1 Introduction

Insects easily outnumber other animal clades in the world inhabiting a diverse range of ecosystems. Insects are, in many cases, dependent on functions provided by bacterial symbionts (Brownlie and Johnson 2009; Basset et al. 2012), including gut microbiota with beneficial functions in nutrition (Salem et al; 2012) and protection (Dillon and Charnley 1986, 1988). Some insects harbor gut communities of specific bacteria, which is especially true in social insects (Engel and Moran 2013). Within the clade of social bees, research on microbial communities has been focusing on honeybees (Martinson et al. 2012; Moran et al. 2012) and bumblebees (Koch and Schmid-Hempel 2011b; Koch et al. 2012; Meeus et al. 2013; Li et al. 2015; Praet et al. 2015) because of their important pollinating role in natural ecosystems and of many agricultural crops, together with a growing evidence of their dramatic declines worldwide (Ghazoul 2005; Lebuhn et al. 2013).

The social nesting behavior of honeybees and bumblebees enables the transfer of gut bacteria over the generations, allowing a close association with the host (Powell et al. 2014). Indeed, social bees, as opposed to solitary bees, generally harbor host-associated bacterial groups, defined as “core set,” and the latter are generally found in the gut microbiota of social bees (Kaltenpoth 2011; Koch and Schmid-Hempel 2011b; Cariveau et al. 2014). Core bacteria dominating the gut of social honeybees and bumblebees have been implicated in beneficial functions such as digestion and protection (Klungness and Peng 1984; Forsgren et al. 2010; Koch et al. 2012; Vásquez et al. 2012; Kwong et al. 2014). Multiple conditions have been reported influencing the bacterial gut community, including the environment, host age, habitat, pathogens, and diet (Dillon et al. 2010; Evans and Schwarz 2011; Hu et al. 2014; Yun et al. 2014).

Currently, the buff-tailed bumblebee Bombus terrestris is reared indoors for the biological pollination in agriculture. This species is widely distributed and native to Europe, coastal North Africa, and West and Central Asia (Goulson 2010). Due to the decline of bees worldwide, research has been focused on diseases (Fürst et al. 2014) and pathogen spillover (Meeus et al. 2011). Besides, there has been little research on the gut microbial community and the dynamics of indoor-reared bumblebees foraging in an environment outdoors.

In this paper, we used 16S rRNA Illumina deep sequencing to identify the gut microbial community in workers of B. terrestris from indoor-reared nests. In a first experiment, we determined the growth and richness of this microbial community when the nests were kept under indoor conditions. Our ambitions were to determine the dominant bacteria in indoor-reared nests and the stability of these bacteria in function of the age of the adults and the nest. In a second experiment, the indoor-reared bumblebee nests were moved to outdoors to investigate the plasticity of the gut microbial community in the workers of different ages. In summary, the new data found in this study are of use to discuss the relationship between core and non-core bacteria in bumblebees and to generate new insights in the gut microbial stability of indoor-reared bumblebee hives when moved to outdoors.

2 Material and methods

2.1 Bumblebee nests and insect labeling

Six nests of B. terrestris were obtained from a bumblebee mass-rearing facility (Biobest, Westerlo, Belgium). All colonies contained one queen, new queen brood but without adult workers. The nests were kept in the lab at standard conditions of 28–30 °C and 60–65 % relative humidity and continuous darkness, and they were fed ad libitum with artificial nectar (Biogluc®, Biobest) and pollen clumps (Soc. Coop. Apihurdes, Pinofranqueado-Caceres, Spain) as energy and protein sources, respectively (Mommaerts et al. 2006). For the six bumblebee nests, we labeled the queen and then also each newborn worker adult on a daily basis with a unique marking disc (Bijenhof, Bissegem, Belgium) to the dorsal thorax. The six nests were used for two experiments where in the first experiment, three nests were kept indoors under the standard conditions as described earlier, while in the second experiment, the other three nests were moved outdoors. The experiments were carried out between end August up to end September and early October 2012.

2.2 Experiment 1: bumblebee nests kept under standard conditions indoors

Three nests were followed for 60 days during the rearing period in the lab under standard conditions. We aimed to investigate the impact of age of the nest and the individual bumblebee worker in the nest. Therefore, we defined two age periods of the nest: young nest (Yn; 4- to 11-day old nest) and old nest (On; 40- to 46-day old nest) and two age periods of the workers: young bee (Yb; 4 to 10 days old) and old bee (Ob; 28 to 35 days old). When combining the age period of the nest with the age of the individual workers, we defined three test groups: young nest, young bee (YnYb); old nest, old bee (OnOb); and old nest, young bee (OnYb). In total, we sampled eight workers per test group. Sampled bumblebees were stored individually at −70 °C until dissection of the gut.

2.3 Experiment 2: indoor-reared bumblebee nests moved to an outdoor environment

In parallel to experiment 1, we kept the three other nests indoors in the lab so that they developed an initial indoor-reared microbiota. When placed outdoors, the three nests developed further under natural conditions during 50 days. The outdoors study site (51° 01ʹ 32.57ʺ N; 3° 42ʹ 46.10ʺ O) was located in a residential area in the city of Ghent (Belgium) with landscape metrics of an urbanized environment (Parmentier et al. 2014). As shown in Figure S1, each nest was placed in a separate polystyrene-box (25 × 35 × 25 cm) with the entrance directed to the east and further treated as described in Parmentier et al. (2014).

In the outdoors experiment, we defined three age periods of the nests: young nest (Yn, 4- to 11-day old nest); middle-aged nest (Mn, 28- to 34-day old nest); and old nest (On, 40- to 46-day old nest), and three age periods of the workers: young bumblebee (Yb, 4 to 10 days old); middle-aged bumblebee (Mb, 18 to 23 days old); and old bumblebee (Ob, 28 to 35 days old). Here, we defined “young” and “old” ages identical as in experiment 1, and the extra “middle” age was introduced in order to sample sufficient forager and nesting bumblebees. We defined a “forager” and “nesting” bumblebee as a worker typically showing a nesting and foraging function, respectively. Thus, in total, we defined five test groups: young nest, young bumblebee (YnYb); old nest, old bumblebee (OnOb); old nest, young bumblebee (OnYb); middle-aged nest, young bumblebee (MnYb); and middle-aged nest, middle-aged bumblebee (MnMb). Following our definition, at nest level, initial sampling in the young nests YnYb was achieved under indoor conditions (same as indoor-reared hives in experiment 1), whereas sampling in all other test groups was achieved outdoors. In total, we sampled 7 ± 1 nesting bees per test group YnYb, OnOb, and OnYb. In the test groups MnYb and MnMb, 6 ± 2 nesting and 6 ± 2 foraging workers were sampled per test group, respectively. Sampled bumblebees were stored individually at −70 °C until dissection of the gut.

To follow and record individually foraging workers leaving or entering the nest, we placed a webcam (Logitech® C920 HD Pro, Lausanne, Switzerland) with a top view on the entrance of each nest and optimized in “move-detection” function (Online Resource Figure S1). The flight duration of individual foragers was calculated and we ranked them daily based on the numbers of flights under good weather conditions. We selected foraging workers (foragers) that had at least five flights per day under good weather conditions (excluding rainy days and temperatures <15 °C).

2.4 Gut samples processing, Illumina sequencing, and identification

The gut, including the crop, was dissected from the different bumblebee workers and transferred individually into a 1.5-mL Eppendorf tube. All equipment used was disinfected with 75 % ethanol between each individual dissection. The extraction of bacterial DNA was done as previously described (Meeus et al. 2013), and then the V4 region of the ribosomal RNA (rRNA) was amplified in triplicate, using the 515F and 806R primers elongated with multiplex identifiers (Caporaso et al. 2011). The composite primers include the 16S primer sites, a different nucleotide barcode (on the 806R primer) and the Illumina adapter sequences necessary for bridge amplification on the Illumina MiSeq flow cell (Illumina, San Diego, CA). Amplicons were normalized, mixed at equimolar concentrations, purified, and further concentrated as previously described (Meeus et al. 2015) to produce a 8-pM sequencing library. To increase sequence diversity, 20 % denaturated Illumina PhiX Control V3 library was admixed. Cluster generation and 2 × 150 paired-end sequencing was performed on one Illumina Miseq flowcell using an Illumina Miseq Reagent Kit v.2. Custom sequencing primers were added in a final concentration of 0.5 μM (Meeus et al. 2015). Base calling and primary quality assessments and de-multiplexing were performed at the Miseq sequencing platform of Laboratory for Pharmaceutical Biotechnology (UGent) using Illumina’s Basespace genomics cloud computing environment.

The pooled data set contained 3,353,558 double read sequences. A unique barcode was used to separate each sequence to its sample of origin. Sequences were analyzed with the Mothur software v.1.31.1 following the standard procedure (Schloss et al. 2009). In brief, the complexity was reduced by a training set of base pairs before clustering into operational taxonomic units (OTUs). The procedure includes denoising of unique sequences (preclustering all sequences with 1 mutation on 100 bp) and removing of chimeras using the UCHIME algorithm (Edgar et al. 2011). The distance matrix was calculated and clustering with a 0.03 cutoff level resulted in 461 OTUs. To exclude sequencing errors, we only kept these OTUs that had more than 0.5 % of the sequence reads in every sample. This procedure resulted in 20 OTUs, representing 99 % of the total reads.

2.5 Classification and verification of new OTUs

The taxonomic identity of each OTU was achieved by alignment of each sequence with the Bacterial SILVA SEED database (training set), supplemented with host-specific sequences (i.e., host Bombus or Apis) to improve classification (Newton and Roeselers 2012). The identity of each bacterial OTU was achieved by a two-way strategy. First, BLASTn searches of the representative sequences were achieved. Second, when a 100 % base pair alignment was not obtained with representative sequences of Bombus hosts of a previous dataset (Meeus et al. 2015) (available at Genbank KM03050545 to KM03050553), a semi-nested PCR with a universal Eub8F or 984yR primer combined with an OTU-specific primer was applied. All new sequences were deposited at Genbank (KP410382 to KP410393) and raw Illumina data reads of all samples were submitted to SRA database of Genbank under project ID PRJNA270053 (Tables I, II).

Table I Taxonomic identification of OTUs, names used, their closest matching strain, and associated source according to GenBank and Papers.
Full size table
Table II OTU-specific primers combined with a universal 16S rDNA primer (Eub8F or 984yR).
Full size table

In our study, we used a classification system of bacterial OTUs to separate core and non-core OTUs (Koch and Schmid-Hempel 2011a; Koch et al. 2013; Cariveau et al. 2014). We defined Neisseriaceae, Orbaceae, Lactobacillaceae, and Bifidobacteriaceae as core families for bumblebees in accordance with Kwong et al. (2014) and Meeus et al. (2015). All other bacteria are regarded as non-core OTUs.

2.6 Statistical analysis

Samples were normalized to the smallest number of reads for a given sample (n = 49,420) using the normalize.shared function (Schloss et al. 2009) and then used to generate the diversity calculators. Calculators “Sobs” (observed number of species), “Chao1” and the Shannon index were used to map the alpha diversity (Schloss et al. 2009). Differences in alpha diversity between pre-defined test groups of indoor-reared and outdoor-developed nests were determined by the non-parametric Mann-Whitney U test (SPSS vs. 21). Significance cutoff values were set at 95 % (α = 0.05). Differences in similarity between test groups (even sampling) were calculated by analysis of similarities (ANOSIM). To visualize differences in the bacterial community (beta diversity), non-metric multidimensional scaling (NMDS) was used based on a Bray-Curtis similarity matrix of the square root transformed normalized relative abundance of the different OTUs per sample (Primer 6 version 6.1.10). An in-depth multivariate analysis was performed in order to evaluate dispersion effects between test groups by using the R package “mvabund” selecting the “anova.manyglm” call. After checking model assumptions by plotting “meanvar” and “manyglm” matrixes, data was further processed as a “negative binomial” distribution, a typical option for abundance count data in the mvabund-package (Wang et al. 2012).

3 Results

3.1 Workers of indoor-reared nests harbored a small microbial gut community of stable composition

In total, we sampled 24 nesting bumblebee workers from three indoor-reared nests, having representative samples of bumblebees in three age groups (YnYb, OnYb, OnOb). Rarefaction curves (Figure S2 in the supplemental material) illustrate that an asymptote was achieved for 49,000 sequence read per sample and seven samples per test group, indicating sufficient sequencing depth. After OTU picking procedure, the microbial community of the indoor-reared workers (experiment 1 and initial indoor-reared bumblebees in experiment 2) clustered in 13 OTUs. Figure 1a shows the relative abundance of each OTU per test group. Most of the sequences are classified under known core OTUs (Table I), i.e., Snodgrasella (Snod.) classified under Neisseriaceae, Gilliamella (Gillia.) and Candidatus Schmidhempelia (Ca. Schmid.) belonging to Orbaceae, Lacto1-Firm4, Lacto2-Firm5, Firm-B1, Firm-E, Firm-W, which are all Firmicutes, and Lacto3 classified under Lactobacillaceae, and Bifido-1 belonging to the Bifidobacteriaceae. A very small number of the reads belonging to Gamma-E1 and Gamma-E4 (both <0.5 %), was classified under Enterobacteriaceae. When looking at alpha diversity calculators such as community richness (Sobs and Chao1) and community diversity (Shannon index) (Fig. 1b), no differences were observed between the test groups in the indoor-reared nests (Mann-Whitney U test; P values are given in Table III). Besides, looking at the 2-dimensional NMDS plots (STRESS1 = 0.16) which is an indicator for beta diversity (Fig. 1c), no differences between test groups (ANOSIM, R = 0.11, P = 0.51) were observed, which was also confirmed by multiple univariate test statistics (all P values >0.05, Table III). However, an in-depth multivariate analysis (MANOVA results using “mvabund” package, Table IV) estimated that there exist small differences (Dev = 22.4, P = 0.045) between the groups, indicating small dispersion effects between young and old bumblebee workers.

Figure 1.
figure 1

a Relative abundance of all OTUs comparing groups YnYb, OnOb, and OnYb of the indoor-reared B. terrestris nests. b Alpha diversity box plots (Sobs, Chao1, and Shannon index). c NMDS plots. P = 0.05; Mann-Whitney U.

Full size image
Table III P values (Mann-Whitney U test, a = 0.05) of alpha diversity indicators Sobs, Chao, and Shannon index between test groups YnYb, OnYb, and OnOb for lab-reared hives.
Full size table
Table IV P values and Wald test scores of gut microbiota OTUs, multivariate and univariate comparison between test groups YcYb, OcOb, and OcYb of indoor-reared hives (P > 0.05; mvabund package. anova.manyglm test).
Full size table

3.2 Newborn workers showed a gut microbiota shift with a major colonization of Enterobacteriaceae when the nests were moved to outdoors

In the second experiment, indoor-reared nests were moved to outdoors, and we sampled again 24 nesting bumblebee workers for the three test groups with sufficient depth (Figure S3). The microbial community in the initial indoor-reared workers YnYb developed a typical microbiota as seen in experiment 1 (Figs. 1a and 2a) but workers from the nests moved to outdoors shifted completely after 6 weeks and new OTUs were taken up (Fig. 2a). Data analysis revealed 20 bacterial OTUs as summarized in Table I. Looking to alpha diversity (Fig. 2b), we detected no differences in community richness (Sobs: P = 0.306 and Chao1: P = 0.507), but a significant difference in community diversity (Shannon: P = 0.002) in the group of young workers born in the old nest (OnYb). Indeed, the evenness of the gut microbiota is different in these young bumblebees due to the major uptake of new OTUs, especially Gamma-E2. Multivariate statistics (anova.manyglm) demonstrated a clear dispersion effect between the test groups (Dev = 585,665, P = 0.001). For the beta diversity (Fig. 2c), as represented in the NMDS plot (STRESS = 0.04), we saw a clear separate cluster (ANOSIM, R = 0.99, P = 0.001), grouping all workers that were newborn in the old nest (OnYb).

Figure 2.
figure 2

a Relative abundance of all OTUs comparing test groups YnYb, OnOb, and OnYb of indoor-reared B. terrestris nests moved to an outdoors environment. b Alpha diversity box plots (Sobs, Chao1, and Shannon index). c NMDS plots. P = 0.05; Mann-Whitney U.

Full size image

When looking to differences between foraging and nesting workers from the nest moved to outdoors, we sampled 14 bumblebees with an equal number of foraging and nesting workers (test groups MnYb and MnMb) with sufficient depth (Figure S4 in supplementary material). Although we observed a significant difference between the age categories MnYb and MnMb (anova.manyglm; Dev = 204,923; P = 0.023), both univariate and multivariate statistics calculated no significant differences between foraging and nesting workers or between nests of both ages (univariate: all P values >0.05, multivariate: anova.manyglm; Dev = 89,929, P = 0.289; see Table S1 in the supplementary material).

Finally, when comparing indoor-reared nests with nests that were moved to outdoors, we observed 11 additional OTUs in the gut microbiota, here identified as Alphaproteobacteria (Alpha-2.1), Enterobacteriaceae (Gamma-E3, Gamma-E4, Gamma-P1), Lactobacillaceae (Lacto-3, Firm-S1, Firm-S2, Firm-E, Firm-B1, Firm-V), and Bifidobacteriaceae (Bifido-4). Especially, one OTU had become very dominant, namely Gamma-E3 with a BLASTn match that is close to the environmental bacteria Rahnella sp. (Genbank JQ522978.1, >99 % identity) and Yersinia kristensenii (Genbank JN167946.1, 100 % identity) (Table I). Remarkably, we observed a dominance of Enterobacteriaceae in all shifted profiles. Indeed, considering all age groups in the old nests outdoors, NMDS visualization showed a cluster of all young bumblebees (OnYb) together with one old bumblebee (OnOb) and two bumblebees of middle age (OnMb) that all showed shifted profiles (ANOSIM, R = 0.99, P = 0.005, data not shown). Moreover, a relative dominance of Enterobacteriaceae correlated with a drop in the relative abundance of the core OTUs, as was clear in the individual profiles of all age groups (Figure S5).

4 Discussion

In this study, we used Illumina deep sequencing technology of the 16S rRNA gene to investigate the gut microbial communities of bumblebee workers in nests reared indoors under controlled conditions in the lab in comparison to nests moved to outdoors after an initial indoor rearing period. We demonstrated that, in indoor-reared nest, the gut microbial community is stable over 6 weeks and all age groups are composed of a small number of identical core bacterial taxa (OTUs) classified under Neisseriaceae, Orbaceae, Lactobacillaceae, and Bifidobacteriaceae which can be grouped under core bacteria in bumblebees (Table I). This observation is in agreement with Meeus et al. (2015) who found that reared B. terrestris nests showed a subset of core bacteria compared to wild bumblebees. Therefore, our in-depth analysis confirms a homeostasis of OTUs and negligible inter- or intra-nest variability when bumblebees are fed pollen and artificial nectar ad libitum and then kept under constant conditions. A homeostasis in gut microbiota of artificial well-fed bumblebees has already been reported, even when bumblebees were primed with diseases (Koch and Schmid-Hempel 2012). Therefore, our results indicate that age of the nest and worker is of minor importance when assessing the core microbiota in the gut of bumblebees under controlled conditions.

Second, we investigated in-depth the gut microbiota of indoor-reared nests moved to an outdoors environment and observed that the alpha diversity calculators showed no significant differences, whereas the beta diversity calculators differed significantly compared to indoor-reared nests. Here, we discuss our results with a study on social and herbivorous (pollen-feeding) Cephalotes varians ants (Hu et al. 2014). In this study, a comparison between laboratory-reared and field-caught workers revealed significant differences in the beta diversity of the gut microbiota. However, it should be remarked that, although the alpha diversity did not change between the bumblebee nests that were moved to outdoors and those that were kept indoors in our experiments, the outdoor-developed bumblebees showed extra OTUs, i.e., Firm-B1, Firm-E, Firm-W, Lacto3, Gamma-E1, and Gamma-E4. Therefore, a possible increase in alpha diversity outdoors could be hidden because the extra OTUs are small and dispersed, as was also the case in the C. varians study (Hu et al. 2014).

In this study, we observed that the profiles can change, mainly due to the colonization of Gamma-E3. A first hypothesis might be that the new bacteria were already present in low numbers in the nest of indoor-reared bumblebees, but that they were suppressed due to constant rearing conditions. Indeed, the outside environment itself with different nectar and pollen sources and other less optimal conditions is very different from the constant rearing conditions indoors promoting a small and stable microbiota (Meeus et al. 2015). In agreement, a shift in the gut bacterial diversity has been demonstrated in starving desert locusts Schistocerca gregaria compared to normal fed individuals (Dillon et al. 2010). Although, a second hypothesis remains possible as our results indicated that most of the new OTUs, including Gamma-E3, have been linked with environmental sources (BLASTn results shown in Table I). Thus, the colonization of Gamma-E3 can also be caused by the uptake of natural bacteria from the environment, probably by foraging workers.

Because the foraging workers had a more direct contact with the new outdoor environment compared to nesting workers, we expected that the former would have a microbiota that shifted earlier compared to the latter. However, in this study, we showed that the microbial community in the foraging workers did not shift earlier than the nesting workers of the same age but that the shift was mainly observed in the new workers (indifferent of their function) born in the old nest. We speculate that, when the workers had an established microbiota typically present in the young or middle-aged nest, they were less susceptible to take up new microbiota. Nonetheless, we believe that the foraging bumblebees can still bring the new bacteria from the outdoor environment into the nest, and this in turn resulted in a new inoculation in the old nest, that produced the newest generation of bumblebees.

Importantly, we observed that shifted profiles showed a high abundance of Enterobacteriaceae at the expense of the core bacteria Snodgrassella sp. and Gilliamella sp. (Fig. 2a). We showed that the Enterobacteriaceae bacteria are not exotic and most of the OTUs correspond with Genbank entries linked with hosts as Apis, Bombus, and other insects (Table I) (Anderson et al. 2013; Engel et al. 2013; Killer et al. 2011). We believe this is an important observation because Enterobacteriaceae are not regarded as core gut microbiota in social bees. However, it should be remarked that both Cariveau et al. (2014) and Meeus et al. (2015) also reported that the relative abundance of Enterobacteriaceae can be dominant (relative abundance up to 90 %) in some wild B. bimaculatus bumblebees and wild collected B. terrestris, respectively. Nevertheless, an Enterobacteriaceae-dominated gut microbial community can have its influence on the fitness of the bumblebee host or its resilience against pathogens (Cariveau et al. 2014). Indeed, a disruption in the homeostasis of mutualistic core bacteria has been linked with dysbiosis and associated with pathogens (Hamdi et al. 2011; Koch and Schmid-Hempel 2012) and this may have happened also in the young workers in the old nests that were moved to outdoors. However, one should take care before making this conclusion not investigating variations in functions between individual bacteria species classified under the same family as was also reported by Hu et al. (2014) in the case of C. varians ants. In our case, the large uptake of Enterobacteriaceae does not necessarily indicate a state of dysbiosis but can also contribute to a host nutritional function, as has been reported in other insects including fruit flies and beetles (Behar et al. 2005; Vasanthakumar et al. 2006). Indeed, our BLASTn searches indicated close matches with Rahnella sp. (>99 % identity) and Y. kristensenii (100 % identity). The latter bacterial species have been isolated from wild flowers from a nature reserve and are associated with Apis mellifera (Vásquez et al. 2012), while the former has been isolated from the gut of the willow leaf beetle Plagiodera versicolora (Demirci et al. 2013). Rahnella aquatilis has also been isolated from the gut of other beetles such as longicorn beetles (Park et al. 2007; Podgwaite et al. 2013) and Vasanthakumar et al. (2006) described that this bacterium plays an important nitrogen-fixing role for developing larvae in the southern pine beetle Dendroctonus frontalis. Thus, we believe that the large uptake of Enterobacteriaceae can also contribute to a host nutritional function in bumblebees. However, further research is necessary before firm conclusions can be drawn. Today, there exist no specific studies dealing with the regarded non-core Enterobacteriaceae in bumblebees or other bee taxa. Since these bacteria are present in the gut of European as well as North American bumblebee species, further research is needed in this family, in relation to host functioning, next to a further interest in the functions of mutualistic core and environmental non-core bee gut microbiota.