Date Published: May 3, 2010
Publisher: Public Library of Science
Author(s): Octavio Martínez, M. Humberto Reyes-Valdés, Luis Herrera-Estrella, Shin-Han Shiu. http://doi.org/10.1371/journal.pone.0010398
Abstract: A central goal of cancer biology is to understand how cells from this family of genetic diseases undergo specific morphological and physiological changes and regress to a de-regulated state of the cell cycle. The fact that tumors are unable to perform most of the specific functions of the original tissue led us to hypothesize that the degree of specialization of the transcriptome of cancerous tissues must be less than their normal counterparts. With the aid of information theory tools, we analyzed four datasets derived from transcriptomes of normal and tumor tissues to quantitatively test the hypothesis that cancer reduces transcriptome specialization. Here, we show that the transcriptional specialization of a tumor is significantly less than the corresponding normal tissue and comparable with the specialization of dedifferentiated embryonic stem cells. Furthermore, we demonstrate that the drop in specialization in cancerous tissues is largely due to a decrease in expression of genes that are highly specific to the normal organ. This approach gives us a better understanding of carcinogenesis and offers new tools for the identification of genes that are highly influential in cancer progression.
Partial Text: Cancer is a complex family of acquired genetic diseases in which a single cell clone and its progeny accumulate heritable changes that cause a malignant phenotype of deregulated cell growth and differentiation . Numerous studies have been performed to better understand the alterations that occur in the transcription profile during the progression of cancer . These experiments have been carried out by directly counting the tags of expressed genes using serial analysis of gene expression (SAGE) , expressed sequence tags (ESTs) , and other counting strategies, or by indirectly measuring the levels of transcription using DNA microarrays . In many cases, these experiments have detected genes that are preferentially expressed in a cancer tumor and can serve as molecular markers of malignancy. In addition, they can also detect significant alterations in the transcription level of sets of genes that participate in complex signaling networks. Changes in these networks represent distortions of the pathways that regulate the physiology of normal cells .
We hypothesized that the morphological and functional changes that occur during cancer progression would lead to substantial changes in the cancer transcriptome, including a reduction in specialization, when compared to that of analogous normal tissues. To test this hypothesis in a broad framework, we selected three collections of gene tags and one microarray experiment. Datasets A and B are selected collections of cDNA libraries from the “Cancer Genome Anatomy Project”  for human and mouse tissues, respectively. Dataset C consists of SAGE libraries from normal human and tumor tissues obtained from the “Human Transcriptome Map” project  and dataset D is a microarray study of human tissues in normal and pre-cancerous states that were paired by patient . Datasets A and B incorporate five embryonic stem cell (ESC) and one hematopoietic stem cell (HSC) libraries and were included in the analysis based on their degree of dedifferentiation. These datasets were subjected to the analysis of information properties of the transcriptome as previously described . In the counting tags datasets, we assessed the statistical significance of the differences in specialization. In each case, we obtained the specificity (Si) and Target Specificity (TSij) for the genes studied in the datasets, which allowed for the selection of putative overexpressed genes in cancer or normal tissues as well as the discrimination of genes preferentially expressed in a given condition.
The use of information theory tools to quantitatively assess changes in steady state transcript abundances allowed us to examine four different datasets to determine whether cancerous tissues have less transcriptome specialization than their normal counterparts. The results obtained from these analyses showed that specialization of the cancer transcriptome decreased when compared to the normal tissue equivalent. The decrease in transcriptome specialization was due mainly to a reduction in the expression level of genes that are tissue-specific and usually expressed at high levels in normal tissue (see Supporting Text S1 and Table S11, Table S12, Table S13, Table S14, Table S15 and Table S16). These results are in agreement with the observation that tumors often show morphologically dedifferentiated cell types in a manner similar to that observed in stem cells . In addition, molecular evidence has shown that poorly differentiated cancer tumors overexpress genes that are enriched in embryonic stem cells . It is not completely clear whether cancer initiates by a process of de-regulation of organ stem cells or by a de-novo dedifferentiation of organ cells driven by the mutations that arise during the development of the tumor .