Modes of Silencing of p16 in Development of Esophageal Squamous Cell Carcinoma Free

Recently, inactivation of tumor suppressor genes as well as amplification/overexpression of proto-oncogenes have been reported to occur in esophageal carcinogenesis (1, 2, 3, 4). Most tumor suppressor genes are related to cyclins, CDKs,² CDK inhibitors, and the retinoblastoma-susceptibility gene (RB1), taking part in control of cell cycle progression from G₁ to S-phase (5). Three important CDK inhibitors p16^INK4a, p15^INK4b, and p14^ARF negatively regulate the G₁-S transition, maintaining pRb in an active (hypophosphorylated) state (6, 7). Among them, p16 gene, also known as MTS-1 (major tumor suppressor 1), INK4a (inhibitor of CDK4a), or CDKN2A (CDK inhibitor 2A), is the major target most commonly involved in carcinogenesis as p53. p16 protein binds to CDK4, inhibits the ability of CDK4 to interact with cyclin D1, and stimulates the passage through the G₁ phase of cell cycle.

Any one or a combination of the following events may silence tumor suppressor genes: deletion, point mutation, and the methylation of their promoter. The point mutation rate in the p16 gene was reported to be low in most of the primary tumors with allelic losses in 9p21 (8, 9), whereas microsatellite analyses using the markers close to the p16 gene have revealed that many tumor types possess a small (<200 kb) biallelic deletion involving the p16 gene (10). Unlike other tumor suppressor genes commonly inactivated by point mutations and deletions after the Knudson’s two-hit model, p16 is often silenced with homozygous deletion. Deletions at the 9p21 loci often encompass and inactivate at least two other nearby targets, p15^INK4band p14^ARF; the latter gives the alternative transcript of the p16 gene. Thus, homozygous deletion has been proposed as a mechanism for simultaneous silencing of two or more genes (the double-target hypothesis; Ref. 11).

Another mechanism for inactivating tumor suppressor genes is the aberrant methylation or hypermethylation of nonmutated promoter regions, which can lead to stable allele-specific loss of transcription function (12). Such methylations tend to occur at the sites of CpG dinucleotides, which are clustered as so-called CpG islands. Several tumor suppressor genes (e.g., RB1, p16, hMLH1, APC, and CDH1) contain CpG islands in their promoter regions, and the CpG islands are often methylated in silencing of the genes (13, 14, 15, 16, 17, 18). The methylation can be confined to one of the alleles (12, 19), and it could thus function as one of the Knudson’s two hits for the development of tumors (17, 20, 21, 22).

In ESCCs, the occurrence of homozygous deletion of the p16 gene was reported (23, 24). On the other hand, there were reports of the p16-promoter methylation in adenocarcinomas (25) but few in ESCCs. Maesawa et al. (26) reported that a p16 silencing commonly occurs either by methylation or by homozygous deletion in advanced ESCCs. But how p16 is silenced during carcinogenesis and progression of ESCCs has not been studied fully. In this study, using 42 cases of the ESCC, we examined the regional distribution of p16 immunoexpression, LOH status at chromosome 9p13–22, and the methylation status of p16 promoter in carcinomatous and dysplastic lesions, and in the surrounding esophageal mucosa around those lesions. We made topographic maps of the changes. From such maps, we retrospectively inferred the sequence of genetic and epigenetic changes on the following premises. Firstly, most of ESCCs are eventually monoclonal, as demonstrated by our group with the HUMARA assay (27). Secondly, in the cell proliferation of solid tumors, daughter cells remain contiguous with each other, and the cells with additional heritable changes constitute well-demarcated subclonal populations. Consequently, a subclone A is considered to precede another subclone B when the distribution of the A surrounds that of the B (28). We adopted laser capture microdissection method to prepare tumor samples with minimal contamination of non-neoplastic cells. Analysis of the methylation status and the expression of p16 in such samples enabled us to discriminate between monoallelic or partial and biallelic methylations.

We used 42 cases of the ESCC (patients: 37 males and 5 females, ranging in age from 46 to 74) resected in the University Hospital, Shiga University of Medical Science. The resected specimens fixed in 4% paraformaldehyde were cut through into the consecutive tissue blocks of 0.5 cm in width and 4 cm in length. These blocks were embedded in paraffin and processed for histological examination based on the Tumor-Node-Metastasis classification (29). We thoroughly examined H&E-stained slides of all of the blocks that were made by step sectioning method in each case.

We used the following mouse monoclonal antibodies: anticyclin D1 (5D4, 1:100; Medical and Biological Laboratories Co., Ltd., Nagoya, Japan), anti-p16 (G175–405, 1:25; PharMingen Becton Dickinson Co., San Diego, CA), and anti-pRb (3H9, 1:100; Medical and Biological Laboratories). Immunoreactivity was detected by an indirect biotin-streptavidin-peroxidase method using a Histofine kit (Nichirei, Tokyo, Japan), followed by diaminobenzidine reaction and counterstained with hematoxylin. The percentage of positive cells was then assigned to one of the five grades: 0 (0%); 1 (≤30%); 2 (30–50%); 3 (50–70%); and 4 (70–100%; Ref. 4). The slides of negative control without primary antibody were processed in parallel. Reactive stromal cells were used as positive internal controls for p16. Immunoreactivity of p16 was examined in at least eight of the tissue blocks with carcinomas and/or dysplasias, in adjacent normal-looking mucosa, and in several sections of remote normal mucosa, whereas cyclin D1 and pRb stains were done in representative sections for cancerous and precancerous lesions, and for the adjacent normal-looking mucosa and the remote normal mucosa. In 4 cases (cases 1, 10, 13, and 14), all of the sections were examined for immunoreactivity of p16, cyclin D1, and pRb.

Serial 5-μm paraffin sections were mounted on coated glass slides (Matsunami Adhesive Slide; Matsunami Glass Ind., Kishiwada, Japan). The first section was for H&E stain. The depth of the mucosal and the submucosal layers were divided into three sublayers of equal thickness: m1, m2, m3, and sm1, sm2, sm3, respectively. Using a laser capture microdissection system (Pixcell 100; Arcturus Engineering, Mountain View, CA) we took 250–350 target cells from an H&E-stained section without mounting coverslip. The sampled cells were immediately suspended in an 80-μl lysis solution containing 20 mm Tris-HCl, 1 mm EDTA, 0.5% Tween 20, and 200 μg/ml proteinase K (pH 8.0). The mixture was incubated for 48 h at 42°C and then boiled for 5 min to inactivate proteinase K. Two μl of the mixture were used for each PCR reaction. DNA was extracted from all of the tissue blocks made in four cases (cases 1, 10, 13, and 14). In the other cases, we extracted DNA from >50% of the cut-through tissue blocks including cancers, dysplasias, the adjacent normal-looking mucosa, and the remote normal mucosa.

Methylation status was analyzed in the upstream of p16 gene that includes the promoter region residing in the 5′ CpG islands. The normal and tumor DNAs modified with sodium bisulfite as described previously and amplified by methylation-specific PCR (30). Three pairs of p16 primers (p16-M, p16-U, and p16-W, which anneal to chemically modified methylated and unmethylated DNAs and any DNA without modification, respectively) were labeled with HEX (p16-W), 6-FAM (p16-U), or TET (p16-M). p16-W served as control for assessing the efficiency of chemical modification.

PCR mixtures contained buffer (GeneAmp 10× PCR Gold; Roche Molecular Systems, Inc., Branchburg, NJ), MgCl₂ (1.5 μm), deoxynucleotide triphosphates (200 μm), a primer pair (10 pmol/each reaction), 1 μl DNA sample, and 1.25 units Taq polymerase (AmpliTaq Gold; Roche Molecular Systems) in a 25-μl final volume. Twenty-two cycles of PCR was performed at 95°C for 1 min, at the each annealing temperature for 30 s, and at 72°C for 30 s with a final extension for 4 min at 72°C in a thermal cycler (PTC-100; MJ Research, Inc., Watertown, MA). After PCR, samples were diluted at 1:7 in formamide, heated to 95°C for 5 min, chilled on ice, and analyzed with Genescan software (27) on a genetic analyzer ABI PRIZM 310 (PE-Applied Biosystem, Foster, CA).

We used six microsatellite markers at chromosome 9p13–22: IFNa, D9S1747, D9S1751, D9S1748, D9S1752, and D9S171 (Fig. 1). The primers designed for these markers based on the Genome Database (IFNa, D9S1747, D9S1751, D9S1748, D9S1752, and D9S171; Ref. 31) were tagged with one of three fluorescent labels: HEX (D9S1747), 6-FAM (IFNa, D9S171, and D9S1751), or TET (D9S1748 and D9S1752). Using these markers, pairs of the normal and the sample (tumor and dysplasia) DNA were PCR amplified. The PCR products were processed and analyzed as described above. Allelic status was assessed by calculating an IF: IF = (T1/T2)/(N1/N2) (32), hence T1/T2 and N1/N2 are the ratios of heights of the peaks of heterozygous alleles in the electrophoretogram of the sample and the normal DNAs, respectively. The sample was scored as LOH when the IF was <0.5 or >2.0. We considered the alleles were homozygously deleted when it was repeatedly found that the peaks of the both alleles of the sample DNA were <10% of the normal peaks of the corresponding alleles, provided that the same DNA samples gave normal PCR products with another primer set.

We assessed statistical significance in the relationships between the presence/absence of methylation or LOH and tumor stages, and between the grades of immunohistochemical stains and tumor stages using Fisher’s exact test and Spearman’s rank correlation, respectively. P < 0.05 was regarded to be significant.

Twenty-six cases had multicentric cancers, up to four cancers in 1 case. Well/poorly differentiated carcinoma was detected in 5/3 of the 23 cases of earlier (I, IIA) stages and 1/4 of the 19 advanced stages. The other cancers were moderately differentiated (Table 1). Of 42 cases, 13, 10, 5, and 14 cases had main cancers with stages I, IIA, IIB, and III, respectively (Table 2). At least one focus of moderate to severe dysplasia was detected in each case. Mild dysplasia as well as basal cell hyperplasia was not included in the dysplasia or in the normal epithelia in this study. The dysplasias were either adjacent to or remote from the main cancers.

The results of immunohistochemistry were displayed in Table 1. We regarded an equivocal stain as negative. Positive immunoreactivity of p16 was confined to the nucleus and was clearly detected in most of the normal esophageal squamous epithelium (Fig. 2,A). The p16 immunoreactivity was virtually absent in dysplasias of 34 cases and in ESCCs of 38 cases (Fig. 2, B and C), whereas 4 (9.5%) of the 42 cases exhibited p16 expression in the cancers and dysplasias.

pRb immunoreactivity was also confined to the nucleus. In the normal epithelium it was invariably found in the parabasal layers and not in the other layers (Fig. 2,D). Most parts of the dysplasias and ESCCs expressed pRb (Fig. 2, E and F). ESCCs in 22 of the cases (46.3%) exhibited pRb overexpression (expression grade 4).

Cyclin D1 immunoreactivity was detected in most of the dysplasias and the ESCCs (Fig. 2, G and H), whereas it was weak and sparse or absent in most of the normal epithelium (Fig. 2,H). The 15 of the 18 of cases (83.3%) with lymph node metastasis (N1) highly expressed cyclin D1 (expression grade 3 or 4). The link between the grade of cyclin D1 immunoreactivity and the tumor stages (Table 1) was statistically significant (P = 0.0026).

ESCCs in 30 cases (71.4%) had methylated DNA in the absence of unmethylated DNA, and invariably showed silenced p16 expression (Table 2). The methylation status was homogenous in all of the tissue blocks examined in each cancer and dysplasia, and common among the multicentric cancers and dysplasias within each case. The normal-looking epithelia adjacent to the cancers/dysplasias in 24 of those 30 cases (80%) had both methylated and unmethylated DNA in the promoter region of p16 with p16 expression; these epithelia with the p16-promoter methylation were distributed in a ring-like zone of up to 2 cm in width around the cancers/dysplasias (Figs. 3, 4, and 5). The cancers in 8 of the 38 cases without p16 expression in the cancer did not have any methylated DNA of the p16 promoter and were surrounded by the normal epithelia without the methylation (Fig. 6). Four of the 8 cases had dysplasias without the methylation. No methylation was detected in the 4 cancers with p16 expression. The methylation data in individual cases are described in Table 1, and the relationship of methylation status to p16 expression, to the LOH at the near-p16 loci, and to the stage of the tumors is summarized in Table 2. No significant correlation was detected between the methylation status and early (I and IIA) and advanced (IIB and III) tumor stages (P = 0.23).

Allelic losses involving at least one of the six microsatellite loci examined were detected in the cancer/dysplasia samples of 32 cases (83.3%), whereas LOH at the near-p16 loci (D9S1748 and D9S1752) was seen in 24 cases (57%; Table 2). Of the 32 cases with the LOH, 11 had multicentric cancers, showing the common pattern of LOH among the cancers in each of 7 cases and discordant LOH pattern in 4 cases. Dysplasias with LOH at the near-p16 loci were found in 4 of the 24 cases that were accompanied by the carcinomas with the near-p16 LOH (Figs. 6 and 7). In these cases, the dysplasias and carcinomas commonly lacked the p16-promoter methylation. In 8 cases, dysplasias showed neither the methylation nor the LOH. The 24 cancers with LOH at the near-p16 loci were more frequently in advanced stages (18 cases: IIB and III, with metastasis to lymph nodes) than in earlier stages (6 cases: I and IIA, without metastasis to lymph nodes), whereas the LOH-negative tumors were mostly in earlier stages (17 of 18). This link between the near-p16 LOH and deeper invasion was statistically significant (P < 0.001), and was also demonstrated in individual tumors.

In case 10 (Fig. 3), LOH at D9S1747 was detected commonly in the superficial part (m1 and m2) and in the deeper parts, whereas LOH at a near-p16 (D9S1748) was only in the deeper part. In case 13 (Fig. 4), LOH at D9S1751 was detected commonly in the superficial and the deeper parts, whereas LOH at D9S1747 was only in the deeper part. In case 14 (Fig. 5), LOH at the 9p loci other than the near-p16 loci (D9S1747, D9S1751, and D9S171) was detected commonly in the superficial and the deeper parts, whereas LOH at the near-p16 loci was only in the deeper part. In case 1 (Fig. 6), LOH at a 9p locus other than the near-p16 loci (IFNα) and LOH at a near-p16 locus (D9S1752) were detected in dysplasias, as well as the superficial and the deeper parts of the carcinomas, whereas LOH at another near-p16 locus (D9S1748) was detected only in the superficial parts of the carcinomas.

Homozygous deletions at the 9p loci were detected in 4 of the 12 cases (33%) bearing cancers without p16-promoter methylation and in 4 of the 30 cases (13%) with the methylation. Homozygous deletions at near-p16 loci were detected in 4 cases (Fig. 7), all of which had cancers in advanced stages, whereas those at the other loci were sporadically found in the cancers both in the early and in the advanced stages (Table 1). No homozygous deletion was seen in the dysplasias examined. Mappings of the regional distribution of homozygous deletion in cancers indicated that the homozygous deletion of the near-p16 loci was detected within the area of LOH of the same locus (Fig. 6).

In 42 cases of the ESCC examined, 4 (9.5%) had cancers expressing p16 without any LOH of the loci examined or p16 promoter methylation. These cancers may have occurred without p16 silencing. The p16 expression was absent or considerably reduced in any part of the cancers in the other 38 cases (90.5%), of which 30 showed the p16-promoter methylation and 16 showed the LOH at the near-p16 loci in addition to the methylation. In these 16 cases, the cancer parts with the LOH at the near-p16 loci were almost completely included in the LOH-negative cancer parts, in which p16 silencing may be caused by p16-promoter methylation alone.

In the ESCC, it was reported that p16 methylation was detected preferentially in advanced stages (26). However, in this study we demonstrated that the p16 methylation was commonly detected not only in earlier stages of ESCCs but also in precancerous lesions and in the background normal-looking epithelia. In 24 of those 30 cases, cancers/dysplasias with the methylation were rimmed with apparently normal mucosa with the methylation and the p16 expression. This pattern was not observed in the normal epithelia remote from the tumor and may represent either monoallelic methylation or the presence of at least 0.1% of methylated alleles among unmethylated ones (30). Such DNA methylation is reported to occur frequently in noncancerous tissue of several organs (33, 34, 35), and it appears that a p16 silencing occurs during the process of changes from the normal-looking, partially methylated mucosa to dysplasia or the earlier ESCCs; the monoallelic or uncommon methylation becomes balladic and common in the promoter region of p16. Accordingly, in the areas of dysplasias as well as in cancers, the methylation was biallelic and homogeneously distributed (Figs. 3,4,5). Thus, the process of p16 silencing can be explained by the modified Knudson’s two-hit model (20), from monoallelic to biallelic methylation, or alternatively by the expansion of a cell population with biallelic methylation.

In 8 of cases with cancers of reduced p16 expression, no p16 methylation but LOH at the near-p16 loci was seen. None of the cancers in these 8 cases were surrounded by the normal-looking mucosa with p16 methylation. In these cancers, p16 silencing may have occurred through LOH either preceded or followed by mutation or microdeletion in the remaining allele of p16. We could not detect any homozygous deletion except in 2 of the 8 cases, probably because the p16 was silenced by mutation and LOH in these tumors, and possibly because the microsatellite markers we used did not include the one within p16. Our mappings of the regional distribution of homozygous deletion and LOH of the same locus in cases 1, 21, 22, 28, and 35 showed that the LOH areas without p16 expression included the areas of homozygous deletion (Fig. 6), suggesting that at least discernible homozygous deletion of p16 is a later event than p16 silencing. The homozygous deletions we detected at the 9p loci were not restricted to the near-p16 loci as reported previously (23), whereas those at the near-p16 loci were found only in advanced cancers. The homozygous deletion itself may occur quite at random, reflecting chromosomal instability. When it occurs incidentally at the near-p16 loci, it may be closely related to an acceleration of tumor progression.

The results of the present study suggest that there are at least two mechanisms of p16 silencing in esophageal carcinogenesis: one through methylation of p16 promoter and another through deletion at the near-p16 loci (including LOH and homozygous deletion) and possible mutation in p16. The former is major and the latter accounts for at most ∼20% of ESCCs. This incidence roughly conforms to the reported incidence of the homozygous deletion of p16 in ESCC (26). The p16 methylation was often followed by additional LOH at the near-p16 loci in the advanced stage as mentioned above. It remains unclear why the LOH occurred even after the p16 silencing by methylation, and how this occurrence is related to tumor progression. Such LOHs may possibly result from inactivation by DNA methylation of the caretaker genes that are responsible for chromosomal stability.

Our histological examination by the step sectioning revealed the presence of at least one moderate to severe dysplasia around the ESCCs in all of the 42 cases examined. In the 30 cases, both the dysplasias and carcinomas commonly showed p16 silencing with the promoter methylation, irrespective of the presence of LOH at the near-p16 loci in the carcinomas. No LOH was detected at the near-p16 loci in any dysplasias of these 30 cases. On the other hand, in 4 of the cases, both the dysplasia and carcinomas showed the LOH at the near-p16 loci without the methylation. Neither methylation nor LOH was detected in the dysplasias of the other 8 cases. Thus, half of the ESCCs with the LOH and without the methylation were accompanied and probably preceded by the dysplasias with the LOH, whereas all of the ESCCs with the promoter methylation were by the dysplasia with the methylation (Table 2). These data suggests that the two mechanisms of p16 silencing mentioned above already operate in the stage of dysplasia. More importantly, when multiple dysplasias and/or multicentric ESCCs coexisted in the one case, they always shared the particular mechanism of p16 silencing (Figs. 4,5,6). It seems that the mode of p16 silencing is not selected stochastically in individual lesions but might be determined in each patient before the occurrence of dysplasias and cancers. This is possibly caused by a carcinogenic microenvironment in the esophagus, which may be subject to environment factors affecting each patient and genetic backgrounds of each patient.

Immunoexpressions of pRb and cyclin D1 were observed in dysplasias as well as in ESCCs, and not in the normal epithelia adjacent to or remote from the tumor. The pRb overexpression may represent a reaction of wild-type pRb secondarily occurring in response to p16 silencing and/or cyclin D1 overexpression as reported previously (4).

Our mapping method was quite laborious, but a powerful method for determining the sequence of genetic and epigenetic events in individual cancers/dysplasias. This also enabled us to point out that the cancers, which developed through the deletion, may take a shorter and more aggressive course than those through the methylation. Follow-up studies are necessary for confirmation of this point.

We thank Drs. Akira Kawaguchi and Dr. Hiroyuki Naitoh (First Department of Surgery, Shiga University of Medical Science) for collecting resected specimens and furnishing clinical data, and Dr. Noriaki Kobayashi (Third Department of Internal Medicine, Kyoto Prefectural University of Medicine) for advice on the methods.