Conceived and designed the experiments: JJB CBH CHW RAC. Performed the experiments: JJB REJ SVD PEM MML JCG RAC. Analyzed the data: JJB CBH PEM PMW RAC. Contributed reagents/materials/analysis tools: SLB KWW RAC. Wrote the paper: JJB CBH CHW MFS AWZ REJ SVD PEM RAS TW DRF SLB RAC. Coordinated overall research team: JJB. Conducted statistical analyses: CBH. Coordinated Army (ECBC) research: CHW. Directed proteomics research: MFS. Organized Army research contribution: AWZ. Conducted proteomics analysis: REJ SVD. Analyzed results using bioinformatics: REJ SVD. Contributed Information Technology guidance: RAS. Set up and annotated proteomics data base: PMW. Consulted on all aspects of Iridoviruses: TW. Coordinated efforts with US Army: DRF. Guided epidemiology approach: EWS. Directed graduate student research: KWW.
Jerry J. Bromenshenk is CEO and a co-owner of Bee Alert Technology, Inc., Missoula, MT; a Montana Board of Regents' Approved Technology transfer company affiliated with The University of Montana. Colin Henderson is a co-owner and Research Vice President of Bee Alert, and Robert Seccomb is a co-owner and Chief Financial Officer of this company. Rabbih E. Jabbour is employed by Science Applications International Corporation, Abingdon, MD; Samir V. Deshpande is employed by Science Technology Corporation, Edgewood, MD; and Patrick E. McCubbin is employed by OptiMetrics, Inc., Abingdon, MD. These latter three companies provide contract research personnel to the US Army. For all four companies, financial support was in the form of salaries for contracted research. This does not alter the authors' adherence to all the PLoS ONE policies on sharing data and materials, as detailed online in our guide for authors.
In 2010 Colony Collapse Disorder (CCD), again devastated honey bee colonies in the USA, indicating that the problem is neither diminishing nor has it been resolved. Many CCD investigations, using sensitive genome-based methods, have found small RNA bee viruses and the microsporidia,
We used Mass spectrometry-based proteomics (MSP) to identify and quantify thousands of proteins from healthy and collapsing bee colonies. MSP revealed two unreported RNA viruses in North American honey bees, Varroa destructor-1 virus and Kakugo virus, and identified an invertebrate iridescent virus (IIV) (
These findings implicate co-infection by IIV and
Colony Collapse Disorder (CCD) continues to impact bee colonies in the USA in 2010 at levels seemingly equal to, or exceeding that of 2007, when this unusual syndrome first received worldwide press coverage
An unabated reappearance of CCD year on year demonstrates lack of progress toward solving the problem. Metagenomics initially identified Israeli acute paralysis virus (IAPV) as a potential marker or cause of CCD
Other studies pointed to a variety of additional possible markers of CCD. A transcriptome study used gene expression-based techniques and identified an abundance and variety of RNA fragments in the gut of bees from CCD colonies. The authors suggested that these fragments were possible markers of CCD. Whether these RNA fragments were from the host or RNA bee viruses is unknown
Given the diversity of potential microbes found in CCD colonies to date, and acknowledged environmental stresses faced by honey bees, some investigators have concluded that CCD is not a specific disease. It is rather a characteristic of colonies collapsing from an assortment of pathogens, physiological stress, or intoxications
In this study, we used mass spectrometry-based proteomics (MSP) and a rigorous sampling method in an attempt to identify potential markers of CCD. MSP offered an orthogonal and complementary approach
Our MSP analyses revealed the presence of two RNA viruses not previously reported in North American bee populations, as well as a highly significant and also unreported co-occurrence of strains of DNA invertebrate iridescent viruses (IIV) with a microsporidian of the genus
MSP analysis was used to survey microbes in bee samples from: (1) CCD colonies from the original event in 2006–2007 from widespread locations in eastern and western parts of the USA (2006–2007 CCD Colonies), (2) a collapsing colony in an observation hive fitted with a bi-directional flight counter and sampled through time as it failed in 2008 (2008 Observation Colony), (3) an independent collapse of bee colonies from CCD in Florida in 2009 (2009 Florida CCD), (4) packages of Australian honey bees delivered to the USA (2007 Australian Reference Group), (5) an isolated non-migratory beekeeping operation in Montana with no history of CCD (Montana Reference Group), and (6) dead bees recovered from inoculation feeding trials with
MSP analysis resulted in a database of more than 3,000 identifiable peptides, representing more than 900 different species of invertebrate-associated microbes. An extensive summary of detected peptides and microbes is presented in a recently completed technical report
Peptides were identified from nine of the approximately 20 known honey bee viruses in the initial sample set (
East Coast – West Coast Colonies, 2006 | Observation Colony, 2007 | Florida Colonies, 2008 | ||||||||
Collapsed n = 8 | Failing n = 10 | Strong n = 13 | Subsamples = 18 | n = 9 | ||||||
Pathogen | Frq | Frq | Frq | Frq | Frq | |||||
2 | 0.3 (0.46) | 5 | 1.5 (2.07) | 5 | 0.9 (1.28) | 13 | 1.3 (1.28) | 7 | 11.6 (12.4) | |
2 | 0.4 (0.74) | 6 | 1.4 (1.8) | 3 | 0.8 (1.54) | 4 | 0.3 (0.57) | 7 | 1.9 (1.5) | |
3 | 0.8 (1.4) | 1 | 0.2 (0.6) | 6 | 0.6 (0.8) | 4 | 0.6 (1.38) | 7 | 15.9 (20.1) | |
8 | 20.9 (28.2) | 10 | 38.0 (39.6) | 9 | 15.6 (22.4) | 18 | 16.1 (12.74) | 9 | 57.6 (23.6) | |
1 | 0.3 (0.7) | 4 | 1.4 (2.3) | 5 | 0.8 (1.3) | 11 | 0.9 (0.96) | 5 | 2.4 (2.8) | |
0 | 0 (0) | 0 | 0 (0) | 3 | 0.3 (.08) | 3 | 0.2 (0.55) | 2 | 0.3 (.04) | |
3 | 0.2 (3.2) | 6 | 1.9 (2.1) | 9 | 1.0 (0.9) | 1 | 1.0 (1.28) | 6 | 3.6 (5.0) | |
2 | 0.9 (1.6) | 4 | 0.9 (1.4) | 6 | 1.2 (2.3) | 11 | 1.3 (1.36) | 6 | 3.8 (7.0) | |
0 | 0 (0) | 1 | 0.2 (0.6) | 1 | 0.2 (0.6) | 4 | 0.4 (1.04) | 5 | 1.3 (1.6) | |
5 | 6.4 (9.1) | 9 | 11.4 (9.6) | 7 | 5.2 (7.7) | 18 | 8.7 (5.74) | 9 | 35.2 (15.3) | |
3 | 0.8 (1.4) | 3 | 0.7 (1.3) | 3 | 0.2 (0.4) | 11 | 1.0 (0.97) | 0 | 0 (0) |
Columns summarize thirty-one colonies from initial CCD study in 2006; eighteen subsamples taken from an observation colony monitored through its collapse from March through August 2007; and a third sample of nine colonies sampled during a CCD incident in Florida in 2008. A hyphen indicates that the value could not be calculated.
The recently-described Varroa destructor virus 1 (VDV-1)
IAPV did not occur frequently and was distributed approximately equally among strong and failing colonies (
The most prevalent viral peptides we detected were identified as invertebrate iridescent viruses (IIV), large double-stranded DNA viruses of the
IIV appeared with 100 percent frequency and at higher peptide counts in failing and collapsed colonies. IIV also occurred in nearly 75 percent of strong colonies, although invariably at lower concentrations. Numerous peptides for
Using those groupings we observed that one group of
Count-weighted occurrence data were subjected to stepwise discriminate function analysis to assess whether strong, failing, or collapsed colonies could be differentiated by specific patterns of pathogen occurrence. The isolated Montana apiary was included as a non-CCD reference group for this analysis. The colonies in this group served as an external control group that was complementary to the strong colonies within the CCD apiaries that served as internal controls.
Discriminate analysis indicated that only two pathogens, IIV and Deformed wing virus (DWV) were necessary for significant discrimination among different colony groups (
Function 1 explains 81 percent of discriminating variance and contrasts higher incidence of iridovirus (IIV),
Standardized Function Coefficients | |||||||||
Function | Eigen value | Var. % | Cum. % | Canonical Correlation | Chi-square | df | IIV-6 | DFW | |
0.68 | 80.6 | 80.6 | 0.64 | 22.8 | 6 | 0.001 | 1.17 | −0.65 | |
0.16 | 19.4 | 100.0 | 0.38 | 5.2 | 2 | 0.076 | 0.05 | 0.98 |
Structure Matrix | ||
Function | ||
Pathogen | 1 | 2 |
Invertebrate iridescent virus 6 | 0.83 | 0.55 |
0.68 | 0.60 | |
0.60 | 0.34 | |
Black queen cell virus | 0.59 | 0.53 |
Acute bee paralysis virus | 0.51 | 0.09 |
Israeli acute paralysis virus | −0.13 | −0.02 |
Deformed wing virus | −0.04 | 0.99 |
Sac brood virus | 0.15 | 0.60 |
Kashmir bee virus | 0.40 | 0.49 |
As expected, the Montana reference group was most distinct and significantly different from the strong condition colonies (
It is, however, likely that the few bees left in colonies at the final stages of collapse are those that are not infected, and thus would be expected to be similar to uninfected bees in strong colonies.
In the 2008 observation colony, as CCD progressed, flight activity exhibited several peaks and crashes until it declined by approximate geometric decay to extinction (
The peptides were detected in dead worker honey bee samples collected from a single collapsing observation hive at the University of Montana, Missoula. Forager flights are absolute counts per day, tabulated by an automated honey bee counter. Peptide counts are the summed counts by day of collection for all unique IIV
Both of these pathogens increased when the bee population decreased most sharply and remained at high levels throughout the remainder of the collapse. None of the other pathogens found in the colony showed a similar pattern. Polymerase chain reaction (PCR) analysis of the
The 2009 Florida CCD samples presented an independent opportunity to corroborate our findings. A group of nine colonies, in sets of three, identified as either strong, failing, or collapsed, were sampled and analyzed. We used the classification functions generated in our original discriminant function analysis from 2007 to classify these new samples as either strong, failing, collapsed, or out-group reference based on virus and
We detected 139 unique peptides in our west and east-coast data that were attributed to IIV-6 with high confidence (match to index ≥0.99). Later samples also indicated an IIV-6-like virus as the dominant virus in collapsing colonies (88 percent of IIV peptides). Furthermore, comparison of IIV peptides among all samples revealed moderate but significant correspondence between the original samples, the laboratory inoculation experiments, and the other field samples (
Sample | Florida Collapse | Inoculation Trial | Collapsing Colony | |
0.26 (0.00) | ||||
Sorensen's Index |
0.18 | |||
0.08 (0.21) | 0.11 (0.07) | |||
Sorensen's Index | 0.21 | 0.18 | ||
0.30 (0.00) | 0.22 (0.00) | 0.03 (0.66) | ||
Sorensen's Index | 0.58 | 0.20 | 0.17 |
Spearman's rank correlation (rho; n = 266).
Sorensen's index of similarity were calculated for each pairwise comparison. East-, West-CCD colonies sampled 2007–2008; Collapsing Observation Colony, 2008; Florida Collapse, 2009; Inoculation Trials, 2009–2010.
These procedures may have identified IIV-6 as the most likely source of peptides because this is the only fully sequenced genome from the genus
Our original wide area bee samples from 2007, the time series samples from the collapsing observation colony from 2008, and the reoccurrence of CCD reflected by Florida samples from 2009 were consistent in that IIV-6 peptides plus some IIV-3 peptides were the only IIV entities detected and were correlated with infections by
Cage trials of 1–3 day old newly-emerged bees demonstrated increased mortality in the experimental group fed both
Newly emerged, 1–3 day old, bees were used in all experiments. Figure represents the combined survival results for 4 biological replications (N = 30 bees in each group for each biological replicate). Bees that perished within 24 hours of inoculation were not included in the survival curve analyses. Deaths in control group were confirmed not to be pathogen related via mass spectroscopy analyses. Inoculum sizes and doses described in materials and
These results revealed mostly an absence of pathogens in the control bees, and the presence of peptides related to IIV-6 or
MS-based proteomics provided an unrestricted and unbiased approach for surveying pathogens in honey bee colonies. Our results detected a DNA virus and two RNA viruses that had not been previously reported in honey bees from the USA. The potential correlation of IIV with CCD may previously have gone unnoticed because these are large DNA viruses, not the small RNA viruses commonly considered to be the cause of most bee diseases. The correlation between IIV and
Interestingly, the presence or absence of IIV in a given honey bee colony may explain why in the USA
Other than the presence of the IIV, only the concurrent absence of Deformed wing virus, another RNA virus, was significant with respect to CCD. We recognize that the iflaviruses VDV-1 and Kakugo virus appear to be variants of Deformed wing virus, whereas the dicistroviruses Kashmir bee virus, Acute bee paralysis virus and IAPV are also closely related to one another
Virtually all of the bees from CCD colonies contained
Since inapparent non-lethal infections by IIVs are common
Large amounts of IIV in failing colonies is consistent with an infection that proliferates in bees but not necessarily to a degree that results in the iridescent coloration of infected bee tissues that is characteristic of IIV disease. The sustained high levels of IIV and
Because of their virulence, IIVs have been investigated as candidates for use as biopesticides
Patent IIV infections are almost invariably lethal but inapparent or covert infections may be common
Other IIVs, such as IIV-6, naturally infect various species of Lepidoptera and commercial colonies of Orthoptera. There is evidence that hymenopteran endoparasitoids can become infected if they develop in an infected caterpillar
There is one known iridescent virus in bees. IIV-24, originally isolated from the Asiatic honey bee
Based on the sequence data generated from MSP, the IIV identified appears to be closely related to IIV-6, possibly because this is the only IIV in the
There is little information about IIVs in bees, although there are historical reports associating IIVs with severe bee losses in India
Transmission of IIV-24 is suspected to occur via eggs, feces, or glandular secretions in food
In addition, an iridescent virus has also been associated with mites, which may act as vectors, and has been implicated in bee losses in the United States. While investigating unusually high losses of bees in the northeastern United States, Camazine and Liu
One or more species of external mites were suspected of being carriers of the virus in Indian bees
These historical findings of IIV, mites, and
The suspected source of
It also implies that if Kashmir bee virus has been in North America for more than twenty years, so might IIV and
Our own work, described here, provides multiple lines of correlative evidence from MSP analysis that associate IIVs and
The fact that IIV-6 inoculated bees experienced increased mortality in the presence of
Moreover, we used a fairly low dose of IIV-6 and
In our studies, we applied six independent scenarios to the assessment of potential causes or markers of CCD and got the same answer, giving us confidence in the results, since this inference approach is approximately analogous to applying the same technique to six different assessments
We anticipate that there also may be questions as to why IIV was detected in our study, but has not been found in any current published research on CCD. And, if these viruses were present, why weren't they seen in infected tissues of the European honey bee,
First, iridescent viruses have been seen before in
The large number of IIV proteins that we identified, 139 in all, represent a significant fraction of the total IIV proteome. The recently published genome for IIV-6
We conclude that the IIV/
We collected sample sets of adult worker honey bees from several areas and years: (1–2) Two initial sample sets of adult honey bees from CCD colonies were obtained in 2006–2007 from twelve beekeeping operations from western, northeastern, and southeastern regions of the USA, (3) Samples from packages of imported Australian bees provided a non-CCD 2007 reference, (4) bees sampled in 2008 from a large, non-migratory beekeeping operation in northwestern Montana with no history of CCD provided a second reference set, (5) bee samples obtained in 2009 from a Florida apiary when 500 colonies suddenly collapsed constituted an independent CCD sample set by location and year.
In each apiary investigated and sampled for this study, based on visible signs of CCD as described by the CCD Working Group
Typically, the largest colony populations had 10–14 frames of adult bees or more, and two or more frames of brood. The collapsed colonies had less than a frame of bees, often no more than the queen and a fist-sized cluster of very young bees. Failing colonies were defined as those that had no more than half the number of bees as the most populous colonies. These colonies often had an excess of bees, and had had far more bees just days or a few weeks before the samples were taken, according to the beekeeper accounts.
Additional reference bees were obtained from packages shipped from Australia to the USA and from the most populous to the weakest colonies from apiaries of a large, commercial beekeeping operation in Montana that is geographically isolated and has no history of CCD. In this case, the weakest bee populations were only about 20 percent smaller than the largest bee populations.
All of the CCD operations were large, migratory beekeeping businesses that transported bees across state borders and rented colonies for pollination of almonds in California. The migratory colonies sampled in 2006 and 2007 represented two different migratory routes, one from the east coast to California, the other from North Dakota to California. In addition, when in California, the east coast and the mid-western colonies were separated by approximately 400 kilometers, so that there was no overlap of either the apiary locations or highways of these two different migratory routes.
Bees were shaken directly into new, clean one quart Ziploc® or one liter Whirl-Pac ® bags. The bags were sealed, placed in a cooler with frozen gel packs, and shipped by overnight express to the U.S. Army Edgewood Chemical and Biological Center (ECBC) laboratory. Bees were often alive when received and were analyzed immediately. In a few cases, bee samples were frozen and stored in a −80°C freezer until analyzed.
Following the same sampling methods, we sampled a repeat of CCD in Florida, where 500 honey bee colonies started from packages in October of 2008, collapsed in January of 2009. As mentioned before, the beekeeper who owned the colonies had experienced CCD in 2006–2007, and had been one of the original beekeeping operations sampled by members of the CCD Working Group. As in 2006–2007, the colonies suddenly collapsed, demonstrating the characteristic signs of CCD
We also observed the progression of CCD in a collapsing colony in an observation hive, taking 18 bee samples of approximately 10–60 bees per sample interval, over a three month period, ending when only a queen and four workers remained.
In the spring of 2008, we lost more than more than 50 of our research colonies to CCD. We took the frames, queen, and the small, surviving population of young bees from one of these collapsed colonies, put them in a five-frame observation hive, and fed them sucrose syrup. This colony soon produced a second queen, and both queens co-existed in the same colony, one on each side of the glass hive, together producing a rapidly increasing combined population of bees. The forager bees had access to both syrup and to abundant food resources from the University of Montana campus, UM's arboretum, and surrounding residential flower gardens. By mid-summer, this bee colony collapsed for the second time. We then began to sample bees from those remaining in the hive.
The number of bees sampled at each time point varied with the health of the colony. We attempted to collect at least 60 bees per sample interval, until the end, when too few bees remained to take even ten bees, which is the minimum sample size used for proteomics analysis. We also recorded forager flight activity and forager losses using a bi-directional digital bee-counter mounted at the entrance of the observation colony.
We are working on isolating the IIV that infects CCD bees for use in inoculations to perform Koch's Postulates. For our preliminary experiments, and because of the high degree of similarity between the CCD-related IIV and IIV-6, based on the MSP data, we elected to conduct inoculation trials using IIV-6 and
Bees were obtained from non-CCD colonies with no detectable levels of
Following emergence from brood frames in an incubator, 1–3 day old bees were placed into sterile cardboard cups in a plant growth chamber with controlled temperature (28°C, relative humidity, and light). Using a 10 µl pipette, each bee was inoculated by feeding it a total of 2 µl in sugar water containing one of four treatments. Only bees that ingested the entire inoculum were used.
The following treatments were given: 1)
Thirty bees were inoculated in each group and the experiment was repeated four separate times for a total of 120 bees in each group. Bees that perished 24 hours after the inoculation were not included in the statistical analysis. Bees were then monitored daily for a period of 14 days. Dead bees were removed immediately upon discovery and frozen at −80°C.
Dead bees from the inoculation experiments were analyzed by PCR and proteomics to detect and confirm infections by
Bee samples were homogenized in 100 mM of ammonium acetate buffer using a tissue homogenizer. The supernatant was filtered to remove large particulates, followed by ultrafiltration at 300 kDa. All filtered bee samples were lysed using an ultra-sonication probe at settings of 20 seconds pulse-ON, 5 seconds pulse-OFF, and 25 percent amplitude for 5 minutes duration. To verify cells were appropriately disrupted, a small portion of lysates was subjected to 1-D gel analysis. The lysates were centrifuged at 14,100× g for 30 minutes to remove all cellular debris. Supernatant was then added to a Microcon YM-3 filter unit (Millipore; USA) and centrifuged at 14,100× g for 30 minutes. Effluent was discarded and the filtrate were denatured by adding 8 M urea and 3µg/µl dithiothreitol (DTT) and incubated for two hours in an orbital shaker set to 50°C and 60 rpm.
A 10 µL volume of 100 percent acetonitrile (ACN) was added to tubes and allowed to sit at room temperature for 5 minutes. Tubes were washed using 100 mM ABC solution and then spun down at 14,100×
A protein database was constructed in a FASTA format using the annotated bacterial and viral proteome sequences derived from all fully sequenced chromosomes of bacteria and viruses, including their sequenced plasmids (as of September 2008)
Each database protein sequence was supplemented with information about a source organism and a genomic position of the respective ORF embedded into a header line. The database of bacterial proteomes was constructed by translating putative protein-coding genes and consists of tens of millions of amino acid sequences of potential tryptic peptides obtained by the
The experimental MS/MS spectral data of bacterial peptides were searched using the SEQUEST® (Thermofisher Scientific, USA) algorithm against a constructed proteome database of microorganisms. SEQUEST thresholds for searching the product ion mass spectra of peptides were Xcorr, deltaCn, Sp, RSp and deltaMpep. These parameters provided a uniform matching score of all candidate peptides
This validating and verification approach uses an expectation-maximization algorithm as described by the Keller
Peptide sequences with probability score of 95 percent and higher were retained in the dataset and used to generate a binary matrix of sequence-to-microbe assignments. The binary matrix assignment was populated by matching the peptides with corresponding proteins in the database and assigned a score of 0 (no-match) or 1 (match). The column in the binary matrix represented the proteome of a given microbe and each row represented a tryptic peptide sequence from the LC-MS/MS analysis.
Bee samples were identified with the virus/bacterium/fungi proteome based on the number of unique peptides that remained after removal of degenerate peptides from the binary matrix. This approach was successfully used for the double-blind characterization of non-genome-sequenced bacteria by mass-spectrometry-based protoemics
Proteomics identified peptides described from a variety of bee viruses, as well nine species of
It is almost certain that the diversity of
We performed forward, stepwise discriminant analysis on square-root or log transformed pathogen counts for
Use of peptide counts as a weighting factor stems from the observation that as total pathogen titer in a sample increases, the number of different peptides that can be identified by proteomics increases in a predictable manner
Four colony groups were discriminated: strong, failing, collapsed, and the Montana reference group. Selection method for variable entry was largest univariate
The analysis was completed after two steps including only IIV-6 and DFW as significant discriminating variables (Final Wilks' lambda = 0.679;
The authors thank Dave Wick of BVS, Inc. for introducing us to the Army team and their advanced analytical technologies and the U.S. beekeepers who allowed us to sample their bees. UM and Bee Alert Technology, Inc. researchers put a tremendous amount of time and energy into the project. Stacy Potter annotated the proteomics database; while Scott Debnam contacted beekeepers and traveled across the U.S. inspecting and collecting bee samples from CCD colonies. Sarah Red-Laird conducted final edits of the manuscript.