Managing the Risk of European Corn Borer Resistance to Transgenic Corn: An Assessment of Refuge Recommendations

Industry and scientists have worked to develop a high-dose refuge management plan that can effectively delay European corn borer (ECB) resistance to new genetically modified pesticidal corn. For a high dose, the corn expresses enough pesticide to kill all but the most resistant corn borers. For refuge, producers plant a traditional corn variety that allows susceptible corn borers to thrive and mate with resistant corn borers, slowing the proliferation of resistance. In general, the more refuge that is planted the less likely resistance. While there is general agreement on the basic premise of the high-dose refuge plan, how much refuge is needed to manage resistance is still being debated. This paper develops a stochastic agricultural production model to assess and provide insight into the reasons why refuge recommendations remain controversial. We find that (a) reducing the risk of resistance requires decreasing agricultural productivity, (b) new technologies that are currently being tested will reduce the risk of resistance, (c) producer noncompliance increases the risk of resistance, and (d) nonrandom mating in the ECB populations increases the risk of resistance. Disagreements over the importance of nonrandom mating and producer compliance can be resolved with additional research. Controversy will remain, however, as long as producers and industry continue to bear most of the financial cost of reducing the risk of resistance through higher refuge recommendations. MANAGING THE RISK OF EUROPEAN CORN BORER RESISTANCE TO TRANSGENIC CORN: AN ASSESSMENT OF REFUGE RECOMMENDATIONS The use of Genetically Modified Organisms (GMOs) in agriculture has been on the rise since 1995. Although these transgenic organisms can embody either innovative product or process characteristics, to this day they have mainly offered new solutions to pest control. In terms of acreage planted, some of the most successful GMOs have been those genetically engineered to kill pests by expressing a protein that is found in the soil bacterium Bacillus thuringiensis (Bt). The potential for pesticidal GMOs to increase agricultural productivity can not be denied; however, empirical evidence suggests these benefits could be diminished by the development of resistance to Bt (Hama et al. 1992; Tabashnik et al. 1992; MartinezRamirez et al. 1995; Tabashnik et al. 1995). As with other highly selective pesticides, controlling pests using a GMO has two important dimensions. First, surviving pests propagate, making the pest population renewable (Regev et al. 1983). Second, pest susceptibility to GMO toxins is generally nonrenewable (Hueth and Regev 1974; Regev et al. 1976; Regev et al. 1983). The increased selection pressure placed on pests by a GMO favors the survival of resistant pests. As resistance develops, the GMO is increasingly less effective for pest control. Because insect pests are mobile, farmers may fail to account for the external benefits of pest suppression efforts and costs of resistance they impose on their neighbors. Concerns about resistance prompted the U.S. Environmental Protection Agency (EPA) to conditionally register the first generation of pesticidal GMOs. The conditional registrations were designed to give industry enough time to develop and implement effective resistance management plans, while also collecting important information on the potential risk of resistance. To manage resistance, industry has focused on a high-dose refuge strategy. For a high dose, the GMO must express extremely high levels of toxin so that all but the most resistant pests survive. For refuge, farmers are expected to plant part of their crop acreage to a crop that does not use the GMOs toxins for pest control. Refuge allows susceptible pests to thrive so they can mate with resistant pests, reducing selection pressure and extending the efficacy of the GMO. Evidence suggests that a high-dose refuge strategy can control resistance effectively in many circumstances provided an adequate amount of refuge is planted. How much refuge is adequate remains a question of debate because of important biological factors that are either random in nature or not precisely 8 / Hurley, Secchi, Babcock, and Hellmich known, and because of economic factors that both determine a stakeholder’s willingness to accept greater risk and influence the effectiveness of the refuge. The purpose of this paper is to develop a model to better understand and provide insight into the debate over refuge recommendations for Bt corn that is used to control the European corn borer in the central and western United States. Currently, recommendations range from 20 to 30 percent when refuge is not treated with pesticides and from 20 to 40 percent when treated with non-Bt pesticides. By exploring key issues of contention within a consistent analytic framework, we provide useful information that can help guide the policy debate over refuge recommendations. Additionally, a Bayesian method is developed that uses pest survival data from the field to quantify the risk of resistance. Incorporating a second gene, which allows us to evaluate the potential of GMOs that express multiple toxins, extends previous bioeconomic models of pest resistance. Pest suppression is explored using alternative stochastic pest population models, and producer compliance with recommendations is addressed using recent survey data. Reviewing the debate over refuge recommendations, we have identified four sources of controversy: (a) the value of reducing the risk of resistance, (b) the potential introduction of new multiple toxin plants, (c) producer compliance, and (d) the degree of nonrandom mating. We find that producers and industry will generally prefer a lower refuge recommendation because they currently bear most of the financial cost of reducing the risk of resistance with higher refuge recommendations. Eventually, plants that express multiple toxins will be commercially available. These new plants will reduce the risk of resistance. Some members of industry support a lower recommendation because they believe these new technologies will soon reduce the need for resistance management. Entomologists and other scientists have supported a higher recommendation because they are less willing to rely on the introduction of these new technologies. Producer noncompliance with refuge recommendations may dictate either a higher or lower recommendation. If a producer’s voluntary compliance decreases as the recommendation increases because of greater compliance cost, a higher refuge recommendation may be necessary to increase the actual size of refuge. However, if a higher recommendation results in lower compliance, an increase in the recommendation can be counterproductive. With the rate of noncompliance we explore, a higher recommendation is appropriate. Nonrandom mating reduces the effectiveness of refuge, resulting in the need for a higher recommendation. Currently, the degree of nonrandom mating is not known. Different assumptions regarding the degree of nonrandom mating have therefore resulted in different recommendations. Managing the Risk of European Corn Borer Resistance to Transgenic Corn / 9 The Conceptual Model Optimally, refuge recommendations should adjust over time as pest susceptibility becomes increasingly scarce and as new information becomes available to resolve unanswered questions. Deriving an optimal exhaustion path ex ante is a daunting task because the optimal path is conditioned on what new information arises in the future. The types of new information that can arise include the realization of unpredictable events such as the level of pest pressure in a particular season, new experimental or field data that elucidates unknown factors such as the current level of resistance to Bt, and the realization of unforeseen events such as the discovery of substitute or complementary technologies. Alternatively, imposing an inflexible, safe minimal recommendation based on the best available information is a more manageable task, though not optimal because it does not incorporate new information as it becomes available or address the increasing scarcity of pest susceptibility. The impracticality of designing temporally optimal refuge recommendations and the inflexibility and suboptimality of a temporally inflexible safe minimal recommendation have resulted in the adoption of an adaptive strategy. The adaptive strategy sets a safe minimal recommendation to satisfy a specific set of objectives, but revises this recommendation as available information warrants. Currently, two objectives that have been identified are (1) the preservation of pest susceptibility and (2) the maintenance of the agricultural benefits provided by Bt corn. Specifically, the International Life Sciences and Health and Environmental Sciences Institutes (ILSI-HESI) in a recent report explicitly defined the preservation of pest susceptibility as the maintenance of resistant allele frequencies below 0.50 for 30 generations. Unfortunately, less specificity has been given to objective (2). The model we develop builds on Hurley et al. (1997) by incorporating important sources of uncertainty related to the current level of resistance, the survival rate of resistant pests, and European corn borer (ECB) population dynamics. Focusing on a simplified production region with a single crop (corn) and pest (ECB) concisely illustrates important issues that influence refuge recommendations. The scope of the region we consider is defined by pest mobility where we assume migration is negligible. There are two varieties of corn planted in the region: a Bt corn event, denoted by Bt, and a non-Bt isoline that is used as refuge, denoted by Non. Define Ν as the proportion of the refuge recommended, while Νt and 1 Νt is the effective proportion of refuge and Bt corn planted in season t. The effective proportion of refuge planted can depend on the refuge recommendation, pest pressure, voluntary compliance rates, and Bt corn adoption rates, for instance. 10 / Hurley, Secchi, Babcock, and Hellmich Genetic Variation Current Bt events utilize one of three toxins (Cry I(A)b, Cry I(A)c, or the more recently registered Cry 9c), but in the future new events may express multiple toxins. To be able to explore the resistance management benefits of multiple toxins, we use a genetic model with two genes (Hartl 1988) that are denoted by a and b. A gene consists of two alleles, one contributed by the mother and the other one by the father. We assume there are only two types of alleles for each gene, one conferring resistance, R and R, with all others conferring susceptibility, S and S. A gamete is a combination of alleles, one for each gene, passed to an offspring by the mother or father. With our two-gene model there are four distinct types of gametes that a parent can pass to its offspring: resistant alleles for both genes, R|R; susceptible alleles for both genes, S|S; or one of two combinations of resistant and susceptible alleles, R|S or S|R such that (∈{R|R, R|S S|R, S|S}. Combining the gametes from the mother and father, (×(, produces the 16 possible allelic combinations illustrated in Table 1. For the θth gene, θ∈{a, b}, the offspring can be either a resistant homozygote, RR; a susceptible homozygote, RR; or a heterozygote, RS or SR. Therefore, although there are 16 different allelic combinations, some are redundant, resulting in nine distinct genetic combinations that may differ in terms of their survival on Bt corn and refuge. Table 1: Punnett square of allelic types Mother’s Contribution Gamete types R|R R|S S|R S|S R|R RR| RR RR| SR SR| RR SR| SR R|S RR| RS RR| SS SR| RS SR| SS S|R RS| RR RS| RS SS| RR SS| SR Father’s Contribution S|S RS| RS RS| SS SS| RS SS| SS Biological Dynamics Midwestern ECB populations are normally bivoltine (i.e., produce two generations per season). The level of ECB pressure experienced in a region varies dramatically depending on pest pressure in the previous generation and other random climatic and environmental factors (Royama 1992). Define Nt g as the average number of ECB per plant emerging at the beginning of a generation, while Nt g S is the average number of ECB per plant surviving to damage crops and reproduce in season t and generation g. (1) Nt+1 1 ∼ N1(Nt 2) and Managing the Risk of European Corn Borer Resistance to Transgenic Corn / 11 (2) Nt 2 ∼ N1(Nt 1). Let Equations (1) and (2) state that the ECB population emerging at the beginning of season t and generation g is randomly distributed conditional on the previous generation’s surviving population. They also assume that reproductive rates are not influenced directly by genetic variation and that the effect of the previous generation on the first generation of ECB may differ from the effect of the second generation on the first, usually due to the overwintering. While reproductive rates are not influenced directly by genetic variation, the size of the surviving pest population is influenced. In particular, the surviving pest population will depend on the emerging pest population, genetic variation, the effective proportion of refuge, and genetic survival rates on Bt corn and refuge that are related to overall fitness and susceptibility to Bt. Define ∋t g j as the proportion of the jth gamete in season t and generation g where ∋t g j ∈ [0.0, 1.0] for j∈γ, Σj∈γ ∋t g j = 1.0, and ∋t g as the vector of these proportions that is currently not precisely known. Define ∆g and ∆g as vectors of survival rates on Bt corn and refuge for the nine distinct genetic combinations in generation g. As with current levels of resistance, these survival rates are not precisely known. Surviving pests on the ith crop, i ∈ {Bt, Non}, can now be written as the average survival rate of pests, which is a function of gamete frequencies and survival rates, multiplied by the emerging pest population: (3) Nt 1 i = ρ(∋t 1, ∆1) Nt 1 and (4) Nt 2 i = ρ (∋t 2, ∆2) Nt 2. All else equal, the surviving population tends to increase as the proportion of gametes with resistant alleles increases, survival rates increase, and the emerging pest population increases. The average number of surviving pests is then an average of pests on Bt corn and refuge weighted by the effective size of the refuge: (5) Nt 1 S = ΝtNt 1 + (1 Νt) Nt 1, and (6) Nt 2 S = ΝtNt 2 + (1 Νt) Nt 2. Finally, the proportion of gametes containing resistant alleles evolves depending on previous levels of resistance, survival rates, and the degree of selection imposed by the effective size of the refuge: 12 / Hurley, Secchi, Babcock, and Hellmich (7) Γt 1 = Γ(Γt-1 2, ∆2, ∆2, Νt-1), and (8) Γt 2 = Γ(Γt 1, ∆1, ∆1, Νt). All else equal, resistance tends to increase as the survival rate of genetic combinations with susceptible alleles decreases and as the effective size of the refuge declines. Biological Objectives The primary biological objective that has been specified for choosing the refuge recommendation is to prolong the efficacy of Bt corn by preserving pest susceptibility over the length of a planning horizon, T. To formalize this objective, we first must specify what preserving pest susceptibility means. The ILSI-HESI report defines preserving pest susceptibility as not allowing resistant gametes to exceed an acceptable threshold. Since resistance is generally treated as irreversible within a reasonable period of time, this objective implies that resistance should not exceed an acceptable level at T. For events expressing a single toxin, the definition of a resistant gamete is straightforward. The appropriate definition of a resistant gamete when events express multiple toxins is more complicated. Since the current level of resistant gametes and survival rates is not precisely known, this definition can only be satisfied in a probabilistic sense; therefore, an acceptable rate of error in meeting this objective must also be specified. An explicit biological objective can be formalized by defining three constraints that must be satisfied by the refuge recommendation: (9) ( ) Ψ − ≤       ≤ Γ + Γ = Ψ 1 Pr , 2 2 a S R T R R T a a b a b a θ θ φ (10) ( ) Ψ − ≤       ≤ Γ + Γ = Ψ 1 Pr , 2 2 b R S T R R T b b b a b a θ θ φ , and (11) ( ) Ψ − ≤        ≤ Γ = Ψ 1 Pr , 2 ab R R T ab ab b a θ θ φ where θ, θ, and θ are the maximum acceptable proportion of resistant alleles at the beginning of T for gene a, gene b, and both genes concurrently; and Ψ is the maximum acceptable rate of error. If an event expresses a single toxin that selects for resistance only at gene a, for example, we can focus on resistance at gene a by setting θ = θ ∈ (0.0, 1.0) and θ = 1.0. For events that express two toxins, the appropriate set of objectives may be (1) θ ∈ (0.0, 1.0), θ ∈ (0.0, 1.0), and θ = Min{θ, θ} Managing the Risk of European Corn Borer Resistance to Transgenic Corn / 13 or (2) θ ∈ (0.0, 1.0) and θ = θ = 1.0. For (1), the constraints imply that the objective is to preserve susceptibility for both genes independently, which might be reasonable if multiple toxin events do not completely replace single toxin events. For (2), the constraints imply an objective of preserving pest susceptibility only for multiple toxin events. Currently, whether (1) or (2) should be used to evaluate multiple toxin events has not been addressed explicitly. Agricultural Benefits The second objective that has been identified for refuge recommendations is the preservation of the agricultural benefits of Bt corn. While an explicit definition of the agricultural benefits of Bt corn has not been articulated, we believe the spirit of this objective is captured by the expected net present value of agricultural production. Assume the expected pest-free yield for both crops is Y bushels/acre, but that actual yields may vary due to pest pressure. Let Dt i = d(Nt 1 , Nt 2 ) be the average proportion of yield loss on the ith crop where Dt i ∈ [0.0, 1.0] is strictly increasing in the number of pest per plant in both generations. The average yield per acre for the ith crop is Yt i = Y(1 – Dt ) in season t. Let C be the average per acre cost of production for the ith crop. Finally, let P be the expected real price of corn and δ be the discount rate. The expected net present value of agricultural production given random pest populations and uncertainty regarding current levels of resistance and survival rates is (12) ( ) ( ) [ ] { ( ) ( ) [ ]} 