The resulting sequences were quality-filtered (Phred quality score cutoff = 20), adapter sequences were removed from the beginning of each read, and 20 bp were trimmed from the end of each read to remove lower-quality bases (Cutadapt version 1.1). Trimmed reads were pooled and duplicates were removed before performing the de novo reference transcriptome assembly. Assembly was performed in CLC Genome Workbench (version 4.9 beta) using default settings for de novo assembly and a minimum contig length of 100 bp. Differentially expressed (DE) transcript identification was performed using CLC Genome Workbench (version 5.0). The sequences from each treatment group were then aligned back to the reference transcriptome and DE transcripts compared with the 12° control group were identified using Baggerly’s test with an FDR of 0.10 (Baggerly et al. 2003). The threshold for significance was set at P < 0.01 (FDR-corrected) and greater than two-fold change in expression. The de novo contigs were annotated with sequence descriptions identified through BLAST (Altschul et al. 1990) searches of the NCBI nucleotide database and assigned with GO terms, enzyme codes, KEGG pathways, and InterPro matches using default parameters in Blast2GO (Conesa et al. 2005).