMPEL* *ETH Zurich, Institute of Integrative Biology (IBZ), Universitatsstrasse 16, CH-8092, Zurich, Switzerland, Ecology and Evolutionary Biology, Yale University, 300 Heffernan Drive, West Haven, CT, 06516, USA, Division of Entomology, Sher-e-Kashmir University of Agricultural Sciences and Technology, Chatha, Jammu (J&K), 180009, India, §Key Laboratory of Pollinating Insect Biology of the Ministry of Agriculture, Chinese Academy of Agricultural Science, Institute of Apicultural Research, Xiangshan, Beijing 100093, China Abstract The animal gut is a habitat for diverse communities of microorganisms (microbiota). Honeybees and bumblebees have recently been shown to harbour a distinct and spe- cies poor microbiota, which may confer protection against parasites. Here, we investi- gate diversity, host specificity and transmission mode of two of the most common, yet poorly known, gut bacteria of honeybees and bumblebees: Snodgrassella alvi (Betapro- teobacteria) and Gilliamella apicola (Gammaproteobacteria). We analysed 16S rRNA gene sequences of these bacteria from diverse bee host species across most of the hon- eybee and bumblebee phylogenetic diversity from North America, Europe and Asia. These focal bacteria were present in 92% of bumblebee species and all honeybee spe- cies but were found to be absent in the two related corbiculate bee tribes, the stingless bees (Meliponini) and orchid bees (Euglossini). Both Snodgrassella alvi and Gilliamel- la apicola phylogenies show significant topological congruence with the phylogeny of their bee hosts, albeit with a considerable degree of putative host switches. Further- more, we found that phylogenetic distances between Gilliamella apicola samples corre- lated with the geographical distance between sampling locations. This tentatively suggests that the environmental transmission rate, as set by geographical distance, affects the distribution of G. apicola infections. We show experimentally that both bac- terial taxa can be vertically transmitted from the mother colony to daughter queens, and social contact with nest mates after emergence from the pupa greatly facilitates this transmission. Therefore, sociality may play an important role in vertical transmis- sion and opens up the potential for co-evolution or at least a close association of gut bacteria with their hosts. Keywords: co-evolution, microbiome, specificity, symbiont, transmission mode Received 6 April 2012; revision received 5 December 2012; accepted 11 December 2012 Introduction Bacteria are ubiquitous partners of animal hosts (e.g. Hosokawa et al. 2006; Fraune & Bosch 2007; Ley et al. 2008) with fundamental importance for their ecology and evolution (Fraune & Bosch 2010; Feldhaar 2011). The intimacy of the relationship between hosts and bac- teria may range from vertically transmitted, intracellu- lar symbionts obligate for host survival and incapable of surviving outside of host cells (Clark et al. 2000), to transient relationships in which bacterial associates reg- ularly grow in the environment outside of their hosts (Gordon 2001). The outcome of these relationships may be beneficial for both partners (mutualism), harmful for the host (parasitism), or only beneficial for the bacteria without affecting the host (commensalism). Of particu- lar importance are the communities of microorganisms (microbiota) found in the gut of almost all animals, with crucial functions, for example, in digestion, immune Correspondence: Hauke Koch, Fax: (001) 203 737 3109; E-mail: [email protected] © 2013 Blackwell Publishing Ltd Molecular Ecology (2013) 22, 2028–2044 doi: 10.1111/mec.12209

Transcript of Diversity and evolutionary patterns of bacterial gut associates of corbiculate bees

Diversity and evolutionary patterns of bacterial gutassociates of corbiculate bees

HAUKE KOCH,*† DHARAM P. ABROL,‡ J ILIAN LI§ and PAUL SCHMID-HEMPEL*

*ETH Z€urich, Institute of Integrative Biology (IBZ), Universit€atsstrasse 16, CH-8092, Z€urich, Switzerland, †Ecology and

Evolutionary Biology, Yale University, 300 Heffernan Drive, West Haven, CT, 06516, USA, ‡Division of Entomology,

Sher-e-Kashmir University of Agricultural Sciences and Technology, Chatha, Jammu (J&K), 180009, India, §Key Laboratory of

Pollinating Insect Biology of the Ministry of Agriculture, Chinese Academy of Agricultural Science, Institute of Apicultural

Research, Xiangshan, Beijing 100093, China

Abstract

The animal gut is a habitat for diverse communities of microorganisms (microbiota).

Honeybees and bumblebees have recently been shown to harbour a distinct and spe-

cies poor microbiota, which may confer protection against parasites. Here, we investi-

gate diversity, host specificity and transmission mode of two of the most common, yet

poorly known, gut bacteria of honeybees and bumblebees: Snodgrassella alvi (Betapro-teobacteria) and Gilliamella apicola (Gammaproteobacteria). We analysed 16S rRNA

gene sequences of these bacteria from diverse bee host species across most of the hon-

eybee and bumblebee phylogenetic diversity from North America, Europe and Asia.

These focal bacteria were present in 92% of bumblebee species and all honeybee spe-

cies but were found to be absent in the two related corbiculate bee tribes, the stingless

bees (Meliponini) and orchid bees (Euglossini). Both Snodgrassella alvi and Gilliamel-la apicola phylogenies show significant topological congruence with the phylogeny of

their bee hosts, albeit with a considerable degree of putative host switches. Further-

more, we found that phylogenetic distances between Gilliamella apicola samples corre-

lated with the geographical distance between sampling locations. This tentatively

suggests that the environmental transmission rate, as set by geographical distance,

affects the distribution of G. apicola infections. We show experimentally that both bac-

terial taxa can be vertically transmitted from the mother colony to daughter queens,

and social contact with nest mates after emergence from the pupa greatly facilitates

this transmission. Therefore, sociality may play an important role in vertical transmis-

sion and opens up the potential for co-evolution or at least a close association of gut

bacteria with their hosts.

Keywords: co-evolution, microbiome, specificity, symbiont, transmission mode

Received 6 April 2012; revision received 5 December 2012; accepted 11 December 2012

Introduction

Bacteria are ubiquitous partners of animal hosts (e.g.

Hosokawa et al. 2006; Fraune & Bosch 2007; Ley et al.

2008) with fundamental importance for their ecology

and evolution (Fraune & Bosch 2010; Feldhaar 2011).

The intimacy of the relationship between hosts and bac-

teria may range from vertically transmitted, intracellu-

lar symbionts obligate for host survival and incapable

of surviving outside of host cells (Clark et al. 2000), to

transient relationships in which bacterial associates reg-

ularly grow in the environment outside of their hosts

(Gordon 2001). The outcome of these relationships may

be beneficial for both partners (mutualism), harmful for

the host (parasitism), or only beneficial for the bacteria

without affecting the host (commensalism). Of particu-

lar importance are the communities of microorganisms

(microbiota) found in the gut of almost all animals, with

crucial functions, for example, in digestion, immuneCorrespondence: Hauke Koch, Fax: (001) 203 737 3109;

E-mail: [email protected]

© 2013 Blackwell Publishing Ltd

Molecular Ecology (2013) 22, 2028–2044 doi: 10.1111/mec.12209

system functioning and organ development (Dillon &

Dillon 2004; Ley et al. 2008; Zilber-Rosenberg & Rosen-

berg 2008; Fraune & Bosch 2010).

The ecological and evolutionary patterns and pro-

cesses shaping the association between these two par-

ties are a central, but poorly understood, research

question. At one end of the spectrum, the microbiota

may co-evolve with their hosts. In these associations,

the host immune system may actively shape the micro-

biota, and members of the microbiota in turn adapt to

different hosts and within-host environments and

potentially provide important functions to the host

(Ochman et al. 2010; Bevins & Salzman 2011; Frese et al.

2011; Brucker & Bordenstein 2012). In this case, gut

microbes will show a high degree of specificity, fre-

quent cospeciation with their host species and signs of

adaptation to different host environments at the level of

the genome (Falush et al. 2003; Hosokawa et al. 2006;

Kikuchi et al. 2009; Oh et al. 2010; Frese et al. 2011).

These microbes also tend to be vertically transmitted to

the offspring and incapable of growing in the environ-

ment outside of the host (Hosokawa et al. 2006; Sachs

et al. 2011). At the other end of the spectrum, gut

microbes may be environmentally (horizontally) trans-

mitted, short-term residents in the animal gut. In this

case, gut bacteria are expected to generally lack evi-

dence of codivergence (cospeciation) with their hosts,

show little or no host species specificity and may be

able to grow in the environment outside of the host

(Gordon 2001; Dunlap et al. 2007; Kikuchi et al. 2007,

2011; Sachs et al. 2011). In spite of the potential diver-

sity and ubiquitous nature of animal gut microbiotas,

few studies have so far examined these above-men-

tioned patterns in a phylogenetic comparative frame-

work of closely related host species.

Because of its accessibility and relative simplicity, the

gut microbiota of corbiculate bees presents a promising

study system to understand the host–microbiota interac-

tions in more detail. In fact, recent surveys have

described a relatively simple and distinct gut microbi-

ota for the corbiculate bee tribes Apini (honeybees) and

Bombini (bumblebees), with both host groups sharing

similar bacterial taxa (Mohr & Tebbe 2006; Koch &

Schmid-Hempel 2011a; Martinson et al. 2011). These

bacteria are apparently absent in solitary bee species,

which appear to possess a more variable, unspecific gut

microbiota (Martinson et al. 2011). Hence, it has been

speculated that sociality is an important factor that

might facilitate the vertical transmission of gut bacteria

and thus allows for co-evolution of host and gut micro-

biota (Martinson et al. 2011). Indeed, honeybee and

bumblebee workers raised in isolation after pupal emer-

gence have been shown to lack this specific gut microbi-

ota (Gilliam 1971; Koch & Schmid-Hempel 2011b;

Martinson et al. 2012), with potentially detrimental con-

sequences such as higher parasite susceptibility (Koch

& Schmid-Hempel 2011b, 2012). A comparative analysis

covering a larger number of species and individuals of

honeybees and bumble bees to assess the taxonomic

distribution and specificity of the members of the gut

microbiota is, however, still missing. Also absent are

studies on the microbiota of the other two related corbi-

culate bees tribes, the eusocial Meliponini (stingless

bees) and the mostly solitary Euglossini (orchid bees),

that could shed light on these issues.

In this study, we focused on two major bacterial taxa

from the bee gut microbiota, recently described as

Snodgrassella alvi (Betaproteobacteria) and Gilliamella api-

cola (Gammaproteobacteria) (Kwong & Moran in press).

These two species were previously found to be domi-

nant members of the gut microbiota of bumblebees

(Koch & Schmid-Hempel 2011a,b; Koch et al. 2012) and

honeybees (Jeyaprakash et al. 2003; Mohr & Tebbe 2006;

Babendreier et al. 2007; Martinson et al. 2011, 2012;

Moran et al. 2012). We screened a wide range of corbi-

culate bee species from Africa, Asia, Europe and North

America for these bacteria and analysed their bacterial

16S rRNA gene sequences. With these data, we investi-

gated the correlation between host phylogeny and geo-

graphical distances with the bacterial phylogeny. We

expect that if, on the one hand, the topology of the bac-

terial phylogeny shows significant congruence to the

host phylogeny, the hypothesis of an intimate co-evolu-

tionary relationship of hosts and gut microbiota would

gain ground. This intimate relationship could evolve

through a stable transmission across generations, poten-

tially mediated by social contact within honeybee and

bumblebee colonies [i.e. a prevalence of vertical trans-

mission to the colony’s offspring (Martinson et al.

2011)]. We therefore experimentally tested the general

ability of Snodgrassella and Gilliamella to be vertically

transmitted to daughter queens. On the other hand, if

no match were found between host and bacterial phy-

logeny, codiversification between hosts and bacteria is

less likely, and bacteria might instead be frequently

transmitted horizontally between host species, for

example via environmental routes such as flowers (Dur-

rer & Schmid-Hempel 1994; McFrederick et al. 2012).

This hypothesis of ecologically driven host–bacterial

associations would be corroborated further by a correla-

tion between the geographical distance between sam-

pling locations and the genetic divergence in the

bacteria. In this case, the bacteria would also be rather

unspecific in their host range, but the environmental

transmission rate may be correlated with geographical

distance, such that individuals in closer proximity to

each other would share more similar strains (Gordon &

Lee 1999; Zamborsky & Nishiguchi 2011). Note that this

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2029

does not exclude the possibility that geographic dis-

tance correlates with host genetic differentiation and

the respective local adaptation of bacteria to their host

populations.

Materials and methods

Collection of samples

Wild bees were collected in pure ethanol in single 2-mL

plastic tubes and stored at �20 °C until further process-

ing (see Table S1, Supporting information for detailed

information on all specimens). We included all bumble-

bee subgenera and honeybee species that were available

to us. If possible, we sampled multiple species per sub-

genus and around five individuals per species to assess

specificity of bacterial strains to individual bee subgen-

era or species. Whenever possible, samples from multi-

ple locations for one host species were included.

However, studies have shown that the chances that sis-

ter workers are heavily over-represented in such field

samples is indeed rather small, and the relatedness

among sampled workers is in fact not different from

zero (e.g. Chapman et al. 2003).

To verify the transmission of microbiota of colonies

to their own daughter queens, we collected several

daughter queens of Bombus terrestris from within colo-

nies kept in the field in 2009 (near Ittingen abbey, TG,

NE Switzerland). The colonies were field-housed in

nesting boxes (Schwegler, Schorndorf/Germany) with

restricted entrance diameters, prohibiting daughter

queens from leaving the nest, while letting workers

pass through. We checked colonies once every week for

new daughter queens and froze them at �20 °C. The

daughter queens of B. terrestris normally spend the first

few days after pupal emergence within their mother

colony (Alford 1975). Therefore, our sampling protocol

allowed us to assess the microbiota acquired by daugh-

ter queens within their mother colony under natural

conditions, while ensuring that they could not have

been taken up outside of the nest.

DNA extraction, PCR, sequencing

Whole guts were removed from bee abdomens by sterile

dissection and crushed in DNA extraction buffer (see

Koch & Schmid-Hempel 2011a). DNA was extracted with

QIAGEN DNeasy Blood and Tissue Kit following the

procedure outlined by Koch & Schmid-Hempel (2011a)

including a predigestion with lysozyme for Gram-posi-

tive bacteria. Sufficient quality of the extracted DNA was

checked via a PCR with the universal 16S rRNA gene eu-

bacterial primers 27Mf (5′-AGA GTT TGA TCM TGG

CTC AG-3′) and 1492 Year (5′-ACG GYT ACC TTG TTA

CGA CTT-3′) with 1.5 lL DNA template in a total of

10 lL PCR volume. The PCR protocol consisted of 30

cycles of 94 °C (30 s), 51 °C (30 s) and 72 °C (1.5 min)

(Koch & Schmid-Hempel 2011a). PCR amplification suc-

cess was verified on a 1.5% agarose gel. Negative samples

were excluded from further analyses. Partial 16S rRNA

gene fragments were PCR amplified with the primer

pair Bprotf (5′-CAGC ACGGAGAGCTTGCTCTC-3′)/

Bprotr (5′-GCATACCGT GTTAAGCGACC-3′), specific

for Snodgrassella alvi and the primer pair Gprotf

(5′-GTATGGGGATCTGCCGA ATG-3′)/Gprotr (5′-AG-

CTATCTACTTCTGGTGCA-3′), specific for Gilliamella

apicola. We chose the specific primers based on an align-

ment of all 16S rRNA gene sequences obtained from

bumblebee guts in a previous study (Koch & Schmid-

Hempel 2011a) and included 16S rRNA gene sequences

of the target bacteria from Apis mellifera (Jeyaprakash

et al. 2003; Babendreier et al. 2007). Primer sequences

were then also cross-checked against 16S rRNA gene

sequences deposited in GenBank for their specificity for

the target bacterial taxa and their match with A. mellifera-

derived sequences. For both primer pairs, PCR condi-

tions were as follows: 35 cycles of 94 °C (30 s), 53 °C(30 s) and 72 °C (1.5 min). The PCR-mix consisted of

0.5 lL of each primer (10 lM), 5 lL 59 PCR buffer,

0.125 lL of each dNTP (50 lM), 0.75U of Taq polymerase,

15.35 lL of ultrapure water and 3 lL DNA template (in

total 25 lL). Samples yielding a single band of appropri-

ate size were sequenced directly from the PCR prod-

uct after incubation with exonuclease I and shrimp

alkaline phosphatase to remove unincorporated primers

and dNTPs. We conducted cycle sequencing in a volume

of 10 lL with 0.8 lL BigDye 3.1, 1.6 lL sequencing

buffer (ABI, Foster City, CA, USA), 0.16 lL primer

(10 lM), 4.94 lL ddH2O and 2.5 lL PCR product. In

addition to the PCR primers, we used the internal

primers F790 (5′-ATT AGA TAC CCT GGT AG-3′) and

R790 (5′-CTA CCA GGG TAT CTA AT-3′) for sequenc-

ing with a twofold coverage from both ends over the

whole amplicon length. Bprotf was replaced with Bprotf-

seq (5′-GAGTAATG CATCGGAACGTAC-3′) for the

sequencing reaction, as the former gave poor sequencing

results. We ran products on an ABI 3130xl capillary

sequencer.

Molecular cloning

All samples from Meliponini and Euglossini screened

negatively for the target bacteria with specific PCR

primers. As the construction of specific Snodgrassella

and Gilliamella PCR primers based on sequences of hon-

eybee and bumblebee bacteria could, however, have led

to a false-negative result for these bacteria in the

Meliponini and Euglossini samples, we selected two

© 2013 Blackwell Publishing Ltd

2030 H. KOCH ET AL.

specimens each of Melipona panamica, Meliponula bocan-

dei and Euglossa imperialis, as well as one specimen of

Eufrisea sp. and Eulaema sp. for the construction of indi-

vidual clone libraries with universal eubacterial 16S

rRNA gene primers. For this, we amplified the 16S

rRNA gene with the universal eubacterial primers 27Mf

and 1492 Year (see above) and cloned PCR products

with the pGEM-T Easy Vector kit (Promega) into

electrocompetent Escherichia coli cells. The molecular

cloning protocol was identical with the one described

by Koch & Schmid-Hempel (2011a). We picked 24

clones for each individual clone library and sequenced

inserts (sequencing protocol see above) with the primer

T7 (5′-TAATACGACTCACTATAGGG-3′). On the basis

of these partial sequences, we then selected different

inserts from each library and sequenced the complete

insert with the additional primers SP6

(5′-CTATTTAGGTGACACTATAG-3′) and F790/R790.

Phylogenetic and statistic analyses

Raw sequences were aligned for each sample with Se-

quencher 5.0 (Gene Codes). Bases with ambiguous

sequence chromatogram peaks were coded with the

appropriate IUPAC nucleotide ambiguity code for fur-

ther analysis. We removed vector and primer sequences

and exported the consensus sequence. The sequences

were checked for chimaeras with Bellerophon (Huber

et al. 2004) and chimeric sequences were removed from

further analysis. We deposited all sequences in Gen-

Bank (Accession nos JQ389879–JQ390033 & JQ405085–

JQ405210, Table S1, Supporting information). We

searched the Ribosomal Database Project (RDP release

10, Cole et al. 2009) for the most similar type strains of

the sequences obtained from the clone libraries using

SeqMatch. We furthermore searched GenBank with

BLASTN for similar sequences with a particular focus

on sequences from bee or other insect associated bacte-

ria. For the analysis of Snodgrassella alvi and Gilliamella

apicola phylogenies, we incorporated a selection of hon-

eybee- and bumblebee-derived sequences from previous

studies (Jeyaprakash et al. 2003; Babendreier et al. 2007;

Olofsson & V�asquez 2008; Koch & Schmid-Hempel

2011a; Martinson et al. 2011; Disayathanoowat et al.

2012; V�asquez et al. 2012). To avoid a bias towards

studies and taxa with deep sampling of 16S, we hap-

hazardly selected a sequence for each bee species from

each study for each phylogenetic dataset. None of the

top hits for the sequences of Snodgrassella alvi and Gil-

liamella apicola were from environmental sources or

hosts other than honeybees or bumblebees, suggesting

these two bacteria to be specific to the latter two bee

host taxa (see also Kwong & Moran in press).

Sequences were aligned with ClustalW (Thompson et al.

1994) (http://align.genome.jp) with standard settings

for DNA (gap opening penalty 15, gap extension pen-

alty 6.66). We deposited the alignments in TreeBASE

(Sanderson et al. 1994; submission ID 13472). We deter-

mined an appropriate model of sequence evolution with

jModelTest 0.1 (Posada 2008), choosing the model with

the lowest Akaike information criterion. For all data

sets (Snodgrassella alvi, Gilliamella apicola, sequences from

clone libraries), we then calculated a maximum-likeli-

hood phylogenetic tree in PhyML 3.0 (Guindon et al.

2010, http://www.atgc-montpellier.fr/phyml/) using a

GTR+c+inv model. Branch support was assessed by

computing 500 bootstrap replicates from the original

matrices (GTR+c+inv model).

We assessed the correlation of the bacterial phyloge-

netic trees with the host phylogenetic tree and the geo-

graphical distance between sampling locations with

Mantel tests (Mantel 1967). For this analysis, we

excluded samples that had either unclear identifications

or originated from laboratory-kept hosts. We also

excluded the B. terrestris samples of daughter queens

sampled from within colonies in Switzerland as they

were sampled with a distinct protocol, and an inclusion

would have overrepresented this species in the analysis.

To run these tests, we first constructed distance matri-

ces for the bacterial phylogenies by computing the sum

of branch lengths between all pairs of taxa (patristic dis-

tance) in R (R Development Core Team 2011) using the

APE package (Paradis et al. 2004). We also calculated

the respective patristic distances for the host species,

using a phylogenetic tree generated from the alignment

of Hines (2008) deposited in TreeBASE (Sanderson et al.

1994; study ID 1927, alignment M2930). We pruned the

alignment of all bee taxa not examined in this study

and conducted a maximum-likelihood search in PhyML

3.0 (Guindon et al. 2010) with a GTR+c+inv model. The

obtained topology was largely identical to the one in

Hines (2008), but a few nodes differed (without boot-

strap support), presumably as a result of using a highly

pruned matrix (the analysis of Hines included 218 bum-

ble bee taxa). To avoid these potential artefacts due to

limited taxon sampling, we subsequently fixed the

topology to the well-supported topology of Hines

(2008). In addition, Apis cerana and Bombus tunicatus

were missing from the alignment of Hines (2008). We

therefore added mitochondrial 16S rRNA gene

sequences for both taxa from Cameron et al. (2007) and

Ram�ırez et al. (2010) to the alignment and fixed the

position of these two taxa with respect to the taxa

included in Hines (2008) according to the results of

Cameron et al. (2007) and Ram�ırez et al. (2010) (i.e. Apis

cerana as sister to Apis mellifera and Bombus tunicatus

as sister to Bombus patagiatus). We then ran a new

maximum-likelihood search in PhyML (GTR+c+inv

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2031

model) using these topological constraints and used the

result for further analyses.

Finally, we computed a geographic distance matrix

for all pairwise comparisons between sample locations

using the Geographic Distance Matrix Generator (Ersts

2012). For bacterial sequences downloaded from Gen-

Bank, we obtained geographical information from the

corresponding publications. We computed Mantel tests

in PASSaGE v2 (Rosenberg & Anderson 2011) using

10 000 replicates. We used partial Mantel tests (Legen-

dre & Troussellier 1988) to control for the possible inter-

dependence of bacteria phylogeny, host phylogeny and

geographical distance. As some sequences obtained

with the Gilliamella primer set fell outside the main

Gilliamella apicola clade and were closer to sequences

from other insects and the Gamma-2 phylotype of Mart-

inson et al. (2011), we analysed the Gilliamella phylogeny

separately excluding this distinct clade. These sequences

(mostly derived from the host subgenus Pyrobombus)

appear to belong to a novel clade within the Orbaceae

and may deserve recognition as a distinct genus.

To further test the congruence between the host and

bacterial trees, we performed cophylogeny reconstruc-

tions in Jane (version 4, Conow et al. 2010). We col-

lapsed branches of zero branch length in the bacterial

phylogenies and treated them as soft polytomies. In this

way, Jane will resolve the bacterial tree into bifurcations

while minimizing the overall cost of the cophylogeny

analysis. If several bacterial sequences from one host

species were present in the resulting polytomy, only

one representative bacterial sequence was maintained.

Furthermore, we collapsed monophyletic groups in the

bacterial phylogeny originating from a single host spe-

cies to one representative tip, to prevent the spurious

inference of duplication events at the tips of the tree.

The absence of a more detailed understanding of the

evolutionary interactions of the host and bacteria in our

system makes it difficult to choose a particular cost

matrix for the cophylogeny analysis. We therefore

evaluated two commonly used settings for the cost

matrix (Conow et al. 2010; Cruaud et al. 2012; Decelle

et al. 2012; Mendlov�a et al. 2012), which have been

implemented as default settings in Jane with a vertex-

based cost model: Setting 1 (cosp. = 0): assuming no

cost for cospeciation and cost = 1 for all other events

(cospeciation = 0, duplication = 1, duplication & host

switch = 2, loss = 1, failure to diverge = 1); Setting 2

(cosp. = 1): assuming cost = 1 for all events (cospecia-

tion = 1, duplication = 1, duplication & host switch = 2,

loss = 1, failure to diverge = 1). The genetic algorithm

was run with 100 generations and a population size of

300, with these settings the algorithm consistently found

solutions of identical costs. Significant matching of host

and bacterial phylogenies was assessed by computing

the costs of 500 replicates with random tip mapping

and comparing the resulting costs to the cost of the ori-

ginal associations.

Transmission experiments

Cocoons of daughter queens from five laboratory col-

onies of Bombus terrestris were removed from their

mother colonies and incubated at 30 °C, 60% rel.

humidity in sterile plastic boxes until emergence of

adults. Half of the emerged daughter queens were

then transferred to individual plastic boxes with auto-

claved cat litter and fed with filter sterilized (0.2 lmpore size) sugar water and heat-treated (85 °C,30 min) pollen. The other half of the daughter queens

were put back into their colony of origin for 3 days.

After this period, they were taken out of the colonies

and kept like the other half of the daughter queens

on sterilized food in individual boxes. Five daughter

queens of each group were frozen at �20 °C after

another 7 days. The remaining daughter queens were

mated with males and put into artificial hibernation

at 4 °C for 3 months. Males were obtained from iso-

lated pupae and had been kept on a sterile diet prior

to mating. After hibernation, daughter queens were

kept in individual boxes on a sterile diet as described

earlier. We had originally intended to let the daughter

queens lay eggs and found a colony. However, after

3 months only one queen had produced a small

amount of brood and, hence, we froze them at

�20 °C for analysis of their gut microbiota. DNA

extraction and specific PCRs for Snodgrassella alvi and

Gilliamella apicola for the daughter queens were identi-

cal to the ones described previously. We furthermore

extracted DNA from honey pot content and wax from

two laboratory colonies to screen for the presence of

the target bacteria in the nest environment.

Results

Phylogeny of Snodgrassella alvi, influence of hostphylogeny and geography

In total, we obtained partial 16S rRNA gene sequences

(approx. 1370 bp in length) from 126 different bee spec-

imens for Snodgrassella alvi. Together with the selection

of sequences from previous studies, this resulted in a

data set of 150 samples from 36 different corbiculate

bee host species. Within the tribe Bombini, we sampled

species from 12 of the 15 subgenera of the simplified

classification system by Williams et al. (2008). Our pri-

mer sets detected both Snodgrassella and Gilliamella in

92% of all sampled Bombus species and in all three

Apis species. In several cases, however, not all tested

© 2013 Blackwell Publishing Ltd

2032 H. KOCH ET AL.

individuals within an infected host species screened

positive. A summary of the presence of Snodgrassella

alvi and Gilliamella apicola in samples of all screened bee

species is provided in Table S2 (Supporting informa-

tion).

The phylogenetic tree of Snodgrassella alvi shows a

clear distinction between the sequences from the three

Apis species and the Bombus species (Fig. 1). On a finer

taxonomic scale, considerable separation can also be

observed for the different host species and subgenera.

For example, Apis mellifera bacterial strains from a wide

geographic range (including samples from Europe,

Asia, North America, and Africa) group together in a

monophyletic clade, different from Snodgrassella of other

0.0

5

Bo_

terr

estr

is_0

9.01

8Q1

Pr_flavescens_YX.238

Ml_lapidarius_BJ07.172

Pr_flavescens_YX.239

Bo

_ter

rest

ris_

09.2

25Q

1

HM

1087

34_T

h_s

on

oru

s

Th_i

mpe

tuos

us_M

G.002

691. 80J B_ mur otr oh_g M

Bo_

spor

adic

us_0

8Sf0

84

A_d

ors

ata_

I10.

059

Pr_flavescens_YX.240

Th_c

alifo

rnic

us_U

08.6

05

HM

2150

20_B

o_t

erre

stri

s_D

St_melanurus_KY08.038HM108516_Pr_im

patiens

Ml_lapidarius_EN88

Th_mesomelas_11.001

St_melanurus_KY08.025

HM21

5018

_Bo_

terre

stris

_D

Ps_

bohe

mic

us_B

J08.

075

Bo_terrestris

_09.350Q2

Th

_cal

ifo

rnic

us_

U08

.764

HM215016_Ml_lapidarius_D

Md_defector_KY08.007

Bo_

terr

estr

is_0

9.01

8Q5

Th_i

mpe

tuos

us_M

G.0

01

Ps_campestris_KY08.252

Ml_lapidarius_BJ08.007

Bo_patagiatus_AR.450

A_mel

lifer

a_I1

0.04

7

HM

1120

94_A

_mel

lifer

a_U

SA

Th

_cal

ifo

rnic

us_

U08

.961

St_appositus_U08.609

Or_haemorrhoidalis_I10.009Ml_lapidarius_BJ07.290

Ml_friseanus_YX.247

Md_defector_KY08.211

Al_

balte

atus

_08S

e079

HM

2150

14_T

h_pa

scuo

rum

_D

Pr_hypnorum_BJ08.268

AY37

0189

_A_m

ellif

era_

RSA

Bo_terrestris

_09.013Q2

Bo_

terr

estr

is_0

9.01

8Q3

Pr_m

ixtu

s_U

08.7

50

Or_haemorrhoidalis_I08.030

A_mellifera

_I10.044

Pr_

mix

tus_

U08

.751

Bo

_tu

nic

atu

s_I1

0.02

3

Pr_

mix

tus_

AK

08.5

89

Bo_terrestri

s_09.013Q4

Bo_ter

rest

ris_B

J08.0

26

A_d

ors

ata_

I10.

058

Ml_friseanus_CHN90

Bo

_tu

nic

atu

s_I1

0.03

2

A_mel

lifer

a_I1

0.04

5

HM215011_Ml_lapidarius_CH

Bo_tunicatus_I08.059

Bo

_tu

nic

atu

s_I1

0.03

1

Pr_m

ixtu

s_U08

.850

Th_c

alifo

rnic

us_U

08.3

58Ps_barbutellus_BJ08.320

Kl_soroeensis_08Sc016Pr

_mix

tus_

AK

08.5

90

A_c

eran

a_I1

0.05

1

Th_c

alifo

rnic

us_U

08.6

00

Ml_rufofasciatus_TZ.412

DQ

8376

21_A

_mel

lifer

a_C

H

Bo_

terr

estr

is_0

9.01

8Q4

Bo

_tu

nic

atu

s_I0

8.10

4

HM215017_Mg_hortorum_CH

Pr_f

lavi

frons

_U08

.618

Bo_lucorum_AR.443

HM

2150

13_T

h_pa

scuo

rum

_CH

Bo_terrestris

_09.013Q3

Bo_

luco

rum

_BJ0

8.23

8

Sb_asiaticus_KY08.134

DQ

8376

17_A

_mel

lifer

a_C

HD

Q83

7620

_A_m

ellif

era_

CH

Or_haemorrhoidalis_I08.024

Ml_friseanus_YX.242 Th_impet

uosus_

MG.003

Ml_friseanus_CHN89

Md_defector_KY08.003

A_c

eran

a_I1

0.05

2

Bo_

terr

estr

is_0

9.01

8Q1x

Th_mes

omela

s_11

.002

Th_p

ascu

orum

_11.

007

HM

1084

92_A

_do

rsat

a_M

AL

St_melanurus_I10.018

Ps_campestris_KY08.242

Or_haemorrhoidalis_I10.002

Or_haemorrhoidalis_I08.040

Th_p

ascu

orum

_11.

005

Pr_mixtu

s_AK08.423

Pr_jo

nellu

s_08

Sf050

Ml_festivus_YX.222

Md_defector_KY08.046

Ml_friseanus_YX.246

Pr_j

onel

lus_

08Sf

052

Pr_hypnorum_BJ08.245

Kl_soroeensis_08Sc012

Ml_friseanus_YX.244

Ml_lapidarius_BJ08.021

Pr_melanopygus_U08.253

Pr_melanopygus_U08.350

Pr_jo

nellus_

AK08.28

8

HM21

5021

_Bo_t

erre

stris

_CH

Bo_

terr

estr

is_0

9.01

8Q2

010. 01I _sil adi ohrr o meah_r O

Pr_j

onel

lus_

08Sf

053

St_melanurus_KY08.030

Th

_cal

ifo

rnic

us_

U08

.602

Bo

_tu

nic

atu

s_I1

0.02

9

HM

2150

12_T

h_pa

scuo

rum

_D

HM

2150

15_T

h_pa

scuo

rum

_DTh

_rem

otus

_CHN32

2

Th_p

ascu

orum

_11.

006

Ps_campestris_BJ08.147

Kl_soroeensis_08Sc011

Pr_jo

nellu

s_AK08

.299

HM215010_Bo_terrestris_CH

HM

2150

19_B

o_t

erre

stri

s_D

Bo

_ter

rest

ris_

09.2

25Q

2

HM1132

26_A

_mell

ifera

_USA

A_mel

lifer

a_I1

0.04

6

St_appositus_U08.610

400.01I_siladiohrromeah_rO

Or_haemorrhoidalis_I10.003

DQ

8376

18_A

_mel

lifer

a_C

H

Pr_

vand

ykei

_U08

.117

Pr_v

osne

sens

kii_

U08

.100

HM

0087

20_A

_mel

lifer

a_T

Bo_terrestris_09.350Q1

A_d

ors

ata_

I10.

060

Bo_patagiatus_AR.451

A_mell

ifera

_I10.0

43

Md_defector_KY08.210

Bo

_ter

rest

ris_

B10

.056

030.01I_sutaci

nut_

oB

EF

2129

16_S

ten

oxyb

acte

r_ac

etiv

ora

ns_

(ter

mit

e g

ut)

Ml_rufofasciatus_TZ.411

A_c

eran

a_I1

0.05

0

HM21

5022

_Bo_

terr

estri

s_D

A_c

eran

a_I1

0.05

3

Bo_terrestris

_B10.055

Bo_terrestris

_09.013Q1

Bo_

terr

estr

is_0

9.01

8Q7

Al_

balte

atus

_08S

f048

Ml_festivus_YX.223

Md_defector_KY08.246T

h_c

alif

orn

icu

s_U

08.7

66

52

50

50

100

99

79

93

90

61

86

36

55

75

50

81

6065

62

64

53

71

100

Sb

: S

ibir

ico

bo

mb

us

Or:

Ori

enta

libo

mb

us

Md

: M

end

acib

om

bu

s

Th

: T

ho

raco

bo

mb

us

Mg

: M

egab

om

bu

s

Ps:

Psi

thyr

us

Pr:

Pyr

ob

om

bu

s

Bo

: B

om

bu

s s.

str

.

St:

Su

bte

rran

eob

om

bu

s

Ml:

Mel

ano

bo

mb

us

Kl:

Kal

lob

om

bu

s

Al:

Alp

ino

bo

mb

us

A: A

pis

Eur

ope

Cen

tral

Asi

a

Chi

na

Indi

a

US

A

Fig. 1 Phylogenetic tree (maximum-likelihood) of 16S rRNA gene sequences from Snodgrassella alvi strains of different bee hosts. Sten-

oxybacter acetivorans (termite gut) was selected as the most closely related outgroup taxon for the root, and the root branch length

was shortened for display purposes. Branch labels are bootstrap values over 50%. Colours indicate different bumblebee subgenera

following Williams et al. (2008) (see legend) and different shades of grey represent different Apis species. Symbols code for geo-

graphic origin of each sample (see legend). Tip labels include species name and sample ID (see Table S1, Supporting information).

For further details, see Methods.

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2033

Apis species. The Bombus-associated Snodgrassella of the

host subgenera Thoracobombus, Subterraneobombus, Pyro-

bombus, Bombus s. str., Melanobombus and Orientalibom-

bus are nonrandomly distributed in the tree, and appear

to cluster to some extent according to the host subge-

nus. As an example, Snodgrassella strains from Bombus

(Subterraneobombus) appositus and B. (St.) melanurus fall

together in a well- supported clade despite having been

collected in the USA and India/Kyrgyzstan, respec-

tively, (with the exception of a single strain from B. (St.)

melanurus, which was closer to strains from B. (Th.) mes-

omelas). Some Snodgrassella also cluster according to host

species, for example for B. (Thoracobombus) mesomelas,

B. (Orientalibombus) haemorrhoidalis, B. (Melanobombus)

festivus or B. (Bombus) tunicatus (with the exception of a

single individual).

In some cases, however, less influence of the host

phylogeny can be observed. For example, Snodgrassella

from bumblebees of the subgenera Mendacibombus,

Psithyrus, Pyrobombus, Megabombus, Sibiricobombus, Mela-

nobombus and Kallobombus were each found in multiple

clades, showing weaker trends of host association.

Furthermore, a well-supported, mixed clade of Snodg-

rassella strains from B. (Melanobombus) lapidarius and

B. (Bombus) terrestris/patagiatus/tunicatus was found.

One sample of B. (Pyrobombus) jonellus and B. (Psithyrus)

bohemicus each were nested within Snodgrassella strains

from the subgenus Bombus sensu stricto.

A ‘tanglegram’ linking the tips of the host phylogeny

and the Snodgrassella phylogeny revealed some phyloge-

netic congruency, with clear exceptions in several lin-

eages (Fig. 2). We used partial Mantel tests to test the

relative influence of the host species phylogeny and

the geographical sampling location on the similarity of

the Snodgrassella 16S rRNA gene sequences. After control-

ling for geographical distance between samples, we

found a strong and highly significant positive correlation

between the Snodgrassella phylogeny and the host phy-

logeny (r = 0.680, two tailed P = 0.0001, 10 000 permuta-

tions, n = 132). In contrast, the correlation between the

Snodgrassella phylogeny and geographical distance after

controlling for the host phylogeny was not significant

with little of the variance explained (r = 0.057, P = 0.186).

The ratio of host switches to cospeciation events in

the cophylogenetic analysis with Jane 4 varied from 1.4

to 1.5 for setting 1 (cost = 0 for cospeciations, cost = 1

for all other events, see methods) to 1.7–2.2 for setting 2

(cost = 1 for all events) (Table 1, Figs S1 & S2, Support-

ing information, host n = 37 species, Snodgrassella

n = 80). For both event cost settings, the costs of the

cophylogeny analyses from randomly generated associ-

ations were markedly higher and did not overlap with

Md. defector

Ps. campestrisPs. barbutellus

Pr. flavescens

Mg. hortorum

Sb. asiaticus

Ml. lapidarius

Kl. soroeensisPr. melanopygusPr. impatiens

Pr. mixtusPr. jonellus

Pr. flavifrons

Pr. vosnesenskii

Pr. vandykeiAl. balteatusBo. lucorum

Bo. terrestris

Bo. tunicatus

Ps. bohemicus

Bo. sporadicus

Th. pascuorum

Th. impetuosusTh. remotus

Th. californicusTh. sonorus

Pr. hypnorum

St. melanurus

Th. mesomelas

St. appositus

Bo. patagiatus

Or. haemorrhoidalis

Ml. festivus

Ml. friseanusMl. rufofasciatus

Apis ceranaApis mellifera

Apis dorsata

Bee host Associations SnodgrassellaFig. 2 Tanglegram linking the bee host

phylogeny (left) to the 16S rRNA gene

phylogeny of Snodgrassella alvi (right)

from the bee gut. Associations (grey lines)

link bacterial sequences with the host

species they were obtained from. Host

names indicate bumblebee subgenera

(Williams et al. 2008; see legend Fig. 1),

tanglegram produced in TreeMap 3

(Charleston (2011), http://sites.google.

com/site/cophylogeny/software).

© 2013 Blackwell Publishing Ltd

2034 H. KOCH ET AL.

the cost of the original associations (Table 1, 500 repli-

cates, P < 0.002), thus indicating a significant, if not per-

fect, congruence between the Snodgrassella and host tree.

Phylogeny of Gilliamella apicola, influence of hostphylogeny and geography

We analysed a matrix of 165 partial 16S rRNA gene

sequences (approx. 1290 bp, 138 sequences new to this

study), from 41 different bee host species (12 out of the 15

subgenera (Williams et al. 2008) of the Bombini). In com-

parison with the tree of Snodgrassella alvi (Fig. 1), the tree

of Gilliamella apicola appeared to be less structured

according to host species and host subgenera (Fig. 3).

Importantly, some of the sequences obtained with our

primers from the bumblebee samples appeared to be

closer to (but distinct from) a separate group of A. mellif-

era bacteria previously referred to as Gamma-2 phylotype

and other insect associated bacteria (red branches in

Fig. 3, Martinson et al. 2011). The remaining sequences

belong to Gilliamella apicola as defined by Martinson et al.

(2012) and Kwong & Moran (in press). Gilliamella apicola

strains from the Apini are mostly distinct from the bum-

blebee strains; however, two strains from Apis cerana fall

together with two bumblebee Gilliamella strains with high

bootstrap support. For the Apis-derived strains of Gilliam-

ella apicola, the different Apis host species do not have

completely distinct bacterial strains, as for example one

A. cerana strain falls within a group of A. mellifera sam-

ples, and some A. mellifera samples are closer to A. dorsata

strains than to other A. mellifera strains.

For the Bombus species, Gilliamella strains from the

host species Bombus (Orientalibombus) haemorrhoidalis

and Bombus (Thoracobombus) californicus are clustered in

well-supported, distinct clades. Furthermore, there is

some clustering between samples from the subgenera

Thoracobombus, Pyrobombus and Melanobombus, all of

which, however, do not form exlusive monophyletic

groups. The Gilliamella apicola tree thus appears consid-

erably less structured according to host species and

phylogeny compared with that of Snodgrassella alvi. This

limited degree of specificity also becomes apparent in

the more ‘tangled’ associations of Gilliamella strains and

host species as shown in Fig. 4 compared with the tan-

glegram of Snodgrassella in (Fig. 2).

A partial Mantel test of the Gilliamella and host phy-

logenies, while controlling for geographic distances

between samples, shows a moderate and significantly

positive correlation (including all sequences (see Fig. 3):

r = 0.332, two tailed P = 0.0001, 10 000 replications,

n = 148; excluding phylotypes outside of the main

Gilliamella apicola clade: r = 0.549, P = 0.0001, n = 128).

A significant but weaker correlation between bacterial

phylogeny and geography, while controlling for host

phylogeny, was also found (all sequences: r = 0.142,

P = 0.0002, excluding phylotypes outside of main Gil-

liamella apicola clade: r = 0.1541, P = 0.0003). When com-

pared to Snodgrassella alvi, the correlation between host

and bacterial associate phylogeny was thus weaker for

Gilliamella apicola, while a stronger correlation of bacte-

rial phylogeny and geography was observed.

For the cophylonenetic analysis in Jane 4, the inferred

ratio of host switches to cospeciation events varied from

1.7 to 2.0 for setting 1 (cost = 0 for cospeciations,

cost = 1 for all other events) to 2.8–3.0 for setting 2

(cost = 1 for all events) (Table 1, Figs S3 & S4, Support-

ing information, host n = 34 species, Gilliamella n = 96).

These predicted ratios of host switches per cospeciation

event were thus slightly higher for Gilliamella than for

Snodgrassella. The costs of the cophylogeny analyses

from randomly generated associations were higher than

the original costs without overlap (Table 1, 500 repli-

cates, P < 0.002), suggesting a significant, though

certainly imperfect, congruence between the Gilliamella

and host trees.

Experimental transmission to daughter queens in themother colony

We detected both Snodgrassella alvi and Gilliamella apico-

la with specific 16S rRNA gene primers in the gut of

almost all of the daughter queens sampled within the

Table 1 Results for the cophylogeny analyses with Jane 4 for Snodgrassella and Gilliamella phylogenies mapped on the bee host tree.

Bacterium Scheme # cosp # d # hs # l Cost # hs/# cosp Rand. rep. cost % rep. < orig. cost

Snodgrassella cosp = 0 20–21 4 29–30 3 67 1.4–1.5 89 0 (~ P < 0.002)

Gilliamella cosp = 0 20–23 9–11 40–44 4–11 102 1.7–2.2 121 0 (~ P < 0.002)

Snodgrassella cosp = 1 16–18 5–6 31–32 1–2 87 1.7–2.0 98 0 (~P < 0.002)

Gilliamella cosp = 1 15–16 13–14 45 1 120 2.8–3.0 134 0 (~P < 0.002)

Scheme: cosp = 0: cost = 0 for cospeciation, cost = 1 for all other events, cosp = 1: cost = 1 for all events. # cosp = number of cospeci-

ation events, d = duplications, hs = host switches, l = losses, no ‘failure to diverge’ events were predicted. Cost = optimal cost for

each analysis. # hs/# cosp = ratio between the numbers of host-switching and cospeciation events. Rand. rep. cost = mean cost of

500 replicates with randomly mapped tips.% rep. < orig. cost = percentage of replicates with lower cost than original solution. See

Figs S1–S4 (Supporting information) for solutions of the analyses.

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2035

0.1

Bo_lucorum_AR.443

Th_pascuorum_11.005

Or_haemorrhoidalis_I08.040St_appositus_U08.610

Th

_cal

ifo

rnic

us_

U08

.358

DQ837611_A_mellifera_CH

Ps_bohemicus_BJ08.075

Bo_lucorum_BJ08.238

HM

0465

73_A

_cer

ana_

T

Th_c

alifo

rnic

us_U

08.6

02M

l_la

pid

ariu

s_B

J07.

172

St_appositus_U08.609

Pr_f

lavi

fron

s_U

08.6

59

HM

2150

30_M

g_ho

rtor

um

Md_

defe

ctor

_KY0

8.21

1

HM215028_Th_pascuorum_D

Th

_cal

ifo

rnic

us_

U08

.766Ml_

fest

ivus

_YX.

222

Th

_cal

ifo

rnic

us_

U08

.961

Th_remotus_CHN322

HM108449_A_dorsata_MAL

Bo_

terr

estr

is_0

9.01

8Q4

200. 01I _sil adi ohrr o meah_r O

Pr_

flave

scen

s_Y

X.2

40

Bo_

tuni

catu

s_I0

8.10

4

HM

0087

18_A

_mel

lifer

a_T

Ml_

frise

anus_

YX.247

Or_haemorrhoidalis_I10.010

Pr_melanopygus_U08.350

Bo_sporadicus_08Sf083

Th_mesomelas_11.002

Or_haemorrhoidalis_I10.003

Pr_

vosn

esen

skii_

U08

.026

Mg_hortorum_BJ08.058

HM215026_Th_pascuorum_CH

St_melanurus_KY08.038

Pr_vosnesenskii_U08.100

A_dors

ata_

I10.06

0

Mg_hortorum_BJ07.058

Bo

_tu

nic

atu

s_I1

0.03

2

Bo_

terr

estr

is_0

9.01

3Q2

Bo_terrestris

_09.225Q2

Ml_

fest

ivus

_YX.

221

Bo_

terr

estr

is_B

10.0

56

Bo

_lu

coru

m_B

J08.

236

HM

2150

32_B

o_te

rres

tris

_D

Pr_

mix

tus_

U08

.850

Ml_

frise

anus

_YX.2

44

Ps_bohemicus_KY08.253

Pr_m

elan

opyu

s_U08

.157

Mg_

horto

rum

_BJ0

8.04

9

Pr_vosnesenskii_U08.774

DQ

8376

03_A

_mel

lifer

a_C

H

Ml_

fest

ivus

_YX

.223

Pr_mixtus_AK08.589

Bo

_pat

agia

tus_

AR

.451

Bo_

spor

adic

us_0

8Sf0

84

Ml_

frise

anus

_CHN

90

Ml_lapidarius_BJ08.021

Ml_

frise

anus

_YX.

242

HM

1120

68_A

_mel

lifer

a_U

SA

Bo_

terr

estr

is_B

10.0

55

Ps_ca

mpe

stris

_KY08

.242

Bo_patagiatus_AR.450Pr_caliginosus_U08.601

Bo_tunicatus_I08.155

HM11

2050

_A_m

ellif

era_

USA

HM

2150

25_G

prot

_Bo_

terr

estr

is_D

Bo_tunicatus_I10.030

150.01I _anar ec_

A

FJ612598_Orbus_herc

ynius

Ps_campestris_KY08.252

AY

3701

92_A

_mel

lifer

a_R

SA

Th

_cal

ifo

rnic

us_

U08

.605

EF6085

32_P

_chalc

ites

Pr_v

osne

sens

kii_

U08

.555

Pr_jonellus_08Sf053

Pr_mixtus_AK08.590

Th_mesomelas_11.001

Pr_m

ixtu

s_U

08.1

78

Th_pascuorum_11.006

Pr_

flav

esce

ns_

YX

.238

Pr_hypnorum_BJ08.268

A_m

ellif

era_

I10.

043

Pr_vosnesenskii_U08.337

Ml_rufo

fasciatus_TZ.412

DQ837610_A_mellifera_CH

Pr_

vosn

esen

skii_

U08

.212

Pr_

mix

tus_

U08

.751

Pr_v

osne

sens

kii_

U08

.305

Pr_

flavi

fron

s_U

08.6

57

Pr_hypnorum_08Sa068

HM215024_Th_pascuorum_D

Bo_terrestris

_09.013Q4Bo_te

rres

tris_

BJ08.

011

Ps_fernaldae_AK08.536

M75

076_

Hae

mo

ph

ilus_

par

aph

rop

ha

Bo_terrestris_09.013Q3

HM

1085

42_P

r_im

patie

ns_U

SA

HM11

3304

_A_m

ellif

era_

USA

Pr_

flav

esce

ns_

YX

.239

Al_balteatus_08Sf048

420.80I_siladiohrromeah_rO

HM215035_Bo_terrestris_D

HM

2150

31_M

l_la

pida

rius

Bo_terrestris_09.018Q2

Pr_jonellus_08Sf052

Ml_

lapi

dari

us_D

08.0

75

Ps_bohemicus_BJ08.264

EF18

7248

_A_m

ellif

era_

Sw

Pr_hypnorum_BJ08.245

Bo_terrestris_09.350Q1

A_dorsata_I10.058

Mg_trif

ascia

tus_

I10.01

6

Ps_sp

_B10

.018

Bo_ter

rest

ris_0

9.01

8Qx

AB480768_A_cerana_JP

St_melanurus_KY08.025

A_m

ellif

era_

I10.

045

HM

2150

33_M

l_la

pid

ariu

s_C

H

Kl_

soro

een

sis_

08S

c016

Kl_

soro

eens

is_0

8Sc0

11

Or_haemorrhoidalis_I10.005

FJ6

5551

6_S

_qu

ercu

s

Bo_lucorum_AR.444

Ps_ca

mpe

stris

_BJ0

8.14

7

FJ6

5551

4_C

inar

a_sp

Pr_

vand

ykei

_U08

.137

Bo_terrestris

_BJ08.026

050.01I_anarec_

A

Ml_

lapi

dari

us_E

N88

Th_pascuorum_11.007

DQ83

7608

_A_m

ellif

era_

CH

Bo

_tu

nic

atu

s_I1

0.02

9

Pr_b

ifariu

s_U08

.252

Bo_te

rres

tris_

BJ08.

033

Ps_fernaldae_AK08.212

Bo_terrestris_09.225Q1

EF1872

49_A

_mel

lifer

a_Sw

EF187250_A_mellifera_Sw

Bo_terrestris_09.013Q1

AB

0508

2_E

dw

ard

siel

la_h

osh

inae

Pr_mixtus_U08.750

Ml_

rufo

fasc

iatu

s_TZ

.413

HM

2150

29_B

o_t

erre

stri

s_C

H

Bo_te

rres

tris_

09.0

18Q

1

Bo_terre

stris_

09.01

8Q3

Bo_ter

rest

ris_0

9.018

Q5

Kl_

soro

een

sis_

08S

c012

Ml_

frise

anus

_YX.2

45

Th_impetuosus_MG.003

Or_haemorrhoidalis_I10.009

Or_haemorrhoidalis_I08.030St_melanurus_I10.018

Ml_fris

eanus_

CHN89

Kl_

soro

een

sis_

08S

c015

Md_defector_KY08.246

Cu

_gri

seo

colli

s_U

08.6

16

Th_impetuosus_MG.002

Bo_patagiatus_AR.452

A_dors

ata_

I10.06

1

Bo_tunicatus_I10.031

St_melanurus_KY08.030

Bo

_tu

nic

atu

s_I0

8.05

9

A_m

ellif

era_

I10.

044

Ml_friseanus_YX.246

A_m

ellif

era_

I10.

047

Th

_cal

ifo

rnic

us_

U08

.600

HM215027_Th_pascuorum_D

Th_impetuosus_MG.001

Ml_rufo

fasc

iatus_

TZ.411

57

77

99

75

99

80

54

93

62

99

66

100

09

69

51

5297

59

100

100

64

99

93

98

88

97

97

53

57

97

84

99

52

54

97

96

54

77

100

56 100

78

79

74

61

53

Eur

ope

Cen

tral

Asi

a

Chi

na

Indi

a

US

AO

r: O

rien

talib

om

bu

s

Md

: M

end

acib

om

bu

s

Th

: T

ho

raco

bo

mb

us

Mg

: M

egab

om

bu

s

Ps:

Psi

thyr

us

Pr:

Pyr

ob

om

bu

s

Bo

: B

om

bu

s s.

str

.

St:

Su

bte

rran

eob

om

bu

s

Ml:

Mel

ano

bo

mb

us

Kl:

Kal

lob

om

bu

sC

u:

Cu

llum

ano

bo

mb

us

Al:

Alp

ino

bo

mb

us

A:

Ap

is

Fig. 3 Phylogenetic tree (maximum-likelihood) of 16S rRNA gene sequences from Gilliamella apicola strains of different bee hosts.

Sequences outside the main Gilliamella apicola clade and closer to the Gamma-2 phylotype (sensu Martinson et al. 2011) are displayed

with dark red branches. Haemophilus paraphropha and Edwardsiella hoshinae were selected as most closely related outgroup taxa for the

root. Branch labels are bootstrap values over 50%. Colours indicate different bumblebee subgenera following Williams et al. (2008)

(see legend) and different shades of grey represent different Apis species. Symbols code for geographic origin of each sample (see leg-

end). Tip labels include species name and sample ID (see Table S1, Supporting information). For further details, see Methods.

© 2013 Blackwell Publishing Ltd

2036 H. KOCH ET AL.

field-placed B. terrestris colonies (Snodgrassella: 15/16

positive, Gilliamella: 16/16 positive). The 16S rRNA gene

sequences from Snodgrassella appeared to be distinct for

the four colonies, however, with overlap between

daughter queens from colonies 09.350 and 09.013 (Fig.

S5, Supporting information). The pattern was less clear

for Gilliamella sequences, with strains from all four colo-

nies being mixed in a phylogenetic tree, with only col-

ony 09.018 showing a distinct genotype for some of the

bacterial strains (Fig. S5, Supporting information). As

these bees remained inside of the colony environment

from eclosion until sampling, our findings indicate a

within-colony transfer of these elements of the gut mic-

robiota under natural conditions.

We further examined the transmission and mainte-

nance during hibernation of the target bacteria in labo-

ratory-reared daughter queens. Daughter queens that

had interacted as adults with other individuals within

the colony all tested positive for Snodgrassella (5/5) or

Gilliamella (5/5) before hibernation. After hibernation,

all surviving daughter queens tested positive for Snodg-

rassella (5/5), while only some of the daughter queens

harboured Gilliamella (2/5). In the group kept in perma-

nent isolation after pupal emergence, only very few

individuals tested positive for the presence of the target

bacteria (Snodgrassella: 1/5 before hibernation, 2/5 after

hibernation; Gilliamella: 1/5 before hibernation, 0/5 after

hibernation). Unfortunately, sample sizes after hiberna-

tion were rather low because out of the 40 daughter

queens originally put into hibernation an unusually low

number of only five individuals from each group (isola-

tion/colony contact) had survived. The three worker

offspring individuals produced by two of the colony

contact queens shared the infection pattern of their

mother queen with Snodgrassella being present and

Gilliamella absent. We also screened samples of honey

pot content and wax from two different laboratory colo-

nies with specific PCRs. Both honey pot samples and

both wax samples were positive for Gilliamella, while no

amplification was observed for Snodgrassella.

Phylogeny of Meliponini and Euglossini bacteria

In our sample of two species of stingless bees and three

species of orchid bees, no positive amplification for the

targeted bacterial species Snodgrassella alvi and Gilliamel-

la apicola was observed with specific primers (Table S2,

Supporting information). A control PCR with universal

eubacterial 16S rRNA gene primers on the bee gut sam-

ples, however, resulted in strong amplification of a

Ml. lapidarius

Kl. soroeensis

Bo. terrestrisBo. tunicatus

Mg. hortorum

Pr. mixtus

Pr. impatiens

Md. defector

Ps. bohemicus

Al. balteatusPr. hypnorum

Bo. lucorumBo. patagiatus

Ps. campestris

Pr. jonellus

Pr. vosnesenskiiPr. melanopygusPr. caliginosus

Ps. fernaldae

Bo. sporadicus

A. ceranaA. dorsata

A. mellifera

Th. californicus

Cu. griseocollisMl. festivus

Ml. rufofasciatusMl. friseanus

Th. impetuosus

Th. pascuorumTh. remotus

Or. haemorrhoidalis

Th. mesomelas

St. melanurusSt. appositus

Mg. trifasciatus

Pr. bifarius

Pr. flavifronsPr. vandykei

Pr. flavescens

Host Associations GilliamellaFig. 4 Tanglegram linking the bee host

phylogeny (left) to the 16S rRNA gene

phylogeny of Gilliamella apicola from the

bee gut (right). Associations (grey lines)

link bacterial sequences with the host

species they were obtained from. Host

names indicate bee genus/subgenus (Wil-

liams et al. 2008; see legend Fig. 3).

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2037

product of the expected size (approx. 1.5 kb) indicating

sufficient quality of the DNA template. A negative

result with the specific 16S rRNA gene primers could

also have come from primer mismatches to Snodgrassella

and Gilliamella in the Meliponini and Euglossini. There-

fore, we cloned the amplification product of the PCR

with universal eubacterial 16S rRNA gene primers for a

selection of samples from both Euglossini and Melipo-

nini. A total of 24 randomly chosen clones were picked

and sequenced for each sample. After removing chime-

ric sequences, bacterial 16S rRNA gene sequences for a

total of 170 clones were obtained, of which 86 were

from the Meliponini and 84 from the Euglossini sam-

ples. None of the clones was similar to Snodgrassella alvi

or Gilliamella apicola from the Bombini and Apini. We

selected clones of distinct sequences from each clone

library for sequencing of the full vector insert.

From these tests, the phylogenetic position of the gut

bacteria of the Meliponini and Euglossini, in compari-

son with their closest described relatives and other pre-

viously sequenced bacteria from bees, is shown in

Fig. 5. In particular, all three Euglossini species had

high numbers of clones from the Acetobacteraceae,

forming a distinct clade related to other Acetobactera-

ceae previously found in flowers (Saccharibacter floricola,

see Jojima et al. 2004) and in other insects including

honeybees and a solitary bee (Colletidae: Caupolicana

yarrowi, around 3–4% divergent to alpha 2.2 sensu

Martinson et al. 2011). The Eulaema sp. sample also har-

boured a bacterium from the Enterobacteraceae close to

Yokenella regensburgi. Both Euglossa imperialis individuals

also had high numbers of clones of an unknown Alpha-

proteobacterium. These clones were closest to a bacte-

rium from the hemipteran Dactylopius opuntiae, but also

showed similarity to a bacterium of the isopod Porcelio

scaber (‘Candidatus Hepatincola porcellionum’) and a

bacterium from Daphnia magna.

Both species of Meliponini were also hosts to differ-

ent strains of Acetobacteraceae. In addition, a wide

variety of bacteria from the order Lactobacillales were

found, including strains close to the genera Streptococ-

cus, Leuconostoc, Aerococcus and Lactobacillus. Some of

these lactobacilli were closely related to bacteria previ-

ously found in honeybees.

Discussion

Snodgrassella alvi and Gilliamella apicola gut bacteriaof bumblebees and honeybees

We obtained and analysed 16S rRNA gene sequences of

bumblebees and honeybees covering most of the phylo-

genetic diversity of the hosts and spanning a wide geo-

graphical range over three continents (North America,

Europe and Asia). The vast majority of bumblebee spe-

cies (92%) and all honeybee species screened positive

for Gilliamella and Snodgrassella.

Our data reject a strict cospeciation of Snodgrassella

and Gilliamella with their bee hosts. The association

between these bacteria and bees appears to allow for a

considerable degree of host switching, contrasting with

the strict cospeciation pattern observed for some obli-

gate intracellular insect symbionts (Clark et al. 2000).

Both Mantel tests and reconciliation analyses suggest a

complex history with a number of possible host

switches. For example, we observed Snodgrassella geno-

types in bumblebees of the subgenus Mendacibombus

identical with genotypes in other distantly related spe-

cies [e.g. Bombus (Pyrobombus) flavescens]. These host

species have probably diverged more than 30 million

years ago (Hines 2008), which would lead to noticeable

16S rRNA gene sequence divergence between bacterial

strains in the case of strict cospeciation (Ochman et al.

1999). Nevertheless, for both Snodgrassella and Gilliamel-

la, we found a significant correlation between the 16S

rRNA gene bacterial phylogeny and the host phylog-

eny. The reconciliation analyses furthermore suggest

the possibility for a number of cospeciation events. In

contrast, geographic distance of the sampled bees corre-

lated only weakly with the Gilliamella tree and not at all

with the Snodgrassella tree. This suggests a close associa-

tion of both bacteria with their hosts, which has been

maintained over large geographical distances and long

periods of time.

Several mechanisms could underlie the topological

congruence of the host and Snodgrassella/Gilliamella

trees. For example, bacterial associates could be more or

less exclusively vertically transmitted (Clark et al. 2000;

Hosokawa et al. 2006). We provide experimental evi-

dence that vertical transmission from the mother colony

to young queens is possible, although the frequency of

horizontal transmission in the field remains to be deter-

mined. The presence of vertical transmission in this sys-

tem may allow for co-evolutionary dynamics in which

host and bacteria exert reciprocal selection on each

other. There is some evidence for a beneficial function

of Snodgrassella and Gilliamella in parasite defence for

Bombus terrestris (Koch & Schmid-Hempel 2011b, 2012),

but whether these members of the microbiota have fit-

ness effects on different honeybee and bumblebee

species remains to be investigated. In this light, it

remains possible that the bacteria have evolved host

specificity without exerting a selective pressure on the

host as commensalistic passengers in the bee gut. A

purely commensalistic relationship would, however, be

somewhat surprising given the high prevalence of both

bacteria across the different targeted bee species and

geographic region. Other explanations for nonrandom

© 2013 Blackwell Publishing Ltd

2038 H. KOCH ET AL.

topological matches between host and bacterial trees

should also be considered. For example, host-switching

event may predominantly occur between closely related

hosts, leading to a match of host and bacterial phylog-

eny in the absence of detectable cospeciation events

(De Vienne et al. 2007).

When compared to Snodgrassella alvi, it emerged that

the association of Gilliamella apicola appears to be less

specific. While we still detect a significant correlation

between host and bacterial associate phylogeny, with

some clades of Gilliamella exclusively restricted to cer-

tain host species (e.g. Bombus haemorrhoidalis, Bombus

0.2

EU034639 Gluconobacter oxydans

GU125609 Lactobacillus sakei

GQ853368 Alphaproteobacterium (Dactylopius opuntiae)

JQ389885 Melipona panamica (B10.043)

AB681877 Yokenella regensburgeri

FJ025127 Sinorhizobium meliloti

EF187245 Lactobacillus (A. mellifera)

JQ389893 Melipona panamica (B10.042)

HM534760 Lactobacillus (Melipona beecheii)

EU096234 Gluconobacter morbifer (Drosophila)

EF187239 Lactobacillus (A. mellifera)

HM534763 Lactobacillus (Melipona beecheii)

HM113169 Lactobacillus (A. mellifera)

NR_024819 Saccharibacter floricola (flower)

HM534761 Lactobacillus (Trigona sp.)

JN392909 Alphaproteobacterium (Daphnia magna)

HQ425688 Aerococcus viridans

AY370183 Lactobacillus (A. mellifera)

HM534779 Lactobacillus (A. cerana)

JQ389891 Meliponula bocandei (B10.046)

Y11374 Lactobacillus kunkeei

HM534811 Lactobacillus (A. florea)

AY189806 “Ca. Hepatincola porcellionum”

AB499842 Acetobacter pasteurianusAB513363 Aceteobacteraceae (flower)

AY773947 Lactobacillus acidophilus

AF009487 Streptococcus suis (pig)

JQ389880 Euglossa imperialis (B10.023)

JQ389890 Melipona panamica (B10.043)

JQ389892 Melipona panamica (B10.043)

HM108484 Acetobacteraceae (A. dorsata)

EF187244 Lactobacillus (A. mellifera)

AB678443 Gluconobacter frateurii

X95421 Lactobacillus lindneri

JQ389894 Meliponula bocandei (B10.046)

AJ971850 Acetobacteraceae (A. mellifera)

DQ068792 Enterobacteriaceae (Myrmeleon)

JN644571 Aerococcus (Culex quinquefasciatus)

HM112107 Lactobacillus (A. mellifera)

JQ389879 Eulaema sp. (B10.049)

HM534814 Lactobacillus (Trigona sp.)

JQ389889 Melipona panamica (B10.042)

AM157443 Streptococcus (human)

JQ389884 Eulaema sp. (B10.049)

HM046576 Lactobacillus (A. cerana)

DQ837625 Acetobacteraceae (A. mellifera)

AB680295 Streptococcus equinus (horse)

AJ419836 Acetobacter peroxydans

HM534764 Lactobacillus (A. koschevnikovi)

AB362721 Leuconostoc citreum

DQ860016 Streptococcus (anchovi)

HM109591 Acetobacteraceae (Caupolicana yarrowi)

AB682268 Klebsiella oxytoca

JQ389887 Meliponula bocandei (B10.046)

AB429371 Lactobacillus kefiri

HM534762 Lactobacillus (Meliponula bocandei)

HM215048 Lactobacillus (B. terrestris)

HM534813 Lactobacillus (Meliponula bocandei)

HM534806 Lactobacillus (Melipona beecheii)

U11014 Rickettsia bellii

JQ389888 Melipona panamica (B10.042)

JQ389883 Eufrisea sp. (B10.050)

JQ389881 Euglossa imperialis (B10.022)

AB326299 Leuconostoc pseudomesenteroides

JQ389895 Melipona panamica (B10.043)

HM112629 Leuconostoc (Diadasia opuntiae)

JQ389882 Euglossa imperialis (B10.022)

EU096229 Acetobacter pomorum (Drosophila)

EF179826 Enterobacteriaceae (Anopheles)

NR_042194 Lactobacillus tucceti

JQ389886 Meliponula bocandei (B10.047)

AB680259 Aerococcus urinaeequi

53

94

96

100

98

94

100

100

78

90

100

100

62

100

100

91

82

92

68

100

99

75

100

100

98

100

100

85

100

74

100

98

97

92

54 97

100

86

92

98

100

69

100

66

100

95

100

100

57

93

88

71

74

90

68

100

Fig. 5 Phylogenetic tree of bacterial 16S rRNA gene sequences from stingless bees and orchid bees. Tip label colours indicate isola-

tion source: from the gut of stingless bees (green), orchid bees (blue), honey bees (red), other bee species (purple) and bacterial type

strains/isolates from nonbee hosts (black). Tip labels include GenBank Accession nos, host species and bee specimen sample num-

bers for sequences obtained from clone libraries in this study. The tree was computed using maximum-likelihood and was midpoint

rooted; branch labels indicate bootstrap values over 50% (for further details, see Methods).

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2039

californicus), inferred host switches appear to be more

common both within honeybees and bumblebees. Fur-

thermore, the influence of geographical sampling loca-

tion appears to be stronger for these bacteria and,

hence, horizontal transfer of these bacteria through

environmental sources may be more common (see

‘Transmission mode’ section of the Discussion).

Potentially, clades within the Snodgrassella and Gil-

liamella tree may also differ in their host specificity,

with some clades apparently present in a number of

distantly related host species, while others seem highly

specialized to distinct hosts. Therefore, considerable

variation may exist in the evolutionary dynamics within

both bacterial taxa, leading to a complex pattern from

frequent host switches to specialized interactions.

Some limitations are associated with the approach of

direct sequencing of 16S rRNA gene fragments in our

study. For example, the 16S rRNA gene is highly con-

served, whereas bumblebees are a relatively recently

diverged group with extant species probably sharing

their last common ancestor only about 35 million years

ago, and speciation within the different subgenera

occurring mostly within the last 15 million years (Hines

2008). The 16S rRNA gene alone may furthermore con-

siderably underestimate the true genetic strain diversity

in the bee gut (Engel et al. 2012). Therefore, an

approach including additional molecular markers such

as multilocus sequence typing (MLST) may be needed

to resolve potential specificity among gut bacteria of

recently diverged hosts (Falush et al. 2003; Oh et al.

2010; Kwong & Moran in press). However, as we never-

theless find strong evidence for a correlation between

host and bacteria phylogeny even at the level of low

resolution with the 16S rRNA gene, this pattern can be

assumed to be robust. As an additional caveat, our

approach of directly sequencing PCR products obtained

with specific primers may bias our analysis towards the

majority strain present in a particular sample, while

rarer strains will be overlooked. Our approach can thus

not exclude the existence of additional rare strains with

different patterns of codiversification. In Apis mellifera,

several strains of Snodgrassella and Gilliamella with dis-

tinct 16S rRNA gene sequences can be present in one

individual (Moran et al. 2012), although individuals in

general harboured a single dominant phylotype. Gener-

ally, in our study, directly sequenced PCR products

gave clear chromatograms of good quality. In some

cases, however, ambiguous bases were detected and

coded as such in the phylogenetic analysis, indicating

the presence of different strains in one individual (0.2%

ambiguous bases for Gilliamella, 0.06% for Snodgrassella).

In addition, queens of single colonies also had multiple,

nonmonophyletic strains (Fig. S5), pointing to within-

colony strain diversity. The extent of strain variation of

Snodgrassella and Gilliamella at the scale of individual

bees thus remains to be investigated.

Transmission mode

We show that both Snodgrassella and Gilliamella are

transmitted to daughter queens within the mother col-

ony in the field. Furthermore, we show that in a labora-

tory environment, social contact of daughter queens

after pupal emergence increases the chances of transmis-

sion of these bacteria and that bacteria can persist in the

daughter queens during hibernation. Similarly, Martin-

son et al. (2012) showed that in honeybees these bacteria

are acquired by newly emerged workers within the col-

ony environment. A comparison of the two bacterial

taxa appears to be in line with the findings of the

across-host species comparison in this study, which

showed a higher degree of specificity for Snodgrassella.

In our small sample, we observed distinct genotypes of

Snodgrassella strains in daughter queens from different

colonies, while this pattern was not apparent for Gilliam-

ella strains. Furthermore, hibernating daughter queens

may be more likely to lose Gilliamella than Snodgrassella

during hibernation. While all daughter queens in our

laboratory sample that had spent time in their mother

colony after emergence harboured both Snodgrassella

and Gilliamella, some of the daughter queens screened

after hibernation lacked Gilliamella, while Snodgrassella

were still present in all individuals. The detection of Gil-

liamella in the content of nectar pots and on wax surfaces

within the nest and their growth on standard culture

media, in contrast to Snodgrassella (Olofsson & V�asquez

2008; Koch & Schmid-Hempel 2011b), further suggests

that Gilliamella is less intimately associated with their

hosts and present in the hive environment.

Taken together, these results suggest that both Snodg-

rassella and Gilliamella can be transmitted ‘vertically’

from the mother colony to the young gynes. Gilliamella,

however, may be transmitted horizontally outside of the

nest more frequently than Snodgrassella. Potentially, the

horizontal transmission could occur on flowers. Recent

studies have shown a considerable diversity of bumble-

bee-transmitted yeasts in flowers (Brysch-Herzberg

2004; Herrera et al. 2008). In a study focusing on lactoba-

cilli associated with bees (McFrederick et al. 2012), flow-

ers were suggested as transmission sites especially for

Lactobacillus spp. of halictid bees. Bacterial communities

in flowers have so far, however, been very poorly stud-

ied, and the few available surveys did not report the

bacteria that we focused on in this study (Junker et al.

2011; Fridman et al. 2012). Bees may furthermore acquire

bacteria horizontally through direct contact with other

bee species. For example, we detected a Snodgrassella

strain from the cuckoo bumblebee Bombus (Psithyrus)

© 2013 Blackwell Publishing Ltd

2040 H. KOCH ET AL.

bohemicus, nested within bacterial strains of its host spe-

cies, that is, members of the subgenus Bombus s.str. As

B. bohemicus parasitizes nests of B. lucorum (Alford 1975)

without producing a worker caste on its own, it may

potentially have acquired bacteria from its host colony.

In conclusion, the relative importance of horizontal

and vertical transmission in the field for the two

groups of target bacteria thus remains to be investigated

further.

Bacteria in Euglossini and Meliponini

Only few studies have previously looked at microor-

ganisms of stingless bees (Meliponini). Rosa et al. (2003)

found a variety of yeasts, Gilliam et al. (1985, 1990)

described several Bacillus spp., and V�asquez et al. (2012)

focussed on lactic acid bacteria associated with stingless

bees. There has not been, to our knowledge, any study

of microorganisms associated with orchid bees (Euglos-

sini) to date.

In our survey of two stingless bee and three orchid

bee species, we detected none of the Snodgrassella alvi or

Gilliamella apicola bacteria commonly found in bumble-

bees and honeybees. This suggests that these microbial

associates may be either rare or absent in the Euglossini

and Meliponini and are potentially unique to the Apini

and Bombini.

Bacteria from the Acetobacteraceae appear to be com-

mon members of the gut microbiota of the Euglossini

and Meliponini. Acetic acid bacteria have previously

been found to be common in the guts of insects relying

on a sugar-based diet, including honeybees, fruit flies,

and mealy bugs (Crotti et al. 2010). They also occur in

flowers (Jojima et al. 2004), suggesting the possibility of

environmental transmission to bees through their food

source. In addition to members of the Acetobacteraceae,

the orchid bee Euglossa imperialis harboured a novel Al-

phaproteobacterium, so far not recorded from bees, and

related to the isopod associated bacterium ‘Candidatus

Hepatincola porcellionum’ (Rickettsiales) (Wang et al.

2004).

The stingless bee gut microbiota furthermore contains

a variety of lactic acid bacteria (Lactobacillales), includ-

ing undescribed members of the genera Lactobacillus,

Leuconostoc, Streptococcus and Aerococcus. In this respect,

the stingless bee microbiota appears to be similar to the

microbiota of honeybees and to a lesser degree also

bumblebees (Olofsson & V�asquez 2008, 2009; Koch &

Schmid-Hempel 2011a; Martinson et al. 2011; V�asquez

et al. 2012). In honeybees, these lactic acid bacteria have

been suggested to play a beneficial role in protection

against pathogens (Forsgren et al. 2010; V�asquez et al.

2012). In line with the findings of V�asquez et al. (2012),

lactic acid bacteria may in conclusion present hitherto

neglected but important associates of stingless bees.

Acknowledgements

G. Cisarovsky, D.W. Roubik, B. Sadd, R. Schmid-Hempel,

E. Stolle, E. Tesch and M. Tognazzo provided help with collect-

ing bee samples (see Table S1, Supporting information). S. Jost,

M. Berchtold and E. Karaus helped with molecular cloning

and sequencing. T. Stadler gave helpful advice on statistical

analyses. Sequence data were generated in the Genetic Diver-

sity Centre (GDC) of ETH Zurich. The study was supported by

grants of the Swiss SNF (grant no. 31003A-116057 to PSH &

no. 140157 to HK), an ERC Advanced Grant (no. 268853 to

PSH), and a Sino-Swiss Science and Technology Cooperation

Agreement (EG13-032009 and 2009DFA32600 to PSH and JL).

References

Alford DV (1975) Bumblebees. Davis-Poynter, London, UK.

Babendreier D, Joller D, Romeis J, Bigler F, Widmer F (2007)

Bacterial community structures in honeybee intestines and

their response to two insecticidal proteins. FEMS Microbiol-

ogy Ecology, 59, 600–610.Bevins CL, Salzman NH (2011) The potter’s wheel: the host’s

role in sculpting its microbiota. Cellular and Molecular Life

Sciences, 68, 3675–3685.

Brucker RM, Bordenstein SR (2012) The roles of host evolution-

ary relationships (Genus:Nasonia) and development in struc-

turing microbial communities. Evolution, 66, 349–362.Brysch-Herzberg M (2004) Ecology of yeasts in plant–bumble-

bee mutualism in Central Europe. FEMS Microbiology Ecol-

ogy, 50, 87–100.

Cameron SA, Hines HM, Williams PH (2007) A comprehensive

phylogeny of the bumble bees (Bombus). Biological Journal of

the Linnean Society, 91, 161–188.Chapman RE, Wang J, Bourke AFG (2003) Genetic analysis of

spatial foraging patterns and resource sharing in bumble bee

pollinators. Molecular Ecology, 12, 2801–2808.Charleston MA (2011) TreeMap 3b. Available from: http://

sites.google.com/site/cophylogeny.

Clark MA, Moran NA, Baumann P, Wernegreen JJ (2000)

Cospeciation between bacterial endosymbionts (Buchnera)

and a recent radiation of aphids (Uroleucon) and pitfalls of

testing for phylogenetic congruence. Evolution, 54, 517–525.Cole JR, Wang Q, Cardenas E et al. (2009) The Ribosomal Data-

base Project: improved alignments and new tools for rRNA

analysis. Nucleic Acids Research, 37, D141–D145.

Conow C, Fielder D, Ovadia Y, Libeskind-Hadas R (2010) Jane:

a new tool for the cophylogeny reconstruction problem.

Algorithms for Molecular Biology, 5, 16.

Crotti E, Rizzi A, Chouaia B et al. (2010) Acetic acid bacteria,

newly emerging symbionts of insects. Applied and Environ-

mental Microbiology, 76, 6963–6970.

Cruaud A, Rønsted N, Chantarasuwan B et al. (2012) An

extreme case of plant-insect co-diversification: figs and fig-

pollinating wasps. Systematic Biology, 61, 1029–1047.De Vienne DM, Giraud T, Shykoff JA (2007) When can host

shifts produce congruent host and parasite phylogenies? A

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2041

simulation approach. Journal of Evolutionary Biology, 20, 1428–

1438.

Decelle J, Probert I, Bittner L et al. (2012) An original mode of

symbiosis in open ocean plankton. Proceedings of the National

Academy of Science, USA, 109, 18000–18005.

Dillon RJ, Dillon VM (2004) The gut bacteria of insects: non-

pathogenic interactions. Annual Review of Entomology, 49,

71–92.Disayathanoowat T, Young JP, Helgason T, Chantawannakul P

(2012) T-RFLP analysis of bacterial communities in the mid-

guts of Apis mellifera and Apis cerana honey bees in Thailand.

FEMS Microbiology Ecology, 79, 273–281.Dunlap PV, Ast JC, Kimura S, Fukui A, Yoshino T, Endo H

(2007) Phylogenetic analysis of host–symbiont specificity and

codivergence in bioluminescent symbioses. Cladistics, 23,

507–532.Durrer S, Schmid-Hempel P (1994) Shared use of flowers leads

to horizontal pathogen transmission. Proceedings of the Royal

Society London B, 258, 299–302.

Engel P, Martinson VC, Moran NA (2012) Functional diversity

within the simple gut microbiota of the honey bee. Proceedings

of the National Academy of Science, USA, 109, 11002–11007.Ersts PJ (2012) Geographic Distance Matrix Generator (version

1.2.3). American Museum of Natural History, Center for

Biodiversity and Conservation, available from http://biodi-

versityinformatics.amnh.org/open_source/gdmg.

Falush D, Wirth T, Linz B et al. (2003) Traces of human migra-

tions in Helicobacter pylori populations. Science, 299, 1582–

1585.

Feldhaar H (2011) Bacterial symbionts as mediators of ecologi-

cally important traits of insect hosts. Ecological Entomology,

36, 533–543.

Forsgren E, Olofsson TC, V�asquez A, Ingemar Fries I (2010)

Novel lactic acid bacteria inhibiting Paenibacillus larvae in

honey bee larvae. Apidologie, 41, 99–108.Fraune S, Bosch TCG (2007) Long-term maintenance of species-

specific bacterial microbiota in the basal metazoan Hydra.

Proceedings of the National Academy of Science, USA, 104, 13146

–13151.Fraune S, Bosch TCG (2010) Why bacteria matter in animal

development and evolution. BioEssays, 32, 571–580.Frese SA, Benson AK, Tannock GW et al. (2011) The evolution

of host specialization in the vertebrate gut symbiont Lactoba-

cillus reuteri. PLoS Genetics, 7, e1001314.

Fridman S, Izhaki I, Gerchman Y, Halpern M (2012) Bacterial

communities in floral nectar. Environmental Microbiology

Reports, 4, 97–104.Gilliam M (1971) Microbial sterility of the intestinal content of

the immature honey bee, Apis mellifera. Annals of the Entomo-

logical Society of America, 64, 315–316.

Gilliam M, Buchmann SL, Lorenz BJ, Roubik DW (1985) Micro-

biology of the larval provisions of the stingless bee, Trigona

hypogea, an obligate necrophage. Biotropica, 17, 28–31.Gilliam M, Roubik DW, Lorenz BJ (1990) Microorganisms asso-

ciated with pollen, honey, and brood provisions in the nest

of a stingless bee, Melipona fasciata. Apidologie, 21, 89–97.

Gordon DM (2001) Geographical structure and host specificity

in bacteria and the implications for tracing the source of coli-

form contamination. Microbiology, 147, 1079–1085.Gordon DM, Lee J (1999) The genetic structure of enteric bacte-

ria from Australian mammals. Microbiology, 145, 2673–2682.

Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W,

Gascuel O (2010) New algorithms and methods to estimate

maximum-likelihood phylogenies: assessing the performance

of PhyML 3.0. Systematic Biology, 59, 307–321.Herrera CM, Garc�ıa IM, P�erez R (2008) Invisible floral larce-

nies: microbial communities degrade floral nectar of bumble

bee-pollinated plants. Ecology, 89, 2369–2376.

Hines HM (2008) Historical biogeography, divergence times,

and diversification patterns of bumble bees (Hymenoptera:

Apidae: Bombus). Systematic Biology, 57, 58–75.Hosokawa T, Kikuchi Y, Nikoh N, Shimada M, Fukatsu T

(2006) Strict host-symbiont cospeciation and reductive gen-

ome evolution in insect gut bacteria. PLoS Biology, 4, e337.

Huber T, Faulkner G, Hugenholtz P (2004) Bellerophon; a pro-

gram to detect chimeric sequences in multiple sequence

alignments. Bioinformatics, 20, 2317–2319.Jeyaprakash A, Hoy MA, Allsopp MH (2003) Bacterial diver-

sity in worker adults of Apis mellifera capensis and Apis

mellifera scutellata (Insecta: Hymenoptera) assessed using

16S rRNA gene sequences. Journal of Invertebrate Pathology,

84, 96–103.

Jojima Y, Mihara Y, Suzuki S, Yokozeki K, Yamanaka S, Fudou

R (2004) Saccharibacter floricola gen. nov., sp. nov., a novel

osmophilic acetic acid bacterium isolated from pollen. Inter-

national Journal of Systematic and Evolutionary Microbiology, 54,

2263–2267.Junker RR, Loewel C, Gross R, D€otterl S, Keller A, Bl€uthgen N

(2011) Composition of epiphytic bacterial communities dif-

fers on petals and leaves. Plant Biology, 13, 918–924.Kikuchi Y, Hosokawa T, Fukatsu T (2007) Insect-microbe

mutualism without vertical transmission: a stinkbug acquires

a beneficial gut symbiont from the environment every gener-

ation. Applied and Environmental Microbiology, 73, 4308–4316.Kikuchi Y, Hosokawa T, Nikoh N, Meng XY, Kamagata Y,

Fukatsu T (2009) Host-symbiont co-speciation and reductive

genome evolution in gut symbiotic bacteria of acanthosoma-

tid stinkbugs. BMC Biology, 7, 2.

Kikuchi Y, Hosokawa T, Fukatsu T (2011) An ancient but pro-

miscuous host–symbiont association between Burkholderia gut

symbionts and their heteropteran hosts. The ISME Journal, 5,

446–460.Koch H, Schmid-Hempel P (2011a) Bacterial communities in

central European bumblebees: low diversity and high speci-

ficity. Microbial Ecology, 62, 121–133.

Koch H, Schmid-Hempel P (2011b) Socially transmitted gut

microbiota protect bumble bees against an intestinal parasite.

Proceedings of the National Academy of Science, USA, 108,

19288–19292.

Koch H, Schmid-Hempel P (2012) Gut microbiota instead of

host genotype drive the specificity in the interaction of a nat-

ural host-parasite system. Ecology Letters, 15, 1095–1103.Koch H, Cisarovsky G, Schmid-Hempel P (2012) Ecological

effects on gut bacterial communities in wild bumblebee colo-

nies. Journal of Animal Ecology, 81, 1202–1210.

Kwong WK, Moran NA Cultivation and characterization of

the gut symbionts of honey bees and bumble bees: Snodg-

rassella alvi gen. nov., sp. nov., a member of the Neisseria-

ceae family of the Betaproteobacteria; and Gilliamella apicola

gen. nov., sp. nov., a member of Orbaceae fam. nov., Orbales

ord. nov., a sister taxon to the Enterobacteriales order of the

Gammaproteobacteria. International Journal of Systematic and

© 2013 Blackwell Publishing Ltd

2042 H. KOCH ET AL.

Evolutionary Microbiology, doi: 10.1099/ljs.0.044875-0, epub

ahead of print.

Legendre P, Troussellier M (1988) Aquatic heterotrophic bacte-

ria: modeling in the presence of spatial autocorrelation. Lim-

nology and Oceanography, 33, 1055–1067.

Ley RE, Hamady M, Lozupone C et al. (2008) Evolution of

mammals and their gut microbes. Science, 320, 1647–1651.

Mantel N (1967) The detection of disease clustering and a gen-

eralized regression approach. Cancer Research, 27, 209–220.

Martinson VG, Danforth BN, Minckley RL, Rueppell O, Tingek

S, Moran NA (2011) A simple and distinctive microbiota

associated with honey bees and bumble bees. Molecular Ecol-

ogy, 20, 619–628.

Martinson VG, Moy J, Moran NA (2012) Establishment of char-

acteristic gut bacteria during development of the honey bee

worker. Applied and Environmental Microbiology, 78, 2830–2840.McFrederick QS, Wcislo WT, Taylor DR, Ishak HD, Dowd SE,

Mueller UG (2012) Environment or kin: whence do bees

obtain acidophilic bacteria? Molecular Ecology, 21, 1754–1768.

Mendlov�a M, Desdevises Y, Civ�a�nov�a K, Pariselle A, �Simkov�a

A (2012) Monogeneans of West African cichlid fish: evolu-

tion and cophylogenetic interactions. PLoS ONE, 7, e37268.

Mohr KI, Tebbe CC (2006) Diversity and phylotype consistency

of bacteria in the guts of three bee species (Apoidea) at an

oilseed rape field. Environmental Microbiology, 8, 258–272.

Moran NA, Hansen AK, Powell JE, Sabree ZL (2012) Distinc-

tive gut microbiota of honey bees assessed using deep sam-

pling from individual worker bees. PLoS ONE, 7, e36393.

Ochman H, Elwyn S, Moran NA (1999) Calibrating bacterial

evolution. Proceedings of the National Academy of Science, USA,

96, 12638–12643.Ochman H, Worobey M, Kuo CH et al. (2010) Evolutionary

relationships of wild hominids recapitulated by gut micro-

bial communities. PLoS Biology, 8, e1000546.

Oh PL, Benson AK, Peterson DA et al. (2010) Diversification of

the gut symbiont Lactobacillus reuteri as a result of host-driven

evolution. The ISME Journal, 4, 377–387.Olofsson TC, V�asquez A (2008) Detection and identification of a

novel lactic acid bacterial flora within the honey stomach of

the honeybee Apis mellifera. Current Microbiology, 57, 356–363.

Olofsson TC, V�asquez A (2009) Phylogenetic comparison of

bacteria isolated from the honey stomachs of honey bees

Apis mellifera and bumble bees Bombus spp. Journal of Apicul-

tural Research, 48, 233–237.

Paradis E, Claude J, Strimmer K (2004) APE: analyses of phy-

logenetics and evolution in R language. Bioinformatics, 20,

289–290.Posada D (2008) jModelTest: phylogenetic model averaging.

Molecular Biology and Evolution, 25, 1253–1256.R Development Core Team (2011) R: A Language and Environment

For Statistical Computing. R Foundation for Statistical Comput-

ing, Vienna, Austria. URL: http://www.R-project.org.

Ram�ırez SR, Nieh JC, Quental TB, Roubik DW, Imperatriz-

Fonseca VL, Pierce NE (2010) A molecular phylogeny of the

stingless bee genus Melipona (Hymenoptera: Apidae). Molec-

ular Phylogenetics and Evolution, 56, 519–525.

Rosa CA, LachanceMA, Silva JOC et al. (2003) Yeast communities

associatedwith stingless bees. FEMSYeast Research, 4, 271–275.

Rosenberg MS, Anderson CD (2011) PASSaGE: Pattern Analy-

sis, Spatial Statistics and Geographic Exegesis. Version 2.

Methods in Ecology and Evolution, 2, 229–232.

Sachs JL, Essenberg CJ, Turcotte MM (2011) New paradigms

for the evolution of beneficial infections. Trends in Ecology &

Evolution, 26, 202–209.

Sanderson MJ, Donoghue MJ, Piel W, Eriksson T (1994) Tree-

BASE: a prototype database of phylogenetic analyses and an

interactive tool for browsing the phylogeny of life. American

Journal of Botany, 81, 183.

Thompson JD, Higgins DG, Gibson TJ (1994) CLUSTAL W:

improving the sensitivity of progressive multiple sequence

alignment through sequence weighting, position-specific gap

penalties and weight matrix choice. Nucleic Acids Research,

22, 4673–4680.V�asquez A, Forsgren E, Fries I et al. (2012) symbionts as major

modulators of insect health: lactic acid bacteria and honey-

bees. PLoS ONE, 7, e33188.

Wang Y, Stingl U, Anton-Erxleben F, Zimmer M, Brune A

(2004) ‘Candidatus Hepatincola porcellionum’ gen. nov., sp.

nov., a new, stalk-forming lineage of Rickettsiales colonizing

the midgut glands of a terrestrial isopod. Archives of Microbi-

ology, 181, 299–304.Williams PH, Cameron SA, Hines HM, Cederberg B, Rasmont

P (2008) A simplified subgeneric classification of the bumble-

bees (genus Bombus). Apidologie, 39, 46–74.

Zamborsky DJ, Nishiguchi MK (2011) Phylogeographical pat-

terns among mediterranean sepiolid squids and their Vibrio

symbionts: environment drives specificity among sympatric

species. Applied and Environmental Microbiology, 77, 642–649.

Zilber-Rosenberg I, Rosenberg E (2008) Role of microorgan-

isms in the evolution of animals and plants: the hologe-

nome theory of evolution. FEMS Microbiology Review, 32,

723–735.

H.K. and P.S.-H. designed research, H.K. performed

laboratory experiments, P.S.-H., H.K., J.L. and D.P.A.

organized collection trips and contributed bee speci-

mens, H.K. analysed data and wrote the paper, H.K.,

P.S.-H. and J.L. contributed to revisions of the original

manuscript.

Data accessibility

DNA sequences: GenBank Accessions JQ389879–

JQ390033 & JQ405085–JQ405210. Alignment of 16S

rRNA gene phylogenies: Treebank submission ID 13472.

Sample information (sample ID, species identification,

sampling locality, collector, corresponding GenBank

Accession nos) in Table S1 (Supporting information).

Supporting information

Additional supporting information may be found in the online ver-

sion of this article.

Fig. S1 Result of reconciliation analysis with Jane (Conow et al.

2010) for Snodgrassella alvi and bee host phylogenies for the

event cost setting cosp. = 0.

© 2013 Blackwell Publishing Ltd

EVOLUTION OF CORBICULATE BEE BACTERIA 2043

Fig. S2 Result of reconciliation analysis for Snodgrassella alvi

and bee host phylogenies for the event cost setting cosp. = 1.

Fig. S3 Result of reconciliation analysis for Gilliamella apicola

and bee host phylogenies for the event cost setting cosp. = 0.

Fig. S4 Result of reconciliation analysis for Gilliamella apicola

and bee host phylogenies for the event cost setting cosp. = 1.

Fig. S5 Unrooted phylogenetic trees (maximum-likelihood) of

16S rRNA gene sequences of Snodgrassella alvi and Gilliamella

apicola strains of the gut from recently emerged (1–7 days old)

young queens collected inside four different Bombus terrestris

field colonies.

Table S1 List of all honey bee and bumble bee specimens

with bee species identification, sample ID, sampling location/

country, sampling date, coordinates in decimal degrees (DD),

collector (BS: Ben Sadd, DPA: Dharam P. Abrol, DWR: David

W. Roubik, ES: Eckart Stolle, HK: Hauke Koch, JL: Jilian Li,

MT: Martina Tognazzo, RSH: Regula Schmid-Hempel) and

GenBank Accession nos of 16S rRNA gene sequences for

Snodgrassella and Gilliamella.

Table S2 Results of the screen of all corbiculate bee species

samples (Table S1, Supporting information) with specific 16S

rRNA gene primers for the target bacteria Snodgrassella alvi

and Gilliamella apicola.

© 2013 Blackwell Publishing Ltd

2044 H. KOCH ET AL.