*Article* **Silicone Wristbands as Passive Samplers in Honey Bee Hives**

### **Emma J. Bullock 1, Alexis M. Schafsnitz 1, Chloe H. Wang 1, Robert L. Broadrup 1, Anthony Macherone 2,3, Christopher Mayack 4,5 and Helen K. White 1,\***


Received: 7 June 2020; Accepted: 2 July 2020; Published: 6 July 2020

**Abstract:** The recent decline of European honey bees (*Apis mellifera*) has prompted a surge in research into their chemical environment, including chemicals produced by bees, as well as chemicals produced by plants and derived from human activity that bees also interact with. This study sought to develop a novel approach to passively sampling honey bee hives using silicone wristbands. Wristbands placed in hives for 24 h captured various compounds, including long-chain hydrocarbons, fatty acids, fatty alcohols, sugars, and sterols with wide ranging octanol–water partition coefficients (Kow) that varied by up to 19 orders of magnitude. Most of the compounds identified from the wristbands are known to be produced by bees or plants. This study indicates that silicone wristbands provide a simple, affordable, and passive method for sampling the chemical environment of honey bees.

**Keywords:** *Apis mellifera*; bee; silicone band; hive; passive sampler

#### **1. Introduction**

The dwindling global population of the honey bee pollinator *Apis mellifera*, the main pollinator species used in agriculture, has driven researchers to investigate honey bees and their responses to different environmental and pathological conditions [1,2]. These insects play a vital role in agricultural systems and almost a third of the global food crop depends exclusively on the European *Apis mellifera* species [3]. Since the 1940s, the number of managed honey bee colonies in the United States has declined, with the most dramatic losses occurring in the past decade [4]. Numerous factors have been implicated in the decline in bee health including exposure to pesticides and viruses [5], the spread of *Varroa destructor* mites [6], and parasites such as *Nosema ceranae* [7]; however, none can be identified as a single underlying causal factor.

Due to the vast number of variables and interactions that impact honey bee health, identifying the cause, impact, and risk associated with individual factors is challenging. Targeted exposure studies assessing chemicals and contaminants including pesticides and pathogens have been examined [6]. Recently, an exposomic approach has explored a more comprehensive approach where external chemicals as well as those produced by honey bees are combined into a single analysis [8]. Some compounds produced by bees, known as semiochemicals, are released by individual honey bees as cues or signals in order to relay information to hive mates, allowing honey bees to coordinate their responses to environmental stimuli as a hive [9]. Semiochemicals can also ward off or attract other honey bees [9], can relay the needs of larvae to nurse bees [10,11], and are involved in social recognition [12,13]. Certain chemicals have the potential to relay specific information about the disease state of an individual. For example, ethyl oleate is a chemical produced by honey bees that is significantly increased by infection of the gut fungus *N. ceranae* [13], and cuticular hydrocarbon profiles of honey bees have been observed to change when exposed to bacterial or *N. ceranae* infections [14,15]. It has also been shown that poor nutrition can lead larvae to release E-β-ocimene into the hive [11]. Monitoring chemicals produced by honey bees is of significant interest as it can provide insight into specific infection status, pesticide exposure, as well as nutrition deficiencies, thus providing insight into the overall health of a bee colony.

Semiochemicals and other volatile and semivolatile chemicals (VOCs and SVOCs) found in honey bee hives have previously been sampled by collecting the air in the hive directly via a syringe and then examining the chemical contents of the air sample via gas chromatography–mass spectrometry (GC–MS) [16]; however, this method does not allow the storage of samples prior to analysis [16]. Alternative methods have utilized adsorbent packing materials (such as Super Q, Hayesep Q, Porapak Q, Tenax TA) combined with vacuum/air tube systems [10,16,17]. These methods have limitations as the vacuum/air tube systems are complicated to construct and their presence may cause vibrations that induce a stress response in the bees and bias the chemical profiles measured [16]. Solid-phase microextraction (SPME) fibers have been used as passive samplers, avoiding the problems involved in complicated air flow setups; however, they are somewhat limited as they are designed to target specific chemicals, and are expensive and fragile to use [16–21].

This study aims to apply silicone wristbands, a previously established method [22–26], as a new approach to passively sample a wide range of chemicals present in honey bee hives. Silicone wristbands are cheap, commercially available, and have been used as passive environmental samplers to provide time-averaged concentrations of human exposures to polycyclic aromatic hydrocarbons (PAHs), consumer products, personal care products, pesticides, phthalates, other industrial compounds [22–24], and organophosphate flame retardants [25,26]. Compounds adsorbed by silicone wristbands have been found to remain on the bands for extended periods of time, unlike compounds on SPME fibers [24–27]. Perhaps the most significant advantage of silicone wristbands is that they are able to sample VOCs and SVOCs with octanol-air partition coefficients (Koa) ranging in 10 orders of magnitude [22–24]. The Koa as well as the octanol–water partition coefficients (Koaw) of a chemical compound are values that represent the hydrophobicity of a compound. Specifically, these values describe the ratio of a compound in octanol divided by the concentration of the same compound in air or water for Koa and Kow, respectively. With respect to passive samplers, these values are used to show the range in hydrophobicity of compounds that a sampler can detect, which, as mentioned, is large for silicone wristbands. Their ability to adsorb such a broad range of compounds makes them ideal for sampling chemicals produced by bees and could potentially collect a broader range of compounds than the aforementioned SPME fibers and polymer samplers.

In this study, commercially available silicone wristbands were pre-cleaned and placed in honey bee hives to sample VOCs and SVOCs. This study presents information about the types of chemicals that can be adsorbed and analyzed to provide a reliable chemical profile of honey bee hives from a variety of urban and suburban locations.

#### **2. Materials and Methods**

#### *2.1. Materials*

All solvents were GC Resolv or Optima grade, obtained from Fisher Scientific (Newark, NJ, USA). Glassware was combusted at 450 ◦C for 8 h before use.

#### *2.2. Band Preparation*

Silicone bands (acquired from 24HourWristbands.com) with a circumference of 20 cm, <sup>≈</sup>40 cm2 surface area, and ≈4 g of sorbent were prepared according to a modified procedure [22]. Briefly, bands were cleaned in previously combusted (450 ◦C, 8 h) glass jars with a 1:1 volume mixture of ethyl acetate (EtOAc) to hexanes and shaken at 80 rpm for at least 2.5 h before the removal of the solvent. This procedure was repeated three more times, at which point the bands were removed from the jar and placed on combusted aluminum foil in a high-vacuum pressure oven at 60 ◦C for 48 h. The bands were stored individually in 40 mL combusted amber glass vials at room temperature prior to use.

#### *2.3. Band Deployment*

During July, August, September, and October of 2016, bands were deployed in 10 apiaries in rural, suburban, and urban areas of southeastern Pennsylvania (details provided in Table S1 of the Electronic Supplementary Materials). At each apiary, three hives were sampled across three time points. In order to capture different parts of the environment inhabited by the honey bees, we placed one silicone band under the top of each hive (in-cover) and placed one inside the entrance of each hive. An additional band was placed on top of each hive (outside cover) to serve as an external control (details shown in Figure 1). The bands were retrieved 24 h later and transferred to individual 40 mL amber vials, placed on dry ice, transported back to the lab, and stored at −20 ◦C.

**Figure 1.** Diagram of a hive (**left**) and photo of a hive in the field (**right**), highlighting the placement of bands. In-cover band for the hive in the field is not shown as it is inside the cover part of the hive. Wristbands shown are 20 cm in circumference and 1 cm width; hives are approximately 1 × 0.5 × 0.4 m.

#### *2.4. Band Extraction*

Silicone bands were cleaned and extracted according to a modified procedure [22]. Briefly, each band was rinsed with Milli-Q water to remove any solids, such as dirt, propolis, or stingers, followed by isopropyl alcohol to remove water. The bands were then placed in individual 500 mL jars with 100 mL EtOAc and shaken at 60 rpm for 2 h. The extract was removed and the extraction

process was repeated twice more before the extracts were combined, reduced in volume via rotary evaporation to approximately 1 mL, and stored at −20 ◦C prior to further analysis. The extracts were then solvent-exchanged into hexane, spiked with perdeuterated *n*-hexadecane as an internal standard (average recoveries were 74%), and charged onto a small glass pipette column (0.5 cm × 6 cm) packed with fully activated silica gel (100–200 mesh). The first fraction (F1) was eluted from the column with 4mL of hexane, followed by the second fraction (F2) eluted with 4 mL of an equal mixture of dichloromethane and methanol containing 1% formic acid. Laboratory blanks run alongside samples contained no detectable compounds.

#### *2.5. Gas Chromatographic Analysis of Band Extracts*

The F1 extracts were analyzed on a 1D Agilent 7890 series gas chromatograph coupled to a flame ionization detector (FID, Santa Clara, CA, USA). Compounds were separated on a J&W DB-XLB capillary column (30 m, 0.25 mm I.D., 0.25 μm film) with helium carrier gas at a constant flow of 1 mL min<sup>−</sup>1. The GC oven had an initial temperature of 40 ◦C (1 min hold) and was ramped at 10 ◦C min−<sup>1</sup> until 160 ◦C (1 min hold), then ramped again at 4 ◦C min−<sup>1</sup> until 320 ◦C (36 min hold). Quantities of *n*-alkanes were calculated using response factors determined from pure standards. Presence of *n*-alkenes in the F1 extracts were confirmed by analyzing select extracts on an Agilent 7890 series gas chromatograph with an Agilent 5975 mass selective detector (MSD) and were noted, though not quantified. All fractions from each in-cover, outside, and entrance band for 10 hives were analyzed via GC–MS with a compound detection limit of 0.1 ng/μL determined from a standard curve using pure standards.

In order to identify unknown compounds, we analyzed F2 extracts via a 1D Agilent 7890 series gas chromatograph coupled to an Agilent 5975 mass selective detector (MSD). Prior to analysis, F2 extracts were derivatized with bis(trimethylsilyl)trifluoroacetamide (BSTFA) in pyridine. Compounds were separated on a J&W DB-XLB capillary column (60 m, 0.25 mm internal diameter (I.D.), 0.25 μm film) with helium carrier gas at a constant flow of 1 mL min−1. The GC oven had an initial temperature of 40 ◦C (1 min hold) and was ramped at 10 ◦C min−<sup>1</sup> until 160 ◦C (1 min hold), then ramped again at 4 ◦C min−<sup>1</sup> until 320 ◦C (36 min hold). The MS was operated in electron-impact (EI) mode with an ionization energy of 70 eV. Spectra were acquired between *m*/*z* 40–650 at a scan rate of 1 cycle s<sup>−</sup>1. Fatty acids and fatty alcohols were identified from mass spectral and retention time characteristics compared to pure standards. All other compounds were tentatively identified from mass spectra and gas chromatographic retention characteristics.

#### **3. Results**

#### *3.1. Chemical Profiles of Hive Air Were Dominated by Hydrocarbon Compounds*

Bands placed in-cover and at the entrance of hives exhibited similar chemical profiles dominated by *n*-alkanes and *n*-alkenes with odd chain lengths between C21–C33 (Figures 2 and 3). The chemical profiles of *n*-alkanes and *n*-alkenes for the entrance bands had more variability between hives (Figure 3), likely due to the fact that entrance bands come into more frequent physical contact with the honey bees than the bands placed in-cover. Bands placed on the outside of hives did not contain compounds at detectable levels (Figure 2), and thus acted as field controls in this study. Differences in the chemical profiles from bands placed in hives at different locations were observed (Figure 3), but were not significant between the location types (rural, suburban, and urban) examined. For the remainder of this study, we chose to focus on the in-cover bands, which have similar chemical profiles to the entrance bands but represent the chemical profile of the hive air with less frequent physical contact, and are practically easier to recover.

**Figure 2.** Representative gas chromatograms of compounds extracted from (**a**) an outside band, (**b**) an entrance band, and (**c**) an in-cover band. Alkanes indicated with dots (•) and alkenes indicated with asterisks (\*). All bands were deployed in the same hive at Awbury Arboretum on 07/15/2016.

#### +\GURFDUERQ&KDLQ/HQJWK

**Figure 3.** Box and whisker plots illustrating the relative abundance of hydrocarbons for all samples collected from (**a**) the entrance and (**b**) in-cover of hives. Boxes contain the interquartile range—50% of data—with center line indicating the median value of data. Whiskers indicate minimum and maximum values within 1.5 times above the 75th percentile or below the 25th percentile, with dots representing outliers.

#### *3.2. Bands Adsorbed Bee-Associated Chemicals*

The predominant compounds detected from the bands (*n*- alkanes and *n*-alkenes with odd chain lengths between C21–C33) are honey bee semiochemicals (Table 1), some of which are known nestmate recognition semiochemicals [28–30]. Four of these compounds—C23- and C25-*n*-alkane, and C23- and C25-*n*-alkene—are also known to be vital for the waggle dance, which relays information about the whereabouts of food [9,31]. Oleic acid, linoleic acid, and α-linolenic acid were also detected in at least half of the samples. These compounds are known to be major constituents of pollen [32] and beeswax [33], and have been detected in worker bees [34,35]. Certain saturated fatty acids and fatty alcohols, previously detected in worker bees [34], drone cocoons [36], or as a constituent of queen retinue pheromone [31], were identified broadly across samples. The alarm pheromone, [Z]-11-eicosenol, which is known to be released by bees, was also observed [31,34]. Glycerol was found in 78% of samples, and is known to participate in the ester biosynthetic pathway in bees [31]. Chrysin is a compound found in honey and was only observed in 13% of samples [37].


**Table 1.** The percentage of silicone bands placed in honey bee hives that contain honey bee semiochemicals and honey bee-associated compounds organized by compound type.

<sup>a</sup> Specific percentages are for individual compounds. A range of values corresponds to the range of compounds described. <sup>b</sup> Log Kow values estimated using the Crippen method EPISuite KOWWIN v1.67 estimate (USEPA) (HSDB [40]). <sup>c</sup> Alkenes identified by molecular weight and fragmentation patterns. Exact location of the double bonds is unknown. <sup>d</sup> Alkenes with two double bonds.

#### *3.3. Bands Adsorb Plant-Derived Compounds*

Plant-derived compounds were also extracted from bands, as well as the fatty acid C9:0, a common non-selective herbicide (Table 2) [41]. Fatty acids with carbon ranges between C10:0–C22:0 are known to be derived from plants and have been identified in pollen, along with oleic, linoleic, and α-linolenic acid, as previously described [32]. Further, the fatty alcohol 1-tritriacontanol, which has been shown to have a plant origin, was detected [42,43]. Plant-originating allelochemicals were also detected, including benzoic acid and cinnamic acid derivatives [44,45]. In fewer samples, sterols and sugars were observed, likely to originate from pollen and nectar, respectively [46,47]. All compounds identified from bands placed in hives are described in Table S2 of the Supplementary Materials.


**Table 2.** The percentage of silicone bands placed in honey bee hives that contain plant-associated compounds organized by compound type.

<sup>a</sup> Specific percentages are for individual compounds. A range of values corresponds to the range of compounds described. <sup>b</sup> Log Kow values estimated using the Crippen method EPISuite KOWWIN v1.67 estimate (USEPA) (HSDB [40]).

#### **4. Discussion**

#### *4.1. Implications for Studies of Honey Bee Health*

In this approach of using silicone wristbands to passively sample chemicals present in honey bee hives, we found that all but one of the chemicals detected were associated with bees and plants as opposed to human activity or viruses. If these types of compounds were present, our inability to detect them likely arises from them being present at lower abundances, giving a smaller signal that is either masked by the much larger signal of the other abundant compounds present, or that it is outside of our detection limits. Our approach, however, is still of use, as several of the compounds detected on the bands have been implicated in studies concerning honey bee health and nutrition.

Compounds related to honey bee health include those produced by bees such as octadecanoic acid and (9Z)-octadecenoic acid. These two compounds have been observed to be produced more in hives that have large numbers of bees infected with *N. ceranae*, possibly as precursors to short fatty acid chains known to have antibiotic properties [8]. Hexadecanoic acid is a known regulator of fatty acid synthesis in bees and has been observed in lower quantities when a hive is infected [8]. Tricosanoic acid and tetracosanoic acid were both detected on bands, and while not produced by honey bees, are known *Varroa destructor* mite semiochemicals [35]. The alkane tricosane has also been shown to be produced more by bees upon infection by Gram-negative bacteria [14]. Overall, it is known that infection from various pathogens causes honey bees to alter their hydrocarbon profiles [15], which could be examined in future studies by analyzing the chemical profiles of bands placed in hives.

The plant-derived compounds identified in this study could also be used to extrapolate information about hive health, as honey bees obtain all necessary proteins, lipids, and vitamins from pollen [32]. It has been shown that not only the quantity of pollen collected, but also the quality and diversity of compounds available in the pollen, determine hive productivity and longevity [32]. As a result, the detection of numerous compounds from contact with pollen could help researchers investigate

the impact of local flora on honey bee hives. These results indicate that deploying silicone bands in honey bee hives may help researchers gain insight into several different aspects of honey bee health, including the regulation of important metabolic pathways, the presence of parasites and pathogens, and the quality of pollen being gathered.

#### *4.2. General Applicability of Bands to Honey Bee Research*

The log octanol–water partition coefficients (Log Kow) for compounds in this study range from −3.38 to 16.6, as reported in the Hazardous Substances Data Bank (HSDB) [40]. This exceeds the capacity of most passive samplers currently in use [24], overcoming the major drawback of currently used passive sampling methods in hives that target certain compound groups [16–19]. The diversity in compounds adsorbed by silicone bands thus removes the necessity for multiple passive samplers when the compounds of interest vary significantly in structure and physical and chemical properties. Despite our expectation that this method would primarily target VOCs and SVOCs, non-volatile compounds were consistently seen across samples, arising from direct contact of the bees with the bands. If future implementation of silicone bands in hives chose to sample VOCs and SVOCs only, then an approach to maintain separation between the bees and bands would be needed (e.g., fine wire mesh cage for the bands).

This study also provides a method that allows researchers to quantitatively compare hive chemical profiles without needing to determine compound concentrations in the air of each hive. Previous studies have shown that the consistent adsorption of silicone bands allow samples to be compared without the need for further calculations, provided the dimensions of the silicone sampler and deployment length are the same [25,48]. Silicone bands are more consistent in their compound adsorption ratios than several other passive samplers such as polyurethane foam, urine sampling, and hand wipes [25,26], potentially because the silicone matrix stabilizes compounds until extraction [24].

Unlike SPME fibers, silicone bands can be used for quantitative analysis of compounds, provided that any comparisons are between band samples, as concentrations adsorbed to the bands do not directly represent environmental concentrations [22,24,25]. It has also been shown that concentrations of volatile compounds adsorbed by the silicone matrix remain stable under transport conditions of 30 ◦C for 7 days and under storage conditions of −20 ◦C for approximately 6 months [24]. To compare concentrations of chemicals from different compound classes between bands, the specific affinities of the silicone matrix to different compounds would need to be examined. In addition, to connect the chemical compounds detected to the health of the hive, a future study would need to also measure specific infections and diseases, such as *N. ceranae*, the *Varroa* destructor mite, deformed wing virus, and European foulbrood, in addition to chemical compounds of interest. Any future study must be designed to be large enough so that statistical testing between chemicals and health to be made.

#### *4.3. Modifications for Future Studies*

One concern with the use of silicone wristbands is the large amount of solvent needed to clean and process the bands; however, in this study, the amount of sample necessary for compounds to be detected via GC–MS and GC–FID was approximately 3 <sup>×</sup> 10−<sup>5</sup> times our original extract per run. This shows that future studies could use fractions of silicone wristbands, rather than the entire band, significantly reducing the amount of solvent needed per sample. Researchers would have to ensure that each section of band being deployed was the same length, width, thickness, and surface area if they wished to compare concentrations of compounds between samples. It would also be necessary to ensure each band is deployed for the exact same duration, as many of the compounds may not reach equilibrium and could exhibit time-dependent concentrations [24].

Since the bands are capable of adsorbing both volatile and non-volatile chemicals, the placement of the bands in the hive will be critical to ensure non-biased quantitative analysis. If researchers are interested in volatile compounds, isolating the bands from the bees using a mesh cage (as previously mentioned) will be essential to obtain non-biased results. If the researcher is interested in non-volatile

compounds, such as hydrocarbons or pesticides, they could pin the band flat in front of the entrance of the hive, requiring bees to walk over it when entering and exiting the hive. This approach would make the quantification of chemicals challenging, but it would provide a broad overview of the compounds worker bees are carrying into the hive.

#### **5. Conclusions**

Considering the ease with which chemicals in bands can be compared, the minimal disturbance to the hive, and the variety of compounds detectable, using silicone bands to investigate the relationship between chemical compounds and honey bees shows great potential. Further, our results show that bands did not collect detectable compounds from outside of the hive, as no compounds were detected on the outside bands. This contrasts with SPME fibers, which are easily contaminated by background volatiles [16] and are quite expensive. Researchers can use bands as samplers in the open hive environment, as was done in this study, as well as in closed sampling containers. In a closed system, it would be possible to sample the volatile chemicals released by bees or adhered to the surface of bees based on certain castes, age groups, or environmental conditions, without the need for complicated air flow systems or filters. As a result, we believe using silicone band passive samplers provides alternative, flexible, more affordable opportunities to explore the chemical ecology of honey bees and the factors that influence their health, behavior, and survival.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2306-7381/7/3/86/s1, Table S1: Locations of hives sampled in this study, Table S2: All compounds identified on bands.

**Author Contributions:** Conceptualization, R.L.B., A.M., C.M., and H.K.W.; methodology, E.J.B., C.H.W., R.L.B., A.M., C.M., A.M.S., and H.K.W.; formal analysis, E.J.B., C.M., and H.K.W.; data curation, E.J.B., C.M., and H.K.W.; writing—original draft preparation, E.J.B. and H.K.W.; writing—review and editing, all authors; funding acquisition, R.L.B., A.M., C.M., and H.K.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by funding from an Agilent Applications and Core Technology University Research Grant (no: 3937) to R.L.B., A.M., C.M., and H.K.W. and a Francis Velay Summer Fellow award to E.J.B.

**Acknowledgments:** The authors acknowledge Julie Brady, Jeff Eckel, Claudia Kent, Don Shump, and Eli St. Amour for access to the hives under their care.

**Conflicts of Interest:** The authors declare no conflict of interest. A.M is an employee of Agilent Technologies, Inc. and contributed to the conceptualization of this study. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Communication* **The Honeybee Gut Mycobiota Cluster by Season versus the Microbiota which Cluster by Gut Segment**

**Jane Ludvigsen, Åsmund Andersen, Linda Hjeljord and Knut Rudi \***

Faculty of Chemistry, Biotechnology and Food Science, Norwegian University of Life Sciences, 1430 Ås, Norway; jane.ludvigsen@nmbu.no (J.L.); aan@mycoteam.no (Å.A.); linda.hjeljord@nmbu.no (L.H.) **\*** Correspondence: knut.rudi@nmbu.no

**Abstract:** Honeybees represent one of the most important insect species we have, particularly due to their pollinating services. Several emerging fungal and bacterial diseases, however, are currently threatening honeybees without known mechanisms of pathogenicity. Therefore, the aim of the current work was to investigate the seasonal (winter, spring, summer, and autumn) fungal and bacterial distribution through different gut segments (crop, midgut, ileum, and rectum). This was done from two hives in Norway. Our main finding was that bacteria clustered by gut segments, while fungi were clustered by season. This knowledge can therefore be important in studying the epidemiology and potential mechanisms of emerging diseases in honeybees, and also serve as a baseline for understanding honeybee health.

**Keywords:** honeybee; gut microbiota; gut mycobiota; season

#### **1. Introduction**

The importance of the gut microbiota (GM) in regulating honeybee (*Apis mellifera*) health has become increasingly evident in recent years. Regarding nutritional and immunological interactions with the host, the bacterial proportion of the GM is most frequently studied [1–6], but studies investigating the fungal part of the microbiota and how some fungi interact with honeybee pathogens have recently emerged [7]. In humans, the fungi part of the GM has recently been linked to human infant development [8].

The GM of honeybees harbor a specific set of 8–9 bacterial groups, which all seem to have evolved and adapted to a life in the guts of honeybees, and they are only found in honeybees or in closely related bees such as bumble bees [9]. These bacterial groups aid in nutritional breakdown of pollen and nectar, interact with the immune system, and contribute to pathogen defense in the gut [10]. Likewise, specific fungi show interactions with pathogens in honeybees [11,12], but little is currently known about the role of fungi in the GM of honeybees.

The honeybee gastrointestinal tract (GI) can be divided into four main parts: crop, midgut, ileum, and rectum. These four parts have been shown to harbor unique bacterial species for which metabolic properties have been elucidated [13]. Although detailed information about each bacterial species community changes, little information is available about how the bacterial community changes in regard to outer stimuli. Only a few studies have described how the GM composition changes according to season [14,15] and developmental stages of the host [16,17]. The fact that specific bacteria change during the season is an indication that diet contributes to variation in composition. Yun et al. 2018 [13] found that foragers harbor a different set of fungi than that of nurse bees, indicating that diet is a source of variation in the fungal composition as well. They also found that queen bees carry an overload of one type of fungi (*Zygosaccharomyces*), which is different from nurse bees, indicating that the fungal part of the microbiota, as seen for the bacterial part, has adapted to different lifestyles, which is reflected in the overall fungal microbiota [13].

**Citation:** Ludvigsen, J.; Andersen, Å.; Hjeljord, L.; Rudi, K. The Honeybee Gut Mycobiota Cluster by Season versus the Microbiota which Cluster by Gut Segment. *Vet. Sci.* **2021**, *8*, 4. https://doi.org/10.3390/ vetsci8010004

Received: 24 November 2020 Accepted: 26 December 2020 Published: 31 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

All previous studies have investigated the microbiota of the total GI (crop to rectum or midgut to rectum), resulting in a lack of information about the microbiota of different gut parts under these scenarios. Since some honeybee pathogens are gut part specific, information about the variation in the microbiota composition in different gut parts is crucial for understanding the microbiota–pathogen–host dynamics in more detail.

Here, we investigate both the bacterial and fungal parts of the GM (by gut parts) crop, midgut, ileum, and rectum—throughout an entire season (longitudinal), from March until November in adult honeybee workers. Our results can aid in understanding specific gut part-specific interactions and help in the design of later in-depth functional and metabolic studies.

#### **2. Materials and Methods**

#### *2.1. Sampling*

The bees were sampled from two neighboring hives located at the Norwegian University of Life Sciences, As, Norway, across seven time points representing before, during and after foraging season. Ten bees where picked randomly from the frame closest to the opening of both hives to represent foragers in April, June, July, and August. In March and November, the bees are not foraging but are clustered together to keep warm, and thus ten bees from both hives were picked from the top of the formed cluster. All bees represent adult bees. All experiments were conducted following Norwegian rules for studies on honeybees [18].

#### *2.2. Gut Dissection*

Bees were sampled and put on ice to induce chill-coma before dissection. Then, the gut was removed from the bee after we had sterilized the bee on the outside by washing it in 50% ethanol. By pulling out the stinger, the gut from midgut to rectum was removed. The crop was dissected out separately with sterile dissecting tools. Unfortunately, we did not collect crops from bees sampled in March. The gut was cut into its respective parts under a dissection microscope and on sterile microscopy slides (washed with 70% ethanol and 1:10 chlorine) by cutting at the transition areas between the different gut parts (Figure S1). The dissection was performed in a drop of PCR water (VWR, Radnor, PA, USA) and the Malpighian tubes were cleaned off the midgut and left attached to the ileum part. The different gut parts were collected in tubes prefilled with a bead matrix consisting of 0.2 g of each < 106 μM acid-washed glass beads, 0.425–0.600 mm glass beads, and 2 2.5–3.5 mm acid-washed glass beads (Sigma-Aldrich, Darmstadt, Germany) and stool transport and recovery (S.T.A.R) buffer (Roche, Mannheim, Germany). For midguts and rectum samples, 300 μL of STAR buffer were used and 200 μL were used for crop and ileum. The dataset contained four gut parts (crop, midgut, ileum, and rectum) from 20 bees over seven months (except from March, which only contained midgut, ileum, and rectum) for a total of 540 samples. The samples were frozen at −20 degrees before DNA extraction.

#### *2.3. DNA Extraction*

DNA was extracted from a total of 540 samples, with 484 samples yielding sufficient DNA for further processing. Mechanical lysis was performed using FastPrep (MP Biomedicals, Santa Ana, CA, USA) at 1800 rpm for 40 s, two times, with a 5 min cool-down between runs. The samples were then centrifuged for 5 min at 13,000 rpm and 50 μL supernatant was transferred to a 96-well plate for DNA extraction using the MagTM mini kit (LGC, Middlesex, UK) following manufacturers recommendations. The extraction was performed on the KingFisherTM Flex Magnetic Particle Processor, (Thermo ScientificTM, Waltham, MA, USA). The extracted DNA was frozen at −20 degrees before subsequent PCR/qPCR and Illumina sequencing.

#### *2.4. qPCR*

Quantification of bacteria and fungi in different gut parts across seasons was performed using qPCR assays targeting the 16S rRNA gene for bacteria and the ITS1 part of the fungal rRNA. Primers targeting the vitellogenin gene of honeybees were used to normalize for possible differences in gut size. Primers used in this study are listed in Table 1. qPCR was performed on LightCycler 480 II (Roche, Mannheim, Germany) using 0.2 <sup>μ</sup>M of forward and reverse primers, 5× HOT FIREPol® EvaGreen qPCRMix Plus (Solis BioDyne, Tartu, Estonia) in 1× concentration, 5 μL gDNA, and adding nuclease-free water (VWR, Radnor, PA, USA) in a total volume of 20 μL. Nuclease-free water (VWR, Radnor, PA, USA) was used as a negative control. PCR conditions for 40 cycles were activated for 15 min at 95 ◦C, annealing for 30 s at 55 ◦C and 54 ◦C for 16S/ITS1 and vitellogenin, respectively, and elongation for 45 s and 30 s for 16S/ITS and vitellogenin, respectively at 72 ◦C.


**Table 1.** Primers applied in this work.

Relative copy numbers for 16S rRNA and ITS1 genes were calculated based on standard curves generated from Ampure® XP (Beckham coulter, Brea, CA, USA) purified PCR amplified targets, which were quantified using Qubit® dsDNA HS assay kit (Life technologies, Carlsbad, CA, USA), both methods performed according to the manufacturer's recommendations. Standard curves were run using 5× HOT FIREPol® Blend Mastermix Ready to Load (Solis BioDyne, Tartu, Estonia) in 1× concentration, with 0.2 μM of forward and reverse primers and 1 μL of gDNA in a total volume of 25 μL. PCR conditions for 30 cycles were as described above for qPCR with an additional 7 min final elongation step at 72 ◦C.

#### *2.5. Illumina Sequencing*

Illumina sequencing of the 16S rRNA gene and the ITS1 region was performed using the same primers as described above (table primers), and we used the same library preparation methodology as described in [22]. For the initial PCR conditions, we used the same conditions as for preparation of standard curves, but we only ran 25 cycles for 16S rRNA gene compared to 30 cycles for the ITS1. We indexed the ITS primers with 16 forward and 36 reverse indexes (table ITS index primer sequences). Pooling and preparation of the Illumina library was performed following the 16S metagenomic sequencing library preparation protocol (Illumina, San Diego, CA, USA). Quantification of pooled library was performed by ddPCR, BIO-RAD QX 200™ droplet reader (BioRad, Hercules, CA, USA) and diluted to 7 pM and sequenced on MiSeq using v3 reagents (Illumina, San Diego, CA, USA).

#### *2.6. Data Analysis*

qPCR copy numbers for all three genes were calculated using SciencePrimer copynumber calculator (Primer 2017), and 16S and ITS were normalized to vitellogenin copynumbers. Samples in each group that fell outside the 1/3 quartile of the median were removed as outliers.

Illumina fastq files were analyzed as described in [22]. In large, dereplicated, filtered, and generated OTUs in USEARCH v 8, then rarefactioned to 4000 sequences and calculated in QIIME for alpha and beta diversity. Identification of bacterial taxa was done using Greengenes V, and the UNITE (2019) database was used for blast searches for fungal OTUs.

All statistical analyses and plots were conducted in R (v. 3.5.1). Statistical significances were tested using the nonparametric Wilcoxon signed-rank test, with a significance threshold of *p* < 0.05. Dimension reduction of multidimensional data was done using nonmetric multidimensional scaling (NMDS). These analyses were performed using the Phyloseq package (v 1.24.2) in R.

#### **3. Results**

The relative abundances of bacterial and fungal taxa with respect to gut segment and season are presented in Figures S2 and S3, in addition to Blast-based taxonomic assignments of fungi (Table S1).

#### *3.1. Higher Bacterial Abundance than Fungal Abundance in All Gut Parts across the Entire Season*

Our dataset enabled us to compare bacterial and fungal abundance in different gut parts and track changes in abundance across the season (before, during, and after foraging). There was a higher bacterial abundance in all gut parts compared to fungal abundance, and this increase became more apparent towards the rear end of the gut (Figure 1). In the rectum, the bacterial load was magnitudes higher on average than the fungal load. Our results are consistent with previous findings of higher bacterial load in the hind gut (ileum and rectum) than in the midgut, as we found that the midgut had the lowest bacterial load across all months.

**Figure 1.** qPCR results showing median relative abundance of bacteria (16S) and fungi (ITS1) across months per gut part. Both 16S and ITS1 gene DNA are normalized against vitellogenin gene DNA. The vitellogenin gene copies were used as a proxy for the weight of the tissue. Error bars represent standard deviations. The asterisks represent statistical significance using the Wilcoxon signed-rank test.

Interestingly, the fungal abundance in the midgut and ileum was substantially influenced by season, as the copy-numbers fluctuated between neighboring months, creating a peak load in these two gut parts in June and July, respectively (Figure 1). The rectum displayed a different trend, with higher fungal load early in the season. The midgut harbored the lowest fungal load compared to the other gut parts (Figure 1).

#### *3.2. Observed Fungal Species Diversity Peaks during Foraging Months*

Numbers of observed fungal species displayed an increasing trend from April, which peaked in June (4×) and returned to baseline in August (Figure 2). This peak in fungal diversity was observed in midgut, ileum, and rectum, and the highest number was detected in the midgut. The crop showed a less clear peak in July, but the trend was present there as well. The observed fungal diversity due to these peaks was 4× higher than the number of observed bacterial species.

**Figure 2.** Observed species for bacteria (16S) and fungi (ITS) shown across months for different gut part. The missing data in the figure are due to lack of information for those datapoints. This particularly relates to fungi in crop, as the levels were very low.

For observed bacterial species diversity, the numbers were 2× higher in the crop than in the other gut parts, although a slight increase in diversity was seen in the late foraging months in midgut samples (Figure 2).

#### *3.3. Fungal Communities Cluster by Months and Bacterial Communities Cluster by Gut Parts*

NMDS plots revealed that the fungal community is highly influenced by season as the different fungal communities each month cluster apart, i.e., fungal communities were not much different between gut parts and the different gut part communities were similar in regard to different months (Figure 3A). This contrasts with the bacterial communities, which clustered by gut parts, i.e., bacterial communities showed highly gut part-specific communities, which were not strongly influenced by season (Figure 3B).

**Figure 3.** Nonmetric multidimensional scaling (NMDS) plots showing clustering of samples from different gut-parts based on (**A**) mycobiota and microbiota (**B**) composition. The stress values in all cases were < 0.2, suggesting a proper dimension reduction. The reason to present panels connected to gut segment for the mycobiota and season for the microbiota is to highlight the main differences visually.

#### **4. Discussion**

Our main finding was that the gut mycobiota changed with season, while the microbiota was mostly affected by the gut segment. This finding can have major implications for understanding the interaction between bacteria and fungi in honeybee health and disease.

Most diseases have a seasonal trend, but the underlying factors determining the seasonality remain unknown [23]. In temperate climates, honeybees stay inside the hive in winter until the weather is warm enough to fly out. This happens around May, and they continue to forage until late August. Our data suggest that the diversity of fungi species is elevated during foraging months, which was demonstrated for all gut parts. Higher species diversity in foragers has previously been shown by Yun et al. 2018 [13], suggesting that only a few fungi are endemic in honeybees and that a large amount of transient fungi might influence honeybees during summer months. Additionally, we saw that during early months there was more yeast-like fungi in midgut, ileum, and rectum. This could be due to bees being inside the hive for a long time in damp conditions. This is congruent with Yun et al. (2018), who found more yeast in nurse bees, which only stay inside the hive as well. The interplay between yeast, fungi, and bacteria could play a role in managing the yeast load in foraging bees, as we saw that the yeast load decreased as more environmental yeast and bacteria were present. Suggestions of this are found in germ-free mice that are highly susceptible to yeast infections [24], and in honeybees, yeast load seems to correlate with Nosema infection, which often peaks early and late in the year [11].

Mortality of honeybees also shows a clear seasonal trend [25]. Therefore, a factor that has not yet been considered in the mortality models is the major change in the fungal population associated with honeybees. Fungal diseases are commonly opportunistic and difficult to trace [26]. The major seasonal trends discovered here highlight the challenge in understanding fungal associated diseases.

In contrast to fungi, bacteria showed a clear association with gut segments through the season. This indicates that bacteria have a crucial role in maintaining honeybee health [27]. Dissection of the different gut parts, as done in this study, can reveal patterns not possible to detect using whole GI tracts. We found in our study that the midgut was more influenced by season than were ileum and rectum parts, which are usually what most studies investigate [14]. The hindgut comprises >90% of the total bacterial load in honeybees and thus will reflect the variation if the whole GI tract is used, and valuable information about which bacteria might be possible transient bacteria will be lost. There were some bacteria in our study that were mostly in the crop and midgut only in foraging months, which thus could be transient bacteria and not part of the endemic honeybee gut microbiota. This type of detailed (both gut part and different foraging months) information can shed light on previous suggestions that these bacteria are a part of the normal flora because they are commonly found in most bees. In addition to obligate pathogens, the absence of health-promoting bacteria can also lead to disease. Such diseases, however, would be much more difficult to detect, as they cannot be linked to specific bacteria [28]. Thus, diseases connected to lack of function could also have a potential role in explaining honeybee diseases, such as the colony collapse disorder (CCD). Thus, the fight towards the obligate pathogens could lead to simultaneous eradication of bacteria that have essential functions, such as vitamin production [29].

A limitation of our study, however, is that we did not consider the microbiota in the mouth part, which could have a substantial influence on both the micro- and mycobiota in the honeybee gastrointestinal tract [30]. Nor did we do specific measurements of diet [31]. Our study is further limited in that we only investigated one location and two hives. Further studies are therefore needed to generalize our findings. Our study also illustrates experimental issues that need to be considered in honeybee studies. Both season and gut segment had a major impact on the gut myco-/microbiota. Without taking the spatiotemporal information into account, misleading conclusions can be drawn related to the association of the honeybee gut myco-/microbiota with health and disease.

#### **5. Conclusions**

In conclusion, we have shown major differences connected to gut segment and seasonal associations of the honeybee gut myco-and microbiota. This knowledge can have major implications for honeybee health and disease.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2306-7 381/8/1/4/s1, Figure S1: Dissection scheme (a) whole gut, (b) the transition between the middle stomach and ileum, and (c) transition between the ileum and the rectum, Figure S2: Relative abundance of the 20 most abundant fungal OTUs across months per gut part, Figure S3: Relative abundance of the bacterial OTUs (with abundance > 1%) across months per gut part, Table S1: Taxonomic assignments of fungal OTUs by Blast.

**Author Contributions:** Conceptualization, J.L. and K.R.; methodology, Å.A., J.L., and L.H.; writing review and editing, all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data is contained within the article or supplementary material.

**Acknowledgments:** We thank NMBU for providing experimental and financial support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


*Article*
