Different Vulnerability among Chromosomes to Numerical Instability in Gastric Carcinogenesis: Stage-dependent Analysis by FISH with the Use of Microwave Irradiation¹

Only a few reports have examined chromosome abnormalities in gastrointestinal neoplasia by FISH³(1, 2). The process of tumorigenesis is assumed to be driven by the progressive accumulation of somatic mutations in a number of genes. Recent studies also suggest that numerical chromosomal aberrations may contribute to tumorigenesis (3, 4), but the chromosomal instability of gastric cancer remains poorly characterized. Gastrointestinal tract tumors are notorious for being difficult to analyze by standard cytogenetic techniques (5, 6, 7, 8). In this report, we extensively studied the numerical chromosomal abnormalities in gastric cancer, using FISH with centromere-specific α-satellite probes for 17 chromosomes. Because frequent necrosis, inflammatory and normal cell contamination of stroma, and poor attachment of tumor cells affected the quality and reliability of the observations even with a recently developed FISH protocol (7), we used MW irradiation (9, 10, 11)for centromere identification by in situ hybridization in gastric cancer cells. This approach may circumvent the problems mentioned above and facilitate an examination of pure cancerous epithelial cells, and thus the accurate detection of chromosomal centromeric instability. We were able to identify the earliest event in the multistep process of chromosomal numerical abnormality in gastric cancer and to propose the sequential alterations in each chromosome according to the progression.

All specimens were obtained from surgically resected materials at Hamamatsu University Hospital and its affiliated hospitals. The samples were prepared from 24 gastric cancers, 20 from men and 4 from women with a median age of 66.2 years (range, 37–87). Touch preparations of the fresh gastric cancerous tissue were used for FISH analysis, and the cancerous tissue quantity and the difference between cytologically normal and cancerous cells were monitored by review of Papanicolaou and Giemsa staining of the serial touch preparation as well as H&E staining of the adjacent tissue section of the touch surface. Sampling of the tumor in early stage cancer was performed from its superficial portion,and those in the advanced stage cancer, from the middle parts of the cancer. We did not analyze the different portions from the same cancer tissue in this study.

All surgical materials were routinely processed in paraffin-embedded,H&E-stained sections. Histological classification was based on the General Rules for Gastric Cancer Study in Japan (12). In terms of histological subcategorization, tubular-type carcinoma was classified as well (tub 1) and moderately (tub 2) differentiated types according to the degree of glandular formation. Poorly differentiated carcinoma was classified as solid type (por 1) and nonsolid type (por 2), which were further divided into tubular type with scirrhous infiltration (tub-sci) mucocellular type with scirrhous infiltration(muc-sci), and signet ring cell carcinoma (sig). The invasiveness of the tumor in the submucosal layer in gastric wall was graded into sm1(slight invasion to submucosal connective tissue), sm2 (invasion to middle layer of the submucosa), and sm3 (invasion to the deep layer of the submucosa). Details of this description are described elsewhere (12).

Fresh touch preparations of the gastric cancer tissue and normal lymphocytes were placed onto the slide glass and subjected to FISH by a modification of the method of Pinkel et al.(13) The samples were denatured for 2 min at 75°C in 70% formamide (Boehringer Mannheim, Mannheim, Germany), 2 × SSC(pH 7.0) and dehydrated. A panel of 17 centromeric α satellite DNA probes (D1Z5, D2Z, D3Z1, D4Z1, D6Z1, D7Z1, D8Z2, D10Z1, D11Z1, D12Z3,D15Z, D16Z2, D17Z1, D18Z1, D20Z1, DXZ1, and DYZ3) derived from chromosomes 1, 2, 3, 4, 6, 7, 8, 10, 11, 12, 15, 16, 17, 18, 20, X, and Y, respectively, was purchased from Oncor, Inc. (Gaithersburg, MD), and all of the probes were labeled with digoxigenin-11-dUTP by nick translation (14). Chromosomal and probe DNAs were denatured simultaneously for 5 min at 75°C in the hybridization mixture on the slide glass. Hybridization was performed at 37°C overnight in a humidified chamber. Digoxigenin-labeled probes were detected with an FITC digoxigenin system, and the nuclei were counterstained with propidium iodide (1 μl/ml; Sigma, St. Louis, MO). One hundred or more interphase nuclei of cancer cells were evaluated for each probe with an Olympus epifluorescence microscope equipped with a WIB filter (Olympus, Tokyo, Japan).

Because nonspecific signal or background noise presumably due to crush and degeneration in the specimen can influence the FISH evaluation, quality control of the procedure was performed. We applied intermittent MW treatment to the slides in addition to the conventional FISH protocol. After treatment with ethanol, a mixed solution of a probe was dripped onto the slide, a coverslip was mounted, and then the slide was treated in a temperature-controllable MW processor (MI-77;Azumaya Co., Tokyo, Japan). The MW was set to beam irradiation in intervals of 3 s on and 7 s off, at a frequency of 2.45 GHz at 300 W of output power with the temperature sensor set to 38°C. To compare the MW-irradiated and conventional FISH (described above), we tested all 17 probes under the two conditions on a lymphocyte control. Each probe was hybridized with MW irradiation for 5 min, 15 min, 30 min, 1 h, 3 h, 6 h, 12 h, and 18 h,respectively, and the results were compared with those for lymphocytes with nonirradiated hybridization. The numbers of nuclei identified by each probe and its signal intensity were simultaneously compared between these differently irradiated preparations. After this standardization, the frequency of diploid cells after hybridization with MW irradiation was compared with those without MW irradiation according to time course (mean ± SD) after fluorescent staining.

We used two indices for evaluation of numerical abnormalities of 17 chromosomes in 100 and more tumor cells in each case. We defined NCAI as the percentage of the tumor cells that do not have normal chromosome numbers (2 for autochromosome and chromosome X and 1 for chromosome Y). Modal number was defined as the number of the chromosomes most prevalent in tumor cells. The most frequent copy numbers (the modal number) of individual chromosomes was considered to represent the ploidy level of the tumor. We set the cutoff level at 20% of nuclei with an abnormal number of signals among all of the tumor nuclei; i.e., we call the cases with numerical abnormality for a given particular probe when the NCAI for that probe is >20%. As for chromosome Y in case 19, the modal number is still 1 (normal), but>49% of tumor cells have abnormal numbers of chromosome Y. In case 23, the two signals for D15Z are still the modal in the tumor cells,but 55% of the tumor cells are not diploid for chromosome 15. Actually, amplification of chromosome 15 based on NCAI exists in case 23. Thus, these two indices differently represent the cases of abnormal chromosomal numbers in the following description (15).

For each probe, statistical differences between the frequency of chromosomal abnormalities and histopathological types, depth, or stage categories were examined by χ² analysis.

The intensity of the signals from the probes hybridized with MW irradiation in control lymphocytes reached plateau values within 30 min. There was a significant difference in the sensitivity of probe detection within a 1-h hybridization between irradiated and nonirradiated samples with all probes (P < 0.001; Fig. 1). We could also obtain a more distinct signal in irradiated cancer cells than in nonirradiated ones (Fig. 2).

Our material consisted of 6 cases of early (tumor invasion limited to mucosal and submucosal layer) and 18 advanced gastric cancers. All of the histopathological data on the 24 patients and the distribution of tumor cells with copy numbers of each of the 17 chromosome centromeres studied are available on request. Numerical aberrations of the chromosomes were detected in both early and advanced stage gastric carcinoma, and the profiles of chromosomal aberration were nonrandom in terms of prevalence and chromosomal specificity. The prevalence of numerical abnormality (represented by NCAI, defined in “Materials and Methods”) in each chromosome was lowest in chromosome 3 (29.2%),followed by chromosome 18 at 33% and then chromosomes 10 and 16 at 45.8% of all cases. Abnormalities most frequently occurred in chromosomes 1 and 2 and were found in all of the cases examined. Abnormalities were found in 87.5% of cases in chromosome 4 and in 75%of all cases in chromosomes 8 and 20 (Table 3). In four early stage cancer cases (up to depth sm2 level), numerical increases were limited to chromosomes 1, 2, 4, and 20. All cases of early cancer had numerical abnormalities in chromosomes 1 and 2 (Fig. 3). When cancer invaded to the lower submucosal layer (sm3) and the proper muscle layer, additional abnormalities in centromere numbers gradually appeared, although changes in chromosomes 6 and 18 occurred in only one of nine cases (11%). Furthermore, chromosome 3 showed disomy in all nine cases, in which the depth of invasion was limited to the proper muscle layer of the gastric wall (Fig. 3). Loss of the Y chromosome was found in only one of five male cases. In contrast,abnormal amplification of chromosomes 1, 2, 4, 8, and 20 in the sm and mp layer occurred frequently (Table 1). Polysomy, represented by modal number, occurred in these 5 chromosomes at a frequency of 89% (8 of 9), 78% (7 of 9), 56% (5 of 9), 56% (5 of 9), and 33% (3 of 9), respectively, and NCAI larger than 20% was found in 100% (9 of 9), 100% (9 of 9), 89% (8 of 9),56% (5 of 9), and 67% (6 of 9) of the cases studied. These values were all significantly greater (P < 0.001) than the corresponding values for each of the other chromosomes (Table 2).

The more extensive involvement of multiple chromosomes became apparent when the tumor invaded beyond the serosa (Fig. 3). The prevalence of amplification in chromosomes 1, 2, 4, and 8 occurs more frequently,when the tumor invaded beyond the gastric wall (depth se and si) was 100% (11 of 11), 100% (11 of 11), 100% (11 of 11), and 72.7% (8 of 11), respectively, for the modal number and 100% (11 of 11), 100% (11 of 11), 100% (11 of 11), and 81.8% (9 of 11) of the cases for NCAI,respectively. The total prevalence of the numerical abnormalities in chromosomes 1, 2, 4, 8, and 20 was greater than that of all of the other chromosomes, 3, 6, 7, 10, 11, 12, 15, 16, 17, 18, X, and Y combined (Table 2; P < 0.001). Abnormalities in chromosomes 3, 6, and 18 were almost entirely restricted to advanced stage gastric cancer (Fig. 3).

Thirteen (65%) of 20 male samples and 1 (25%) of 4 females samples exhibited the gain of an X chromosome, whereas loss of X was not found in any sample. Gain of a Y chromosome by modal number was found in 3 of 20 male cases (15%). Five cases (25%) were considered to have a gain of Y chromosome based on NCAI. Loss of the Y chromosome was found in 4 samples (20%) by modal number and 5 cases (25%) by NCAI.

Significant loss in the other autosomal chromosomes was for chromosome 17 in 25% of tumor cells in case 12, in 37% of tumor cells for chromosome 18 in case 10, in 26% for chromosome 20 in case 17, and in 25% for chromosome 18 in case 19. Loss of chromosomes other than these were all <20% of tumor cells. There were less abnormalities in chromosomes in tumors of female patients, with no abnormalities being found in chromosomes 16 and 18 in female cases.

The prevalence of chromosomal numerical abnormality according to pathological subtype is shown in Table 3. The signet ring cell type (three cases) had the lowest rate of chromosome number changes in terms of NCAI (32%), followed by muc-sci(46%). The highest rate of alterations were seen in the tub-sci type(92.6%) followed by tub 2 (77.6%). There was a significant difference in chromosomal instability between the mucocellular (sig and muc-sci)and tubular-type (tub 1, tub 2, por 1 and tub-sci) carcinomas in total(39.0% versus 68.0%: P < 0.001; Table 3). Even in advanced stage, six samples of sig or muc-sci (cases 1, 2, 8,16, 18, and 24) had relatively fewer numerical abnormalities, with chromosomes 18 and Y being normal in all cases. Tub 2, por 1, and tub-sci, which all have severe morphological atypism (cytologically as well), showed a high frequency of abnormalities encompassing all of the chromosomes investigated, with amplification being found preferentially in chromosomes 1, 2, 8, and 20 in all 11 cases.

Although we only studied a few cases with distant metastasis, severe chromosomal abnormality was observed in these cases. In one who had multiple liver metastasis at operation, numerical abnormalities were found in all chromosomes, and in the other two cases who died due to peritoneal dissemination a few months after the operation, only one or two chromosomes with normal numbers could be found for either case. In our results, both the prevalence and the numbers of numerical abnormalities of the chromosomes increased as the stage progressed(Table 4).

*\. Paradoxically, stage IIIa had the highest abnormality rate (82%) and stage IIIb had the lowest rate (31%) in our series. However, no significant difference was found among the chromosomes involved with regard to lymph node metastasis. There were no significant differences based on the age of the subjects either. The rate of abnormality per chromosome depending on the depth of the tumor is shown in Fig. 3. On the basis of these results, we propose that the instability of the chromosomal centromeric copy number occurs in a multistep process for m-sm 2, sm3-mp, and ss-si according to the progression of the cancer(Figs. 3 and 4).

The only systematic analysis of abnormalities in chromosome copy numbers to date was reported by Rao et al. (2)in his analysis of 6 cases of gastric cancer using 15 types of centromeric probe. They studied fresh biopsy specimens and showed a gain in many chromosomes. However, their study was hampered by frequent samples that could not be analyzed and low detectability of abnormalities, probably due to contamination from normal cells and uncontrollable variations in staining quality. They did not find any trends for changes in number of specific chromosomes.

In our series of 24 surgically resected gastric cancer samples, the quality and quantity of tumor cells based on cytology and histology were well controlled by the new protocol (11). We were able to identify chromosomes susceptible to CIN in gastric cancer during the course of gastric cancer progression. Furthermore, we demonstrated a cumulative process of chromosomal centromeric abnormalities based on 17 currently available chromosome-specific probes. Fresh material taken from touch preparations of cancerous tissue was used for FISH analysis, and the cancer cell quantity and the difference between cytologically normal and cancer cells were monitored by using Papanicolaou and Giemsa staining of touch tissue as well as HE staining of touched sections. This procedure made it possible to detect abnormalities in chromosome number in tumor cells with greater accuracy.

Our stage-dependent study is the first to document the relationship between depth of invasion in the gastric wall and chromosomal numerical changes. Cancer in an early stage of invasion (up to depth sm 2) was predominantly close to diploid; a DNA ploidy level of 3 or more was seen in tub 2, por 1, and tub-sci, but not in the mucocellular type. Stage-dependent analysis (Fig. 3) revealed that abnormalities were limited to chromosomes 1, 2, 4, and 20 in the less invasive stages of m and sm1. When the cancer invaded to the sm3 level, abnormalities were more widely seen in other chromosomes. However, in some of the highly invasive cases (depth se), numerical abnormalities detected were limited to only four or five chromosomes including 1, 2, and 4 in all three cases. Therefore, we can conclude that chromosomes 1, 2, and 4 are the most susceptible to CIN in gastric carcinogenesis in both early and advanced stages. This raises the possibilities that fewer abnormalities may be an indicator of better prognosis.

On the basis of these observations, we depicted the scheme of sequential chromosomal numerical abnormalities during the course of progression of gastric cancer in Figs. 3 and 4. The numerical increase of centromeres of chromosomes 1 and 2 occurs in the earliest stage in any subtypes of gastric cancer. As shown in this scheme, the biological behavior of gastric cancer can be divided into those invading within two-thirds3 of the submucosal layer and those invading to the deeper layer of the gastric wall from the standpoint of CIN. In other words,up to sm 2, there are few abnormalities in a limited chromosomes (1, 2, 4, and 20), whereas abnormalities occur with a higher frequency once tumors start to penetrate through the submucosa (sm3), and then abnormal amplification expands to other chromosomes. After increases in chromosomes 1 and 2, increased numbers of chromosomes 4, 8, and 20 occur in a relatively earlier stage and at a high frequency. These may be the most characteristic chromosomes of gastric cancer. In the next stage, the abnormalities spread to chromosomes 7, 10, 11 12, 15, 16,17, and X. The abnormalities in chromosome 3, 6, 18, and Y are found only in a relatively late stage and are less likely to occur.

Chromosomes 3 and 18 seem to be strongly protected from centromeric numerical abnormality. Chromosome 3 showed virtually no abnormalities in nine cases limited to depth mp, which may be a useful marker of progression. Gain and loss of chromosome Y occur at an equal frequency. The loss of Y has previously been reported in conventional cytogenetic analysis of gastric cancers by Ochi et al. (5)and by Rodriguez et al. (16) in three of four and six of seven male tumor cases, respectively. Our results were consistent with their observations. Furthermore, van Dekkan et al. (1) and Rao et al. (2)reported loss of Y chromosome by FISH in five of eight and four of six cases, respectively. Because the stage of these tumors is not given in their reports, the significance of this high prevalence is hard to interpret. Rao et al. (2) also documented the loss of Y in three of three cases of esophageal cancer. In our data,loss of Y was found in 5 of 20 cases (25%), a lower rate than in previous reports. However, we are one of the first to show clear evidence of Y gain in three gastric cancer cases.

Chromosomal abnormalities according to histological subtypes are shown in Table 3. Subtypes tub 2, por 1, and tub-sci had a high frequency of abnormalities for each chromosomes, whereas the mucocellular type had a significantly lower number of abnormalities (Table 3). The order of the sequential chromosomal abnormalities during the progression seems to be common in both tubular and mucocellular subtypes; however, the extent of aneuploidal cancer cells in each case were quite different among subtypes. The extent of these abnormalities possibly reflects the biological behavior of the cancer within these subtypes. As shown in Table 3, abnormality in any chromosomes was 32%in sig, 46% in muc-sci, and in 50% in tub 1. In the tub-sci subtype,most of the tumor cells (92.6%) had numerical aberrations, and this profile is identical with that of scirrhous carcinoma.

The genetic events unique to gastric carcinogenesis remain unclear (6, 17, 18, 19). A high frequency of loss of heterozygosity for chromosome 1 has been observed. A tumor suppressor gene, p73, has also recently been discovered on the short arm of chromosome 1 (20). CIN of chromosome 1 has recently been detected in breast (21, 22) and bladder cancer (23) with FISH. Our data may add another type of human cancer to this list. It would be interesting to see the relationship between loss of heterozygosity and centromere amplification (24).

Some specific genes have been reported to be associated with the pathogenesis of gastric cancer including c-met on chromosome 7p (24), K-ras-2 on 12q (25), and c-erbB2 (26, 27, 28), E-cadherin (4),and p53 on 17p (29, 30, 31). Unexpectedly, FISH analysis showed no outstanding gain or loss in these chromosomes. In recent analyses of gastric cancer with CGH (32, 33), a considerably different profile of long and short arm abnormalities on the chromosomes has been documented. Because evaluation of the intensity in a chromosomal arm in CGH procedure is usually performed without assessment of numerical changes of that chromosome, our data would be complimentary to the CGH approach and should validate some of these studies.

Interestingly, we found that numerical abnormalities of chromosome 3 and 18 were rare during gastric carcinogenesis. Loss of the arms of these chromosomes is associated with lung (34), gastric,and colon cancer (35). More recently, CGH analysis suggests gain of 3p and 3q in gastric cancer (36). The DCC gene, loss of which is assumed to be related to colon cancer progression, is also on chromosome 18. In these lines, we could presume that centromere stability and arm instability may complementarily exist in chromosomes 18 and 3, which is consistent with Lengauer’s hypothesis of different pathways of CIN and microsatellite instability (3, 37).

The mechanistic basis for CIN observed here is challenging. Lengauer et al. (3, 37) advocate that centromeric instability represented by numerical centromeric abnormality can occur due to mechanisms different from microsatellite instability, which is caused by a mismatch repair deficiency. They further argue that the processes of carcinogenesis and cancer progression can arise through these two distinct pathways.

Conventional karyotyping of solid tumors is notoriously difficult. Even when tumor cells can be cultured, karyotypes may be modified by in vitro expansion of particular clones and artifactual chromosomal changes during the procedure; and these karyotypic changes obtained do not always reflect changes in vivo. The analysis of chromosomal abnormalities with FISH like ours can reveal naturally occurring alterations and cannot be simply compared with the previous karyotyping results.

In this study, we analyzed the chromosomal centromeric instability of 17 chromosomes and propose a stepwise process of CIN in gastric cancer progression for the first time. A chromosome-wide search for numerical abnormalities covering almost all of the chromosomes may provide new perspectives on genetic changes in gastric cancer. Our observations suggest that chromosomal gain is much more frequent than loss in terms of centromere numbers; thus, it could be postulated that oncogenic changes may contribute to gastric carcinogenesis more often than that by defective suppressor genes. However, some of these types of numerical abnormalities, on the other hand, may also occur as secondary changes during the course of carcinogenesis. Actually, the limitation of our argument is that the numbers of centromere cannot reflect the changes in chromosome arm. Intratumor heterogeneity cannot be assessed by the procedure in this study. We could not study chromosomes 5, 9,13, 14, 19, 21, and 22 because of the lack of appropriate specific probes. Because there are a number of genes located on chromosomes 5,9, and 13 that are presumed to be involved in human carcinogenesis, the status of the centromere gain or loss of these chromosomes will be of additional interest.