Stacked Bt maize and arthropod predators: exposure to insecticidal Cry proteins and potential hazards

Abstract

Genetically engineered (GE) crops with stacked insecticidal traits expose arthropods to multiple Cry proteins from Bacillus thuringiensis (Bt). One concern is that the different Cry proteins may interact and lead to unexpected adverse effects on non-target species. Bi- and tri-trophic experiments with SmartStax maize, herbivorous spider mites (Tetranychus urticae), aphids (Rhopalosiphum padi), predatory spiders (Phylloneta impressa), ladybeetles (Harmonia axyridis) and lacewings (Chrysoperla carnea) were conducted. Cry1A.105, Cry1F, Cry3Bb1 and Cry34Ab1 moved in a similar pattern through the arthropod food chain. By contrast, Cry2Ab2 had highest concentrations in maize leaves, but lowest in pollen, and lowest acquisition rates by herbivores and predators. While spider mites contained Cry protein concentrations exceeding the values in leaves (except Cry2Ab2), aphids contained only traces of some Cry protein. Predators contained lower concentrations than their food. Among the different predators, ladybeetle larvae showed higher concentrations than lacewing larvae and juvenile spiders. Acute effects of SmartStax maize on predator survival, development and weight were not observed. The study thus provides evidence that the different Cry proteins do not interact in a way that poses a risk to the investigated non-target species under controlled laboratory conditions.

1. Introduction

Commercialized insect-resistant genetically engineered (GE) crops express genes from the bacterium Bacillus thuringiensis (Bt) that encode crystalline proteins. Those Cry proteins provide protection against certain Lepidoptera or Coleoptera pests. One concern with the growing of Bt crops is that valued non-target species might be harmed. Those include natural enemies that help to control herbivores in crop fields and thus contribute to sustainable pest management [1,2]. The environmental risk assessment that precedes any release of GE crops for cultivation addresses the toxicity of the produced insecticidal compound for non-target species (hazard) as well as the concentrations that non-targets ingest in the field (exposure) [3].

Many studies used for regulatory dossiers as well as studies from public sector scientists on non-target effects of Bt crops have been conducted in the past 20+ years. Those include laboratory studies with purified Cry proteins or GE plant material and field studies. Meta-analyses of the large body of data concluded that beneficial arthropods, including predators, are not adversely affected by the current Bt crops and the Cry proteins that they produce [4–6]. Laboratory studies measuring the flow of Cry proteins in the food chain confirmed that most herbivores ingest the Cry proteins and pass them on to their natural enemies [7–9]. The Cry proteins are, however, diluted from lower to higher trophic levels because of excretion and digestion. The presence of Cry proteins in herbivores and, at lower levels, in natural enemies has been confirmed by field collections in different Bt crops [10–14].

Most of the previously published studies worked with individual purified Cry proteins or plants producing one Cry protein. In modern Bt crops, however, several Cry and other Bt proteins have been stacked, either by simultaneously transferring multiple genes or by conventional crossing of plants containing individual transgenes. The production of multiple insecticidal proteins with different modes of action increases the efficacy and target pest spectrum and reduces the probability of resistance evolution in the target pests [15]. In addition, plants with multiple traits, such as herbicide tolerance and insect resistance, have become increasingly important and occupied 41% of the global 185 million hectares of GM crops in 2016 [16]. While stacked Bt crops expose arthropods to multiple Cry proteins simultaneously, the exposure of herbivores to the different Cry proteins depends on the expression levels in the consumed plant tissues. Subsequently, exposure of predators to the individual Cry proteins depends on the concentrations in the consumed herbivores, which are also influenced by degradation and digestion rates. One concern with stacked Bt crops is that the different Cry proteins in the food of herbivores and predators could interact in a synergistic rather than additive way, which may lead to unexpected and unpredictable effects on non-target species, even if they are not susceptible to the individual Cry proteins [17].

In 2010, SmartStax was commercialized in the USA and Canada. This maize product combines six insect-protection genes and two herbicide-tolerance genes and provides the unique opportunity to study the flow and fate as well as potential non-target effects of multiple Cry proteins simultaneously. This is an advantage over previous studies with individual Cry proteins, where comparisons on the protein fate and flow through the arthropod foodweb could only be done indirectly, suffering from differences in study design and execution. SmartStax produces Cry34Ab1, Cry35Ab1 and Cry3Bb1 that are active against Diabrotica species (Coleoptera: Chrysomelidae) and Cry1F, Cry1A.105 and Cry2Ab2 that provide protection against various lepidopteran pests [18,19]. This study reports bi- and tritrophic experiments with SmartStax maize, two herbivores with different modes of feeding (spider mite and aphid) and three generalist predators (spider, ladybeetle and lacewing). The aims of the study were: (i) to measure and compare the movement of the different Cry proteins from the plant to higher trophic levels and (ii) to investigate if ingestion of multiple Cry proteins leads to acute lethal or sublethal effects in the predators.

2. Material and methods

(a). Plants and insects

SmartStax^® maize (Monsanto, St Louis, USA, referred to as ‘Bt maize’) and the genetically closest conventional hybrid EXP258 (non-Bt maize) were grown in a glasshouse according to Meissle et al. [20]. Maize pollen was used as a plant-derived food source that can be directly consumed by the predators under study. When plants started to flower, pollen was collected every other day, dried at ambient conditions for 2 days, sieved (mesh size: 0.2 mm) to remove anthers and contaminants, and stored at −80°C for a maximum of six months [20]. Pollen used for the experiments was pooled from several plants, which were grown in several consecutive time periods. Bt and non-Bt plants were always grown in parallel. Before pollen was used for the experiments, it was rehydrated for 2–3 h in a plastic box (13 × 10 × 5 cm) lined with a wet paper towel.

Spider mites, Tetranychus urticae (Acari, Prostigmata: Tetranychidae), from a starting culture provided by Syngenta (Stein, Switzerland), were reared on maize plants in a climate chamber with fluctuating daily temperature (22–28°C) and humidity (85–45% RH) and a 16 L : 8 D regime. A starting culture of aphids, Rhopalosiphum padi (Hemiptera: Aphididae), was also provided by Syngenta. Aphids were reared on maize plants in a climate chamber at constant 25°C, 50% RH and 16 L : 8 D. Cultures of both herbivore species were maintained on Bt and non-Bt maize (approx. 10 leaves stage) in the same climate chamber, but spatially separated by 2–3 m to minimize exchange (leaves of Bt and non-Bt plants not touching each other). Spider mites were collected by beating the maize plants over a plastic tray. Aphids were collected with a paint brush. The herbivores have a different mode of feeding. While spider mites feed on mesophyll cells and are known to contain relatively high concentrations of Cry proteins, aphids feed mainly on phloem sap and ingest very little Cry protein when feeding on Bt maize [12,21].

Gravid females of the web-building spider Phylloneta impressa (Araneae: Theridiidae) were collected in conventional maize fields near the Agroscope research station (Zurich, Switzerland) in June 2014 and reared in a climate chamber according to Meissle & Romeis [12]. When juveniles dispersed, the feeding experiments were started. Eggs of the lacewing Chrysoperla carnea (Neuroptera: Chrysopidae) were obtained from our own long-term culture at Agroscope (Zurich), which was established in 1993 [22]. Eggs of the ladybeetle Harmonia axyridis (Coleoptera: Coccinellidae) were also derived from our own continuous culture, which was started with beetles collected around Zurich in 2013.

(b). Bioassays

All bioassays were conducted in a climate chamber at 25 ± 1°C, 70 ± 5% RH and 16 L : 8 D. Lacewing eggs from several females laid overnight were placed individually in plastic dishes (5 cm diameter, 1 cm high) with ventilated lids. Individual neonates were divided in six ad libitum food treatments: (i) aphids from Bt maize, (ii) aphids from non-Bt maize, (iii) spider mites from Bt maize, (iv) spider mites from non-Bt maize, (v) pollen from Bt maize and (vi) pollen from non-Bt maize. Depending on the size of the predators, each feeding consisted of 20–50 aphids, 100–300 spider mites or 5–15 mg pollen. In the spider mites and pollen treatments, a drop of water was provided together with the food. Aphids contained sufficient water. Every day, new food was provided and larval stage and mortality of C. carnea larvae were recorded. Larvae reaching a new instar were weighed. Dishes were changed or cleaned every other day. Final (third) instars were fed for two more days and on the following day, they were weighed again and frozen at −80°C. A total of 65–81 individuals (replicates) were set up for each of the six food treatments. To obtain this total number, the experiment was conducted in two consecutive time periods with approximately half of the replicates in each period. For Cry protein measurements, three lacewing larvae from the same treatment and time period were pooled to one sample, resulting in a total number of 19 analysed samples for the Bt aphid treatment, eight for the Bt pollen treatment and 20 for the Bt spider mite treatment. From the respective non-Bt treatments, two to four samples were analysed to test for potential matrix effects. A further experiment was set up in the same way as described above except that larvae were weighed and frozen on the third day after reaching the second instar (N = 33–36 per treatment). This was done to examine if the Cry protein concentrations in the larvae change during the larval development period. Ten samples of second instars (three individuals pooled) were analysed in the Bt aphid treatment, 10 in the Bt pollen treatment, five in the Bt spider mite treatment and two to five in the respective non-Bt treatments.

A similar experiment was performed with ladybeetle larvae. Neonates that had left their egg batch and started searching for food were placed individually in plastic dishes and fed as described previously. Larvae were frozen 1 day after reaching the third instar. This day was selected as the endpoint for the ladybeetles because it was not possible to feed them successfully to the final (fourth) instar with spider mites. In total, 41–52 individuals (replicates) were set up for each of the six food treatments described previously. To obtain this total number, replicates were conducted in two consecutive time periods for aphids as food and three time periods for pollen and spider mites as food. Cry protein measurements were conducted with individual ladybeetle larvae (40–41 larvae in each Bt food treatment and nine larvae in each non-Bt treatment).

Juvenile spiders from several mothers were distributed to the six food treatments described previously once they dispersed after the second moult. Spiders were weighed and kept individually in ventilated transparent plastic cylinders (2.8 cm diameter, 3 cm high), where they produced webs. Food was provided every 3–4 days. In the pollen treatments, the cylinders were gently shaken to ensure that pollen got stuck in the spider webs. Mortality was recorded whenever food was provided. After four weeks, all cylinders were cleaned. One day after the 11th feeding event (day 37), the spiders were weighed and frozen at −80°C. After this (long) feeding period, six (pollen) and 23–26 (aphids, spider mites) individuals per treatment survived. Additional spiders were fed as described previously, but collected between days 7 and 12 (short feeding period, 10–17 individuals per treatment) or between days 20 and 26 (medium feeding period, 8–34 individuals per treatment). Overall, 51 spiders were fed with aphids from each Bt and non-Bt maize, 80–81 with spider mites and 87–94 with pollen. Those replicates were set up in a period of two months. Because juvenile spiders were rather small, three to seven surviving individuals from the same treatment and mother were combined to one sample for Cry protein measurements. This resulted in 10 samples from the Bt aphid treatment, seven from the Bt pollen treatment, 18 from the Bt spider mite treatment and a similar number from the respective non-Bt treatments.

Throughout the period of the bioassays with the three predators, samples of the food items were taken. From Bt maize, 31 samples of spider mites (approx. 500 individuals per sample), 25 samples of aphids (approx. 100 individuals per sample) and 44 subsamples of the pooled maize pollen (approx. 10 mg per sample) were collected. In addition, 21 spider mite samples, 12 aphid samples and 26 pollen subsamples were collected from non-Bt maize. Maize leaf samples were collected in the glasshouse when 10–11 unfolded true leaves were present. One square centimetre was cut from the middle part of each leaf. Altogether, 108 leaf-samples from 11 Bt plants and 12 samples from three non-Bt plants were collected. All samples were weighed and frozen at −80°C.

(c). Quantification of Cry proteins

Samples from plants (leaves, pollen), herbivores (aphids, spider mites) and predators (lacewings, ladybeetles, spiders) were lyophilized in a CentriVap (Labconco Corporation, Kansas City, MO, USA) for 24–48 h depending on sample weight and the dry weight was determined. Leaf samples were cut into small pieces using a pair of scissors. For protein extraction, 600–1000 μl of PBS (phosphate-buffered saline) with 0.55% TWEEN-20 and one 3 mm tungsten carbide ball was added to each sample and tissues were macerated in a bead mill (MM300, Retsch, Haan, Germany) for 2–3 min at 25 Hz. After centrifugation at 13 000g for 5 min at 4°C, the supernatants were diluted with extraction buffer according to the expected concentrations of Cry proteins. Samples were kept on ice throughout processing. Cry proteins were quantified with commercial double antibody sandwich enzyme-linked immunosorbent assay (DAS-ELISA) kits (Agdia, Elkhart, IN, USA). Individual kits were used for the detection of Cry1A.105 (Cry1Ab/1Ac kit), Cry1F and Cry34Ab1. Cry2Ab2 and Cry3Bb1 were measured either with a dual kit (lacewings, spiders) or with individual kits (ladybeetles). The dual kit was chosen initially to save costs and effort, but was replaced by individual kits later in the project for more flexibility with dilutions. One of the Cry proteins produced by SmartStax, Cry35Ab1, could not be measured because no kit was available. Purified Cry proteins of certified quality provided by Monsanto or Dow AgroSciences (Indianapolis, IN, USA) were used to construct standard curves with nine concentrations.

All antibody-coated plates were loaded with the appropriate enzyme conjugates, diluted sample extracts and Cry protein standards according to the manufacture's protocol. The plates were incubated overnight at 4°C and washed seven times with PBS 0.05% TWEEN-20. The colour reagens were added according to protocol and optical densities were measured after 20 min with a plate reader (Spectrafluor-Plus, Tecan, Switzerland). The concentrations of Cry proteins were determined with the standard curves using regression analyses based on a hyperbola model. The limit of detection (LOD) for the clear separation of positive readings from the variation of controls was calculated according to Meissle & Romeis [12]. The lack of cross-reactivity of ELISA plates was verified by comparing measured concentrations in a mixture of all purified Cry proteins with the individual solutions.

(d). Data analysis

Concentrations of Cry proteins are presented as medians. Cry protein concentrations in spiders did not change consistently with feeding duration. Therefore, the values from the short, medium and long feeding periods were analysed together. Data for the different leaves of maize plants were also pooled. Statistical analyses of life table parameters were performed with Statistica 12 (StatSoft, Tulsa, OK, USA). Mortality data for each predator species were analysed separately by treatment (pollen, aphids, spider mites) using log-rank tests. Data for all feeding durations were included in the survival analyses and individuals destructively sampled for Cry protein measurements were treated as censored (missing) from the date of removal. Weight data of newly moulted second and third instar lacewings and ladybeetles were log transformed and analysed with two-way ANOVAs (independent variables: food and Bt). Owing to high mortality of spiders in some food treatments (in particular pollen) and the limited availability of juveniles, the spiders were not homogeneously distributed among the different food treatments and feeding durations. As a consequence, the initial weight of juvenile spiders was statistically different among food types and feeding durations (three-way ANOVA, full factorial model, independent variables: duration, food, Bt). There was, however, no difference in initial spider weight between Bt and non-Bt treatments for any food type (p = 0.7). Therefore, the weight change rather than the absolute weight after the feeding period was calculated (weight when frozen/weight at start), log-transformed and analysed with a three-way ANOVA (feeding duration, food, Bt). For lacewings and ladybeetles, the number of days from hatching to second and third instar was analysed with generalized linear models using a Poisson distribution.

3. Results

(a). Transfer of Cry proteins along the trophic chain

The highest median Cry protein concentration in Bt maize leaves was 125.4 µg g⁻¹ dry weight Cry2Ab2 and the lowest was 17.6 µg g⁻¹ Cry1F (figure 1). Concentrations in Bt maize pollen were generally lower than in leaves except for Cry1F (32.6 µg g⁻¹). The lowest median in Bt maize pollen was measured for Cry2Ab2 (0.46 µg g⁻¹). Aphids feeding on Bt maize contained no measurable amounts of Cry1A.105 and Cry2Ab2, and concentrations lower than 0.3% of the leaves for Cry1F, Cry3Bb1 and Cry34Ab1. Spider mites feeding on Bt maize generally contained higher median concentrations than the leaves (140–200%) except for Cry2Ab2 (13%). Estimated median Cry protein concentrations in plant material and herbivores from non-Bt maize were lower than 0.05 µg g⁻¹, indicating very limited exchange of herbivores in the climate chamber cultures, or cross-contamination of samples between Bt and non-Bt maize.

Figure 1.

Median concentrations of (a) Cry1A.105, (b) Cry1F, (c) Cry2Ab2, (d) Cry3Bb1 and (e) Cry34Ab1 in SmartStax maize leaves and pollen, in spider mites (T. urticae) and aphids (R. padi) feeding on whole maize plants, and in lacewing larvae (C. carnea, second or third instar), ladybeetle larvae (H. axyridis, third instar) and juvenile spiders (P. impressa) feeding on maize pollen, aphids or spider mites. Median Cry protein concentrations in µg g⁻¹ dry weight are presented in the boxes. Colour codes were assigned to each order of magnitude to improve visibility. Arrow thickness represents the percentage of Cry protein in the higher compared with the next lower trophic level.

Predators fed with aphids contained either no detectable Cry protein or values lower than 0.1 µg g⁻¹. At those low concentrations, however, several medians of the estimated Cry-protein concentrations of predators were in the same order of magnitude as the estimates for the aphids. The highest Cry protein concentration in the aphid treatments was measured for Cry34Ab1 in ladybeetles. Ladybeetle larvae generally contained the highest concentrations of all predators in the different food treatments. In the pollen and spider mite treatments, ladybeetle larvae contained 20–71% of the concentration of Cry proteins found in the food except for Cry2Ab2, where larvae contained only 1 and 9% of the concentration found in spider mites and pollen, respectively. Second and third instar lacewings contained Cry proteins at similar levels. However, Cry3Bb1 was higher in third instars fed with pollen and Cry34Ab1 was higher in second instars fed with spider mites. The highest measured median concentration in lacewings was Cry34Ab1 in third instars fed with pollen (23% of the pollen value). By contrast, low concentrations of Cry2Ab2 were detected when third instars were fed with spider mites (0.2% of the spider mite value) and no Cry2Ab2 was measurable in lacewings feeding on pollen. Cry protein concentrations in spiders were of the same order of magnitude as those in lacewings or lower. No Cry1A.105, Cry2Ab2 or Cry3Bb1 was detected in the pollen treatment.

Most predator samples from the non-Bt treatments contained no measurable Cry protein (median below the LOD). Trace amounts (median of 0.005 µg g⁻¹) were measured for Cry1F in lacewings fed with non-Bt pollen or spider mites, and for Cry34Ab1 in spiders fed with aphids or spider mites from non-Bt plants. ELISA measurements indicated the presence of Cry 1A.105, Cry1F, Cry3Bb1 and Cry34Ab1 in ladybeetles fed with spider mites (0.14–0.45 µg g⁻¹) and of Cry34Ab1 in ladybeetles fed with aphids (0.023 µg g⁻¹) or pollen (0.23 µg g⁻¹) from non-Bt maize.

(b). Acute toxic effects of Bt maize on predators

Juvenile spiders, lacewing larvae and ladybeetle larvae fed with aphids, spider mites or pollen from Bt maize did not show significant differences in mortality, weight or weight change, or development time compared with those fed with food from non-Bt maize (table 1; electronic supplementary material, table S1). By contrast, the type of food was significant in all statistical comparisons (p < 0.05). Juvenile spiders increased body weight when fed with aphids or spider mites, but lost weight when fed with pollen. There was no difference in weight change among feeding durations in the pollen and aphid treatments (Tukey's HSD, p > 0.05), indicating that the spider weights plateaued after the first one or two weeks. Spiders fed with spider mites for the long period, however, had a larger weight increase than those fed for the short or medium period (p < 0.001). Spider mortality was low in the aphid and spider mite treatments and high in the pollen treatment. Similarly, lacewing and ladybeetle larvae developed best when fed with aphids (highest weight, lowest development time, lowest mortality), intermediate with spider mites and worst with pollen. There was no significant interaction of the factors Bt (Bt versus non-Bt) and food type in any of the statistical analyses (p > 0.05).

Table 1.

Life table parameters of lacewing larvae (C. carnea), ladybeetle larvae (H. axyridis) and juvenile spiders (P. impressa) in bi- and tritrophic feeding experiments with SmartStax Bt maize or the genetically closest non-Bt maize.

		mean ± s.e. (N)		significance
parametera	foodb	control	Bt	Btc	foodd
P. impressa juveniles
weight increase (long)	aphids	2.24 ± 0.186 (26)	2.14 ± 0.127 (25)	n.s.	B
	spider mites	3.58 ± 0.532 (23)	3.47 ± 0.397 (26)		A
	pollen	0.72 ± 0.086 (6)	0.69 ± 0.082 (6)		C
weight increase (medium)	aphids	2.01 ± 0.296 (8)	1.28 ± 0.238 (8)		B
	spider mites	2.58 ± 0.454 (33)	2.30 ± 0.346 (34)		B
	pollen	0.58 ± 0.051 (8)	0.66 ± 0.080 (11)		C
weight increase (short)	aphids	2.48 ± 0.231 (11)	2.66 ± 0.256 (10)		AB
	spider mites	2.05 ± 0.251 (17)	2.33 ± 0.537 (17)		B
	pollen	0.77 ± 0.044 (10)	0.65 ± 0.097 (15)		C
mortality	aphids	5.88% (51)	3.92% (51)	n.s.	—
	spider mites	6.17% (81)	2.50% (80)	n.s.	—
	pollen	72.40% (87)	65.96% (94)	n.s.	—
C. carnea larvae
weight second instar	aphids	1.31 ± 0.050 mg (95)	1.16 ± 0.039 mg (93)	n.s.	A
	spider mites	0.89 ± 0.035 mg (74)	0.83 ± 0.030 mg (81)		B
	pollen	0.74 ± 0.017 mg (83)	0.77 ± 0.017 mg (78)		C
weight third instar	aphids	3.99 ± 0.160 mg (64)	3.91 ± 0.151 mg (58)	n.s.	A
	spider mites	3.19 ± 0.078 mg (60)	3.18 ± 0.064 mg (63)		B
	pollen	2.93 ± 0.099 mg (19)	3.10 ± 0.085 mg (24)		B
time hatching to second instar	aphids	3.88 ± 0.049 days (95)	3.94 ± 0.040 days (93)	n.s.	C
	spider mites	5.17 ± 0.163 days (75)	5.24 ± 0.141 days (82)		B
	pollen	5.99 ± 0.156 days (85)	5.94 ± 0.155 days (79)		A
time hatching to third instar	aphids	6.88 ± 0.098 days (64)	7.03 ± 0.097 days (59)	n.s.	C
	spider mites	10.07 ± 0.163 days (60)	9.81 ± 0.152 days (62)		B
	pollen	12.40 ± 0.489 days (20)	13.54 ± 0.605 days (24)		A
mortality	aphids	5.10% (98)	8.16% (98)	n.s.	—
	spider mites	34.58% (107)	26.85% (108)	n.s.	—
	pollen	60.68% (117)	51.28% (117)	n.s.	—
H. axyridis larvae
weight second instar	aphids	1.84 ± 0.150 mg (37)	1.65 ± 0.090 mg (39)	n.s.	A
	spider mites	1.12 ± 0.065 mg (41)	1.13 ± 0.065 mg (46)		B
	pollen	0.96 ± 0.041 mg (47)	1.00 ± 0.044 mg (48)		B
weight third instar	aphids	7.06 ± 0.434 mg (36)	6.43 ± 0.466 mg (37)	n.s.	A
	spider mites	3.07 ± 0.151 mg (41)	3.17 ± 0.151 mg (38)		B
	pollen	3.17 ± 0.163 mg (33)	3.12 ± 0.199 mg (39)		B
time hatching to second instar	aphids	2.00 ± 0.039 days (37)	2.05 ± 0.036 days (39)	n.s.	C
	spider mites	3.15 ± 0.089 days (41)	3.00 ± 0.082 days (46)		B
	pollen	4.00 ± 0.129 days (47)	3.69 ± 0.130 days (48)		A
time hatching to third instar	aphids	3.46 ± 0.085 days (35)	3.44 ± 0.084 days (36)	n.s.	C
	spider mites	5.53 ± 0.154 days (38)	5.37 ± 0.133 days (38)		B
	pollen	6.47 ± 0.135 days (32)	6.44 ± 0.155 days (39)		A
mortality	aphids	7.32 (41)	0 (41)	n.s.	—
	spider mites	15.69 (51)	15.69 (51)	n.s.	—
	pollen	7.69 (52)	17.31 (52)	n.s.	—

^aWeight increase for P. impressa = end weight/start weight for long (37 days), medium (20–26 days) or short (7–12 days) feeding duration. Weight increase (log transformed) was analysed with three-way ANOVA (factors feeding duration, food, Bt). Feeding duration and the food × duration interaction were significant. Weight of second and third instars for C. carnea and H. axyridis was analysed with two-way ANOVA (factors food, Bt). Mortality was analysed with log-rank tests for each food treatment (factor Bt). Time data were analysed with generalized linear models with Poisson distribution and log-link function (factors food, Bt).

^bPredatory species were fed pollen from SmartStax (Bt) or non-Bt maize, aphids (R. padi) or spider mites (T. urticae) feeding on whole Bt or non-Bt plants.

^cn.s. denotes no significance for the factor Bt (for details, see electronic supplementary material, table S1 and footnote a).

^dFood type was significant in all statistical comparisons. Different letters denote significant differences within each analysis (Tukey's HSD for weight data, 95% confidence intervals for times). No significant interactions of food type with Bt (Bt versus non-Bt) were observed in any of the analyses (for details, see electronic supplementary material, table S1).

4. Discussion

In our study, Bt maize (SmartStax) leaves contained relatively high levels of all of the five measured Cry proteins, although the concentration of Cry1F was almost one order of magnitude lower than that of Cry2Ab2. By contrast, Bt maize pollen contained concentrations of Cry1F and Cry34Ab1 that were similar to leaves, while the values for Cry1A.105 and Cry3Bb1 were one order of magnitude, and Cry2Ab2 more than two orders of magnitude lower than in leaves. Literature values of Cry protein concentrations in SmartStax are only available from regulatory documents for the USA [23]. While Cry1F and Cry2Ab2 concentrations in leaves and pollen in our study were in the range of the data reported from five field sites in the USA, our values for Cry1A.105, Cry3Bb1 and Cry34Ab1 were approximately one-half to one-third of the reported field values. Expression levels of those Cry proteins in our glasshouse-grown plants may have been lower than in field-grown plants because of different nutritional, climatic and light conditions or the use of different varieties. In addition, the antibodies in the commercial ELISA kits might have been less sensitive in capturing the Cry proteins from the tissue extracts than the kits used by the biotech companies.

Spider mites from Bt maize contained all Cry proteins, except Cry2Ab2, in higher concentrations than in leaves. High Cry protein concentrations in spider mites, of the same order of magnitude as the leaves, were reported for maize containing Cry1Ab [21,24,25] or Cry3Bb1 [12,25,26]. In Bt cotton, spider mites contained even 4–17 times more Cry1Ac than leaves [27–29], while Cry2Ab2 concentrations were lower than in leaves [27]. Spider mites suck out mesophyll cells and Dutton et al. [30] have shown that mesophyll cells contain high concentrations of Cry1Ab. Thus, spider mites ingest large amounts of the Cry protein relative to their body size. ELISA-based protein concentration estimates for spider mites cannot be directly compared with those of leaves, as ELISA-based estimates are measured per gram dry weight of the whole sample. The macerated leaf tissue also includes structures expected to contain little Cry protein, such as cell walls or vascular cells, which reduces the overall concentration encountered compared with the cell content of mesophyll cells. Our results confirm, for the five Cry proteins examined, that spider mites are the arthropods with the highest measured concentrations with the consequence that their natural enemies are also exposed to high concentrations. By contrast, aphids are known to ingest only traces of Cry protein [31]. This was also evident in our study where all five Cry proteins were measured in very small or undetectable concentrations. Aphids feed on the phloem which has been shown to transport almost no Cry1Ab protein [32]. Thus, predators feeding on aphids are exposed to negligible amounts of Cry protein. Predators feeding on aphids from Bt maize in our study often contained no measureable Cry protein, but some measurements showed low concentrations similar or even higher than those in aphids (e.g. Cry1F, Cry3Bb1 and Cry34Ab1 in ladybeetles). Compared with the Bt maize leaves, however, those concentrations represent a dilution factor of four orders of magnitude and measurements were often near the LOD of the ELISA, which resulted in relatively high variation and thus uncertainty. ELISAs of ladybeetles feeding on aphids from non-Bt maize indicated a median concentration of 0.023 µg g⁻¹ Cry34Ab1, which was higher than the value in their prey. The ELISA kits for Cry34Ab1 had the highest sensitivity compared with the other kits and they indicated positive values for several samples from non-Bt maize. The relatively high Cry protein concentration estimates in ladybeetles feeding on aphids could therefore be a result of contamination or of matrix effects, i.e. antibodies binding to compounds in the extract other than the specific Cry protein. An alternative hypothesis was recently raised by Paula & Andow [33], who suggested that H. axyridis might accumulate a certain (albeit small) amount of Cry proteins in their bodies.

Among the predator species, ladybeetles contained around one order of magnitude higher Cry protein concentrations than lacewings and spiders when consuming food from Bt maize. Between 20 and 70% of the Cry protein in spider mites and Bt maize pollen were measured in ladybeetle larvae, except for Cry2Ab2 which was present in much smaller amounts. In lacewings and spiders, concentrations ranged from less than 1 to 23%. Spiders ingest liquidized food after extra-intestinal digestion [34] and have a rather slow metabolism and growth, which could explain the comparatively low concentrations of measured Cry proteins. With their piercing-sucking mouthparts, lacewing larvae also ingest liquid food, although extra-intestinal digestion seems to play a minor role. In addition, lacewing larvae cannot excrete faeces because their midgut is not connected to the hindgut [35]. Similar to spiders, however, lacewing larvae contained comparatively low Cry protein concentrations, which may be explained by limited ingestion (the prey's gut where the Cry proteins are located may be only partly consumed) and by high proteolytic activity in the lacewing gut. By contrast, larvae of H. axyridis have chewing mouthparts that are used to consume large amounts of food items, and excrete the undigestable remains [36]. This mode of feeding might explain the comparatively high concentrations of Cry proteins.

In conclusion, the different Cry proteins measured simultaneously in our study showed a generally similar flow through the food chain. An exception was Cry2Ab2 with highest concentrations in the Bt maize leaves, lowest concentrations in Bt maize pollen and lowest acquisition rates by aphids and predators when compared with the other measured Cry proteins. Similar findings were reported for cotton containing Cry1Ac and Cry2Ab2 [27]. This demonstrates that Cry proteins differ in digestability and stability when moving through the food chain. Depending on the consumed food, predators can be exposed to Cry protein levels next to, or even higher, than in plants as shown in the spider mites treatment, or to no measureable concentrations, as shown in the aphids treatment. Thus, generalists feeding on different food sources, including Bt plant material and prey, are likely to ingest highly variable Cry protein concentrations. The average exposure of generalist predators, however, is likely to be much lower than the Cry protein levels measured in plants. In addition to the prey and food spectrum, Cry protein concentrations in predators may also depend on feeding preferences, the mode of feeding, the rate of food uptake, the physiology of digestion, digestibility of the food and finally excretion. Field collections in Bt maize [11,12], cotton [7], soya bean [13] and rice [14] confirmed that Cry protein concentrations in predators are substantially lower than in leaves and depend on the species and life stage. In any case, estimates of Cry protein concentrations based on ELISA measurements give no indication of the biological activity of the measured proteins or immuno-active fragments. However, studies with sensitive insects confirmed that the biological activity of Cry-proteins was consistent with ELISA measurements. Data are available for Cry1Ab contained in Spodoptera littoralis caterpillars, for Cry3Bb1 contained in Diabrotica v. virgifera beetles and for Cry1Ab and Cry3Bb1 in spider mites feeding on Bt maize [37,38].

Independent of the different levels of exposure through the different food sources, predators in the Bt maize treatments did not perform differently from those feeding on food from non-Bt maize in our study. Neither survival nor weight or development time were affected by Bt maize. By contrast, the food types had a different nutritional quality for the predators with pollen being of lowest value for all three predator species, confirming earlier studies with P. impressa [12] and C. carnea [20]. Consequently, our study provides evidence that the ingestion of six different Cry proteins does not lead to unexpected synergistic effects that would be expected to cause adverse impacts on the investigated non-target species. Similarly, no negative effect of SmartStax was observed on termites [39] and stacked Bt maize producing Cry1A.105, Cry2Ab2 and Cry3Bb1 did not harm honeybees [40,41]. Also field evaluations provided evidence that adverse effects on non-target arthropods did not occur when stacked Bt maize producing Cry1A.105, Cry2Ab2 and Cry3Bb1 [42], and Cry1F, Cry34Ab1 and Cry35Ab1 [43,44] was compared with the genetically closest non-Bt maize. This confirms earlier findings showing no adverse effects of single-gene Bt maize on non-target arthropods in the field [3–6,45–50].

In conclusion, our results show that different Cry proteins move through the arthropod food chain with a similar pattern, although the transfer efficiency can differ. Large differences in acquisition rates exist between arthropod species depending on the mode and site of feeding. In general, Cry protein concentrations in higher trophic levels are considerably lower than in plant material. Our study revealed no unexpected acute lethal or sublethal effects of stacked Bt maize on predators and provides evidence that synergistic interactions among individual Cry proteins that may affect beneficial non-target species do not occur.

Data accessibility

The datasets supporting this article are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.51d4h [51].

Authors' contributions

Z.S., O.S.H., J.R. and M.M. participated in the design of the study. Z.S., Y.S. and M.M. carried out the data collection and immunological measurements. M.M. performed the statistical analyses and all authors contributed to the drafting of the manuscript. All authors gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by the Scientific Exchange Programme (SciEx-NMS.CH), Project Code 13.228, and the Chinese Scholarship Council, File No. 201408440098, with fellowships for Z.S. and Y.S., respectively.