Effects of ingested essential oils and propolis extracts on honey bee (Hymenoptera: Apidae) health and gut microbiota

Abstract

Managed honey bee (Hymenoptera: Apidae: Apis mellifera Linnaeus) hives require frequent human inputs to maintain colony health and productivity. A variety of plant natural products (PNPs) are delivered via feeding to control diseases and reduce the use of synthetic chemical treatments. However, despite their prevalent use in beekeeping, there is limited information regarding the impact of ingested PNPs on bee health. Here, we tested the effects of different essential oils and propolis extracts on honey bee life span, nutrient assimilation, xenobiotic detoxification, and gut microbiota abundance. Brazilian propolis extract lengthened worker life span, while the other PNPs (Louisiana propolis extract, lemongrass oil, spearmint oil, and thyme oil) exerted variable and dose-dependent effects on life span. Vitellogenin (vg) gene expression was reduced by Brazilian propolis extract at high doses. Expression of CYP6AS1, a detoxification-related gene, was reduced by low doses of thyme oil. The abundances of 8 core gut microbiota taxa were largely unaffected by host consumption of PNPs. Our results suggest that in addition to propolis’s structural and immunomodulatory roles in the colony, it may also exert beneficial health effects when ingested. Thyme oil, a commonly used hive treatment, was toxic at field-realistic dosages, and its use as a feed additive should be viewed with caution until its effects on bee health are more thoroughly investigated. We conclude that the tested propolis extracts, lemongrass oil, and spearmint oil are generally safe for bee consumption, with some apparent health-promoting effects.

Introduction

The biggest current threats to honey bee health in the global beekeeping industry include parasites and pathogens (Genersch 2010, Cornman et al. 2012), malnutrition (Brodschneider and Crailsheim 2010, Branchiccela et al. 2019), and pesticides (Johnson et al. 2010, O’Neal et al. 2018). Many beekeepers apply synthetic chemicals to control pests and diseases. However, several of these treatments, including oxytetracycline (antibiotic), tau-fluvalinate and coumaphos (miticides), and fumagillin (fungicide), have seemingly negative consequences on bee health (Liu 1990, Van Den Heever et al. 2014, Rangel and Tarpy 2015, Ortiz-Alvarado et al. 2020). Increased understanding of the risks of agrochemical use and the emergence of miticide resistance (Rinkevich 2020) have created an urgent need for more sustainable inputs in beekeeping.

Several plant natural products (PNPs) are used by beekeepers as sustainable alternatives to synthetic chemical treatments (Tauber et al. 2019, Boncristiani et al. 2021). Certain essential oils (EOs) have been used for controlling mites (Keskin and Özgör 2018) and fighting microbial infections (Albo et al. 2003, Gende et al. 2009, Boudegga et al. 2010, Costa et al. 2010, Ansari et al. 2017, Tutun et al. 2018, Colin et al. 2019, Hýbl et al. 2021, Pusceddu et al. 2021), albeit with mixed results. Thyme, lemongrass, and spearmint oils are some of the most widely used EOs administered to bee colonies in sugar syrup or patty form (Ricigliano et al. 2022). Thymol, the main constituent of thyme oil, serves as the active ingredient for commercial products aimed at controlling Varroa mites (Gregorc 2005, Gregorc and Planinc 2005, Giacomelli et al. 2015). In vitro assays have shown that thyme oil and its constituents inhibit the growth of several fungal and bacterial honey bee pathogens (Alippi et al. 1996, Kloucek et al. 2012, Wiese et al. 2018, Krutmuang et al. 2022). Lemongrass oil, which is predominantly composed of neral and geronial (Mukarram et al. 2022), and spearmint oil, which contains high amounts of carvone and limonene (Snoussi et al. 2015), are used as feeding stimulants and as bee feed additives. As with thyme oil, some in vitro work demonstrates the inhibition of growth of some honey bee pathogens by both lemongrass oil (Alippi et al. 1996, Kloucek et al. 2012, Krutmuang et al. 2022) and spearmint oil (Ansari et al. 2017). Essential oils show some potential for beekeepers as natural treatments, but little work has been done to explore their full effects on honey bee health.

Propolis is another PNP that confers health benefits to honey bees (Simone-Finstrom and Spivak 2010, Nicodemo et al. 2014, Borba et al. 2015, Simone-Finstrom et al. 2017). Honey bees manufacture propolis by mixing collected plant resins with varying amounts of wax. This mixture is deposited at the nest entrance and the nest interior, surrounding the entire inner surface area in what is called a propolis envelope (Seeley and Morse 1976, Ghisalberti 1979, Simone-Finstrom and Spivak 2010). Propolis is comprised of flavonoids, phenolics, esters, and terpenes, but the exact phytochemical composition depends on the local flora (Bankova et al. 1983, Savka et al. 2015, Wilson et al. 2015, Sun et al. 2019, dos Santos et al. 2021). Ingestion of propolis extracts induces expression of several cytochrome P450 CYP6AS subfamily of genes (Johnson et al. 2012), which are involved in the metabolism of phytochemicals obtained from the bee diet (Mao et al. 2009, 2013, 2015). Quercetin, a polyphenol flavonoid found abundantly in propolis, nectar, and honey, acts as a substrate for the pesticide-detoxifying CYP9Q enzymes as well as the CYP6AS enzymes (Mao et al. 2009, 2011).

In vitro assays and field studies have shown that propolis and its constituents can reduce levels of pathogens, including Paenibacillus larvae (Bastos et al. 2008, Wilson et al. 2015, Borba and Spivak 2017), Ascosphaera apis (Simone-Finstrom and Spivak 2012, Voigt and Rademacher 2015, Wilson et al. 2015), Vairimorpha spp. (formally known as Nosema spp., see Tokarev et al. 2020) (Yemor et al. 2015, Suwannapong et al. 2018, Burnham et al. 2020, Mura et al. 2020, Naree et al. 2021), and some viruses (Drescher et al. 2017, Yosri et al. 2021). As such, it is hypothesized that propolis deposits in the hive may serve as an extension of the bee immune system, allowing bees to save energy by reducing costly immune functions (Simone et al. 2009, Simone-Finstrom and Spivak 2010, Simone-Finstrom et al. 2017, Spivak et al. 2019). There is some evidence that honey bee colonies “self-medicate” with propolis by resin foraging in response to a pathogen challenge, supporting the idea that propolis plays a role in social immunity (Simone et al. 2009, Simone-Finstrom and Spivak 2012).

In a colony setting, a propolis envelope has been shown to result in lower (Simone et al. 2009, Borba et al. 2015) expression of immunity-related genes in 7-day-old worker bees across seasons. Although it is generally thought that honey bees do not naturally consume propolis, feeding propolis extracts may impact bee health. One study showed that bees fed propolis extracts had increased expression of antimicrobial peptides (defensin-1, abaecin, hymenoptaecin, and apidaecin) following a bacterial challenge (Turcatto et al. 2018). This finding suggests that propolis consumption enabled a stronger induction of the bee’s innate immune response (Turcatto et al. 2018). Furthermore, when challenged with Vairimorpha, propolis-fed bees had lower spore levels and improved survival (Suwannapong et al. 2018, Burnham et al. 2020, Mura et al. 2020, Naree et al. 2021). However, the full effects of ingested propolis on honey bee health remain largely understudied.

Here, we conducted a laboratory study to investigate the effects of ingested PNPs (3 EOs—lemongrass, spearmint, and thyme—and 2 propolis extracts from Louisiana and Brazil) on worker bee health. We tested the impacts of ingested PNPs on the gut microbiota, which plays a vital role in host immunity (Kwong et al. 2017), pathogen defense (Steele et al. 2021), metabolism (Zheng et al. 2017), detoxification (Wu et al. 2020), and can be affected by ingested phytochemicals (Geldert et al. 2021). Prior work suggests that exposure to propolis can stabilize the honey bee microbiome (Dalenberg et al. 2020, Saelao et al. 2020), but little research has been done on how ingestion might influence the gut microbiota. Here, we examined the impacts of ingested PNPs on (i) bee lifespan, (ii) bodyweight, (iii) absolute quantities of 8 core gut microbiota taxa, and (iv) transcript expression of 6 genes associated with detoxification, immunity, and nutritional status, and overall health. Overall, this study aimed to provide a more holistic understanding of how PNP feed additives impact honey bee health.

Methods

Honey Bees

Frames of emerging brood were collected from six honey bee colonies managed by the USDA Honey Bee Breeding, Genetics and Physiology Lab in Baton Rouge, LA, in September 2021 and left overnight in an incubator at 34 °C and 50% relative humidity (RH). Prior to the experiment, all six colonies had been treated with Apivar® (amitraz) strips to control Varroa destructor mites.

Propolis Extracts

The ethanolic propolis extracts were made with fresh propolis scraped from a Baton Rouge colony (“Louisiana propolis”) and from Brazilian green propolis, mainly derived from Baccharis dracunculifolia (family Asteraceae) and recovered from storage at −20 °C (“Brazilian propolis”) (Salatino et al. 2005). The extracts were prepared based on modified methods outlined in Bankova et al. (2019). Propolis was ground to a powder and covered with 70% ethanol in a glass jar and left to soak for at least two weeks in the dark, allowing the active compounds to solubilize in the ethanol. Extracts were filtered to remove particulates and stored at room-temperature and left unexposed to light until further use.

Experimental Design

The experiment followed established methods of a typical cage feeding trial (Ricigliano et al. 2021). Approximately 50 newly emerged adult bees (<24 hours old) were sorted into steel cages with pollen paste (50% sucrose syrup w/v mixed with bee-collected pollen) and a syringe of treatment syrup. To make the syrups, the treatment was mixed into 50% sucrose syrup (w/v) with 70% ethanol (776 µl per 50 ml) and soy lecithin (0.21g per 50 ml syrup) as emulsifiers. The treatments included ethanolic Louisiana (LP) and Brazilian (BP) propolis extracts, lemongrass oil (LM, Cymbopogon flexuosus, www.revive-eo.com), spearmint oil (SP, Mentha spicata, Sigma–Aldrich), and thyme oil (TY, Thymus vulgaris, www.revive-eo.com). A sugar syrup containing just the emulsifiers was used as a treatment control (CTL). Currently, there is little standardization among beekeepers regarding the dose of EOs that may be used as feed additives for bees. Therefore, the PNP treatments in this study were administered at a “high” or “low” concentration: 3,000 or 100 µg/ml, respectively, for both the propolis extracts and EOs. These concentrations were chosen to represent doses higher and lower than the recommended dose of one commercially available EO-based feed additive: 1,000 µg EOs per 1 ml sugar syrup. Two doses were tested to see how, if at all, the effects of the PNPs varied by dose. Our low dose of EOs is approximately equivalent to the concentration used in Pătruică et al. (2023). Three replicate cages were assigned to each dose of each treatment, as well as three cages for the control group.

All cages were stored in the same incubator in the dark at 30 °C with a tub of water below to maintain ~50% RH. Mortality was recorded daily, and any dead bees were removed. All cages were provided with their respective treatment syrup and pollen paste ad libitum for the first seven days of the trial to ensure complete development of gut microbiota (Powell et al. 2014), with the consumption of both syrup and pollen measured through day 7. Given the nature of our experimental design, we measured the consumption of pollen and treatment syrup for groups of bees in each cage instead of for each bee sampled. On day 7, six bees were sampled from each cage over dry ice. The heads and thoraces of the sampled bees were dried following standard methods and weighed to measure nutrient assimilation (Brodschneider and Crailsheim 2010, Human et al. 2013) and the abdomens were stored at −80 °C for subsequent molecular analyses. All cages were switched to a diet of only 50% sucrose syrup for the remainder of the trial while mortality was monitored.

Gut Microbiota and Gene Expression

Gut microbiota analyses of bees fed low- and high-dose PNP treatments focused on 8 prominent bacterial taxa that comprise the honey bee gut microbiota: Snodgrassella alvi (Gram negative, class Betaproteobacteria), Gilliamella apicola (Gram negative, class Gammaproteobacteria), Lactobacillus Firm-4 (Gram positive, class Bacilli), Lactobacillus Firm-5 (Gram positive, class Bacilli), Bifidobacterium asteroides (Gram positive, class Actinomycetia), Frischella perrara (Gram negative, class Gammaproteobacteria), Bartonella apis (Gram negative, class Alphaproteobacteria), and Alpha 2.1 (Gram negative, class Alphaproteobacteria) (Martinson et al. 2011, Moran et al. 2012, Kwong et al. 2014). qPCR (Bio-Rad SsoAdvanced Universal SYBR Green Supermix) was performed on DNA extracted from the abdomens (Qiagen All-Prep DNA/RNA Mini Kit) targeting the 16S region of each of the bacterial taxa (primer sequences in Supplementary Table S1). All DNA samples were diluted to 15.0 ng ± 1.0 ng/μl prior to qPCR. Plasmid standard curves were included to achieve absolute quantification of each bacterial target. The number of gene copies detected in each sample for each bacterial target was normalized via a natural log (ln) transformation prior to statistical analyses.

RT-qPCR was also carried out using cDNA synthesized (LunaScript RT SuperMix Kit) from the RNA extracted (Qiagen All-Prep DNA/RNA Mini Kit) from the above abdomens (n = 6 per cage, 18 per treatment group) to analyze the expression of various genes related to detoxification and nutrition. The targeted detoxification genes included CYP6AS14, CYP6AS1, CYP9Q3, and CYP306A1 (primer sequences listed in Supplementary Table S1). Each of these genes has been shown to be differentially regulated in response to PNPs such as thymol and components of propolis (Boncristiani et al. 2012, Johnson et al. 2012, Mao et al. 2013). Other genes targeted in this experiment included vitellogenin (vg), which acts as a general health status indicator, and abaecin, an immune gene that encodes an antimicrobial peptide. Following established methods from similar studies, rp49 served as the housekeeping gene for the analyses (Lourenço et al. 2008, Moon et al. 2018, Hyang Jeon et al. 2020), and it was ensured that the expression of this gene was consistent across all groups. The relative expressions for each gene (calculated as 2^−ΔΔCt using the mean of the controls as the calibrator) were normalized via log₁₀ transformation prior to statistical analysis.

Statistical Analysis

All statistical analyses were conducted in R Studio version 4.3.0 (R Core Team 2023). All variables measured, except for lifespan, were analyzed individually via linear mixed models with treatment, dose, and treatment–dose interaction as fixed effects and cage as a random effect. Marginal means were estimated with dose nested within treatment. Pairwise differences were assessed via post hoc Tukey HSD tests performed on the estimated marginal means of each group. These analyses were performed using the lme4, emmeans, and multcomp packages in R (Hothorn et al. 2008, Bates et al. 2015, Lenth 2023).

Survival Analysis

The bees sampled on day 7 (n = 6 bees per cage) were censored from the survival analyses. The median and mean worker lifespans and standard error (SE) were calculated within each treatment–dose group across all 3 replicate cages. The Kaplan–Meier survival curves were constructed using the survival package and then visualized using the survminer package (Kassambara et al. 2021). These curves were compared using a Log-Rank test followed by specific pairwise log-rank comparisons, also via survminer (Bland and Altman 2004, Kassambara et al. 2021). The resulting P-values of the pairwise comparisons were adjusted using the Benjamini–Hochberg method.

Holistic Analyses

To see how all output variables correlated with each other across all bees sampled, we constructed a Pearson correlation matrix using the R package corrgram (Wright 2021). Pearson correlations and associated P-values were obtained using the Hmisc package (Harrell 2023).

To model how the bees from each treatment and dose group (n = 18 bees per group) could be generally grouped based on their core gut microbiota and gene expression profiles, we performed linear discriminant analysis (LDA). We created 4 separate models to visualize the groupings of the gut microbiota and gene expression profiles of the sampled bees, separated by low and high dose, with the controls included in all models. Separating by dose allowed us to better visualize any potential treatment group clustering within each dose. The LDA models were created using the MASS package, and the plots were generated using the ggord package (Venables and Ripley 2002, Beck 2022).

Results

Life Span and Survival

We compared the median lifespans and survival curves of all treatment–dose groups against the control group (CTL). Overall, there were significant differences among the survival curves of all treatment–dose groups (log-rank test, χ² = 615, df = 10, P < 2e-16). Propolis extract treatments generally improved survival and increased average life span. Compared to control bees, which had a median life span of 16 days (Table 1), bees fed the low doses of both Brazilian and Louisiana propolis extracts (BP-L and LP-L) lived 2 and 3 days longer, respectively (Table 1), and had significantly better survival (BP-L: Pairwise Log-Rank Test, P = 0.0007; LP-L: P = 5e-05; Fig. 1). Bees fed the high dose of Brazilian propolis extract (BP-H) also had improved survival (P = 0.0053), living 1 day longer than control bees on average (Table 1), but LP-H bees had no difference in survival or median life span relative to controls (Fig. 1).

Table 1.

The median and mean honey bee worker life spans varied among treatment and dose groups. Average life spans were calculated for bees from all 3 replicate cages of a treatment–dose group, excluding the 6 bees sampled from each cage for analysis

Treatment	Dose	Median life span (days)	Mean life span (days)	Standard error (SE)	n
Control (CTL)	—	16	15.65	0.50	104
Brazilian propolis (BP)	Low	18	18.06	0.72	95
	High	17	17.28	0.73	88
Louisiana propolis (LP)	Low	19	19.30	0.71	85
	High	16	15.81	0.52	93
Lemongrass oil (LM)	Low	17	15.81	0.47	96
	High	21	20.27	0.86	90
Spearmint oil (SP)	Low	18.5	18.71	0.78	98
	High	14	14.74	0.45	118
Thyme oil (TY)	Low	9	8.57	0.23	114
	High	14	15.74	0.77	104

Fig. 1.

Honey bee survival was affected by some of the PNP treatments administered, shown here in the Kaplan–Meier survival curves. The treatments administered were Brazilian propolis extract (BP), Louisiana propolis extract (LP), lemongrass oil (LM), spearmint oil (SP), thyme oil (TY), or control (CTL). All treatments, except for CTL, were administered at either a low (L) or high (H) dose. Lifespans were recorded from all bees from all 3 replicate cages of each group, with the 6 bees sampled from each cage censored from the analysis. Significant differences among these survival curves were detected via a pairwise Log-Rank test (α = 0.05). Survival was significantly increased in the BP-L (P = 0.0007), BP-H (P = 0.0053), LP-L (P = 5e-05), SP-L (P = 0.0001), and LM-H (P = 3e-08) groups and decreased in the SP-H (P = 0.0468) and TY-L (P < 2e-16) groups.

Effects of the three EOs were highly variable and different between the two doses. Of all the treatment groups, bees fed the low dose of thyme oil (TY-L) had the shortest lifespans (median = 9 days; Table 1) and a much worse survival curve compared to controls (P < 2e-16; Fig. 1). Bees in the TY-H group had a slightly shorter median lifespan (14 days) than the controls (16 days), but there was no significant difference in their survival curve relative to the controls (Fig. 1). The low dose of spearmint oil (SP-L) (median lifespan = 18.5 days) improved survival (P = 0.0001), while SP-H (median lifespan = 14 days; Table 1) worsened survival (P = 0.0468; Fig. 1). While the low dose of lemongrass oil (LM-L) showed little impact on median lifespan (17 days) or survival, LM-H yielded the highest median lifespan (21 days; Table 1) and greatly improved survival relative to controls (P = 3e-08; Fig. 1).

Head and Thorax Weight

The overall mean head and thorax weight across all sampled bees was 11.40 ± 0.07 (SE) mg. No significant differences in weight were observed among the treatment–dose groups (n = 18 bees per group, P > 0.05, Supplementary Fig. S1).

Pollen and Syrup Consumption

Across all treatment–dose groups (n = 3 cages per group), each cage of bees consumed a mean of 2.14 ± 0.07 (SE) g of pollen paste and 8.23 ± 0.24 (SE) g of treatment syrup during the 7 days of the trial. There were no significant differences in consumption of either pollen or syrup detected among any of the groups (P > 0.05;Fig. S2).

Gut Microbiota

For the 8 core gut microbiota taxa measured in this study, our models detected no significant differences (α = 0.05) in bacterial abundance among the treatment–dose groups. However, some several near-significant trends were observed. Bees fed the SP-L syrup trended toward higher quantities of Snodgrassella alvi (21.9, 95% CI [21, 22.8]) relative to controls (19.9, 95% CI [19.0, 20.8], P = 0.1050; Fig. 2A), but lower quantities of Gilliamella apicola (18.8, 95% CI [17.9, 19.8], P = 0.1001; Fig. 2B). TY-H bees trended toward higher quantities of Lactobacillus Firm-5 (20.2, 95% CI [19.9, 20.5]) compared with controls (19.6, 95% CI [19.3, 19.9], P = 0.1445; Fig. 2D). TY-H bees also trended toward lower amounts of Alpha 2.1 (16.1, 95% CI [14.4, 17.8]) relative to controls (19.8, 95% CI [18.1, 21.5], P = 0.0994; Fig. 2E).

Fig. 2.

The effects of the PNP treatments on 8 members of the core gut microbiota: (A) Snodgrassella alvi, (B) Gilliamella apicola, (C) Lactobacillus Firm-4, (D) Lactobacillus Firm-5, (E) Alpha 2.1, (F) Bifidobacterium asteroides, (G) Bartonella apis, and (H) Frischella perrara. The treatments administered were Brazilian propolis extract (BP), Louisiana propolis extract (LP), lemongrass oil (LM), spearmint oil (SP), thyme oil (TY), or control (CTL). All treatments, except for CTL, were administered at either a low (L) or high (H) dose. The orange triangles indicate the group mean natural log (ln)-transformed bacterial abundance (n = 18 bees per treatment–dose group), and the horizontal black line within each box represents the group median. The boxes encompass the middle 50% of the group values and the whiskers delineate the top and bottom 25% of values, with outliers indicated as black dots. The letters above each box indicate significant differences (α = 0.05) detected by a Tukey HSD test performed on the estimated marginal means of the groups.

Neither the Brazilian nor Louisiana propolis extracts resulted significant differences in the abundances of any bacteria targeted at either the low or the high dose (P > 0.05; Fig. 2).

Gene Expression

Expression of abaecin was stable across all treatment groups with no significant differences detected among any of the treatment–dose groups (P > 0.05; Fig. 3A). Expression of vg was similar among all treatment groups, except for BP-H, in which it was significantly reduced (−2.34, 95% CI [−2.98, −1.71], P = 0.0009; Fig. 3B).

Fig. 3.

The effects of the PNP treatments on the relative expression (2^−ΔΔCt, log-transformed value) of genes (A) abaecin, an antimicrobial peptide, (B) vg, vitellogenin, and four detoxification-related cytochrome P450 (CYP) genes: (C) CYP6AS1, (D) CYP6AS14, (E) CYP306A1, and (F) CYP9Q3, shown in the above box-and-whisker plots. The treatments administered were Brazilian propolis extract (BP), Louisiana propolis extract (LP), lemongrass oil (LM), spearmint oil (SP), thyme oil (TY), or control (CTL). All treatments, except for CTL, were administered at either a low (L) or high (H) dose. The orange triangles indicate the group mean (n = 18 bees per treatment–dose group), and the horizontal black line within each box represents the group median. The boxes encompass the middle 50% of the group values and the whiskers delineate the top and bottom 25% of values, with outliers indicated as black dots. The letters above each box indicate significant differences (α = 0.05) detected by a Tukey HSD test performed on the estimated marginal means of the groups.

Only one of the four targeted CYP genes was significantly (P < 0.05) reduced in response to PNP treatment. TY-L bees had lower expression of CYP6AS1 (−0.70, 95% CI [−0.96, −0.44]) than controls (0.0044, 95% CI [−0.26, 0.26], P = 0.0249; Fig. 3C). Additionally, CYP9Q3 expression trended lower in LP-H bees (−0.60, 95% CI [−0.84, −0.35]) compared with controls (−0.03, 95% CI [−0.27, 0.21], P = 0.0658; Fig. 3F). No significant differences were observed in the expression of either CYP6AS14 or CYP306A1 among any of the treatment groups relative to the control group (P > 0.05; Fig. 3D and E).

Holistic Analyses

Correlation matrix

We generated a Pearson’s correlation matrix to calculate the pairwise relationships among all 18 tested variables for all bees sampled. Median lifespan had moderate to weak positive correlations with pollen consumption (R = 0.39, P < 0.0001), CYP6AS1 expression (R = 0.21, P = 0.0032) and CYP9Q3 expression (R = 0.20, P = 0.0038), and weak negative correlations with syrup consumption (R = −0.20, P =0.0051), G. apicola abundance (R = −0.24, P = 0.0007), and B. apis abundance (R = −0.18, P = 0.0112) (Figure 4). Expression of vg correlated positively with pollen consumption (R = 0.39, P < 0.0001), negatively with syrup consumption (R = −0.27, P = 0.0001) and positively with Lactobacillus Firm-4 abundance (R = 0.30, P < 0.0001), but there was no correlation detected between individual vg expression levels of the bees sampled and cage-wide median lifespan (Fig. 4). There were no significant correlations between abaecin expression and any other variable measured (Fig. 4).

Fig. 4.

A Pearson’s correlation matrix of all 18 output variables among all individual bees (n = 192 complete observations) sampled across all treatment and dose groups. A dark red cell indicates a stronger negative correlation between the 2 intersecting variables and a dark blue cell indicates a stronger positive correlation, while the lighter cells indicate weak or no correlations. The R value is displayed in each cell with asterisks indicating statistical significance (*P < 0.05, **P < 0.001, ***P < 0.0001).

Linear discriminant analysis

Linear discriminant analysis (LDA) was performed to obtain a broader picture of how each PNP tested may have influenced gut microbiota and gene expression.

Within both models (low- and high-dose treatment groups), all groups appeared to have similar microbiota compositions regardless of PNP treatment, as indicated by the overlapping 95% confidence interval ellipses (Fig. 5A and C). For the low-dose gut microbiota model, the LD1 and LD2 axes achieved 47.62% and 27.89% of the separation among individual samples, respectively, with the abundances of Lactobacillus Firm-5 and Firm-4 contributing strongly to that separation (Fig. 5A). For the high-dose gut microbiota model, the LD1 and LD2 axes were associated with 51.73% and 23.23% of the separation among samples, respectively, with none of the 8 measured bacteria appearing to contribute more than the others to the separation (Fig. 5C).

Fig. 5.

LDA plots show clustering of sampled bees from each treatment group by their core gut microbiota (A—low dose, C—high dose) and gene expression (B—low dose, D—high dose). The treatments administered were Brazilian propolis extract (BP), Louisiana propolis extract (LP), lemongrass oil (LM), spearmint oil (SP), thyme oil (TY), or control (CTL). Each point represents one bee (n = 18 bees per treatment and dose group; 108 complete observations per model). The colored ellipses represent 95% confidence intervals for samples within each treatment group. The arrows represent each variable contributing to the model. The length of the arrow indicates the strength of its contribution to the model and the direction indicates along which axis or axes the arrow contributes to the separation of the samples.

Gene expression models for both the low- and high-dose treatment groups were similar to the gut microbiota models in that treatment groups did not separate from one another, indicating overall similarity in gene expression profiles of the genes targeted in this study (Fig. 5B and D). For the low-dose gene expression model, the LD1 and LD2 axes were associated with 62.41% and 24.21% of the separation among samples, respectively (Fig. 5B). Expression of CYP6AS1 appeared to contribute most to separation of samples along the LD1 axes, while the expression of CYP6AS14, CYP306A1, and CYP9Q3 appeared to contribute strongly to separation of the samples along the LD2 axis (Fig. 5B). The high-dose gene expression model showed slightly more separation among the treatment groups compared to the low-dose gene expression model. For the high-dose model, the LD1 and LD2 axes achieved 63.34% and 28.05% of the separation among the samples, respectively (Fig. 5D). The expression of all four CYP genes (CYP6AS1, CYP6AS14, CYP306A1, and CYP9Q3) and vg appeared to contribute strongly to separation among samples along both the LD1 and LD2 axes (Fig. 5D). The 95% confidence interval ellipses for LP-H and BP-H appeared to clearly separate away from LM-H and SP-H, with the LM-H and SP-H ellipses almost completely overlapping with each other and the LP-H and BP-H ellipses only narrowly overlapping with each other (Fig. 5D). This suggests that of the genes targeted in this analysis, the overall expression profiles were very similar among the LM-H- and SP-H-treated bees, and although the expression profiles were not very similar among the two propolis groups (LP-H and BP-H), their profiles were both different from those of LM-H and SP-H (Fig. 5D). The 95% confidence interval ellipse for TY-H did not overlap with the other two EO groups (LM-H and SP-H), but rather with LP-H, indicating similarity between those two groups (Fig. 5D). The control (CTL) ellipse overlapped at least some with all the PNP ellipses, suggesting that none of the PNP treatments, even at high doses, affected the bees enough to completely change their gene expression profiles compared to the controls (Fig. 5D).

Discussion

This study investigated the effects of ingested PNPs on various aspects of honey bee health. Despite their prevalent use in beekeeping, few studies have explored the effects of ingested EOs on bee health (Conrad 2010, Pătruică et al. 2023). Although research on the effects of propolis and its potential to improve bee health is growing (Simone-Finstrom et al. 2017, Pusceddu et al. 2021), more research is necessary to understand its influence on bee health and gut microbiota, which can be considered an extension of bee physiology. This is particularly relevant since there has been increased research focus aimed at propolis use as a feeding treatment (Simone-Finstrom et al. 2017, Suwannapong et al. 2018, Turcatto et al. 2018, Burnham et al. 2020, Mura et al. 2020, Naree et al. 2021). We hypothesized that ingested EOs and propolis extracts would modulate worker life spans and nutrition. Furthermore, we aimed to test how gut microbiota and gene expression are impacted by the ingestion of these PNPs. Given that honey bee-associated microbes have evolved in the presence of a propolis envelope, and that these microbial communities appear to be stabilized by exposure to propolis in the nest environment (Dalenberg et al. 2020, Saelao et al. 2020), we hypothesized that the tested extracts would not impact honey bee gut microbiota.

We found that propolis extracts, lemongrass, and spearmint oils have potential benefits as dietary supplements due to their generally positive effects on lifespan at the doses tested. In a colony setting, longer worker lifespans are beneficial because that would increase the resiliency of colonies by reducing impacts of perturbations in worker age dynamics (i.e., due to acute pesticide exposure of foragers) and preventing colony population decline from workers transitioning from in-hive to foraging related tasks too early (Khoury et al. 2011).

The only treatment that significantly shortened worker lifespan was the low concentration of thyme oil, suggesting the oral toxicity of this EO. However, it was unclear why the high dose of thyme oil did not have an equally strong or stronger negative effect on lifespan. Thyme oil’s acute oral toxicity against honey bees has been demonstrated in other studies (Albo et al. 2003, da Silva et al. 2020), and reduced survival has also been observed in other thymol-fed bees in cage studies (Palmer-Young et al. 2017, Canché-Collí et al. 2021). This toxicity may be due to thymol’s ability to inhibit the enzyme acetylcholinesterase (Jukic et al. 2007), which is the same mechanism attributed to some major classes of pesticides (e.g., organophosphates and carbamates). In its current most widely recognized role as an active ingredient in a miticide product (Gregorc and Planinc 2005), thymol is not intended to be consumed by bees. Rather, as the bees perform their cleaning behaviors, they purportedly transport the thymol-containing product throughout the hive and the product’s vapors kill mites. However, results from our study and others may warrant a thorough investigation into whether bees may in fact be consuming thymol-based miticides and how that may affect their health.

Our results suggest that the tested PNPs have no effect on nutrient assimilation based on head and thorax weights and gene expression of the nutritional storage protein vitellogenin. The only potential negative effect of this nature was observed in the BP-H group, which had reduced vg expression, but few other differences in health measures as compared to controls. All cages were provided with an identical source of protein and lipids (bee-collected pollen) ad libitum. There were no differences in pollen consumption or bodyweights among any of the treatment groups. Therefore, the ingested PNPs did not negatively impact bee nutrient assimilation. Furthermore, the PNP treatments did not increase food consumption, despite the prevalent use of lemongrass and spearmint oils as feeding stimulants. Among beekeepers, lemongrass oil is thought to be useful in attracting swarms to hives, and potentially acting as a feeding stimulant, because its two main chemical constituents, geranial (E-citral) and neral (Z-citral) (Gao et al. 2020), are also found abundantly in the honey bee Nasonov pheromone, an attractant pheromone (Pickett et al. 1981). However, our pollen and syrup consumption measurements suggest no difference in appetite between the lemongrass oil-fed bees and the controls. Although the number of replicate cages in each group (n = 3) limited the power of our statistical test, these PNPs did not exert biologically meaningful effects on consumption in our study.

Our analyses of bee gut microbiota revealed that core bacteria are resilient to the tested EOs and propolis extracts, despite these PNPs possessing known antimicrobial activities. None of the bacterial taxa included in our analyses varied in abundance among the treatment groups at either the low or high dose. Our LDA models also indicated overall similarity in the core gut microbiota among all treatment groups and controls. Several trends were observed, however, which we interpret cautiously. Most near-significant differences resulted from either thyme or spearmint oil, suggesting that these EOs may disrupt the gut microbiota more so than lemongrass oil or ethanolic propolis extracts. One of the targeted bacterial species, Bartonella apis, has been shown to be reduced in bees exposed to a rich propolis envelope (Saelao et al. 2020); however, we saw no statistically significant reduction of these bacteria in our test. Another bacterial species targeted in our study was Frischella perrara, which was not affected by any of the PNP treatments. Frischella perrara dominates the pylorus (the region of the gut between the midgut and the start of the hindgut) (Engel et al. 2015) and has been shown to activate the honey bee immune system (Emery et al. 2017).

At field-realistic doses, ingestion of spearmint oil resulted in a slightly increased abundance of S. alvi and reduced abundance of G. apicola. It is unclear whether the observed shifts in abundance could have health consequences. Together, these twogram negative bacteria typically dominate the anterior segment of the hindgut - the ileum - in a biofilm (Martinson et al. 2012, Kwong and Moran 2013). These two gut symbionts occupy different, yet complementary, metabolic niches (Kwong et al. 2014, Anderson and Ricigliano 2017, Zhang et al. 2022, Kešnerová et al. 2017). S. alvi forms a biofilm along the epithelial layer of the gut, on top of which G. apicola colonizes (Martinson et al. 2012). S. alvi metabolizes the products of fermentation from other gut bacteria (Kwong and Moran 2016). This bacterium has also been shown to stimulate its host’s immune system, allowing the bee to better defend itself against invading pathogens (Horak et al. 2020). G. apicola has been shown to metabolize sugars that are harmful to its host (Zheng et al. 2016). Other studies have associated increases in quantities of G. apicola with gut dysbiosis (Kakumanu et al. 2016, Maes et al. 2016, Anderson and Ricigliano 2017), so lower abundance of G. apicola may not present a cause for concern. Despite these potential shifts in these prominent gut symbionts, bees fed the low dose of spearmint oil lived longer than controls, indicating no negative effect. This is supported by our correlation matrix, which revealed a general negative correlation between G. apicola abundance and lifespan across all bees sampled.

Our LDA models suggest that the low dose of the PNP treatments resulted in little change to the overall gene expression profile, while the high doses of the treatments may have affected the lemongrass oil and spearmint oil-treated bees differently than the propolis and thyme oil-treated bees differently. Regarding the expression of the 4 CYP genes targeted in our analysis (CYP6AS14, CYP6AS1, CYP9Q3, and CYP306A1), none of the PNPs appeared to significantly modulate the expression of any of these detoxification genes. In fact, the expression of these genes trended lower in many of the PNP treatment groups. Given the increased mortality in the thyme oil-fed bees, which suggested toxicity, we initially expected to see increased expression of the detoxification-related CYP genes in response to thyme oil ingestion. However, our thyme-oil-fed bees had lower expression of CYP6AS1, which aligns with findings from another study that saw reduced CYP gene expression in response to thymol consumption (Boncristiani et al. 2012). Our findings suggest that most of the PNP treatments administered in our study yielded no apparent negative health effects, with the potential exceptions being thyme oil, which shortened median lifespan at both doses, and the high dose of Brazilian propolis extract, which reduced vg expression.

We found the expression of abaecin consistent across most groups, but it trended higher in bees fed the low dose of Brazilian propolis extract. Generally, propolis has been found to either have no effect on immune response when ingested (Turcatto et al. 2018), or it may decrease immune gene expression when bees are exposed to propolis in the nest environment (Simone et al. 2009, Borba et al. 2015). However, ingestion of propolis has been associated with an increase in antimicrobial peptide expression of bees after challenges with bacterial immune elicitors (Simone-Finstrom et al. 2017, Turcatto et al. 2018). These findings and their implications for colony health need to be validated in future work.

While our findings revealed some promising health benefits of ingested PNPs on bees, the effects we saw were dose dependent. For example, the low dose of Louisiana propolis extract significantly increased worker life span, while the high dose had no effect. The reasons for this phenomenon were unclear, but additional investigations would help clarify what concentration(s) may reliably increase worker lifespan. Additionally, our study only tested the effects of these PNPs on newly emerged workers in a highly controlled lab environment. Little work has been done to investigate the effects of ingested PNPs on overall colony health, but one recent study reported an increased amount of brood and higher honey yields in colonies provided with sugar syrup containing EOs (Pătruică et al. 2023). More field studies are needed to better understand how PNPs affect colony health.

Switching from traditionally used synthetic chemicals to natural products may present an economic incentive for beekeepers who seek to market their honey and other bee products as organic. The global market for organic honey was valued at $605 million USD in 2020, and its projected compound annual growth rate is 5.5% from 2021 to 2030 (Pathak et al. 2022). Although the USDA standards for certified organic honey are not as clearly outlined as those from Canada or the European Union, the US National Organic Standards Board voted in favor of an Organic Apiculture Recommendation in 2010 stating that beekeepers must not use any antibiotics or synthetic miticides to treat their honey bees to sell organic honey (Giacomini 2010). Our findings here suggest that, when administered at field-relevant doses, lemongrass and spearmint EOs and propolis extracts used as feed additives may benefit bee health by increasing lifespan without negative impacts on gut microbiota or expression of several health-related genes. Further work is needed to better understand the full potential of these PNPs as hive treatments and their ability to support colony health and disease resistance.

Author Contributions

Allyson Martin Ewert (Conceptualization [Equal], Data curation [Equal], Formal analysis [Equal], Investigation [Equal], Visualization [Equal], Writing – original draft [Equal]), Michael Simone-Finstrom (Conceptualization [Equal], Funding acquisition [Equal], Methodology [Equal], Project administration [Equal], Resources [Equal], Writing – review & editing [Equal]), Quentin Read (Formal analysis [Equal], Methodology [Equal], Writing – review & editing [Equal]), Claudia Husseneder (Writing – review & editing [Equal]), and Vincent Ricigliano (Conceptualization [Equal], Funding acquisition [Equal], Methodology [Equal], Project administration [Equal], Resources [Equal], Writing – review & editing [Equal])