Total RNA extracted from a pool of 1 to 2 mm antral follicles, determined as atretic using microscopic examination after Feulgen staining of some granulosa cells, originating from sow 61030
annotation pf/mf/gf: PF date hybridation-genovul: 2008-03-25 status: A animal-number_genovul: 61030
Extracted molecule
total RNA
Extraction protocol
Once ovaries collected, all visible antral follicles larger than 1 mm in diameter were isolated carefully using a binocular microscope. After dissection, follicle diameter was measured and each follicle was classified according to size, animal and follicular quality. Follicles measuring 1 to 2 mm were kept and granulosa cells were recovered from all individual follicles as described previously, and they were stored at -80°C until RNA extraction. For each follicle, a sample of granulosa cells was smeared on a histological slide, and then Feulgen stained to determine follicular quality by microscopic examination. Both healthy follicles (presence of mitosis and absence of pycnosis in granulosa cells), noted SHF, and atretic follicles (absence of mitosis and presence of pycnosis in granulosa cells), noted SAF, were kept for this study. RNA was extracted from granulosa cells according to the technique described by Chomczynski and Sacchi with minor modifications using pools of granulosa cells from the same follicle size class, status and the same animal. The quality of each RNA sample was checked through the Bioanalyser Agilent 2100 (Agilent Technologies, Massy, France) and low-quality RNA preparations were discarded.
Label
P33
Label protocol
The arrays were hybridized with 33P-labelled complex probes synthesized from 5 µg of each RNA sample, using SuperScript II RNAse H-reverse transcriptase (Invitrogen, Cergy-Pontoise, France)
Hybridization protocol
A porcine microarray, published in GEO database as GPL3729, described by Bonnet et al. and produced by CRB-GADIE was used to hybridize the RNA samples. Micro-arrays were first hybridized with a 33P- labelled oligonucleotide sequence present in all PCR products to control the quality of the spotting process and to evaluate the amount of DNA in each spot. After stripping, the arrays were hybridized with 33P- labelled complex probes synthesized from 5 µg of each RNA sample, using SuperScript II RNAse H- reverse transcriptase (Invitrogen, Cergy-Pontoise, France). Each complex probe has been hybridized on membrane exposed 1, 3 or 7 days to radioisotopic-sensitive imaging plates (BAS-2025, Fujifilm, Raytest, Courbevoie, France). The imaging plates were scanned thereafter with a phosphor imaging system at 25µm resolution (BAS-5000, Fujifilm, Raytest, Courbevoie, France).
Scan protocol
Hybridization images obtained from oligonucleotide and complex probes were quantified using the semi-automated AGScan software.
Description
Reanalysis of GSM595692. A pig nylon micro-array is hybridized with a complex 33P labelled probe
Data processing
Data exploration and analysis were performed using R software. The data coming from complex probes hybridization were analysed on the logarithmic scale. As a first step, negative spot with >7 intensity values were checked on images and replaced by the median of negative spots in case of obvious overshining effect or hybridization stain. Then, luciferase positive controls were removed and the correlations of the data from each membrane were examined intra condition. Hybridizations with a lower than 0.80 correlation coefficient were discarded. Spots with low signal value (below the average of empty spots + 3 standard deviations) were considered as unexpressed and were excluded from the analysis. Finally, the negative control spots were removed and the remaining data were centred for each membrane. Thereafter, a linear model was fitted on these data, using the lm procedure on follicle type factor and differentially expressed genes were selected with R statistical software. We tested the significance of the follicle class on the gene expression using t-test (Student test) followed by the Benjamini-Hochberg procedure controlling False Discovery Rate (FDR) for each cDNA. A set of predictive genes was identified using the balanced Random Forests approach (RFs). To obtain a stable selection of genes, fifty of balanced RFs were launched with each of 15000 trees. For each cDNA, the average value of its rank in the different random forest and was calculated for two criteria of importance calculated in the Random Forest procedure: Mean decrease Gini and Mean decrease accuracy. The average rank values were plotted in a decreasing order and groups of genes with homogeneous values were made. The most important cDNAs were selected as the group of genes having the higher average rank. Finally, the relevance of the two selections was evaluated via unsupervised hierarchical clustering using the Ward method and Euclidian distance with the R functions hclust and heatmap.
