Field data collection
To study variation in the plant–insect trophic network with elevation, we established six elevation transects that covered the diversity of environmental conditions of the Swiss Alps, differing in local climate and bedrock – i.e. in the areas of Bex, Calanda, Faido, Grindelwald, Martigny and Salgesch (Supporting information, Appendix S1, Fig. S1). Each transect was divided into eight sites, spanning elevations from 578 to 2417 m a.s.l., located on average 240 m of elevation apart from each other. Sites were chosen to be open grasslands with a limited impact from anthropogenic activities. At each site, we defined a 10 m x 10 m survey plot representative of the homogeneous composition of the surrounding vegetation. Orthopteran surveys were conducted under sunny weather conditions during the summer at insect peak activity times. We focused on Caelifera and Ensifera suborders that are known to feed on living plant material (Baur et al. 2006). We identified orthoptera by visual inspection, caught on average 10 individuals per species, kept them in falcon tubes for c. 2 h, for collecting fecal excretions, before releasing them again all at once. We performed the vegetation surveys in a 9 m2 circular plot located in the most homogeneous zone of the 100 m2 plot and searched for additional rare species within the 100 m2plot. We used temperature as the main environmental variable that changed along the elevation gradient (Appendix S1, Fig. S2). Soil temperature data were collected for half of the sites per transect using temperature loggers (DS1921G-F5 HomeChip, Newton Longville, England) that were parameterized at a 0.5°C resolution with a sampling rate of 240 minutes, wrapped in parafilm, protected by a silicone capsule and buried 4 cm deep in the ground at each site (from October 2017 to October 2018). Summer soil temperature was extrapolated for unmonitored sites by linear regression (Appendix S1, Fig. S3). To study the keystone plant species, we measured plant functional traits that related to physical resistance or nutrient content: specific leaf area (SLA), leaf dry matter content (LDMC), force required to pierce the leaf lamina (punch), and carbon-to-nitrogen ratio (C/N). We sampled well-developed, healthy leaves to measure the traits of all species with a minimum of three replicates across their elevation range (76% of the total number of surveyed species). SLA and LDMC were measured using standard procedures (Pérez-Harguindeguy et al. 2013). Punch was calculated using a digital force gauge (IMADA CO., LTD. Toyohashi, Japan), following Sanson et al. (2001). C/N was determined by dry combustion of ground leaf material (4mg +/- 0.2mg) of intraspecific replicates pooled to equal weight using an elemental analyzer (NC-2500 from CE Instruments, Wigan, Lancashire, United Kingdom). The collection of trait data was completed with published datasets (Kattge et al. 2011; Körner et al. 2016; Descombes et al. 2017).