Interacting Antagonisms: Parasite Infection Alters Bombus impatiens (Hymenoptera: Apidae) Responses to Herbivory on Tomato Plants

Abstract

Little is known about how simultaneous antagonistic interactions on plants and pollinators affect pollination services, even though herbivory can alter floral traits and parasites can change pollinator learning, perception, or behavior. We investigated how a common herbivore and bumble bee (Bombus spp.) parasite impact pollination in tomatoes (Solanum lycopersicum L.) (Solanales: Solanaceae). We exposed half the plants to low-intensity herbivory by the specialist Manduca sexta L. (Lepidoptera: Sphigidae), and observed bumble bee visits and time spent on flowers of damaged and control plants. Following observations, we caught the foraging bees and assessed infection by the common gut parasite, Crithidia bombi Lipa & Triggiani (Trypanosomatida: Trypanosomatidae). Interestingly, we found an interactive effect between herbivory and Crithidia infection; bees with higher parasite loads spent less time foraging on damaged plants compared to control plants. However, bees did not visit higher proportions of flowers on damaged or control plants, regardless of infection status. Our study demonstrates that multiple antagonists can have synergistic negative effects on the duration of pollinator visits, such that the consequences of herbivory may depend on the infection status of pollinators. If pollinator parasites indeed exacerbate the negative effects of herbivory on pollination services, this suggests the importance of incorporating bee health management practices to maximize crop production.

Reproduction and yield in many crops depend on interactions with pollinators (Klein et al. 2007, Garibaldi et al. 2013). However, pollination services can be negatively impacted by antagonists, such as herbivores on plants and parasites of pollinators. In agricultural settings herbivores may have direct negative effects on yield (Oerke 2006), but can also reduce fruit and seed yield indirectly by disrupting pollination services (Moreira et al. 2019). Similarly, pollinator parasites pose direct threats to pollinator health (Goulson et al. 2015), but their effects on pollination services remain understudied (but see Gillespie et al. 2015, Koch et al. 2017). Furthermore, no studies have assessed the combined effects of plant and pollinator antagonists on plant-pollinator interactions, despite the frequency of these overlapping stressors (Pedersen and Fenton 2007).

Commercially important crops such as tomatoes (Solanum lycopersicum L.) (Solanales: Solanaceae) often benefit from insect pollination, but herbivores cause floral changes that could reduce pollination and limit production. Tomatoes are a self-compatible crop, but maximal fruit and seed production depend on releasing pollen from poricidal anthers by a specialized behavior called ‘buzz pollination’ (Morandin et al. 2001). Tomato flowers attract pollinators via visual and scent displays that advertise pollen as a reward (Morse et al. 2012). However, in other systems, herbivory can induce changes in floral morphology (Moreira et al. 2019), floral scent (Hoffmeister et al. 2016), floral tissue chemistry (Strauss and Irwin 2004), and pollen qualities (Aguirre et al. 2020). In wild tomato, Solanum peruvianium, herbivory induced floral volatiles that reduced pollinator attraction and subsequent seed set (Kessler and Halitschke 2009, Kessler et al. 2011). Thus, plants that have experienced herbivory may be less attractive to pollinators and receive fewer or shorter visits, with consequences for plant reproduction.

Bumble bees (Bombus spp.) are important pollinators for tomato plants (Strange 2015, Toni et al. 2021) that are capable of buzz pollination (Cooley and Vallejo-Marín 2021), but their efficiency as pollinators could be inhibited by parasite infections. Parasites, such as the gut endoparasite Crithidia bombi Lipa & Triggiani (Trypanosoma: Trypanosomatidae; hereafter Crithidia), appear to cause cognitive declines that hamper bumble bee ability to process visual and chemical cues, and learn novel flower handling behaviors (Gegear et al. 2005, 2006). Consequently, parasite infection may result in reduced pollination services; bumble bees infected with Crithidia visit artificial flowers at lower rates (Otterstatter et al. 2005), are less likely to carry pollen on their corbiculae (Shykoff and Schmid-Hempel 1991) and some plants in natural communities with higher Crithidia prevalence received fewer bumble bee visits (Gillespie and Adler 2013). Parasite-induced bee cognitive declines could be particularly relevant for the pollination of tomatoes because buzz pollination may require a learning period (L.A.A., personal observation) and bees with challenged cognition could take longer to collect pollen (i.e., slower handling times). Paradoxically, longer foraging times on a flower could theoretically result in more pollen deposited on stigma.

We investigated how an herbivore and pollinator parasite interact to affect pollination. Specifically, we measured pollination service quality as the duration of floral visits. We focused on flower visit duration because visit duration is a reliable predictor of pollen transfer in sweet pepper (Capsicum anuum), another Solanaceous crop (Jarlan et al. 1997), and the number of ‘buzzes’ and the bruising levels on tomato anthers, both of which increase with visit duration, are reliable predictors of fruit set and fruit weight (Morandin et al. 2001, Hogendoorn et al. 2006). We hypothesized that herbivory would reduce visit duration because bees are less attracted to herbivory-induced wild tomato flowers (Solanum peruvianum) (Kessler and Halitschke 2009, Kessler et al. 2011). Conversely, we hypothesized that Crithidia infection would increase visit duration because Crithidia-induced cognitive impairments would increase flower handling times (Otterstatter et al. 2005). We further predicted that when both antagonists co-occur, their opposing effects would offset each other, resulting in little net change in visit duration. Additionally, we assessed whether infection altered bee preference for visiting flowers on damaged or undamaged plants (hereafter, referred to as control plants) by calculating the proportions of visits to flowers on damaged plants (i.e., number of floral visits on damaged plants/total number of visits) for individual bees and the number of consecutive visits to the same plant treatment. We hypothesized that infected and uninfected bees would both prefer control plants, visiting control flowers in a higher proportion, and for longer bouts of consecutive visits.

Methods

Tomato seedlings (Solanum lycopersicum, ‘Matt’s Wild Cherry’ variety, n = 110) were obtained from a commercial supplier on 15 May 2017 and were maintained in their original 6-plug trays until all plants were at least 0.25 m tall. On 17 July, plants were transplanted onto an open field plot at the UMass Crop Animal Research Farm in South Deerfield, MA, USA (42° 28′45.53ʺ N 72° 34′46.06ʺ W). Tomatoes were planted in a 10 × 11 grid (1.8 m distance between plants and rows) and supplemented with 3.5 g of ‘Scott’s Osmocote Classic’ slow-release fertilizer (The Scott’s MiracleGro Company, Marysville, OH). On 10 August, herbivory treatments were applied on half of the plants; treatments were applied uniformly, with every other plant subjected to the herbivory treatment. The herbivory treatment was carried out by placing three-fifth-instar tobacco hornworms (Manduca sexta L., Lepidoptera: Sphingidae; Great Lakes Hornworm, Romeo, MI) in mesh bags enclosing three fully extended leaves on separate branches; in agricultural settings, tobacco hornworms can completely defoliate tomato plants (Delannay et al. 1989). Hornworms were removed two days later, when all leaf tissues were consumed. To control for manipulation effects, three leaves on control plants were also enclosed with mesh bags for the same period.

To take a ‘snapshot’ of bee behavior over a five-day period, we observed wild foraging bees visiting our farm plot every day from 0800 to 1300 h, between 22 and 26 August. Observations of individual bees lasted 2–5 min, and we only included observations in which bees sonicated the anthers and visited both damaged and control plants to enable direct comparisons of behavior between damage and control treatments for each bee. An observation began when a free-flying bee approached a flower near the center of the plot and continued until it exhibited grooming behaviors, an indication of preparation to leave. We recorded visits to flowers by recording when a bee landed by saying ‘start’ into a voice recorder, and when the bee left by saying ‘end’. Recordings were transcribed later using the ‘GarageBand’ application from MacOS to measure the time intervals between each start/end sequence. Altogether, 91 bees were observed making n = 2,410 flower visits, with individual bees making 11–59 flower visits each. All bees were identified as Bombus impatiens Cresson workers.

Following observations, bees were caught, placed on ice, and transported to the lab, where they were dissected to assess infection by Crithidia following the protocol in Richardson et al. (2015). Briefly, each bee’s hindgut was ground in a microcentrifuge tube with 300 µl of ¼ strength Ringer’s solution, homogenized with a tissue grinder, vortexed, and left standing for 4 h to allow the tissues to settle. Then, 10 µl of the supernatant was micro-pipetted onto a hemocytometer and the number of live Crithidia cells in a 0. 02 µl aliquot was counted under a microscope at 400×.

To investigate the effects of herbivory and bee infection status on pollinator visit duration, data were fitted using a global model with a Gamma error distribution that included herbivory treatment, Crithidia cell count per 0. 02 µl of gut solution, Julian date (as a categorical variable), wing marginal cell size (a proxy for bee size; Nooten and Rehan 2020) and all possible interactions for the first three fixed effects, and bee individual as a random effect. We performed model selection by removing nonsignificant terms and comparing unreduced and reduced models. Better models were selected using Akaike Information Criterion (AIC) and parsimony (i.e., more complex models were deemed better only if their AIC values were 2 units lower than simpler models). The best model included three predictors (herbivory, Crithidia count, and date) and the interactions between herbivory and count, and herbivory and date, as well as the bee individual as a random effect. Lastly, we conducted Type III Wald χ² tests for the remaining terms in the final model. Model assumptions were checked using the DHARMa package (Hartig 2021).

To assess whether bees preferred damaged or control plants (i.e., proportion of visits to flowers on damaged plants and number of consecutive visits to plants in the same treatment), data were fitted using global models with a Gaussian (visit ratios) and a Poisson error distribution (number of consecutive visits). Because no predictor of interest was retained in the final models for either response, we report the analysis and results in the Supplementary Material and do not discuss them further.

All statistical analyses were performed using R ver. 4.1.0 (R Core Team 2021) and the packages glmmTMB (Brooks et al. 2017) and DHARMa (Hartig 2021). Plots were created with the packages emmeans (Lenth 2019), ggplot2 (Wickham 2016), and patchwork (Pedersen 2020).

Results

Overall, 48 out of 91 bees were infected with Crithidia, providing a balanced sample size to assess effects of infection. We found a significant interaction between herbivory and parasite infection intensity (Fig. 1; Table 1), such that infection intensity reduced visit duration on flowers of damaged but not control plants. This resulted in bees with the highest infections spending approximately 25% less time (i.e., ~1 s) than uninfected bees on flowers from damaged plants. Additionally, bumble bee visits to control and damaged plants were similar in duration when the bees were not infected with Crithidia (Fig. 1; Table 1). However, there was an interaction between effect of herbivory and date (Table 1), such that visit durations were longer on damaged plants but only on the first day of observations (Supp Fig. 3 [online only]).

Discussion

Our results indicate that herbivores and pollinator parasites can act in a negatively synergistic manner (Fig. 1). To our knowledge this is the first study demonstrating that pollinator parasites may exacerbate the negative effects of herbivory on plants. Although we do not know the mechanism underlying this observation, we hypothesize that infected bees are either less tolerant of secondary compounds produced in response to herbivory or less able to extract pollen from flowers on damaged plants. Floral tissues contain relatively high concentrations of herbivore-deterrent secondary compounds (Cook et al. 2013) that can increase after herbivory (Strauss et al. 2004, Barber and Soper Gorden 2014). Consequently, bumble bees may become exposed to secondary compounds in the anthers as they break epidermis while anchoring themselves onto the flower. Uninfected bees may tolerate these compounds, but when bee health is compromised their tolerance may be lowered, as many secondary compounds can have toxicity effects (Detzel and Wink 1993). Alternatively, herbivory could reduce either the amount of pollen produced per flower (Lehtilä and Strauss 1999) or modify the anther structure, making pollen collection more energetically expensive. If Crithidia infection also has an energetic cost, then infected bees may leave the flower sooner because they have a lower giving-up density that they reach more quickly in flowers from damaged plants. These hypotheses, however, are speculative.

Our results did not support our original hypothesis that herbivory alone would reduce visit duration, and that infection alone would increase visit duration. While herbivory on wild tomato relatives (i.e., Solanum peruvianum) reduces pollinator visitation (Kessler and Halitschke 2009, Kessler et al. 2011), we did not find the negative effect of damage on visit duration that we expected. This result may be due to damaged plants compensating for herbivore damage by increasing floral signals or resources (Paige and Whitham 1987) or the effect may be too weak to measure. Meanwhile, infection levels did not affect the duration of visits (to flowers from control plants). This suggests that infected bees do not suffer cognitive challenges significant enough to affect flower handling times. Thus, Crithidia may only reduce pollination services when combined with other stressors, such as herbivory.

The strong correlations between the number of ‘buzzes’ performed by a bee and fruit size for tomato plants (Hogendoorn et al. 2006), suggest that the negative effects of herbivory and pollinator parasites could lower crop yield by reducing visit duration. Thus, incorporating bee health management practices may improve crop production. Several factors suggest the value of examining these interactions more broadly. First, approximately 70% of crop species are estimated to depend on animal-mediated pollination (Klein et al. 2007), the economic value of which is calculated to be approximately 180 billion USD annually (Gallai et al. 2009). Secondly, bee parasites and parasites are prevalent in agricultural settings. For example, Crithidia prevalence in wild bumblebees can be as high as 80% in the region where our study took place (western Massachusetts, USA); other parasites and parasites, such as Nosema bombi and conopids, are also common (Gillespie 2010, see also Averill et al. 2021). Lastly, the effects of crop antagonists such as herbivores and pollinator parasites are both projected to increase as a consequence of climate change (O’Connor 2009, West et al. 2012) and continued parasite spillover from commercial activities (Otterstatter and Thomson 2008, Graystock et al. 2016), respectively. Thus, it is critical to understand how antagonists interact to affect crop yield, as the economic repercussions of negative synergisms may be increasingly significant under future climate scenarios.

Acknowledgments

We thank J.K. Davis and the South Deerfield Agricultural Station staff for logistical support, and the UMass Quantitative Statistics Group for feedback on the statistical analysis. This work received financial support from the Lotta M. Crabtree Fellowship from the University of Massachusetts, Amherst and the National Science Foundation Graduate Research Fellowship (grant numbers NSF 1451513; NSF 1938059) to LAA and the United States Department of Agriculture's National Institute of Food and Agriculture and Co-op Research and Extension Services (Multi-state Hatch; grant numbers USDA-NIFA-2016-07962; MAS00497) to LSA.

Author Contributions

LAA and LSA designed the study. LAA carried out the experiment, conducted the statistical analysis and drafted the first version of the manuscript. Both authors contributed to the final manuscript.

Data Availability

All data and R scripts are available at GitHub (https://github.com/laguir3/2017_tomato_visits).

References Cited