Host specificity of gut microbiota associated with social bees: patterns and processes

SUMMARY

Gut microbes provide benefits to some animals, but their distribution and effects across diverse hosts are still poorly described. There is accumulating evidence for host specificity (i.e., a pattern where different microbes tend to associate with distinct host lineages), but the causes and consequences of this pattern are unclear. Combining experimental tests in the laboratory with broad surveys in the wild is a promising approach to gaining a comprehensive and mechanistic understanding of host specificity prevalence, origin, and importance. Social bees represent an ideal testbed for this endeavor because they are phylogenetically and functionally diverse, with host-specific, stable, and tractable gut microbiota. Furthermore, the western honeybee (Apis mellifera) is an emerging experimental model system for studying microbiota–host interactions. In this review, we summarize data on the prevalence and strength of host specificity of the social bee gut microbiota (bumblebees, stingless bees, and honeybees), as well as the potential and proven ecological and molecular mechanisms that maintain host specificity. Overall, we found that host specificity in bees is relatively strong and likely results from several processes, including host filtering mediated by the immune system and priority effects. However, more research is needed across multiple social bee species to confirm these findings. To help future research, we summarize emerging hypotheses in the field and propose several experimental and comparative tests. Finally, we conclude this review by highlighting the need to understand how host specificity can influence host health.

INTRODUCTION

The gut microbiota has been shown to influence development, health, and fitness in some animals (1, 2). For example, gut microbes can increase the host’s digestive abilities (3, 4), improve immune development (5), protect against pathogens (6), and may improve tolerance, acclimation, and adaptation to environmental and lifestyle changes (7).

While molecular, cellular, and physiological processes underpinning animal–microbe symbioses in the gut are starting to be better understood in humans and animal models like the laboratory mouse, we know surprisingly little about the ecological and evolutionary dynamics shaping gut ecosystems in animals more broadly (2, 8–10). Previous work has shown that the gut microbiota varies tremendously in structure and function across the animal tree of life and is composed of particular microbes that are often only found in animal guts but are also sometimes free-living (2, 11, 12). However, it is still unclear to what extent gut microbes show specificity to their host (i.e., microbiota structure and function are correlated to host identity and/or phylogeny, as defined in box #1 and illustrated in Fig. 1A), as often observed in mammals (13–15) but not necessarily in all animals (16, 17). Furthermore, it is currently unknown how ancient this association is (i.e., whether host and microbes share a common, faithful evolutionary history as represented by co-phylogenies or whether symbionts often switch between host lineages) (18) (Fig. 1C). While these questions have recently gained a lot of attention in humans (19) and other animals (16, 20, 21), there is no consensus on the prevalence and mechanisms of host specificity of gut microbes, and so three main questions emerge:

Fig 1.

Scales and patterns of host specificity. (A) Panel depicts two host species associated with distinct (host-specific) microbial communities (i.e., a set of microbial species) mostly composed of distinct (host-specific) individual symbionts (green or yellow symbionts), as well as one generalist member (blue member). (B, C) Patterns of host specificity encompass both community- (B) and individual- (C) level scales. Community-level patterns can be represented with multidimensional ordination (e.g., principal coordinate analysis or non-metric multidimensional scaling) and beta-diversity comparisons (panel B). Individual-symbiont patterns include host ranges (e.g., number of colonized hosts) and co-phylogenetic measures (panel C). (D) Theoretical relationships between single-nucleotide polymorphism (SNP) counts and average nucleotide identity (ANI) score and relationship between ANI and taxonomic levels (strain, species, genus). Suggested ANI cut-off between species lies around 95%, but this depends on species (22). The link between ANI and strain definition is less clear (23). (E) Tentative (hypothetical) link between bacterial genomic scale and host relatedness by illustrating the taxonomic relationship expected of a pair of bacterial cells from the gut across and within hosts. Correspondences between scales and patterns are depicted by horizontal grey arrows.

1. To what extent do animals host a specific gut microbiota, and how does specificity vary between gut symbionts?

2. What are the processes underlying this specificity or lack thereof?

3. Are there any beneficial consequences of this specificity for the host?

Answering these questions requires a multidisciplinary research program that combines three complementary approaches:

1. Generate expectations using theories inspired by the fields of ecology and evolution of symbiosis, parasitism, and macroorganism (e.g., plant) communities.

2. Establish the prevalence of specificity in nature across diverse hosts using observational and correlative approaches.

3. Probe the ecological, cellular, and molecular mechanisms behind specificity using both comparative methods and experiments.

This research program requires the careful choice of a model system with two main constraints. The model system should allow broad ecological surveys in natural settings to establish the prevalence and strength of specificity and be experimentally tractable (i.e., include one or more host species that can be maintained in the lab and host a stable and tractable gut microbiota amenable to experimental manipulation).

Multiple model systems have been proposed to study host specificity in the last decade, including rodents (24), squids (25), stinkbugs (26), termites (27), wasps (28), marine worms (29), and Drosophila (30). In parallel, a less-classical animal model system has emerged: eusocial corbiculate bees and their gut microbiota (Box 2). Eusociality, defined by “an overlap of generations, cooperative brood care and reproductive division of labour” (31) (see also Box 1) is an important innovation of bees that seems to have evolved from solitary ancestors around 50–80 mya (32) (see Fig. 2A and B for the position of eusocial corbiculate bees among other bee families). This relatively new system represents a promising model system to study animal gut symbioses in general (33) and host specificity in particular for the following reasons (34) (Box 2):

Box 1. Definitions and history.

• Host specificity (s. stricto): a term originating from the field of parasitology (37) that was initially defined as the degree to which a symbiont colonizes a restricted set of hosts. It was traditionally defined as the inverse of the host “range” (=number of colonized host species). This definition was extended to encompass various other components, such as geography and host phylogeny (38).

• Host specificity (s. lato): The definition of host specificity has been recently extended to also include the specificity of sets of microbes (=microbiota or microbial “communities”) as the degree to which microbiota composition correlates to host identity and host phylogeny. This broad definition of specificity thus requires the choice of (i) a biological entity to consider (e.g., either the entire microbiota or individual symbionts, Fig. 1A) and (ii) a point of comparison (e.g., microbiota or individual symbionts from the non-host environment, distantly related species, or closely related subspecies).

• Bees (Anthophila): hymenopteran insects closely related to wasps and ants best known as pollinators. In contrast to their closest wasp relatives, bees usually use pollen as the protein source of their larvae.

• Sociality: the tendency of individuals to form groups. There is a varying degree of sociality observed in animals, particularly in bees.

• Corbiculate bees: bees harboring a pollen basket or corbicula, a structure of the tibia used to harvest and transport pollen. Corbiculate bees are composed of four tribes: Euglossini (orchid bees), Bombini (bumblebees), Apini (honeybees), and Meliponini (stingless bees).

• Eusocial corbiculate bees: corbiculate bees, excluding Euglossini tribe (orchid bees)

• Microbiota-deprived (MD) bees: “newly emerged bees that have been extracted from brood frames at the pupal stage and raised under aseptic conditions in the lab” (39)

• Co-diversification: “the synchronized bifurcation of two or more lineages of organisms” (40). Co-diversification can result from repeated co-speciation events.

• Co-speciation: the concomitant speciation of host and symbiont.

• Co-phylogeny: a pattern of phylogenetic congruency between host and individual symbiont phylogenies.

• Co-evolution (s. stricto): “an evolutionary change in a trait of the individuals in one population in response to a trait of the individuals of the second population, followed by an evolutionary response by the second population to the change in the first” (41)

• Phylosymbiosis: a measure of phylogenetic host specificity at the microbiome (community) level.

• Transient gut microbe: a species that does not persist in the gut without the current re-supply of cells from outside of the gut (e.g., on food).

Fig 2.

Bee diversity and ecology. (A) Phylogenetic relationships of the seven bee families and position of “eusocial” corbiculate bees in Apidae. (B) Phylogenetic relationships of the three tribes of eusocial corbiculate bees (Melopinini or stingless bees, Bombini or bumblebees, and Apini or honeybees). (C) Honeybees (genus Apis, tribe Apini) are represented by 10 species and include the domestic Western (European) honeybee Apis mellifera for which some features are presented. Phylogenetic branch lengths are not to scale, but the triangular area is proportional to species richness of the family (panel A) and tribe (panel B). Bee richness estimations are derived from reference 36. Phylogenetic relationships are from reference 35.

Box 2. Pros and cons of corbiculate eusocial bees as a model system to study host specificity.

+Pros:

    Simple and abundant bacterial community

    Microbiota has proven beneficial effects on the host.

    Microbes are not transient.

    Microbiota-deprived bees are easily produced.

    Most gut microbes are culturable.

−Cons:

    Simple bacterial community that may not be representative of other hosts (e.g., mammals).

    Only microbiota-“deprived” bees can be produced but not entirely germ-free individuals.

    Limited genetic tools are available for the host (42, 43) and are only emerging for the gut bacteria (44–46)

    The western honeybee (Apis mellifera) has been domesticated over centuries, with a high mixture of genetic background and complex dispersal history and maintained in conditions very different than their natural ecological conditions (e.g., artificially large colony sizes, supplemented food during low seasons, higher susceptibility to pathogens, etc.).

1. Eusocial corbiculate bees are highly diverse phylogenetically and in their ecology (diet, sociality, size, etc.) (Fig. 2) (35).

2. Eusocial corbiculate bees host a relatively simple and tractable gut microbial community (Fig. 3). The community harbors a small and consistent number of species. Most of the microbes can be isolated and maintained in laboratory conditions (33).

3. The western honeybee (A. mellifera) and its gut microbiota comprises an experimentally tractable model system (Fig. 3A) (i.e., one can generate microbiota-deprived bees under laboratory conditions and colonize them with bacterial strains isolated from diverse social bees). It is also widely distributed globally as a domesticated species and in the wild across different parts of Africa (36).

Fig 3.

Basics of the eusocial bee microbiota. (A) General structure of the western honeybee as a reference (most prevalent genera and their rough estimated number of species) and their primary localization along the digestive tract (colors of the compartments correspond to color bar alongside genus names). (B) Prevalence (% of samples where the bacteria are present) of symbionts (genera) of A. mellifera in other eusocial corbiculate bees. (C) Phylogenetic positioning of Lactobacillus isolates derived from eusocial bees. Compartmentalization (panel A) is derived from reference 47, and prevalence data (panel B) are derived from reference 48. Phylogeny (panel C) is adapted from reference 49.

PATTERNS: ARE BEE MICROBIOTA HOST SPECIFIC?

Eusocial corbiculate bees harbor a unique microbiota not found in other environments

Distinct from the non-host environment

Honeybee gut bacteria from the approximately five to seven most prevalent genera (~10–20 species, Fig. 3A) seem to exclusively grow in the honeybee gut and not outside of the host, such as on the hive frames, in the diet or on flowers (33). For example, Lactobacillus and Bifidobacterium species found in the honeybee gut form clades that are well differentiated from bacterial species of the same genera living in other environments. In contrast, a number of other accessory symbionts, such as Bombella (Acetobacteraceae, also known as Alpha 2.2 or Parasaccharibacter) (50) or Commensalibacter species (Acetobacteraceae, also known as Alpha 2.1) (51), can grow outside of the gut, for example in bee bread, royal jelly, or the hive wall. Apart from these few exceptions, the existence of a largely “endemic” gut microbiota in bees stands in contrast with microbiota from other insects, such as the fruit flies Drosophila melanogaster (52), or Lepidoptera caterpillars, such as Manduca sexta (53), where gut microbes can also grow on the host food outside of the digestive system. In bees, more research is needed to confirm the extent to which bacteria associated with other eusocial bees, especially stingless and bumblebees, can survive and grow outside of the bee gut environment. Importantly, while the most prevalent bee gut symbiont populations are probably not able to grow outside the gut, they are likely able to persist for some time in non-gut environments, for example, allowing inter-host transmission.

Somewhat distinct from “solitary” bees

“Solitary” bees represent a highly diverse paraphyletic group of bees that encompass most described bee species (Fig. 2), with a minority of species showing some degree of sociality. Solitary bees offer a useful point of comparison for the microbiota of the corbiculate eusocial bees (54–56) (see Fig. 2; Box 1). The gut microbiota of corbiculate eusocial bees, while clearly not homogenous, is relatively distinct in structure and composition from the solitary bee microbiota. For example, most of the few solitary bees surveyed rarely harbor the most prevalent symbionts found in eusocial bees, such as Lactobacillus species related to the ones found in social bees (e.g., Lactobacillus melliventris) or Snodgrassella and Gilliamella species. However, some recent work has revealed that the situation may not be as clear cut for two reasons. First, partial 16S rRNA gene sequences from some typical eusocial corbiculate bee gut symbionts, such as Snodgrassella and Gilliamella species, were detected in some solitary bee species (notably orchid bees or Euglossini), although it is unclear yet if these sequences are derived from bacteria actively replicating in solitary bee guts (54). Second, some eusocial bee species seem to lack them, in particular stingless bees (48, 54, 57) (Fig. 3B), suggesting that the gut microbiota of corbiculate eusocial bees is not homogeneous. In any case, in contrast to eusocial bees, solitary bees host diverse and more variable gut microbiota, for example, composed of Acetobacteraceae, Lactobacillaceae, or Bacillaceae (reviewed in [58]). In addition, 16S rRNA gene profiling of the gut tissue sometimes mostly recovers endosymbionts like Wolbachia sp., suggesting that resident gut microbes are in low abundance in some solitary bee species (59). One major feature of the solitary bee gut microbiota seems to be its relatively variable composition across individuals of the same species in contrast to corbiculate bees that host relatively more stable microbiota (54). In addition, the microbiota of corbiculate eusocial bees is distinct from close relatives in the tribe Euglossini (neotropical orchid bees that show variable but generally weakly social species forming the sister lineage to the three other corbiculate tribes), where the most prevalent symbionts of corbiculate eusocial bees are at a very low abundance and taxa of Caulobacteraceae, Enterobacteriaceae, and Lactobacillaceae are found instead (59). Surprisingly, some symbionts of a more distantly related family of Apidae, the Xylocopinae (carpenter bees, genus Xylocopa), bear unexpected similarities with the eusocial corbiculate gut microbiota that remain to be elucidated (60–62). Finally, the gut microbiota of corbiculate eusocial bees seems to be, on average, much larger (higher population size or “load”) than solitary bee microbiota, but this information is often lacking for solitary bees. For example, one study identified low and extremely variable abundance of Lactobacilli in megachilid bees: from 10³ to 10⁶ cells compared to 10⁸–10⁹ in honeybees (63). In summary, eusocial corbiculate bees host a unique gut microbiota distinct from distantly related solitary bee species. More generally, symbionts of the social bee microbiota usually form monophyletic clades, with environmental or solitary bee bacteria as sister groups (Fig. 3C for Lactobacillus spp.).

Host specificity of the eusocial corbiculate bee microbiota

Definitions

Host specificity refers to the degree to which gut microbes’ distribution correlates to host identity and can be measured at the level of the community (Fig. 1A and B) or the individual symbionts (Fig. 1A and C), with or without consideration of the host and microbial phylogenies. In both cases, host specificity measurement requires the careful choices of the biological unit of analysis, which may be informed by both taxonomic and functional data. The most common taxonomic units used so far to quantify host specificity rely on operational definitions based on marker genes (usually partial or full 16S rRNA gene) like amplicon sequence variants or operational taxonomic units (64). These units (and the associated amplicon data) have the advantage of being relatively easy and affordable to produce, but come at the expense of taxonomic and/or functional resolution. More recently, delineation of units based on more than one gene has also been proposed. These include clustering approaches relying on average nucleotide identity (ANI, Fig. 1D, especially useful for metagenomic data sets) (65) or patterns of gene flow (especially useful for isolate genome data sets) (66, 67). These approaches provide high-resolution data and intend to delineate true species boundaries (22, 66) but come at the expense of computational tractability (e.g., bioinformatic approaches, such as metagenome-assembled genome reconstruction) and at higher cost. A more recent approach proposes to ignore taxonomic units (e.g., species) and to use genes (or orthologous gene families, such as Kyoto Encyclopedia of Genes and Genomes orthologs) as the unit of analysis, independent of the genome they belong to (65, 68). This approach has the advantage of focusing on functional potential at the expense of complex bioinformatic analysis and higher cost.

Prevalence and strength

Host species specificity and phylosymbiosis

On average, in corbiculate eusocial bees, the gut microbiota composition clusters by host species [i.e., similarity in microbiota composition is higher between conspecifics (individuals from the same host species) than between heterospecifics (individuals from different host species)]. This is, for example, the case between the Apis (48, 69, 70), bumblebee (71, 72), and stingless bee species (48, 49). Interestingly, differences in microbiota composition between closely related species like Apis spp. are often only apparent at finer microbial genetic scales (see Fig. 1D and E). For example, the most prevalent bacterial genera (and even sometimes species) are often shared among closely related host species, and only different species of these genera (or different strain clusters within species) are found in different bee species (65, 69, 70). In some studies, closely related bee host species also share more similar microbiota than distantly related species, a pattern called “phylosymbiosis” (17, 48, 73). However, this is not always the case (55, 65), and further work is needed to elucidate the prevalence and strength of phylosymbiosis in eusocial corbiculate bees.

Lineage specificity

Community-wide host specificity emerges at the scale of the individual gut microbiota member: a symbiont is said to be host-specific (or “specialist” in the ecological vocabulary, in contrast to “generalist”) when the number of hosts colonized by a given symbiont is low (Fig. 1A and D). This pattern is not uncommon in bees when individual symbionts are defined at the species or strain cluster level (Fig. 1E) (48, 69). This is, for example, the case for Lactobacillus (74), Snodgrassella (75–77), Gilliamella (67, 76, 78), Bifidobacterium (79), and Frischella (80) species. This is also exemplified by the impressive pace at which new bacterial species are being discovered in understudied bees (49) (e.g., new Gilliamella [81] or Bifidobacterium [82] species recently described).

Host specificity can have deep evolutionary roots: if the host and symbiont intimately evolved together, we expect them to have co-speciated and co-diversified (18, 37). One hallmark of this joint evolutionary history in present-day species genomes is the congruence of their phylogenies (40, 83) (Fig. 1C; Box 1). In contrast, incongruency of symbiont and host phylogenies suggests repeated shuffling of partners during evolution or symbionts switching host lineage (40, 83). While quantitative tests of co-phylogeny are starting to emerge in other gut microbiota systems, such as in hominids (19, 21, 68), in bees, only very few quantitative tests have been proposed so far. One seminal study (76) coupled an extensive sampling of honey and bumblebee species (n = 36–41 species) with nearly full-length 16S rRNA gene-based phylogenies for two gut microbes (Snodgrassella and Gilliamella) and formally tested phylogenetic congruence using tree reconciliation methods and mantel tests. While statistically significant congruences were recovered between the host and both symbionts’ phylogenies, the rate of host switches was usually nearly twice that of co-speciation, suggesting that switching is more common than co-speciation. Other work using either whole-genome (49) or marker gene (48) phylogenies also found (qualitatively) mixed support for phylogenetic congruency, but a new recent metagenomic study in honeybees (genus Apis) did not detect co-phylogenetic signal (65). In summary, additional formal tests of phylogenetic congruency across a wide range of hosts and symbionts are needed to establish the prevalence and strength of this pattern in bees. While marker gene-based studies allow broad sampling of many species and genera, they are limited in resolution. Whole genome-based analysis provides greater resolution but is limited by throughput, as it requires isolation and sequencing. Metagenomic approaches can combine the advantages of both approaches, especially when combined with long-read sequencing. However, an important limitation of these topology-only based tests is they cannot distinguish between co-speciation and host switches that occur preferentially between related host species (83). Discriminating between these two scenarios requires measuring the congruence in time (i.e., the timing of clade divergence, for example, in millions of years) matches between the two partners (83). This is usually not possible in symbiosis involving microbes because microbial phylogenies are extremely difficult to calibrate given the lack of host-associated microbial fossils. However, recent methodological developments using the relative timing of divergence across microbial clades show great promise to test whether host and microbes co-diversify or not (40).

Comparison with solitary bees

Comparisons of host specificity in eusocial corbiculate bees with related clades (e.g., solitary bees) are key to a better understand the evolution of host specificity in bees. Overall, solitary bees seem to lack strong patterns of host specificity, at least in comparison with eusocial corbiculate bees. At the level of individual symbionts, for example, an important study (74) established that the specificity of Lactobacilli from (solitary) halictid species (Fig. 2) was weak compared to Lactobacillus species in the western honeybee gut. At the scale of the whole microbiota, only a few studies simultaneously measured host specificity in eusocial corbiculate and solitary bees. In one study focusing on bees from Central American tropical forests, the effect of host species on microbiota composition was twice as large in stingless bees than solitary bees (54). In summary, while these studies suggest that (eu)sociality is linked to higher host specificity, more comparative studies across solitary and social bees are needed.

Comparison with other animals

Social bees seem to harbor relatively strong specificity of their gut microbiota in comparison to other animals (17). In animals, host specificity is far from universal, and its strength varies greatly across hosts and tissues (84, 85). For example, at the scale of the whole microbiota, phylosymbiosis seems relatively strong in non-volant mammals, sponges, corals, and insects but weaker in amphibians, marine invertebrates, or Actinopterygians (ray-finned fishes) (17). However, most surveyed host taxa are only represented by a handful of studies, and further work is clearly needed to establish trends across the animal tree of life. Open data and high reproducibility will be key in enabling quantitative comparison of host specificity strength across systems.

General summary

Corbiculate eusocial bees host a relatively highly host-specific gut microbiota, at least in comparison to solitary species. However, more in situ (observational) studies are warranted to establish solid estimates of prevalence, variability, and strength of the pattern across bees and test whether the present-day host specificity has deep evolutionary roots (co-phylogeny).

PROCESSES: WHY ARE BEE MICROBIOTA HOST SPECIFIC?

Proximal ecological mechanisms

Host specificity is conceptually related to the notions of ecological specialization and ecological niche breadth, which have benefited from intense theoretical developments and synthesis in ecology and evolution (37, 86–88). One of the insights from these developments, the “environmental filtering metaphor,” (89) can be applied to host specificity (17, 37) to derive useful predictions and hypotheses. In this framework, host specificity can emerge from three non-mutually exclusive ecological scenarios: symbionts could be found restricted to a given host species and are not found in non-native host species (Fig. 4A) because of three potential (non-mutually exclusive) barriers or “filters” (Fig. 4B):

Fig 4.

Ecological processes underpinning host specificity. A conceptual framework inspired by the filtering metaphor (37, 89) is presented and illustrated with one native western honeybee gut bacteria (in green), as well as three non-native (i.e., not found in western honeybees) bacteria, from three different bee species. We assume bacteria are specific to their host (panel A, left side), which raises the question as to why the non-native bacteria are not found in the western honeybee in the field (panel A, right side). A conceptual framework is presented with three successive filters (panel B), each one preventing the colonization of one non-native bacterium in the western honeybee gut as illustrated by the stopping arrow. Two experimental tests of the theories (mono or co-colonization of non-native bacteria with native ones in microbiota-deprived western honeybees) are presented in panel C and applied to the native and non-native gut bacteria.

• –the symbiont did not have the opportunity to colonize a different host species due to dispersal limitation, even if it is physiologically able to colonize this new host (scenario 1, “dispersal filter,” row #2 in Fig. 4);
• –the symbiont is physiologically unable to colonize the host (scenario 2, “host filter,” row #3 in Fig. 4); or
• –the symbiont is outcompeted by the native microbiota (scenario 3, competitive or “biotic filter,” row #4 in Fig. 4).

Experimental work is key to testing and differentiating between these ecological scenarios. Results of mono-inoculation of non-native symbionts in a focal microbiota-deprived (see Box 1 for definition) or germ-free host species (Fig. 4C) can differentiate between scenarios 1 and 2/3, while those of co-inoculation of non-native and native can differentiate between scenarios 2 and 3 (Fig. 4C).

Such in vivo experiments have been carried out in bees, mostly using the western honeybee as the focal host (with sometimes bumblebees as a focal host) and a few non-native symbionts from bumblebees and various other honeybees, but not stingless bees yet. For example, reference 90 found that microbiota-deprived western honeybees were successfully colonized by Lactobacillus spp. strains isolated from Bombus hosts, albeit to a much lower load than native strains (about 10–100 times on average less abundant, 10^7–8 vs 10⁹ CFU/gut). This result partially rejects scenario 2 (strict host filter) and suggests that the barriers to cross-host species colonization are only partially physiological. However, these non-native strains were outcompeted and lost when co-inoculated with the native strains, suggesting that native strains are better adapted to the focal native host than non-native strains and validating scenario 3 (i.e. competitive filter, Fig. 4B). The same scenario has also been suggested for Gilliamella spp. (91) as well as Snodgrassella spp., when competing in the western honey bee a strain isolated from honey bees vs. a strain isolated from bumble bees (48, 92). In contrast, Snodgrassella strains isolated from closely related hosts (i.e., different honeybee species) could colonize the western honeybee, as well as native strains, suggesting that they can pass the host filter and are adapted to live in non-native closely related hosts, rejecting scenario 2. Intriguingly, some of these non-native strains could co-exist and even outcompete the native strain, rejecting scenario 3. These findings are, therefore, compatible with a model where dispersal limitation (scenario 1) between different honeybee species maintains host specificity of some gut symbionts at least between closely related hosts. More work is needed within species (genotype “matching” [93]) and between related species to confirm these findings. However, they are in line with two important in situ observations: (i) preliminary comparisons of bee and gut microbiota phylogenies show that host switching does occur over long evolutionary time scales (see Section 1), and (ii) the fact that some bee gut symbionts are not always restricted to only one host species but can be generalist (48, 65). In addition, the previous results (48) suggest that the success of colonization of non-native gut symbionts is at least partially coupled with the phylogenetic distance between the native and non-native hosts. This might indicate that host phylogenetic distances represent a proxy of environmental dissimilarities experienced by the gut microbiota, and that as bacteria specialize in each host, they progressively lose their ability to thrive in increasingly dissimilar environments maybe due to the existence of a trade-off between specialization and fitness.

In summary, cross-inoculation experiments suggest that both host filtering and competition with native microbes are important processes that maintain host specificity in bee gut microbiota, but it opens the question of which host and microbial traits and molecules mediate this filtering. In addition, dispersal limitation seems to play a role in host specificity, at least for some symbionts and between closely related host species. However, more formal tests are required to test the dispersal filter, and more symbiont strains and lineages need to be tested to get a better understanding of which processes are most important.

Traits and molecules mediating host filtering

Several host traits can maintain host specificity via filtering: either in an active, controlled, and adaptive way encoded in the host genome or in an indirect, passive way as a by-product of another host function that did not evolve to filter or control the microbiome (84). For example, host immunity represents a promising candidate that can actively filter non-native microbes either mediated by negative selection (i.e., the immune system eliminating some microbes [94]) and/or positive selection (i.e., the immune system facilitating the development of some microbes [95]). In contrast, host diet, gut anatomy, or physiology could indirectly filter a subset of specific gut microbes, as these traits can differ between host species (36). In other animal–extracellular bacteria symbiosis, such as between the squid Euprymna scolopes and Vibrio fisheri (96), between the mouse and the gut bacteria Lactobacillus reuteri (97), or between cnidarian Hydra and their bacterial associates (98), the molecular mechanism behind host filtering is starting to be unraveled (reviewed in reference 34). We review below the current knowledge of the bee system.

Host diet and secreted carbohydrates

As the primary nutrient source for gut bacteria, the host diet is known to heavily influence microbiota composition (99, 100) and host specificity, especially in mammals (101). Bees mostly feed on plant pollen and nectar collected by foragers (and their transformed version, honey, beebread, and royal jelly), and different bee species can feed on different resources (e.g., different plant species), although eusocial bees tend to have a relatively generalist diet. Diet is a promising candidate for host specificity in bees because bee gut bacteria are known to harbor a rich metabolic machinery to degrade and utilize nutrients and complex carbohydrates from the bee diet (78, 102, 103). In addition, a pollen diet is known to promote co-existence of closely related Lactobacillus species within a single host individual via resource partitioning of complex carbohydrates (104).

Within a host species, experimental manipulation of the bee diet and characterization of corresponding microbiota changes have provided valuable insights. For example, in western honeybees, the manipulation of the bee diet during winter (eight treatments corresponding to eight complex commercial diets, some composed of pollen but not all) showed a strong effect on member relative abundances, notably Bifidobacterium and Gilliamella (105). Another study compared a simple, “poor” pollen diet (one plant species) to a diverse polyfloral pollen diet and found significant changes in Lactobacillus and Bifidobacterium species (106). Comparison of the microbiota of Apis cerana fed sugar water with that fed beebread revealed a significant change in microbiota composition (106). Similar experiments have been conducted in bumblebees but with conflicting results across studies. For example, reference 107 found that different complex pollen cocktails complemented with different sugar syrups impacted gut microbiota composition, in particular Bifidobacterium spp. and Lactobacillaceae species. In contrast, reference 108 found that feeding bumblebees with mixes of pollen from three different plant species did not alter microbiota composition. The relative simplicity and lack of variation between pollen mixes in the latter study (108) compared to the former (107) might explain why the diet effect is not apparent in the former. In summary, experimental manipulation of bee diet within a host species seems to impact gut microbiota composition, as observed (to a greater extent) in vertebrates. However, it remains unclear to what extent diet can impact microbiota composition and foster host specificity because studies of the effect of diet changes on microbiota across bee species are lacking.

Multiple research avenues exist to explore the effect of diet on host specificity. First, the rearing of different wild species in the same controlled environment with identical diet, for example, bee bread isolated from different host species, can be used to measure the decrease in host specificity when the diet is kept identical across host species. Second, in situ functional characterization of the natural diet across species using plant metabarcoding of gut content or bee bread (109) combined with prediction of microbiota metabolic potential (e.g., using metagenomics or whole genome sequencing of isolated strains) presents a valuable approach to explore the relationship between diet and microbiota composition. For example, reference 49 sequenced 46 new deeply divergent strains isolated from stingless bees and conducted a comparative genomic analysis of strains found across honeybees, bumblebees, and stingless bees, particularly focusing on genes responsible for carbohydrate degradation and uptake. They found large variability in the carbohydrate–active enzyme (CAZyme) coding gene repertoire across strains from the same or different bee tribes, such as glycoside hydrolases involved in the degradation of pollen-derived carbohydrates. To what extent this variability is linked to the specific diet of the native host and whether this (partially) drives the observed host specificity is currently unknown, and further studies comparing the CAZyme repertoire found across bacteria found in diverse hosts will be key to unravelling the effect of carbohydrate metabolism on host specificity.

Another study (110) showed that the host also directly feeds a member of the gut microbiota (Snodgrasella alvi) with secreted organic acids, including citrate or 3-hydroxymethylglutarate. Whether such direct interactions between bees and their gut microbiota mediate host specificity remains to be clarified but represents an exciting new hypothesis. To further explore this idea, it would first be important to establish if the secreted metabolites are also present in other host species (e.g., using targeted metabolomics).

Gut anatomy, physiology, and biofilm formation

Some gut bacteria can physically attach to a secreted non-cellular layer of their host epithelium (e.g., mucus layer in mammals), sometimes even forming a biofilm (i.e., a community of microbes “adhering to each other or a substrate and encased in an extracellular matrix” [111]). Theory predicts that this physical intimate interaction between host and microbes (with or without biofilm formation) can act as a host-filtering mechanism (112) and therefore emerges as a natural candidate for mediating host specificity (113). Successful physical attachment and colonization of host tissues require (i) that symbiont growth rate and attachment dynamics match the speed at which the gut content is moved downstream the gastro-intestinal tract (retention time), and (ii) there is also a match between host and symbiont molecular mediators of the interaction.

Furthermore, retention time can play a role in host specificity if it varies across host species. Empirical evidence suggests that retention time differs between honeybees and bumblebees, and that this impacts the success of colonization of a honeybee pathogen (Nosema ceranae) in bumblebees (114). This suggests that retention time is one possible mechanism that filters native from non-native microbes, but further work is needed to test this hypothesis for gut microbes and across a wider range of hosts.

Physical interactions between host and symbiont require molecular effectors. In mammals, candidates for host molecular mediators of this adhesion include mucosal glycans and immunoglobulin A (IgA, see also section on the immune system below) (112), while on the bacterial side, capsular polysaccharides have been shown to be key for effective adhesion to the host mucus layer (115), along with other candidates, such as serine-rich surface adhesins (113). In bees, the host hindgut (ileum + rectum) epithelium is coated with a chitin layer referred to as the cuticle (Fig. 5). Microscopy work suggests that some gut microbes, especially Gilliamella sp. and Snodgrassella sp. can attach to it and form biofilms (116). Both Gilliamella sp. and Snodgrassella sp. from A. mellifera encode proteins involved in the formation of capsular polysaccharides (92, 117), and a recent study has shown that this even extends to strains isolated from stingless bees (49). Critically, mutagenesis of Snodgrassella from A. mellifera revealed that host colonization determinants include cell surface structures potentially implicated in physical interaction with the host, such as type IV pili, LPS O-antigens, and trimeric autotransporters (118), and this interaction may underlie host specificity (119). While the importance of bacteria–host interactions for native host colonization is increasingly recognized, it is currently unclear whether this can mediate host specificity in the bee gut. Intriguingly, in pathogenic mammal-associated bacteria (Bartonella sp.) that are closely related to bee symbionts, type IV secretion systems have been found to mediate host specificity via an adhesion mechanism (120). Further experimental work, such as targeted mutagenesis or reciprocal insertion of candidate genes from bacteria isolated from different hosts, will be key to testing whether a similar mechanism mediates host specificity in bee gut bacteria.

Fig 5.

Molecular processes underpinning filtering in host specificity. The figure is hypothetical (if annotated with an interrogation mark) and suggested molecular effectors of host filtering relevant for host specificity. A native host is presented along with focal native (green) and non-native (yellow) bacteria, as well as native diet and resident gut community (represented by a single member for simplicity).

Host immunity

The immune system is well known to shape gut microbiota composition in mammals, as suggested by mutant mice deficient in key immune system molecules like bacterial sensors (e.g., Toll-like receptors) that have an altered microbiota (121). It has also been suggested that the immune system could be a key factor controlling host specificity (17, 122), and this has been empirically validated for innate immunity in the Cnidarian Hydra (98) but rejected for adaptive immunity in mammals (24).

Bees possess an innate immune system with humoral and cellular effectors (123, 124), and the gut microbiota is known to stimulate the immune system (125). A recent important study (91) sought to investigate whether the observed host filtering of non-native Gilliamella spp. strains (described in the “ecological processes” section above) was due to a differential host immune response or sensitivity between native and non-native strains. The authors elegantly demonstrated that the poor ability of non-native strains to colonize non-native hosts was partly due to an increased immune response of the host in reaction to the presence of non-native Gilliamella strains. This inhibitory effect was mediated by an increase of reactive oxygen species (ROS) in the gut through immune deficiency and Toll pathways but not via anti-microbial peptides (simplified representation provided in Fig. 5). However, the inhibitory effects of ROS only partially explain the poor ability of non-native strains to colonize non-native hosts: when this immune pathway is experimentally inhibited (via RNAi or supplementation of vitamin C), non-native Gilliamella load only partially reaches native load levels. A back-of-the-envelope estimate suggests that this pathway contributes about 1% to the colonization defect of non-native Gilliamella (non-native load ≈ 10⁶ marker gene copy/gut, native load ≈ 10⁸ marker gene copy/gut, and non-native load with immune pathway inhibited ≈ 10^6.5 marker gene copy/gut; 10^6.5/(10⁸–10⁶) ≈ 1%). This suggests that other mechanisms promote host specificity for Gilliamella spp., and/or that the experimental inhibition of the specific immune pathway was incomplete.

Interaction with other microbes

One potential process that could maintain host specificity but has not been empirically tested so far relates to the effect of the native community on native and non-native focal symbionts. For example, the positive effect (e.g., cross-feeding) of the native microbiota on a native member could be decreased or absent with non-native symbionts, maybe preventing non-native colonization success. While there are multiple lines of evidence suggesting that both positive (103) and negative (126) interactions are common in the bee gut (33), it remains unknown whether these interactions are specific to native symbionts or if they extend to non-native symbionts. To test this hypothesis, future studies could complement the colonization assays described above by adding one or multiple other native and non-native symbionts. If the hypothesis is true, we expect that the co-culture of focal species with native symbionts would facilitate colonization, while co-culture with non-native symbionts would hamper colonization.

Transmission mode, dispersal, and priority effects

In addition to host or competitive filters, dispersal limitation can maintain host specificity if microbial exchange between allospecific individuals (from different host species) is low compared to that between conspecific individuals (from the same host species). This is particularly expected in eusocial bee species where pupae emerging into adults mostly interact exclusively with the hive environment and conspecifics for days until they transition into foragers and leave the hive on foraging trips, where they can potentially encounter other social bee species (e.g., on flowers).

The route of acquisition and exchange of the gut microbiota in bees has been mostly experimentally studied in honeybees and some bumblebee species. In honeybee larvae, the shedding of the gut lining prevents larval microbiota from remaining in the gut. The new adult emerges with little to no gut bacteria and is colonized within a couple of days (33). Microbes transmitted to newly emerged bees are thought to originate from conspecifics but not entirely through trophallaxis (food transfer between conspecifics) (127) and from the hive environment (e.g., brood frame), especially in the case of rectum symbionts (Lactobacillus sp., Bombilactobacillus sp., and Bifidobacterium sp.) (128, 129). The exact mechanism and vector of transmission remain unclear, but one recent study (130) suggests that transfer from fecal material is sufficient to recapitulate native gut microbiota structure, although it remains unclear how often adult bees come in contact with the feces of their nestmates in the hive. In contrast to honeybees, bumblebee workers do not overwinter (i.e., they all die at the end of the warm season, and only the queen survives the cold season), and the queen establishes a new colony each year, which constrains the transmission of gut bacteria across generations (71, 131, 132).

While it is expected that transmission route and limited transfer between allospecifics foster host specificity, there are very few tests of this hypothesis. One potential indirect test of this hypothesis could be to compare the gut microbiota composition of parasitic social bees (e.g., “cuckoo” bumblebees) to that of their bumblebee host. If dispersal limitation contributes to host specificity, we expect a parasitic bumblebee to host a gut microbiota very similar to its host (76).

Interaction between dispersal and filtering mechanisms is also likely to reinforce host specificity via priority effects. Priority effects are defined as “scenarios where the outcomes of species interactions depend on their arrival times or initial abundances” (133). As native microbes are expected to be more likely first arrivers in recently emerged adult bees, we expect that they have an advantage in colonizing their native host over non-native gut symbionts. The potential importance of this arrival order advantage can be tested by observing the competition between native and non-native gut bacteria across a range of initial population size ratios. For example, reference 48 has elegantly shown that such initial inoculum size differences between native and non-native strains can drastically impact the outcome of microbial interaction. While Snodgrassella strains native to A. mellifera can be outcompeted by some non-native strains (isolated from other Apis species) in A. mellifera, this outcome is less pronounced when the native strain has an initial inoculum size advantage. This suggests that the priority effect could play a key role in maintaining host specificity when physiological differences between strains are small, which is most likely the case between recently diverged hosts. Future experimental studies of cross-inoculation between closely related hosts at various inoculum ratios and/or introducing inoculum of similar size but at different time points will be key to estimating the importance of priority effects in host specificity.

OUTLOOK: CONSEQUENCES OF HOST SPECIFICITY

Gut microbiota can have a net positive effect on many animal hosts, including bees (39), as well as neutral effects (9). In honeybees and bumblebees, the gut microbiota seems to confer resistance to pathogens (134, 135), increas learning and memory (136, 137), and impact social interactions (138). An important question is to what extent the native microbiota is needed to provide maximal benefits for the host and whether non-native microbiota can also provide these benefits. Preliminary results in germ-free rodents inoculated with non-native bacteria suggest that benefits are higher when conferred by native microbes (13, 139, 140), but little is currently known in bees. These results could arise from non-native bacteria being unable to confer the benefits native bacteria do or a lack of efficient colonization by non-native bacteria. We expect bees to represent a promising model to explore the consequence of host specificity because non-native bacteria can colonize A. mellifera at various levels; well-defined native communities are easy to produce; and standard host phenotypic readouts are starting to emerge. Furthermore, a key aspect will be to measure to what extent these different types of benefits can only be provided by the native microbiota as opposed to any other community of microbes. For example, we could expect pathogen colonization resistance to be a general benefit provided by a dense and diverse microbiota community but not necessarily only the native one (141). In contrast, benefits, such as memory or social interaction, that may require more intimate molecular dialogue between the host and its symbiont might only be provided by the native microbiota.

Biographies

Florent Mazel received a Ph.D. in Ecology from University of Grenoble Alpes (France) in 2016. Since 2017 he has mainly focused on animal microbiome evolution. He developed his microbiome interest during a first postdoc at the University of British Columbia (Canada) and second one at University of Lausanne (Switzerland). He will continue exploring this topic as an assistant professor in 2025 at the University of Lausanne.

Aiswarya Prasad is a Ph.D. candidate in Quantitative Biology at the Department of Fundamental Microbiology, University of Lausanne (Switzerland), focusing on the ecology and evolution of gut microbiota using honeybees as a model organism. She earned her bachelor’s and master’s degrees from the Indian Institute of Science (India), where she developed and applied ‘omics approaches to study the human gut microbiome in clinical samples. Since 2019, she has investigated gut microbiota diversity and function in honeybees using bioinformatics and experimental approaches, aiming to advance our understanding of microbial ecology toward enabling microbiome engineering.

Philipp Engel received a Ph.D. in Infection Biology from University of Basel (Switzerland) in 2010. During his postdoc with Nancy Moran at Yale University he became interested in the study of microbial symbiosis. This is when he started to work on the gut microbiota of honeybees. Since 2014 he holds a professorship at the University of Lausanne, where his group studies the ecology and evolution of host-associated microbial communities using social bees as model and combining computational and experimental approaches.