Research ArticleEVOLUTIONARY BIOLOGY

Dynamic microbiome evolution in social bees

See allHide authors and affiliations

Science Advances  29 Mar 2017:
Vol. 3, no. 3, e1600513
DOI: 10.1126/sciadv.1600513
  • Fig. 1 Distribution of the eusocial corbiculate bees and sample collection sites.

    The honey bees (Apis spp.) are largely restricted to South and East Asia, with the exception of A. mellifera and an introduced population of A. cerana in Australia. Stingless bees (pictured; H. itama) are found in tropical and subtropical regions, including Africa. Bumble bees (pictured; B. impatiens) in this study were collected from several sites in the United States. Biogeographical data are from previous studies (34, 60, 120, 121).

    Photo credits: W. K. Kwong and J. S. Ascher/www.discoverlife.org
  • Fig. 2 Gut microbial community profiles of sampled bee species.

    Bacterial taxa were categorized on the basis of the OTU97 taxonomic assignments (see data file S2). Relative abundances, as averaged over all samples for each host species, are given as percentages. Prevalence heat map indicates the proportion of samples per host species carrying a bacterial taxon in >0.5% abundance. Absolute abundances of 16S rRNA gene copies represent means of quantitative polymerase chain reaction (qPCR) measures (see data file S1 and fig. S1). n = number of individuals per bee species sampled. Bombus data are from this study and the work of Powell et al. (93). *, species identified on the basis of the closest mitochondrial DNA match but unverified by morphology.

  • Fig. 3 Gut community similarities, compared using PCoA.

    (A) All corbiculate bee samples, compared at a depth of 1900 reads per sample [see fig. S2A for plot excluding Powell et al.’s set (93)]. (B) A. cerana samples from different geographic locations, compared at a depth of 4500 reads. (C) Comparison of gut communities in sympatric bees, at a depth of 2900 (Austin, USA) or 4500 reads. See figs. S2 and S3 for additional PCoA and NMDS plots. All plots are derived from binary Sørensen-Dice dissimilarities of samples at OTU97 clustering. Analysis of similarities (ANOSIM) and permutational multivariate analysis of variance (PERMANOVA) were used for statistical testing of group similarities. The proportion of variation explained for each axis is given in parentheses. Tick marks, 0.2.

  • Fig. 4 Gut microbiome evolution.

    (A) Microbiome dissimilarity as a function of host divergence, showing that closely related bee species have more similar gut communities. (B) Phylogenetic relationships of sampled bee species, with parsimony-inferred shifts in the microbiota over corbiculate evolution. Only gains and losses of the corbiculate core and corbiculate-specific lineages are plotted. Taxa with <1% abundance and <20% prevalence are considered losses. See Materials and Methods for details on tree inference and microbiome reconstruction.

  • Fig. 5 Strain-level diversity in the bee gut microbiota is largely host-specific.

    Relative abundances of all reads per bee species (rows) corresponding to particular OTUs99.5 (columns) are shown for selected bacterial taxa. See data file S3 for OTU99.5 diversity summaries of other taxa. For Lactobacillus Firm-5, the conserved, single-copy tuf gene was cloned and sequenced from representative samples to visualize Firm-5 diversification and relationship with host taxa. Dashed lines and coloring indicate host origins. Clades containing previously cultured members are denoted by “*” [Lactobacillus apis (40)] and “†” (41, 42). Bayesian posterior probabilities and maximum likelihood bootstrap supports, in that order, are shown at select nodes as percentage values. Displayed tree is based on Bayesian inference; bar represents nucleotide substitutions per site. See figs. S7 and S8 for uncondensed trees. Tip taxa in host phylogeny are in the same order as in Fig. 4. Cophylogenetic analysis of hosts and Firm-5 is shown in fig. S10.

  • Fig. 6 Diversity of Snodgrassella alvi and its impact on host switching.

    (A) Bayesian phylogeny of S. alvi clones and cultured isolates is shown, with the strains used for colonization assays indicated by colored symbols. Bayesian posterior probabilities and maximum likelihood bootstrap supports, in that order, are shown at select nodes as percentages. Bar represents nucleotide substitutions per site. (B) Monoinoculations demonstrate that S. alvi strains from other Apis spp. are able to colonize A. mellifera [bars, mean; detection limit, 5 × 105 colony-forming units (CFUs)]. (C) Co-inoculations suggest variability in competitiveness and that A. mellifera–derived strains are not always dominant (error bars, SEM).

  • Fig. 7 Relationship between host identity, bacterial load, and microbiota diversity.

    (A) Number of OTUs at 97% clustering for each bee species. The Bombus microbiota is, on average, less diverse than that of Apis, whereas there is substantial variation in the Meliponini (nested ANOVA). (B) Gut community evenness as measured by Shannon’s E on OTUs97 for each bee species. The Bombus microbiota is more erratic than that of Apis, with more variation in evenness (Levene’s test). n.s., not significant. (C) Mean bacterial abundance versus average worker size for each bee species. Larger bees harbor more bacteria (F test, excluding noncorbiculates). Dashed lines, 95% confidence interval. (D) Gut microbial diversity, as measured by Shannon’s H on OTUs99.5, versus mean bacterial abundance for each bee species. There is a weak positive correlation between diversity and community size when each corbiculate tribe is considered separately [ordinary least squares (OLS) F test, colored lines] but not together (OLS F test, black line). Correcting for phylogeny improves fit [phylogenetic generalized least squares (PGLS)]. Host taxa in (A) and (B) are in the same order as Fig. 2 and fig. S1; boxes and whiskers represent 25th to 75th and 5th to 95th percentiles, respectively, of individual samples within each bee species; bar indicates median. See fig. S11 and S12 for additional plots.

  • Fig. 8 Large bee colony size is associated with higher microbiome diversity.

    (A) Plot of microbiome diversity (number of OTUs at OTU97 clustering) against gut community size (16S rRNA gene copies) × bee colony size. (B) Plot of microbiome diversity (Shannon’s H at OTU99.5 clustering) against gut community size (16S rRNA gene copies) × bee colony size. Large dots, species averages; small dots, individual samples. Best-fit lines for different regression analyses are shown. (C) Linear modeling of the effects of gut community size (16S rRNA gene copies), host phylogenetic affiliation (tribe), and bee colony size on microbiome diversity (Shannon’s H at OTU99.5 clustering). Models that include an interaction of gut community size and bee colony size have the greatest relative likelihood. Goodness of fit between models was compared with likelihood ratio tests. See fig. S13 for linear model details.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/3/e1600513/DC1

    fig. S1. Absolute bacterial abundance per bee gut, as measured by 16S rRNA gene copies.

    fig. S2. Gut community similarities, compared using PCoA of binary Sørensen-Dice dissimilarities at OTU97 clustering.

    fig. S3. NMDS plots of gut community Sørensen-Dice dissimilarities at OTU97 clustering.

    fig. S4. UPGMA clustering of communities based on Sørensen-Dice and unweighted UniFrac distances.

    fig. S5. Linear mixed models testing of the effects of sampling location and host identity on microbiome composition.

    fig. S6. Ancestral microbiome reconstructions using parsimony for the major corbiculate bacterial phylotypes.

    fig. S7. Bayesian phylogeny of Lactobacillus Firm-5 based on the tuf gene.

    fig. S8. Maximum likelihood phylogeny of Lactobacillus Firm-5 based on the tuf gene.

    fig. S9. Phylogeny of Lactobacillus Firm-5 based on rpoA, and geographic associations of strains.

    fig. S10. Tanglegram of host bee and Lactobacillus Firm-5 phylogenetic topologies.

    fig. S11. Gut microbial diversity, as measured by Shannon’s H on OTUs clustered at 99.5% similarity.

    fig. S12. Relationships between bee body size, gut community size (16S copies), and microbiome diversity.

    fig. S13. Linear models testing of the effects of bee body size, gut community size (16S copies), host phylogenetic affiliation (tribe), and bee colony size on microbiome diversity.

    table S1. Bee size and colony size estimation and literature sources.

    table S2. Snodgrassella co-inoculation trials count data.

    data file S1. List of samples and associated metadata used in this study.

    data file S2. OTU97 tables and curated taxonomic assignments.

    data file S3. OTU99.5 tables summarized by bacterial taxa, from samples rarefied to 5000 reads.

    data file S4. List of sequences deposited in GenBank.

    data file S5. Representative sequences from OTU97 and OTU99.5 clustering.

    References (122148)

  • Supplementary Materials

    This PDF file includes:

    • fig. S1. Absolute bacterial abundance per bee gut, as measured by 16S rRNA gene copies.
    • fig. S2. Gut community similarities, compared using PCoA of binary Sørensen-Dice dissimilarities at OTU97 clustering.
    • fig. S3. NMDS plots of gut community Sørensen-Dice dissimilarities at OTU97 clustering.
    • fig. S4. UPGMA clustering of communities based on Sørensen-Dice and unweighted UniFrac distances.
    • fig. S5. Linear mixed models testing of the effects of sampling location and host identity on microbiome composition.
    • fig. S6. Ancestral microbiome reconstructions using parsimony for the major corbiculate bacterial phylotypes.
    • fig. S7. Bayesian phylogeny of Lactobacillus Firm-5 based on the tuf gene.
    • fig. S8. Maximum likelihood phylogeny of Lactobacillus Firm-5 based on the tuf gene.
    • fig. S9. Phylogeny of Lactobacillus Firm-5 based on rpoA, and geographic associations of strains.
    • fig. S10. Tanglegram of host bee and Lactobacillus Firm-5 phylogenetic topologies.
    • fig. S11. Gut microbial diversity, as measured by Shannon's H on OTUs clustered at 99.5% similarity.
    • fig. S12. Relationships between bee body size, gut community size (16S copies), and microbiome diversity.
    • fig. S13. Linear models testing of the effects of bee body size, gut community size (16S copies), host phylogenetic affiliation (tribe), and bee colony size on microbiome diversity.
    • table S1. Bee size and colony size estimation and literature sources.
    • table S2. Snodgrassella co-inoculation trials count data.
    • Legends for data files S1 to S5
    • References (122–148)

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • data file S1. List of samples and associated metadata used in this study.
    • data file S2. OTU97 tables and curated taxonomic assignments.
    • data file S3. OTU99.5 tables summarized by bacterial taxa, from samples rarefied to 5000 reads.
    • data file S4. List of sequences deposited in GenBank.
    • data file S5. Representative sequences from OTU97 and OTU99.5 clustering.

    Files in this Data Supplement:

Navigate This Article