Journal of Animal Ecology 2012
doi: 10.1111/j.1365-2656.2012.02004.x
Ecological effects on gut bacterial communities in wild bumblebee colonies Hauke Koch*, Gabriel Cisarovsky and Paul Schmid-Hempel ETH Zu¨rich, Institute of Integrative Biology, Universita¨tstrasse 16, 8092 Zu¨rich, Switzerland
Summary 1. Animal hosts harbour diverse and often specific bacterial communities (microbiota) in their gut. These microbiota can provide crucial services to the host such as aiding in digestion of food and immune defence. However, the ecological factors correlating with and eventually shaping these microbiota under natural conditions are poorly understood. 2. Bumblebees have recently been shown to possess simple and highly specific microbiota. We here examine the dynamics of these microbiota in field colonies of the bumblebee Bombus terrestris over one season. The gut bacteria were assessed with culture-independent methods, that is, with terminal restriction fragment length profiles of the 16S rRNA gene. 3. To further understand the factors that affect the microbiota, we experimentally manipulated field-placed colonies in a fully factorial experiment by providing additional food or by priming the workers’ immune system by injecting heat-killed bacteria. We furthermore looked at possible correlates of diversity and composition of the microbiota for (i) natural infections with the microbial parasites Crithidia bombi and Nosema bombi, (ii) bumblebee worker size, (iii) colony identity, and (iv) colony age. 4. We found an increase in diversity of the microbiota in individuals naturally infected with either C. bombi or N. bombi. Crithidia bombi infections, however, appear to be only indirectly linked with higher microbial diversity when comparing colonies. The treatments of priming the immune system with heat-killed bacteria and additional food supply, as well as host body size, had no effect on the diversity or composition of the microbiota. Host colony identity had only a weak effect on the composition of the microbiota at the level of resolution of our method. We found both significant increases and decreases in the relative abundance of selected bacterial taxa over the season. 5. We present the first study on the ecological dynamics of gut microbiota in bumblebees and identify parasite infections, colony identity and colony age as important factors influencing the diversity and composition of the bacterial communities. The absence of an effect of our otherwise effective experimental treatments suggests a remarkable ability of the host to maintain a homoeostasis in this community under widely different environments. Key-words: field experiment, immune system, microsporidia, social insects, species–area relationship, specificity, symbiont, Trypanosomatidae Introduction Specific microbial communities (microbiota) colonizing animal hosts are ubiquitous in nature and can play important roles, such as for digestive functions, immune defence and organ formation (Fraune & Bosch 2010). However, we know little about the dynamics and the ecological factors shaping these communities in the field. The field situation is of course especially interesting, as it can differ fundamentally from the situation in laboratory-reared hosts (e.g. Xiang et al. 2006). As a case study, we here investigate the bacterial microbiota *Correspondence author. E-mail:
[email protected]
of important pollinators in temperate regions, the bumblebees (Bombus spp.) (Bingham & Orthner 1998). The case is likely of practical relevance, as bumblebee populations have been declining world-wide (Goulson, Lye & Darvill 2008; Williams & Osborne 2009), and this has been linked to pathogens (Williams & Osborne 2009; Cameron et al. 2011). With regard to their microbiota, recent evidence shows that bumblebees and honeybees have a distinct and relatively speciespoor microbiota in the gut (Koch & Schmid-Hempel 2011a; Martinson et al. 2011). These bacteria might play an important role in defence against pathogens (Forsgren et al. 2010; Koch & Schmid-Hempel 2011b) and the digestion of food (Gilliam 1997; Va´squez & Olofsson 2009). But the field situa-
2012 The Authors. Journal of Animal Ecology 2012 British Ecological Society
2 H. Koch, G. Cisarovsky & P. Schmid-Hempel tion, where different ecological factors affect the microbiota, and the respective functions remain poorly studied (Hamdi et al. 2011). Plausible ecological factors that affect the microbiota include nutritional status of the host (Dillon et al. 2010) and the activation of the immune system upon parasitic infections (Ryu et al. 2008; Lazzaro & Rolff 2011). We tested these factors by experimentally manipulating colonies of B. terrestris (Linnaeus) either by stimulating the immune system with heat-killed bacteria or by providing additional food (sugar water) to field-placed colonies in a full-factorial design. We also looked at the following questions: 1. At the individual level: (a) Does the microbiota change when natural infections by the two common microbial parasites C. bombi Lipa & Triggiani and N. bombi Fantham & Porter are present? An infection could alter the gut microbiota, either through direct interaction or through activation of the immune system by parasites, disturbing the homoeostasis with resident microbiota (Ryu et al. 2008; Lazzaro & Rolff 2011). (b) Does host body size influence the diversity of gut microbiota (allometry)? For macroscopic parasites, host size tends to be correlated with higher parasite diversity (Poulin & Morand 2000). Hosts size can be expected to be related to gut volume (Yang & Joern 1994) and therefore to the size of the potential habitat for gut microbiota within a host. The diversity of free-living bacteria shows a species–area relationship comparable to that found in macroscopic organisms (Bell et al. 2005). However, whether such a relationship also exists for microbial communities within hosts of different sizes has so far not been examined. Bumblebees present a good model to study this question, as highly related workers within one colony can vary up to 10-fold in body mass (Couvillon et al. 2010). 2. At the colony level: (a) Do different colonies harbour distinct microbiota? The microbiota may be colony-specific as has been observed in termites (Minkley et al. 2006). Colony specificity of microbiota could arise through predominantly vertical transmission within the nest in contrast to environmental transmission in the field. If the host genotype can influence the microbiota (Ley, Peterson & Gordon 2006), microbiota should differ between colonies. (b) Does the microbiota change over the colony cycle? While the microbiota of isolated bumblebee individuals have been characterized (Koch & Schmid-Hempel 2011a; Martinson et al. 2011), there is no information on a possible change in microbiota throughout the colony cycle. We looked both at changes in overall microbial diversity and at changes in relative abundance of specific bacterial taxa.
Materials and methods DESCRIPTION OF FIELD EXPERIMENT ⁄ SAMPLING
The B. terrestris colonies used in this experiment originated from queens collected in Neunforn (Thurgau, Switzerland) near the
experimental field site in spring 2009. The queens were first taken to the laboratory, fed ad libitum with pollen and sugar water and allowed to establish colonies of c. 10 workers. Forty colonies were then taken to the field site (around Ittingen abbey, TG, Switzerland, c. 6 km from the queen collection site) and kept in nest boxes (Schwegler, Schorndorf, Germany). Ten of the 40 colonies were each assigned arbitrarily to one of the following four treatment groups:
1. Priming (Pr): Once every week, all workers from the colony were
2.
3. 4.
injected with 2 lL of a cocktail of heat-killed bacteria (Arthrobacter globiformis (Conn) and Escherichia coli (Escherich), at 0Æ5 · 108 cells mL)1 each in insect Ringer) into the haemolymph with a sterile glass capillary. Food supply (Fs): Colonies received 60 mL of 50% Apiinvert sugar water (Su¨dzucker AG Mannheim ⁄ Ochsenfurt), offered inside the colony, every week. Primed and food supply (PrFs): Both treatments (i) and (ii) combined. Naı¨ ve (N): None of the above.
Individuals in treatment groups (2) and (4) were also pricked with a sterile needle but without injecting heat-killed bacteria, to control for the effect of experimental manipulation when injecting the bacteria. If excreted, the weekly injection of 2 lL insect Ringer may lead to a small increase in the amount of liquid passing through the gut. Nevertheless, this is unlikely to affect gut physiology, as bumblebees expel an amount of liquid close to their own body weight every day (c. 140 mg, Bertsch 1984). Every week, 10% of the workers from each colony with more than five individuals were collected randomly, removed and frozen for further analysis (total n = 354 individuals). A detailed analysis of the effects of the experimental manipulation on colony development and infection with C. bombi parasites will be published in an accompanying paper (Cisarovsky, Koch & Schmid-Hempel 2012). To examine the relationship between host body size and gut size, we sampled 20 worker B. terrestris from two laboratory colonies. Workers were selected to be evenly distributed across the whole range of body sizes in these colonies. We used mass as a measure of body size and gut size. Bees were killed by freezing and weighed on a microbalance (model UMT2, Mettler-Toledo, Switzerland). Whole guts were then dissected out and weighed separately.
BACTERIAL CULTURE FOR THE IMMUNE CHALLENGE
To activate the Toll and Imd pathways – two major pathways of insect immune defence (Hoffmann 2003) – heat-killed Gram-positive A. globiformis (DSM 20124) and Gram-negative E. coli (DSM 498) bacteria were injected into the haemolymph. For this, bacteria were cultured in liquid broth (10 g bacto-tryptone, 5 g yeast extract, 10 g NaCl in 1000 mL distilled water, pH 7) at 30 C and 24 h for A. globiformis and 37 C and overnight for E. coli. One millilitre of culture was washed three times with insect Ringer (previously autoclaved and filtered though a 0Æ2-lm filter) by centrifuging 10 min at 3000 r.p.m., removing the supernatant and replacing with fresh Ringer. Bacteria were then counted in Neubauer counting chambers and concentrations adjusted with insect Ringer to 108 cells mL)1 of each bacterium. Both bacterial suspensions were then mixed in equal proportions and stored in 2 mL aliquots at )80 C. Bacteria were heat-killed by incubating thawed aliquots at 90 C for 15 min before use.
2012 The Authors. Journal of Animal Ecology 2012 British Ecological Society, Journal of Animal Ecology
Ecological effects on gut bacterial communities 3 MOLECULAR METHODS
The protocol for DNA extraction, PCR and terminal restriction fragment length polymorphism (TRFLP) analysis followed the method described in detail in Koch & Schmid-Hempel (2011a). Briefly, after weighing individual bees, the whole guts were dissected out and genomic DNA was extracted with a Qiagen DNAeasy kit (Qiagen, Hilden, Germany) following the protocol for blood and tissue samples with an additional digest with lysozyme. The bacterial 16S rRNA gene was then PCR-amplified with the nearly universal eubacterial primers 27f (AGA GTT TGA TCM TGG CTC AG, FAM labelled for TRFLP analysis) and 1492r (ACG GYT ACC TTG TTA CGA CTT) (Weisburg et al. 1991). PCR products were cleaned with Sephadex G-50 (GE Healthcare, Little Chalfont, UK), digested with HaeIII, cleaned again with Sephadex G-50 and run on a MegaBACE capillary sequencer.
DETECTING INFECTIONS WITH CRITHIDIA AND NOSEMA
Infection status (infected vs. uninfected) with the intestinal parasite C. bombi and the intracellular parasite N. bombi was detected by PCR specific for either parasite using the DNA from the extracted guts as template. For C. bombi, a part of the 18S rRNA gene was amplified using the Crithidia-specific primers CB-SSUrRNA-F2 and CB-SSUrRNA-B4 following the study of Schmid-Hempel & Tognazzo (2010). To detect infections with N. bombi, the Nosema-specific primer pair 18f (CACCAGGTTGATTCTGCC) and 1537r (TTATGATCCTGCTAATGGTTC) (Baker et al. 1995) was used. For both reactions, the presence of a product of the right size was checked on a 1Æ5% agarose gel.
DATA ANALYSIS
Analysis of the TRFLP profiles followed the procedure outlined in the study of Koch & Schmid-Hempel (2011a). In short, after raw data are processed and sized in Fragment Profiler version 1.2 (MegaBACE, GE Healthcare, Little Chalfont, UK), the TRFLPs of a sample yield a profile with peaks corresponding to the various taxonomic units. Peaks are then filtered from baseline noise and binned between samples as described in the study of Abdo et al. (2006). The individual peak area was divided by the total peak area of all peaks in a sample to standardize between samples. A dissimilarity matrix was computed from these data using the Bray–Curtis coefficient (Bray & Curtis 1957) to compare the similarity of TRFLP profiles with a twodimensional non-metric multidimensional scaling (NMDS) analysis with the PROXSCAL module in SPSS 19 (IBM). To assess the goodness-of-fit of the NMDS solution, a Shepard plot and Kruskal’s STRESS1 were examined. To test for significant differences between (i) the treatment groups, (ii) bumblebee colonies and (iii) individuals either infected or uninfected with Nosema or Crithidia, a one-way analysis of similarity (anosim) (Clarke 1993; Rees et al. 2004) was carried out in PAST (Hammer, Harper & Ryan 2001) with 10 000 permutations on the Bray–Curtis dissimilarity matrix; P-values were corrected for multiple testing by a Bonferroni’s correction. Positive R-values indicate a higher similarity between samples within one group than between groups, with values around R = 0 indicating no difference in similarity between samples within and between groups. Values of R > 0Æ75 are generally interpreted as indicating strong separation between groups, R > 0Æ5 as separation with overlap and R < 0Æ25 as barely separable (Ramette 2007). For reasons of simplicity and to exclude potential treatment effects, we restricted the between-colony comparison of the bumblebee microbiota to the
largest treatment group (Fs). To compare diversities of microbial communities, the number of peaks in individual profiles was counted as a measure of the richness of bacterial taxa in the gut of an individual bee. Additionally, the Shannon diversity index (Krebs 1989) was calculated. For this, the peak areas were first standardized by dividing the area of each peak by the total peak area of all peaks in an individual profile to account for different quantities of labelled DNA in each sample. For each sample, the standardized peak area was then treated as abundance of a bacterial taxon, while the number of peaks in total was treated as the number of bacterial taxa. The Shannon diversity index was calculated according to the formula in the study of Kuehl et al. (2005). The statistical analyses of the relationship of number of taxa and Shannon diversity with experimental treatment and infection with parasites were carried out on the residuals of the colonies, to account for the colony effect. The colony residuals for the respective variables were extracted from an anova with the colony of origin as a random effect. A histogram of the colony residuals for the number of taxa and the Shannon diversity index showed them to be approximately normally distributed. We then constructed a linear model including experimental treatment and parasite infection status with Crithidia and Nosema as factors and sampling week as a covariate. Non-significant interaction terms were removed from the model. We also looked at the correlation between the average diversity of the microbiota of individuals within one colony and the infection prevalence with either Nosema or Crithidia using a Spearman’s correlation test. For this, colonies that had died within the first 4 weeks of the experiment were excluded, as the number of sampled individuals for these colonies was very low. For the analyses not looking at the effects of the experimental treatments, we restricted the analysis to the food-supplied groups (Fs & PrFs). The colonies of the two treatment groups without food supply (N & Pr) had mostly died in the first weeks, therefore not providing useful data on the change in the microbiota through the season and inflating the sampling for the first weeks. We combined the Fs and PrFs groups for the analyses, as the immune priming treatment did not show a significant effect on composition or diversity of the microbiota (see Results). To test the effect of individual body size and colony age on the diversity of microbiota, the residuals of the colonies were correlated with individual body mass or week of collection in a Spearman’s rank correlation analysis, respectively. The identification of the bacterial taxa corresponding with the TRFLP peaks is based on the survey of bacterial communities in bumblebee guts in the study of Koch & Schmid-Hempel (2011a), which used the same protocol for the TRFLP analysis and identified the predominant taxa with 16S sequences from clone libraries. As some of the clone libraries were from bumblebee individuals collected from a site in the vicinity of the field site of this study and included B. terrestris, we assume the peak identifications of Koch & Schmid-Hempel (2011a) to be valid for this study as well. We correlated the colony residuals of the relative signal intensity of the identified peaks with the week of collection, to look for changes in relative abundance of these taxa over the season. To visualize potential changes in the contribution of these taxa to the whole community in the different treatment groups and individuals infected or uninfected with parasites, we displayed the relative signal intensities in a heat map. Only the data for the first 5 weeks are displayed for the different treatment groups, as we did not have samples for all treatment groups for the following weeks.
Results Analysing the average for all individuals from all experi mental colonies, the number of bacterial taxa and bacterial
2012 The Authors. Journal of Animal Ecology 2012 British Ecological Society, Journal of Animal Ecology
4 H. Koch, G. Cisarovsky & P. Schmid-Hempel (a)
(b)
49
16
10 28 8
6 Fs
PrFs
N
Pr
Mean no. bacterial taxa
12 43
154
287
10
TFLP peak
12 Mean no. bacterial taxa
(a)
(b)
67
200
8
6
Crith(–) Crith(+) Nos(–) Nos(+)
Treatment group (first 5 weeks)
Parasite infection status
Fig. 1. Mean number of bacterial taxa of the microbiota from individuals of different treatment groups and parasite infection statuses. (a) Treatments are: Fs: food supply; PrFs: immune priming and food supply; N: naı¨ ve (control); Pr: immune priming; data restricted to weeks 1–4. (b) Individuals infected (+) or uninfected ()) with Crithidia (Crith) and Nosema (Nos). Error bars: ±1 SE. Number of individuals in each group on top of each bar. Shannon diversity of bacterial communities showed the same patterns (not shown).
Table 1. anova (type II) of linear models for the number of bacterial taxa and Shannon diversity index of bacterial communities Number of taxa (colony residuals)
Shannon–Wiener diversity (colony residuals)
Factor
d.f.
F
P
Factor
d.f.
F
P
Week Fs Pr Crithidia Nosema
1 1 1 1 1
1Æ6 0Æ49 0Æ01 0Æ04 3Æ53
0Æ21 0Æ48 0Æ93 0Æ83 0Æ06
Week Fs Pr Crithidia Nosema
1 1 1 1 1
0Æ34 0Æ15 0Æ01). A visualization of the relative signal intensity of the main TRFLP peaks in a heatmap similarly shows no differences between the treatment groups (Fig. 2a). Hence, these ecological factors seemed not to affect the microbiota in the average individual.
38 bp
38 bp
202 bp
202 bp
224 bp
224 bp
246 bp
246 bp
257 bp
257 bp
307 bp
307 bp 321 bp
321 bp Fs
PrFs N Pr Treatment group Mean standardized peak area
Crith(–) Crith(+) Nos(–) Nos(+) Parasite infection status
0·30 0·25 0·20 0·15
0·10 0·05 0·00
Fig. 2. Heatmap with the relative contribution of the dominant bacterial taxa to the whole bacterial gut microbiota in different treatment groups (a) or infection statuses (b). (a) Treatments are: Fs: food supply; PrFs: immune priming and food supply; N: naı¨ ve (Control); Pr: immune priming; data restricted to weeks 1–4. (b) Individuals infected (+) or uninfected ()) with Crithidia (Crith) and Nosema (Nos). Putative identifications of TRFLP peaks are (with clade no. from the study of Koch & Schmid-Hempel 2011a,b: peak 38 bp: taxon Bacteroidetes (clade IV), 202 bp: ‘Candidatus Gilliamella apicola’ (I), 224 bp: ‘Candidatus Snodgrassella alvi’ (III), 246 bp: Lactobacillus sp. (VI), 257 bp: Bombiscardovia coagulans (IX), 307 bp: Fructobacillus sp. (VIII), 321 bp: Firmicutes (V). Numbers in peak labels indicate the terminal restriction fragment length of the respective peak in base pairs. The shade of the heatmap cells is the average proportion of the corresponding peak area (taxonomic group) relative to the peak area of all peaks in respective profiles. INCREASED DIVERSITY OF GUT MICROBIOTA IN PARASITE-INFECTED INDIVIDUALS
Whereas no significant effects of the treatments on the microbiota were observed, natural infections with Crithidia and Nosema acquired by the colonies in the field were linked to both a higher number of bacterial taxa in the gut and a higher Shannon diversity index of the community (no. of bacterial taxa: Fig. 1b, data for Shannon diversity show identical pattern). However, when accounting for colony identity as a potentially confounding factor, this increased diversity was only linked to Nosema infections (Table 1). While Crithidia infections were not directly linked to higher gut bacterial diversity (Table 1), when comparing all colonies the average per-capita diversity of the microbiota for members of a given colony was positively and significantly correlated with the infection prevalence with Crithidia (i.e. percentage of workers infected) of the same colony (mean number of bacterial taxa: Spearman’s rs = 0Æ571, P = 0Æ0133; mean Shannon diversity index: rs = 0Æ576, P = 0Æ0123). This correlation was not observed for the Nosema infection prevalence (mean number of bacterial taxa: Spearman’s rs = 0Æ159, P = 0Æ529; mean Shannon diversity index: rs = 0Æ161, P = 0Æ524). The community composition significantly differed in Crithidia-infected individuals as compared to uninfected ones (anosim: P < 0Æ0001). These two groups were, however,
2012 The Authors. Journal of Animal Ecology 2012 British Ecological Society, Journal of Animal Ecology
Ecological effects on gut bacterial communities 5 Table 2. R-values of analysis of similarity (anosim) comparing the similarity of the composition of the microbiota between different colonies (only Fs treatment) Colony no.
13
134
160
171
172
18
196
202
225
13 134 160 171 172 18 196 202 225 243
0Æ163* 0Æ036 0Æ207** 0Æ225 0Æ113 0Æ229
