Statistical analyses
Results are expressed as the mean ± se of n replicates used per treatment. The normality of the distribution and homogeneity of variances were previously tested (Shapiro-Wilk, Bartlett and Levene tests respectively). Two-Way ANOVA was performed for temperature (T), S, and T x S interaction effects on the measured variables, and classification of temperature modalities (T-modalities) was carried out by comparing averages (Tukey test). The differences between HS and LS for a given T-modality were tested with a t-test. These statistical analyses were performed with XLStat software. To determine the non-additive effects of temperature events (Figure 1a), we compared the relative effect of the events-combined sequence (mod1, CES) with the relative cumulated effects of the successive single-event sequence (SES, mod2 and mod3) on the measured variables under both S supplies. The relative effect was calculated as the increase or decrease relative to the control value. The relative cumulated effect of the successive SESs was then calculated by multiplying the two single relative effects (mod2 x mod3).