Experimental evidence for the absence of a functional gut microbiome in the solitary bee Megachile tosticauda

Abstract

The eusocial honey bee is a model for insect microbiome research, with socially transmitted gut communities that play key roles in health and development. In contrast, solitary bees lack social transmission pathways and often have an environmentally acquired microbiome, which may or may not be functionally important. Using 16S rRNA gene metabarcoding, previous work has described the bacterial communities in the solitary resin bee Megachile tosticauda pollen provisions, brood, and adult bees, but their functional significance has yet to be studied. Here, we investigate the importance of live bacteria for larval development and survival with an antibiotic-feeding experiment, and test whether bacteria are present in adult guts using scanning electron microscopy. Removing live bacteria in the food and gut of feeding larvae had no significant effect on survival or growth. Microscopy revealed no bacterial colonization of the adult gut, and the dominant taxon detected in larval pollen could not be cultured under targeted conditions. These results suggest an absence of a beneficial gut microbiome in M. tosticauda. Based on our findings, we propose that some bacteria detected by DNA-based methods in M. tosticauda may represent relic DNA, surface-associated or transient bacteria. Our findings highlight fundamental differences in the bacterial associations between social and solitary bees.

This study shows that a solitary Australian bee lacks live gut bacteria and develops normally without them, supporting the view that animal reliance on microbiomes exists along a continuum.

Introduction

Insects can form host-specific symbiotic relationships with bacteria, such as gut microbiomes, and these symbioses can be critical to the host’s health and development. A well-known example is the eusocial honey bee (Apis mellifera), which has communities of host-specific gut bacteria that remain stable independent of location and diet (Zheng et al. 2018). The honey bee microbiome plays a role in metabolism, immune function, protection against pathogens, growth and development (Raymann and Moran 2018). The honey bee is often considered a model species for microbiome research, as it offers unique experimental opportunities to study the assembly of specialized communities and assess the influence these communities have on host biology (Zheng et al. 2018). Importantly, the honey bee microbiome is driven by sociality. Trophallaxis and the shared nest environment facilitates the transmission of microorganisms between individuals and generations, which has allowed for the co-evolution and the formation of mutualist relationships (Powell et al. 2014, Tierney et al. 2025). Consequently, fundamental lessons from the honey bee model cannot be directly applied to most other bees, because more than 90% of species are not social (i.e. are solitary; Danforth et al. 2013).

Despite lacking social inheritance, the bacterial communities of solitary bees often depend on the species, are consistent across individuals, and independent of location (Martinson et al. 2011, Voulgari-Kokota et al. 2019). Whilst these are typical traits of host-specific bacteria, they could also be related to dietary differences, supported by the observation that adult microbiomes frequently resemble the floral substrates they forage on and larval microbiomes reflect those of their pollen provision (larval food) (Williamson et al. 2025a, McFrederick et al. 2012, Kapheim et al. 2021, Keller et al. 2021, Nguyen and Rehan 2022, Martin et al. 2025). Some studies suggest bacteria in pollen provisions are essential for larval development, either by making nutrients available or serving as a food source (Steffan et al. 2019, Christensen et al. 2021, Leonhardt et al. 2022 Dharampal et al. 2019). More research is needed on the characterization and functional importance of solitary bee microbiomes.

Studies of insect microbiomes (and those of animals more broadly) typically involve sampling organisms, extracting DNA, performing PCR amplification targeting the 16S rRNA gene (a marker for bacteria), and gene sequencing to identify microbial communities (Caporaso et al. 2011). Yet the detection of bacterial sequences alone does not confirm their ecological role, and not all bacteria detected through 16S rRNA gene sequencing are part of an insect’s microbiome. In some cases, sequences may represent contaminants, and the lack of contamination control in the field suggests that a considerable proportion of bacteria in the insect-microbiome literature are contaminants (Williamson et al. 2025b). Moreover, the presence of bacterial sequences does not necessarily imply that these microbes are active symbiotic bacteria. While consensus on the terminology used within the field is still developing (Berg et al. 2020), we define ‘beneficial bacterial symbionts’ as bacteria that provide bioactive functions that are beneficial for the host’s survival and health, but the relationship is not necessarily reciprocal (not mutualistic). We define the microbiome as containing all microorganisms including what we will collectively refer to as an ‘incidental microbiome’, i.e. commensals, transients, or relic DNA.

Commensals are microbes that replicate within the host environment, without detriment or obvious benefit to the host (Hooper and Gordon 2001). Transient microbes do not replicate within a host, but pass through and may have positive, negative, or negligible effects on insect hosts (Hammer et al. 2019). Transient bacteria may be found within the insect, as well as on its exoskeleton. While surface sterilization is commonly deployed to eliminate external bacteria, in some instances this method does not significantly alter the bacterial communities detected by sequencing (Hammer et al. 2015). Relic DNA is genetic material left behind by dead cells that can still be amplified and sequenced (Lennon et al. 2018). Both transients and relic DNA have been shown to influence ecological assessments of microbial communities as they interact differently with hosts compared to beneficial symbionts and pathogens (Lennon et al. 2018, Snell Taylor et al. 2018, Berg et al. 2020). Finally, some insect species may altogether lack beneficial bacterial symbionts (Hammer et al. 2019).

Determining whether DNA detected using 16S rRNA gene sequencing indicates the presence of beneficial bacterial symbionts requires multiple lines of evidence. Experimental application of antibiotics can be used to assess whether live bacteria are actively contributing to host metabolism or immune function. In honey bees, the application of antibiotics has been shown to increase mortality in larvae, prepupae, and adults (Peng et al. 1992, Raymann and Moran 2018, Duan et al. 2021). The location of bacteria, i.e. their presence in a certain organ where physical structures separate microbes in space, is also a useful tool for identifying symbionts (Fronk and Sachs 2022). In honey bees, microscopy has been used to describe the location of core bacteria in the gut, with specific bacteria occupying the ileum and rectum (Callegari et al. 2021). Finally, culturing is another tool in microbiota assessments, although limited by presence of non-culturable microbes and the constraints posed by media and incubation conditions. Culturing is frequently applied to already identified beneficial symbionts, such as for metabolic experimentation as well as whole genome assessments from pure isolates which facilitate transcriptomic work. Additionally, culturing can be used to compare community profiles between culture dependent and independent assays, helping to link community structure to viability of microbes (Nocker et al. 2007, Emerson et al. 2017).

Our recent study used 16S rRNA gene sequencing to identify the bacteria associated with the Australian native solitary bee Megachile tosticauda, throughout its developmental cycle (Williamson et al. 2025a). We found that the bacterial community composition of M. tosticauda specimens was consistent over two years of sampling and depended on developmental stage (Williamson et al. 2025a). The communities associated with adults and pollen provisions were similar, consisting primarily of Acinetobacter, which decreased in relative abundance with larval age. Older larvae and prepupae harboured mostly Tyzzerella—a taxon that has been reported in brood of other Megachilidae such as Osmia bicornis (Mohr and Tebbe 2006) and O. excavate (Wang et al. 2024) as well as honeybee larvae (Maigoro et al. 2024). The consistency and abundance of Tyzzerella within samples of M. tosticauda brood, along with its frequent associations with other juvenile bee species, has flagged it as a potential beneficial symbiotic partner. Here, we investigate the possible role and function of the described bacterial communities. Using an antibiotic diet modification experiment, we assess whether removal of live bacteria from the food supply and gut of developing larvae influences the weight and survival. In addition, we use scanning electron microscopy (SEM) to investigate whether, similar to honey bees, adult females possess specific bacteria localized in the gut.

Materials and methods

Diet modification experiment

We applied antibiotics to pollen provisions to determine whether the suppression of live bacteria influences larval development and survival. To obtain young brood, we observed the provisioning behaviour of adult female M. tosticauda in a bee hotel in mid-March 2023 (Eden Hills, South Australia), and sampled nests (n = 10) within a day after they were observed to be completed and closed off with resin. Detailed observations on nesting and foraging behaviour are provided in Williamson et al. (2025c). Each nest was carefully opened in a laminar flow hood, by making a slit along the side with a sterile scalpel. The nests contained eggs or larvae of different ages. The number of brood per nest was counted, and the developmental stage recorded, and the length of the larvae and their respective cells was measured.

Each brood and its pollen provision were weighed and then transferred into a well of a sterile 96-well plate. Treatments were assigned at the nest level, with all brood within a given nest receiving the same treatment. Allocation to treatment was done in such a way to achieve a similar distribution of brood numbers per group. The two treatments were; (1) ‘the antibiotic group’ where the pollen provisions of brood (n = 25; 5 eggs and 20 larvae from five nests) were treated by pipetting 2 μl of 25% antibiotic solution (penicillin 5590 units/ml, sreptomycin 5 mg/l, neomycin 10 mg/l; Sigma–Aldrich #P4083) onto the pollen provisions and, (2) ‘the control group’, where brood (n = 28; 9 eggs and 19 larvae from five nests) had 2μl of sterile water (Sigma–Aldrich) pipetted onto each provision.

In both treatments, the liquid pipetted onto the provision would soak into the pollen ball, which were moist and viscous inside, but contained a dry outer layer (Williamson et al 2025c). The treatments were reapplied every 3 days and the 96-well plates were incubated at 24°C and kept in total darkness: standardized conditions recommended by Becker and Keller (2016). We chose to use antibiotics over gamma irradiation of larval food due to logistical problems: harvesting the pollen from the brood cells, sending it away for sterilization and then feeding to newly harvested young brood would have more than doubled the number of brood cells required, and does not preclude introduction of bacteria when the brood is placed on the pollen supplies. The antibiotic solution we chose has been used in other insect microbiome studies and is understood to remove bacteria without having adverse health effects on the treated insect (e.g. in mosquitoes; Contreras-Garduño et al. 2015, Alvarado-Delgado et al. 2016, Alvarado-Delgado et al. 2021). The brood was checked daily to assess survival. The experiment ended when brood was observed to have transitioned to final instar larvae (i.e. when they had finished eating their provisions, started voiding, and spinning cocoons), which provided a consistent and easily identifiable developmental point for sampling across treatments, minimizing variation between samples. The final instar larvae were weighed and sampled for subsequent 16S rRNA gene sequencing, to verify that the antibiotics had inhibited bacterial growth inside the developing larvae. We did this by comparing our treated samples to those that had developed undisturbed inside four nests, referred to henceforth as ‘natural larvae’ (n = 10) which allowed us to obtain a baseline of the natural bacterial load.

Culturing pollen provisions

Prior to commencing the diet modification experiment, we cultured pollen provisions (n = 5) on tryptone soy agar (TSA; ROWE scientific) that had been treated with our antibiotic solution (details on the solution are given above) at concentrations of 100%, 25%, 10%, 1%, and 0.1% (Fig. S1), diluted with sterile water (Sigma–Aldrich #W3500). This allowed us to verify that we could successfully inhibit bacterial growth, and to determine a minimum concentration of antibiotics that could be used. We chose TSA as the media as it has been successfully used in previous studies to grow Acinetobacter, the dominant bacteria in our pollen provisions as shown by 16S rRNA gene metabarcoding (Williamson et al. 2025a, Kim et al. 2014, Christensen et al. 2021).

To investigate the identity of the live bacterial strains in the pollen provisions, we also cultured pollen provisions without antibiotics. Samples of pollen provisions were prepared for culturing by mixing one provision with 1 ml of sterile phosphate-buffered solution (PBS), followed by a five-step serial dilution. Each dilution was plated on TSA and repeated in triplicate. Single colonies were repeatedly transferred to fresh plates using aseptic technique to isolate pure strains. Pure colonies were then sampled and made into glycerol stocks (Kim et al. 2014), and each isolate was included in our PCR and 16S rRNA gene sequencing. This process also allowed us to ascertain that provisioning females had not added antibacterial gland secretions whilst preparing the provision.

DNA extractions, PCR, and 16S rRNA gene sequencing

To investigate the effect of our antibiotic treatments on the bacterial communities, and identify cultured isolates, we used 16S rRNA gene sequencing. Our protocol for DNA extractions, PCR, and sequencing are detailed in Williamson et al. (2025a) and summarized as follows. All samples were stored at −80°C until DNA extraction. We extracted DNA in a pre-PCR laboratory from homogenized insects using the ZymoBIOMICS DNA miniprep kit. For each batch of DNA extractions, we included an extraction blank control (EBC), which was included in all subsequent PCR and sequencing steps. For isolated bacterial strains, we did not perform DNA extractions and instead added 2 μl of the microbial glycerol stock and combined it with 14 μl of water, and 2 μl of this solution was then added to the PCR master mix.

PCR at the V4 region of the 16S rRNA gene (Caporaso et al. 2010) used the following parameters: single reactions of 2.5 μl ×10 HiFi buffer, 0.1 μl Platinum™ Taq DNA Polymerase (Thermo Fisher), 19.2 μl dH2O, 0.2 μl 100 mM dNTP mix, 0.5 μl each of 10 μM forward (515F 5' -GTGCC AGCMGCCGCGGTAA- 3') and uniquely barcoded 806R reverse primer (806R 5' -GGACTACHVGGGTWTCTAAT- 3'), and 1 μl extracted DNA. Amplification involved an initial denaturation at 94°C for 3 min, followed by 35 cycles of 94°C for 45 s, 50°C for 1 min, 68°C for 90 s, with final adenylation for 10 min at 68°C. The barcoded amplicons were pooled and sequenced on an Illumina MiSeq (v2, 2 × 150 bp).

Quantitative PCR (qPCR) for treated final instar larvae, pollen provisions, and adult bees

We performed quantitative PCRs (qPCRs) on all brood samples, four negative controls and three honey bees as positive controls, and for comparison with our new data we also incorporated six DNA extractions of pollen provisions and adult M. tosticauda samples from a previous study (Williamson et al. 2025a). In addition, we included single reactions of ZymoBiomics bacterial DNA standards containing E. coli at varying concentrations (E2006-2, ZymoBiomics, USA) to generate a standard curve. The qPCRs used the same primers that targeted the V4 region of the 16S rRNA gene and were performed with the following parameters: Single reactions of 7 μl PowerUp™ SYBR® Green Master Mix (Thermo Fisher Scientific, USA), 1 μl each of 5 μM forward and reverse primer, and 1 μl DNA. The DNA was amplified and analyzed on a QuantStudio™ 6 Flex Real-Time PCR System, using an initial hold at 50°C for 2 min, denaturation at 95°C for 2 min, followed by 40 cycles of denaturation at 95°C for 15 s, and a combined annealing and elongation step at 60°C for 1 min. All qPCRs were repeated three times to obtain an average.

Using the bacterial standards, we plotted the standard curve using the CT scores and the logarithm of the total 16S rRNA gene copies, calculated based on E. coli’s genome size (bp) and 16S rRNA gene copy number. The equation of the line from the standard curve was then used to obtain the total gene copies for each biological sample (qPCR calculations provided in the supplementary materials). A figure of the absolute abundance per sample type was made in R (v.4.1.2) using ggplot2. Plots for the relative abundance according to qPCR results were scaled using the total 16S rRNA gene copy number (log scale). The total 16S copy number assumes that number of 16S rRNA genes in E. coli (n = 7, ZymoBiomics) is representative of all taxa in the samples, but in reality this would vary.

Bioinformatics pipeline and statistical analysis

Demultiplexed sequences were analysed using QIIME 2’s pipeline software (v2-2022.8) (Bolyen et al. 2019). Forward and reverse reads were joined using VSEARCH (Rognes et al. 2016). Sequences underwent quality control using the quality-filter q-score command, were denoized with deblur (Amir et al. 2017) and then trimmed to 250 bp. Taxonomy was assigned using a trained SILVA reference classifier with a 99% confidence score (Quast et al. 2013). A phylogenetic tree was made using SEPP phylogenetic placement and fragment insertion (Janssen et al. 2018).

Analysis of bacterial communities at the genus level was done in R (v.4.1.2) and visualized with the ggplot2 package (v.3.4.0). The Decontam package (v.1.14.0) was used to identify putative contaminants using a threshold of 0.5 (Davis et al. 2018). Decontam classifies sequences as contaminants or true microbes of interest using statistical methods based on the frequency and prevalence of ASVs in biological samples compared to negative controls. To explore β-diversity, we used principal coordinate analysis (PCoA) and plotted the ordination of unweighted UniFrac distances (Lozupone and Knight 2005). For relative abundances we plotted the 20 most abundant genera in the treated (antibiotic and control) and in naturally occurring final instar larvae. For some plots we removed reads from contaminants, chloroplasts and mitochondria (i.e. off-target DNA), and rarified to 500 (Fig. S2). We present both filtered and unfiltered data for the final instar larval samples to demonstrate that our samples were low in microbial biomass (Fig. 1).We did not filter out off-target DNA for the qPCR results as the amplification from these assays would represent all the 16S rRNA genes present in the samples. On the filtered/rarified data, we used a dispersion test using vegan package (v 2.6–4) which confirmed our community dispersals were not significantly different between treatments and validated the use of a PERMANOVA in phyloseq to determine whether the bacterial communities between the three larval groups were significantly different.

Figure 1.

The relative abundance of the top 20 most abundant genera in the bacterial communities of final instar Megachile tosticauda larvae. The three larval groups include those sampled from their nest/cocoon after ‘Natural’ development, and those that were reared on ‘Control’ and ‘Antibiotic’ diets. (A) Includes non-target DNA from chloroplasts, mitochondria, and possible contaminants, while in (B) these have been filtered out. (C) The distribution of filtered samples based on unweighted UniFrac distances, where each point represents one sample and ellipses represent the 95% confidence interval around each group.

Statistical analysis of the antibiotic experiment was also done in R. Because brood sex (females typically occupy larger cells than males), development time, provision volume, and final weight differ between cell lengths, we tested potential sex bias by comparing the length of the cells between treatments with an ANOVA. We then used a Chi-square test to determine whether the treatment had a statistically significant effect on the brood’s survival and a t-test to assess whether the treatment influenced the brood’s final weight. At the onset of the experiment, brood were in different larval stages and thus received different periods of antibiotic applications. To control for the differences in the frequency of antibiotic application, we performed the same analyses exclusively using the data from brood that started the experiment as an egg (the antibiotic group, n = 5; the control group, n = 9).

Scanning electron microscopy of the gut tissues and pollen samples

To investigate potential presence of beneficial symbiotic gut bacteria, we caught provisioning female M. tosticauda (n = 3) upon return to their nest. Worker honey bees (n = 3) were caught and processed alongside as a positive control. Within an hour after the bees were caught, the heads were removed with a sharp scalpel and the bodies were carefully dissected under a dissecting microscope to obtain the intact digestive tract, including the oesophagus, crop, midgut, ileum, and rectum.

Once dissected, the digestive tracts were immediately moved under a fume hood and fixed overnight in 4.0% Paraformaldehyde/1.25% Glutaraldehyde in PBS, 4% sucrose (pH 7.2). After 24 h, the guts were washed three times for 5 mins in PBS, and then dehydrated in ethanol using a stepwise approach of 30% ethanol for 30 mins, 50% ethanol for 30 mins, 70% ethanol for 30 mins, 80% ethanol for 30 mins, 95% ethanol for 15 mins repeated three times, and 100% ethanol for 15 mins repeated three times. The guts were then mounted on aluminium stubs with sticky carbon tabs, making sure that the gut lay linearly, and that the five sections were clearly visible. Using sterile pointed forceps, we opened a part of each gut section, so that we could observe the interior structures and features. If partially digested pollen was present inside the rectum, it was gently brushed to the side to ensure visibility of the lining.

In addition to the gut samples, we also obtained pollen from the abdominal scopa of three females, and three small subsamples from pollen provisions. These pollen samples were simply placed onto the carbon tabs and allowed to dry at room temperature for 24 h. All samples were then coated in platinum and visualized using the HitachiFLEX SEM at spot 30. We used the SEM photos to assess the presence of bacteria, and compared photos taken from the honey bee and M. tosticauda gut tissue. We also compared the state of the pollen kernels from M. tosticauda’s scopa to those after being processed into their pollen provisions, and those observed within the digestive tract, to describe whether there were differences in the prevalence of bacterial cells at various stages.

Results

Diet modification experiment

The mean cell length between treatments did not differ (ANOVA: F[1, 51] = 1.41, P = 0.241), indicating no evidence of sex bias in the results. We found no statistically significant difference in the survival and final weight of brood between the antibiotic and control treatments. In total, 86% of the control group survived (24/28) and 64% of the antibiotic group survived (16/25; survival: χ² = 2.29, P-value = 0.13). The application of antibiotics did not affect the brood’s final weight (t = −0.50, P-value = 0.62). The same was found for brood that had received treatments from the egg stage onward (survival: χ² = 0.01, P = 0.93; weight: t = 0.13; P = 0.90).

In total 22 final instar larvae from the control group and 16 from the antibiotic group were sampled for 16S rRNA gene sequencing, to further verify that our antibiotic treatment suppressed bacterial growth relative to the controls and natural larvae. A total of 2 241 188 reads were generated from our samples. One antibiotic treated larva and one natural larva yielded no sequencing reads. Two surviving larvae from the control group were not included for 16S rRNA gene sequencing as they were instead used for culturing and microscopy assays (data unpublished). After filtering and removal of off-target DNA (from chloroplasts, mitochondria and contamination, 45.9% of all sequencing reads) 8/9 natural, 5/15 antibiotic, and all control samples had sufficient reads to be included in the analysis (Fig. 1). The alpha diversity was higher in the antibiotic-treated group compared with the control and natural group and a PERMANOVA test showed that the communities of bacteria differed significantly between the three larval groups (F = 2.9293, P < 0.05; Fig. 1).

The difference in bacterial community, bacterial load and the culturing assays verified that the antibiotics successfully inhibited bacterial growth (Fig. 2). Comparing the qPCR results from the natural and antibiotic-treated final instar larvae, the control group had a higher absolute abundance, with sequencing showing a proliferation of Tyzzerella (Fig. 2).

Figure 2.

(A) Absolute abundance and (B) relative abundance of the top 20 bacterial genera across experimental groups: EBCs, three final instar larval groups (natural, control diet, and antibiotic diet), pollen provisions, adult Megachile tosticauda, and honey bees as a positive control. The top 20 genera account for 91.97% of total sequencing reads. Data were not filtered as amplification from the qPCRs represent all DNA present, including off-target DNA.

The microbial composition of adult female M. tosticauda and pollen was similar. Samples contained high amounts of off-target DNA from chloroplasts and mitochondria, but the relative and scaled absolute abundance of Acinetobacter was higher in adults than in larval pollen provisions (Fig. 2). The bacterial communities of co-foraging honey bees differed substantially from that of M. tosticauda.

Culturing pollen provisions

From the pure bacterial colonies from pollen provisions, we obtained isolates of strains of Erwinia, Enterobacteriaceae, and unknown Enterobacterales (labelled NA_unclassified, Fig. 2), each of which were part of the pollen provision bacterial community as determined by 16S rRNA gene sequencing (Williamson et al. 2025a, photos of the isolates are provided in the supplementary materials, Fig. S3). However, we were unsuccessful in cultivating the most abundant taxon in the pollen provisions, Acinetobacter.

Scanning electron photographs

SEM was conducted on the guts of M. tosticauda and as well as various pollen samples. These samples contain high relative abundances of Acinetobacter (Williamson et al. 2025a). The SEM of honey bee guts validated our protocol as we were able to clearly observe resident bacteria in the lining of the ileum and rectum (Fig. 3), and on pollen grains within the honey bee rectum (Fig. 4). However, there were no visual signs of bacteria within the gut of M. tosticauda (Fig. 3). Furthermore, no bacteria were visible in photos taken of pollen from the scopa of provisioning females, from smears of completed pollen provisions, or from pollen in the rectum of M. tosticauda (Fig. 4).

Figure 3.

Scanning electron microscopic images taken of the adult gut of M. tosticauda (top) and honey bees (bottom). Images were taken for each of the four sections of the gut; the crop, midgut, ileum, and rectum.

Figure 4.

Scanning electron microscopic images taken at 150✕, 1000✕ and 2000✕ magnifications of pollen. From left to right the pollen samples are from the abdominal scopa of provisioning females, from completed larva provisions, from the rectum of adult M. tosticauda, and from the rectum of adult honey bees.

Discussion

Our antibiotic experiment showed that the suppression of live bacteria in the food and gut did not significantly affect brood survival or weight upon entering the final instar larval stage. While 16S rRNA gene sequencing found an abundance of Acinetobacter in adults, these bacteria were not observed in the gut lining. We hypothesize that despite being high in relative and absolute abundance, Acinetobacter may not have been alive in our samples and instead formed part of an incidental microbiome. These findings are explored in more details below.

The natural final instar larvae that had developed in their nests undisturbed harboured a more diverse but less abundant bacterial community than the control group, where brood developed in petri dishes and the pollen supplies were treated with sterile water (Figs. 1 and 2). This difference in the community composition and abundance may be the result of 1) the natural larvae picking up bacteria from the nesting substrate, or 2) the rearing conditions in the sterile wells favouring the proliferation of some microbes (i.e. Tyzzerella). These two possibilities are not mutually exclusive. The former would be consistent with our previous study, which demonstrated that the adult microbiome is determined by the substrates they encounter (Williamson et al. 2025a). The latter has been found for Lepidopteran samples (Lighthart 1988), and a potentially favourable condition in the sterile wells compared to the nests was the absence of resin, which could have antimicrobial properties (Shanahan and Spivak 2021, Chui et al. 2022). Additionally, the higher bacterial diversity in antibiotic-treated final instar larvae can be attributed to low bacterial biomass, which leads to over-representation of contaminants (Paniagua Voirol et al. 2021). The comparison in bacterial absolute abundance between our negative control and antibiotic treatment samples supports this effect of contaminants (Fig. 2).

Our culturing assays and comparison between the absolute abundance of bacteria in antibiotic-treated vs natural larvae demonstrated that our antibiotic treatment successfully inhibited the growth of live bacteria in pollen provisions and larval guts, but this inhibition did not significantly affect the survival or weight of developing brood. Therefore, despite their presence in high relative abundance, we find no support for the hypothesis that Acinetobacter in pollen provisions or Tyzzerella inside brood are beneficial bacterial symbionts involved in the development of larval M. tosticauda. It is possible that Acinetobacter served as a food source for developing brood. This would be consistent with suggestions that bees are omnivorous (Steffan et al. 2019, Dharampal et al. 2020). While the antibiotics stopped bacterial growth and replication, relic DNA and cellular materials from dead and/or inactive bacteria may persist (Lennon et al. 2018) and therefore might still fulfill the nutritional requirements of larvae.

It is unlikely that the bacteria in the pollen provisions of M. tosticauda served as an ‘external rumen’ (Keller et al. 2021), pre-digesting the floral resources and providing larvae access to the nutrients therein, as has been reported in Osmia species (Dharampal et al. 2019, Dharampal et al. 2020, Dharampal et al. 2022). These contradictive results could be due to methodological differences, where the Osmia pollen was sterilized by soaking in ethanol and exposure to UV light overnight, which could impact on the nutritional quality of the pollen. Similar to our findings, Crowley and Schaeffer (2024) showed no significant effect on the larval development of O. lignaria when fed pollen sterilized using ethylene oxide (EO). Additionally, Brar et al. (2024) found that the weight and survival of M. rotundata was unaffected when larvae were fed pollen provisions sterilized using gamma-irradiation (Brar et al. 2024). Taken together, there is mounting evidence that, in at least some cases, environmentally acquired bacteria are non-essential larval development.

While we did not find any statistically significant effect of the antibiotic treatment on larval development, our diet modification experiment has several limitations. First, we did not investigate the long-term consequences of removing bacteria. A study on the ground-nesting solitary bee Anthophora bomboides standfordiana showed that immature bee microbiomes may provide key benefits such as pathogen protection and cold tolerance during overwintering (Christensen et al. 2024). It is also possible that, because our laboratory conditions did not reflect natural environmental conditions or pressures, any benefits provided by bacteria were not evident. For example, if some bacteria provide pathogen protection, antibiotics may have masked this effect by inhibiting both beneficial and pathogenic bacterial growth. Furthermore, we assessed survival between the two groups using a Chi-squared test, but ideally a Cox proportional hazards model would be used for a more accurate outcome. This was not used as the exact time of larval death could not be assessed. Lastly, treatments were assigned to the brood at the nest level, which does not account for between nest variation in survival rates and weight gain. Thus, based on our results, it seems unlikely that removal of bacteria by antibiotics affects brood health, but we cannot unequivocally state this.

Our results also do not preclude the possibility that the bacterial community of developing brood or pollen provisions was transient, commensal, or relic, forming a purely incidental microbiome. Transient bacteria are those acquired from the environment that may survive temporarily but do not establish; commensals may live passively without harm or benefit to the host; and relic DNA represents bacterial DNA from dead or lysed cells. Lepidopteran larvae treated with antibiotics have shown a similar independence from beneficial bacterial symbionts despite harbouring specific microbial communities, exemplifying that gut bacteria are not always essential for insect survival (Phalnikar et al. 2019, Brar et al. 2024).

We were able to successfully isolate Erwinia, Enterobacteriaceae, and an unclassified Enterobacteriales, all of which were consistently present in pollen provisions, though in lower abundance than Acinetobacter (Williamson et al. 2025a). This indicates that our culture conditions were suitable for at least some bacteria from the provision environment. While it is possible that this specific strain of Acinetobacter was unculturable (Bodor et al. 2020), we used the medium TSA as it has been successfully used to culture Acinetobacter from bees and floral sources in previous studies (e.g. Álvarez-Pérez and Herrera 2013, Kim et al. 2014, Álvarez-Pérez et al. 2021, Christensen et al. 2021). Alternatively, it is possible that Acinetobacter did not survive the conditions imposed by foraging bees or in the pollen provisions. Acinetobacter’s likely source is provisioned nectar (Vannette 2020), which prior to being mixed with pollen to prepare the larval food may undergo processing, including the concentration of nectar (Portman et al. 2021), and/or the addition of gland secretions. Bacteria may be killed during or because of these processes (Morales-Poole et al. 2023). It is possible that surviving bacteria (like those cultured) are those from pollen transported via the scopa, but this requires research.

In our case, the hypothesis that Acinetobacter reflects relic DNA in this case is supported by the observation that the relative abundance decreased over time in brood and is absent in frass (Williamson et al. 2025a), as well as by the SEM photos, which showed an absence of bacteria in M. tosticauda gut and pollen samples. There was a stark difference between M. tosticauda and honey bees in the bacterial presence in the ileum, rectum, and in partially digested pollen (Figs. 3 and 4), where honey bee samples showed an abundance of bacteria. Together, the lack of observable bacteria in the gut lining, absence of culturable Acinetobacter in pollen provisions, and the finding that antibiotics did not impact larval development suggest that the Acinetobacter picked up from 16S rRNA gene sequencing was relic DNA rather than active bacteria.

Overall, the microbiome identified through 16S rRNA gene sequencing did not indicate the presence of beneficial bacterial symbionts in pollen provisions, feeding larvae, or adult bee guts. This study is the first on an Australian solitary resin bee described as having no gut microbiome and adds to the body of evidence that at least some solitary bees have not evolved specialized gut symbionts (Dobson and Peng 1997, Martinson et al. 2011). However, to close cells and nests, M. tosticauda uses resin, which has antibacterial properties. Therefore further research is needed to determine whether this pattern holds across other solitary bees with different life histories. Our study supports the emerging hypothesis that social behaviour in bees has been key for the evolution, development, and maintenance of specialized gut microbiomes, largely due to the vertical transmission route it provides between generations (Engel and Moran 2013, Kwong and Moran 2015, Kwong et al. 2017).

Our study also exemplifies the importance of distinguishing between beneficial symbionts and the incidental microbiome. The genus Acinetobacter has been reported at high relative abundance (>15% total bacterial community) in several solitary bee species, including Osmia lignaria (Cohen et al. 2020), Svastra obliquua, Megachile fortis, Habropoda laboriosa (Holley et al. 2022), Centris atripes (Kwong et al. 2017), Lithurgus littoralis (McFrederick et al. 2017), and Ceratina calcarata (Nguyen et al. 2023). Yet, whether its presence reflects a symbiotic relationship or an incidental, transient association like we observed in M. tosticauda, remains unclear. Without such distinctions, it is unknown how widespread the absence of a gut microbiome is across solitary bees. How often are 16S rRNA gene sequences from solitary bees transient or relic? Do most solitary bees lack a gut microbiome? Future work should not only describe bacterial communities using 16S rRNA gene sequencing but also investigate the presence of these DNA sequences using microscopy and culturing.

Animals that do not harbour gut microbes have evolved mechanisms that prevent bacterial establishment through chemical, mechanical or immunological means (Hammer et al. 2019). For example, caterpillars possess digestive physiologies that are inhospitable to microbial growth (Hammer et al. 2017). Studying these animals offer unique opportunities to research alternative models of digestion, immune defence, and animal-microbe interactions, but there is a current lack of studies on animals that do not depend on gut microbiomes. When organisms without gut microbiomes are studied, they are usually artificial ‘germ-free’ hosts that are compared to ‘natural’ hosts in an experimental setting (Henry et al. 2021). These are helpful in determining the role of the gut microbiome in animal health but might lack valuable information about the baseline host response. Studying organisms that naturally lack beneficial gut bacteria (i.e. null model organisms) could help distinguish between intrinsic genetic contributions vs microbial contributions to phenotypic traits (Henry et al. 2021) and further our understanding on how and when microbiomes are evolutionarily relevant.

Importantly, the finding that M. tosticauda does not harbour a beneficial gut microbiome does not mean their evolution is entirely decoupled from microbial influences, as no animal has evolved independently from microbes. While some animals have evolved to host mutualistic or commensal bacteria that offer protection from pathogens, like honey bees, animals without beneficial gut bacteria also need protection and could form interactions with external, environmental microbes for this reason. Here we investigated whether bacteria associated with M. tosticauda form bioactive symbioses in the food, larval, and adult gut. Future work should investigate whether Tyzzerella in the nest offers pathogen protection to overwintering brood, and whether Acinetobacter is a food source for developing larvae. Additionally, the cultures from pollen provisions, used to verify the antibiotics successfully inhibited bacterial growth and to isolate pure colonies, revealed a presence of fungi that were unaffected by the antibiotics (Fig. S1). Future research should investigate the potential benefits of fungi in pollen provisions, as they may have an essential role in nutrition, detoxification, and/or pathogen protection (Rutkowski et al. 2023, Christensen et al. 2024).

Conflicts of interest

None declared.

Funding

This work was supported by the Linnean Society of NSW’s William Macleay Microbiology research fund (Grant number 0006013282) and the Royal Society of South Australia small research grant (Grant number 0006014057). This research was supported by the Charles John (CJ) Everard Scholarship, awarded by The University of Adelaide to Elisabeth Williamson. The author gratefully acknowledges this financial support throughout the duration of their PhD.

Data availability statement

All DNA sequences have been uploaded to the NCBI sequence read archive under bioproject PRJNA1146331.