Synergistic negative effects between a fungicide and high temperatures on homing behaviours in honeybees

Abstract

Interactions between environmental stressors may contribute to ongoing pollinator declines, but have not been extensively studied. Here, we examined the interaction between the agricultural fungicide Pristine (active ingredients: 25.2% boscalid, 12.8% pyraclostrobin) and high temperatures on critical honeybee behaviours. We have previously shown that consumption of field-realistic levels of this fungicide shortens worker lifespan in the field and impairs associative learning performance in a laboratory-based assay. We hypothesized that Pristine would also impair homing and foraging behaviours in the field, and that an interaction with hot weather would exacerbate this effect. Both field-relevant Pristine exposure and higher air temperatures reduced the probability of successful return on their own. Together, the two factors synergistically reduced the probability of return and increased the time required for bees to return to the hive. Pristine did not affect the masses of pollen or volumes of nectar or water brought back to the hive by foragers, and it did not affect the ratio of forager types in a colony. However, Pristine-fed bees brought more concentrated nectar back to the hive. As both agrochemical usage and heat waves increase, additive and synergistic negative effects may pose major threats to pollinators and sustainable agriculture.

1. Background

Pollinator populations have been reported to be under threat across the globe [1–3]. In parallel, managed honeybee populations in the US and Europe are experiencing high rates of annual colony loss [4,5]. These losses are important because honeybees and other animal pollinators contribute substantially to agricultural production. Bee-pollinated crops constitute one-third of the global human food supply [6]. In the US, animal pollination services are worth $14.2–23.8 billion annually [7]. There is also evidence that certain US crops, including apples, cherries and blueberries, are already limited by lack of pollinator visitation [8]. Given the economic and societal importance of pollinators, it is essential to understand the causes of their population declines.

Among the likely causes of pollinator population declines are habitat loss and fragmentation, agrochemicals, pathogens, non-native species and climate change [3]. While environmental stressors are often studied in isolation, interactions between these stressors are likely. For example, poor nutrition can make bees more vulnerable to viruses [9] as can pesticide exposure [10,11]. Of particular interest are interactions between pesticides and extreme temperatures, such as the heat waves that are expected to occur as a result of climate change. Neonicotinoids reduce the ability of honeybees to survive exceptionally warm or cold temperatures [12,13], with greater differences in gene expression found in bees exposed to both stressors as opposed to one or the other [12,14,15]. A recent study also found that the neonicotinoid imidacloprid interacted with high temperatures to reduce the distances that bumblebees were able to fly in a tethered flight mill [16].

Navigation and foraging are complex tasks for honeybees that require associative learning [17]. Honeybees are central place foragers, which means they need to navigate to floral resources outside the colony and bring nectar and pollen home [18]. Successful navigation requires bees to learn about their surroundings using multiple sensory modalities [19], integrating visual landmarks [20], polarized light [21], and optic flow [22] cues to understand the direction and distance needed to travel in order to return to the hive. Odour learning also plays a role in navigation and foraging, as workers learn floral odours from their nest-mates in the hive, which helps them locate and exploit the source later [23].

Because navigation and foraging are complex tasks that involve associative learning, stressors such as high temperatures [24] and pesticides [25,26] that impair associative learning also negatively affect navigation and foraging behaviours. Neonicotinoids and other insecticides can reduce the frequency and increase the average duration of foraging trips [27–30], reduce the average age of first foraging [31,32] and alter the amount of pollen and nectar resources brought back to the hive [33]. Neonicotinoids [34–37], pyrethroids [38] and the herbicide glyphosate [39] can also negatively impact the ability of bees to successfully return to the hive after foraging, and these homing failures can contribute to colony collapse over time [36]. Heatwave-like temperatures can lead to altered foraging activity in bumblebees [40,41] and can induce precocious foraging and reduce longevity in honeybees [42]. No studies have experimentally investigated the synergistic effects of high temperature and pesticides on navigation and foraging in bees, although some have found that exposure to neonicotinoids during cold weather can reduce homing success [43,44].

We have previously shown that the fungicide Pristine impairs olfactory associative learning performance in honeybees [45]. We therefore hypothesized that it would negatively impact homing and foraging behaviours, which depend on the bee's learning ability. Pristine's active ingredients are boscalid (25.2%) and pyraclostrobin (12.8%), both of which interfere with the electron transport chain in fungal cellular respiration [46]. We used a combination of manual methods and radio-frequency identification (RFID) tracking technology to test whether field-relevant Pristine exposure reduced homing success and the amounts of various types of resources collected by foragers. The homing experiments were conducted across a wide range of air temperatures, from approximately 20–40°C, enabling us to determine whether air temperature influenced any potential negative effect of Pristine on the ability of honeybees to return to their hives.

2. Methods

(a) . Honeybee colonies and fungicide exposure

Colony initiation and exposure protocols were similar to those described in previous studies [45,47]. For the 2021 homing experiment, we obtained six 1.4 kg Italian honeybee (Apis mellifera ligustica) packages from Pendell Apiaries in Stonyford, California in April. The packages were used to initiate new colonies in Apimaye plastic hive boxes (Kaftan LLC, Tempe, AZ) at the Arizona State University Bee Lab in Mesa, Arizona (33.293173, −111.684520).

For the 2022 homing and foraging experiments, we took approximately 1.4 kg of adult bees from the 2021 hives and paired them with new queen bees purchased from Pendell Apiaries. In April, the bees and new queens were used to initiate new colonies (with new frames and hive boxes). Since this procedure kept only adult bees from the initial 2021 colonies (which were also quickly replaced by newer brood), it prevented the 2022 colonies from being influenced by any potentially leftover fungicide residues from the 2021 experiments.

For the 2021 experiments, fungicide exposure began in September. Pollen traps were placed on hive entrances to limit the amount of outside pollen that foragers were able to bring in. A random number generator (random.org) was used to assign three hives each to the fungicide treatment and control groups. Pollen patties containing either plain deionized water (control) or deionized water mixed with 23 ppm Pristine (treatment group) (BASF Corporation, Research Triangle Park, NC) were placed inside the hive ad libitum. Fifty grams of the pollen patty mixture were given to each colony at a time, and hives were checked every other day so they could be replenished as needed. We chose 23 ppm Pristine as the focal concentration because bees could realistically be exposed to that amount while foraging in a treated almond orchard [47], and this concentration also impairs associative learning performance in the laboratory [45].

In 2022, fungicide exposure began in July. Hives were assigned to treatment groups and fed in the same manner as in 2021. Fungicide exposure continued until experiments concluded in October. This corresponded to a three-month exposure period. Colonies involved in almond pollination are typically exposed for at least four weeks (the length of the almond bloom). A three-month exposure period is not unlikely because these colonies are migratory and Pristine is sprayed on a number of bee-pollinated crops [47].

To ensure that all focal bees in our experiment would be of similar age, we completed an age-marking process similar to the one described in Fisher et al. [48], once in 2021 (late September) and twice in 2022 (early August and late September). Capped brood frames were removed from each hive and placed in separate wire cages in an incubator (34°C, 90% relative humidity) overnight. The following day, newly emerged adults were marked on the mesonotum with a paint colour corresponding to their date of emergence, and were then returned to their hive of origin. This process was repeated until there were at least 550 marked bees in each hive. The homing and foraging experiments began after marked bees from all hives had begun foraging (approx. two weeks later).

(b) . 2021 Homing success

Marked foragers were collected in glass vials as they exited the hive. They were brought inside, anesthetized on ice, and then wood glue (Gorilla Glue, Cincinnati, OH) was used to attach a queen tag (Mann Lake Bee & Ag Supply, Hackensack, MN) with a unique colour and number to their mesonotum. Workers were then placed back in their glass vials along with a piece of cotton soaked in a 1.0 M sucrose solution. They were allowed to feed ad libitum on the sucrose during the time between marking and their release (approx. 20 min).

Bees were released from a location 1 km east of the ASU Bee Lab, as determined using the measurement tool in Google Maps. This distance was chosen because it was comparable to the distance honeybees would usually cover when foraging [49], and it has been used in similar homing studies [36]. The release site was located along Old Pecos Road, which runs through vacant ASU-owned property and is closed to vehicle traffic. This direction was chosen because, compared to other locations around the Bee Lab, it contained relatively few landmarks, which can influence results [43]. Weather data (cloud cover, air temperature, wind speed, and wind direction) for the time of release were taken from a weather station at the Phoenix-Mesa Gateway Airport (https://www.wunderground.com/weather/KIWA), which was located approximately 2 km from the release site. Bees were released by removing the caps from the glass vials, placing them on the ground, and allowing individuals to exit on their own. The vials were left open for five minutes, after which any remaining bees were considered unmotivated to fly and removed from the experiment.

The colony entrances were partially covered with mesh wire at all times, which slowed down returning foragers and made it easier for the observers to see their tag numbers. Observers watched the colony entrances starting when the bees were released and ending one hour later. The length of the observation period was chosen because previous studies suggested that the majority of bees to return would do so within the first hour [36]. Observers noted the time that each marked bee landed at the colony entrance.

Experiments were performed during a total of 14 bouts, taking place from 7 October to 4 November. The number of hives sampled during each bout ranged from 1 to 4 depending on the number of observers available (with one observer per hive). Hives to be sampled from were selected randomly each time, although care was taken to ensure that all hives had been equally sampled from throughout the experiment. Around 15 bees from each hive were captured at a time.

(c) . 2022 Homing success

The capture and marking process was the same as in 2021, except bees were marked with RFID tags (BEE-TAG mic3Q1.6, microsensys GmbH, Erfurt, Germany) instead of queen tags. Each unique tag was assigned to a treatment group (fungicide or control) and hive using a specialized RFID device (iIDPENsolid UHFcc, microsensys GmbH). The release site and methods were the same as in 2021, although only two colonies (one control and one treatment) were sampled during each bout. Before each bout of the experiment, RFID readers (iIDscience reader device AEB-03.C2D, Microsensys GmbH) were installed at hive entrances. They remained installed for three hours following the release time. Whenever the readers were not installed on the hives, three-dimensional printed ‘sham’ readers were present instead so that the entrance would always look the same to the bees. As bees returned, their tag IDs were recorded along with timestamps on a specialized system controller (iIDBEEcontroller, microsensys GmbH).

The experiment took place across two separate repetitions (one in August/September and one in October). There were 10 runs in August/September between 29 August and 15 September. There were 11 runs in October, taking place 10–29 October.

(d) . 2022 Forager resource collection

Forager sample collection occurred at the same time as both of the 2022 homing repetitions but was not conducted in 2021. Temperature, cloud cover and wind speed/direction were recorded using the same source as in the homing experiment. Marked foragers were captured in glass vials as they returned to the hive. They were brought inside and immediately euthanized in a mixture of ethanol and dry ice, which prevented regurgitation of crop contents.

Bees were classified as pollen, nectar or water foragers following a similar methodology to the one outlined in Prado et al. [33]. If the bee was carrying pollen, the pollen was scraped off using forceps and weighed on an analytical balance. The bee's abdomen was cut off with dissecting scissors and its crop contents were collected using 2, 5 and 10 µl microcapillary tubes (Drummond Scientific, Broomall, PA) to measure crop volume. The crop contents were then transferred to a BRIX refractometer (VLT032, V-Resourcing) in order to measure sugar concentration. If the sugar concentration was 10% Brix or higher, the bee was counted as a nectar forager, otherwise, it was classified as a water forager [33].

(e) . Statistical analysis—homing

The datasets and code for all experiments are available in Dryad [50]. Data were analysed in R version 4.2.2 [51]. We analysed two dependent variables: whether the bee returned to the hive after it was released (which we refer to as probability of return) and the time it took successful bees to return. For both variables, we considered the effect of treatment group (Pristine versus control) and the air temperature at the time of release. We included hive, wind speed and wind direction (a numeric variable between 0° and 360°) in all models as random effects. We analysed the two years separately as well as combined (with year as a fixed effect in the combined models); we report both results here. The pooled analysis gave us more statistical power overall and allowed us to compare the results directly between years, while the separate analyses control for the different methodologies between years. Reporting both provides the most complete picture of the results.

To evaluate the probability that bees successfully returned to the hive, we ran generalized linear mixed models (logit link function, binomial family) using the lme4 package [52]. For the model that combined the two years, we first tested for a three-way interaction between treatment, temperature and year, as well as for effects of all those variables separately. Hive was included as a random effect. We then dropped non-significant interactions and re-ran the model. For the models that analysed the years separately, we tested for an interaction between treatment and temperature, as well as effects of each variable separately.

To evaluate the time it took bees to return to the hive, we ran mixed effects Cox models using the coxme package [53]. The procedure was the same as described above for the probability of return GLMMs. We included hive as a random effect and tested for three-way interactions between treatment, temperature and year in the combined model. In the models for each separate year, we tested for interactions between treatment and temperature.

(f) . Statistical analysis—foraging

Data were analysed in R version 4.2.2 [51]. Using the lme4 package [52], we created four linear mixed effects models to test for effects of treatment group and repetition on corbicular pollen mass, nectar forager crop volume, water forager crop volume and nectar forager crop sugar concentration. Hive was included in all models as a random effect. Pearson's chi-squared tests were performed using the base stats package to analyse the difference in proportions of foragers carrying different resource types (water, nectar, pollen or nothing) between treatment groups and repetitions.

3. Results

(a) . Homing—probability of return

The model that combined years produced statistically significant effects of treatment (χ² = 6.50, p = 0.0108) and temperature (χ² = 22.0, p < 0.001), meaning that Pristine bees were less likely to return, and all bees were less likely to return at higher temperatures. Year did not produce a statistically significant effect (χ² = 0.715, p = 0.398). In 2021, treatment alone produced a significant effect (χ² = 4.16, p = 0.0415), meaning that Pristine bees were less likely to return. However, temperature (χ² = 0.178, p = 0.673) and the interaction between treatment and temperature (χ² < 0.001, p = 0.979) were nonsignificant (figure 1b). In 2022, treatment alone did not produce a significant effect (χ² = 2.28, p = 0.131), but temperature (χ² = 15.1, p < 0.001) and the interaction between treatment and temperature (χ² = 4.69, p = 0.0303, figure 1e) did, meaning that all bees, but especially those exposed to Pristine, were less likely to return at higher temperatures.

Figure 1.

Effects of Pristine and air temperature on homing performance during experiments in 2021 (top row) and 2022 (bottom row). (a) Cumulative percentage of bees that returned to the hive over time during 2021 experiments. n = 63 control bees, 70 Pristine bees. (b) Plot showing the potential interaction between air temperature and bee probability of return in 2021. In this case, the lines do not cross, indicating that there was no interaction between temperature and treatment group. R² = 0.0427. n = 63 control bees, 70 Pristine bees. (c) Plot showing the potential interaction between temperature and the time it took bees to return in 2021. In this case, the lines do cross, indicating a likely interaction (Pristine bees took longer to return to the hive at higher temperatures). R² = 0.0512. n = 31 control bees, 22 Pristine bees (sample sizes are lower in panels (c) and (f) because only the bees that successfully returned were counted). (d) Cumulative percentage of bees that returned to the hive over time during 2022 experiments. n = 109 control bees, 105 Pristine bees. (e) Plot showing the interaction between temperature and probability of return in 2022. Lines cross, indicating an interaction (Pristine bees were less likely to return to the hive at higher temperatures). The grey panel represents the range of temperatures measured in 2021 to highlight the expanded range of temperatures sampled in 2022. R² = 0.243. n = 109 control bees, 105 Pristine bees. (f) Plot showing the interaction between temperature and time to return in 2022. Lines cross, indicating an interaction (Pristine bees took longer to return to the hive at higher temperatures). R² = 0.213. n = 48 control bees, 36 Pristine bees.

(b) . Homing—time to return

The model that combined the years produced statistically significant effects of year (χ² = 17.7, p < 0.001, likely because of the longer observation period in 2022) and the interaction between treatment and temperature (χ² = 6.94, p = 0.00843), with Pristine bees taking longer to return than controls at higher temperatures. Neither treatment (χ² < 0.001, p = 0.997) nor temperature (χ² = 0.0643, p = 0.800) alone produced significant effects. In 2021, neither treatment (χ² = 0.0362, p = 0.849), temperature (χ² = 0.173, p = 0.677), nor the interaction between them (χ² = 2.69, p = 0.101) was significant, although the graph showed a trend towards an interaction (figure 1c). In 2022, neither treatment (χ² = 0.00770, p = 0.930) nor temperature (χ² = 0.0579, p = 0.810) produced a statistically significant effect on their own, but the interaction between the two was significant (χ² = 4.88, p = 0.0272), again with Pristine bees taking longer to return than controls at higher temperatures (figure 1f).

(c) . Foraging—pollen, water and nectar loads

Repetition (August versus October) had no significant effects on anything except for the sugar concentration of nectar forager crop contents (χ² = 5.20, p = 0.0226), with higher concentrations recorded in August (mean = 44.5%, 95% CI = 5.86%) compared to October (mean = 36.8%, 95% CI = 3.76%).

Fungicide treatment had no significant effect on corbicular pollen mass (χ² = 1.87, p = 0.645) (figure 2a), volume of nectar forager crop contents, (χ² = 0.128, p = 0.721) (figure 2b), or volume of water forager crop contents (χ² = 2.00, p = 0.158) (figure 2c). Fungicide treatment did produce a significant effect on the sugar concentration of nectar forger crop contents (χ² = 5.03, p = 0.0250) (figure 2d), with higher concentrations recorded in the Pristine group (mean = 43.6%, 95% CI = 4.80%) compared to the control group (mean = 36.3%, 95% CI = 4.35%).

Figure 2.

Comparisons of nectar, pollen and water amounts brought back to the hive between control and Pristine treatment groups. Bars indicate means, while error bars indicate 95% confidence intervals. Asterisks denote significant differences between groups. (a) Comparison of corbicular pollen masses between groups. n = 24 control bees, 26 Pristine bees. (b) Comparison of the volume of nectar forager crop contents between groups n = 39 control bees, 36 Pristine bees. (c) Comparison of the volume of water forager crop contents between groups. n = 27 control bees, 28 Pristine bees. (d) Comparison of the sugar concentration of nectar forager crop contents between groups. n = 39 control bees, 36 Pristine bees.

(d) . Foraging—proportion of foragers collecting each resource type

Repetition (August versus October) had a significant effect on the proportion of foragers carrying nectar, water, pollen or nothing (χ² = 22.9, p < 0.001), with more pollen foragers collected in October. Treatment group (Pristine versus control) had no significant effect on the proportion of foragers carrying each resource type (χ² = 0.420, p = 0.936) (figure 3).

Figure 3.

Percentages of foragers in both the control and Pristine treatment groups that were captured at hive entrances carrying each type of resource (water, pollen or nectar). ‘Empty’ foragers were not carrying any provisions. n = 98 control bees, 100 Pristine bees.

4. Discussion

In the homing experiments, we found that Pristine exposure reduced the probability that a bee would successfully return to the colony. Fungicide treatment alone produced a significant effect in 2021 (figure 1a), as well as in the model with combined years (with a larger sample size and more power). To our knowledge, this is the first study to show a direct effect of a fungicide on homing, a critical bee behaviour. Aside from the herbicide glyphosate [39], most of the compounds found to affect honeybee homing have been neuroactive insecticides [34–38]. Our study adds to a growing body of literature suggesting that Pristine is not safe for honeybees [45,47,48,54–58]. Of particular interest is a prior study by Fisher et al. [47], which showed that Pristine reduced adult population size and worker longevity in chronically exposed colonies. Our results provide a potential underlying mechanism, as bees who do not successfully return to the hive after foraging likely do not survive, which can lead to reduced worker populations and ultimately colony collapse over time [36].

We also found that these homing failures were exacerbated when paired with high temperatures. We define ‘high temperatures’ as those above 32°C, which were sampled in our 2022 experiments but not 2021, and may be experienced during an Arizona summer or during heat waves in many parts of the world. The model that measured probability of return with the two years combined showed a significant effect of temperature (suggesting that all bees, regardless of treatment group, were less likely to successfully return to the hive at higher temperatures), but did not show a significant interaction between treatment and temperature. When the two years were analysed separately, 2022 showed a significant interaction between fungicide treatment and temperature (figure 1e), while 2021 did not (figure 1b). The interaction becomes most prevalent at temperatures above 32°C, which were sampled in 2022 but not 2021. It is important to note that 32°C by itself is not necessarily considered a stressful temperature for honeybees, as they have been observed foraging without major performance limitations in the Sonoran Desert at temperatures above 40°C [59]. However, it may simply be that Pristine exposure makes bees more susceptible to these elevated temperatures.

We also found that Pristine interacted with high temperatures to increase the amount of time it took the bees to return. This effect was significant in the model with combined years and in the 2022 model (figure 1f). The effect was not statistically significant in the 2021 model (figure 1c), although the lines crossed in the relevant graph, indicating a trend towards an interaction. We conclude that the interactive effect between fungicide treatment and temperature on time to return was present to some degree in both years, but was most apparent when the higher temperatures were sampled in 2022.

There is a potential confounding factor in our results—the observation period was three hours long in 2022, versus only one hour in 2021. When comparing the return rates of foragers at one hour, there is substantial variation between years. In 2021, 49% of bees in the control group successfully returned to the hive after one hour, compared to 31% of Pristine bees. In 2022, 22% of control bees had returned after one hour, compared to 17% of Pristine bees (a much smaller difference). We believe that this difference is in large part due to the higher temperatures sampled in 2022, which caused lower return rates in bees from both treatment groups. Regardless, we encourage caution when comparing the results between years due to the differences in observation time and return rates.

It is also worth noting that there could be some indirect factor correlated with temperature at play (e.g. a change in the availability of flowers/nectar sources, water sources or shade). In this case, it would not be temperature itself producing the effect but rather something else, which would be difficult to fully separate from temperature in this field setup. We discuss interactions between the fungicide and temperatures throughout this paper, but we encourage caution when interpreting the results.

The synergistic effects between the fungicide and high temperatures are worrisome given current trends of climactic warming. Pristine has been used extensively on almond trees in California. Although temperatures are unlikely to climb as high as 32°C during the February almond bloom in California's Central Valley in the near term (with average temperatures ranging from approximately 5–15°C and all-time highs at approximately 25°C) [60], Pristine is used on a variety of other honeybee-pollinated crops, some of which bloom in the summer. Normal summer temperatures usually peak around 30°C in the Central Valley, but heat waves can bring temperatures in excess of 40°C [60]. Especially as heat waves become more frequent as a result of climate change [61], growers and regulators should exercise caution when deciding which agrochemicals are safe to use on blooming crops.

A handful of other studies have shown interactions between pesticides and temperature extremes. Neonicotinoids reduce the ability of honeybees to survive exceptionally warm or cold temperatures [12,13]. A recent study also found that the neonicotinoid imidacloprid interacted with high temperatures to reduce the distances that bumblebees were able to fly in a tethered flight mill [16]. Our study focused on homing success, but also required bees to fly one kilometre, a distance which may have been challenging for Pristine-exposed bees, particularly at higher temperatures. No study to date has shown an interactive effect between high temperatures and pesticides on homing success specifically, although the neonicotinoid thiamethoxam can interact with low temperatures to exacerbate honeybee homing failure [43,44].

The reasons underlying these homing failures are unclear. We present a few possible mechanisms here, but concede that further research is needed to determine a definite cause. One possible mechanism underlying this interaction could involve an effect of the fungicide on energetic metabolism, causing exposed bees to tire more quickly than controls, especially at high temperatures. Pristine directly inhibits cellular respiration in isolated honeybee mitochondria [54], and it also lowers ATP levels in honeybee flight muscles [55]. However, supporting evidence for this hypothesis is mixed, as some studies have reported no effects of Pristine on carbon dioxide production and thorax temperatures during flight [54]. Glass et al. [58] reported a negative effect of Pristine on maximal flight performance, but only when fed at 230 ppm, which is 10 times higher than the concentration used in our experiments.

Another possible mechanistic explanation for these homing results involves Pristine's effects on nutrient absorption. The active ingredient pyraclostrobin damages the honeybee midgut [62,63] and the formulation interferes with protein and carbohydrate absorption [55,57]. These changes could potentially result in Pristine-fed bees lacking the fuels necessary to sustain flight or impaired development of some tissues. Glass et al. [58] found that Pristine bees had smaller thoraxes than controls, suggesting that flight muscles were not as well-developed in these bees.

A third possible explanation for the increase in homing failures is that bees could have reduced cognitive abilities as a result of exposure to Pristine, high temperatures, or both. As Pristine is able to pass through the gut and into the hemolymph [64], it is plausible that it could also pass through the sheath surrounding the brain and poison mitochondria. This explanation goes along with our previous finding that Pristine impairs olfactory associative learning performance [45], which provides a basic measurement of a bee's ability to learn about relevant environmental stimuli, a necessary ability for successful homing. Likewise, short-term exposure to high temperatures can impair both learning [24] and foraging [40] behaviours in bumble bees, although the underlying mechanisms remain unknown.

In the foraging experiments, bees brought back similar masses of corbicular pollen and similar volumes of nectar and water regardless of Pristine exposure, but foragers from the fungicide treatment group returned with more concentrated nectar than controls. Nectar was also more concentrated in August compared to October. We also observed different proportions of forager types in August compared to October, with more pollen foragers observed in October. However, we observed no differences between treatment groups. Interestingly, this result differs from a previous study involving Pristine, which found that the fungicide increased colony pollen collection and consumption [47]. Overall, our results suggest that Pristine does not significantly impair forager resource collection, as Pristine bees returned to their colonies with comparable amounts of nectar, pollen and water and even greater concentrations of nectar when compared to controls.

The difference in sugar content of nectar is an interesting result, and although it may not produce any negative effects on its own, it is possibly suggestive of broader sensory effects of the fungicide. Again, more research is needed to determine the causes and consequences underlying this result. Sucrose response threshold is the lowest concentration of sucrose that will elicit proboscis extension [65–67]. There is some precedence for fungicides affecting sucrose responsiveness in honeybees [68]. Bees with higher sucrose response thresholds may be more likely to return to the colony after foraging with higher concentrations of nectar [65,66], so it follows that perhaps Pristine-exposed bees have higher sucrose response thresholds compared to controls. Despite this, a previous study found that sucrose responsiveness in a laboratory assay was not affected by Pristine consumption [45]. However, sucrose responsiveness in laboratory assays does not always correlate with sucrose responsiveness in the field [69], so it is still possible that the fungicide could be affecting this critical sensory trait.

Another possible explanation for the difference in nectar concentrations between the two treatment groups could be related to the fungicide's effects on nutritional status in honeybees. Pyraclostrobin damages the cells lining the midgut [62,63], and the Pristine formulation interferes with nutrient absorption [55,57]. If Pristine-exposed bees have trouble processing and/or absorbing sugar, collecting more concentrated nectar may help them compensate for this deficit and avoid the most serious effects of nutrient deprivation.

5. Conclusion

We found that the fungicide Pristine reduced the proportion of bees that successfully returned to the colony during a homing test, and that this effect was exacerbated when paired with temperatures above 32°C. We found limited effects of the fungicide on forager resource collection overall, although Pristine-exposed bees returned with more concentrated nectar in their crops than controls. Our results reinforce the idea that fungicides are not safe for pollinating insects [70], even if they act via non-neurotoxic mechanisms. They also highlight the importance of testing for sublethal effects of agrochemicals in the field [71], in part because important interactive effects become apparent, such as the interaction between Pristine and high temperatures. In order to ensure the stability of pollination services and our food supply, and especially as heat waves become more common due to climate change, future research should seek to develop a better understanding of the interactive effects between agrochemicals and extreme temperatures on pollinators.

Ethics

No specific ethical approval was required, as the study involved the use of invertebrates.

Data accessibility

The datasets and R code associated with this article are available in Dryad: https://doi.org/10.5061/dryad.jh9w0vth8 [50].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

N.S.D.: conceptualization, formal analysis, methodology, visualization, writing—original draft, writing—review and editing; E.K.C.: formal analysis, investigation, writing—review and editing; C.O.: investigation, methodology, writing—review and editing; C.M.L.: formal analysis, visualization, writing—review and editing; J.F.H.: conceptualization, funding acquisition, methodology, supervision, writing—review and editing; B.H.S.: conceptualization, funding acquisition, methodology, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the United States Department of Agriculture (grant nos. 2017-68004-26322 and 2022-67013-36285).