Plant sampling and herbivore bioassays
To analyse constitutive leaf GSL production, the 14 species ofCardamine were sampled at the flowering stage, from May until August, following the natural phenology of the plants. The flowering stage for sampling was chosen in order to avoid potential ontogenetic effects on plant chemistry (Barton & Koricheva 2010), since most species flower very rapidly and for long periods, sometimes throughout the whole growing season. All plants (n = 10 per species) were collected directly from the field in each species’ optimal habitat (Aeschimannet al. 2004).
For each plant species, we performed insect resistance bio-assays (N = four insect species × 10 replicates = 40/plant spp.) with P. brassicae, S. littoralis, B. brassicae and M. persicae . To this end, individual plants that were visibly not damaged by herbivores, and with a minimum distance of 10 m apart, were carefully excavated, transplanted in cylindrical 20 cm diameter plastic pots by adding common potting soil where needed (Ricoter AG, Aarberg, Switzerland), and placed in a climate-controlled chamber (14:10 hrs and 23:15 °C day: night, and 55% relative humidity). For assays with chewing herbivores, we used one 6-days old caterpillar per plant, while for sap-feeding aphids we used one adult per plant. For each herbivore species, we randomly choose two fully-expanded leaves per plant and placed them in a Petri dish on a filter paper moisten with one drop of distilled water. After five days of feeding, we estimated plant resistance against the two different feeding guilds by calculating larval gain weight for caterpillars using the formula: ln (final fresh weight - initial fresh weight), and the number of progenies for aphids. We specifically used detached leaves in order to avoid plant defence induction due to feeding, since here the aim was to correlate the measured chemical diversity of GSLs across different species (see below) with insect resistance.