In the article on how L1-CAM and ADAM10 induce metastasis in colon cancer in the August 15, 2007 issue of Cancer Research (1), there was an error in the computer code that assigns P-values to the difference between the experimental data and control. The authors obtained a new list of differentiating genes that overlaps by more than 70% with the previous list; all the genes that were validated and discussed in the article also appear on the new list. All the results and conclusions of the article remain valid. The new list of differentiating genes (Supplementary Table 1) is available as supplementary data at Cancer Research Online,1
and the new Methods section and the new version of Fig. 6 appear below.
L1-CAM expression in Ls174T cells increases the levels of a cluster of genes also induced in human colon carcinoma tissue. U133A Affymetrix GeneChips were used to analyze the expression of genes elevated in two independently derived Ls174T cell lines (L1/1, L1/2) and compared with control neor Ls174T cells. In the list of genes that were significantly different from the control (see Methods), 150 unique genes (represented by 191 probesets on the microarray) were identified (Supplementary Table 1). The expression levels of these probe sets in a dataset of U133A microarrays, consisting of 170 human colon carcinoma and 43 normal colon epithelial tissue samples, are shown in C in pseudo color; red is high and blue is low expression level. Each row corresponds to a probe set and each column to a sample; the tissue type is identified by the color bar below the expression matrix, with blue and peach marking carcinoma and normal tissue, respectively. Rows and columns were ordered by the SPIN algorithm. The resulting distance matrix of the probe sets, shown in A, reveals clusters with different expression profiles. A group of 60 genes (blue square marked by a box), listed in B, that were overexpressed in L1-CAM transfected cells, was found to also be elevated in a significant number of human carcinoma patients, compared with normal colon epithelium. D, semiquantitative RT-PCR for some of the genes elevated in this cluster (individually boxed in B) was carried out using samples of the two Ls174T cell lines expressing L1-CAM and compared with that in neor control Ls174T cells.
L1-CAM expression in Ls174T cells increases the levels of a cluster of genes also induced in human colon carcinoma tissue. U133A Affymetrix GeneChips were used to analyze the expression of genes elevated in two independently derived Ls174T cell lines (L1/1, L1/2) and compared with control neor Ls174T cells. In the list of genes that were significantly different from the control (see Methods), 150 unique genes (represented by 191 probesets on the microarray) were identified (Supplementary Table 1). The expression levels of these probe sets in a dataset of U133A microarrays, consisting of 170 human colon carcinoma and 43 normal colon epithelial tissue samples, are shown in C in pseudo color; red is high and blue is low expression level. Each row corresponds to a probe set and each column to a sample; the tissue type is identified by the color bar below the expression matrix, with blue and peach marking carcinoma and normal tissue, respectively. Rows and columns were ordered by the SPIN algorithm. The resulting distance matrix of the probe sets, shown in A, reveals clusters with different expression profiles. A group of 60 genes (blue square marked by a box), listed in B, that were overexpressed in L1-CAM transfected cells, was found to also be elevated in a significant number of human carcinoma patients, compared with normal colon epithelium. D, semiquantitative RT-PCR for some of the genes elevated in this cluster (individually boxed in B) was carried out using samples of the two Ls174T cell lines expressing L1-CAM and compared with that in neor control Ls174T cells.
Methods
The chips were processed and scaled with Affymetrix MAS5, using the following preprocessing steps: (a) genes that were marked by Affymetrix as ‘present’ in all samples were kept, leaving a group of 11331 probe sets; (b) the two chips of Ls174T cell clones expressing L1-CAM (A1,A2) were averaged, producing a single vector of expression levels, called Ā; (c) lowess correction was implemented on the control (C) and Ā; (d) threshold to 1, i. e., any expression value below 1 was changed to 1; (e) log2 transformation. The approach for analyzing the differences between the average of two experimental chips (for each Ls174T cell clone expressing L1-CAM) and the neor control was a generalization of the standard fold-change approach: because the differences between the log expression values of two microarrays diminish with increasing intensities, a single fold change cutoff cannot be used to determine which probe sets are significantly different from the control. We used a method that calculates P-values for each gene, based on the distribution of the differences in expression of genes that have a similar average expression. The null hypothesis is that there is no difference between Āi (gene i in Ā), Ci (gene i in control C). There are two main assumptions in this method: 1), Ā,C have normal distribution with the same mean and variance: Ā,C∼N(μ,σ2); 2), The noise of genes that have a similar average expression level is similar. Let
The significantly different probe sets were analyzed using a second data set of U133A Affymetrix GeneChips from colon cancer patients. This analysis was done on 170 human colon carcinomas and 43 normal colon epithelial tissues, obtained in the framework of a collaborative National Cancer Institute/NIH program project grant. For visualization purposes, the data were centered and normalized. The expression matrix of the over expressed probe sets was two-way sorted by sorting points into neighborhoods (SPIN; once for the genes and once for the samples; ref. 2). Functional annotation was determined using the NIH program DAVID (3, 4).
RSS Feeds