Glyphosate perturbs the gut microbiota of honey bees

Increased mortality of honey bee colonies has been attributed to several factors but is not fully understood. The herbicide glyphosate is expected to be innocuous to animals, including bees, because it targets an enzyme only found in plants and microorganisms. However, bees rely on a specialized gut microbiota that benefits growth and provides defense against pathogens. Most bee gut bacteria contain the enzyme targeted by glyphosate, but vary in whether they possess susceptible versions and, correspondingly, in tolerance to glyphosate. Exposing bees to glyphosate alters the bee gut community and increases susceptibility to infection by opportunistic pathogens. Understanding how glyphosate impacts bee gut symbionts and bee health will help elucidate a possible role of this chemical in colony decline.

Glyphosate, the primary herbicide used globally for weed control, targets the 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS) enzyme in the shikimate pathway found in plants and some microorganisms. Thus, glyphosate may affect bacterial symbionts of animals living near agricultural sites, including pollinators such as bees. The honey bee gut microbiota is dominated by eight bacterial species that promote weight gain and reduce pathogen susceptibility. The gene encoding EPSPS is present in almost all sequenced genomes of bee gut bacteria, indicating that they are potentially susceptible to glyphosate. We demonstrated that the relative and absolute abundances of dominant gut microbiota species are decreased in bees exposed to glyphosate at concentrations documented in the environment. Glyphosate exposure of young workers increased mortality of bees subsequently exposed to the opportunistic pathogen Serratia marcescens. Members of the bee gut microbiota varied in susceptibility to glyphosate, largely corresponding to whether they possessed an EPSPS of class I (sensitive to glyphosate) or class II (insensitive to glyphosate). This basis for differences in sensitivity was confirmed using in vitro experiments in which the EPSPS gene from bee gut bacteria was cloned into Escherichia coli. All strains of the core bee gut species, Snodgrassella alvi, encode a sensitive class I EPSPS, and reduction in S. alvi levels was a consistent experimental result. However, some S. alvi strains appear to possess an alternative mechanism of glyphosate resistance. Thus, exposure of bees to glyphosate can perturb their beneficial gut microbiota, potentially affecting bee health and their effectiveness as pollinators.

The broad-spectrum herbicide glyphosate [N-(phosphonomethyl)glycine] has long been the primary weed management system, and its use is growing in connection with crops genetically engineered to be resistant to glyphosate (1, 2). Its mechanism of action, inhibition of 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS), an enzyme in the shikimate pathway, prevents the biosynthesis of aromatic amino acids and other secondary metabolites in plants and some microorganisms (3). EPSPS catalyzes the reaction between phosphoenolpyruvate (PEP) and shikimate 3-phosphate (S3P) (4), and glyphosate is a competitive inhibitor that blocks the PEP-binding site (5). EPSPS enzymes from different organisms vary in molecular weight (46–178 kDa) and sequence homology (6) and form two phylogenetic clusters that differ in tolerance to glyphosate. Class I enzymes are sensitive to glyphosate and are present in all plants and in some bacteria, such as Escherichia coli (4); class II enzymes are only found in some bacteria, such as Staphylococcus aureus, and can tolerate high concentrations of glyphosate (7, 8).

Animals lack the shikimate pathway, which is why glyphosate is considered one of the least toxic pesticides used in agriculture (9). However, some evidence suggests that glyphosate affects nontarget organisms, for example, changing the behavior of honey bees (10), reducing reproduction of soil-dwelling earthworms (11), and affecting the growth of microalgae and aquatic bacteria (12). Glyphosate is also associated with changes in plant endophytic and rhizosphere microbiomes (2) and with disturbances of gut microbiota of animals living near agricultural sites (13).

Honey bees and bumble bees are major pollinators of flowering plants, including many crops. When foraging, they can be exposed to a variety of xenobiotics, such as glyphosate. This herbicide is known to affect the growth of microorganisms (13⇓–15), and the health of bees is intrinsically related to their distinct gut microbial community (16, 17). The honey bee gut microbiota is dominated by eight bacterial species: Lactobacillus spp. Firm-4, Lactobacillus spp. Firm-5 (phylum Firmicutes), Bifidobacterium spp. (phylum Actinobacteria), Snodgrassella alvi, Gilliamella apicola, Frischella perrara, Bartonella apis, and Alpha 2.1 (phylum Proteobacteria) (18). Each of these species exhibits strain diversity corresponding to differences in metabolic capabilities and tolerances to xenobiotics (19, 20). Newly emerged workers (NEWs) are nearly free of gut bacteria and acquire their normal microbial community orally through social interactions with other workers during the first few days after emergence (21). Bees deprived of their normal microbiota show reduced weight gain and altered metabolism (22), increased pathogen susceptibility (17), and increased mortality within hives (23).

In this study, we investigated the effects of glyphosate exposure on the size and composition of the honey bee gut microbiome. We found the microbiome was affected by glyphosate exposure during and after gut colonization, and that glyphosate exposure during early gut colonization increased mortality of bees exposed to an opportunistic pathogen. Additionally, bee gut bacteria differ in glyphosate susceptibility. We explored the molecular mechanisms of this variability in glyphosate tolerance by expressing the EPSPS of bee gut symbionts in E. coli. Some bee gut bacteria tolerate glyphosate by virtue of a class II EPSPS, but a few strains with susceptible class I EPSPS depend on other, yet unknown, mechanisms for tolerance. Overall, our results show that glyphosate exposure can perturb the gut microbiota of honey bees, and that compositional shifts typically favor species tolerant to glyphosate and disfavor sensitive species.

Glyphosate Perturbs the Honey Bee Gut Bacterial Community. Hundreds of adult worker bees were collected from a single hive, treated with either 5 mg/L glyphosate (G-5), 10 mg/L glyphosate (G-10) or sterile sucrose syrup (control) for 5 d, and returned to their original hive. Bees were marked on the thorax with paint to make them distinguishable in the hive. Glyphosate concentrations were chosen to mimic environmental levels, which typically range between 1.4 and 7.6 mg/L (24), and may be encountered by bees foraging at flowering weeds. To determine the effects of glyphosate on the size and composition of the gut microbiome, 15 bees were sampled from each group before reintroduction to the hive (day 0) and postreintroduction (day 3), and relative and absolute abundances of gut bacteria were assessed using deep amplicon sequencing of the V4 region of the bacterial 16S rRNA gene and quantitative PCR (qPCR). At day 0, glyphosate exposure had little effect on the bee gut microbiome size, but the absolute and relative abundances of the core species, S. alvi, were significantly lower in the G-10 group (Fig. 1 and SI Appendix, Fig. S1). The effects of glyphosate exposure on the bee gut microbiome were more prominent at day 3, after treated bees were returned to the hive. The total number of gut bacteria decreased for both treatment groups, relative to control, but this drop was significant only for the G-5 group, which also exhibited more severe compositional shifts (Fig. 1). The absolute abundances of four dominant gut bacteria, S. alvi, Bifidobacterium, Lactobacillus Firm-4 and Firm-5 were decreased (Fig. 1), and the relative abundance of G. apicola increased in the G-5 group (SI Appendix, Fig. S1). Surprisingly, only Lactobacillus Firm-5 decreased in absolute abundance in the G-10 group (Fig. 1). This experiment was repeated using bees from a different hive and season, and similar trends were observed (SI Appendix, Fig. S2). As in the first experiment, significant reductions in abundance were observed for S. alvi in bees treated with glyphosate (SI Appendix, Fig. S2). Fig. 1. Changes in gut microbiota composition following glyphosate exposure of honey bees with established gut communities. (A) Stacked column graph showing the relative and absolute abundances of gut bacterial species in control bees and bees treated with 5 mg/L or 10 mg/L glyphosate at posttreatment days 0 and 3. Each column represents one bee. (B) Boxplots of bacterial 16S rDNA copies for control (C) and glyphosate-treated (G-5 and G-10) bees at posttreatment days 0 and 3 (n = 15 for each group and time point). Box-and-whisker plots show high, low, and median values, with lower and upper edges of each box denoting first and third quartiles, respectively. *P < 0.05 and **P < 0.01, Wilcoxon rank sum test followed by Bonferroni correction. The relative lack of effects of the G-10 treatment on the microbiota composition at day 3 posttreatment is unexplained, but may reflect other effects of glyphosate on bees. Our recapture method fails to sample bees that died or abandoned the hive. Since bees exposed to glyphosate may exhibit impaired spatial processing, compromising their return to hives (10, 24), bees in the G-10 group that consumed more glyphosate-laced sugar syrup before reintroduction to the hive may have been less likely to return to the hive after foraging. Since fewer than 20% of bees reintroduced to the hive were recovered, recovered bees may not represent the total effect of glyphosate on treatment groups.

Glyphosate Affects Early Gut Bacterial Colonization. Glyphosate arrests bacterial growth without directly killing the cells, so we hypothesized that it would have a greater effect on actively dividing cells present during early gut colonization. To test this, NEWs, which are nearly free of gut bacteria (21), were simultaneously exposed to an inoculum consisting of their normal microbial community and to glyphosate. This simultaneous exposure is relevant to field situations, since glyphosate has been detected in hives and honey samples (25, 26), indicating that honey bee foragers can transport residues of this herbicide to the colony and contaminate other bees, including NEWs, and food resources. Also, glyphosate is a stable, water-soluble chemical that can persist in the environment for long periods (10). Assessment of gut microbiomes, as described in the previous section, identified all eight core gut taxa in both control and treatment groups (Fig. 2A), showing that glyphosate does not eliminate colonization by any core member. Average total bacterial abundance was slightly lower in glyphosate-treated bees, but this was not statistically significant (Fig. 2B). S. alvi was the most strongly affected member of the gut microbiota and decreased in both absolute and relative abundance, while Lactobacillus Firm-4 increased in relative abundance (Fig. 2 C–E and SI Appendix, Fig. S3). Based on relative abundance, gut community compositions of glyphosate-treated bees differed from those of controls (principal coordinate analysis of weighted UniFrac) (27), permanova test with 9,999 permuations; P = 0.0078, pseudo-F statistic = 6.66) (Fig. 2F). Thus, glyphosate exposure during early development of the gut community can interfere with normal colonization by altering the abundance of beneficial bacterial species. Fig. 2. Changes in gut microbiota composition following glyphosate exposure of young honey bees and susceptibility to Serratia infection. (A) Stacked column graph showing the relative and absolute abundances of gut bacterial species in control and glyphosate-treated bees. Each column represents one bee. (B–E) Boxplots of total bacterial 16S rDNA copies and of absolute and relative abundances of two gut bacterial species for control (n = 14) and glyphosate-treated (n = 11) bees. **P < 0.01, and ***P < 0.001, Wilcoxon rank sum test followed by Bonferroni correction. (F) Principal coordinate analysis of gut community composition using weighted UniFrac (permanova test with 9,999 permuations; P = 0.0078, pseudo-F statistic = 6.66). (G) The percent survival of age-controlled bees after Serratia kz19 exposure, shown as a Kaplan–Meier survival curve. ***P < 0.001, coxph model implemented in the “survival” package in R. GH, gut homogenate-exposed bees; Gly, glyphosate treatment; MF, microbiota-free bees; Ser, Serratia challenge. Typically, captive honey bees do not defecate, and dead bacterial cells and the released DNA accumulate in the gut (23). Thus, we also analyzed changes in bacterial abundance after glyphosate exposure by extracting both DNA and RNA from the guts of treatment and control bees in a second colonization experiment. We included a positive control group, in which bees were exposed to tylosin, an antibiotic used in beekeeping. This antibiotic treatment was expected to perturb the microbiota, but the decrease was significant only for RNA samples (SI Appendix, Fig. S4). Glyphosate exposure resulted in nonsignificant decreases in total bacteria for both DNA and RNA assays. We also checked changes in absolute abundance for three core bacterial species, S. alvi, Lactobacillus Firm-4, and Lactobacillus Firm-5. Tylosin treatment resulted in reductions for 16S rRNA copies (SI Appendix, Fig. S4). Effects of glyphosate treatment were specific to S. alvi, which was the only assayed species showing significant reductions in absolute abundance, observed for both DNA and RNA assays (SI Appendix, Fig. S4). This experiment suggests that measures based on DNA are partly obscured by DNA from dead bacterial cells, although this effect does not entirely mask shifts in bacterial abundance.

Glyphosate Exposure Makes Young Worker Bees More Susceptible to Serratia. To determine whether glyphosate-induced perturbation of microbiota colonization affects host health, we measured the susceptibility of glyphosate-treated bees to an opportunistic bacterial pathogen. NEWs were exposed to glyphosate in the stage of acquiring their normal microbial community. After 5 d of treatment, bees were challenged with Serratia marcescens kz19, an opportunistic pathogen commonly detected at very low frequencies in the bee gut (28, 29). For bees lacking gut microbiota, Serratia challenge resulted in increased mortality relative to that observed for bees with a conventional gut microbiota, regardless of glyphosate exposure (Fig. 2G and SI Appendix, Fig. S5). For bees with a conventional gut microbiota, glyphosate treatment resulted in increased mortality after Serratia challenge. To determine whether this increased mortality was attributable to the effects of glyphosate on the gut microbiota or to direct effects of glyphosate on bees, we included control groups not challenged with Serratia. In bees exposed to glyphosate, but not challenged with Serratia, survival rates were not significantly affected by glyphosate and were much higher than in the Serratia-challenged groups (Fig. 2G), demonstrating that a direct effect of glyphosate on bees is not the basis of the high mortality of glyphosate-exposed, pathogen-challenged bees. Our results show that glyphosate reduces the protective effect of the gut microbiota against opportunistic pathogens and that S. alvi is the bacterial species most negatively affected by glyphosate exposure. By itself, S. alvi appears to give some protection, but not as much as the whole gut microbiota (SI Appendix, Fig. S6). S. alvi forms a biofim on the wall of the gut ileum (18, 21, 30), which may function as a mechanical barrier against pathogen invasion. Some S. alvi strains encode type VI secretion systems (31), which could contribute to colonization resistance through contact-dependent inhibition of Serratia. Furthermore, host expression of antimicrobial peptides is upregulated after S. alvi colonization (32), which could increase resistance to infection by pathogens. Besides a direct protective effect, S. alvi may be critical in enabling the full microbiota to assemble, thus enabling greater protection.

The Bee Gut Contains Bacterial Species with both Sensitive and Insensitive Types of EPSPS. Bacterial EPSPS exists as two main types, corresponding to two phylogenetic clusters, that differ in sensitivity to glyphosate: Class I is naturally sensitive, whereas class II is insensitive (8). To identify the EPSPS types present in the bee gut microbiota, a phylogenetic tree was constructed using the EPSPS protein of bacterial strains isolated from honey bee and bumble bee guts and of other representative organisms (Fig. 3). EPSPS sequences from S. alvi, G. apicola, F. perrara, Bifidobacterium, and Apibacter adventoris (phylum Bacteroidetes) (33) clustered with those from other organisms containing a class I EPSPS, and thus these bacteria are predicted to be sensitive to glyphosate. In contrast, sequences from B. apis and Lactobacillus Firm-4 clustered with other bacteria containing a class II EPSPS, as did Parasaccharibacter apium (Alpha 2.2), a bacterium commonly detected in honey bee larvae and hives, but rare in the guts of workers (34), and Paenibacillus larvae, the agent of American foulbrood in honey bee larvae (35); these bacteria are predicted to be unaffected by glyphosate exposure. Lactobacillus Firm-5 strains for which genomes are available lack the gene encoding EPSPS and were excluded from our analysis. Fig. 3. Maximum-likelihood phylogeny based on amino acid sequences of EPSPS (PhyML 3.1, LG model + Gamma4, 100 bootstrap replicates). Bee gut bacterial strains, other bacteria, and some plant species are represented in the phylogeny. The heatmap represents the growth of some bee bacterial strains in the presence/absence of 10 mM glyphosate at three time points (24, 36, and 48 h). Glyphosate was dissolved in the culture media (InsectaGro or MRS broth, based on bacterial preferences). A value of 1 indicates that growth is the same in the presence or absence of glyphosate.

Bee Gut Bacteria Vary in Glyphosate Sensitivity at the Species and Strain Levels. Several bee gut-associated bacterial strains isolated from honey bees and bumble bees were cultured in vitro in the presence or absence of a high dose of glyphosate. Most S. alvi and G. apicola strains tested, which contain a class I EPSPS, either do not grow or have a delay in growth when cultured in the presence of glyphosate; no such effect is observed for strains containing a class II EPSPS, Lactobacillus Firm-4 and B. apis (Fig. 3 and SI Appendix, Fig. S7). However, S. alvi strains wkB2 and wkB298, despite containing a class I EPSPS, grow as well in the presence of glyphosate as in its absence, with no initial delay in growth. We looked for potential single-site mutations in the EPSPS active site of these strains, which is known to confer tolerance to glyphosate (36), but no mutations were observed, indicating that the resistance in these S. alvi strains results from other mechanisms. A previous study of the genes required by S. alvi to live in the bee hindgut showed that the aromatic amino acid biosynthetic pathway is required for growth in this niche (30), which is consistent with low aromatic amino acid concentrations in the hindgut (37). Thus, bee gut bacterial strains having a glyphosate-susceptible EPSPS are predicted to drop in abundance following exposure, as observed for S. alvi (Fig. 1 and SI Appendix, Fig. S2) and Bifidobacterium (Fig. 1) in the hive experiments. Lactobacillus Firm-4, which encodes a class II EPSPS, and Firm-5, which does not contain the target enzyme of glyphosate, also had their abundances reduced in the hive experiment (Fig. 1), which was not expected. This may be explained by the fact that these strains lack the aromatic amino acid biosynthetic pathway (18), relying on uptake of aromatic amino acids released by other bacterial species, such as S. alvi, in the bee gut environment. The increase in G. apicola relative abundance (SI Appendix, Fig. S1) was unpredicted, but was also observed in a previous study on microbial community responses to antibiotic perturbation (23).

Glyphosate Resistance Is Independent of EPSPS Class in Some Bee Gut Strains. To understand the mechanism that prevents some bee gut bacterial strains from growing in the presence of glyphosate, we complemented E. coli ΔaroA with aroA genes cloned from bee gut bacterial strains as well as with the E. coli K12 aroA, which is known to be sensitive to glyphosate. E. coli ΔaroA cannot synthesize aromatic amino acids and does not grow in minimal media, but grows normally when transformed with an arabinose-inducible plasmid carrying the intact E. coli aroA gene (Fig. 4). Fig. 4. Growth curves of E. coli ΔaroA BW25113 expressing the aroA gene from different bee-associated bacterial strains (B. apis in red, E. coli in black, G. apicola in blue, and S. alvi in green) cultured in minimal media in the presence or absence of 10 mM glyphosate. Transformants carrying the aroA gene from S. alvi, G. apicola, and B. apis were able to grow in minimal media at a similar rate to the transformant carrying the aroA gene from E. coli (Fig. 4). The addition of 10 mM glyphosate to the media resulted in a delay in growth of ∼48–72 h for transformants carrying the aroA gene from all S. alvi and G. apicola strains tested (Fig. 4). This is expected if glyphosate binds to a susceptible EPSPS, blocking the shikimate pathway and preventing bacterial growth until the concentration of PEP or EPSPS exceeds that of glyphosate, allowing the transformants to resume growth. On the other hand, the transformant carrying the aroA gene from B. apis did not exhibit the growth delay in the presence of glyphosate (Fig. 4), as predicted since this aroA version encodes an insensitive class II EPSPS. Moreover, the addition of increased concentrations of arabinose in the media or reduction in glyphosate concentration sped up the growth of all transformant strains (SI Appendix, Fig. S8), which corroborates the reversible mechanism of EPSPS inhibition by glyphosate. Although S. alvi strains wkB2 and wkB298 were resistant to glyphosate (Fig. 3 and SI Appendix, Fig. S7), their aroA versions were sensitive to glyphosate (Fig. 4). Potentially, some bee gut microbes may have evolved alternative glyphosate resistance mechanisms due to a history of exposure, similar to the resistance observed for the antibiotic tetracycline used in beekeeping (38). Therefore, we overexpressed, in WT E. coli, certain genes encoding transporters from some bee gut bacteria, including wkB2 and wkB298 strains, that could be involved in glyphosate resistance: yhhS, which encodes a membrane transporter conferring glyphosate tolerance when overexpressed in E. coli (39), and tetC, which encodes an efflux pump that provides tetracycline resistance to S. alvi wkB2 (38). However, these transporters were not able to reverse the delay in E. coli growth caused by glyphosate (SI Appendix, Fig. S9). As the glyphosate tolerance exhibited by some S. alvi strains does not appear to be due to a resistant EPSPS or to transport by YhhS or TetC, these strains are likely to employ a novel mechanism of glyphosate resistance. Future studies might identify this mechanism and determine the evolutionary origin of resistance.