The 'high dose-refuge' (HDR) strategy is commonly recommended and currently used for delaying or preventing pest adaptation to transgenic plants producing Bacillus thuringiensis (Bt) toxins. The efficiency of this strategy depends, among other factors, on the initial frequency of Bt resistance alleles and on the fitness costs associated with these alleles. Two years ago, an allele conferring resistance to Bt poplar was detected in a French population of the poplar pest Chrysomela tremulae F. Although this pest had never been subjected to Bt selection pressure due to human activities, the frequency of this allele was estimated at 0.0037, with a 95% credible (CI) interval of 0.00045-0.0080. We investigated the frequency of this allele in a second sample of C. tremulae collected more than 500 km from the site of the initial population. The estimated frequency in this sample was 0.0113 (95% CI 0.0031-0.0247), reinforcing the conclusion that resistance to Bt plants may be present at detectable frequencies in pest populations before selection resulting from pest management by humans. The frequency of the Bt resistance allele over the two samples was 0.0049 (95% CI 0.0020-0.0091). We also followed five laboratory lines in which the frequency of this allele was initially fixed at 0.500. After five generations maintained on non-Bt poplar leaves, the frequency of this allele decreased in all lines, whereas allelic frequencies at a neutral locus were unaffected. Thus, the Bt resistance allele detected in French populations of C. tremulae is probably associated with a fitness cost.

1. Metabarcoding has revolutionized the study of ecological communities, but PCR bias hampers quantitative analyses, as required in studies of trophic interactions. Direct DNA shotgun sequencing can avoid the amplification step when unassembled reads are mapped to a reference database, enabling identification and quantification of prey items. 2. Two feeding bioassays tested the precision and accuracy of quantitative assessments with the coccinellids Harmonia axyridis and Hippodamia convergens, the chrysopid Chrysoperla externa, and the aphid Myzus persicae. Guts were dissected and the total DNA extracted was directly sent to sequence by Illumina HiSeq2500 (insert 350 bp, PE 250). Predator gut content reads were blasted against an arthropod mitochondrial DNA reference database for taxonomic assignment and the matches curated for false positive prey identification using a series of bioinformatics pipelines, mostly in R. Taxonomic assignment through KrakenUnique, which is based on the counts of unique k-mers, was compared. 3. In a Prey Quantity bioassay, the number of prey reads was correlated to the amount of prey consumed and the elapsed time since consumption. In a Direct and Indirect Predation bioassay, prey was detectable in the predator 6 h after feeding (direct predation) and the prior prey of the prey was detectable 3 h after feeding (indirect predation). Detection of indirect predation was related to the species-specific decay rates of the two predators, but not to the order of predation. 4. We demonstrate that degraded prey DNA was quantifiable across trophic levels with high accuracy (98.4% positive predictive value) and taxonomic resolution.