*For offprint requests: juliette@pondside.uchicago.edu; tel: (773) 702-3856; fax: (773) 702-9740

SUMMARY

In order to use computer models as useful risk assessment tools, we need field experiments to test their predictions. Last year we introduced a predictive resistance management model that compared the performance of three cropping strategies that varied the way in which two pesticidal transgenes were distributed in a crop plant. This past year we have worked to develop a field experiment that will test the predictive value of this model. We bred Brassica napus, the crop that produces canola oil, to possess one or both of two insecticidal transgenes homozygously (Bacillus thuringiensis -endotoxin and potato proteinase inhibitor II). We also collected and laboratory-reared two populations of diamondback moth (Plutella xylostella), a widespread pest of canola. One of the two moth populations comes from central Florida, a region where the insects have developed field resistance to Bt -endotoxin. We assessed the initial lethality of our transgenic Brassica plants on these two moth populations. This is a crucial starting point for field experiments designed to detect changes in that lethality over time (evolution). These field experiments are now underway. We will discuss the design of these experiments and present some preliminary results.

INTRODUCTION

The specter of dangerous pesticides has long been a source of public concern. One manifestation of their danger is that pesticides cause the evolution of pesticide resistant pests. It is becoming increasingly difficult for humans to compete successfully with these resistant pests for our food. In particular, pesticide-resistant pests pose two important risks. First, because we cannot control them, they have the potential to run rampant on our crops. Second, these pests may cause unspecified damage in natural plant communities due to their altered ecology. Can biotechnology circumvent these risks, and if not, can biotechnology be managed so as to minimize these risks?

Genetically engineered pesticidal plants are at least as risky as pesticides with respect to the evolution of resistant pests. The pesticides in transgenic plants, unlike conventional pesticides, do not wear off or degrade, thus selection for the evolution of resistance will be more unrelenting. As a result, resistant pests are likely to appear sooner. Furthermore, many of the genes used to make transgenic plants pesticidal come from other plants. If plants in natural communities use these genes to protect themselves from infestation, the plants will be newly susceptible to resistant pests. Thus, biotechnology will not circumvent the risk of resistant pests posed by pesticides. The best we can hope to do is manage genetically engineered plants to minimize these risks.

One very useful tool to assess the scope of the risks of pest evolution is computer modeling. We have developed a model with which we compare three management strategies for the use of two pesticidal transgenes, and we explore the rate and consequence of pest evolution within these strategies.

MODEL STRUCTURE AND PREDICTIONS

There are three comparable ways to envision the distribution of two pesticidal transgenes within a particular crop. First, each individual plant in the crop can express both transgenes. This multiple toxins approach represents a homogeneous distribution of the transgenes. Alternatively, the two pesticidal genes can be distributed heterogeneously in the crop. One way to do this is by growing a spatial mosaic of individuals with one or the other transgene. The other way to do this is to alternate plantings of these two single-transgene varieties over time, in a temporal rotation.

Our model simulates the evolution of a pest population on a transgenically pesticidal crops by cycling through several generations of the pest. In each pest generation, there is a chance for a mutation to arise that confers resistance to one or the other of the plant toxins. Individual pests with that mutation always survive, whereas susceptible pests have a reduced chance of survival. The reduction is based on how strongly pesticidal the expressed transgenes are.

There are several simplifying assumptions incorporated into this model (see Winterer, 1995 for a discussion of the importance of these assumptions, and what happens when they are relaxed).

There is no pleiotropy, that is, no cost to the pest associated with carrying resistant alleles.

Generations of the pest are discrete.

The pest population size is very large.

There is a gene-for-gene relationship between pest susceptibility and the pesticidal effect of a transgene.

Forward and backward mutation rates are the same.

The pest colonizes host plants without regard to host genotypes.

Resistance genes in the pest occur on different chromosomes, so that genes assort independently. Furthermore, mating is random with respect to pest genotype.

The pest population is closed and well mixed.

Over time the average fitness of the pest population increases because the frequency of individuals with resistance alleles increases (Figure 1). There comes a time when the level of resistance in the pest population enables the pest to damage the crop as if it were unprotected by pesticidal transgenes. At this threshold, the durability of the crop fails. Before crop failure, the pest population is able to do some, though perhaps not much, damage to the crop. The amount of damage is proportional to the fitness of the pest population. The precise function relating pest fitness to crop damage will vary from system to system. The effectiveness of a cropping strategy is a function of its durability and its ability to suppress pest fitness while it is durable.

In all three strategies, pests experience both toxins after two crop generations. The pattern of the experience, though, is very different (Figure 2). Pests exposed to multiple toxins crops evolve resistance quickly to both pesticidal transgenes. Pests exposed to spatial mosaics evolve more slowly. In rotations, pests evolve in response to each pesticidal gene independently, a process that is also slow. The ability of the multiple toxins approach to suppress the fitness of the pest population before they evolve is better than the other two strategies. During this time the crop is well protected from damage.

Which strategy is most effective depends on the pesticidal strength of transgene expression (Figure 3). Our model suggests that weak transgenes are best used in a multiple toxins strategy. Strongly pesticidal transgenes are more effectively used in a heterogeneous cropping strategy, either rotation or spatial mosaic. This general prediction is robust, that is, few parameters alter it. We have spent the past year setting up an experiment that will test the usefulness of this model as a predictive tool.

EXPERIMENT DEVELOPMENT

Plant transformation and breeding. Brassica napus var. Westar (oilseed rape) is an annual crop plant cultivated widely in many temperate climates including the United States and Canada. It represents an interesting model system for questions of risk assessment for several reasons. It has several weedy wild relatives with which it is genetically compatible, it has both generalist and specialist herbivores, and it is transformable.

Brassica napus was genetically engineered to express the insect resistance proteins potato proteinase inhibitor II and Bacillus thuringiensis -endotoxin (cry IA(c) gene). Transformation of plants was achieved through use of a disarmed Agrobacterium tumefaciens binary vector system. The plasmid used to transform lines to express the Bt -endotoxin protein was obtained from Calgene Inc., Davis, California. This construct contained the coding sequence of a truncated crystal protein cryIA(c) from Bacillus thuringiensis subsp. kurstaki flanked by the enhanced mannopine synthase (mac) promoter from A. tumefaciens and CaMV, and by the mas 3' terminator from A. tumefaciens. The genetic construct used to transform lines to express the potato proteinase inhibitor II protein (plasmid pRJ12) was obtained from Dr. C.A. Ryan of Washington State University, Pullman, Washington. It contained the proteinase inhibitor II gene from potato flanked by the 35S 5' promoter from cauliflower mosaic virus (CaMV), and by the 3' terminator region of gene 7 from A. tumefaciens T-DNA. Both constructs also included a selectable marker, the kanR gene, encoding APH(3')II from Tn5. This gene confers resistance to the antibiotic kanamycin.

Brassica napus was transformed using a stem section explant procedure (Fry et al., 1987). Seeds from regenerated plants were germinated on selective media and scored for the resistant phenotype.

The transformed seedlings were transplanted and grown in a greenhouse in the spring of 1995. At maturity, these plants were backcrossed by hand pollination to untransformed Brassica napus var. Westar. Tissue from the offspring of this backcross was analyzed by southern blot to isolate lineages with single copies of the transgenes. The Bt and PI-II-expressing lines were crossed to generate four phenotypic lineages, distinguished using ELISAs. Thus, the offspring expressed Bt alone, PI-II alone, both Bt and PI-II, or neither (null segregants). These crosses were replicated so that each phenotype was represented by at least two independently transformed lineages. Finally, each of these lineages was self pollinated to generate homozygous genotypes.

The unmodified parental plants and null segregants show no expression of either Bt or PI-II proteins. In the transformed plants, expression of these proteins ranges from 20 to 100 ng per mg total protein. Transformed plants have the same morphological and structural characteristics as untransformed plants.

Moth populations and bioassay. The diamondback moth, Plutella xylostella (L.) is an economic pest of several Brassica crops including B. napus. Beginning as leaf miners and continuing on the outside of leaves, the larvae are voracious folivores. The generation time extends three to four weeks, and a crop can experience several generations during its growth. The insects do not diapause but instead recolonize colder regions annually (Talekar and Shelton, 1993). Because diamondback moths are specialists on cruciferous plants, and because they have a very short generation time, they are a good model system for studying evolutionary responses to changes in their host plant. The are particularly interesting as they are the only insect for which field resistance to Bacillus thuringiensis has been documented (Tabashnik et al., 1990).

We collected moths from two populations, one in central Illinois (University of Illinois at Urbana-Champagne), and one from central Florida. The Florida population is reputed to evolve resistance to Bt every year in this population. In a controlled environment chamber (22C, 16L/8D) we reared larvae in petri dishes with leaves of Brassica napus var. Westar. After eclosion we transferred adults to flight cages for oviposition on live plant material. We attempted to maintain population sizes greater than 200 for each group.

In order to determine how larvae in the two populations performed on the various transformed lineages of plants we developed a bioassay. In this assay, late third and early fourth instar larvae were placed in groups of ten to twenty (low density) in petri dishes containing a leaf from a transformed plant or null segregant. These leaves were sampled just prior to the assay for later ELISA analysis. Larvae were confined to these food sources for two days, then transferred to untransformed B. napus leaves. Performance of the larvae in the dishes was measured by how many adults eclosed. Illinois and Florida populations both showed the same probability of survival when fed null plants and when fed proteinase inhibitor expressing plants, but differed in their ability to survive on the two plant phenotypes expressing Bt -endotoxin (Figure 4). The Illinois population was entirely susceptible to the presence of Bt, but one quarter of the Florida population was able to resist the effect of the toxin. The resolution of the bioassay was insufficient to distinguish any effect of proteinase inhibitor expressed alone or in combination with Bt.

EXPERIMENT

To test the predictions of the computer model, we designed an experiment that is now underway in a controlled environment chamber at the University of Chicago. In this experiment we compare the performance of the three cropping strategies (rotations, spatial mosaics, and multiple toxins) when in the presence of moths capable of evolving resistance to the toxins. One replicate of this experiment consists of six cropping treatments and two moth treatments, all replicated twice for a total of twelve twenty-four treatments. Each treatment is an enclosure containing 32 plants in a 0.6 m x 0.5 m x l m wood frame and surrounded by a cover of cloth (summer weight garden cover, Gardener's Supply Company, Burlington, Vermont) that permits light and water entry, but prevents moth escape. The contents of each of cage are outlined in Table 1. The experiment is replicated twice.

Twenty-eight moths were released in each moth treatment in mid-September 1996. In late October their offspring were transferred to cages of fresh plant material that represented the same lines as in the previous generation except in the case of the Rotation treatment. In mid-December and in late January 1997 the moths will be transferred again. After every two moth generations we will conduct a bioassay of the larvae to look for changes in the proportion of resistant individuals (evolution). We are also censusing moth population size every generation.

In the absence of moths the plants are undamaged, and we measure the leaf areas of the plants and their above-ground biomass to compare to the plants in the presence of moths. It is expected that transformed plants will enjoy reduced damage both because they resist damage and because they suppress moth population growth.

CONCLUSIONS

Widespread commercialization of transgenic crops will inevitably lead to the acquisition of resistance in plant pests. This year was the first year in which Bt corn and Bt soybean were planted commercially in the United States. In some regions, Bt cotton was unable to suppress cotton bollworm damage, and there is already speculation that field resistance is beginning to evolve in this species (Fox, 1996). Other nations express hesitation in regard to the commercialization of these transformed crops based on the possibility of pest resistance evolution. If we hope to use these crops for more than a few years, adequate management strategies such as the ones evaluated here must be implemented immediately.

ACKNOWLEDGMENTS

Thanks to 0. Joost and D. Droste for transformations. Thanks to C. Palm for southern blot analysis of backcrossed B. napus transformants. Thanks to C. Eastman at the University of Illinois, Urbana-Champagne and J. Massimino of central Florida for permission to collect moths from their cabbage crops. This research was funded by USDA grant 94-33120-0799 to J.B. and J.W.

REFERENCES

Fox, J. L. 1996. Bt cotton infestations renew resistance concerns. Nature Biotechnology 14:1070.

Fry, J., A. Barnason and R. B. Horsch. 1987. Transformation of Brassica napus with Agrobacterium tumefaciens based vectors. Plant Cell Reports 6:321-325.

Tabashnik, B.E., N.L. Cushing, N. Finson, and M.W. Johnson. 1990. Field development of resistance to Bacillus thuringiensis in diamondback mothError! Bookmark not defined. (Lepidoptera: Plutellidae). J. Econ. Entomol. 83:1671-1676.

Talekar, N.S., and A.M. Shelton. 1993. Biology, ecology, and management of the diamondback moth. Annu. Rev. Entomol. 38:275-301.

Winterer, J. 1995. The ecology and evolution of plant defense, herbivore tolerance, and disease virulence. Doctoral dissertation. University of Washington, Seattle, Washington, USA.

Table 1. Genotypes used for each experimental treatment.

Treatment	Generation 1	Generation 2	Generation 3	Generation 4
Susceptible control	null	null	null	null
Bt control	Bt	Bt	Bt	Bt
PI control	PI	PI	PI	PI
Rotation (rep 1)	Bt	PI	Bt	PI
Rotation (rep 2)	PI	Bt	PI	Bt
Mosaic	Bt + PI	Bt + PI	Bt + PI	Bt + PI
Multiple toxins	BtPI	BtPI	BtPI	BtPI

Figure 1. Evolutionary trajectory. Pest population mean fitness increases as the proportion of resistant genotypes increase.

Figure 2. Strategy trajectories. Evolutionary trajectories of pest fitness exposed to three different plantings of crops expressing two pesticidal transgenes differ. In this example, the genes are lethal to 60% of the susceptible pest population every generation and a spatial mosaic is the most effective strategy.

Figure 3. Strategy comparisons. The best strategy for deploying two toxins depends on how well they kill pests. Weak toxins are most effective in multiply toxic plants (dashed line); strong toxins are most effective when spread heterogenously in a crop, either in rotation (solid line) or in a spatial mosaic (dotted line).

Figure 4. Bioassay results. Illinois (open circles) and Florida (closed circles) moth populations differ in their tolerance of host plants that have Bt toxin.