So far, genetic engineering of plants in the context of insect pest control has involved insertion of genes that code for toxins, and may be characterized as the incorporation of biopesticides into classical plant breeding. In the context of pesticide usage in pest control, natural enemies of herbivores have received increasing attention, because carnivorous arthropods are an important component of insect pest control. However, in plant breeding programmes, natural enemies of herbivores have largely been ignored, although there are many examples that show that plant breeding affects the effectiveness of biological control. Negative influences of modified plant characteristics on carnivorous arthropods may induce population growth of new, even more harmful pest species that had no pest status prior to the pesticide treatment. Sustainable pest management will only be possible when negative effects on non-target, beneficial arthropods are minimized. In this review, we summarize the effects of insect-resistant crops and insect-resistant transgenic crops, especially Bt crops, from a food web perspective. As food web components, we distinguish target herbivores, non-target herbivores, pollinators, parasitoids and predators. Below-ground organisms such as Collembola, nematodes and earthworms should also be included in risk assessment studies, but have received little attention. The toxins produced in Bt plants retain their toxicity when bound to the soil, so accumulation of these toxins is likely to occur. Earthworms ingest the bound toxins but are not affected by them. However, earthworms may function as intermediaries through which the toxins are passed on to other trophic levels. In studies where effects of insect-resistant (Bt) plants on natural enemies were considered, positive, negative and no effects have been found. So far, most studies have concentrated on natural enemies of target herbivores. However, Bt toxins are structurally rearranged when they bind to midgut receptors, so that they are likely to lose their toxicity inside target herbivores. What happens to the toxins in non-target herbivores, and whether these herbivores may act as intermediaries through which the toxins may be passed on to the natural enemies, remains to be studied.
Agro-ecosystems consist of organisms that interact in so-called food webs (Dicke and Vet, 1999; Janssen et al., 1998; Mayse and Price, 1978; Poppy, 1997; Price, 1981; Price, 1997; Price et al., 1980; Weires and Chiang, 1973) (Figure 1). Herbivores and carnivores may compete for the same food sources or may partition resources through niche segregation. While lepidopterans are usually economically important pests, insects from other orders such as aphids and thrips or mites may also be important pests. Potential pest arthropods may be naturally suppressed by their natural enemies in agro-ecosystems, or biological control agents may be specifically introduced to control pest arthropods. Biological control is an important component of durable crop protection programmes (DeBach, 1974; Lewis et al., 1998; Van Lenteren, 1993). Hampering the effects of carnivorous arthropods as a side-effect of pest control methods may induce new pests (DeBach and Rosen, 1991; Van Lenteren, 1993) and should therefore be avoided. Natural enemies of herbivores commonly live and/or search for herbivores on plants. Consequently, plant characteristics can affect the behaviour, physiology and ecology of carnivorous arthropods.
Crop protection may be achieved by host plant resistance and biological control. Conventional plant breeding and biological control have been developed independently, but they may interact in various ways (Hare, 1992). When aiming for durable crop protection, one should address these interactions (Dicke, 1999; Hare, 1992; Schuler et al., 1999c). With the development of genetic engineering technology, a whole new array of possibilities has been made available to develop pest-resistant plants, i.e. the introduction of genetic material from any species into crop plants (Duck and Evola, 1997). In the context of insect pest control, this technology has been applied by inserting genes that code for toxins. The δ-endotoxins produced by Bacillus thuringiensis (Bt) are the best known example, and genes encoding these endotoxins have been transferred to major crops (Jenkins, 1999; Merrit, 1998). Although genes from other groups of organisms and plants are currently being explored for their potential use in developing pest-resistant plants, so far commercialized insect-resistant transgenic plants only contain genes coding for (an activated form of) a Bt toxin. Consequently, this review is focused on transgenic Bt plants.
When insect-resistant transgenic plants are evaluated, possible effects of host plant resistance on biological control should be considered in detail. The toxins that insect-resistant transgenic plants produce may have similar effects on carnivores as toxins that are naturally produced in a plant. Therefore, this review includes the effects of plant allelochemicals on biological control agents. This review does not include (possible) interactions with microbial organisms.
Conventional host plant resistance
Secondary plant metabolites (allelochemicals) that are part of natural plant defence have been used in breeding programmes to develop cultivars that produce larger quantities to protect the plants from herbivorous arthropods (Panda and Khush, 1995; Smith et al., 1994). Generalist herbivores usually cannot detoxify or sequester specific secondary plant compounds (Barbosa, 1988; Dicke, 1999). Specialist herbivores may overcome the natural plant defence by sequestration of the secondary plant metabolites and use them for their own defence against natural enemies (Barbosa, 1988; Duffey and Scudder, 1972; Ferguson and Metcalf, 1985; Krieger et al., 1971; Pasteels and Daloze, 1977; Pasteels et al., 1986). When very large amounts of allelochemicals are produced by a new cultivar, specialist herbivores may be negatively affected by these compounds as well. Also, these allelochemicals may be passed on to biological control agents either directly or indirectly, in the same way as Bt toxins in transgenic plants, and (positively or negatively) affect the natural enemies as well.
There is increasing evidence that a change of plant characteristics in a breeding programme affects natural enemies of herbivores. These effects may be positive, negative or neutral (see Table 1 for a review of the literature). Such specific breeding programmes have aimed for (a) physical plant barriers, such as an altered wax layer or glandular trichomes, (b) plant allelochemicals that negatively affect herbivore performance, and (c) specifically in cotton: nectariless cultivars, as nectar may attract herbivores (Table 1).
Table 1. Effects of host plant resistance on biological control agents
An example of positive effects of host plant characteristics on carnivores are plant digestibility reducers that slow down herbivore development. Because the herbivores remain in a stage that is susceptible to their enemies for a longer period of time, this may increase carnivore effectiveness (Dicke, 1999; Haggstrom and Larsson, 1995; Loader and Damman, 1991; but see Benrey and Denno, 1997). Plant structures such as domatia that are used by ants or mites to provide shelter, may also enhance the effectiveness of carnivores (O'Dowd and Willson, 1989).
Negative effects of host plant resistance on carnivores may be found when herbivores sequester plant toxins and subsequently employ them in their defence against their carnivorous enemies (Barbosa, 1988; Björkman and Larsson, 1991; Mendel et al., 1992; Pasteels et al., 1988; Peterson et al., 1987). This may especially occur in specialist herbivores (Dicke, 1999). Specialist carnivores in turn are generally better able to cope with herbivore-sequestered plant toxins than generalist carnivores (Barbosa, 1988; Price, 1991). Natural enemies may also come into contact directly with plant toxins, because they may use various plant tissues and plant products, such as floral and extra-floral nectar, plant sap and pollen, as sources of nutrition (Dicke, 1999).
Transgenic host plant resistance
Bt crops may be considered as having a specific form of host plant resistance, as a toxin is produced in the plant for defence against herbivores. However, there are a few important differences from conventional plant breeding, apart from the fact that in Bt crops a gene from another organism (the bacterium) is incorporated into the genome of the plant. In conventional plant breeding, closely related plant species or cultivars are crossed, and it is largely unknown which or how many genes are involved, or even which plant characteristics are responsible for increased insect resistance. Consequently, it is difficult to specifically determine their environmental impact. In Bt crops, host plant resistance is determined by one major gene (i.e. vertical resistance), and this gene and its product are understood in great detail. Another major difference is the expression level of the insect-resistant plant characteristics in the newly developed lines. In conventional plant breeding, the expression level of plant allelochemicals, for example, is lower than the maximum occurring in nature. In Bt crops, the expression level of the gene encoding the Bt toxin is determined by the (modified) promoter (Koziel et al., 1993; Perlak et al., 1990; Perlak et al., 1991) and the number of gene copies inserted in the plant genome (Kota et al., 1999). In this way, the level of Bt toxins in transgenic plants may become as high as 2–5% of the total protein level in the plants.
The main crops into which Bt genes have been incorporated to control certain pest insects are maize, rice, cotton, potato, tomato, tobacco, soybean and Brassica species (cabbage, oilseed). Bt toxins are expressed constitutively throughout the Bt plant. Bt toxins are only active in insects when ingested orally, as their mode of action is in the midgut. What causes the specificity of the Bt toxins has not yet been completely elucidated. When insects come into contact with and ingest the crystalline inclusions, these inclusions as well as the toxins present in these inclusions (i.e. cry(stal) protoxins; Cry) first need to be proteolytically processed before binding to receptors is possible (Boucias and Pendland, 1998; Pietrantonio and Gill, 1996; Watson et al., 1992). Hence, the specificity of Bt strains may be caused by the presence or absence of specific proteases in an insect midgut. This is an important difference from the mode of action of Bt toxins produced in a Bt plant. Transgenic Bt plants produce toxins instead of protoxins, as only the part of the gene that codes for the active toxin is inserted in the plant genome. As a result, after ingestion, the δ-endotoxins may bind directly to specific receptors on the midgut membrane, without the involvement of insect proteolytic enzymes. In the case of Bt plants, specificity may be based on the presence of specific receptors. So far, the identified receptors to which δ-endotoxins may bind are general receptors, i.e. aminopeptidase-N proteins, cadherin-like proteins or glycolipids (Denolf, 1996; Ferréet al., 1991; Gahan et al., 2001; Van Rie et al., 1990). Binding of the toxins to the receptors is a prerequisite for their insecticidal activity, although binding alone is not sufficient for toxicity (De Maagd et al., 1999, 2001; Denolf, 1996; Schnepf et al., 1998). The rate at which the ligand is inserted into the target membrane decisively influences its relative toxicity (Boucias and Pendland, 1998). However, only when binding is followed by pore formation in the midgut membrane, are midgut cells disrupted and the insect dies (De Maagd et al., 1999; De Maagd et al., 2001; Jenkins et al., 2000; Liang et al., 1995; Schnepf et al., 1998).
How may non-target arthropods come into contact with Bt toxins?
Beneficial, non-target arthropods may come into contact with toxins produced in transgenic plants in several ways; by feeding on plant parts themselves, through feeding on target or non-target herbivorous insects, or via the environment, i.e. the soil when toxins persist and do not lose their toxicity after plant parts or insects have died.
Through feeding on the plants themselves
Carnivorous arthropods such as predators and parasitoids may feed on plant parts or pollen, and thus may directly ingest the toxin from these plant parts (Godfray, 1994; Hagen, 1986; Weiser and Stamp, 1998). Generally, floral nectar and pollen are fed upon by many different groups of carnivorous insects and mites (see Van Rijn et al., 2002 and references therein) that rely on these food sources for energy and in some cases also for egg production (Hagen, 1986). Dipteran and hymenopteran parasitoids are entomophagous in their larval stage, but the adults usually feed on (extra)floral nectar or honeydew (Godfray, 1994; Hagen, 1986). Bt toxins are not present in (extra)floral nectar and honeydew, and thus adult parasitoids are not likely to come into contact with Bt toxins. A number of predators (e.g. coccinellid beetles, tachinid flies, pentatomid bugs) may feed on pollen or green plant parts when prey is not available or present in low densities (e.g. Andow, 1990; Coll and Bottrell, 1991; Coll and Bottrell, 1992). Thus, these predators will consume Bt toxins as adults.
In conclusion, natural enemies may feed directly on plant parts in which Bt toxins are present; however, nectar does not contain Bt toxins.
Through feeding on insects that have ingested the toxin
Carnivorous arthropods may also ingest the Bt toxins in a more indirect way when they feed on herbivorous insects that have ingested the toxins from the plant. In target insect herbivores, the toxins are bound to receptors in the midgut epithelium, after which they are structurally rearranged (De Maagd et al., 1999; De Maagd et al., 2001; Gazit et al., 1998; Masson et al., 1999) and thus most likely lose their toxicity to natural enemies. In non-target herbivores, however, the toxins do not bind to midgut receptor cells or do not cause pore formation. They may or may not be digested by proteolytic enzymes in the digestive tract, so that the toxins may remain active and subsequently affect entomophagous natural enemies.
In conclusion, natural enemies may not come into contact with Bt toxins via target herbivores, but via non-target herbivores.
Bt toxins may not only get into the soil via decaying plant material or insects, but also via root exudates, as shown by Saxena et al. (1999). They tested the toxicity of root exudates of Bt corn producing Cry1A(b), using M. sexta, and found a 90–95% mortality after 5 days. Soil samples remained toxic even after 25 days. The concentration of Cry1A(b) protein in soil samples from the rhizosphere of seedlings was about 95 µg g−1 soil (Saxena et al., 1999), which is much more than Sims and Ream (1997) calculated for Bt toxins that enter the soil via decaying Bt plants. Sims and Ream (1997) calculated a maximum expected field load of 1.6 µg g−1 soil when Bt cotton with 60 000 plants per acre is uniformly incorporated into the top 7.6 cm of soil, because they found a Cry2A concentration of about 34 µg g−1 fresh weight in Bt cotton tissue, which amounted to 8 mg of toxin per plant. When root exudates are taken into account as well, the maximum field load may be 60 times higher.
As Bt toxins may persist in the soil for at least as long as 234 days (the longest period analysed; Tapp and Stotzky, 1998), soil organisms such as Collembola and Carabidae are exposed to the toxins for prolonged periods of time. Due to accumulation of toxins over time during degradation of plant biomass, the doses of Bt toxin to which these soil organisms are exposed may increase with time as well. Soil bacteria, fungi, protozoa, nematodes and earthworms do not seem to be affected by Bt toxins, as Saxena and Stotzky (2001) did not find mortality or weight differences between these organisms in Bt soil and non-Bt soil. However, earthworms do ingest Bt toxin bound to the soil: Saxena and Stotzky (2001) detected Cry1A(b) in the guts and casts of earthworms that were kept in jars with soil containing biomass of Bt corn for 45 days. Hence, similar to non-target herbivores, these earthworms may serve as intermediates through which Bt toxins may be passed on to organisms feeding on these earthworms.
In conclusion, Bt toxins are only extractable from soils for a very short period, after which they bind to the soil. These bound toxins retain their toxicity, as has been shown by all relevant studies published since 1996, even after frost. Whether soil organisms are affected by these bound toxins remains to be studied. The bound Bt toxins may be passed on to other organisms feeding on earthworms.
Effects of Bt crops on different food web components
The Bt genes in crops are generally used to control specific target pest insects belonging to one of the three insect orders Lepidoptera, Coleoptera or Diptera. As mentioned before, many other (secondary) pests may occur in these crops apart from the target pests for which Bt crops are developed. Non-target pests will feed on the same plants and may be influenced by the Bt toxins. If they are not influenced, biological control agents may be important to reduce the possibility of a new pest developing. Also, many species belonging to the Coleoptera and Diptera are not phytophagous but carnivorous or saprophagous, and a number of them are important biological control agents. Specifically, about 35% of coleopteran species are phytophagous while 65% are carnivorous or saprophagous, and only about 30% of the dipteran species are phytophagous while 70% are carnivorous or saprophagous (Strong et al., 1984). Of the Lepidoptera, more than 99% are phytophagous. However, this order of insects includes not only pest species but also species that are important in terms of conservation, such as the monarch butterfly (Danaus plexippus) in the United States and swallowtails (Papilio spp.) in the United States and Europe. In determining possible effects of Bt plants on non-target insects, it is important to recognize the food webs and their components that may be present. Non-target arthropods comprise non-target Lepidoptera, other non-target herbivorous pests, pollinators, and parasitoids and predators.
The study that drew a lot of attention to (possible) effects of Bt plants on butterflies that are important for conservation was that by Losey et al. (1999). They reasoned that, as Bt corn also expresses the toxin in the pollen, and pollen can be dispersed to and deposited on other plants near corn fields, non-target organisms that feed on these other plants may ingest Bt toxins and be affected as well. When milkweed leaves were dusted with pollen from N4640 Bt corn (Bt11 event), D. plexippus larvae consumed significantly less from these leaves compared to leaves dusted with untransformed pollen, and in 4 days almost half of the tested larvae died, which was significantly more than on the leaves with untransformed pollen where none of the tested larvae died.
As the amounts of toxin differ in the three Bt corn events, the effect of pollen from these corn plants on monarchs differed as well. Pollen from Bt11 and Mon810 did not have larvicidal effects on monarchs, while pollen from event 176 did (Hellmich et al., 2001; Stanley-Horn et al., 2001). Monarch larvae were significantly smaller when feeding on leaf discs treated with pollen densities as low as 5–10 grains cm−2 from event 176 (Hellmich et al., 2001). In event 176 corn fields, fewer than 7% of monarch larvae survived, although this rate of decline slowed after systematically removing predatory arthropods (Zangerl et al., 2001). For swallowtails, pollen from this event resulted in significant mortality when present at densities of 100 grains cm−2 and higher (Zangerl et al., 2001).
Monarch butterflies naturally occur in agricultural habitats throughout their breeding season, where their densities were as high as or higher than in non-agricultural habitats (Oberhauser et al., 2001). Thus, agricultural practices in general could have a large impact on monarch populations. In the case of Bt corn event 176, the amount of grains deposited on milkweed leaves in and around corn fields is sufficient to detrimentally affect monarch larvae. Seed companies producing event 176 chose not to seek US EPA re-registration in 2001 (http://www.epa.gov/pesticides/biopesticides/factsheets/fs006458t.htm), which means that this type of corn will be commercially available only until 2003.
In conclusion, monarch larvae are affected by the Bt toxin Cry1A(b). Swallowtails seem to be affected by Bt toxins in pollen only when offered at higher doses. One other study determined the effect of Bt on non-target Lepidoptera, i.e. the tree-feeding Papilio glaucus, P. canadensis and Callosamia promethea (Johnson et al., 1995). This study relates to sprays of Bt var. kurstaki. In forests in the US and Canada, Btk sprays are used against the forest defoliator, the gypsy moth Lymantria dispar, and therefore non-target lepidopterans in these areas may be affected as well. When first- and early-second-instar larvae of the three non-target Lepidoptera were placed on host trees that were or were not sprayed with Bt, significantly fewer caterpillars were alive on the Bt-treated trees after 5 days. With P. glaucus, an extra experiment was conducted, in which first- and second-instar larvae were placed on Bt-treated trees 0, 10, 20 and 30 days after spraying. This species showed significantly higher mortality after 8 days for all treatments, i.e. after 8, 18, 28 and 38 days, respectively, compared to controls (Johnson et al., 1995), which indicates that the Bt sprays used in this study were more persistent than usually assumed.
It is important to note that all above-mentioned studies refer to non-target lepidopterans that do not feed on Bt plants themselves. Green plant parts contain much higher levels of Bt toxins (Hellmich et al., 2001). Non-target lepidopterans that do feed on Bt plant parts but are not affected by the toxin ingest these toxins and may subsequently act as intermediates through which these toxins are passed on to the third trophic level, their predators and parasitoids, in a similar way as naturally occurring plant allelochemicals.
Apart from the Lepidoptera, Diptera and Coleoptera, there are pest arthropods from other orders as well, the most important ones being aphids and other sucking insects (Homoptera, Hemiptera), thrips (Thysanoptera), and spider mites (Tetranychidae). Most herbivorous insects and mites will ingest the toxins when feeding on the transgenic plants. The major exceptions are aphids, because they do not feed on plant cells but on phloem sap, in which Bt toxins are not expressed (Raps et al., 2001). After ingestion, herbivorous non-target arthropods may be affected by the Bt toxin synthesized in the plant, depending on the presence of receptor cells in the midgut epithelium and whether binding to receptors subsequently results in pore formation. Although specific receptors for the toxic proteins on the midgut epithelium appear to be necessary, so far it is unclear whether such receptors are present in non-target organisms (De Maagd et al., 2001; Tapp et al., 1994).
One major unanswered question so far is what happens to Bt toxins after ingestion by non-target herbivores. As these toxins may not bind to receptors so that their structure remains unchanged, it is conceivable that the toxins will remain intact and therefore toxic. If so, natural enemies parasitizing or feeding on these herbivores may come into contact with and be affected by the toxins. An example illustrating this relates to spider mite–predatory mite interactions. Spider mites do not seem to be affected by Bt toxins (Chapman and Hoy, 1991), although they do ingest the toxins when feeding from transgenic plants (Dutton et al., 2002). However, Chapman and Hoy (1991) did find a toxic effect on the predatory mite Metaseiulus occidentalis when leaf disks with Tetranychus urticae were sprayed with suspensions of Bt var. tenebrionis (i.e. protoxin). Predatory mites are important biological control agents of spider mites (Helle and Sabelis, 1985). Thus, if toxins from transgenic plants are passed on via the herbivore to the predator, the predatory mites may be negatively affected when feeding on Bt-fed spider mites which may result in outbreaks of spider mites.
In conclusion, so far it is largely unknown what happens to Bt toxins in non-target herbivores and/or whether these herbivores may act as intermediaries through which the toxins may be passed on to predators and parasitoids.
Pollinators are essential for most fruit and vegetable growers. The honey bee Apis mellifera is the pollinator most commonly used, although solitary bees such as Osmia spp. are also used. In the current transgenic crops, Bt toxins are produced in the pollen as well; therefore it is important to determine whether honey bees are affected by these toxins. In collecting pollen, bees visit many flowers per flight, so that their body is often covered with pollen grains. Because of this behaviour, gene flow between transgenic crops and non-transgenic plants may occur (e.g. Ramsay et al., 1999). However, it is beyond the scope of this review to cover the possibilities and dangers of gene flow between transgenic and non-transgenic plants. The bees that may be affected by the Bt toxins are the larvae and younger bees (up to 10 days of age) for which the pollen are collected. The bees collecting the pollen feed from floral nectaries (nectar is also collected for the production of honey). As transgene products have not been detected in nectar, bees are not exposed to the toxin when feeding on this plant product (Malone and Pham-Delègue, 2001). Toxicity of Bt toxins should thus be determined for larvae and young bees of up to 10 days old.
Several studies have been conducted to determine the toxicity of Bt toxins to pollinators. Vandenberg (1990) tested a solution of Bt (Bt var. tenebrionis) in sucrose that was fed to newly emerged (< 24 h) adult A. mellifera in cages, and found no significant differences in mortality after 5 days when compared to adults fed plain sucrose. Sims (1995, 1997) conducted toxicity tests on larval and adult A. mellifera that were fed sugar solutions with the activated Cry1A(c) toxin (Sims, 1995) or the activated Cry2A toxin (Sims, 1997). In neither case were significant differences found in mortality between bees fed the Bt sugar solutions or the sugar solution without Bt toxins. Arpaia (1997) mixed Cry3B toxin in supplemental syrup that was offered to A. mellifera colonies. Each week, larval survival and pupal dry weight were recorded, and no toxic effects were found for this toxin either. Neither did Malone and Pham-Delègue (2001) find differences in longevity and flight activity when A. mellifera bees were fed with purified Cry1B(a) toxins.
In conclusion, there is no evidence that A. mellifera larvae or adults are affected by Bt toxins. Other pollinators have not been tested so far.
Parasitoids and predators
When the third trophic level is considered, primary parasitoids and predators may feed on one or several herbivorous insect species, depending on their degree of specialization. Parasitoids are usually specialists and thus will mostly parasitize only a few species belonging to one family (e.g. Hawkins, 1994). When the population of target insects diminishes due to the Bt crop, their parasitoids may move to other sites where more hosts will be encountered (Croft and Messing, 1990). This effect will be similar for all pest control methods that lead to a reduction in a pest population.
When investigating whether effects of herbivore population reductions on natural enemies occur, it is important to first verify whether treated plots contain a significantly smaller population of the target herbivore. This is not always the case. For example, Orr and Landis (1997) and Pilcher et al. (1997) did not find a difference in population density of predators between transgenic corn (containing Cry1A(b) toxin) and isogenic corn. However, no significant differences in egg densities and egg masses of the target herbivore Ostrinia nubilalis were found either (Orr and Landis, 1997; Pilcher et al., 1997), but the percentage of individual eggs within egg masses that hatched was significantly lower in transgenic plots (Orr and Landis, 1997).
Only a few field studies have been conducted that aimed to determine not the overall population density differences of parasitoids and predators in general, but of specific predator or parasitoid species between Bt and non-Bt crops. For example, Riddick et al. (1998) determined differences in abundance of Lebia grandis and Coleomegilla maculata adults between isogenic, 100% Cry3A-transgenic potato fields, and mixed fields. The specialist predator L. grandis was significantly less abundant in seed-mixed and pure 100% transgenic potato fields than in non-transgenic potato fields, and probably dispersed from transgenic fields due to low densities of leptinotarsa decemlineata (Riddick et al., 1998). The generalist predator C. maculata was as abundant in seed-mixed and pure 100% transgenic potato fields as in the isogenic potato fields. When L. decemlineata is present in low densities, most likely C. maculata turns to alternative prey or to plant pollen (Riddick et al., 1998). However, when feeding on pollen, this predator may still encounter Bt toxin.
Effects on carnivorous arthropods may especially be expected when Bt toxins are used in crops against dipteran and coleopteran pests, because many Diptera and Coleoptera are important generalist predators. For example, Bt toxins in potato aimed against the coleopteran Leptinotarsa decemlineata may affect an important group of predators that prey upon potential non-target pest species, such as aphids. Spiders are also important generalist predators of a large number of insect species, although their role is not always recognized (Bogya, 1999; Mayse and Price, 1978; Riechert and Lockley, 1984; Schmaedick and Shelton, 2000).
Published effects of Bt toxins on predators and parasitoids are summarized in Table 2, and discussed below.
Table 2. Effects of Bt on biological control agents
a Host fed Bt toxin present in plant; b host fed Btk in artificial diet; c field plots treated with Btk sprays; d Leaves dipped in Btk solution; e Bt var. san diego; f plus pollen offered; g three Cry toxins tested (separately): Cry1A(b) toxin, Cry1A(b) protoxin, Cry2A protoxin: effects for the three toxins were similar; h Fitness: survival, aphid consumption, development, reproduction; i L, Laboratory experiments; F, field experiments; j +, Bt had a positive (indirect) effect on effectiveness of biological control agent(s); –, Bt had a negative (indirect) effect on effectiveness of biological control agent(s); 0, Bt had no effect on biological control agent(s). *No difference in hosts between treated and untreated plots. 1Johnson & Gould 1992; 2Johnson 1997; 3Atwood et al., 1997; 4Young et al., 1997; 5Chenot & Raffa 1998; 6Schuler et al., 1999a; 7Schuler et al., 2001; 8Schuler et al., 1999b; 9Orr & Landis 1997; 10Blumberg et al., 1997; 11Schuster 1994; 12Johnson et al., 1997b; 13Riggin-Bucci & Gould 1997; 14Sims 1995; 15Sims 1997; 16Campbell et al., 1991; 17Lopez & Ferro 1995; 18Riddick & Mills 1995; 19Lozzia & Riggamonti 1998; 20Riddick et al., 1998; 21Giroux et al., 1994; 22Zwahlen et al., 2000; 23Pilcher et al., 1997; 24Riddick & Barbosa 1998; 25Keller & Langenbruch 1993; 26Dogan et al., 1996; 27Cloutier & Jean 1998; 28Hilbeck et al., 1998a; 29Hilbeck et al., 1998b; 30Hilbeck et al., 1999; 31Dutton et al., 2002; 32Meier & Hilbeck 2001; 33Painter et al., 1996; 34Merrit et al., 1991; 35Chapman & Hoy 1991.
Effects of different Bt sprays on target and non-target arthropods have been reviewed by Krieg and Langenbruch (1981), Flexner et al. (1986), and Glare and O'Callaghan (2000). In determining effects of δ-endotoxins on carnivorous arthropods, the results of many of these studies are of limited use, as β-exotoxins were present in many of the Bt products that were used until the late 1980s (Glare and O'Callaghan, 2000). Because β-exotoxins were found to be toxic to many species, including mammals, Bt sprays are now required to be free from β-exotoxins in most European countries, the US and Canada (Glare and O'Callaghan, 2000). The addition of Ca2+ ions has been used to precipitate β-exotoxins. Unfortunately, most studies did not specify the exact content of the Bt sprays used in the experiments. Therefore, we chose to review only the studies conducted in Europe, the US and Canada from 1990 onwards (see Table 2).
The toxins that are produced in Bt plants may be passed on to biological control agents via plant material (pollen or in some cases also plant leaves) or via target or non-target herbivores, depending on which life stage comes into contact with the toxin via the plant or via the herbivore in the field. So far, few studies have been conducted to determine the effects of Bt toxins on natural enemies of both target and non-target herbivores. Most studies have concentrated on natural enemies of target herbivores (see Table 2).
Bt toxins may affect natural enemies either directly, or indirectly by feeding from sub-optimal food. Hilbeck et al. (1998a) tried to distinguish between these effects by feeding two different kinds of prey to Chrysoperla carnea, i.e. one that is highly susceptible to the toxin (Ostrinia nubilalis) and one that is less sensitive (Spodoptera littoralis). Significant differences in the overall developmental time of C. carnea were found between predators reared on Bt-fed and non-Bt-fed O. nubilalis larvae, while the overall developmental time was similar between predators reared on Bt-fed and non-Bt-fed S. littoralis. This suggests that the nutritional deficiency of Bt-affected prey has a larger effect on C. carnea than the Bt toxins themselves (Hilbeck et al., 1998a). However, C. carnea reared on Bt-fed prey had a higher mortality than predators reared on non-Bt-fed prey. Also, Schuler et al. (1999a) found a negative effect of Bt hosts on the development of Cotesia plutellae, but this was due to the fact that all hosts died before C. plutellae could eclose from these hosts. Pilcher et al. (1997) did not find fitness effects of Cry1A(b) toxins from transgenic corn on the predator Coleomegilla maculata when fed either pea aphids (Acyrthosiphon pisum), non-Bt pollen, or Bt pollen. As Cry1A(b) should be specifically toxic to Lepidoptera, it is not surprising that this coleopteran predator was unaffected by this toxin. However, as long as the specificity of Bt toxins is not completely clarified, effects of each Bt toxin on insects belonging to other orders cannot be excluded.
Direct toxicity tests with specific Bt toxins were conducted by Schuster (1994) and Sims (1995, 1997). Schuster (1994) examined a number of insecticides, including Dipel (Btk), by dipping tomato leaflets in an aqueous preparation of the insecticide, air-drying them, and then letting the parasitoids survive on them. With Dipel, adult Diglyphus intermedius (Hymenoptera, Eulophidae) showed significant mortality compared to the control, but the larval stage, and the larvae and adults of Neochrysocharis punctiventris (of the same family) did not show higher mortality. However, direct toxicity tests are of limited value, because it remains unclear whether natural enemies would be exposed to the toxin in this way in the field, and, even when no direct toxicity is found, the overall fitness and parasitism or predation rate may still be affected.
In conclusion, it seems that the fitness of parasitoids and predators is not affected by Bt toxins expressed in Bt plants, but there is an indirect effect by feeding from sub-optimal food or because of host death.
So far, it seems that predators and parasitoids do not distinguish between Bt-intoxicated prey and non-intoxicated prey. Parasitoids and predators of the Colorado potato beetle (L. decemlineata) did not discriminate between Bt-intoxicated and non-intoxicated individuals (Lopez and Ferro, 1995; Riddick and Barbosa, 1998). Neither did the parasitic fly Myiopharus doryphorae discriminate between Bt-intoxicated hosts and non-intoxicated hosts; larvae were laid in control L. decemlineata as well as lethally or sub-lethally infected L. decemlineata, although it took about twice as long to larviposit in the control larvae (Lopez and Ferro, 1995). However, third-instar Chrysoperla carnea did feed significantly less from Bt-fed S. littoralis than from non-Bt-fed S. littoralis, at least when probings lasting up to a minute were not taken into account (with the idea that no substantial amount of food can be ingested in such brief time; Meier and Hilbeck, 2001).
In conclusion, positive effects of Bt toxins on the performance of parasitoids and predators were found when these toxins caused prolonged development in target hosts or prey. Natural enemies are likely to be less attracted to Bt plants infested with Bt-susceptible prey, because reduced feeding by these herbivores causes a reduced production of volatile cues.
Effects of plant allelochemicals on biological control agents
As found with Bt toxins, herbivores, especially generalists, may be influenced by plant allelochemicals in terms of their behaviour or development such that predators or parasitoids can attack them more easily or for prolonged periods (see Table 1). A good example are hydroxamic acids that are present in wheat and other Poaceae. On wheat cultivars with increased levels of hydroxamic acids, aphids showed a longer developmental time and a smaller body size, which negatively influenced their defensive reactions to parasitoids, i.e. kicking behaviour to prevent parasitoid stabbing (Fuentes-Contreras and Niemeyer, 1998).
Nicotine in tobacco may be sequestered by herbivores in their haemolymph, which alters their suitability for natural enemies (Barbosa, 1988; Rowell-Rahier and Pasteels, 1992; Thomas and Waage, 1996). Negative effects of increased levels of nicotine have been found for the parasitoids Cotesia congregata and Hyposoter annulipes, both parasitizing Manduca sexta, a specialist herbivore of Solanaceae (see Table 1). The specialist parasitoid C. congregata was less affected by nicotine-fed hosts than H. annulipes, a generalist parasitoid of a variety of noctuids (Barbosa et al., 1986; El-Heneidy et al., 1988). However, C. congregata was detrimentally affected by the nicotine-fed hosts just prior to and after larval parasitoid emergence, although the amounts of nicotine used caused no significant mortality in the hosts (Barbosa et al., 1986). Also, in the field, a significantly smaller proportion of male and female C. congregata adults emerged from M. sexta hosts feeding on a high-nicotine tobacco variety compared to a low-nicotine tobacco variety (Thorpe and Barbosa, 1986). These studies indicate that nicotine is passed on via the herbivore to parasitoids, which are subsequently negatively affected by this plant allelochemical.
In tomato, several plant metabolites have been found to reduce herbivory: α-tomatine (Campbell and Duffey, 1979; Traugott and Stamp, 1996), the methyl ketones 2-tridecanone and 2-undecanone (Barbour et al., 1993; Farrar and Kennedy, 1993; Farrar et al., 1992), and chlorogenic acid and rutin (Stamp et al., 1997). α-Tomatine, a steroidal alkaloid, reduces the growth of various caterpillar species. As for other steroidal alkaloids, tomatine binds to sterols, making them unavailable for structural purposes, steroid hormone synthesis, and reproductive tissues (Bernays, 1989; Campbell and Duffey, 1981). Insect predators rely on their prey as a source of cholesterol or utilizable sterols (Svoboda et al., 1978). Campbell and Duffey, 1979) determined the effect of α-tomatine on the parasitoid Hyposoter exiguae when fed Helicoverpa zea. When H. exiguae was fed on α-tomatine-fed hosts, its larval period was prolonged and percentages of pupal eclosion, adult weight and adult longevity were significantly reduced. Higher concentrations of tomatine resulted in smaller adult parasitoids. α-Tomatine was found in the larval tissues of H. exiguae, indicating that it was absorbed from the host.
When herbivores sequester plant toxins, this can affect their susceptibility to pathogens. For instance, tobacco budworm larvae fed diets with low nicotine concentration are highly susceptible to Bt, whereas they are unaffected by the bacterium when feeding on a diet with a high nicotine concentration (Krischik et al., 1988). Similar effects were recorded for the influence of tannins on the susceptibility of gypsy moth or corn earworm larvae to viruses or Bt toxin, respectively (Hunter and Schultz, 1993; Navon et al., 1993).
In conclusion, as with Bt toxins, plant allelochemicals can positively affect the performance of natural enemies due to prolonged developmental time or reduced defensive reactions of the prey. Nicotine and α-tomatine can be passed on via the herbivore to parasitoids, which subsequently may be negatively affected by these plant allelochemicals.
Influence of natural enemies on resistance development in target insects
Natural enemies of herbivorous insects can increase or decrease the rate of herbivore adaptation to host plant resistance (Gould et al., 1991; Gould, 1994). For example, generalist predators may feed on the few early-pupating (i.e. faster-developing and thus resistant) individuals, but become satiated as the masses of (more susceptible) individuals pupate. Such a response would slow the rate of adaptation in the herbivore. By contrast, predators or parasitoids may have a search image for pupae when they are abundant and not when they are rare. In this case, susceptible individuals are more subject to predation than the faster-developing resistant individuals. Although such synergism between plant resistance and natural enemies may be favourable in terms of the initial efficacy of pest control, it is not favourable for the durability of control. Due to the removal of susceptible individuals, adaptation of the herbivore to insect-resistant (Bt) plants will be accelerated (Johnson and Gould, 1992). Hence, as long as adapted (resistant) herbivorous individuals are less prone to natural enemy mortality than unadapted susceptible individuals, natural enemies will increase the rate of adaptation of the pest population (Gould et al., 1991; Gould, 1994).
Biological control is an important component of durable crop protection strategies, and therefore negative effects of (transgenic) insect resistance on carnivorous arthropods should be avoided. At present, it is difficult to compare different studies of the effects of Bt toxins on non-target, beneficial insects, because in Bt sprays and different transgenic lines the toxin type and toxin dose are reported differently. To be able to compare possible non-target effects of (activated) Bt toxins expressed in different transgenic plant lines, a standardized measure is needed of the amount of toxin in different plant parts. There is a need for dose–response curves. Because of the development of a new generation of Bt crops with much higher expression levels (Kota et al., 1999; McBride et al., 1995), the effects on non-target organisms reported so far are likely to be an under-estimate.
Non-target, beneficial arthropods may come into contact with toxins produced in insect-resistant (transgenic) plants by feeding directly on plant parts in which the toxins are present, with the exception of nectar which does not contain Bt toxins. Bt toxins are persistent in the soil, where they bind especially to clay particles. These bound toxins retain their toxicity, even after frost, as shown by all relevant studies published since 1996. Whether soil insects are affected by these bound toxins remains to be studied. The bound Bt toxins may be passed on to other organisms feeding on earthworms.
When feeding from pollen of Bt corn (event 176), monarch larvae are affected by the Bt toxin Cry1A(b). Swallowtails seem to be affected by Bt toxins in pollen only when offered in high concentrations. There is no evidence that honey bee A. mellifera larvae or adults are affected by Bt toxins. Other pollinators have not been tested so far.
Natural enemies may not come into contact with Bt toxins via target herbivores, but via non-target herbivores, because the toxins do not bind to receptors on the midgut membrane in the non-target herbivores. So far it seems that the fitness of parasitoids and predators is not directly affected by Bt toxins expressed in Bt plants, only indirectly by feeding from sub-optimal food or because of host death. The parasitism or predation rate were positively affected by Bt sprays or toxins expressed in Bt plants when this caused prolonged developmental time or reduced defensive reactions in hosts or prey. However, such an initial positive effect on the efficacy of natural enemies has negative effects on the durability of control. Such removal means that susceptible individuals are selectively removed from the population, which accelerates resistance development in target insects. Negative effects on the parasitism/predation rate have been found in some predators/parasitoids when they came directly into contact with Bt sprays. Other predators were not affected in this way. Differences in overall natural enemy densities between Bt and non-Bt fields are difficult to interpret. Most likely, generalist natural enemies remain in Bt fields as long as alternative food is present, while specialist natural enemies emigrate to sites with higher prey densities.
To make a true contribution to durable crop protection through transgenic plants, it is essential to take food web consequences into account. Therefore, studies on the multi-trophic aspects of transgenic insect-resistant crops are badly needed.
†Present address: Department of Entomology, North Carolina State University, Raleigh, NC 27695, USA
This work was supported by a grant from the Dutch Ministry of Housing, Spatial Planning and the Environment. We thank Willem Brandenburg, Marco Gielkens, Fred Gould, Michel Haring, Hans den Nijs, Joerg Romeis, Jan Sevenster and three anonymous reviewers for constructive comments on a previous version of the manuscript.