Heather pollen is not necessarily a healthy diet for bumble bees

There is evidence that specialised metabolites of flowering plants occur in both vegetative parts and floral resources (i.e., pollen and nectar), exposing pollinators to their biological activities. While such metabolites may be toxic to bees, it may also help them to deal with environmental stressors. One example is heather nectar which has been shown to limit bumble bee infection by a trypanosomatid parasite, Crithidia sp., because of callunene activity. Besides in nectar, heather harbours high content of specialised metabolites in pollen such as flavonoids but they have been poorly investigated. In this study, we aimed to assess the impact of Crithidia sp., heather pollen and its flavonoids on bumble bees using non-parasitised and parasitised microcolonies fed either control pollen diet (i.e., willow pollen), heather pollen diet, or flavonoid-supplemented pollen diet. We found that heather pollen and its flavonoids significantly affected microcolonies by decreasing pollen collection as well as offspring production, and by increasing male fat body content while parasite exposure had no significant effect except for an increase in male fat body. We did not highlight any medicinal effect of heather pollen or its flavonoids on parasitised bumble bees. Our results provide insight into the impact of pollen specialised metabolites in heather-bumble bee-parasite interactions. They underline the contrasting roles for bumble bees of the two floral resources and highlight the importance of considering both nectar and pollen when addressing medicinal effects of a plant towards pollinators.


Introduction 46
For their own subsistence and that of their offsprings, bee females mostly forage on two floral 47 resources, namely nectar as main source of carbohydrates (Nicolson & Thornburg, 2007), and pollen as 48 main source of proteins and lipids (Campos et al., 2008). Among these nutritional resources, the pollen 49 chemical composition is particularly complex and highly variable among plant species (Vaudo et al., 2020). 50 While pollen central metabolites, for instance the protein-to-lipid ratio, play crucial role in bee health, 51 development, and fitness (Di Pasquale et al., 2013), pollen also contains numerous specialised metabolites 52 (e.g., alkaloids, flavonoids and terpenoids, Irwin  activities of these metabolites are multiple so that they may be involved in protecting pollen from abiotic 54 factors, such as UVs (Li et al., 1993), but also from biotic factors, acting as antibacterial, antifungal or 55 insecticidal compounds (Pusztahelyi et al., 2015;Zaynab et al., 2018). When ingesting pollen, bees are then 56 exposed to all these biological activities that may be beneficial, for instance by reducing parasite load decreasing offspring size and production (Arnold et al., 2014), inducing larvae or imago death (Hendriksma 60 et al., 2011;Weber, 2004), and altering immune system (Gekière et al., 2022a). Given these opposite 61 effects on bees, it is essential to question how specific specialised metabolites may impact bee health, 62 especially in a changing world with multiple environmental pressures.

64
In the current context of biodiversity erosion (Butchart et al., 2010), bees are unfortunately not 65 exception, and many threats have been pinpointed as responsible for their negative population trends 66 (Dicks et

72
Social bee species such as bumble bees (Apidae; Bombus spp.) are particularly impacted by parasites, the 73 latter benefiting from their social system to readily infect numerous individuals (Folly et al., 2017). One of 74 the most prevalent parasites in wild bumble bee populations is the gut trypanosomatid Crithidia bombi 75 Lipa & Triggiani, 1980 (Euglenozoa: Trypanosomatidae; Schmid-Hempel, 2001). Despite its generally 76 moderate impacts, it can decrease foraging effectiveness (Otterstatter et al., 2005), offspring production 77 (Schmid-Hempel, 1998), queen survival through hibernation (Fauser et al., 2017), and increase mortality in 78 synergy with other stresses (Brown et al., 2000). Queenless microcolonies of five workers were exposed to specific diet treatments ( Fig.1 pollen candies (i.e., pollen mixed with a 65% sugar solution) for 35 days, with pollen candies being freshly 130 prepared and renewed every two days. When workers died, they were discarded, weighted and replaced 131 by a worker from the same queenright colony, which was marked with a colour dot on the scutum. Larvae 132 ejected from the brood were also checked every day, counted and discarded from the microcolonies. 133 Microcolonies were handled under red right to minimise disturbance. 134  the other half served for massive extraction of flavonoids. Flavonoids were extracted using a Soxhlet 149 extraction for approximatively 40 cycles with methanol as solvent, at 100°C. Extract was then vacuum 150 filtered and evaporated to dryness (rotavapor IKA RV8). For flavonoid purification, extract was solubilized 151 in water with a minimal amount of methanol, and placed in a separatory funnel with dichloromethane. The 152 funnel was shaken and left to settle overnight before recovering the aqueous phase. The purified extract 153 was then dried using rotary evaporator and dissolved in aqueous ethanol solution (1:1, v/v) before addition 154 to the control diet to prepare a flavonoid supplemented diet. Control and heather pollen diets were also 155 supplemented with a similar amount of ethanol to avoid any bias (for details see Appendix A, Table S1).

164
The inoculum was homogenised, brought to 1 mL with 0.9% NaCl solution, and purified by a triangulation 165 method (Cole, 1970) adapted by Baron  Resource collection was assessed by weighting every two days in each microcolony the syrup container, 179 as well as the recovered pollen candy and the newly introduced one. These data were corrected for 180 evaporation using controls, as well as divided by the total worker mass by microcolony to avoid bias due 181 to worker activity. To evaluate the reproductive success, all microcolonies were dissected at the end of the 182 experiment (Day 35) to weigh the total hatched brood mass, as well as the individual mass of each emerged 183 male used as reference of viable offspring at the end of development (Goulson, 2010). Offspring masses 184 were divided by the total worker mass by microcolony to avoid any bias due to worker care. Regarding 185 stress response, we assessed worker mortality, larval ejection, pollen dilution (ratio between the collection 186 of syrup and pollen) as well as pollen efficiency (ratio between offspring mass and pollen collection; Tasei  187 & Aupinel, 2008) that highlights when a micro-colony needs to consume more pollen to produce offspring.

189
For the individual health parameters, fat body content was measured at the end of the bioassays on 190 two males and two workers per microcolony (40 individuals per treatment) following Ellers (1996). The 191 abdomens were cut and dehydrated in an incubator at 70°C for three days before being weighed. They 192 were then placed for one day in 2mL of diethyl ether to solubilise lipids constituting the fat body. The 193 abdomens were then washed twice with diethyl ether, and incubated at 70°C for seven days before being 194 weighed. Fat body content was defined as the mass difference between dry abdomen before and after 195 lipid solubilisation, divided by the dry abdomen mass prior to solubilisation. Moreover, in infected 196 treatments, we repeatedly monitored the parasite load within microcolonies using the same marked 197 worker along the bioassays. The first measurement was made three days post-inoculation (day 4) to enable 198 Crithidia sp. to multiply, and ensure its presence in the faeces (

229
The last parameter measured was the parasite load at different time points within infected 230 microcolonies. As infection dynamics is a discrete time series, it was analysed using a generalised additive 231 mixed-effect model (GAMM; Wood, 2006). Parasite load were square root-transformed and fitted using a 232 Gaussian distribution with a log link. Diet and day were set as fixed factor and microcolony nested within 233 colony as random factor. The model assumptions were tested using diagnostic graphs and tests.

235
Contrasts were then performed on the models to determine whether uninfected control differed from 236 the infected control, whether effects on uninfected or infected microcolonies differed among diets 237 (emmeans function from the emmeans R-package v.

244
Parasite impact 245 Comparison of microcolonies fed with control pollen between parasitised and non-parasitised 246 treatments showed that Crithidia sp. infection did not impact the parameters related to resource collection 247 (Fig 2A), reproductive success (Fig 2B-2C), or stress response (Fig 3A-C) (p > 0.05, Fig. 2 and 3). However, 248 fat body content in newly emerged males was significantly higher with a mean that increased by 56% in 249 infected microcolonies fed the control diet compared to uninfected ones fed the same diet (t = -3.828, p = 250 0.0012; Fig. 4B). The estimates (mean ± standard error) of our variables for each treatment are available in 251 the appendices (Appendix B, Table S2) 252 253 Effect of heather pollen and its flavonoids on healthy bumble bees 254 Regarding resource collection, total pollen collection was significantly lower in microcolonies fed the 255 supplemented diet compared to those fed the other diets (control vs supplemented: 43% less pollen 256 collected, t = -5.672, p <0.001; heather vs supplemented: 33% less pollen collected, t = 3.924, p < 0.001; 257 Fig. 2A). With regards to the reproductive success, microcolonies fed the supplemented and heather diets 258 produced a significantly lower brood mass compared to microcolonies fed the control diet (control vs 259 supplemented: brood mass 52% lower, t = 3.890, p < 0.001; control vs heather: brood mass 32% lower, t = 260 2.189, p = 0.0331; Fig. 2B), as well as significantly smaller emerged males (control vs supplemented: t = 261 2.350, p = 0.0192; control vs heather: t = 2.925, p = 0.0036; Fig. 2C) 262 263 In terms of stress responses, pollen dilution was significantly higher in microcolonies fed the 264 supplemented diet compared to those fed the other diets (control vs supplemented: t = 2.282, p = 0.0268; 265 heather vs supplemented: t = -3.191, p = 0.0025; Fig. 3A). Microcolonies fed the heather or supplemented 266 diets also displayed a lower pollen efficacy than the microcolonies fed the control diet (control vs 267 supplemented t = -2.741, p = 0.0085; control vs heather: t = -3.025, p = 0.0039; Fig. 3B). On the contrary, 268 no significant difference was detected for larval ejection (p > 0.05) or for worker mortality (p < 0.05, Fig.  269 3C). 270 Regarding individual health, while no difference was detected in worker fat body content among diet 272 treatments (p > 0.05; Fig. 4A), fat body content in newly emerged males was significantly higher in 273 microcolonies fed the supplemented or heather diets compared to those fed the control diet (control vs 274 supplemented: fat body content 62% higher, t = -3.891, p = 0.0012; control vs heather: fat body content 275 41% higher, t = 2.850, p = 0.0223; Fig. 4B). 276

Effect of heather pollen and its flavonoids on parasitised bumble bees 277
Similarly as previous results with uninfected microcolonies, total pollen collection was significantly 278 lower in infected microcolonies fed the supplemented diet than in microcolonies fed the control diet (36% 279 less pollen collected, t = -4.414, p < 0.001), but also in infected microcolonies fed heather diet compared 280 to those fed the control diet (16% less pollen collected, t = -2.866, p = 0.0061) ( Fig. 2A). Regarding the 281 reproductive success, as observed in uninfected microcolonies, microcolonies fed the supplemented and 282 heather diets produced a significantly lower brood mass compared to microcolonies fed the control diet 283 (control vs supplemented: brood mass 51% lower, t = 3.784, p < 0.001; control vs heather: brood mass 284 41% lower, t = 3.551, p < 0.001; Fig. 2B). However, no significant difference was detected for the mass of 285 newly emerged males among diet treatments (p < 0.05; Fig. 2C).

297
In

348
Besides our results showed that Crithidia sp. induced larger fat body content in males emerged in 349 infected microcolonies compared to those emerged in uninfected ones, whereas it has not impact on 350 worker fat body content. We propose two rationales to explain such a Crithidia-induced difference in fat 351 body content only in newly emerged males and not in workers. First, newly emerged males and workers 352 were likely not inoculated at the same age. Indeed, workers developed in healthy colonies and were 353 inoculated at the adult stage (most likely > 2 days old) for the establishment of infected microcolonies. microcolonies is unlikely to have arisen from a difference in brood care (i.e., no significant difference in 358 pollen efficacy), we cannot rule out that infected workers displayed specific brood caring behaviour. For 359 instance, they could have added peculiar nutrients or microorganisms to larval food from their 360 hypopharyngeal and mandibular glands and/or stomach to prepare their offspring to face infection (e.g., 361 addition of sterols, Svoboda et al., 1986). Such an increase in offspring fat body content through adapted 362 larval feeding by workers could be interpreted as a trans-generational prophylactic behaviour. Indeed, 363 enhanced fat body content has been assumed to correspond to a specific allocation of resources to 364 counteract parasites by mounting an immune response (Brown et al., 2003). It would be interesting to test 365 whether infected workers provide their larvae with specific central and specialised metabolites. We should 366 however underline that the relationship between fat body content and immunity has become controversial 367 since some results have been contradictory (e.g., Brown  Heather pollen harbours kaempferol flavonoids linked to one/two coumaroyl groups which are also 378 linked to one/two hexosides (Gekière et al., in prep). Herein, we have shown that these heather flavonoids 379 reduced the total offspring production, as indicated by a decreased pollen collection and a lower pollen 380 efficacy, as well as a reduced mass of newly emerged males, thereby altering microcolony performance. 381 Indeed, drone mass is known to impact flight distances, but also reproductive abilities, affecting the 382 dissemination and reproductive success of bumble bee populations (Greenleaf et al., 2007;Amin et al., 383 2012). Such poor quality of heather pollen for the maintenance of buff-tailed bumble bee microcolonies 384 has already been pinpointed (Vanderplanck et al., 2014). While it was partly attributed to its nutritional 385 content (i.e., low concentration of amino acids and abundance of δ-7-avenasterol and δ-7-stigmasterol, 386 Huang et al., 2011, Vanderplanck et al., 2014, our study demonstrated that specialised metabolites may 387 also impact heather pollen quality, regardless of its nutritional content (i.e., central metabolites).

389
Both heather pollen and its flavonoids showed detrimental effects (i.e. reduction of offspring 390 production, pollen efficacy). However, heather flavonoids seemed to induce a higher stress response than 391 heather pollen as dilution behaviour was significantly higher in microcolonies fed the supplemented diet 392 compared to those fed the control diet (i.e., mixing behaviour to mitigate unfavourable diet properties, 393 Berenbaum & Johnson, 2015; Vanderplanck et al., 2018) while such a difference was not observed for 394 microcolonies fed the heather diet. The reason of this discrepancy is not obvious, as both diets harbour the 395 same flavonoids and should therefore lead to similar dilution behaviour. Two hypotheses could be 396 proposed to explain this difference: (i) flavonoids were more bioavailable in the supplemented diets 397 (outside pollen grains after the chemical extraction) and then more easily absorbed by the workers, which 398 ultimately reduced the diet palatability (Wang et al., 2019); and (ii) as flavonoid extract was added to the 399 control diet (i.e., willow pollen) that already contained flavonoids, the supplemented diet was richer in 400 flavonoids than the other diets, reaching a threshold that ultimately reduced the diet palatability. 401 Unfortunately, it is not possible to unravel these hypotheses without additional experiments. Another 402 observation supporting the potential toxicity of heather flavonoids is the increase in fat body content in 403 males emerging from microcolonies fed heather and supplemented diets compared to those emerging in 404 microcolonies fed the control diet. Indeed, such an increase could be interpreted as a specific allocation of 405 resources to the fat body for performing a detoxification activity (Li et al., 2019). In that way, flavonoid 406 assimilation is known to induce the activation of defence mechanisms based on cytochrome P450 407 monooxygenase, a molecule that is highly active in the fat body (Scott et al., 1998). This increase in fat 408 body content was not observed in workers, which could be explained by the different exposure to 409 flavonoids during their life stages. Indeed, workers within microcolonies mainly fed on syrup, while males 410 fed on pollen during their whole larval development and were then more exposed to specialised 411 metabolites. Moreover, it is highly possible that sensitivity to pollen specialised metabolites is higher in 412 larvae than in adults, as already demonstrated in honey bees (Lucchetti et al., 2018). 413 The complex response of parasitised bumble bees to heather pollen and its flavonoids 414 Flavonoids were associated with an increase in parasite load, which has been also observed for other

421
The nutritional stress due to heather pollen feeding could then increase the effect of Crithidia sp. that could 422 be more effective under stressful conditions (Brown et al., 2000;Brown et al., 2003).

424
We found that mortality in infected microcolony was lower in microcolonies fed the heather diet 425 compared to those fed the control diet. Although infection was not shown to have a significant impact on 426 mortality in microcolonies fed the control diet, heather pollen could then increase host tolerance to the 427 parasite but this effect is unlikely due to its flavonoid content as mortality in infected workers fed the 428 supplemented diet did not significantly differ from those fed the control diet. Regarding fat body content 429 in newly emerged males, males that emerged in microcolonies fed the heather or supplemented diets 430 displayed higher fat body content than those that emerged in microcolonies fed the control diet, but this 431 diet effect was not significant anymore in infected microcolonies, probably because, as discussed before, 432 the parasitic stress also increased fat body content in microcolonies fed the control diet. 433

Conclusion 434
How heather pollen and its specialised metabolites impact the buff-tailed bumble bee, and how they 435 modulate the interaction with its obligate gut parasite Crithidia sp. are complex questions given the 436 diversity of specialised metabolites found in the floral resources of this species. Previous studies have 437 found that heather nectar does not contain any flavonoids (Gekière et al., in prep) but protected the 438 pollinator from its parasite Crithidia sp. through callunene activity (Koch et al., 2019). In this study, we 439 found that the occurrence of flavonoids in heather pollen reduced its collection, as well as the bumble bee 440 fitness. Moreover, heather pollen did not help to counteract the parasite but rather appeared to induce an 441 additional stress that could potentially increase the parasite effect. Actually, our results complete the 442 understanding of the bumble bee-heather-parasite relationship by underlining that heather pollen is not 443 suitable to buff-tailed bumble bee performance and does not display any therapeutic effect. This study 444 highlights the complexity of the plant-pollinator interaction by illustrating the distinct roles and effects of 445 specialised metabolites found either in nectar or pollen. We strongly encourage the consideration of these 446 two floral resources in future studies investigating the medicinal effects of plant species, especially when 447 defining pollinator conservation strategies. 448

Acknowledgements 449
We would like to thank D. Evrard and L. Marin for their help in colony and microcolony maintenance, 450 as well as the numerous people that helped for microcolony dissection. We are very grateful to the 451 Laboratory of Cellular Biology (UMons) and the Laboratory of Therapeutic Chemistry and Pharmacognosy 452 (UMons) for their advice and the access to the device needed to the microscope analyses and the diet 453 preparation. We also thank François Dittlo for providing us with heather pollen, as well as the family of the 454 first author (I., C., C., and C. Tourbez) for their help. 455