Weed-biocontrol insects reduce native-plant recruitment through second-order apparent competition.

Small-mammal seed predation is an important force structuring native-plant communities that may also influence exotic-plant invasions. In the intermountain West, deer mice (Peromyscus maniculatus) are prominent predators of native-plant seeds, but they avoid consuming seeds of certain widespread invasives like spotted knapweed (Centaurea maculosa). These mice also consume the biological-control insects Urophora spp. introduced to control C. maculosa, and this food resource substantially increases deer mouse populations. Thus, mice may play an important role in the invasion and management of C. maculosa through food-web interactions. We examined deer mouse seed predation and its effects on seedling emergence and establishment of a dominant native grass, Pseudoroegneria spicata, and forb, Balsamorhiza sagittata, in C. maculosa-invaded grasslands that were treated with herbicide to suppress C. maculosa or left untreated as controls. Deer mice readily took seeds of both native plants but removed 2-20 times more of the larger B. sagittata seeds than the smaller P. spicata seeds. Seed predation reduced emergence and establishment of both species but had greater impacts on B. sagittata. The intensity of seed predation corresponded with annual and seasonal changes in deer mouse abundance, suggesting that abundance largely determined mouse impacts on native-plant seeds. Accordingly, herbicide treatments that reduced mouse abundance by suppressing C. maculosa and its associated biocontrol food subsidies to mice also reduced seed predation and decreased the impact of deer mice on B. sagittata establishment. These results provide evidence that Urophora biocontrol agents may exacerbate the negative effects of C. maculosa on native plants through a form of second-order apparent competition-a biocontrol indirect effect that has not been previously documented. Herbicide suppressed C. maculosa and Urophora, reducing mouse populations and moderating seed predation on native plants, but the herbicide's direct negative effects on native forb seedlings overwhelmed the indirect positive effect of reducing deer mouse seed predation. By manipulating this four-level food chain, we illustrate that host-specific biological control agents may impact nontarget plant species through food-web interactions, and herbicides may influence management outcomes through indirect trophic interactions in addition to their direct effects on plants.