Osmia bicornis is rarely an adequate regulatory surrogate species. Comparing its acute sensitivity towards multiple insecticides with regulatory Apis mellifera endpoints

Bee species provide essential ecosystem services and maintain floral biodiversity. However, there is an ongoing decline of wild and domesticated bee species. Since agricultural pesticide use is a key driver of this process, there is a need for a protective risk assessment. To achieve a more protective registration process, two wild bee species, Osmia bicornis and Bombus terrestris, were proposed by the European Food Safety Authority as additional test surrogates. We investigated the acute toxicity (median lethal dose, LD50) of multiple commercial insecticide formulations towards the red mason bee (O. bicornis) and compared these values to honey bee (Apis mellifera) regulatory endpoints. In two thirds of all cases O. bicornis was less sensitive than the honey bee. By applying an assessment factor of 10 on the honey bee endpoint a protective level was achieved for 87% (13 out 15) of all evaluated products. Our results show that O. bicornis as a non-sensitive species is rarely an adequate additional surrogate species for lower tier risk assessment. Given the currently limited database, the honey bee seems sufficiently protective in acute scenarios as long as a reasonable assessment factor is applied. However, additional surrogate species such as O. bicornis and B. terrestris are still relevant for ecologically meaningful higher tier studies.

Bees are important pollinators of wild and cultivated flora which makes them essential 2 providers of ecosystem services and maintainers of floral biodiversity [1,2]. Aside from 3 the honey bee Apis mellifera there is a broad spectrum of wild bee species that In the European agricultural landscape, bees are exposed to a variety of pesticides 15 that target all major pests, e.g. herbicides, fungicides, insecticides [7,8]. They are not 16 only contaminated during foraging on crops but also from visitations of field-adjacent 17 wild flowers [9]. Bees are exposed to pesticides by direct overspray as well as oral uptake 18 of and contact to nectar and pollen while foraging. There are also fed contaminated 19 pollen and nectar as larvae. Furthermore, there is potential uptake of soil residues by 20 adults and larvae of soil-nesting species [10]. Moreover, consumption of non-nectar 21 fluids such as puddle water, guttation droplets or extrafloral nectar may also lead to 22 contamination [11,12]. Consequently, bee species are exposed to pesticides through 23 various environmental matrices throughout their lifespan. 24 To prevent adverse impact of pesticide applications on wild bee populations, toxic 25 effects of these substances on bee species need to be understood. However, the majority 26 of toxicity testing in laboratory and field setups has been performed using the honey 27 bee, a bred livestock species, whereas all other bee species are far less well-understood 28 in their sensitivity [10]. 29 Furthermore, the honey bee is the only pollinator species that is tested for its 30 reaction towards pesticides in the current risk assessment scheme after Regulation (EC) 31 1107/2009 [13]. However, wild bee species (i.e. bumble bees, solitary bees) may show 32 quite different responses to pesticide exposure due to differences in physiology and 33 ecology [14]. As a reaction to the information scarcity regarding the sensitivity of 34 bumble bees and solitary bees, the European Food Safety Authority (EFSA) proposed 35 the inclusion of the buff-tailed bumblebee Bombus terrestris and the red mason bee 36 Osmia bicornis into EU pesticide risk assessment as additional surrogate species [15]. 37 However, there has been reasonable doubt that these two species are adequate to 38 provide additional safety in lower tier risk assessment. Uhl [14]. They could not collect O. bicornis data but other Osmia species 49 (O. cornifrons, O. lignaria) were usually also more resistant than A. mellifera. 50 Moreover, EFSA (2013) proposed an assessment factor of 10 to account for interspecific 51 differences when testing only honey bees [15]. This approach proofed to be protective in 52 95% of cases in the meta-analysis by Arena & Sgolastra (2014) [14]. It is unclear, 53 however, if this factor would be protective for both proposed test species due to the slim 54 database of their sensitivity [16,17].

55
There is a need to assess the suitability of the new test species that EFSA proposed. 56 Only sensitive species will reduce uncertainty in lower tier risk assessment. However, 57 with the current database it is not possible to properly evaluate if the proposed species 58 are adequate. Therefore, we tested one of these two species, O. bicornis, with 59 commercial formulations of multiple common insecticides. We performed acute contact 60 toxicity laboratory tests to derive 48h contact median lethal doses (LD50s). We wanted 61 to assess the acute toxic potency of several insecticides from various classes on O. 62 bicornis. Furthermore, our goal was to compare those toxicity endpoints to honey bee 63 data from pesticide regulation. This enabled us to evaluate if O. bicornis is usually 64 more sensitive than the honey bee which would make it a suitable additional surrogate 65 species. Additionally, we examined if an assessment factor of 10 is protective when 66 July 4, 2018 2/12 comparing honey bee to O. bicornis sensitivity.

68
Insecticides 69 The majority of tested insecticides were chosen with respect to the application frequency 70 of their commercial products in apple, grapes and winter oilseed rape (Table 1) which 71 represent three main cultivation types in Germany [18]. Additionally, formulations of 72 four insecticides that are not frequently applied were included because of the following 73 reasons: Imidacloprid has been implicated as a major factor in bee decline [10].

74
Dimethoate is often used as a toxic reference in bee ecotoxicity studies. Chlorpyrifos 75 was chosen to include another organophosphate insecticide aside from dimethoate.

76
Furthermore, flupyradifurone is a relatively new insecticide with low acute toxicity 77 towards honey bees that has been applied for registration in multiple EU countries [19]. 78 Insecticides were assigned to pesticide classes according to the Compendium of Pesticide 79 Common Names [20]. Representative formulated products that contain those pesticides 80 as active ingredients (a.i.) were chosen for testing ( Table 1). Most of these formulations 81 are or were registered in Germany in recent years aside from Pyrinex ® (a.i.  Table 1 for a list of all tested formulated products and corresponding active ingredients. 85  [21]. This protocol is a precursor of a standardised testing guideline. 95 Prior to the experiments, bee cocoons were collected from the refrigerator and placed in 96 an environmental chamber at test conditions which are explained below to hatch. Male 97 bees were also collected but only used for range finding tests. Female bees' eclosion time 98 was usually between five to seven days. Afterwards hatched females were again stored at 99 4°C until one day before application. At this date, they were transferred in to the 100 environmental chamber in test cages (1 L plastic boxes sealed with a perforated lid) and 101 fed ad libitum with sucrose solution 50% (w/w) through 2 mL plastic syringes to 102 acclimatize overnight. Twenty bees were assigned to each treatment (usually 5 per cage, 103 n = 4). Please see the raw data for details on individual study setups [22]. an account of the raw data [22]. Models were chosen by visual data inspection and using 140 Akaike information criterion (AIC). Furthermore, it was ensured that appropriate model 141 were used for tests with control mortality (no fixed lower limit In two thirds of all cases O. bicornis was less sensitive than the honey bee (15 out of 163 16 insecticides could be evaluated). When applying an assessment factor of 10 on the 164 respective honey bee endpoint, it was lower than the O. bicornis endpoint for 87% of all 165 tested substances ( Table 2)   endpoints should generally also be reported in a weight-normalised format (see Table 2). 187 To create a more protective environmental risk assessment for bees, EFSA (2013) 188 proposed the inclusion of two additional wild bee species as surrogates (B. terrestris, O. 189 bicornis) [15]. These species should accompany the previous sole test species, the honey 190 bee. However, in acute toxicity testing the addition of new species is only reasonable if 191 they are generally more sensitive than the test species already in place. For two thirds 192 of the insecticides we tested O. bicornis was indeed less sensitive than the honey bee 193 ( pesticides, cadmium and arsenic [17]. Their results were inconclusive as to whether the 200 wild bee species or the honey bee was acutely more sensitive. However, they could show 201 that both newly proposed test species were less sensitive in 40% of comparisons across 202 time. When evaluating this combined information it becomes evident that O. bicornis 203 (and possibly B. terrestris) is seldomly an adequate supplementary surrogate species for 204 acute testing of pesticides since its inclusion would not provide additional safety for the 205 risk assessment process for most pesticides. As postulated by Uhl et al. (2016), test 206 species should be chosen according to their sensitivity in acute effect studies [16].

207
However, both proposed test species were selected because they are bred for 208 commercially pollination, can therefore be obtained easily in large numbers and can cope 209 well with laboratory conditions. While those criteria are important for the conduction 210 July 4, 2018 7/12 of laboratory experiments in general, they should not be decisive for the selection of 211 surrogate species. The honey bee may be a better choice in acute toxicity tests since the 212 not fully matured cuticle of young workers makes it more susceptible towards pesticides 213 compared to solitary bees [25,26]. Furthermore, there are differences in the immune 214 response of young adults. In honey bees the individual detoxification capacity is 215 relatively low and increases from thereon as they age [27,28]. However, antioxidant 216 enzyme levels already rise in O. bicornis adults before eclosion which is further evidence 217 that they are more resistant than honey bees at least at this life stage [26].

218
Furthermore, we could show that for 87% of the tested insecticides an assessment 219 factor of 10 when applied to the honey bee endpoint is sufficient to cover O. bicornis' 220 sensitivity (Fig. 1). This assessment factor was found to be protective in 95% of all 221 cases that were analysed in the meta-analysis of Arena & Sgolastra (2014) [14]. After testing as long as a reasonable assessment factor is applied [17]. However, they also note 226 that there are exceptions for some substances, e.g. neonicotinoids. Arena & Sgolastra 227 (2014) already mentioned that for this class wild bee species showed equal to higher 228 sensitivity than the honey bee [14]. This trend is also visible in our data: O. bicornis 229 was more sensitive towards all three tested neonicotinoids (acetamiprid, imidacloprid, 230 thiacloprid) than the honey bee ( Fig. 1).

231
Consequently, the honey bee is a sufficient surrogate species to assess acute toxicity 232 of most pesticides. In some cases (e.g. neonicotinoids) it might be necessary to increase 233 the assessment factor to >10 to achieve a proper level of safety. To distinguish these 234 substances that are relatively more harmful to wild bees than to honey bees, a 235 comprehensive ecotoxicological database should be established that includes a 236 representative amount of species and pesticides. Such a database would also be helpful 237 for choosing suitable additional test species if necessary. Moreover, regulatory reporting 238 standards should be improved. Our search for honey bee endpoints that were used in 239 the registration process presented quite complicated. We partly received contrasting 240 information from several sources. A solution for this problem would be the creation of a 241 transparent and publicly available database of regulatory data. Those data could be 242 then complemented by non-regulatory study results to further not only the open science 243 idea but also establish a more transparent regulation process.

244
Despite only rarely providing additional safety for lower tier risk assessment it 245 should be noted that both proposed test species may be more valuable surrogates in 246 more realistic experimental setups in higher tier risk assessment. Due to their ecological 247 differences to the honey bee, populations of O. bicornis and B. terrestris may react 248 quite differently in (semi-)field studies. Such divergent effects have been shown in a 249 Swedish field study where clothianidin/beta-cyfluthrin treatment of oilseed rape had no 250 adverse effects on honey bee colonies, yet substantial impact on O. bicornis' and 251 B. terrestris' population development [29]. Due to their properties they are good For the majority of substances we tested, the honey bee was more sensitive than 257 O. bicornis. We therefore agree with Heard et al. (2017) that A. mellifera is a sufficient 258 proxy for wild bee species in laboratory acute mortality testing [17]. However, it is still 259 necessary to investigate less well-known issues such as effects of pesticides 260 mixtures [32,33], prolonged pesticides exposure [17] or effects of pesticide adjuvants [34]. 261 An assessment factor of 10 proved to be protective for O. bicornis when applied honey 262 bee endpoints for nearly all tested insecticides. There might be exceptions (e.g. 263 neonicotinoids) where this assessment factor needs to be increased. Therefore, our study 264 provides further evidence that O. bicornis is rarely an adequate surrogate species that 265 will usually not improve lower tier risk assessment. Unnecessary acute studies with 266 non-sensitive species should not be conducted. Only sensitive species should be chosen 267 as additional surrogates to reduce overall uncertainty. However, we agree that both 268 proposed test species can be very relevant in higher tier risk assessment. In complex 269 field settings ecological differences between the honey bee, bumble bees and solitary 270 bees are more relevant as shown by Rundlöf et al. (2015) [29]. Therefore, such realistic 271 experiments are better suited to evaluate the overall impact of pesticides on wild bee 272 species. Consequently, we believe that (semi-)field data should be relied upon to a 273 greater extent than laboratory results in wild bee risk assessment.