normalization data transformation protocol
Median raw spot intensity data were log2 transformed, lowess-normalized to correct for intensity bias (JMP-Genomics; SAS Institute Inc., Cary, NC), then normalized using linear mixed models to account for fixed (dye) effects and random (array) effects, then quantile normalized.
array scanning protocol
slide images were captured using the Packard Bioscience ScanArray Express (PerkinElmer Inc., Waltham, MA), and spot intensities were assayed for each channel using ImaGene software (Biodiscovery Inc., El Segundo, CA)
nucleic acid hybridization to array protocol
Samples were competitively hybridized to Agilent custom oligonucleotide microarrays (Earray design ID 027999) and incubated at 60 degrees C for 18 hrs in a SureHyb microarray hybridization chamber (Agilent Technologies, Santa Clara, CA).
nucleic acid labeling protocol
Samples were coupled with Alexa fluor¬ dyes (Alexa fluor 555 and 647; Molecular Probes, Inc.)
Gill tissues were immediately dissected from adult animals captured from field sites and stored in RNAlater
We compared trajectories of gene expression change through time among fish collected from different field sites using mixed model analysis of variance in JMP Genomics, where main effects were 'species', 'population', and 'time', and all three two-way interactions and one three-way interaction terms were specified. Five biological replicate fish were included per treatment; replicates were not pooled. 'Dye' was treated as a fixed effect, 'array' was treated as a random effect, and replicate individuals within treatment (time-by-population-by-species) was treated as a random effect.
nucleic acid extraction protocol
Gill RNA (following tissue preservation in RNA later) was extracted following tissue homogenization in Trizol reagent (Invitrogen Inc.) and spin column purification.