Microarray images were visually inspected and quality control of the data was done using the Affymetrix array QC pipeline at ArrayAnalysis.org (www.arrayanalysis.org, BiGCaT Maastricht University, the Netherlands) together with RMA normalization using the Brainarray custom CDF version 15 for annotation (http://brainarray.mbni.med.umich.edu). Expression values were log2-transformed. Principal Component Analyses (PCA) (Pearson, 1901) was performed with Qlucore Omics Explorer 2.2 on the log2-transformed normalized data to look for nuisance effects of experimental variables 'week batch' and 'RNA isolation batch'. These are in the column 'Comment[EXP]'. A clustering of samples was observed according to 'week batch'. For this reason we fitted a linear model (Kerr et al., 2000; Wolfinger et al., 2001; Cui and Churchill, 2003) with compound and concentration group and 'week batch' as fixed factors, as 'week batch' did not have enough levels to effectively model as random. We used the microarray analysis of variance (MAANOVA) package (Wu et al., 2003) in R (version 2.14.2) for this and removed the variance appointed to 'week batch' from the data. As all RA treated samples were treated in week batch 3 and this could affect the apparent week batch 3 effect, we fitted the model without RA and transferred variance appointed to week batch 3 to RA data afterwards. The ratio of Treated vs control was taken, and results are in the file Sannedata3AllTreat_Comp_NoWeekfixedResicoeffslogratio.txt as uploaded to ArrayExpress.
The hybridization, washing and staining was performed with the Affymetrix GeneTitan Hybridization, Wash, and Stain Kit for WT Array Plates (901622) (Affymetrix, Inc, Santa Clara, CA). In total, 96 arrays were used for all experimental groups.
RNA labelling and microarray hybridization were performed by ServiceXS B.V. (Leiden, Netherlands). 100 ng total RNA was used to synthesize cRNA (minimal yield 10 ug) followed by sense stranded cDNA (ss cDNA) with the Ambion WT Expression kit (#4411974). Fragmentation and terminal labelling was done with 5.5 ug of ss cDNA using the Affymetrix Terminal Labelling Kit (901524). 25 ng/ul fragmented ss cDNA was utilized for hybridization on Affymetrix GeneChip Zebrafish 1.0 ST Array.
In short, 30 minutes after spawning we collected fertilized batches of eggs (>90% fertilisation) and washed them several times in Dutch Standard Water (DSW; demineralized water supplemented with 100 mg/l NaHCO3, 20 mg/l KHCO3, 200 mg/l CaCl2 2H2O, and 180 mg/l MgSO4 7H2O and then aerated for 24h at 27C). Eggs were evenly distributed among the test concentrations after removal of coagulated eggs or other debris. Next, we selected embryos within the 4- to 64-cell stage and transferred them to a 24-well plate containing the same test medium. One embryo was transferred to one well. For the next 24 or 72 hours embryos were allowed to develop in an incubator at 26.5C +- 1C with a photoperiod of 14h light: 10h dark.
Embryos were exposed to four concentrations of caffeine (CAFF; CAS No. 58-08-2), carbamazepine (CBZ; CAS No. 298-46-4), all-trans-retinoic acid (RA; CAS No. 302-79-4) and valproic acid sodium salt (VPA; CAS No. 1069-66-5). With concentration-range finding studies, we determined the benchmark concentration (BMCGMS) of a 1, 5, 10 and 20% decrease in GMS. These concentrations were selected for this study. For CAFF 75, 215, 640 and 1750 uM were used. CBZ was tested at 70, 130, 250 and 480 uM. RA was tested in the nanomolar range at 0.4, 1.1, 3 and 8 nM. Exposure concentrations for VPA were 150, 320, 730 and 1500 uM. We also included two negative control compounds, D-mannitol (DMAN; CAS No. 69-65-8) and saccharin sodium salt hydrate (SACCH; CAS No. 82385-42-0, all from Sigma-Aldrich, Zwijndrecht, the Netherlands), both at 10 mM. RA and CBZ were dissolved in DMSO and further diluted in DSW to their final concentrations with 0.2% DMSO (vol/vol) present. CAFF, VPA, DMAN and SACCH were directly dissolved in DSW and further diluted and DMSO was added to obtain a similar concentration of solvent. Controls were exposed to DSW with 0.2% DMSO.
For transcriptomics analysis, samples of 20 pooled embryos were used for RNA isolation. 8 control samples, 8 samples for DMAN and 8 for SACCH were obtained. For each developmental toxic compound we used 4 samples per concentration, except for the lowest and highest concentration. For these concentrations we used 5 samples. Sample collection was done in three independent series with all compounds and concentrations present in one series, except for RA which was tested only in the last series. All samples were collected in RNAlater (Ambion, Austin, TX) and kept at 4C for at least one day to stabilize RNA. Before RNA isolation, RNAlater was removed and embryos were homogenized using a pestle in an Eppendorf tube. Following homogenization, RNA was isolated using the RNeasy mini kit with an additional DNAse treatment (RNase free DNase set) (both from Qiagen, Venlo, the Netherlands) according to the manufacturers instructions. RNA isolation was performed in ten different batches due to the amount of samples. RNA concentration was measured on the NanoDrop spectrophotometer (ND-1000, NanoDrop Technologies Inc., Wilmington, DE). RNA integrity was assessed on the Bioanalyzer 2100 (Agilent Technologies, Amstelveen, The Netherlands) using the RNA 6000 Nano Chip kit (Agilent technologies) by automated gel electrophoresis. Samples with a RIN value >7 were used for further analyses. RNA was stored at -80 C until labelling and hybridization.