Defining a Common Serum Response Program in Fibroblasts

For defining a common serum response program in fibroblasts, global gene expression patterns in 50 fibroblast cultures derived from ten anatomic sites, cultured in the presence of 10% or 0.1% FBS, were characterized by DNA microarray hybridization (Chang et al. 2002 (link)). We selected for further analysis genes for which the corresponding array elements had fluorescent hybridization signals at least 1.5-fold greater than the local background fluorescence in the reference channel, and we further restricted our analyses to genes for which technically adequate data were obtained in at least 80% of experiments. These filtered genes were then analyzed by the multiclass Significance Analysis of Microarrays (SAM) algorithm (Tusher et al. 2001 (link)) to select a set of genes whose expression levels had a significant correlation with the presence of serum in the medium, with a false discovery rate (FDR) of less than 0.02%. The corresponding expression patterns were organized by hierarchical clustering (Eisen et al. 1998 (link)). Genes that were coordinately induced or repressed in response to serum in most samples (Pearson correlation, greater than 90%) were identified. This set of 677 genes, represented by 772 cDNA probes, of which 611 are uniquely identified by UniGene (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=unigene), was termed the common fibroblast serum response gene set. To identify the subset of these 677 genes whose variation in expression was directly related to cell cycle progression, we compared this set of genes to a published set of genes periodically expressed during the HeLa cell cycle (Whitfield et al. 2002 (link)). Because both datasets were generated using similar cDNA microarrays, we tracked genes by the IMAGE number of the cDNA clones on the microarrays. The majority of the genes in the fibroblast serum response gene set showed no evidence of periodic expression during the HeLa cell cycle. One hundred sixty-five genes, represented by 199 cDNA clones, overlapped with the cell cycle gene list; the remaining 512 genes, represented by 573 clones, of which 459 are uniquely identified in UniGene, was termed the CSR gene set.
The patterns of expression in human tumors of the 512 genes of the fibroblast CSR gene set were analyzed using data from published tumor expression profiles. Detailed methods and primary datasets are available as Datasets S1 and S2 and on our Web site (http://microarray-pubs.stanford.edu/wound). We used the Unigene unique identifier (build 158, release date January18, 2003) to match genes represented in different microarray platforms. For cDNA microarrays, genes with fluorescent hybridization signals at least 1.5-fold greater than the local background fluorescent signal in the reference channel (Cy3) were considered adequately measured and were selected for further analyses. For Affymetrix data, signal intensity values were first transformed into ratios, using for each gene the mean values of the normalized fluorescence signals across all the samples analyzed as the denominators (Bhattacharjee et al. 2001 (link)). The genes for which technically adequate measurements were obtained from at least 80% of the samples in a given dataset were centered by mean value within each dataset, and average linkage clustering was carried out using the Cluster software (Eisen et al. 1998 (link)). In each set of patient samples, the samples were segregated into two classes based on the first bifurcation in the hierarchical clustering dendrogram. For the datasets shown, the clustering and reciprocal expression of serum-induced and serum-repressed genes in the tumor expression data allowed two classes to be unambiguously assigned. Samples with generally high levels of expression of the serum-induced genes and low levels of expression of the serum-repressed genes were classified as “activated”; conversely, samples with generally high levels of expression of serum-repressed genes and low levels of expression of the serum-induced genes were classified as “quiescent.” Survival analysis by a Cox–Mantel test was performed in the program Winstat (R. Fitch Software).