Cytogenetic heterogeneity and progression of esophageal squamous cell carcinoma

Abstract

It is widely believed that most human tumors, including esophageal squamous cell carcinoma (ESCC), arise through multistep genetic and cytogenetic alterations. The time sequence of these alterations, however, is still unknown. The present study was designed to differentiate common early changes from uncommon later ones with combined comparative genomic hybridization (CGH) and ploidy analyses in multiple or single samples of 12 ESCCs. We first demonstrated that the mean copy numbers of chromosomes 3 and 11, determined directly by fluorescence in situ hybridization, showed linear correlation with the mean copy numbers calculated from the G/R ratio of CGH and DNA ploidy (R² = 0.714, P<0.0001). On this basis, we estimated the absolute copy numbers of chromosomal parts by applying the ploidy-dependent threshold criteria to the G/R ratio data after the criteria were corrected by the percentage of tumor cells in each sample. One-copy changes in the DNA-diploid stage may give large shifts of the G/R ratio, even after tetraploidization, whereas those after tetraploidization undergo small shift. Using the tumors with multiple samples, it was actually demonstrated that most of the gains common to the samples in individual tumors showed the large shifts. Though early changes varied from tumor to tumor in the nine informative cases, it was found that gains of 3q (5/7: number of cases with large-shift 3q+/total number of cases with 3q+), 8q (3/4), 11q13 (4/5), and 14q (3/4) were early events, while losses of 3p (2/8), 5q (1/5), 13q (1/5), and 21q (1/5), and gains of 1p (1/4) and Xq (1/4) were later events in progression of individual tumors.

Introduction

Esophageal cancer ranks among the nine most frequent cancers in the world [1]. The majority of esophageal carcinomas are classified as squamous cell carcinoma (SCC) and adenocarcinoma. SCC has more often been identified in populations of African descent, while adenocarcinoma is more prevalent in the Caucasian population [2]. Esophageal SCC (ESCC) has one of the poorest prognoses among the malignancies of the gastrointestinal tract. Despite recent reports documenting alterations of some oncogenes and tumor suppressor genes in ESCC [3], the molecular and cytogenetic bases of esophageal tumorigenesis remain largely unknown.

Thus far, SCC of head and neck, lung, and esophagus were thought to develop in a mode of field carcinogenesis [4], [5], while it is widely believed that most cancers arise from a single cell [6]. Recent molecular cytogenetic analyses using microsatellite markers, comparative genomic hybridization (CGH), and other methods are based on this clonal nature of the tumor. Using the human androgen receptor gene assay, our group has showed recently that a large part of ESCC is of monoclonal constitution [7].

In ESCC, only a few cytogenetic studies have been reported [8]. Some molecular cytogenetic alterations associated with ESCC have been identified recently, including loss of heterozygosity (LOH) at chromosomal loci on 3p, 5q, 9p, 9q, 13q, 17p, 17q, and 18q [9], [10] and amplifications of c-MYC (8q24), FGFR (17q21∼q22), and CCND1 (11q13) [9]. CGH has revealed several chromosomal regions that contain amplified cellular oncogenes as well as loss or gain of chromosomal regions in esophageal adenocarcinoma and SCC [11], [12], [13], [14], [15], [16], [17], [18]. The time sequence of these genetic events is still unknown, however, though it is now widely believed that most human tumors, including ESCC, result from multistep cytogenetic alterations. It may be partly because most of the available molecular genetic data, except for a few cases [19], were obtained by a series of single-sample analyses of a large number of tumors in various pathological stages with different grades of malignancy. Multiple sampling may enable us to discriminate common early changes from later regional changes [19]; so far, the CGH data of each tumor have given us no information about the relative importance of each gain/loss of chromosomal material in the tumor.

What CGH clarifies as the shift of the green-to-red (G/R) ratio is relative gain/loss from the mean copy number of chromosomes [20], [21]. Though it has been believed that the shift distance cannot be assessed in CGH [21], it was demonstrated by combined CGH, ploidy, and fluorescent in situ hybridization (FISH) analyses that the shift distance was linearly correlated with the actual copy number within the same ploidy class [22], [23]. It remains open, however, whether such linearity is also demonstrated in primary tumors in vivo, in which contamination of the normal component and heterogeneity of tumor cells have to be taken into consideration. In this study, we first tested whether there is a linear relationship between the centromere numbers directly determined by FISH and those estimated by the mean DNA ploidy and the G/R ratio value, which was corrected by the percentage of tumor cells in the sample.

On this basis, this study was designed to determine the absolute copy number of the altered chromosomal part by the DNA ploidy and the shift distance of the G/R ratio in primary ESCC. With this absolute copy number we would discriminate duplicated and single-copy changes, which are inferred to occur earlier and later, respectively, than tetraploidization. To test the feasibility of this approach, we carried out multiple sampling from individual tumors and tested whether the chromosomal alterations commonly shared by all the samples in a tumor coincide with those inferred (from the shift distance of the G/R ratio) to have occurred before tetraploidization.