456 Opinion TRENDS in Ecology and Evolution Vol.18 No.9 September 2003 Indirect effects of host-specific biological control agents Dean E. Pearson1,2 and Ragan M. Callaway1 1Division of Biological Sciences, University of Montana, Missoula, MT 59812, USA 2Current address: USDA Forest Service, Rocky Mountain Research Station, Box 8089, Missoula, MT 59807, USA Biological control is a crucial tool in the battle against discuss the indirect effects of biocontrol agents in the biological invasions, but biocontrol agents can have a context of the current theory and practice of biocontrol and deleterious impact on native species. Recognition of recent advances in community ecology to illustrate that risks associated with host shifting has increased the host-specific biocontrol agents can exhibit nontarget emphasis on host specificity of biocontrol agents for effects on native species and ecosystems. We focus our invasive weeds. However, recent studies indicate host- discussion on insect biocontrol of invasive plants, but the specific biocontrol agents can also exhibit substantial general concepts also apply to biocontrol of invertebrate nontarget effects through indirect interactions and pests [21]. food-web subsidies. Based on an evaluation of these studies, we conclude that the interaction strength Biocontrol in theory and practice between biocontrol agents and their hosts is at least as The conceptual model underlying biocontrol was derived important as host specificity for determining the out- from predator–prey theory (Fig. 1) [22,23] and is based on come of biocontrol introductions. Host-specific biocon- the notion that exotic species become invasive by escaping trol agents that establish, but fail to reduce the the controlling influence of their natural enemies [24–26]. densities of their hosts can facilitate bottom-up effects We refer to this model as the ‘natural enemies model’. In that link the target weed to other native organisms this model, control of the invasive species is achieved when through food webs, thereby expanding the impacts of the introduction of its natural enemies reduces its mean the invasive weed. We believe that indirect nontarget equilibrium density below some economically or ecologi- effects of host-specific biocontrol agents arising from cally defined threshold [22,23]. The model predicts a direct food-web subsidies could prove more deleterious to negative effect of the biocontrol agent on its intended host native species than are the direct nontarget effects cur- that translates into an indirect positive effect on native rently recognized from host shifting. species and a negative feedback that reduces and regulates its own populations (Fig. 2). Biological invasions increasingly threaten the biological When biocontrol is successful, it is truly elegant. Cases diversity and ecological integrity of natural ecosystems [1]. of successful biocontrol demonstrate that top-down control Classic biological control (referred to henceforth as biocontrol), the importation of natural enemies for the control of exotic, invasive species, is a powerful manage- ment tool that has proven effective at controlling numer- Mean equilibrium density ous invasive species [2,3]. However, biocontrol also poses significant risks to nontarget, native taxa [4–8]. The most notable examples of nontarget effects occur when biocon- trol agents expand their host range to attack native Threshold density organisms (host shifting) [3,5,6,9–12]. Recognition of the risks associated with host shifting has spawned an intense density Pest Mean equilibrium density debate over the importance of the host specificity of biocontrol agents [2,3,5–7,13–17]. We believe that this Biocontrol release debate has improved the field of biocontrol of exotic plants by establishing host specificity-requirements for biocon- Time trol agents that reduce risks for nontarget species [2,3,14]. However, recent studies examining insect biocontrol of TRENDS in Ecology & Evolution invasive plants indicate that even host-specific biocontrol Fig. 1. Theoretical control of pest species using biocontrol agents, illustrating how agents can incur significant nontarget effects on native pest densities might fluctuate over time before and after the introduction of a suc- species [18–20]. Yet, indirect threats have not been cessful biocontrol agent. Before the introduction of the biocontrol agent, the pest density fluctuates around a mean equilibrium density that is above a threshold of considered in biocontrol theory or practice. Here, we economic or ecological impact. Following the introduction of the biocontrol agent, the pest densities stabilize at a new equilibrium level that is below the threshold of Corresponding author: Dean E. Pearson ([email protected]). impact. Adapted, with permission, from [23]. http://tree.trends.com 0169-5347/$ - see front matter. Published by Elsevier Ltd. doi:10.1016/S0169-5347(03)00188-5 Opinion TRENDS in Ecology and Evolution Vol.18 No.9 September 2003 457 cases, in biologically significant negative impacts on those species [5,6,9–12]. Recognition of the risks associated with host shifting has led to an increased emphasis on host Biocontrol insect specificity in screening prospective biocontrol agents [14]. However, most biocontrol introductions result in the establishment of host-specific biocontrol agents that + exhibit weak negative effects on their host [30]. Such agents are presumed safe because host specificity is + – believed to ensure their neutrality toward nontarget species [2,3,22]. This assumption has contributed to the ‘multiple release’ strategy in biocontrol [4,31,32] that advocates introducing multiple agents for each target – species with little regard for interaction strength. The ironic result of this approach is that exotic biocontrol Invasive plant Native plants – insects now far outnumber the exotic plants that they were introduced to control [30,31], and new research [18–20] suggests that host-specific biocontrol agents can incur TRENDS in Ecology & Evolution strong nontarget effects through indirect interactions associated with ecological replacement, compensatory Fig. 2. Theoretical control of pest species using biocontrol agents presented in the context of the natural enemies model, illustrating direct (straight lines) and indirect responses and food-web subsidies. effects (curved lines) predicted by the model. Line weight indicates interaction strength. Dotted lines indicate empty niche of primary consumer and postulated Nontarget effects of host-specific biocontrol agents effects of introducing a biocontrol agent. This illustrates how the natural enemies model focuses on the direct negative effect of the biocontrol on the target pest and Ecological replacement the resulting indirect positive effect on the native species, but ignores other com- Indirect nontarget effects can occur through ecological munity interactions that might arise within the system (i.e., they are not addressed by the model). replacement when a biocontrol is used against a pest that has become integrated into the native community by can be achieved over invasive species using natural physically or functionally replacing native species. enemies when the biocontrol behaves as a keystone species Although biocontrol in cases involving ecological replace- (an organism with community effects that are dispropor- ment can result in indirect nontarget effects on native tionately large relative to its abundance) [27]. This point is species, this is not a failure of the biocontrol or the natural well illustrated by the control of klamath weed Hypericum enemies model. The natural enemies model predicts that perforatum by the chrysomelid beetle Chrysolina quad- those organisms directly interacting with the pest species rigemina [28]. Klamath weed is an exotic forb that had will be affected by its control (Fig. 2), and well established invaded nearly 1 million ha of rangeland in northern exotic species can be expected to develop interactions with California by the mid 1900s, causing dramatic reductions native organisms. For example, biocontrol of exotic in the biomass and diversity of native species. The European rabbits Oryctolagus cuniculus in the UK is introduction of C. quadrigemina reduced Klamath weed believed to have resulted in the extirpation of the large to ,1% of its peak invasion densities and facilitated blue butterfly Maculina arion through a series of indirect recovery of the native system within 12 years of the initial effects that fatally linked this species to the rabbits [33]. release. These two species currently persist at densities The large blue required nests of the ant Myrmica sabuleti well below the threshold of ecological impact. Presumably, for the development of their larvae. These ants in turn experimental removal of C. quadrigemina would result in were dependent upon rabbit grazing to maintain open klamath weed recovering its former range at great expense habitat for their nests, so biocontrol of the rabbits with the to the native community. This keystone phenomenon is Myxoma virus initiated a cascade of interactions believed also demonstrated by other successful biocontrol agents in to have led to the extinction of the large blue. Recognition natural systems [2,3] and illustrates the importance of of the risks associated with indirect nontarget effects from interaction strength (the population-level impact of one ecological replacement has helped to avoid repeating the species on another [29]) for achieving successful control. story of the large