Oecologia (1991) 86:177-182 Oecologia 9 Springer-Verlag 1991 Tritrophic interactions between aphids (Aphis jacobaeae Schrank), ant species, Tyria jacobaeae L., and Senecio jacobaea L. lead to maintenance of genetic variation in pyrrolizidine alkaloid concentration* Klaas Vrieling, Wouter Smit, and Ed van der Meijden Department of Population Biology,Research Group Ecologyof Plants and Herbivores, University of Leiden, P. O. Box 9516, 2300 RA Leiden, The Netherlands Received March 30, 1990 / Accepted in revised form October 19, 1990 Summary. We hypothesize that the tritrophic interaction between ants, the aphid Aphisjacobaeae, the moth Tyria jacobaeae, and the plant Seneciojacobaea can explain the genetic variation observed in pyrrolizidine alkaloid con- centration in natural populations of S. jacobaea. The ant Lasius niger effectively defends S. jacobaea plants infest- ed with A. jaeobaeae against larvae of T. jacobaeae. S. jacobaea plants with A. jacobaeae which are defended by ants escape regular defoliation by T.jacobaeae. Plants with aphids and ants have a lower pyrrolizidine alkaloid concentration than plants without aphids and ants. When these data are fitted to an existing theoretical model for temporal variation in fitness it is shown that varying herbivore pressure by T. jacobaeae in interaction with ants defending aphid-infested plants with a low pyrrolizidine alkaloid concentration can lead to a stable polymorphism in pyrrolizidine alkaloid concentration. Costs of the production and maintenance of pyr- rolizidine alkaloids are not accounted for in the model. Key words: Seneciojacobaea Tritrophic interactions - Genetic variation Pyrrolizidine alkaloids - Herbivory There is substantial individual variation in the concentra- tion of secondary plant products in natural populations (Coley 1983; van der Meijden et al. 1984; Berenbaum et al. 1986; Hartmann and Zimmer 1986; Cates and Redak 1989; Von Borstel et al. 1989). It has been demonstrated for several of these compounds that this variation is partially genetically based (Hanover et al. 1966; Beren- baum et al. 1986; Ostrem 1987). One popular theory explains the maintenance of this variation by the opera- tion of two opposing selective forces, namely: (1) benefits to the plant, because the secondary compound acts as a chemical defence against herbivores and (2) costs asso- ciated with the production and/or maintenance of this * Publication of the "Meijendel-comit6", new series no. 114 Offprint requests to: K. Vrieling chemical defence system (Krischik and Denno 1983; Simms and Rausher 1987, 1989; Rausher and Simms 1989). Although the empirical data are limited, some studies support this theory, for example the monoter- penes in Pinus monticola Dougl. (Hanover 1966), furano- coumarins in Pastinaca sativa L. (Berenbaum et al. 1986, 1989), tannins in Cecropia peltata L. (Coley 1986) and cyanogenic glycosides in Trifolium repens L. (Kakes 1989). In other studies, however, no costs of defence were detected (Gould 1983; Simms and Rausher 1987, 1989; Brown 1988; Rausher and Simms 1989). It may be that in some systems there is indeed a negligible cost to de- fence and that some other form of selective balance is involved. The present study investigates this possibility. In natural populations of the ragwort Senecio jaco~ baea L. the concentration of pyrrolizidine alkaloids (Pa's) varies tenfold (van der Meijden et al. 1984) and is partially genetically determined (Vrieling et al. un- published work). The costs of the production of Pa's expressed as reduction of growth can only explain a small amount of this variation in Pa concentration (Vrieling et al. unpublished work). In 1986 we observed that only those individuals of S. jacobaea that were protected by ants managed to escape total defoliation by the larvae of the moth Tyria jacobaeae L. and were able to produce seeds. These ants were exploiting aphid colonies of Aphis jacobaeae Schrank. If aphids or ants select plants with a low Pa concentration, then, in years with high herbivore pressure by T. jacobaeae these plants should have a higher fitness because they are the only seed-producing individuals. On the other hand, in years with a low herbivore pressure, plants with aphids might have a low- er fitness because of a reduced seed output due to feeding by these aphids. We therefore hypothesize that the genet- ic variation in Pa concentration in S. jacobaea is main- tained through an interaction between aphid-ant mutual- ism and the main herbivore of this plant in the dune system, T. jacobaeae. We will, therefore, address specific- ally the following questions: do ants protect S. jacobaea plants infested with A. jacobaeae?; do aphids select S. jacobaea plants with a low Pa concentration?; can varia-