Hepatopancreatic multi-transcript expression patterns in the crayfish Cherax quadricarinatus during the molt cycle

Variations of hepatopancreatic transcript expression related to induced molt cycle were identified in male Cherax quadricarinatus using a cDNA microarray estimated to contain 2180 unique sequences. Molt induction was performed by X-organ sinus gland extirpation or by repeated 20-hydroxyecdysone injections. Manipulated males were sacrificed at premolt or early postmolt and a reference population at intermolt. Their isolated hepatopancreatic mRNA was hybridized onto the microarray and differentially-expressed genes were identified, sequenced, and annotated. Clusters of similarly expressed transcripts, in correlation with the four combinations of induction methods and molt stages were identified, containing both known and novel genes. Biologically interesting clusters were characterized by general shift of expression throughout premolt and early postmolt vs. intermolt, or by different premolt vs. postmolt expression. Several genes were differentially expressed in 20-hydroxyecdysone injected crayfish vs. X-organ sinus gland extirpated ones. Keywords: Physiological state analysis The experiment submitted in this series includes male Cherax quadricarinatus in five experimental conditions. Individual variability was integrated into the hybridization design, by using RNA populations of 3 individual crayfish males for each of four experimental conditions: injected-premolted, injected-postmolted, extirpated-premolted and extirpated-postmolted males. Three pools of reference intermolt individuals were used: Ref pool 1 contained RNA from 6 individuals whereas each of Ref pools 2 and 3 contained 3 individuals. Three hybridization sets which are biological replicates, each composed of 8 slides, were carried out. In each set, one individual from each treatment was compared to one of the Ref pools, and in addition the treated individuals were compared to each other in a loop design, totaling 8 slides. The design was balanced with respect to the dye labeling of the RNA populations, such that each individual RNA population was labeled with one dye on one slide and with the other dye in two other slides (see graphic presentation of the hybridization design in linked file GSE6947_Hybridization_design.pdf). The entire 24 slides set was analyzed as one integrative experiment, resulting with eight binary comparisons among experimental conditions as follows: extirpated-premolted vs. intermolt reference (Epre_Ref); extirpated-postmolted vs. intermolt reference (Epost_Ref); injected-premolted vs. intermolt reference (Ipre_Ref); injected-postmolted vs. intermolt reference (Ipost_Ref); extirpated-premolted vs. injected-premolted (Epre_Ipre); extirpated-premolted vs. extirpated-postmolted (Epre_Epost); injected-premolted vs. injected-postmolted (Ipre_Ipost); extirpated-postmolted vs. injected-postmolted (Epost_Ipost). Data analysis was conducted under the R/BioConductor environment (Gentleman et al., 2005). Calculation and statistical analysis of log2 expression ratios (VALUE column of the sample data table) of each unique spot across a binary comparison was carried out using the LIMMA package (Linear Models for Microarray data Analysis; Smyth, 2004). GENEPIX quantified image files were imported into LIMMA, spots with quality flags < -49 were marked as unreliable, and the Cy5 and Cy3 intensities within each slide were normalized using the print-tip Loess method. No background correction was applied. LIMMA calculated for each spot on an individual slide an M-value [M=log2(Cy5)/Cy3); Cy5 / Cy3 are the normalized emission intensities of the spot] and an emission intensity A-value [A=(log2(Cy5)*Cy3))/2]. Subsequently, LIMMA fitted a linear model to the expression data followed by Empirical Bayes smoothing. Inter-duplicate correlation method was used at the linear modeling step. A meanA value, which is the average spot emission intensity across the analyzed experiment, was calculated for each unique spot and only spots with meanA > 8.5 were included in the analysis, avoiding faint emissions. The LIMMA summary statistics of each unique spot included an estimated average M-value across each binary comparison associated to two statistical parameters: B statistic, the log-odds that the spot is differentially expressed and p-value, derived from a moderated t-statistic after multiple testing correction according to Benjamini and Hochberg (1995). All these parameters are found in the sample data tables.