Identification of expressed genes linked to malignancy of human colorectal carcinoma by parametric clustering of quantitative expression data

Abstract

Background: Individual human carcinomas have distinct biological and clinical properties: gene-expression profiling is expected to unveil the underlying molecular features. Particular interest has been focused on potential diagnostic and therapeutic applications. Solid tumors, such as colorectal carcinoma, present additional obstacles for experimental and data analysis.

Results: We analyzed the expression levels of 1,536 genes in 100 colorectal cancer and 11 normal tissues using adaptor-tagged competitive PCR, a high-throughput reverse transcription-PCR technique. A parametric clustering method using the Gaussian mixture model and the Bayes inference revealed three groups of expressed genes. Two contained large numbers of genes. One of these groups correlated well with both the differences between tumor and normal tissues and the presence or absence of distant metastasis, whereas the other correlated only with the tumor/normal difference. The third group comprised a small number of genes. Approximately half showed an identical expression pattern, and cancer tissues were classified into two groups by their expression levels. The high-expression group had strong correlation with distant metastasis, and a poorer survival rate than the low-expression group, indicating possible clinical applications of these genes. In addition to c-yes, a homolog of a viral oncogene, prognostic indicators included genes specific to glial cells, which gives a new link between malignancy and ectopic gene expression.

Conclusions: The malignancy of human colorectal carcinoma is correlated with a unique expression pattern of a specific group of genes, allowing the classification of tumor tissues into two clinically distinct groups.

Figure 1

An example of an ATAC-PCR electropherogram. The height of each peak represents fluorescence intensity. Blue, PCR product signals; red, size marker signals. Starting from the left, the marker sizes are 35, 50, 75 and 100 bases. The seven blue peaks on the right correspond to ATAC-PCR products. From the left, the first peak corresponds to 10 equivalents of control cDNA, the second corresponds to two equivalents of control cDNA, and the last five peaks correspond to five equivalents of sample cDNA.

Figure 2

Distributions of the first to tenth factor scores obtained by principal component analysis. Vertical axis, number of genes; horizontal axis, factor scores. The interval of each column is 0.3, and the range of the central column is between -0.15 and 0.15. Red curves are normal distributions fitted to the columns. Percent variance explained by each component is as follows, from the first to the tenth: 12.33, 4.96, 4.76, 4.22, 3.28, 3.08, 2.57, 2.51, 2.27, 2.17.

Figure 3

Parametric clustering of genes using the Gaussian mixture model. The dots represent a two-dimensional matrix of genes generated by principal component analysis of the gene-expression data. Horizontal axis, factor score of the first component extracted from the expression patterns of each gene; vertical axis, factor score of the second component. Green, group GM-A; blue, group GM-B; red, group GM-C.

Figure 4

Hierarchical cluster analysis of the genes on the basis of expression patterns. A total of 1,536 genes are aligned horizontally. One hundred cancer and 11 normal tissue samples are vertically aligned in the same order as in Figure 6a. The bottom dots indicate genes grouped by parametric clustering. Green, group GM-A; blue, group GM-B; red, group GM-C. The black bar corresponds to cluster 43.

Figure 5

Correlation of gene expression with cancer phenotype. The vertical axis represents the correlation ratio (CR) of the differences between tumor and normal tissues in (a) group GM-A and (b) GM-B; or the presence or absence of distant metastasis in (c) GM-A and (d) GM-B; or lymph node metastasis in (e) GM-A and (f) GM-B. The horizontal axis represents the genes sorted by CR. Red, original data; blue, trials of permuted data.

Figure 6

Expression of TCL (tumor-classifier) genes was correlated with the malignant potential of colorectal cancer. Results used the 100 cancer and 11 normal tissues. (a) Correlation with distant metastasis. We aligned the 12 genes listed at the top of Table 1 from left to right horizontally. All samples were then sorted vertically by the average of the gene-expression levels of the 12 genes. The border of groups 1 and 2 is set at 0. The right bar indicates remote metastasis status, with red, green and blue representing remote metastasis positive, remote metastasis negative, and normal tissues, respectively. (b) Kaplan-Meier plot analysis of groups 1 and 2. Vertical axis, fraction of survival; horizontal axis, survival time in years. The groups of patients analyzed consisted of either all stages (left) or Dukes' B and C stages only (right). All the expression data and relevant clinical information are available as additional data files.