Floral diversity in tropical landscapes affects fatty acid composition in bee bread

Honeybee diets, particularly fatty acids in tropical forests, are not well known. This study explored fatty acid concentration and diversity, measured as methyl ester derivatives (FAMEs), in bee bread from tropical landscapes with and without forest cover. Bee bread samples from 24 hives in Papua New Guinea were analysed using GCMS and DNA metabarcoding. Fatty acids consisted of 43% saturated and 57% unsaturated lipids. Myristic, palmitoleic, and stearic acids were higher in forest landscapes, providing antimicrobial properties and important compounds for wax production. Lauric acid was higher in landscapes without forest, which may benefit honeybee nutrition in disease resistance. FAME concentrations were high in omega-6 and omega-9 but low in omega-3, and total FAME diversity increased with floral diversity. This study highlights the importance of diverse pollen sources to maintain a complete fatty acid profile for honeybee health.

European honeybees (Apis mellifera) are a polylectic bee species whose diet consists entirely of nectar and pollen collected from floral sources within the surrounding landscape (Westrich, 1996). Nectar provides carbohydrates, minerals, and vitamins essential for energy metabolism, while pollen supplies proteins in the form of amino acids and lipids in the form of fatty acids, both of which are critical for larval development, adult maintenance, and successful foraging and reproduction (Robinson and Nation, 1970; Wright et al., 2018). Pollen and nectar are stored and consumed as bee bread, a fermented mixture preserved within hive cells for later use (Wright et al., 2018). Because individual plant species differ markedly in their nutritional composition, colonies forage from a diversity of floral sources to avoid nutritional deficiencies and to balance essential nutrients (Hendriksma and Shafir, 2016; Mayda et al., 2020; Zarchin et al., 2017).

Among these nutrients, lipids—and particularly fatty acids—play a critical role in determining pollen quality and colony health. Lipids are consumed in pollen and bee bread (Wright et al., 2018), and the sugar-rich, enzyme-containing matrix of honey helps preserve lipid fractions during storage (Čeksterytė et al., 2020). The lipid fraction of pollen consists primarily of fatty acids and typically comprises between 2% and 20% of total pollen dry mass, depending on the botanical source (Nicolson, 2011). Experimental evidence demonstrates that honeybees increase pollen collection and consumption in response to higher lipid content in floral sources (Arien et al., 2018; Pernal and Currie, 2002; Ruedenauer et al., 2021; Zarchin et al., 2017). For example, in feeding studies, highly preferred pollen species containing 17%–20% lipids trigger stronger stimulatory responses compared to species with 11%–12% lipids (Singh et al., 1999). Such responses influence foraging strategies and facilitate nutritional balancing by allowing colonies to regulate total lipid intake relative to protein and other essential nutrients, while also benefiting from the diversity and composition of fatty acids that determine lipid quality (Schmidt, 1985; Schmidt and Hanna, 2006). Together, these findings suggest that both total lipid content and fatty acid diversity across pollen sources are important determinants of colony nutrition.

Fatty acids in pollen lipids are commonly classified as essential, intermediate, or minor contributors to honeybee nutrition (Manning, 2001; Robinson and Nation, 1970). Essential fatty acids cannot be synthesised by bees and must be obtained from plant sources. These typically include palmitoleic (omega-7), linoleic (omega-6), and linolenic (omega-3) acids, which together can contribute 60%–80% of total pollen fatty acid content, depending on the botanical origin (Mayda et al., 2020). These fatty acids have been linked to brood development, immune defence, and disease resistance (Arien et al., 2015; Manning, 2001). For example, omega-6 to omega-3 ratios close to 1:1 increase successfully capped brood in artificial feeding experiments (Arien et al., 2020), a balance that is likely achieved through mixing pollen species with complementary fatty acid profiles (Čeksterytė et al., 2008). Intermediate fatty acids, including decanoic, dodecanoic, oleic, and myristic acids, can be synthesised physiologically but confer additional health benefits when consumed in pollen form (Bryś et al., 2021; Huang, 2012). Oleic acid (omega-9), for instance, ranges from 5% to 65% of pollen lipids depending on the botanical source and constitutes a major component of lipid reserves in larvae and adult bees (Manning, 2001; Robinson and Nation, 1970). Monofloral or low-diversity diets have been shown to reduce honeybee health in experimental studies (Manning, 2001), further underscoring the importance of fatty acid diversity within pollen diets.

Because floral diversity determines the range of pollen sources available to colonies, landscape composition may indirectly shape the diversity and balance of fatty acids present in bee bread. The floral diversity within 5–10 km of an apiary influences the botanical composition of collected pollen and, consequently, honeybee nutrition (Filipiak et al., 2022; Steffan-Dewenter and Kuhn, 2003). However, land-use change and habitat destruction for agriculture and urban development have reduced floral diversity in many regions (Decourtye et al., 2019; Parreño et al., 2021). In temperate environments where forests have been largely cleared, honeybees increasingly forage on remnant trees (Donkersley et al., 2014). Similar patterns have been observed in tropical systems (Cannizzaro et al., 2022), where rates of deforestation are particularly high (Hoang and Kanemoto, 2021). In temperate landscapes, lipid diversity and quality have been linked to floral diversity (Donkersley et al., 2017; Trinkl et al., 2020), yet in tropical environments, this relationship remains poorly understood. Although recent work demonstrated higher floral diversity in bee bread collected from forested tropical landscapes compared to non-forested areas (Cannizzaro et al., 2022), no study has directly examined how tropical forest presence influences the diversity and concentration of fatty acids within bee bread.

In this study, we investigated the lipid profiles of bee bread in tropical landscapes, comparing areas with and without forest cover. Specifically, we addressed the following research questions: 1) What fatty acid methyl esters (FAMEs) are present in bee bread collected from tropical landscapes, how diverse are they and what are their concentrations? 2) How do these concentrations differ in relation to forest presence or absence? 3) How are the botanical origins and diversity of bee bread associated with the concentration and diversity of its lipid content? By integrating fatty acid profiling with botanical composition data, this study provides insight into how landscape-driven differences in floral diversity influence the nutritional quality of bee bread in tropical environments, with implications for the health of managed honeybees.

Study site and bee bread collection

Bee bread was collected in March 2019 from eight apiary sites in the Eastern Highlands of Papua New Guinea (Figure 1). Bee bread was selected rather than freshly collected pollen because it represents pollen that has undergone microbial processing and storage within the hive and is therefore the form consumed by developing brood and adult bees. As bee bread accumulates over time, it provides a time-integrated representation of dietary intake and reduces short-term variability associated with individual foraging bouts. We classified sites as either forest or no-forest landscapes using Google Earth Engine and Earth Observation Research Centre satellite data from the ALOS PALSAR global forest/no-forest maps (Gorelick et al., 2017; Shimada et al., 2014). Four sites were classified as ‘Forest’ and contained forest cover >20% of total land area, and the remaining four sites were classified as ‘No Forest’ and contained no-forest cover within a 2.5-km radius from the centre (apiary) of each site.

We collected six bee bread samples per hive from three hives per site (except for one site where only two hives were accessible) for a total of n = 138 bee bread samples. Each bee bread sample was transferred to a 5-mL glass vial containing n -hexane to preserve the samples and maintain lipid content. Samples were vortexed and centrifuged at 14,000 rpm (Eppendorf 5810R) for 15 min, and n -hexane was removed from each sample using nitrogen evaporation under a fume hood through an air pore tape. Dried samples were then homogenised, weighed, and stored at −20°C.

Metabarcoding and bioinformatics

The taxonomic composition of pollen in bee bread was determined using DNA metabarcoding. Approximately 30 mg of dry bee bread (mean ± SD: 31.6 ± 1.03 mg) was flash-frozen in liquid nitrogen and homogenised in a TissueLyser II (Qiagen) for 6 min at 30 Hz using a mixture of 0.1, 1.0, and 2.3 mm zirconia/silica beads. Genomic DNA was extracted from the lysate using the DNeasy Plant Mini Kit (Qiagen, Germany) according to the manufacturer’s instructions.

Metabarcoding followed the protocol of Wilson et al. (2021). We targeted the chloroplast rbcL marker due to its broad amplification across angiosperms, extensive representation in public reference databases (e.g., GenBank and BOLD), and reliable genus-level taxonomic resolution for most flowering plants. The conserved priming regions of rbcL facilitate consistent amplification from mixed pollen samples, making it well-suited to characterising diverse floral assemblages in bee bread and aligning with the aims of this study. Absolute read abundances are prone to distortion due to biases among species at any step in the metabarcoding process, e.g., one species may generate many more reads than another during the PCR process. Relative read abundances are considered more reliable, and the proportion of reads can be considered a reasonable approximation of the proportion of species in a mixture (Bell et al., 2023). Thus, we used relative read abundances for all statistical tests.

The rbcL region was amplified from bee bread DNA using the rbcL2 (Palmieri et al., 2009) and rbcLa-R (Kress and Erickson, 2007) primer pair, adapted for Illumina amplicon sequencing (Kozich et al., 2013). Amplifications were conducted in triplicate 10 μL reactions per sample to minimise PCR bias. Each reaction contained 5 μL of KAPA HiFi HotStart ReadyMix (Kapa Biosystems, MA), 0.6 μL of each primer, and 3.8 μL of DNA template. Thermal cycling conditions were as follows: 95°C for 3 min; 35 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 1 min; followed by a final extension at 72°C for 10 min. Triplicate PCR products were pooled and normalised using the SequalPrep Normalisation Plate Kit (Invitrogen, CA, USA). Libraries were quantified using a Bioanalyzer High Sensitivity DNA assay (Agilent Technologies, Santa Clara, CA, USA) and a dsDNA High Sensitivity assay (Life Technologies GmbH, Darmstadt, Germany), pooled equimolarly, and sequenced on an Illumina MiSeq platform with v3 600-cycle chemistry (Illumina Inc., San Diego, CA, USA) using the FASTQ workflow.

A curated plant DNA reference database was compiled using BCdatabaser (Keller et al., 2020) to support taxonomic assignment of bee bread sequences. Bioinformatic processing was conducted using VSEARCH v2.1.4.2, and sequences were quality filtered, clustered, and taxonomically assigned following the protocols described by Campos et al. (2021).

Diversity metrics, including Shannon–Wiener diversity (H′), Pielou’s evenness (J), and species richness (S), were calculated as described in Cannizzaro et al. (2022).

Sugar content of bee bread

The Sugar content of all samples was quantified to identify honey and nectar components of bee bread. We weighed and then suspended each dried bee bread sample in 300 µL of Milli-Q water and vortexed the samples for 3 min to separate the sugar from the pollen. We then centrifuged the samples for 3 min at 3,000 rpm and used a hand-held refractometer (ATAGO IC-PAL-1 refractometer) to measure the degree Brix percentage of the supernatant. Degree Brix measurements were then converted to µg per mg of bee bread.

Oil extraction

Oil was extracted from bee bread samples to determine the lipid fraction and fatty acid content in bee bread samples. Bee bread samples, each weighing 50 mg, were transferred to 2-mL Eppendorf tubes, and 2.3 mm of zirconia beads and 0.1 mm of zirconia sand were added. Tubes were frozen in liquid nitrogen and shaken in a TissueLyser II at 30 Hz for 3 min; this step was then repeated twice.

Ground bee bread samples were transferred to glass vials and suspended in 4 mL of hexane. Glass vials were vortexed, covered in tin foil, and allowed to stand for 12 h at 25°C. Samples were centrifuged at 2,000 rpm for 1 min. The hexane–oil phase was transferred to a new 4-mL glass vial, and hexane was evaporated under nitrogen gas until only oil remained. All remaining oils were weighed and stored at −20°C until methyl esterification.

Derivatisation

Samples were prepared for methyl esterification by adding 700 µL of a BHT solution (100 mL of methanol CH 3 OH and 15 mg of butylated hydroxytoluene C 15 H 24 O) to each sample followed by 25 µL of hydrochloric acid (HCl 33%). Samples were placed in an oven and left to derivatise overnight at 60°C. Samples were removed from the oven and 500 µL of hexane was added to each sample followed by 500 µL of Milli-Q water. Samples were mixed for 30 s, and the water–methanol layer was removed. An additional 500 µL of Milli-Q water was added to each sample, mixed, and removed. Anhydrous sodium sulphate (Na 2 SO 4) was then added to each sample, and the hexane phase (~300 µL per sample) was transferred to a 350-µL glass insert in a 2-mL brown GC vial and stored at −20°C.

FAME analysis

Fatty acid methyl ester analyses were performed on a PerkinElmer Clarus 580 GC coupled to PerkinElmer Clarus SQ8S MS. The column used was an Elite-5MS (30 m × 0.25 mm × 0.25 μm). The helium carrier gas had a constant flow of 1 mL/min. The injection port was 280°C with a split ratio of 100:1 from −0.50 to +1.00 min and then set to 30:1. The temperature programme was operated from 50°C for 0.5 min, ramping at 10°C/min until 300°C and holding for 1.0 min. The mass spectrometer analyzed a mass range from 40 to 400 (m / z), from 3.1 to 26.5 min at 70 eV. Compounds were identified by comparison of retention times to authentic standards and comparison of mass spectra against the National Institute of Standards and Technology (NIST) (08) MS library match. Quantitation was via integration of the total ion current chromatogram.

Statistics

We tested how fatty acid compositions are affected by landscapes with generalised linear mixed models (hereafter GLMM) using the R package glmmTMB (Brooks et al., 2017). We constructed GLMMs with a Tweedie distribution and log-link function to account for continuous, zero-inflated data. We used a hierarchical approach for each GLMM by beginning with a model composed of the following explanatory variables: landscape (forest or no forest), bee bread weight, and extracted bee bread oil weight when variables were not significantly correlated. Random effects were the number of hives per site and hive identity (nested within sampled sites) to account for any variance from stratified sampling methods. We used the R package DHARMa to assess fitted model residuals to test for dispersion from normality and deviations from homoscedasticity (Hartig, 2020). P -values for explanatory variables were obtained by maximum likelihood (adaptive Gauss–Hermite quadrature) ratio tests using the package afex (Singmann et al., 2015). Explanatory variables were tested for interactions and, where not significant, were dropped at each model iteration before comparing all models by analysis of variance using the R package stats. We selected the best model fit by the AIC score and likelihood ratio tests (χ 2).

Spearman’s rank correlations were performed to test the relationships between total FAMEs, Shannon–Wiener diversity, and Pielou’s evenness of bee bread using the R package stats. Spearman’s rank correlations were also performed to test the relationships between relative read abundances of plant taxa in bee bread and µg per mg of fatty acid concentrations. Weak correlations, where Spearman’s ρ values were below <0.2, were removed, and bee bread taxa that did not show any statistically significant correlations were removed from the final correlation matrix.

We identified 12 fatty acids in total from bee bread samples. The mean fatty acids consisted of 43% saturated and 57% unsaturated lipids of bee bread. There was no significant difference between saturated and unsaturated fats between landscapes. The fatty acids that were at detectable concentrations above >2.5 µg/mg (linoleic, palmitic, and oleic acids) showed no significant relationships to landscape, hives per site, or bee bread weight (Figure 2). Elaidic, behenic, linolenic, eicosenoic, and arachidic acids were present in concentrations below <1.25 mg/mg and showed no significant relationships to landscape, hives per site, or bee bread weight (Figure 3). The fatty acids with the highest concentrations were oleic acid (omega-9) in bee bread samples followed by palmitic acid, linoleic acid, and stearic acid (Figures 2, 4).

The mean ratio across all samples of essential omega-6 to omega-3 was 3.8:1, higher in omega-6, and did not differ significantly between forest and no-forest landscapes. Oil yields extracted from bee bread samples were not significantly different between forest (mean and SE = 2.02 ± 0.03 µg/mg) and no-forest (mean and SE = 2.27 ± 0.06 µg/mg) landscapes. Brix values of sugar content in bee bread samples were also not significantly different between forest (mean and SE = 0.33 ± 0.01) and no-forest (mean and SE = 0.30 ± 0.01) landscapes.

Lauric acid was higher in no-forest landscapes compared with forested landscapes (GLMM: χ 2 = 5.73, Df =1, P = 0.01*; Figure 4A). In contrast, myristic acid was higher in forest landscapes compared with no-forest landscapes (GLMM: χ 2 = 4.95, Df =1, P = 0.05; Figure 4B), as was palmitoleic acid (GLMM: χ 2 = 6.57, Df =1, P = 0.05; Figure 4C) and stearic acid (GLMM: χ 2 = 6.33, Df =1, P = 0.05; Figure 4D).

Total fatty acid diversity in bee bread increased significantly with Shannon–Wiener diversity of bee bread (rho = 0.21, P = 0.02; Figure 5A) and Pielou’s evenness (rho = 0.23, P = 0.008*; Figure 5B), but both relationships were weak, and there was no significant increase with species richness and total fatty acid diversity.

Metabarcoding yielded a total of 5,174,738 sequencing reads, with 5,166,928 reads retained following quality filtering, representing 80 plant taxa across all samples. Concentrations from 8 of the 12 fatty acids identified in this study were positively correlated with relative read abundances of 23 plant taxa detected in bee bread samples. Statistically significant correlations between relative read abundances of bee bread taxa and fatty acid concentrations ranged from rho = 0.20 to 0.52 (Figure 6).

In total, bee bread from seven tree taxa, including Calophyllum soulattri, Leucaena leucocephala, Pinus sp., Coffea arabica, Dillenia papuana, Callicarpa arboreta, and Melanolepis multiglandulosa, was correlated with one or more fatty acids. A further 12 herbaceous taxa, including Sida rhombifolia, Loranthaceae, Piper recessum, Hylodesmum nudiflorum, Ageratum conyzoides, Bidens pilosa, Mimosa pudica, Piper sp., Piper gibbilimbum, Asteraceae sp., Euphorbia glyptospermia, and Erigeron canadensis, were also correlated with one or more fatty acids. In addition, spores from bee bread samples representing four fern taxa (Sticherus sp., Gleicheniaceae sp., Cyatheales sp., and Angiopteris sp.) showed significant correlations with fatty acid concentrations (Figure 6).

Our results show that bee bread fatty acids in tropical environments have high omega-9 and low omega-3 concentrations compared to temperate environments (Avni et al., 2014; Dong et al., 2015). We found that the total number of fatty acids (completeness) increases with the diversity of plant taxa in bee bread, regardless of landscape. Our study demonstrates that more diverse pollen diets were associated with a more complete fatty acid profile in honeybee diets, supporting the role of plant diversity in the landscape for honeybee nutrition. We also found that the concentrations of some minor fatty acids were higher in landscapes with greater forest cover, indicating a positive association between forested habitats and fatty acid diversity.

Bee bread in our study had high concentrations of linoleic (omega-6), palmitic (omega-7), and oleic (omega-9) fatty acids but low concentrations of linolenic acid (omega-3) compared with temperate environments (Avni et al., 2014; Dong et al., 2015). To our knowledge, few studies have examined fatty acid concentrations in bee bread in the tropics, and no studies have examined the correlations between fatty acids and plants across landscapes. The few studies conducted in tropical environments found low levels of linolenic acid (~1.4%–3.8% of pollen), supporting our results that contrast with temperate environments, where it is one of the most abundant fatty acids in pollen, ranging from ~20% to 30% (Bastos et al., 2004; de Arruda et al., 2013; Odoux et al., 2012). The range of linolenic acid concentrations in pollen is dependent on the botanical sources they are derived from, varying between 0.9% and 61% in pollen from temperate flora, and in some species being entirely absent (Avni et al., 2014; Manning, 2001; Trinkl et al., 2020). In these studies, the abundant plant taxa in bee bread samples in the tropics were Eucalyptus (Myrtaceae) and Eupatorium (Asteraceae), while in temperate environments, the abundant taxa were Trifolium (Fabaceae), Impatiens (Balsaminaceae), and other herbaceous species (Bastos et al., 2004; Donkersley et al., 2017). We found that oleic acid (omega-9) was the most abundant fatty acid in bee bread samples, and linoleic acid (omega-6) was also among the higher concentration fatty acids. This is supported by other tropical and temperate studies that show levels of oleic acid (omega-9) ranging up to ~14.39% of bee bread and linoleic acid (omega-6) ranging up to ~49.7% of fatty acids in bee bread (Bastos et al., 2004; Čeksterytė et al., 2008). Interestingly, the same studies found substantially higher concentrations of arachidic acid, ranging from ~12.1% to 42.7%, compared to trace amounts found in this study (Bastos et al., 2004; Čeksterytė et al., 2008). Arachidic acid is high in pollen from temperate taxa, including Brassica, Zea mays, Prunis dulcis, and other temperate species (Manning, 2001). Ratios of omega-6:3 are important for larval development and overall health, and a 1:1 ratio has been shown to provide the best health outcomes for bee development and survival (Arien et al., 2020). Our results show a ratio of linoleic:linolenic acid (omega-6:3) of 3.8:1, suggesting that plant taxa in tropical rainforests may be deficient in omega-3 fatty acids but high in other important fats such as omega-6 and omega-9. Based on our findings, we suggest that tropical environments can be a rich source of unsaturated essential fats for honeybee health, and bees may benefit from supplementing plant sources rich in omega-3 fatty acids.

We found that the completeness of fatty acids per bee bread sample increased with Shannon–Wiener diversity and species evenness of plant sources in bee bread samples although the correlation coefficients were low. Previous studies have found lipid concentration increases with species richness surrounding apiaries and the diversity of pollen in subtropical environments, but our study is the first to find a relationship between the number of fatty acids and the diversity in bee bread collected by bees (Trinkl et al., 2020). Bees are able to detect fatty acid concentration and select pollen sources that are higher in lipid content (Ruedenauer et al., 2021). This demonstrates honeybees’ capacity to actively seek out a diversity of plant sources to achieve a balanced diet for the colony (Singh et al., 1999; Wright et al., 2018; Zarchin et al., 2017). In a tropical context, our results demonstrate that tropical landscapes provide sufficient flora that enables nutritional balancing of fatty acids, which may aid in antimicrobial protection of larvae and provide complementary fatty acids necessary for honeybee health.

Fatty acid concentrations were positively correlated with the relative read abundances of numerous bee bread taxa. Notably, relative reads of B. pilosa were highly correlated with stearic, palmitic, myristic, and eicosenoic acids. These results are supported by a recent study that examined the fatty acid chemistry of B. pilosa pollen and reported similarly high concentrations (referenced as dominant fatty acids) of stearic and palmitic acids but found myristic and eicosenoic acids to be non-dominant fatty acids (Hsu et al., 2021). Interestingly, three Piper species (P. recessum, P. gibbilimbum, and Piper sp.) plants were correlated with elaidic acid, and two (P. recessum and Piper sp.) were both correlated with linoleic acid. We found four fern taxa that were correlated with palmitic and stearic fatty acids, both of which have been identified in previous studies exploring the lipid chemistry of fern spores (Gemmrich, 1977; Nekrasov et al., 2019). Fern taxa are not generally associated with honeybee flora but have been identified in bee bread in several DNA barcoding studies (Gous et al., 2019; Hawkins et al., 2015; Olga et al., 2012). Ferns release spores containing starches, proteins, and lipids at similar quantities relative to their total mass, as pollen grains do (Gómez-Noguez et al., 2016; Pedrero‐López et al., 2023). The abundance and specific fatty acid profile of spores, however, have been reported to be related more to sporulation abundance rather than the species of fern (Nekrasov et al., 2019). In tropical environments where ferns are easily accessible, honeybees have been observed making use of these resources (Roubik and Patiño, 2018). We found that pine pollen in bee bread was also correlated with fatty acids: palmitoleic and eicosenoic acids. The difference between fatty acid profiles in pollen of gymnosperms compared to angiosperms has been quantified with one recent study showing that certain species of gymnosperms (Picea sp.) contain significantly lower lipid and fatty acid concentrations compared to angiosperms (Liliaceae and Solanaceae), although there is less evidence for the specific fatty acid difference between species or compared to other gymnosperms (Breygina et al., 2023; Manning, 2001). The native rainforest trees Calophyllum soulattri, Dillenia papuana, Callicarpa arboreta, and Melanolepis multiglandulosa also provided significant lipid nutrition, mostly in the form of myristic, lauric, and behenic acids.

Herbaceous, introduced species represented 52% of the plant taxa relative read abundance that was significantly correlated with bee bread fatty acids, despite the majority of relative abundance plant taxa in bee bread samples coming from trees. This demonstrates how herbaceous species and weeds provide an important supplementary source of nutrition for honeybees in tropical landscapes.

Our previous study revealed that greater diversity of plant taxa in bee bread was associated with forest landscapes in tropical environments (Cannizzaro et al., 2022). The present study shows that forest cover in the landscape results in higher concentrations of myristic acid, palmitoleic acid, and stearic acid, while lauric acid was higher in non-forested landscapes. In contrast, most essential and intermediate fatty acids were similar between landscapes. Myristic acid was substantially greater in bee bread from forest landscapes. Myristic acid also has strong antimicrobial properties and is an effective inhibitor of the honeybee pathogen Paenibacillus larvae (American foulbrood) (Feldlaufer et al., 1993). Palmitoleic and stearic acids were also higher in forest landscapes. Palmitoleic acid (omega-7) is considered a strong complementary fat to omega-3 found in tropical nuts and is a good source of unsaturated fat (Hu et al., 2019). Stearic acid is important in wax production that supports colony growth and is an important precursor compound used in synthesising mandibular excretions in queens and worker bees (Buchwald et al., 2009; Plettner et al., 1996; YuQi et al., 2015). High concentrations of lauric acid have been identified in several studies in certain species of Asteraceae (Nicolson and Human, 2013; Standifer, 1966). In our study, bee bread samples contained a high proportion of herbaceous taxa from Asteraceae (~28% of herbaceous taxa) that may have contributed to high levels of lauric acid. Lauric acid has high antimicrobial activity against P. larvae (American foulbrood) and may benefit honeybee nutrition in disease resistance in sites without forest plant sources (Feldlaufer et al., 1993).

Our results demonstrate that higher floral diversity in the landscape increased the diversity of fatty acid profiles in bee bread. Thus, the diversity of flora in the landscape can improve nutritional balancing and support honeybee health. We show that fatty acid concentrations in tropical bee bread are high in omega-6 and omega-9 but low in omega-3 and that minor sources of fatty acids are higher in landscapes with forest surrounding hives. We recommend maintaining floral diversity in tropical landscapes to support the health of honeybees.