Combining Exposomics with Genetic Disease Screening to Better 
Understand Honey Bee Health

Anthony Macherone^{1,2, 1}Agilent Technologies, Santa Clara, California, USA, 2The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA.

The western honey bee population has succumbed to a host of environmental stressors. Although many investigations offer insight into the reasons for the global health decline of honey bees, this complex combination of stressors has made it difficult to pinpoint key features of disease causality. This article describes a pilot study of hives in seven geographical locations in eastern Pennsylvania that combined 1) exposomic profiling using gas chromatography–quadrupole time‑of‑flight mass spectrometry (GC–QTOF-MS) with 2) semi-quantitative polymerase chain reaction (PCR) identification of Nosema ceranae (a common honey bee parasite) and 3) organoleptic identifiers of honey bee health. Chemical exposures associated with N. ceranae were screened against known honey bee biological pathways. These results offer insight into new methods that could be used to prospectively monitor honey bee health.

The phenotype - an individual’s observable health outcome - is a complex combination of genetic expression, external and endogenous chemical exposures, epigenetics, and stochastic events. The exposome represents cumulative exposures over an individuals’ lifetime. It includes but is not limited to, environmental pollutants, diet, drugs, individuals’ internal biochemistry, and psycho-social factors that elicit biological responses (for example, noise pollution, stress disorders, lifestyle choices). The exposome paradigm associates exposures with biological response pathways and provides a framework for the understanding of the causative factors of chronic human diseases - much in the same manner that genomics has been used to advance the fields of genetics and biology. Exposomics is the application of omics tools, such as metabolomics workflows using mass spectrometry (MS) and sophisticated bioinformatics software, to characterize and measure the exposome. Using exposomics to identify chemical markers of chronic disease causation has only recently been shown as practical, but it is already making a significant impact in human disease and epidemiological research.

Background

A variety of environmental stressors including viruses, parasites, pesticides, and loss of foraging habitat may be key factors in the global health decline of western honey bees (Apis mellifera). It has recently been reported that spraying the insecticide 1,2-dibromo-2,2-dichloroethyl dimethyl phosphate as a response to fight the spread of Zika virus in South Carolina killed millions of bees (1). As is evident by these and so many more undetermined threats to bee survival, the potential causes for the global decline in bee health are manifold. To address this highly complex and intertwined genetics‑exposure problem, we hypothesized that the use of the exposome paradigm coupled with semi‑quantitative polymerase chain reaction (PCR) may identify chemical profiles associated with disease and provide proof‑of‑principle for the development of tools to prospectively monitor bee health.

Evidence of environmental stressors like viruses, parasites, or pesticides on hive health in the form of a targetable biomarker or array of biomarkers may exist but have yet to be identified. To determine if chemical signatures (exposomes) associated with disease do in fact exist, a collaborative pilot study including researchers from Agilent Technologies (Santa Clara, California, USA), Haverford College (Haverford, Pennsylvania, USA), and Swarthmore College (Swarthmore, Pennsylvania) was conducted in the fall of 2015. The study was designed to characterize the exposome profiles of hives located in Philadelphia and surrounding counties in the eastern Pennsylvania region. Discovery-based (non-targeted) gas chromatography–mass spectrometry (GC–MS) methods were used to measure the exposomes present in bee extracts. Multiplex semi-quantitative PCR analysis of bee samples collected at the same time and place was performed to quantify Nosema ceranae. The exposome data was correlated with the N. ceranae genetic information using statistical techniques to determine exposure associations. Two primary hypotheses were defined:
1. The mean increase in parasite infestation as measured by PCR moves in a positive direction from zero and is associated with a decline in overall bee health.
2. There will be a measurable change in the exposome chemical profile associated with parasite infestation.

Experimental

Sample Collection: Honey bee samples were collected with permission of the beekeepers and land owners. Biological samples were collected from 29 unique hives in seven geographical locations. These consisted of one 30–50 bees sample from each hive for exposomics profiling, and one 30–50 bees sample from each hive for PCR analysis. Each random sample therefore represents the hive. Bees were euthanized by placing in a -80 oC freezer as is common practice.

Sample Extraction: Samples for mass spectrometry (MS) were extracted by pulverizing 3 g of bees in 27 mL of water–acetonitrile–acetic acid (44%:55%:1%). QuEChERS extraction (Agilent Technologies) was performed as follows: 6 g of magnesium sulphate and 1.5 g sodium acetate was added. Tubes were sealed, shaken, and centrifuged at 3000 rpm for 5 min. Two millilitres of the supernatant were applied to a conditioned solid-phase extraction (SPE) cartridge (Agilent Technologies). Analytes were eluted from the SPE cartridge with acetone/toluene (7:3 v/v).

Samples for PCR screening were prepared by adding 6 mL of RNase free water to 30 bees collected from each hive. The bees were macerated with a mortar and pestle to thoroughly homogenize the contents. An aliquot of 150 µL was used for multiplex PCR disease screening protocol.

PCR Analysis: Specific primers have been designed to identify and distinguish N. apis and N. ceranae based on a unique sequence found in a highly conserved ribosomal gene (2). Phenol:Chloroform DNA extraction was used along with multiplex PCR to amplify the target sequences. After PCR, a 1.1% gel electrophoresis was performed to semiquantify each Nosema species based on product band intensity using open-source ImageJ (3). Semiquantitative values were averaged over three replicate analyses and compared to house controls.

Mass Spectrometry: An Agilent 7200 GC-QTOF system was used for discovery chemical profiling of bee exposomes. The GC was configured with a 40 m × 0.25 mm, 0.25‑µm DB-5MS Duraguard column (Agilent J&W 122‑5532G). A 0.2 µL pulsed, splitless injection at 280 ^oC was made. The oven programme ranged from 80 ^oC to 310 ^oC and the transfer line temperature was 300 ^oC. The quadrupole time-of-flight mass spectrometer was operated in electron ionization mode with a source temperature of 275 ^oC. High resolution, accurate mass spectral information was collected at 5 Hz over a mass range of 50 Da to 800 Da. Each sample was analyzed in duplicate.

Data Analysis: To extract chemical features from the raw mass spectra, chromatographic deconvolution was performed using the Agilent MassHunter Unknowns Analysis software package. Briefly, chemical features were identified if the calculated signal-to-noise of a chromatographic peak was >3:1 and at least three extracted ion chromatograms could be aligned based on peak shape and retention time. Each identified feature was tested against a known commercial spectral library (NIST11) (4) for mass, number of ions, and ion ratio similarities. If a match score >0.7 was determined, the feature was annotated with chemical names, CAS number, and other pertinent information.

Statistical Analyses: The average of the replicate ratio for each sample, as determined in the PCR experiments, was associated with the corresponding mass spectral data files and used as a single condition for covariate analysis. The resulting multiomic dataset was statistically analyzed using Agilent MassProfiler Professional (MPP) bioinformatics software. Associations were made between chemical profiles of samples with N. ceranae ratio ≤0.10 compared to N. ceranae ratio >0.10. Chemicals identified as significant (p <0.005) via ANOVA test were screened against known apis mellifera biological pathways downloaded from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (5,6) using Agilent Pathway Architect tools.

Results and Discussion

General Information: Three of the 29 sample extracts used for PCR analysis of N. ceranae load elicited extremely high values (N. ceranae load >20) and were excluded from these analyses. These samples will be examined more comprehensively in follow-up studies.

PCR: N = 26 samples were analyzed in triplicate and the ratio of N. ceranae load to control were determined. The data were divided into two groups based on N. ceranae ratio. Summary statistics are presented in Table 1.

Mass Spectrometry: N = 26 samples were run in duplicate. Retention time variation was <0.15 min over the course of the analysis. A total of 2390 chemical features were identified by chromatographic deconvolution and were annotated using the NIST spectral library.

Statistical Analysis: The data were aligned in the bioinformatics software and allowed for retention time variation of 0.15 min and a mass extraction window of +/- 20 ppm. All raw data was Log2 transformed. The baseline of the data was determined by using the baseline to the median of all log transformed response values across all samples. Significance testing was performed using features of the software that firstly Filter by Flags wherein chemical entities were retained if they were present in at least 2 out of the 52 data files (26 samples × 2 technical replicates, each) followed by Filter by Frequency to retain chemical entities that appear in 100% of samples within the defined condition of N. ceranae load. One-way ANOVA then yielded 158 chemical entities with a p-value <0.005 and a fold-change (distance from median value) >2.

Discussion: The data identified multiple chemical biomarkers that may affect honey bee health. Many of the annotated chemicals were associated with known biological pathways of the western honey bee including phenylalanine metabolism, caffeine metabolism, fatty acid biosynthesis, and nicotinate and nicotinamide metabolism. One pathway in particular, associated with ubiquinone and other terpenoid-quinone biosynthesis, revealed endogenous chemicals significantly associated (p <0.005) with N. ceranae infestation that may perturb the pathway and downstream biological processes. In particular, a putative relationship between N. ceranae load and tocopherols where lower N. ceranae loads may align with a down-regulation of tocopherol and higher loads may align with increased tocopherol abundance in the samples was observed.

The initial results from the pilot study tend to confirm the hypotheses that as parasite infestation increases there is a measurable change in the exposome chemical profile. The identified chemicals can be associated with known biological pathways of the western honey bees, and there is putative evidence of a “dose-response” relationship where higher N. ceranae loads appear to correlate with a higher abundance of tocopherols in bee extracts. We acknowledge the small sample size of the pilot study may not elicit the appropriate effect size needed to have high confidence in the observations and view the findings as preliminary at best. We caution that these results need to be validated through more rigorous studies using much larger datasets. All experimental designs and results are presented for research use only and not for use in diagnostic procedures.

Future Work

In support of the putative results of the pilot study, Agilent Technologies has awarded a $50,000 Core Technologies-University Relations grant to Haverford College to explore the exposome of western honey bees. The award includes a partnership with Swarthmore College, also located in eastern Pennsylvania. Scheduled for July 2016 through June 2017, the project will implement a multiomics approach including state-of-the-art chemical measurement (mass spectrometry) and DNA characterization (PCR) techniques that combines genomic and chemical exposure information with sophisticated bioinformatics tools to interpret the resulting dataset. The study will include beehives situated within a natural diversity of settings such as urban, rural, farm, and orchard environments and will monitor the hives prospectively and longitudinally over time and generations. The experimental design intends to associate parasite infestation and viral infection, exposome profiles, and external exposure profiles as collected on silicone bands, with honey bee and hive health status. From this work the partners intend to establish chemical profiles associated with bees and hive health and develop a noninvasive tool to allow beekeepers to prospectively determine the health status of their colonies.

Acknowledgements

The author gratefully acknowledges the contributions of Robert Broadrup and Helen White of Haverford College and Christopher Mayack of Swarthmore College.

References

1. S. LaMotte, Zika spraying kills millions of honeybees. Retrieved from http://www.cnn.com/2016/09/01/health/zika-spraying-honeybees/index.html (2 September 2016).
2. R. Martín-Hernández, A. Meana, L. Prieto, A. Martínez Salvador, E. Garrido-Bailón, and M. Higes, Applied and Environmental Microbiology73(20), 6331–6338 (2007).
3. C.A. Schneider, W.S. Rasband, and K.W. Eliceiri, Nature methods9(7), 671–675 (2012).
4. The National Institute of Standards and Technology, NIST Standard Reference Database 1A. Retrieved 17 February 2013, from NIST.gov: http://www.nist.gov/srd/nist1a.cfm#2011.
5. M. Kanehisa, Y. Sato, M. Kawashima, M. Furumichi, and M. Tanabe, Nucleic Acids Res.44, D457–D462 (2016).
6. M. Kanehisa and S. Goto, Nucleic Acids Res.28, 27–30 (2000).

Anthony Macherone is a Senior Scientist with Agilent Technologies and a Visiting Scientist in the Department of Biological Chemistry at the Johns Hopkins University School of Medicine. He has demonstrated leadership in the global exposome initiative since 2013 and has spoken at multiple venues about the subject. To further espouse the exposome paradigm, he established and chairs an Exposomics Interest Group within the American Society for Mass Spectrometry (ASMS) and has chaired exposome workshops and oral platform sessions at the Hamner Institute and SETAC 2014 and 2015. He is currently engaged in a number of privately funded exposomics research collaborations with principle investigators at major universities including the University of California, Berkeley, Harvard Medical School, and Memorial Sloan Kettering Cancer Center where his expertise in liquid and gas phase chromatography and exposomics affords him unique insight into experimental design and analytical methodologies.