Floral microbes provisioned by Osmia lignaria establish in larval food stores, but do not affect bee development or survival

Abstract

Microbial dispersal and subsequent establishment among linked habitats can be used to examine drivers of community assembly and function. Flowers host microbial communities that can be acquired and vectored by bees to new flowers, establish within the adult bee gut, and enter food stores (e.g. pollen provisions) of developing larvae. Yet, whether microbes vectored by insects or applied for biocontrol can establish across these habitats and if they affect bee fitness remain unknown. Here, we applied microbes to flowers visited by blue orchard bees (Osmia lignaria) and compared microbial communities in flowers, adult bee guts, and pollen provisions before and after inoculation to determine microbial establishment, environmental filtering, and overlap across habitat types. We also inoculated provisions with microbes to test their effects on larval survival, development, and weight. Experimentally inoculated microbes were detected in all habitats, demonstrating that flowers are a source of microbial acquisition for adult and larval bees; however, the tested larval health metrics were largely unaffected by microbe supplementation.

Osmia lignaria acquire microbes from flowers and transfer them to their offspring, though there are minimal effects of floral microbes on larval health metrics. Created in BioRender. Martin, L. (2006)

Introduction

Determining the transmission and establishment of microbial species between habitats is key to understanding their life history and adaptations, and can inform the broader ecological effects of microbial biocontrol application. Like macro-organisms, microbial establishment and subsequent growth depend on dispersal, diversification, selection, and drift (Vellend 2010, Nemergut et al. 2013). After dispersal to a new habitat, microbes encounter biotic factors (e.g. priority effects, competition, symbiosis) and abiotic factors (e.g. pH, oxygen, available nutrients) that can act as environmental filters (sometimes also called habitat filters), selecting for species that can survive and proliferate within specific conditions (Kraft et al. 2015, Mittelbach and Schemske 2015). However, teasing apart the role of these factors in shaping community assembly and function when using microbial surveys is challenging due to sampling limitations, detection biases, incomplete species knowledge, and variation in dispersal routes (Peay 2014, Kraemer and Boynton 2017). Here, we use flowers, bees, and pollen provisions (food stores for developing bee offspring) as a tractable model system to investigate the establishment of microbes across uniquely demanding habitats with the same microbial food resources (pollen and nectar).

Flowers host microbial communities within pollen and nectar (reviewed by Vannette 2020). When bees visit flowers to collect these resources for themselves or their offspring, they also acquire microbial communities established in these resources (Herrera et al. 2010, Corby-Harris et al. 2014, Rothman et al. 2019, Russell et al. 2019). Although social bee species exchange microbes through intraspecific interactions (Kwong and Moran 2016), most bee species (∼77%) are solitary (i.e. each female provisions her own offspring) and lack social interactions. As a result, most bees are thought to acquire a majority of their microbial associates from environmental sources that they interact with, like flowers and materials used for nest-building (e.g. soil, leaves, wood, etc.) (Keller et al. 2013, Voulgari-Kokota et al. 2019, Nguyen and Rehan 2023). It is not entirely understood which environmentally acquired microbes establish in bee-associated habitats (flowers, guts, pollen stores) or how they interact with bees. However, current research suggests environmental microbes may modify floral resources (Herrera et al. 2008, Vannette and Fukami 2016, 2018, Christensen et al. 2021) and impact bee nutrition (Dharampal et al. 2020), survival (Nguyen and Rehan 2025), development (Dharampal et al. 2022), and behavior (Good et al. 2014, Rering et al. 2018, Schaeffer et al. 2019, Pozo et al. 2020) (reviewed by Martin et al. 2022).

Of the solitary bees that have been studied to date, many show evidence of environmental acquisition of microbes within adult bee guts and pollen provisions. For instance, the gut microbiomes of adult blue orchard bees (Osmia lignaria) vary across locations (Cohen et al. 2020) and are closely correlated with pollen composition (Vannette et al. 2025), suggesting that they are primarily environmentally derived. Similarly, other species of megachilid bees show microbial overlap between flowers and bee guts (McFrederick et al. 2017). Microbes identified in the pollen provisions of O. lignaria and Osmia ribifloris are commonly found in flowers and soil (Rothman et al. 2019). Together, these studies strongly suggest that flowers are an important microbial source for both bee gut and pollen provision microbiomes. Yet, the extent to which these microbes successfully disperse between, and establish within, each bee-associated habitat is poorly understood.

Once microbes enter flowers, bee guts, or pollen provisions, they encounter the novel conditions associated with each of these habitats. As a result, microbial species that consistently colonize these habitats can be highly specialized, harboring adaptations to survive, utilize pollen nutrients, and interact with other microbes (Herrera et al. 2008, Dhami et al. 2016, Pozo and Jacquemyn 2019, Alvarez-Perez et al. 2021, Christensen et al. 2021). For instance, a microbe that is adapted to low oxygen environments may be able to establish well in the bee gut, but unable to compete in aerobic pollen provisions. However, no studies have directly tracked the movement and subsequent establishment of these specialized microbes into adult bee guts and their pollen provisions.

The pollen provision is the main route through which solitary bee larvae obtain microbes, so larval microbiomes contain a subset of the microbes consumed (Voulgari-Kokota, Steffan-Dewenter and Keller 2020, Nguyen and Rehan 2022), Kueneman et al. 2023). In some cases, these microbes offer increased survival and development (Dharampal et al. 2019, Dharampal et al. 2020, Dharampal et al. 2022). For O. ribifloris, larvae-fed pollen containing the natural microbiome had increased prepupal biomass, took less time to develop, and had increased overall survival when compared to bees fed sterilized pollen provisions (Dharampal et al. 2020). The provision microbiome may also offer benefits for bees via protection against pathogens. Anthophora bomboides pollen provisions contain the bacteria Streptomyces, which was able to inhibit the growth of the bee fungal pathogens Ascosphaera and Aspergillus in plate-based assays (Christensen et al. 2024). However, additional manipulative lab-based experiments will be useful in further uncovering specific effects of microbes in bee development.

Here, we leverage a tractable flower-bee-microbe model system to directly trace the establishment and dynamics of a microbial community in multiple bee-associated habitats, as well as the subsequent impacts of these communities on larval survival, development, and weight. Specifically, we (1) assessed whether flowers act as sources for microbes in adult bees and pollen provisions, (2) evaluated the role that environmental filtering plays in microbial community establishment across habitat types, (3) examined the extent to which microbial communities overlap between habitat types, and (4) investigated the effect of microbial supplementation on offspring health. To do this, we sprayed an inoculum of bee- and flower- associated microbes onto flowers and then sampled the microbial communities within the flowers, adult O. lignaria guts, and pollen provisions before and after the treatment. We also applied microbial inoculum directly to pollen provisions and monitored larval survival and development. We hypothesized that flowers would serve as a microbial source for bees and their pollen provisions, but microbial species would vary in their ability to establish among habitat types depending on their initial isolation source (i.e. flower, bee gut, pollen provision). Additionally, if flowers are the only source of microbial acquisition, we expected to find very few microbes unique to guts and provisions. Finally, if microbes within the inoculum positively impact larval bee health, we expected to observe increased survival, increased emergence rates, and less weight loss during non-feeding developmental stages.

Methods

Study system

Osmia lignaria (Say, 1837), the blue orchard bee, is a solitary, cavity nesting mason bee species native to North America (Fig. 1a, Rust 1974). As solitary bees, each female creates her own nest. After locating a suitable nest-site, an O. lignaria female collects pollen and nectar from flowers to create a pollen provision, lays an egg on top of the provision, and seals the brood-cell with a mud wall (Fig. 1b, Torchio 1989). Each offspring spends its entire larval and pupal stages in this nest, ecloses as an adult in fall and overwinters in diapause within its cocoon. New adults emerge in spring to mate, and females repeat the nesting cycle annually (Torchio 1989). Populations of O. lignaria are both wild and commercially managed for use in crop pollination (Torchio 1991).

Figure 1.

(a) Adult O. lignaria female on a Phacelia tanacetifolia flower. (b) Cross-section of an O. lignaria nest filled with pollen provisions and eggs, separated by mud walls. (c) Hoop house planted with P. tanacetifolia. (d) Schematic of sample collection, where all steps were repeated for two rounds except bee collection(*). Photos taken by Alexia Martin and Rachel Vannette.

We performed two experiments: one hoop house experiment to investigate acquisition and community assembly of floral microbes (Aim 1–3) and one lab experiment to test larval performance in response to microbe addition to pollen provisions (Aim 4). To investigate acquisition and community assembly, six microbes were chosen as focal species to inoculate into flowers (“focal microbes”): two bacteria (Acinetobacter pollinis and Apilactobacillus micheneri) and four fungi (Metschnikowia reukaufii, Starmerella bombicola, Debaryomyces hansenii, and Aureobasidium pullulans) (Table 1). These microbes are all frequently found in association with both bees and flowers (McFrederick et al. 2017, Alvarez-Perez et al. 2021, Rutkowski et al. 2023); however, their primary habitats within these systems remain unclear, and movement between these habitats is poorly understood. Additionally, these microbes were not previously detected in Phacelia tanacetifolia flowers under semi-controlled conditions experienced within the experimental hoop houses used for this study (RL Vannette and NM Williams, Unpublished Data).

Table 1.

Focal microbes used to inoculate flowers, the isolation source of the strain included in the study, the inoculation round they were used, and their recovery in post-spray samples.

Microbe type	Microbe name	Strain ID	Strain isolation source	Spray round used	Recovered in samples?
Bacteria	Acinetobacter pollinis	SCC477	Scrophularia californica flower	1, 2	Yes
	Apilactobacillus micheneri	HV60	Megachile rotundata gut	1, 2	Yes
Fungi	Metschnikowia reukaufii	EC52	Epilobium canum flower	1, 2	Yes—through sanger sequencing
	Debaryomyces hansenii	UCDFST 67–77	Bombus impatiens gut	2	Yes
	Starmerella bombicola	UCDFST 02–305	Apis mellifera gut	2	No
	Aureobasidium pullulans	Gall_Forb_Edge_N_Y1	Prunus dulcis flower	2	No
	Vishniacozyma victoriae	N/A	Environmental*	1,2	Yes

^** = Can be found in the phyllosphere (Gouka et al. 2022), but was not included in the inoculum.

We chose a community of three bacteria to inoculate into pollen provisions (Apilactobacillus kunkeei, Acinetobacter pollinis, and Pantoea agglomerans). The presence of these bacteria is variable among pollen provisions (Voulgari-Kokota et al. 2019, Dew et al. 2020, Kueneman et al. 2023, Vannette et al. 2025), but their effects on bees are of interest due to their role as a potential probiotic (Chege et al. 2023, Daisley et al. 2023, Nguyen and Rehan 2025, Usta et al. 2025) or biocontrol agent in agricultural systems (e.g. Stockwell et al. 2002, Kim et al. 2012).

Osmia rearing in hoop houses

This experiment was conducted in the spring of 2022 at the University of California, Davis Harry H. Laidlaw Jr. Honey Bee Research Facility in two 6.3 m × 11 m × 2.04 m screen hoop houses. The hoop house frames were constructed using a wooden base and metal beams, then screened in with a well-fitting fine mesh fabric (Reaco Brand Cage, Santa Rosa, CA). The California native wildflower Phacelia tanacetifolia Benth. was grown from the seed bank, as these flowers had been planted within the hoop houses in previous years (Fig. 1c; Stuligross and Williams 2020, 2021, Melone et al. 2024). Phacelia tanacetifolia is used by O. lignaria and provides high-quality pollen and nectar resources (Williams 2003, Boyle et al. 2020). When flowers approached full bloom in April 2022, adult male and female O. lignaria were released into each hoop house. Approximately 8–15 females were active at any given time, which was enough to ensure consistent nesting. Bees were provided with wooden nest blocks lined with new paper straws to mimic preferred nest sites and facilitate sampling. Nesting blocks were housed in corrugated plastic shade boxes called “fiesta boxes” (Fig. 1c; Boyle and Pitts-Singer 2019). Bees were also provided with a consistent mud source to use for nest construction.

Microbial suspension preparation and application

To prepare microbial treatments, microbes were plated from glycerol stocks onto Yeast Media Agar (M. reukaufii, S. bombicola, D. hansenii, Au. pullulans), Tryptic Soy Agar (Ac. pollinis) or De Man–Rogosa–Sharpe agar + 2% Fructose (Ap. micheneri). Plates were incubated for 3–5 days at 28°C. On the day of floral inoculation, microbes were suspended in 1 l sterile water at ∼4000 cells/μl per microbe (determined through cell counts on a hemocytometer). The suspension was mixed by gentle inversion and transferred into two clean spray bottles. Microbes were applied directly onto open flowers (∼80% of all open flowers) in each hoop house using the spray bottle, with 500 ml of inoculum applied to plant surfaces per hoop house. Individual flowers are open for three days (Williams 1997), with flowers opening sequentially on an inflorescence. Individual bees can live for up to six weeks and presumably vector microbes among individual flowers. The inoculation was performed in two rounds, with Spray Round 1 (three microbes, ∼865 million cells/m²) occurring on 22 April 2022 and Spray Round 2 (six microbes, ∼1.7 billion cells/m²) occurring 26 April 2022 and 28 April 2022 (Table 1). Two rounds of inoculation and sampling allowed us to increase replication and examine if focal microbes were maintained in the system after inoculation (likely via bee vectoring among flowers).

Flower, adult bee, and provision sample collection

Prior to spray inoculation, pooled flower samples (3–4 individual flowers per sample; n = 9) and adult female bees (n = 6) were collected from both of the hoop houses (Fig. 1d). Additionally, completed cells within each nest were marked with permanent marker and assigned as being collected pre-spray inoculation. Pooled flowers (3–4 individual flowers) were collected 1–3 days following each spray round (n = 12), and adult female bees were collected 1–3 days following Spray Round 2 (n = 12). Nest progress was monitored every 1–3 days for 22 total days with completed cells marked as being created post-spray inoculation. Following completion, nests (n = 8) were destructively sampled to obtain pollen provisions, which were assigned as being created pre- or post-spray inoculation (n_pre=21; n_post=19).

DNA extraction and sequencing

Whole guts (including the crop, midgut, ileum, and rectum) were sterilely dissected from adult bees. Provisions were dissected from nesting straws using sterile technique and offspring (eggs, first instar larvae, or second instar larvae) were removed. Flower samples were prepared for DNA extractions by adding the pooled flowers to a 5 ml tube, covering with 1–2 ml Phosphate Buffered Saline, vortexing for 30–60 s, sonicating at 50% power for 20 s (Branson Bransonic Ultrasonic Bath 2800), pipetting off the liquid, centrifuging the liquid for 3 mins at 14 000 rcf to obtain a pellet, pipetting off 0.5–1.5 ml of the liquid, and resuspending the pellet in the remaining 500 μl of PBS (protocol based on Ushio et al. 2015). DNA was extracted from full guts, approximately 1/3 of each pollen provision, and the processed flower samples using the QIAGEN DNEasy PowerSoil Kit according to kit instructions with two modifications. First, the bead beating step was performed on a Benchmark BeadBlaster 24 and included four 20 s cycles. Second, metal beads were added to gut samples during bead beating to improve tissue maceration. The inocula from the second spray round were included as positive controls, and two kit blanks were used as negative controls. DNA extract was sent to the Integrated Microbiome Resources at Dalhousie University in Halifax, NS for sequencing of the 16S rRNA (V5/V6) region using 799F (5′-AACMGGATTAGATACCCKG-3′) and 1115R (5′-AGGGTTGCGCTCGTTG-3′) primers (Chelius and Triplett 2001, Anguita-Maeso et al. 2022, Christensen et al. 2024) and ITS region using ITS86F (5′-GTGAATCATCGAATCTTTGAA-3′; Turenne et al. 1999) and ITS4R (5′-TCCTCCGCTTATTGATATGC-3′; White et al. 1990) primers via Illumina MiSeq (2×300 bp PE).

Sequence processing

Sequence data was processed in R (Version 4.2.2 “Innocent and Trusting”; R Core Team 2022) using the standard dada2 pipeline (Callahan et al. 2016) to remove primers, merge forward and reverse sequences, and remove chimeras. For the ITS pipeline, primers were removed via “cutadapt” (Martin 2011), and the filtering criteria “maximum errors allotted” was set to five for reverse sequences to allow more to pass through the pipeline. Taxonomy was assigned using the Silva database v138.1 (Quast et al. 2013) for the 16S data and the Hybrid Unite Database (format 2; Nilsson et al. 2019) for the ITS data. Chloroplast and mitochondrial sequences were removed from the dataset using the Phyloseq package (McMurdie and Holmes 2013). Excluding positive and negative controls, an average of 92% of reads per sample made it past quality filtering for bacteria; however, only an average of 8.8% of reads per sample remained after filtering out chloroplasts, mitochondria, and contaminants with most cuts occurring after the mitochondrial removal step. On average, there were 2169 bacterial reads per sample following filtering. For fungi, an average of 78% of reads per sample made it through both quality and taxonomic filtering with an average of 25 730 reads per sample following filtering. Detailed results from the bacterial and fungal pipeline can be found in the supplemental information. Two sequenced kit blanks were used with the “Decontam” package (Davis et al. 2018) to identify contaminant sequences based on their prevalence across samples and negative controls, resulting in four contaminants removed from the bacterial dataset and no contaminants identified in the fungal dataset. Sufficient sequencing depth was assessed using sampling curves with the “rarecurve” function in the “vegan” package (Oksanen et al. 2024). Bacterial samples with fewer than 200 reads (n = 11) and fungal samples with fewer than 500 reads (n = 2) were removed from their respective datasets (Fig. S1). For bacteria and fungi, unassigned ASVs in the top 30 most abundant sequences were manually assigned using NCBI BLAST (Altschul et al. 1990) with a 98% cutoff. Any ASVs with multiple equal percentage matches were assigned to the lowest taxonomic level in common between matches. For the fungal pipeline, 42% of ASVs were unassigned at the Kingdom level. To account for this, any ASVs that were present in higher than 0.01% abundance in the entire dataset were manually assigned using NCBI BLAST at a 98% cutoff. Any BLASTed sequences which were assigned as anything other than fungi were removed from the dataset. Because only one (D. hansenii) out of the four applied fungi was detected via the Illumina amplicon data, PCR and gel electrophoresis were performed on DNA extracted from all of the samples to assess potential bias in the primer set and SnapGene was used to assess primer binding (See Supplemental  Methods).

qPCR to determine bacterial and fungal abundance

Extracted DNA from all sample types was used to quantify bacterial and fungal rRNA copy number via qPCR. To determine the appropriate concentration to use, six representative samples were diluted 1 : 10, 1 : 100, and 1 : 1000. These dilutions, along with the stock concentration, were run and the dilution with Cq values within optimal range (20–30, closest to 25) was chosen. For bacteria, the 1 : 100 dilution was used for flowers and the 1 : 1000 dilution was used for guts and provisions. For fungi, the 1 : 10 dilution was used for guts and flowers and the 1 : 1000 dilution was used for provisions. For some fungal samples, the final Cq value was outside the 20–30 range, so they were re-run at an adjusted dilution to try to get them into this range. A bacterial standard curve was generated using a plasmid created from the partial 16S rRNA of Ap. micheneri (LC318485.1), while a fungal standard curve was generated using a plasmid created from the partial 18S rRNA of Moniliella oedocephalis (NG_062174.1).

Samples were randomized across plates using the sample() function in R to help account for plate effects. For each bacterial reaction, 5 μl of SYBR, 3.4 μl of PCR water, 0.3 μl of forward primer 799F (10 μM), and 0.3 μl of the reverse primer 1115R (10 μM) were used. Thermocycler conditions were as follows: 95°C for 3 min, followed by 35 cycles of 95°C for 30 s, 52°C for 30 s, and 72°C for 1 min. To quantify fungal abundance, the FungiQuant system was used (Liu et al. 2012). For each reaction, 5.0 μl PCR Biosystems qPCRBIO Probe Mix, 0.3 μl FQ-F primer (10 μM), 0.3 μl FQ-R primer (10 μM), 0.03 μl of probe, and 3.37 μl PCR water was used. We used the thermocycler conditions described in Liu et al. 2012. Each sample was run in triplicate.

Rearing experiment

To investigate the impacts of supplemented floral microbes on larval health, we performed a larval rearing experiment from April 2024 to May 2025. The experiment included two treatments: Control (no additional microbes applied) and Microbe Supplementation (P. agglomerans, Ac. pollinis, and Ap. kunkeei added). We note that two genera in the inoculum are shared with the floral application study above; P. agglomerans was not applied to hoop houses but can inhabit nectar (e.g. Kong et al. 2021, Cusumano et al. 2023, Mueller et al. 2023) and is of interest as a microbial biocontrol agent (e.g. Stockwell et al. 2002, Kim et al. 2012). To prepare the Microbe Supplementation treatment, each microbe was first cultured on its preferred medium (P. agglomerans and Ac. pollinis on Tryptic-Soy Agar, Ap. kunkeei on De Man–Rogosa–Sharpe agar + 2% Fructose). Next, a bolus of each microbe was added to 2 ml of 30% sucrose individually, diluted, and counted on a hemocytometer. A set of glycerol stocks was created by combining all microbes with sterile 15% glycerol to reach a final concentration of 5000 cells of each microbe per dosage (15 000 total cells applied per bee larva). The control treatment stocks consisted of the same volume of 30% sucrose and 15% glycerol, but without microbe supplementation. Both the microbe supplementation treatment stocks and control treatment stocks were stored at −80°C until use and were not refrozen after usage. To confirm microbe viability, we also created individual glycerol stocks for each microbe to a final concentration of 5000 cells/μl. Next, 20 μl of the individual microbe stocks, the mixed microbe stock, and the control stock were plated (individual stocks: preferred media type as described above, mixed stock and control stock: Tryptic-Soy Agar and De Man–Rogosa–Sharpe agar + 2% Fructose). Growth was observed for all microbial stocks plated, while no growth was documented for the control stocks.

The hoop house set-up described above was prepared as above with two minor differences. First, the hoop houses were seeded with additional P. tanacetifolia seeds in order to ensure adequate availability of flowers. Second, large plastic tote boxes were used to house nesting blocks instead of the fiesta boxes. Hoop houses were stocked with enough adult male and female O. lignaria to ensure consistent nesting. Nesting progress was checked daily, and completed nests were brought to the laboratory. Within three days of collection, nests were dissected, and non-feeding developmental stages (egg or first-instar larva) were randomly assigned one of the two treatments and inoculated. For each bee, sex was predicted based on its position within the nest and the size of the pollen provision. Rearing protocols followed Williams et al. 2025. In brief, larvae were maintained at 25°C and 60% humidity, with survival and development monitored daily until cocoon spinning was complete. In July 2024, all cocoons were weighed and transferred to gelatin capsules (pierced for air circulation) for pupation monitoring via X-ray. Starting in November 2024, cocoons were re-weighed to determine weight loss through pupation and summer dormancy, then moved into winter conditions (4°C, 25% humidity). In April 2025, cocoons were removed from winter conditions, re-weighed to determine overwintering weight loss, and transferred to scintillation vials at room temperature (∼25°C). For three weeks, cocoons were checked daily for adult emergence, and sex was noted (Williams et al. 2025). For bees that did not emerge, the cocoon was dissected to determine the sex of the bee. If the bee did not develop to the adult stage following cocoon spinning (n = 5), the predicted sex was used in subsequent analysis as there was 79% accuracy in correctly predicting sex.

Statistical analysis

General sample description

To compare the alpha diversity of microbes among habitat types (flower, bee gut, pollen provision), we used a linear model with microbial Shannon diversity as a response variable and treatment (pre- or post- spray) and habitat type as predictors, with separate models for bacteria and fungi (“stats” package, R Core Team 2022). Then, pairwise tests using Tukey’s HSD were performed between each of the treatment types. To compare if microbial inoculation modified the communities found in flowers, adult bee guts or pollen provisions, we used permutational multivariate analysis of variance (perMANOVA) with Bray–Curtis distance using the adonis2 function (“vegan” package; Oksanen et al. 2024). Separate models were run for each habitat type, with treatment and spray round, their interaction, and hoop house identity as independent variables.

Aim 1: Are flowers a source of microbes for adult bee guts and pollen provisions?

First, the relative abundance of each focal microbe was determined by dividing the number of reads of each focal microbe by the total number of reads within each sample. Then to determine the effect of inoculation on focal microbe presence (>0 read count) and relative abundance, we used generalized linear models and assessed significance using chi-squared and F-tests, respectively. Separate models were run for each microbe detected using the amplicon data set (Ac. pollinis, Ap. micheneri, D. hansenii, and V. victoriae) and each habitat type (flower, adult bee gut, pollen provisions). For flowers and pollen provisions, hoop house and the interaction between treatment and spray round were included as fixed effects. For gut samples, only treatment and spray round were included as fixed effects, since guts were not collected before and after each spray. The models for chi-squared tests, in this and all subsequent analyses, were run with family = binomial. The degree of multicollinearity was assessed using the vif() function in the “car” package (Fox and Weisberg 2019), but since all values were below two, all predictors were retained in the model.

The generated bacterial and fungal standard curves were used to translate the Cq value to log copy number. For the bacterial data where we used identical primers, we accounted for the percentage of reads identified as mitochondrial, chloroplast, or microbial contamination by combining the qPCR data with the 16S sequencing results. First, the calculated log copy number was translated into copy number by taking the exponential. Then, the copy number was multiplied by the percentage of reads remaining in each sample after filtering. Finally, we took the log of the adjusted copy number to use in statistical analysis.

To determine if treatment affected the copy number of DNA in the samples, we performed linear mixed models (LMMs) separated by habitat type (“lme4” package, Bates et al. 2015). The structure of each model was as follows: log(Adjusted Copy Number) ∼ Treatment + (1|Plate ID). We included plate ID as a random effect in order to account for any plate effects. We did not compare copy numbers across habitat types, as we believe that the amount of each sample used for DNA extractions is not directly comparable. After processing the data, one flower outlier was removed.

Aim 2: Does environmental filtering impact microbial establishment across habitats?

To determine if environmental filtering occurred for each focal microbe among habitat types, chi-square and F-tests were performed on a generalized linear model subsetted to only post-spray samples. In these models, inoculated microbe presence or relative abundance was included as the response variables and habitat type was included as a fixed effect.

Aim 3: Do microbial communities overlap between habitat types?

To determine the overlap between bacterial and fungal communities found among habitat types, the bacterial and fungal datasets were subsetted to the top 15 and top 30 most abundant ASVs for ease of visualization, respectively. Next, heatmaps were used to visualize identity of and number of taxa shared across each habitat type (“Phyloseq” package, McMurdie and Holmes 2013). A taxa was considered present in a habitat type if it had a greater than 0 read count in at least one sample. Finally, we used DESeq2 (Love et al. 2014) to identify genera that were differentially abundant amongst habitat types.

Aim 4: Does microbial supplementation impact larval health?

We aimed to investigate whether microbial inoculation affected survival, development time, weight loss during development, pupation success, and adult emergence. To assess survival, we performed a Cox proportional-hazards model (“survival” package; Therneau and Grambsch 2000, Therneau 2024) with survival status (dead, alive) as the response variable and treatment (control, microbe supplemented) as the predictor variable. To determine the effect on development time, we calculated the number of days it took each larva to progress from their second instar to cocoon completion. We analyzed total development time using a LMM, with treatment and sex (male, female) as fixed effects, and the rearing plate and nest of origin as random effects on development time. To investigate effects on two key developmental stages, we also analyzed time to complete the fifth instar (from defecation to cocoon initiation) and time to spin their cocoon (from cocoon initiation to cocoon completion) using the same model structure as above. To assess weight loss, we calculated (1) total weight lost between cocoon spinning and adult emergence (total weight lost), (2) weight lost during summer dormancy (July–October) (summer dormancy weight loss), and (3) weight lost during overwintering (October to April) (overwintering weight loss). We analyzed each using a LMM with the same fixed and random effects. Pupation success (yes/no) was analyzed using a GLM with a chi-squared test, with treatment as the predictor. To examine emergence, we performed two models: a chi-square test on a GLM with emergence success (emerged, did not emerge) as the response, and a LMM with days to emergence as the response variable.

Results

Microbial diversity and composition vary among habitats and are altered by experimental addition

Alpha diversity differed within bacterial and fungal communities based on habitat type (Bacteria: F = 6.1, df = 2, P < 0.01; Fungi: F = 34.8, df = 2, P < 0.001), with the highest alpha diversity in flowers (Bacteria Mean = 4.18; Fungal Mean = 2.44), followed by provisions (Bacteria Mean = 3.80; Fungal Mean = 1.95), and then bee guts (Bacteria Mean = 3.08; Fungal Mean = 1.65) (Fig. 2a and b). Bacterial community composition was highly variable among samples, with few taxa consistently present across all samples within a habitat type; however, fungal communities were much more consistent. Flowers and bee guts were dominated by Mycosphaerella and Cladosporium, and larval provisions were dominated by those two taxa as well as Alternaria and Podosphaera (Figs. 2c, d, and 3).

Figure 2.

Alpha diversity among habitat types for (a) bacteria and (b) fungi based on amplicon sequencing. Letters signify significant differences between treatments (P < 0.05). PCoA based on habitat type for (c) bacteria and (d) fungi, where light purple circles denote flower samples, blue squares denote bee gut samples, and dark purple squares denote pollen provision samples.

Figure 3.

Relative proportional abundance plots of the (a) top 15 most abundant bacterial ASVs and (b) top 30 most abundant fungal ASVs within each sample. Taxa within the plots are at the genus level, though some taxa were included at the class or family level if there was no consensus. “IN” is the inoculum for the second spray round. The white space above bars represents taxa that were found in the sample but were not included in the most abundant taxa.

Experimental inoculation altered microbial communities but only in specific environments, with significant differences detected in fungal communities within flowers and adult bee guts and bacterial communities within pollen provisions (Tables 2 and 3). Bacterial communities in pollen provisions were also influenced by the spray round, hoop house, and the interaction between spray round and treatment. Fungal communities in flowers and pollen provisions were affected by spray round. Additionally, some groups differed in variance for the fungal communities. This included adult bee guts between treatments (betadisper F = 7.4, df = 1, P = 0.011) and pollen provisions between spray rounds (betadisper F = 5.4, df = 1, P = 0.027).

Table 2.

Permanova examining effects of treatment, spray round and hoophouse on bacterial community composition within each habitat. Results Permanova table based on a Bray–Curtis Distance Matrix for the 16S (bacterial) data.

Habitat type	Effect	DF	Sum of squares	R2	F	Pr(>F)
Flower	Treatment	1	0.42	0.059	1.23	0.094
	Spray round	1	0.43	0.059	1.24	0.062
	Hoop house	1	0.44	0.062	1.29	0.056
	Treatment × spray round	1	0.37	0.052	1.087	0.286
	Residual	16	5.51	0.77
	Total	20	7.18	1.00
Gut	Treatment	1	0.60	0.14	1.46	0.11
	Hoop house	1	0.50	0.11	1.21	0.26
	Residual	8	3.27	0.75
	Total	10	4.36	1.00
Provision	Treatment	1	0.58	0.049	2.10	0.006
	Spray round	1	0.74	0.062	2.67	0.002
	Hoop house	1	0.42	0.036	1.53	0.045
	Treatment × spray round	1	0.50	0.042	1.83	0.017
	Residual	35	9.65	0.81
	Total	39	11.90	1.00

Table 3.

Permanova examining effects of treatment, spray round, and hoophouse on Fungal community composition within each habitat. Results of Permanovas based on a Bray–Curtis Distance Matrix for the ITS (fungal) data.

Habitat type	Effect	DF	Sum of squares	R2	F	Pr(>F)
Flower	Treatment	1	0.28	0.088	2.035	0.043
	Spray round	1	0.39	0.12	2.80	0.012
	Hoop house	1	0.23	0.071	1.65	0.11
	Treatment × spray round	1	0.087	0.027	0.63	0.81
	Residual	16	2.21	0.69
	Total	20	3.19	1.00
Gut	Treatment	1	0.58	0.24	4.40	0.007
	Hoop house	1	0.14	0.055	1.014	0.46
	Residual	13	1.74	0.71
	Total	15	2.47	1.00
Provision	Treatment	1	0.12	0.032	1.58	0.18
	Spray round	1	0.54	0.14	7.00	0.002
	Hoop house	1	0.089	0.023	1.14	0.27
	Treatment × spray round	1	0.048	0.013	0.62	0.57
	Residual	39	3.024	0.79
	Total	43	3.83	1.00

Aim 1: Focal microbes were detected in flowers, adult bee guts, and pollen provisions suggesting that flowers are a source of microbes

Of the six microbes that were sprayed onto flowers, only three were detected using amplicon sequencing, including both bacteria Ac. pollinis, Ap. micheneri, and a single fungal species (D. hansenii). Additionally, the environmental fungus Vishniacozyma victoriae was overwhelmingly detected in the inoculum despite not being purposefully included. We do not know where V. victoriae originated, but because it is present within the environment it may have been present in the spray bottles (Gouka et al. 2022, Rush et al. 2022, Kristjuhan et al. 2024). Regardless, it was included in the analysis because it was applied to the flowers. The yeasts M. reukaufii, Au. pullulans, and S. bombi were not detected in the inoculum or any of the samples, potentially due to primer bias (See Supplemental Methods). This bias could have also been a reason why V. victoriae seemed so abundant within our inoculum. Of the four inoculated microbes that were detected, none were present in any samples prior to the first spray (Fig. 4).

Figure 4.

Relative proportional abundances of focal microbes across the three sampled habitat types (flowers, adult bee guts, and pollen provisions). Ac. pollinis and Ap. micheneri were used in both spray rounds, while D. hansenii was only used in spray round 2. V. victoriae was found in the fungal inoculum.

The presence and relative abundance of Ac. pollinis increased in provisions following flower inoculation, with relative abundance 5.38 times higher following inoculum application (Tables 4 and 5). Its presence differed between spray rounds, increasing in presence after the first round and remaining in flowers before and after the second round. The presence of Ap. micheneri increased in flowers, adult bee guts, and provisions following inoculation, with differences in provisions between spray rounds (Tables 4 and 5). There was also a 3.83-fold increase in relative abundance of Ap. micheneri within provisions, while it was only detected in guts and flowers following inoculation. The presence and relative abundance of V. victoriae within pollen provisions only differed based on spray round and hoop house (Tables 4 and 5); however, there was no effect of treatment. Finally, D. hansenii was not found in any flowers pre- or post-spray but was detected in 1/11 post-spray guts. Additionally, D. hansenii was detected in only 4.5% of pre-spray provisions (1/22) and 18% of post-spray provisions (4/22); however, this difference was not statistically significant.

Table 4.

Chi-square tests for examining differences in presence of Acinetobacter pollinis, Apilactobacillus micheneri, and Vishniacozyma victoriae across habitat types (flowers, adult bee guts, and pollen provisions).

	Acinetobacter pollinis		Apilactobacillus kunkeei		Vishniacozyma victoriae
	LRT	Pr(>Chi)	LRT	Pr(>Chi)	LRT	Pr(>Chi)
Flower
Treatment	3.045	0.0816	11.96	0.00054	2.96	0.085
Spray round	0.39	0.53	0.20	0.66	0.0051	0.94
Hoop house	0.056	0.81	3.68	0.055	0.18	0.66
Treatment × spray round	0.00	1.00	0.00	1.00	0.00	1.00
Gut
Treatment	1.24	0.26	5.062	0.024	2.27	0.13
Hoop house	1.53	0.22	2.093	0.15	3.70	0.054
Pollen provision
Treatment	12.98	0.00032	16.41	<0.0001	0.00	1.00
Spray round	1.84	0.17	1.16	0.28	11.49	<0.001
Hoop house	0.0001	0.99	0.093	0.76	5.66	0.017
Treatment × spray round	5.22	0.022	6.47	0.011	0.00	1.00

Table 5.

Results of F-tests for examining the relative abundance of Acinetobacter pollinis and Apilactobacillus micheneri across the three habitat types (flowers, adult bee guts, and pollen provisions)

	Acinetobacter pollinis		Apilactobacillus kunkeei		Vishniacozyma victoriae
	F-value	Pr(>F)	F-value	Pr(>F)	F-value	Pr(>F)
Flower
Treatment	1.65	0.22	13.23	0.0022	2.66	0.12
Spray round	1.50	0.24	0.049	0.49	0.93	0.35
Hoop house	0.40	0.53	5.077	0.039	2.33	0.15
Treatment × spray round	0.21	0.65	1.22	0.29	0.49	0.49
Gut
Treatment	1.14	0.32	1.058	0.33	1.61	0.23
Hoop house	1.33	0.28	0.0085	0.93	2.92	0.11
Pollen provision
Treatment	5.15	0.030	5.76	0.022	0.0051	0.94
Spray round	0.32	0.57	0.12	0.73	13.96	<0.001
Hoop house	1.85	0.18	0.82	0.37	5.096	0.030
Treatment × spray round	0.93	0.34	1.31	0.26	3.62	0.065

Despite not being detected in the inoculum or any samples in the Illumina sequencing data, M. reukaufii was detected using Metschnikowia-specific primers (See Supplemental Methods). It was found in both inocula (using gel electrophoresis only), 7 flower samples (1 pre-spray, 6 post-spray), 6 bee gut samples (none pre-spray, 6 post-spray), and 14 provision samples (3 pre-spray, 11 post-spray) (Supplemental Fig. 2). Additionally, the presence of M. significantly increased with treatment for both guts (P < 0.05) and provisions (P < 0.05). Metschnikowia presence did not differ across habitat types. We did not detect a significant effect of treatment on microbial abundance measured by the adjusted log copy number for bacteria nor fungi in any of the habitat types (Supplemental Fig. 3).

Aim 2: Inoculated microbes differed in relative abundance across habitats, suggesting environmental filtering occurs

To investigate if environmental filtering occurred, presence and relative abundance of each inoculated microbe was compared across habitats. Both Ac. pollinis and Ap. micheneri differed in relative abundance across habitats (Ac. pollinis: F_2,37 = 3.6, P = 0.04; Ap. micheneri: F_2,37 = 4.0, P = 0.03), attaining the highest relative abundance in pollen provisions. No differences in presence or relative abundance were detected for V. victoriae or D. hansenii.

Aim 3: Bacterial and fungal taxa were shared across habitat types

Of the top 15 most abundant bacterial ASVs, 14 were shared across all three habitat types (Fig. 5a). Specifically, Phaseolibacter was not found within flowers. Of the top 30 most abundant fungal ASVs, 26 were shared across all three habitat types (Fig. 5b). Debaryomyces was not found within flowers, while Buckleyzyma, Aureobasidium, and Hormonema were not found within adult bee guts. All bacteria and fungi present within pollen provisions were found in at least one other habitat type. Additionally, there were no universally present bacteria, but the fungi Mycospherella and Cladosporium were found in every sequenced sample.

Figure 5.

Heatmap of the taxa shared between flowers, adult bee guts, and pollen provisions, subsetted to the (a) top 15 bacterial ASVs and (b) top 30 fungal ASVs. White indicates the taxa was absent, while darker shading indicates the taxa was present in higher relative abundance within the sample. Each tick mark on the x-axis represents a single sample, with all samples grouped based on spray round.

For both bacteria and fungi, multiple genera were found to be differentially abundant (FDR < 0.05) across the habitat types though there were more differentially abundant taxa for fungi (n = 20) than bacteria (n = 4) (Supplemental Figs. 4 and 5). The highest differential abundances were found between provisions and guts for bacteria and in provisions for fungi.

Aim 4: Microbial supplementation had little impact on most bee health metrics

Microbial supplementation of pollen provisions did not significantly affect larval survival (P = 0.38). Although developmental timing differed by sex—females took longer to complete the fifth instar (t = −2.6, df = 120.57, P < 0.01) and complete total development (t = −7.4, df = 113.41, P < 0.001)—treatment did not affect any of the analyzed developmental stages (Fifth Instar: t = 0.27, df = 94.28, P = 0.78; Cocoon Spinning: t = 0.74, df = 93.40, P = 0.46, Total Development: t = −1.17, df = 91.22, P = 0.25) (Fig. 6a). Similarly, weight loss during development varied by sex (Summer Dormancy: t = 12.8, df = 110.2, P < 0.001; Overwintering: t = 6.4, df = 121.4, P < 0.001; Total Weight Loss: t = 11.27, df = 121.67, P < 0.001) (Fig. 6b). Treatment had a small but significant effect on weight loss during summer dormancy with bees in the microbe supplementation treatment losing less weight (t = 2.01, df = 86.04, P < 0.05), but not during overwintering (t = −0.18, df = 83.2, P = 0.86) or overall (t = 0.95, df = 85.13, P = 0.35). Nearly all larvae successfully pupated (131/135), with no effect of treatment on pupation success (P = 0.95). Finally, microbial treatment did not impact emergence success (χ²: P = 0.73), but both sex and treatment influenced emergence timing. Males emerged 1.3 days earlier than females, and bees in the microbe-supplemented treatment emerged almost one day later than controls (Sex: t = −3.81, df = 60.26, P < 0.001; Treatment: t = 2.62, df = 2.64, P < 0.05) (Fig. 6c).

Figure 6.

Microbial effects on larval health metrics based on treatment (control, diet microbe supplementation) and confirmed offspring sex (female, male) (a) Total number of days it took for the offspring to develop from first instar to cocoon spinning. (b) Total weight lost between summer dormancy and adult emergence in October 2024. (c) The number of days it took for larvae to emerge after cocoons were warmed to room temperature.

Discussion

Our controlled hoop house experiment provides clear empirical evidence that flowers are a source of microbial acquisition for bee guts and pollen provisions, but microbial communities are subsequently filtered within pollen provisions and adult bee guts. Consistent with previous studies on the gut microbiota of O. lignaria, we found that the microbial ASVs in adult bee guts and pollen provisions reflect the microbial ASVs present on flowers (Cohen et al. 2020). Adult O. lignaria have low microbial read counts at adult emergence (Crowley and Schaeffer 2024), which suggests that these microbes are not carried over from their larval stage, during which they consume the pollen provision. Together, these results support the growing body of literature that adult O. lignaria acquire their microbiome from environmental sources. Yet, our data show that flower microbes differ in their ability to establish across bee-associated habitats. Unlike social bee species that transfer established microbes between individuals (Kwong and Moran 2016), establishment of microbes in the adult female gut does not seem to be a requirement for establishment within the pollen provision. Following inoculation, we saw that the wild bee-associate Ap. micheneri was present within mother bees and pollen provisions, while the flower-associate Ac. pollinis was only present within pollen provisions.

Although bacterial communities were extremely variable among and within habitats, the fungal communities were remarkably consistent. The presence of the same fungal taxa within and across habitat types could be because these taxa are core to the plant species, because they were dominant in the experimental hoop house environment, or because they are endophytic within the pollen (Hodgson et al. 2014). Notably, a recently published geographic survey of O. lignaria pollen provision across California found high relative abundances of different fungal taxa (Mycosphaerella and Cladosporium) in pollen provisions but strong associations between fungal and pollen composition (Vannette et al. 2025), suggesting fungal acquisition from pollen. Although we did not examine interactions between them, bacteria and fungi co-occur in each habitat, such that they may inhibit, facilitate, or not impact each other’s growth. For instance, the common floral bacterium Asaia astilbes can decrease growth of the yeast M. reukaufii (Rering et al. 2020). As a result, it is possible that the variation of bacterial communities could be due to competitive dominance by the fungi in each habitat. Additional experiments investigating the co-growth of microbes in bee-associated habitats may offer further insight into this quandary.

The decrease in alpha diversity from flowers to provisions to guts suggests that environmental filtering of microbial communities occurs between these habitat types. Abiotic and biotic factors of each habitat may influence which microbes survive and reproduce in each habitat, such that increased strength of filtering could reduce microbial diversity (Nemergut et al. 2013, Mittelbach and Schemske 2015, Moran et al. 2019). Each of the investigated habitats has unique conditions that could differentially influence microbial growth, such as sugar concentration, nitrogen availability, and oxygen content. As a consequence, microbes that dominate the communities in each of these habitats may be specialized for the associated conditions (Herrera et al. 2010, Dhami et al. 2016, Pozo and Jacquemyn 2019, Alvarez-Perez et al. 2021, Christensen et al. 2021), and not all microbes may be equally capable of colonizing each habitat type. Future comparative genomic or transcriptomic studies will be useful in elucidating the mechanisms involved in microbial habitat specialization (e.g. Vuong and Mcfrederick 2019).

Although its introduction was not intentional, the presence of V. victoriae provides an intriguing comparison of a microbe not adapted to this system. Despite its overwhelming presence in the fungal inoculum, V. victoriae was not able to significantly establish in any of the bee-associated habitats following its inoculation. This further supports the hypothesis that the microbes that dominate this system are specially adapted to survive in these novel environments.

Four of the focal microbes were identified across all three habitat types, which supports the hypothesis that flowers are a source of non-pathogenic microbes for bees and pollen provisions (McFrederick et al. 2017, Argueta-Guzmán et al. 2025); however, the unequal presence and abundance among habitat types also suggests environmental filtering. Larger differences in weather conditions across the two spray rounds, timing of sample collection, lack of a control hoop house, or differences in initial microbial communities within each hoop house could have led to the differences we saw in presence of focal microbes across spray rounds and hoop houses. In addition, the low relative abundance of focal microbes could suggest the presence of priority effects, which have previously been found in nectar microbial communities (Toju et al. 2018, Chappell et al. 2022, Noroian, Martin, and Vannette, unpublished data). The samples in this study were not sterile prior to microbial inoculation, as pre-spray samples were colonized by a diverse bacterial and fungal community. The pre-existing community of microbes in each habitat type could have prevented the inoculated microbes from establishing in high numbers, potentially through direct competition for space/resources (e.g. Herrera et al. 2008, Vannette and Fukami 2016, 2018) or via modification of the environment (e.g. acidification, temperature changes, etc.; Herrera and Pozo 2010, Lee et al. 2019). Of the focal microbes, Ap. micheneri was found in the highest number of samples. Previous genomic analysis has revealed that Ap. micheneri is likely adapted to the bee and floral environments (Vuong and Mcfrederick 2019). For instance, Ap. micheneri contains genes associated with pollen digestion and host adherence, which could make it a better competitor (Vuong and Mcfrederick 2019). Despite these adaptations, it still did not reach high abundances in any of the sample types, further suggesting the presence of priority effects.

Though the ITS86F/ITS4R primer set has been successful in previous ITS primer comparisons (Op De Beeck et al. 2014), the results from our fungal sequencing suggest that this primer set was biased against M. reukaufii and was unable to bind to Au. pullulans and S. bombicola (See Supplemental Methods). This finding also highlights the importance of including a positive control or mock community in sequencing submissions in order to maximize accuracy of results (e.g. The Human Microbiome Project Consortium 2012, Kozich et al. 2013, Bokulich et al. 2016, Karstens et al. 2019), as well as a reminder of the detection biases of primer sets (e.g. Bellemain et al. 2010, Tedersoo and Lindahl 2016). It should be noted that primer bias could have led to low abundance of our detected focal yeasts (rather than environmental filtering) and the presence of V. victoriae.

The presence of unique taxa found in both the gut and the provision indicates that flowers are not the only source of microbial acquisition for bees. Rather, adult bees may pick up microbes from other sources including their cocoons, natal nests, soil, water, and other nesting substrate (Keller et al. 2013, Rothman et al. 2019, de Sousa 2021). The environmental conditions of each habitat may contribute to the overlap in ASVs. For instance, flowers and pollen provisions are both exposed to oxygen, while adult bee guts typically have no or low oxygen content (Zheng et al. 2017). Microbes shared between bee guts, flowers, and pollen provisions would thus need to survive in these contrasting environments, which may explain why all ASVs were not consistently present in every habitat type. Instead, it seems more likely that microbes specialize within one of the habitats and use the others as temporary transport between new preferred habitats (Moran et al. 2019, Vannette 2020).

When we supplemented pollen provisions with additional microbes, we found minimal effects on summer dormancy weight loss and emergence timing, but no effect on larval survival, development, and pupation. A lack of initial pollen sterilization could explain why we did not observe the survival benefits previously documented in other studies (Dharampal et al. 2022). Additionally, the inoculated microbes may not have colonized the larval gut, which could have contributed to the null effects we observed. Regardless, this finding provides groundwork for future studies on the effects of P. agglomerans, a microbial biocontrol agent used for controlling the plant pathogen fire blight (Erwinia amylovora) in orchard crops (e.g. Pusey et al. 2011, Kim et al. 2012), on bee health. Understanding the off-target impacts of such agents on beneficial insects is crucial to maintaining sustainable agroecosystems; therefore, future field-grounded studies can build upon our null result to ensure usage of this microbe is safe.

In conclusion, this study provides evidence that adult bees acquire microbes from flowers and vector them to pollen provisions for their offspring, with no detectable effect on larval health. These microbes do not need to be ingested or established in the female gut to reach the pollen provisions, suggesting that females may serve only as microbial vectors for their offspring. Flowers are often considered to be the “dirty doorknobs” of the bee world—it has long been posited and supported that pathogenic microbes are acquired from flowers, and this study finds that commensal microbes are also acquired from this transmission route. Although this study focused on a single flower species and one solitary bee species, we reveal that pollination networks serve as pathways for microbial exchange and highlight variation in microbial establishment across distinct bee-associated habitats. By directly tracking inoculated microbes through unique and complex environments, we observed the direct route of microbial transmission and differential establishment within this tractable model system. This work is broadly applicable in relation to microbial control efforts, since bees have the potential to vector insect and plant pathogens (e.g. Durrer and Schmid-Hempel 1994, Mukhtar et al. 2024), but also suggests that microbial control agents could be incorporated into the pollen provision. The traditional dogma proposed by the Dutch microbiologist Martinus Wilhelm Beijerinck suggests that “everything is everywhere, but the environment selects” (O’Malley 2008); however, this study also highlights the importance of dispersal via vectors in transferring microbes between habitats. Overall, these findings suggest that factors of an environment—including pre-existing microbial communities, climate, and habitat conditions—may lead to differential transmission and establishment among microbial species.

Author contributions

Alexia N. Martin (Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing), Clara Stuligross (Resources, Writing – review & editing), Neal M. Williams (Resources, Supervision, Writing – review & editing), Helen M. Noroian (Investigation, Writing – review & editing), Rachel L Vannette (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing).

Conflicts of interest

The authors have no conflicts of interest to report.

Funding

This work was supported by a National Science Foundation Graduate Research Fellowship and the Phil and Karen C. Drayer Wildlife Health Center Fellowship awarded to ANM, as well as a National Science Foundation grant [DEB-1929516] awarded to RLV.

Data availability

The scripts used to process this data and all data files are available on github at the following URL: https://github.com/lexienichole/Osmia-Hoophouse-2025. The sequencing data used in this study were submitted to the NCBI Sequence Reads Archive (SRA) under the Bioproject ID PRJNA1298349. All files will be made publicly available upon publication.