Testing the Potential for Cut Flower Pollen to Improve Bumble Bee Health

Final report for LNE21-424R

Project Type: Research Only
Funds awarded in 2021: $98,539.00
Projected End Date: 11/30/2024
Grant Recipient: University of Massachusetts Amherst
Region: Northeast
State: Massachusetts
Project Leader:
Dr. Lynn Adler
University of Massachusetts Amherst
Expand All

Project Information

Summary:

Farmers, landscapers, conservation land managers, government and the public are all invested in providing floral resources to promote pollinator abundance, diversity and health. Pollen and nectar from certain flowers can reduce pollinator infection by detrimental pathogens, but this has been examined for only a few plant species and almost entirely in laboratory settings. The Adler lab discovered that sunflower pollen dramatically and consistently reduced a common detrimental gut pathogen of bumble bees in the lab, and that farms with more sunflower had bumble bees with reduced infections. If pollen from other flowers in the sunflower family (Asteraceae) similarly reduces infections in bumble bees, promoting the planting of Asteraceae cut flowers could be an avenue for farmers, landscapers, horticultural growers and land managers to improve bee health while providing a source of income. We tested this hypothesis using a two-pronged approach. We worked with growers at three ‘cut flower’ farms (at least 3/4 acre of cut flowers from our target species) and compared with 3 ‘control’ farms (1/10 acre cut flowers or less). We deployed honey bee hives with pollen traps at the cut flower farms to collect pollen that we sorted, identified to species using DNA barcoding, and then tested in the laboratory to determine effects of sunflower-family pollen on pathogen infection. On the farms, we surveyed three times to assess abundance of flowers from Asteraceae species, observe bumble bee preference for Asteraceae cut flowers and bordering wild flowers, and sample bumble bees to assess pathogen infections, with the goal of determining whether cut flower plantings on farms reduce bee pathogens, as we found for sunflower. All farms had honey bee hives, to avoid confounding cut flower treatment with pathogen transmission. We consulted with farmers on study design, cut flower selection and potential obstacles during proposal preparation, and engaged with farmers who host the cut flower and control sites.

The primary bumble bee we collected was Bombus impatiens, and so we focused our analyses on that species. Across all six farms, 46% of the 532 B. impatiens workers collected were positive for C. bombi infection. However, none of the floral abundance variables significantly explained the variation observed in C. bombi infection intensity, indicating no significant relationship between cut or bordering wild Asteraceae floral abundance and C. bombi infection. Silphium and Eutrochium had the highest pollinator visits per observation period, indicating that both cut flowers tended to be less attractive than wild Asteraceae bordering the farms. Although we were able to sort the honey bee-collected pollen to concentrate Asteraceae samples, nearly all of it was from wild species rather than cut flowers. Furthermore, sorting by color still resulted in pollens that were mixes of several plant species, primarily composed of different proportions of Ericameria, Ambrosia, Silphium and Eutrochium. In bioassays, four out of five Asteraceae pollen diets significantly reduced C. bombi infection compared to the negative control, buckwheat pollen. While this study adds new species to the list of Asteraceae whose pollen reduced C. bombi infections in the laboratory, under realistic field-conditions the available Asteraceae pollen was not abundant or not foraged enough to reduce infections in local bee populations. Thus, although previous work demonstrated that sunflowers on farms can decrease B. impatiens infection and increase reproduction, within the limited scope of the current study (six farms, compared to 20 in the sunflower study) we do not see a similar benefit of growing other cut flowers in the Asteraceae.

Project Objective:

Our goal is to evaluate whether pollen from cut flowers in the sunflower family (Asteraceae) can reduce disease in bumble bees, based on our discovery of this effect in sunflower pollen. We will collect a diversity of pollen from farms growing Asteraceae cut flowers to test its effects on bee pathogens in lab trials, and will collect bumble bees at ‘cut flower’ and ‘control’ farms to assess impacts of cut flowers on wild bees. Our work is novel for innovating sustainable, low-input approaches to manage pollinator health by growing specific crops whose pollen could reduce pollinator pathogens while providing income.

Introduction:

Pollinator decline is attributed to many factors, including pathogens and lack of food resources. Such declines can have consequences for food security, with lack of pollination limiting yield in several major US crops. Many farms are planting flowering strips to promote healthy and diverse pollinator communities and increase pollination. However, these floral resources could increase or decrease pathogen infection, depending on their relative roles in facilitating disease spread versus providing medicinal benefits.

In western Massachusetts many small and large farms grow cut flowers as a primary crop, a small part of a diverse planting plan, or to host weddings or agro-tourism. Many of these cut flowers are in the sunflower family (Asteraceae). My lab discovered that pollen from many sunflower varieties as well as goldenrod (also in the Asteraceae), all dramatically reduced common eastern bumble bee (Bombus impatiens) infection by Crithidia bombi. This common gut pathogen ('Crithidia' hereafter) can reduce bumble bee queen colony-founding success and colony size. Because many Asteraceae flower in late summer when daughter queens emerge and forage before overwintering, impacts on queen health are particularly important.  

Our research fills two major gaps by assessing whether (a) pollen from several cut flower species reduces pollinator infection in the lab and (b) floral plantings can reduce bee infection on farms. This work may have broad implications for the use of cut flowers in agroecoystems, the landscape industry, and designing pollinator habitat.

If we found evidence that pollen from Asteraceae cut flowers reduces bee pathogens in the lab and on farms, this could lead to changes in practice for several stakeholder groups. In our previous farm research, farms growing cut flowers averaged 29-35% of blooms from Asteraceae cut flower species (other than sunflower), indicating strong interest in our species but also room to increase investment. I requested input and assistance from Tawny Simisky, Sue Scheufele and Jason Lanier (all UMass Extension) to develop an informational survey to assess interest in our novel research approach. The survey was shared by these UMass Extension professionals to listservs reaching vegetable farmers, landscapers, turf managers, arborists, nurseries, garden centers, and greenhouse producers. We received 2384 responses; over 95% (2260 respondents) described themselves as ‘highly’ or ‘very highly’ interested in using flowering plants to improve pollinator health in their business/land ownership. Responses were resoundingly positive in terms of both number of individuals who participated and their overwhelming enthusiasm for the applications of our research, and clearly demonstrate strong enthusiasm for our novel approach.

 

Cooperators

Click linked name(s) to expand/collapse or show everyone's info

Research

Hypothesis:

We hypothesize that (1) pollen from Asteraceae flowers will reduce pathogen infection in the common eastern bumble bee compared to mixed wildflower pollen in laboratory trials, and (2) farms with more Asteraceae cut flowers will have bumble bees with lower pathogen infection than bees in control farms. Because there is currently almost no research on how pollen from specific plant species affects bee pathogens in the laboratory or field beyond our work with sunflower pollen, this research may provide novel and important information about cut flowers that could be used as low-impact, non-chemical methods for stakeholders to reduce bee disease.

Materials and methods:

Bumble bee collected in a vial at a farm.

Researchers measuring flower abundance at a cut flower farm.

On-Farm Study

We conducted the on-farm study during summer 2021. We compared pathogen infection in wild-caught bumble bees at farms with high and low Asteraceae cut flower acreage. We compared farms with at least ¾ acre of Asteraceae cut flowers (“cut flower farms” hereafter) compared to control farms with no more than 1/10 acre of Asteraceae cut flowers. ‘Cut flowers’ included any Asteraceae species that produced pollen; these comprised on average 29-35% of cut flowers in our 2019 farm surveys. We did not include sunflowers, since cut-flower varieties do not make pollen and we have already extensively tested sunflower pollen. Cut flower farms included Laurenitis, Red Fire, and LaSalle farms; these farms range from 12.5-77.5 acres and grew between 0.75-3.5 acres of our target cut flowers. Control farms were Simple Gifts, Red Fire (a separate field 3km away from the cut flower site), and Golonka Farm, which range from 14-29.2 acres and grew 0.1 acres or less of our target cut flowers. All farms also grow vegetable crops, except for LaSalle (only cut flowers). Only one farm (Simple Gifts) is organic, but all attempt to minimize pesticide use. All farm sites are separated by at least 1 km to ensure independence. We offered compensation to farmers that allocated resources towards growing cut flowers for our research, and farmers were welcome to sell their cut flowers according to their usual practices.

We visited each farm three times during summer 2021, exceeding the proposal goal of visiting each farm twice. Whenever possible we collected from a control and cut flower farm on the same day to avoid confounding treatment with date, and alternated morning vs. afternoon sampling for control vs. cut flower farms.

At each visit we collected bumble bees to sample for pathogens. We expanded the scope of the original study by collecting all the common bumble bees at the farms, including Bombus vagans, bimaculatus and grisecollis, as well as B. impatiens as originally proposed.  Bees were individually stored in vials in a cooler, taken to the lab, and dissected to assess Crithidia infection following our standard protocols, and counted Crithidia cells in a 0.02 µL gut sample of each bee using a microscope at 400x. We ultimately collected data from 687 bumble bees, of which 600 were B. impatiens and the rest were relatively evenly spread between the other three species. Furthermore, some of the bees turned out to be male, leaving 532 B. impatiens workers. Due to the low sample sizes for other species, we only conducted statistical analyses for Bombus impatiens.

We quantified the abundance of each cut flower species at each farm, plus all sunflower and wild Asteraceae such as goldenrod (Solidago) and Joe-Pye weed (Eutrochium), to isolate their impacts on bee pathogens. In addition, because we saw little relationship between cut flower abundance and infection at the first visit, for the second and third visits we also assessed bumble bee visitation to cut flowers, in case the lack of relationship was due to failure to visit these species. 

We spent fall 2021 measuring the marginal cell length from the field-collected bees as an estimate of body size, which often correlates with infection intensity, and entering data; the floral resource data set was large and this took quite a bit of time.  Data analysis took place during spring and fall of 2022. Because farms ended up varying quite a bit in the amount of cut flowers, we used cut flower abundance (count of floral heads and inflorescences) per visit as a predictor, rather than the cut vs. control farm categories. We used generalized linear models to test the effect of cut flower abundance on infection prevalence (infected/uninfected) and infection intensity in infected bees (cells/0.02 µL). Because farms were sampled three times, farm was included in the model as a random factor, along with covariates of Julian date, bee size (estimated from wing marginal cell length) and abundance of sunflower, goldenrod and Joe-Pye weed. We calculated VIFs and confirmed low collinearity between variables, and used AIC to select the final best-fit model for analysis.

Laboratory Bioassay

We also tested the ability of pollen from Asteraceae cut flowers to reduce Crithidia infection in the common eastern bumble bee, Bombus impatiens, compared to control wildflower pollen in the laboratory. We collected pollen from the three cut flower farms using pollen traps in honey bee hives (honey bees are generalist foragers, and traps are not developed for bumble bees). All control farms also had apiaries, so we did not confound honey bee presence with farm treatments. Pollen was collected twice weekly and stored at -20oC. We spent fall 2021 doing a preliminary sorting to ascertain which colors of pollen were Asteraceae (had spines) in each collection. We planned to ship the pollen to my colleague Dr. Rebecca Irwin to sort by color using a new pollen sorting machine she ordered just as the COVID pandemic began. Due to COVID-related and other logistical delays, the machine to sort the pollen automatically by color was not operational until spring of 2024. After sorting was complete, we sent 18 samples for barcoding in July 2024 to identify plant species (1-3 replicates from the farm sites and collection dates with the highest representation of Asteraceae) at the Penn State University Honey and Pollen Diagnostics Lab. Sorting by color still resulted in pollens that were mixes of several plant species, primarily composed of different proportions of Ericameria, Ambrosia, Silphium and Eutrochium.

We compared the effect of pollen from five sorted samples and two control species. Pollen types samples were given an identifier A-E and were composed of different relative proportions of Asteraceae cut and border flower genera.  Sunflower pollen (Helianthus annuus) and buckwheat pollen (Fagopyrum esculentum) were used as the positive and negative controls, respectively. To prepare diets, pollen pellets were ground using a coffee grinder then mixed with distilled water until a smooth consistency was reached. All pollen diets were produced in bulk and saved in plastic containers in -7°C freezer. 

Crithidia bombi cells (origin, Hadley, Massachusetts: 42.363911 N, −72.567747 W) were maintained in commercial B. impatiens colonies and used for infection trails. Inoculum was prepared the day of each experimental trial by dissecting up to 8 bees from infected colonies and using our standard lab protocols to create a final inoculum of 600 cells/uL and 15% sucrose solution. Bombus impatiens workers from three uninfected colonies were transferred to individual vials, starved for 2-3 hours and then presented a 15 uL droplet of inoculum (~9,000 C. bombi cells) and observed until the drop was consumed. Only bees that consumed the full droplet were included in experimental trials. 

The bioassay experiments were conducted across 12 trials from April 19 to May 20, 2024 using 486 B. impatiens workers from five commercially reared colonies. Of the 486 workers, 97 died and 4 escaped; final treatment sample sizes varied from n = 53-61, except for pollen diet treatment A, which had less total pollen, and was used in 6 trials for a sample size of n = 37. For each experimental trial, 36-48 bees were inoculated and randomly assigned to one of seven diet treatments. Each trial started with an even number of bees per colony per treatment (two or three bees). After inoculation, bees were placed individually in containers (clear plastic 8 oz deli cup with perforated lids and mesh bottoms) with access to 30% sucrose and pollen treatment ad libitum and housed in an incubator in darkness at 27°C and 55–60% humidity. Fresh pollen patties and sucrose were administered every 48 hours. Survival was also recorded every 48 hours. After seven days, live bees were dissected, and C. bombi cells were quantified according to our standard protocols to determine infection intensity after pollen diet treatments. 

Pollen starvation reduces C. bombi infection and increases mortality of infected bumble bee workers. Therefore, pollen consumption was measured for 48 hours during the experiment for each bee. Pollen diets (0.15 mg ± 0.036) were administered in microcentrifuge tube caps that were weighed without and then with fresh pollen before delivering to an experimental bee. After 48 hours the cap was removed, placed in drying oven at 45°C and then reweighed. Data analysis took place in fall 2024. We assessed the effect of pollen diet treatments on the intensity of C. bombi infection (cells per 0.02 µL) in B. impatiens workers using a zero-inflated negative binomial mixed model. We used a mixed effects Cox Proportional Hazards model to determine the difference in survival. For the full model, fixed effects included pollen diet treatment, amount of pollen consumed, their interaction, colony identity, and marginal cell length to estimate bee size, and the date of inoculation as a random effect. Pairwise comparisons between treatments were performed using Tukey’s honestly significant difference test.

Research results and discussion:

For the farm study, we assessed the influence of Asteaceae cut flower abundance, border flower abundance and total flower abundance on the prevalence and intensity of C. bombi infection (cell count via visual inspection) in sampled B. impatiens using a zero-inflated binomial mixed model, including effects of farm, date and bee size (wing marginal cell length). Across all six farms, 46% of the B. impatiens workers collected were positive for C. bombi infection. None of the floral abundance variables significantly explained variation in C. bombi cell count. Larger bees had higher infection levels (χ2 = 6.2, p = 0.0128), and infection decreased over time (χ2 = 4.8, p = 0.028).

We also analyzed bumblebee visitation per observation period to 12 Asteraceae genera in foraging Bombus spp. using a negative binomial mixed model.  Visits were pooled across bumblebee species and sex because we were not always able to identify species and sexes on the wing, although we mostly observed B. impatiens workers (approximately 85% of total visits). Bumblebee visits to flowers differed with plant genus (χ2 = 82.197, p < 0.0001). Silphium, and Eutrochium sp. had the highest estimated visits per observation period.

For the laboratory bioassays, we assessed the effect of pollen diet treatments on the intensity of C. bombi infection (cells per 0.02 µL) in B. impatiens workers using a zero-inflated negative binomial mixed model. We assessed the effect of pollen diet treatment on bee survival using a mixed effects Cox Proportional Hazards model. For both models we also included trial date, bee size, bee parent colony and pollen consumption as predictors.

Compared to the negative control treatment (buckwheat pollen), all pollen diets significantly reduced C. bombi infection intensity, except pollen diet B. Parent and marginal cell length, a proxy for bee size, affected C. bombi infection intensity. Pollen consumption (χ2 = 31.9, df = 1, p < 0.0001) and source colony (χ2 =39.1, df = 4, p < 0.0001) affected bee survival. 

Floral products from a diversity of plants mediate disease in bee pollinator communities. Discovering antimicrobial floral traits shared between plant species and across plant families is valuable information to integrate while planning future pollinator habitats. We assessed effects of a wide range of Asteraceae pollen diets from cultivated cut and wildflowers on a common bee gut parasite, Crithidia bombi. Under laboratory settings, pollen from various Asteraceae plant species significantly reduced C. bombi infection in bumblebees. This effect has been described in sunflowers and other Asteraceae wild flowers in previous research, and our study expands the anti-parasitic effect of Asteraceae pollen to more species within the family, suggesting a broad effect of Asteraceae pollen in reducing C. bombi infection. 

While the laboratory results are suggestive, we did not find a significant relationship between size of commercial plantings of cut Asteraceae flowers nor abundance of wildflowers bordering the farm field margins (border flowers) on C. bombi infection in foraging bumblebees. All farm sites had varying levels of cut flower abundance but only one farm planted a full acre of flowers, comparable to the size of local cultivated sunflower plantations. While one acre of sunflower commercial plantings was correlated with a reduction of C. bombi infection in foraging Bombus impatiens, we did not find this effect on the farm that planted the most cultivated cut flowers. At the field level, growing enough Asteraceae cut flowers to reduce bee infection may not be feasible for farmers, especially since small farms characteristically grow a diversity of crops. Creating pollinator habitat that directly financially benefits farmers via planting commercial cut flowers is an interesting strategy to incentivize supporting bee health, but we found that it is not as effective at reducing infection as directly feeding bumblebees an Asteraceae pollen diet. While a considerable portion of the agricultural industry in Massachusetts, cut flowers most likely do not occupy enough land and therefore will not be able to produce sufficient pollen to have a significant effect on infection levels in the local bee community.

Research conclusions:

Our previous research found that sunflower abundance on farms was related to lower C. bombi infection and higher reproduction in commercial B. impatiens on the farms, but in the current study we did not find a relationship between the abundance of other cut flowers in the sunflower family and bumble bee infection. Although our sample size of farms was small (six farms), one farm grow over an acre of cut flowers and did not have infection levels noticeably lower than other farms. This suggests that although cut flowers can provide food resources, growing cut flowers is not an effective strategy for New England growers to manage bumble bee pathogen infection.

We furthermore found that honey bees collected very little flower from cut flowers, although they did collect abundant pollen from wild Asteraceae plants growing on the farm margin, particularly Ericameria (rabbitbrush), Ambrosia (ragweed), Silphium (cup plant) and Eutrochium (Joe-Pye weed). In the laboratory, four of five of the pollens tested significantly reduced C. bombi infection in common eastern bumble bee workers compared to buckwheat, the negative control pollen. These pollens were as effective in reducing infection as sunflower pollen, suggesting widespread effects of Asteraceae pollen in reducing C. bombi infection. Although we found no relationship between wild Asteraceae and infection of wild bees, our study was not initially designed to test this and sample sizes were small. While our results are not robust enough to make strong recommendations, we suggest allowing bordering wild plants to flower when possible to support pollinator health.

Participation Summary
6 Farmers participating in research

Education & Outreach Activities and Participation Summary

Educational activities:

2 Webinars / talks / presentations
1 Workshop field days
2 Other educational activities: Training graduate undergraduate students in farm and laboratory research

Participation Summary:

130 Farmers participated
10 Number of agricultural educator or service providers reached through education and outreach activities
Outreach description:

Because we found no relationship between cut flower abundance and bumble bee disease, we have not broadly disseminated our results. However, we did participate in the annual South Deerfield Farm day in 2021, which I estimate was attended by 30 farmers and 10 extension educators, and I discussed the hypotheses and goals of this project. Although I participated again in subsequent summers, our presentation focused on other research. 

I trained 3 undergraduates and 2 graduate students who conducted the field work with me in summer 2021; an additional undergraduate researcher helped sort all the pollen in fall 2021, two more helped hand-sort in fall 2023 and conduct bioassays in spring 2024. This may not be what is meant by an 'educational activity,' but I strongly believe in the power of these experiences to train a new generation of potential agricultural researchers.

I ran our first Advisory Board meeting in April 2022 to describe our project context and get feedback on how to make it most effective. It was attended by two extension educators and three farmers (the members of the Advisory Board).  Our second meeting happened in March 2023 to report on results so far; I reported on final results via email in March 2025 at project completion. We also conducted our research on 6 farm fields (run by 5 farms) and discussed the premise with all of the participating farmers. 

I am gave a research talk for UVM extension on Feb 17 2022, and discussed this project amongst others. In addition, I presented on research to 100 beekeepers at the TriCounty Beekeepers Association February 28 – March 1 2025. Although that talk focused on sunflower effects, beekeepers asked about other plants and so I discussed results from this project; at least one beekeeper identified as a cut flower farmer who was very interested in our research for this project.

Data analysis is complete and a first draft of a manuscript incorporating all results has been written, with planned submission by the end of spring 2025.

Learning Outcomes

Key areas in which farmers reported changes in knowledge, attitude, skills and/or awareness:

I did not collect data at outreach events.

Project Outcomes

1 Grant applied for that built upon this project
1 Grant received that built upon this project
$2,430,711.00 Dollar amount of grant received that built upon this project
3 New working collaborations
Additional Outcomes:

I submitted a large National Science Foundation Integrative Biology grant as the lead PI that was funded. To be transparent, the NSF proposal was submitted before the NE SARE project officially began. However, both build upon our hypotheses that Asteraceae pollen can have disproportionate effects reducing pathogen infection in bumble bees. We collected additional pilot data, inspired by the submission of the NE SARE grant, that was important for the success of the larger NSF grant. This grant involves collaborations with 5 different institutions, including 3 that are new for me (Shalene Jha at UT Austin, Megan Povelones at Villanova, and Henry Suarez of the School of Education at UMass Amherst). 

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and should not be construed to represent any official USDA or U.S. Government determination or policy.