Multiple pollen types were collected for the in vitro pollen feeding experiment (Table 1). Pollen of field grown maize varieties were collected by shaking flowering maize tassels in paper bags. The freshly collected maize pollen was sieved (Ø 0.32 mm). Preceding storage at −80° Celsius, the pollen was dehydrated for 24 hours at room temperature to prevent the grains to burst at freezing.

Pollen of the single transgenic Bt-maize event Mon810 (DKc7565, cultivar Novelis, Monsanto Co.) was collected on July 24^th 2008 near Kitzingen (Lower Franconia, Germany). This maize variety expresses Cry1Ab proteins for the control of stemborers such as the European corn borer Ostrinia nubilalis (Hübner) (Lepidoptera: Crambidae) at concentrations of 1–97 ng/g fresh weight (fwt) in pollen ^[25].

Pollen of a stacked Bt-variety was collected in the week of August 4^th 2008 near Braunschweig (Germany). This maize variety expresses three genes for insect resistance and one gene for herbicide tolerance and was obtained by a traditional cross of the maize varieties Mon89034 and Mon88017. Line Mon89034 confers resistance to a wide range of butterflies and moths, such as the fall armyworm (Spodoptera sp.), the black cutworm (Agrotis ipsilon), the european corn borer (Ostrinia nubilalis) and the corn earworm (Helicoverpa zea) by expression of the Bt-proteins Cry1A.105 at a level of mean 4.24 µg/g (n = 16, fwt in pollen, Sauer and Jehle, pers. comm.) and Cry2Ab2 at a level of mean 1.19 µg/g (n = 16, fwt in pollen, Sauer and Jehle, pers. comm.). Cry1A.105 is a chimeric gene synthesized by combining 4 native Bt-gene domains of cry1Ab, cry1F and cry1Ac^[26]. This chimeric protein provides an increased activity against lepidopteran species compared to the original Cry1Ab protein as expressed in Mon810. The other parental line, Mon88017 (DKc5143), confers resistance to coleopteran pests, the Western, Northern and Mexican corn rootworms Diabrotica spp. (Coleoptera: Chrysomelidae) by the expression of the Bt-protein Cry3Bb1 at levels of mean 6.95 µg/g (n = 16, fwt in pollen, Sauer and Jehle, pers. comm.) (trademark YieldGard ® Rootworm). Mon88017 also expresses an Agrobacterium sp. CP4 derived 5-enolpyruvylshikimate-3-phosphate synthase (CP4 epsps) confering tolerance against glyphosate, the active ingredient of the herbicide Roundup (trademark Roundup Ready®) at an expression level of 170 µg/g (fwt in pollen; www.gmo-compass.org/pdf/regulation/maize/MON89034MON88017_application.pdf).

Stacked Bt-maize pollen and also control pollen of three conventional maize varieties was collected in the week of August 4^th 2008 near Braunschweig (Germany). These maize varieties were grown on an experimental field in a randomized block-design with eight replications. Samples were collected from all 30×40 m subplots, pooling the pollen into one representative sample per variety. The non-GM variety DKc5340 (Monsanto Co.) is near-isogenic to the tested stacked Bt-maize variety, DKc4250 (Monsanto Co.) is more distantly related and Benicia (Pioneer HiBred, Johnston, Iowa, USA) is totally unrelated to the stacked event (Table 1).

Pollen of the neotropical plant Heliconia rostrata was collected June 23^rd 2009 from the greenhouse in the botanical garden of the University of Bayreuth (Upper-Franconia, Germany). The Heliconia family is known to have chemical defenses against herbivores ^[27] and anecdotal brood mortality is known for Heliconia foraging honey bee colonies. The pollen of the flowers was collected in a 1.5 ml tube by shaking and scraping pollen from the anthers with a scalpel (45 mg pollen from 41 flowers).

The rearing of larvae upon hatchment under laboratory conditions was performed following the protocols by Aupinel et al. and Hendriksma et al. ^[22], ^[23]. Six donor honey bee colonies were selected from different Upper-Franconian apiaries, choosing naturally mated non-sibling queens (Apis mellifera carnica). By means of an excluder lid, the queens were trapped within their colonies on artificial combs (Nicoplast^©) (day 1; D1, 25^th June 2009). After 91 hours, without grafting manipulation, larvae within plastic queen cups were collected from the combs. Considering a 72 hours development time of the embryos until the hatchment of eggs, the larvae had a mean chronological age of 9:30 hours (D4; min. 0 to max. 19 hours old) and were typically first instars ^[23].

The subsequent laboratory rearing was performed with larvae in queen-cups mounted in culture plates, placed in a hermetic plexiglass desiccator within an incubator at 35° Celsius. The larvae were fed once a day over D4 to D9 with a 10-, 10- 20-, 30-, 40-, 50-µl semi-artificial diet, respectively ^[22]. The daily diets were administered with pipettes, adding each new diet into the diet in which individual larvae were floating. Each larva was fed the total amount of 160 µl, since no diet was removed during or after the feeding period. The diet consisted of 50% royal jelly (Le Rucher du Buzard certified organic apiary, Sospel, France) mixed with a 50% aqueous solution. The yeast extract/glucose/fructose proportion in the aqueous solutions was respectively 2/12/12 percent at D4 and D5; 3/15/15 percent at D6; and 4/18/18 percent at D7, D8 and D9 ^[22]. During larval development, relative humidity in the incubator was kept at 96% using a saturated solution of K₂SO₄. Further development upon hatching took place in 80% humidity, maintained using a saturated solution of NaCl. The survival of larvae preceding treatment was 97% (D4 to D6). For more details about the method please see Hendriksma et al. 2011 ^[23].

For each treatment, a stock solution of 50 mg pollen per 500 µl D6-diet was made. This application is in agreement with empirical findings that the food of larvae contains pollen from the third instar stage onwards (D6) ^[28], ^[29]. In this way each 20 µl treatment diet contained a 2 mg pollen dose per larvae ^[14]. Mean pollen numbers per dose were obtained by 8 sample counts per stock solution using a Neubauer improved counting chamber and a light-microscope (Table 1). The larvae were only once given a dietary pollen dose (D6). Because the larvae did not finish their daily dietary amounts within 24 h, the pollen were consumed over the remaining total exposure time, until the diet was completely finished at the non-feeding days D10/D11. The maize pollen varieties were tested on N = 20 larvae per treatment (5 colonies×4 larvae). Heliconia pollen and a no-pollen control treatment were performed on N = 10 and N = 12 larvae respectively (Table 1).

The survival of larvae during the experiment was noted daily, to assess possible lethal effects of Bt-maize pollen during the 120 hours of dietary exposure. By weighing the prepupae after defecation (D11), a potential sublethal effect was monitored. As larvae defecate and molt their intestine at this stage, both the exposure and the potential Bt-protein-receptor based mechanism are physically terminated. Hence, the effective gain in weight can only measured after defecation. Every prepupa was transplanted with soft metal tweezers into a new clean cell on an analytical microbalance to measure the weight to the nearest 0.001 g.

The data were analyzed with mixed models using different packages of the open source statistic software R version 2.11.1 ^[30]. The identity of the replicate donor colonies was included as a random factor in the models to take the non-independence of larvae from individual colonies into account ^[23].

Prepupae weights were analyzed with linear mixed effects models using the package nlme^[31]. The survival dynamics of larvae were analyzed with Cox proportional hazards regression models ^[32] using the R packages survival and survnnet^[33], ^[34]. A dynamic survival analysis is not applicable when all individuals of a group survive; in that case a Chi-square analysis was used.

Three test levels were considered. An overall sort-effect was tested over the five maize varieties. All treatments were also tested individually, paired to one another, to indicate sort effects. The significance of P values (α = 0.05) of multiple comparison were determined with an α-correction using the sequential Holm-Bonferroni procedure ^[35]. In case of no detectable difference, the treatment comparisons were summarized by evaluating the pooled data on Bt-maize pollen with control maize pollen data, also pooled.

All 40 larvae fed with Bt-maize pollen survived the 120 hours of dietary exposure upon the prepupae phase (Fig. 1). The survival rate of the conventional maize pollen fed larvae did not differ significantly from Bt-maize pollen fed larvae {C: 56 out of 59; 95%} (Chisq = 0.72, df = 1, P = 0.40). Of all the maize pollen fed larvae (N = 99), in total 97% survived until the prepupal phase. Specific survival rates were: for stacked Bt-maize 100%, near-isogenic line 100%, Mon810 100%, DKc4250 95%, and for Benicia 90%. Thus, no significant difference among the five maize pollen varieties was found (Chisq = 5.41, df = 4, P = 0.28).

Among the larvae fed with diets without pollen, the individual survival dynamics and the survival rate of 92% did not differ compared to larvae fed with maize pollen enriched diets (all P values≥0.64). In contrast, significantly fewer larvae survived the larval phase when they were fed with H. rostrata pollen compared to the other six treatments (P values≤0.01, all significant with an α/6 sequential Holm-Bonferoni correction) (Fig. 1).

With a mean of 142.3 mg, prepupae weights of Bt-maize pollen fed larvae were almost identical to the mean weight of conventional maize pollen fed larvae (142.6 mg; t = −0.20, df = 1, P = 0.82) (Table 2). A general variety-effect, considering possible differences between the five maize varieties, was not found (F = 0.26, df = 4, P = 0.90) thus the weight distributions of the transgenic and non-transgenic maize pollen treatments were all alike (Fig. 2).

Individual comparison shows that mean prepupae weights differed neither between stacked Bt-pollen and pollen from the near-isogenic line (t = 0.83, df = 33, P = 0.41), nor between the stacked Bt-variety and the single Bt-variety (t = 0.81, df = 34, P = 0.42) (Table 2). In contrast, H. rostrata pollen fed larvae showed a significantly lower mean prepupae weight compared to all the other treatments (mean 87.7 mg±21.0 SD; P values≤0.001) (Table 2).

One recent trend in plant biotechnology is stacking of multiple insect resistance traits in a single cultivar ^[6]. Honey bees are exposed to mass flowering GM crops and not a single published study deals with the effect of stacked Bt-cultivars on bees. The results of this study did not indicate adverse effects of the consumption of single and stacked Bt-maize pollen on the survival and prepupae weight of in vitro reared A. mellifera larvae. At a realistic exposure dose, the 120 h survivorship of Bt-pollen treated larvae was 100% until the prepupae phase (Fig. 1). At the prepupae stage, where larvae had terminated feeding, digesting and growing, were no indications of a sublethal Bt-pollen effect on the weight of the prepupae (Fig. 2).

The outcome of our data on stacked Bt effects are in line with earlier brood tests under colony conditions on single insect resistant Bt-maize pollen ^[36] or single purified Bt-proteins ^[7], ^[37]. In contrast to these colony level studies, the current results are achieved by testing under controlled laboratory conditions, with minimum control mortality. Compared to single Bt-proteins in pollen or in purified form, our plant produced stacked Bt-proteins, with the chimeric Bt-protein Cry1A.105, indicate a similar level of safety. In accordance, a stacked maize variety, expressing Bt proteins VIP3A and Cry1Ab, also caused no adverse effects on the biodiversity of arthropods during a 3 year ERA field experiment ^[38]. A stacked cotton cultivar, expressing cowpea trypsin inhibitor (CpTI) and Cry1Ac in pollen carried no lethal risk for honey bees, though a worst case feeding regime did cause feeding inhibition ^[39]. However, in studies comparing Cry1 with transgenic protease inhibitors, it was found that only the latter was causing reduced feeding effects ^[12], ^[40], ^[41], ^[42].

The stacking of insect resistance traits in one crop aims to enhance the effectiveness towards target pest insects, to cause an additive or synergistic toxicity. Among target pest insects, synergistic effects between e.g. Cry1Ab, Cry1Ac, Cry1F and/or Cry2Ab2 have been reported ^[43], ^[44]. Involved in toxicant synergies are mostly uptake, transportation or degradation pathways ^[45], causing a higher toxicity and a lower selectivity. Hence, potential synergistic effects on non-target insect also deserve consideration. The honey bee, a key non-target insect, has never shown lethality to Bt-proteins ^[7] and our data support the notion that, synergistic effects by stacking Bt-proteins at plant produced levels are unlikely a risk to bees. However, sublethal effects ^[46] on feeding, learning performance and foraging behavior might occur ^[47], ^[48]. Indeed, the in vitro approach covers the opportunity of testing of potential sublethal effects, by a subsequent behavioral tests on hatched bees ^[49].

In order to examine a potential effect of increased protein expression levels, two Bt-maize varieties with different expression levels were compared. Bt-maize variety Mon89034×Mon88017 has compared to Mon810 a 10² to 10⁴ times increased Bt-protein expression level in pollen (see material and methods). Hypothetically, Mon810 could have had Bt-protein levels under a toxic threshold, but the larvae remain unharmed by the multifold Bt-protein of stacked Bt-maize pollen.

The current bioassay tests GM plant material directly and realistically, by reflecting a natural consumption and digestion of pollen by A. mellifera larvae. It closes an important knowledge gap between in vivo colony experiments ^[7], ^[14], ^[36], ^[37], ^[38] and in vitro experiments with purified transgenic proteins ^[50], ^[51], ^[52]. Although purified proteins are ideal to test worst case exposure scenarios ^[9], ^[48], the E. coli produced purified Bt-substances do not represent a field situation. And although field experiments have realistic pollen exposure conditions, a down-side is a variety of uncontrolled environmental factors. In addition, pollinator field studies have to be synchronized to the flowering period and they are space and time consuming and therefore relatively costly ^[53]. Finally, within a bee colony many factors such as colony size, diseases, and nutrition could have an influence on the brood development. The presented robust bioassay minimizes any environmental effect on larval development and allows a good control of dietary pollen amounts (Table 1).

The conventional non GM maize cultivars (Table 1) allow a secure assessment of the impact of the introduced transgenic traits ^[54]. It makes assessments comprehensive, since it enables a reliable estimate of naturally occurring variation within the crop species. Though having tested the total of five maize varieties, no maize-sort related differences were found. Nevertheless, the toxic control treatment and the power analysis indicated that monitoring discernable effects of pollen on honey bee larvae was effective (Fig. S1, Fig. S2, Power analysis S1). At the given sample sizes, the test was able to distinguish effects on both the survival and the weight endpoint.

The functionality of the pollen bioassay is proven by the feeding dose of 1600 H. rostrata pollen, which caused significant lethal and sublethal effects on larvae. This dose caused 50% of the larvae to die in 72 hours (LT₅₀) and 100% to die in 7 days (LT₁₀₀). The results demonstrate the usefulness of positive controls in order to i) validate the ingestion of pollen treatments, ii) to demonstrate the capacity of detecting treatment effects and iii) to allow comparisons with other studies ^[9].

Precise and robust ERA methods are needed for honey bees ^[55]. Our bioassay is well suited to monitor environmental pollution of pollen or natural pollen toxicity (Pictures S1). Of genuine concern are systemic, lipophilic chemicals (e.g. neonicotinoids) as used in agriculture, because the plant pollen are a carrier of pesticides into honeybee colonies. Such pesticides may cause (sub-) lethal effects and can be extremely persistent ^[46]. Our pollen test is widely applicable and it fits international tiered risk assessment schemes for regulatory biosafety assessments of any new transgenic trait. Hence, we propose the in vitro bioassay for consideration as a standard pre-release test for all polleniferous transgenic crops.