Non-choice feeding bioassays
To study the effect of short and long term attack, MCB larvae were fed with stems of plants infested 48 hours (short-term attack) or nine days (long-term attack) before bioassay establishment or non-infested plants. 1st instar MCB larvae were initially weighted and individually placed in plastic tubes on 2 cm sections of the stem below the main ear. Larvae were previously fed on maize-based artificial diet and maintained at starvation for 24 h, presenting weights of 1-3 mg at the beginning of the bioassay. Sixty larvae per treatment and genotype were set and maintained in a growth chamber under controlled conditions of temperature and humidity (22 °C, 80% RH) and a photoperiod of 16L:8D. When necessary, new fresh stem portions of plants infested 48 hours and 9 days before and control plants were provided to the bioassay larvae. Larval weights and data related to dead larvae were recorded at 7, 11, 14, 18, 22, and 26 after bioassay establishment.
A repeated-measure analysis was performed to test differences for larval weights using the PROC GLIMMIX procedure of SAS software (SAS, 2008, Stroup, 2013). Initial larval weight was included as covariate, genotype and treatment were set as fixed factors, and a first-order autoregressive covariance structure (AR-1) was chosen in the within-subject correlation. Within each genotype, differences for larval weight of treatments were tested at each time using least square (LS) means adjusted by the initial larval weight. Additionally, linear and quadratic coefficients of regression of larval weight on time were obtained for each treatment-genotype combination. Within each genotype, the comparison of the larval growth curves of treatments was performed by making orthogonal contrasts among treatment regression parameters (intercept, linear and quadratic components, respectively) (p ≤ 0.05)(Littell, Milliken, Stroup, Wolfinger & O, 2006). The PROC LIFETEST procedure of SAS software was used to test differences of larval survival among treatments applied to the same genotype using the Kaplan-Meier method (SAS, 2008). The death of larvae was the event of interest and the missing and alive larvae at the end of the bioassay were treated as censored data. The homogeneity of the survival distributions was tested using the Šidák multiple-comparison adjustment for log-rank test (p ≤ 0.05).