, Sunnyvale, CA). The median backgrounds were subtracted from the median Cy3 and Cy5 foreground intensity reads for each spot and were log 2-transformed. Technical control spots and spots exhibiting an average signal to noise ratio of less than 3 over all 36 arrays were excluded from further analysis. Signal to noise ratios were calculated as (median foreground−median background)/background standard deviation for each dye. Locally Weighted Scatterplot Smoothing (LOWESS) procedure was utilized to correct for intensity dependent dye bias [17]. A linear mixed model approach

was used to estimate treatment means by fitting the difference of Cy3 and Cy5 normalized signal intensities with each treatment groups’ parameterization for each gene as described by Sandford et al. BAY 73-4506 [64]. Only one random effect, experimental replicate, was included in the model, as likelihood ratio testing determined no effect of slide or array position. P values were obtained for contrasts of interest. False discovery rate was controlled by converting P values to q values using the R package q value [67]. Gene ontology analysis of biological processes for significant genes was performed using the Database for CP-690550 in vivo Annotation, Visualization, and Integrated Discovery (DAVID) [32] and [33]. Quantitative real time PCR (qRT-PCR) was performed as described by [59] to confirm microarray results. The following fifteen genes were selected because of significance

in the microarray study: clusters of differentiation (CD) 3ε, CD4, CD5, CD28, toll-like receptors (TLR) 7, TLR15,

TLR21, heat shock protein 70 (HSP70), P20K, Rab11a, avian beta-defensins (AvBD) 2, AvBD4, AvBD5, AvBD6, and AvBD7. 28S was utilized as a housekeeping gene to normalize for starting concentration of RNA. Primer sequences for CD4, CD5, TLR7, Rab11a, AvBD2, AvBD4, AvBD5, AvBD6, and AvBD7 were designed using sequences from NCBI and PRIMER3 [62]. Primer sequences for 28S, TLR15, and TLR21 have been previously reported (28S [38]; TLR15 [31]; TLR21 [9]). CD3ε, CD28, and P20K were previously utilized [73] but primer sequences were not published. All unpublished primer sequences can be found learn more in Table 1. Each sample was run in three wells. Cycle threshold (CT) values were recorded for each well and each sample triplicate was averaged. Slopes representing reaction efficiency for each gene were generated through amplification of a serial dilution. CT values were adjusted for RNA concentration and reaction efficiency using the formula: 40−[Sample Mean CT Target Gene+(Median 28S for All Samples−Sample Mean 28S)×(Slope Target Gene/Slope 28S)]. Adjusted CT values were analyzed using the Fit Model procedure in JMP software (SAS Institute Inc., Cary, NC). Validation was carried out utilizing RNA extracted from different birds than those included in the microarray analysis representing the same treatment groups and replicates, allowing for both technical and biological replication.