DNA copy number profiles of gastric cancer precursor lesions

Background Chromosomal instability (CIN) is the most prevalent type of genomic instability in gastric tumours, but its role in malignant transformation of the gastric mucosa is still obscure. In the present study, we set out to study whether two morphologically distinct categories of gastric cancer precursor lesions, i.e. intestinal-type and pyloric gland adenomas, would carry different patterns of DNA copy number changes, possibly reflecting distinct genetic pathways of gastric carcinogenesis in these two adenoma types. Results Using a 5K BAC array CGH platform, we showed that the most common aberrations shared by the 11 intestinal-type and 10 pyloric gland adenomas were gains of chromosomes 9 (29%), 11q (29%) and 20 (33%), and losses of chromosomes 13q (48%), 6(48%), 5(43%) and 10 (33%). The most frequent aberrations in intestinal-type gastric adenoma were gains on 11q, 9q and 8, and losses on chromosomes 5q, 6, 10 and 13, whereas in pyloric gland gastric adenomas these were gains on chromosome 20 and losses on 5q and 6. However, no significant differences were observed between the two adenoma types. Conclusion The results suggest that gains on chromosomes 8, 9q, 11q and 20, and losses on chromosomes 5q, 6, 10 and 13, likely represent early events in gastric carcinogenesis. The phenotypical entities, intestinal-type and pyloric gland adenomas, however, do not differ significantly (P = 0.8) at the level of DNA copy number changes.


Background
Gastric cancer is the second most frequent malignancy worldwide and the prognosis of this malignancy remains very poor [1]. Gastric cancer incidence and mortality rates differ between different countries within the European Union [2]. In the Netherlands it ranks fifth as a cause of cancer death, with approximately 2,200 new cases each year [3]. Surgery with curative intent is the treatment of choice in advanced cases of gastric cancer, whereas local endoscopic mucosectomy can be curative in early gastric cancer. Detection and removal of gastric neoplasias in an early or even premalignant state will contribute to reduce death due to gastric cancer. To achieve this goal, better tests for early detection of gastric cancer are needed, and an improved understanding of the biology of gastric cancer progression is crucial in this respect.
According to the Correa model, pathogenesis of intestinal-type gastric adenocarcinoma follows a pathway of chronic active gastritis due to Helicobacter pylori infection, leading to mucosal atrophy, intestinal metaplasia followed by intraepithelial neoplasia and finally invasive adenocarcinoma [4]. Genetic characterization of tissue samples in intraepithelial neoplasia stage would substantially contribute to our understanding of the molecular pathogenesis of gastric cancer. However, these lesions are only rarely detected, possibly due to rapid progression through this stage towards cancer, and are usually present only in parts of biopsy specimens, hampering genomic analysis of these lesions. Analysis of alternative precursor lesions could therefore, at least partly, be a substitute. Development of gastric cancer through an adenoma stage, although less common, is such alternative route. These adenomas are occasionally detected during gastroscopy and present as large lesions that histologically show intraepithelial neoplasia, which makes them suitable for genomic analysis. Gastric adenomas have a direct malignant potential and account for approximately 20% of all epithelial polyps [5,6]. Gastric adenomas can have a classic tubular, tubulovillous, or villous morphology with a predominantly intestinal-type epithelium, but can also appear as pyloric gland adenomas [6]. Pyloric gland adenomas arise from deep mucoid glands in the stomach and are strongly positive for mucin 6 [7,8]. A substantial number of gastric adenomas already show progression to adenocarcinoma. On first diagnosis around 30-40% of all pyloric gland adenomas already show a focus of carcinoma [9,10]. For intestinal-type adenomas this number is lower and varies from 28,5% for villous adenomas and 29.4% for tubulovillous type adenomas to only 5.4% in the tubular adenomas [11]. Both adenocarcinomas, ex intestinal-type adenomas and ex pyloric gland adenomas, show glandular structures, in contrast to diffuse type gastric cancer.
In the present study, we set out to study whether two morphologically distinct categories of gastric cancer precursor lesions, i.e. intestinal-type and pyloric gland adenomas, would carry different patterns of DNA copy number changes, possibly reflecting distinct genetic pathways of gastric carcinogenesis in the two adenoma types.
In the intestinal-type gastric adenomas, the most common aberrations observed were gains on chromosomes 8, 9q and 11q, and losses on chromosomes 5q, 6, 10 and 13. In four adenomas (36.4%), gain of chromosome 11q23.3 was observed with a common region of overlap of 2.6 Mb. Gain of chromosome 9q was observed in four adenomas (36.4%) with a 12.6 Mb common region of overlap located on chromosome 9q33.1-q34. 13. Gain of chromosome 8 was observed in three adenomas (31%), two of which adenomas showed gain of whole chromosome 8, and the third adenoma showed a gain of chromosome 8p-q22.3 with an additional 28.7 Mb gain on chromosome 8q24.11-qter. In addition, gains were observed on chromosomes 1, 3, 6p, 7, 11p, 12p, 13q, 16, 17, 19, 20 and 22q. No amplifications were seen in the intestinal-type adenomas.
Deletions on chromosome 13 were observed in seven intestinal-type adenomas (64%). Of these, five showed a 11.9 Mb deletion of chromosome 13q21.2-21.33 with an additional 7.7 Mb deletion on chromosome 13q31. 1-31.3. The other two adenomas showed a 16.6 Mb deletion of 13q14.3-31. A deletion on chromosome 6 was observed in six adenomas (55%), with an overlapping region of 68.9 Mb located on 6cen-q22.1. A deletion of chromosome 5q was observed in four adenomas (36%) with a common region of overlap located on chromosome 5q22.1-q23.2. In addition, a deletion of whole chromosome 10 was observed in four adenomas (36%). Other losses observed in intestinal-type adenomas were located on chromosomes 8q, 9p, 10, 12q, 20q and 21. An overview of all DNA copy number aberrations of the intestinal-type adenomas is shown in Table 1.
The most common aberrations shared by both intestinaltype and pyloric gland adenomas were gain of chromosome 9q (29%), 11q (29%), and 20q (33%) and loss of chromosome 5 (43%), 6 (48%), 10 (33%) and 13q (48%). By comparing intestinal-type and pyloric gland adenomas, CGH Multiarray revealed eight clones to be significantly different, six of which were located at chromosome 6q14-q21 (p = 0.02 to 0.05) and two clones on chromosome 9p22-p23 (p = 0.02 and 0.04, respectively) ( Figure 1). No genes located in the regions covered by these clones have been known to be involved in cancer related biological processes. Yet, CGH Multiarray Region, after correction for multiplicity, yielded a false discovery rate (FDR) of 1 for all these regions, indicating no significant differences between the two different types of adenomas at the chromosomal level. Unsupervised hierarchical cluster analysis yielded 2 clusters. No significant associations were found here (p = 0.8).

Discussion
Given the heterogeneous phenotype of gastric cancer, the present study primarily aimed to compare copy number changes between intestinal-type adenomas and pyloric gland adenomas, in order to find leads towards genetic pathways involved in the pathogenesis of gastric cancer. Adenoma-to-carcinoma progression is observed in 30-40% of the pyloric gland adenomas and in approximately 5-30% of the intestinal-type adenomas (varying from about 5% in tubular adenomas to almost 30% for tubulovillous and villous adenomas) [9][10][11], indicating the direct malignant potential of these two adenoma types and making gastric adenomas a suitable model for detecting early events in gastric carcinogenesis.
Pyloric gland adenomas constitute a recently recognized entity [8,26]. To the best of our knowledge, this type of adenomas has never been analyzed by array CGH before. The mean number of events in this type of adenoma was 5.4 (0-9), with 2.4 (0-7) gains and 3 (0-7) losses. This is comparable with the mean number of aberrations in intestinal-type adenomas (6.5 (0-18), 3.4 (0-14) and 3.1 (0-7) respectively). In pyloric gland adenomas, frequent events were gain on chromosome 20 and losses on chromosomes 5q and 6, while intestinal-type adenomas mainly showed gain on chromosomes 8, 9q, and 11q, and losses on chromosomes 5q, 6, 10 and 13. In the present study, gain of chromosome 7 was less common than found previously [16]. Although these frequently altered regions differ between the two types of adenomas, hierarchical cluster analyses did not separate the groups. In addition, CGH Multiarray Region did not reveal any significant differences after correction for multiple comparisons. This lack of statistically significant differences could be due to the limited sample size combined with the fact that in general, adenomas show little chromosomal aberrations. On the other hand, it could simply be that these morphologically different entities do not differ in terms of chromosomal gains and losses. Finding no significant differences at the chromosomal level does not preclude other genetic and biological differences such as mutation or promoter methylation status of specific genes. Aberrations already detected in adenomas may be early events in the stepwise process of accumulating changes which may cause progression of adenoma to carcinoma. As expected, the mean number of chromosomal events was lower in adenomas compared to the carcinomas [13,14,27]. Moreover, high level amplifications are uncommon in adenomas, while carcinomas frequently show high level amplifications [13,16].
The aberrations found in both intestinal-type and pyloric gland adenomas, such as losses on chromosome 5q, are also frequently detected in gastric carcinomas [15,19,28]. Previous CGH results showed a significantly higher number of chromosome 5q losses in intestinal-type carcinoma compared to diffuse type carcinoma [29]. Chromosome 6, also lost in both types of adenomas, frequently is deleted in gastric carcinomas as determined by LOH studies [30,31]. Moreover, chromosome 6q deletion has been reported to be involved in an early stage of gastric carcinogenesis, since chromosome 6q deletions are frequently detected in early gastric cancer and also in intestinal metaplasia [31,32]. Losses of chromosomes 10 and 13 have been previously observed in adenomas at lower frequencies. In gastric carcinomas, both gains and losses of chromosome 10 and 13 have been observed by previous CGH studies [15,19,21,33]. Chromosome 10 harbors the oncogene FGFR2 (10q26) and tumour suppressor genes PTEN/ MMAC1 (10q23) and DMBT1 (10q25-q26), both involved in carcinogenesis, which could explain the observation of both gains and losses of chromosomes 10 in gastric carcinomas [34][35][36]. Indeed chromosome 13 harbors tumour suppressor genes such as BRCA2 (13q12.3) and retinoblastoma gene (RB1) (13q14). In contrast, gain of chromosome 13q has been correlated to colorectal adenoma-to-carcinoma progression, and amplification of chromosome 13 has been observed in gastric adenomas with severe intraepithelial neoplasia [14,37]. The precise role of chromosome 13 aberration in gastric cancer therefore remains to be resolved.
Most frequent copy number gains were observed on chromosomes 8, 9q, 11q and 20. Especially gains of chromosomes 8 and 20 are consistent with previous (array) CGH studies in both gastric adenomas and gastric carcinomas [13][14][15][16]19,25], implicating this as early events in tumourigenesis. Although gain of chromosome 11q has not been described as a frequent event in adenomas, in carcinomas gain or amplification on chromosome 11q is common [13][14][15][16]. In the present study gain of chromosome 11q was frequently observed in the adenomas, implying the malignant potential of these adenomas.

Conclusion
These data indicate that gains on chromosomes 8, 9q, 11q and 20 and losses on chromosomes 5q, 6, 10 and 13 are early events in gastric carcinogenesis. Despite the pheno-

Material
Twenty-one paraffin-embedded gastric adenomas, 11 intestinal-type and 10 pyloric gland adenomas, were included in this study (Figure 2A and 2B). Tumour and patient data are given in Table 3. For each case, a tumour area consisting for at least 70% of tumour cells was demarcated on a 4 µm hematoxylin and eosin stained tissue section. Adjacent 10-15 serial tissue sections of 10 µm were stained with hematoxylin and the corresponding tumour area was microdissected using a surgical blade. A final 4µm "sandwich" section was made and stained with hemotoxylin and eosin, to compare with the first slide as a control. After deparaffinization, DNA was extracted by a column-based method (QIAamp DNA mini kit; Qiagen, Westburg, Leusden, NL) [38].
Genomic DNA obtained from peripheral blood from ten normal individuals was pooled (either ten females or ten males, depending on the gender of the patient from which the adenoma was obtained) and used as control reference DNA.

Array CGH
Array CGH was performed essentially as described previously [39]. Briefly, 300 ng tumour and reference DNAs, sex-mismatch as experimental control, were labelled by random priming (Bioprime DNA Labelling System, Invitrogen, Breda, NL), each in a volume of 50µL. Non incorporated nucleotides were removed using ProbeQuant G-50 microcolumns (Amersham Biosciences  Comparison of DNA copy number alterations in intestinal and pyloric gland type gastric adenomas Figure 1 Comparison of DNA copy number alterations in intestinal and pyloric gland type gastric adenomas. A p-value (Y-axis) was calculated for every clone, based on a Wilcoxon test with ties, and plotted in chromosomal order from chromosome 1 to 22 (X-axis). Eight clones reached the level of significance (p < 0.05), but failed to maintain a significantly low false discovery rate after correction for multiple comparison.

Image acquisition and data analysis
Images of the arrays were acquired by scanning (Agilent DNA Microarray scanner-Agilent technologies, Palo Alto, USA) and quantification of the signal and background intensities for each spot for the two channels Cy3 and Cy5 was performed by Imagene 5.6 software (Biodiscovery Ltd, Marina del Rey, CA, USA). Local background was subtracted from the signal median intensities and tumours to reference ratios were calculated. The ratios were normalized against the mode of the ratios of all autosomes.
Clones with poor quality of one of the triplicates and hybridization with a standard deviation (SD) ≤ 0.22 and clones with > 50% missing values in all adenomas were excluded, leaving 4648 clones for further analysis. All subsequent analyses were done considering the clone position from the UCSC May2004 freeze of the Human Golden Path.
Array CGH smooth [42,43], was used for automated detection of breakpoints to determine copy number gains and losses. Since variation in quality is observed in DNA obtained from formalin-fixed paraffin-embedded gastric tissues, different smoothing parameters were applied, depending on the quality of the hybridization. For array CGH profiles with a standard deviation smaller or equal to 0.15, between 0.15 and 0.20 or between 0.20 and 0.22, the applied smoothing parameters to determine gains and losses were 0.10, 0.15 and 0.20 respectively. Log 2 tumour to reference ratio above 1 was regarded as amplification.

Statistical analysis
Unsupervised hierarchical cluster analysis was performed to analyze the distributions of the genomic profiles of all adenomas using TMEV software 3.0. 3 [44]. Based on normalized smoothed log 2 tumour to normal fluorescence intensity ratios, a hierarchical tree was constructed using the parameters complete linkage and euclidean distance. Pearson Chi-square test was used for analyzing correlations between cluster membership and adenoma type (SPSS 11.5.0 for windows, SPSS Inc, Chicago, IL, USA). Pvalues less than 0.05 were considered to be significant.
Supervised analysis was used for identifying chromosomal regions specific for the two adenoma types using CGH Multiarray and CGH Multiarray Region [45,46]. Based on normalized smoothed log 2 tumour to normal fluorescence intensity ratios, p-values were calculated for the significance of difference of values for each clone between pyloric gland and intestinal-type adenomas, using a Wilcoxon test with ties. To correct for multiple testing, a permutation-based false discovery rate (FDR) was calculated [47].

Competing interests
The author(s) declares that there are no competing interests.
Haematoxilin and eosin staining (original magnification ×400) of intestinal-type (A) and pyloric gland (B) gastric adenomas Figure 2 Haematoxilin and eosin staining (original magnification ×400) of intestinal-type (A) and pyloric gland (B) gastric adenomas. A. Intestinal-type adenoma of the stomach composed of irregularly arranged glands composed of intestinal-type epithelium with eosinophilic cytoplasm and enlarged nuclei. B. Pyloric gland adenoma of the stomach composed of densely back to back packed glands consisting of cells with pale cytoplasm and small round hyperchromatic nuclei.