Changes in the glycosylation patterns of cell surface carbohydrates have been demonstrated in most human cancers. Many recent studies have indicated that aberrant glycosylation occurs early during oncogenic transformation and may represent a key event in invasion and metastasis.1 The blood group ABH antigens are cell surface carbohydrates distributed abundantly in epithelial tissues. Loss or reduction of A and B epitopes in human cancers is well documented, and loss of A and B antigens is correlated with the degree of malignancy and metastatic potential in lung, bladder and oral carcinomas.2, 3, 4, 5, 6, 7In vitro studies have demonstrated that loss or addition of a single glycosyl residue, for example, αGalNAc on A antigen or αGal on B antigen, may affect tumor cell motility by altering glycosylation of α3 and α6 integrin receptors and their interaction with β1 integrin. This may explain the observed correlation between glycosylation and prognosis.8
The A and B antigens are terminal carbohydrates synthesized step-by-step by the addition of single sugars catalyzed by a series of specific glycosyltransferases. The phenotypic difference between A and B antigens is due to differences in substrate specificity of A and B glycosyltransferases. Phenotypic O is characterized by the absence of A and B glycosyltransferases and presumably by the presence of an inert protein encoded by the O allele.9, 10, 11, 12, 13, 14, 15 Expression of A/B antigens in tumors is directly correlated with A and B glycosyltransferase activity.16 It is, however, not clear how the activity of A and B glycosyltransferases is suppressed in carcinomas. In vitro studies have shown that A antigen negative tumors have reduced levels of ABO transcript as compared to A antigen positive tumors.17 Loss of heterozygosity (LOH) studies of bladder carcinomas have suggested frequent deletion at chromosome 9q34, which contains the ABO gene.18 Alternatively, in vitro studies have shown that A antigen negative tumor cell lines have a higher level of DNA methylation at the ABO promotor region compared to A antigen positive cell lines, indicating that ABO promotor hypermethylation could be responsible for transcriptional inactivation.19, 20
As lack of ABH antigen expression is seen in the majority of oral squamous cell carcinomas and is correlated to patient survival, it is of major importance to elucidate the underlying molecular mechanism(s). The aim of the present study was therefore to study genetic and epigenetic alterations of the ABO locus in oral squamous cell carcinomas, including LOH at 9q34, loss of specific ABO alleles, and hypermethylation of the proximal ABO promoter.
MATERIAL AND METHODS
Frozen samples from 30 patients with oral squamous cell carcinoma were obtained from 2 different places. Group 1 of 14 patients was from the Department of Dentistry, Mackay Memorial Hospital, Taipei. Tissues from these patients were cut as frozen sections followed by fixation for 10 min in 99% ethanol before shipment. Group 2 of 16 patients was from Odense University Hospital, Denmark. Informed consent and approval by the Ethics Committee were obtained according to Danish legislation. Tissues from these patients were stored in −80°C and used as frozen unfixed sections. Histologically normal tissues adjacent to the tumors were included. The age range of patients was 37–89 (59.28 ± 12.20) years; there were 5 women and 25 men. Blood group types of patients were established by routine procedures at hospitals and further evaluated by immunohistochemical staining of normal epithelium and vessels in the connective tissues. There were 21 blood group A patients, 4 blood group B patients and 5 blood group O patients. Frozen biopsies from buccal and gingival epithelia, as well as epidermis were obtained from 2 healthy individuals aged 52 and 59 years. Both were blood group A individuals.
To investigate the histological distribution of A/B blood group antigens and A/B transferases, immunohistochemical stainings were carried out. All sections were incubated at 4°C overnight with undiluted supernatant anti-A monoclonal antibody HH6 (IgM),21 anti-B monoclonal antibody (IgM; DAKO, Copenhagen, Denmark) at a dilution of 1:20 and undiluted supernatant antibody 2F7 (IgG1) for A/B transferase,21 then washed with PBS and incubated at 4°C for 1 hr with FITC-conjugated rabbit anti-mouse immunoglobulins (DAKO) diluted 1:70 in PBS. The control reaction included omission of the first layer and staining with another monoclonal antibody of irrelevant specificity but of the same isotype as the specific antibody.
Microdissection and DNA extraction
A laser capture microdissection system (P.A.L.M.) was used to separate tumor and normal tissues, as well as blood group A or B antigen positive and A or B negative tumor cells if both of them were seen by IHC staining. The cells were transferred to sterile microcentrifuge tubes. Epithelia, obtained from the biopsies of 2 healthy individuals, were separated from connective tissues by EDTA as described previously.22 The epithelial tissues were put into the microcentrifuge tubes for DNA extraction. DNA was extracted by routine procedures by use of a DNeasy Kit (Qiagen, Venlo, The Netherlands).
PCR was carried out in a final volume of 25 μl including 1× PCR buffer, 0.15 mM each dNTP, 12.5 pmol of each primer (FY46, 5′-GAATTCACTCGCCA CTGCCTGGGTCTC-3′ and FY57, 5′-GAATTCATGTGGGTGGCACCCTGCCA-3′), 0.5 U of Taq polymerase, and 10–20 ng of DNA. The PCR amplifications were done for 35 cycles for 60 sec at 95°C, 1 sec at 60°C and 30 sec at 72°C with an addition of 2 sec/cycle to the extension step, using a block thermocycler (GeneAmp PCR System 2400; Perkin-Elmer, Foster City, CA). Briefly, digestion of the PCR products with KpnI at nt258 revealed the O allele, whereas digestion with BstEII showed the presence of either A or B alleles. Blood group O type samples and water were used as positive and negative controls, respectively.
Matched pairs of tumor and normal DNA samples from 27 patients were screened for LOH at the 9q34 region, using the microsatellite markers D9S1847, D9S915, D9S2157, D9S1198, D9S1793, D9S66, and D9S1818 (Fig. 1). The primer sequences were obtained from the Genome Database (http://www.gdb.org/) and the mapping distance data were from http://www.ncbi.nlm.nih.gov/mapview/maps.cgi (Fig. 1). A radioactive PCR was carried out in a volume of 10 μl in microtiter plates (Hybaid, Franklin, MA), 20 ng of template DNA, 1–3 mM nonlabeled and 33P end labeled primer, 0.1 mM each dNTP, 1.5 mM MgCl2 and 0.2 U of Taq DNA polymerase in buffer (Promega, Madison, WI). The amplification was run in a programmable heating block (Omniegene, Hybaid) for 30 cycles at 96°C for 30 sec, 48°C for 30 sec and 73°C for 40 sec. Chain elongation was continued for 5 min after the last cycle. The amplified fragments were separated by electrophoresis for 1–3 hr at 100 W in 6% denaturing polyacrylamide sequencing gels.23 LOH was determined as at least a 50% reduction in the relative intensity of one allele compared to the normal control; microsatellite instability (MSI) was identified by the presence of one or more novel microsatellite alleles in tumor samples.24 Control samples were included during the whole procedure.
Genomic DNA was treated with sodium bisulfite as described previously.25 Briefly, 1 μg of DNA was denatured in a volume of 78 μl 0.3 N NaOH for 15 min at 37°C, followed by 5 min at 99°C, and then cooled immediately on ice. One milliliter of a freshly prepared solution containing 4.2 M sodium bisulfite (Sigma, St. Louis, MO) and 0.5 mM hydroquinone (Sigma) was added. After incubation at 55°C for 16 hr, the modified DNA was recovered using glass milk from the GeneClean II Kit (Bio 101 Inc., Carlsbad, CA). The reaction was completed by incubating the recovered DNA in 0.3 M NaOH for 15 min at 37°C, followed by ethanol-precipitation. DNA was resuspended in Tris-EDTA and used immediately or stored at −80°C until use. DNA treated with SssI methyltransferase (New England Biolabs, Beverly, MA) was used as the methylated control.
Methylation-specific PCR (MS-PCR) was carried out as described by Herman et al.26 The primer sequences used for amplification of the unmethylated and methylated ABO proximal promoter and DAPK1 promoter were as described previously by Kominato et al.19 and Katzenellenbogen et al.,27 respectively. The primers for the distal promoter of the ABO gene were dABO-MF, 5′-GGGGTTTTGTTACGGATCGC-3′ and dABO-MR, 5′-AAACTCTACGTCCCGCAAACTAAA-3′ for the methylated reaction, and dABO-UF, 5′-AGTGGGGTTTTGTTTATGGATTGT-3′ and dABO-UR, 5′-ACAAAACTCTACATCCCACAAACTAAA-3′ for the unmethylated reaction. These primers cover a region from position −650 to position −400 designed, according to Kominato et al.28 Reaction mixtures (25 μl) contained 10 pmol of each primer, 0.2 mM each dNTP, 1 U of HotStarTaq DNA polymerase (Qiagen), 1× PCR buffer (Qiagen), 0.2 mM cresol red, 12% (w/v) sucrose, and 50–100 ng of bisulfite-treated DNA. Reactions for the ABO promoter were started with initial denaturation at 95°C for 10 min, followed by 35 cycles at 95°C for 1 min, the appropriate annealing temperature for 1 min and 72°C for 2 min. The annealing temperatures for amplification of the unmethylated and methylated ABO proximal promoter were 60°C and 65°C, respectively; for the distal promoter they were 60°C and 67°C, respectively. The DAPK1 promoter was amplified by using the program described by Katzenellenbogen et al.27 The PCR products were resolved on 2% agarose gels.
Methylation-specific melting curve analysis (MS-MCA) was carried out as described previously.29 Amplification was carried out using the Lightcycler (Roche Diagnostics, Mannheim, Germany) in 20-μL reactions containing 2 μL of 10× Lightcycler FastSTART DNA Master SYBR Green I reagent (Roche), 3 mM MgCl2, 50–100 ng of bisulfite-treated DNA, and 0.5 μM each primer. PCR was initiated by incubation for 10 min at 95°C, followed by 35–40 cycles of 10 sec at 95°C, 10 sec at 65°C and 40 sec at 72°C. The primer sequences used for amplification of the ABO proximal promoter were ABO-LC-F, 5′-[CCCGCCC]GGGGGAGTAGGATGTTAGGGGGT-3′ and ABO-LC-R, 5′-[GCCCGCCGCGCCCCGCGCCCGTCCCGCCGCCCCCGCCCGCCCCT]TAACCCTAATACCTTAAAACCCTATC-3′. Nucleotides in brackets denote a GC-clamp. DNA melting curves were acquired immediately after amplification by measuring the fluorescence of SYBR Green I during a linear temperature transition from 70°C to 98°C at 0.05°C/sec. Fluorescence data were converted into melting peaks by the Lightcycler software (Ver. 3.50) to plot the negative derivative of fluorescence over temperature vs. temperature (−dF/dT vs. T).
Loss of A/B antigen expression in oral carcinomas
Immunohistochemical staining of tissue sections from 25 oral squamous cell carcinomas using anti-A (HH6) and anti-B monoclonal antibodies showed that 21 of the tumors (84%) had lost expression of blood group A or B antigens (Table I). Twenty of these tumors had entirely lost A/B expression (Fig. 2d) and one showed a mixture of positive and negative staining. The 4 remaining tumors showed positive staining of A/B antigens, in some cells the staining was accumulated in the cytoplasm whereas other cells exhibited membrane staining as in normal epithelium (Fig. 2a). In one of the B-negative cases (#15374), dysplastic epithelium adjacent to the tumor also showed B antigen negative staining (Fig. 2c). In 5 cases, the histologically normal epithelia showed patchy loss of A/B-antigen expression. As expected, all tumors and adjacent epithelia from the 5 blood group O individuals stained negative.
Table I. IHC Staining, Genotyping, LOH and Methylation Analysis Results in Oral Carcinomas1
LOH or MSI
IHC of A/B enzyme
IHC of A/B antigen
NA, no available information; ABO-P, ABO proximal promoter; ABO-D, ABO distal promoter; +/−, positive cells were more numerous than negative ones; −/+ more negative cells than positive cells. T, E following the case numbers represent tumour and adjacent epithelium. Epithelia in 15374 showed dysplasia, in 27088 showed hyperplasia. In 31572, T1 represents as well-differentiated tumor cells adjacent to normal epithelium, T2 represents poor-differentiated tumor cells far from normal epithelium. LOH 1–7 represent markers D9S1847, D9S915, D9S2157, D9S1198, D9S1793, D9S66, D9S1818 respectively.
Hypermethylation was found in both normal epithelia and tumor cells.
Unfixed, frozen tissues from 16 cases were examined for expression of A and B glycosyltransferases, using the monoclonal antibody, 2F7. A/B transferase expression was seen as a granular peri-nuclear staining. In both normal and tumor tissues, expression of the A/B transferases was directly related to the expression of A/B antigens (Fig. 2a,b). Accordingly, all cases with A/B transferase expression showed expression of A/B antigens, and all cases with expression of A/B antigens showed expression of A/B transferases. This indicated that expression of blood group A/B antigen in oral carcinomas was controlled at the enzyme level.
Genotyping and LOH analyses at the ABO locus
To determine the genotype of blood group A/B patients, DNA from normal epithelia and tumor-free connective tissues was amplified by PCR and digested with KpnI and BstEII. Twenty cases carried an AO/BO genotype and 5 carried an AA/BB genotype (Table I). Genotype analysis of tumor DNA showed A/B allele loss in 3 of 20 AO/BO cases (15%) (Fig. 3). This loss corresponded to loss of A/B-antigen expression. In one case with patchy loss of B antigen expression (#33103), the B allele was lost in some areas of the tumor with negative staining for B antigen, but was retained in other B antigen-negative areas and in B-antigen positive area. Three tumors had lost the O allele and retained the A/B allele (Fig. 3). Two of these tumors showed loss of A/B antigen expression. The remaining 14 cases had an apparently intact ABO locus, but only 3 of them showed positive A/B antigen staining. No allelic loss was seen in normal and dysplastic epithelia.
LOH analysis at the 9q34 region using 7 microsatellite markers showed allelic loss in 7 of 27 oral carcinomas (26%), and one case showed microsatellite instability (MSI) at 3 markers (Figs. 1,3; Table I). The most frequent loci showing LOH in these tumors were 9S2157, 9S1847 and 9S1198. All 8 patients showing LOH or MSI in their tumors belonged to blood group A (5 AO and 3 AA). Except for one case (#17093), both LOH and MSI were associated with negative staining of A antigen. In AO patients with LOH, one showed A allele loss and one showed O allele loss by genotyping (Fig. 3). Of 5 homozygous AA/BB patients, 3 showed LOH.
Hypermethylation of the proximal ABO gene promoter
To examine the methylation status of the proximal ABO gene promoter, 72 samples from 30 patients were initially investigated by MS-PCR (Table I). Aberrant methylation of the ABO promoter region was present in 13 samples from 10 patients. In one case (#31572), ABO hypermethylation was found both in well-differentiated tumor cells adjacent to normal epithelium (T1) and in poor-differentiated tumor cells far away from normal epithelium (T2) (Fig. 4). In 2 cases, hypermethylation was also found in epithelial tissues adjacent to the tumors, including dysplastic tissue showing loss of B antigen expression (#15374) (Fig. 4) and hyperplastic tissue from a blood group O individual (#27088).
To further characterize ABO methylation patterns in oral carcinomas, all 13 samples showing a positive signal for methylated alleles using MS-PCR were investigated by MS-MCA. Aberrant methylation was confirmed in all cases. Two of these samples contained high amounts of fully methylated ABO alleles (#CT18 and #31572T2); 2 contained high amounts of heterogeneously methylated ABO alleles (27088T, 15374T), which matched with a weak reaction for unmethylated amplification using MS-PCR; and 9 contained low amounts of heterogeneously methylated ABO alleles, including the 2 epithelial tissues (27088E and 15374E) (Fig. 4). Nine of 10 tumor samples with ABO hypermethylation were from blood group A/B patients, and all of them showed loss of A/B-antigen expression. The remaining sample was from a blood group O person. No aberrant methylation was found in four samples with positive blood group A/B antigen staining.
Hypermethylation of the distal ABO gene promoter was found in tumors from 13 cases. Hypermethylation, however, was also found in epithelium and connective tissues from 8 and 2 of these cases, respectively. No relationship was found between hypermethylation of the distal ABO promoter and loss of A/B antigen expression (Table I). Five of 30 cases harbored hypermethylation of the DAPK1 gene located proximal to the ABO gene, but there was no correlation between the methylation status of the DAPK1 and ABO genes (Table I).
A summary of ABO alterations that may account for loss of A/B antigen expression in 14/21 oral squamous cell carcinomas is shown in Figure 5.
It is well known that loss of expression of histo-blood group A/B antigens is a frequent event in oral squamous cell carcinoma, and that this loss is caused by loss of ABO glycosyltransferase activity.16, 21, 30 The present work confirms these previous results as 84% of the investigated tumors from blood group A/B patients showed loss of A/B antigen and corresponding lack of A/B glycosyltransferase staining. We found that only 3 of 20 samples with AO/BO genotype showed loss of A or B alleles. These results are in agreement with previous results based on paraffin embedded, formalin fixed oral carcinoma tissue.31 The chromosomal region 9q34, which is the locus for the ABO gene, is one of the regions that most frequently shows alterations in transitional cell carcinoma of the bladder,32, 33, 34, 35, 36 even though allelic loss would explain loss of A/B antigen expression in only 5 of 26 cases.18 These results are similar to our findings in oral carcinomas, showing allelic loss in 3 of 21 cases with loss of A/B antigen expression.
Previous studies of bladder cancer indicate that loss of A/B expression involves deletion of a large chromosomal region including the ABO locus at 9q34.1–2,18, 32 which has been suggested to contain one or more tumor suppressor genes. We therefore investigated alterations of the 9q34.1–2 region by microsatellite analysis and found LOH or MSI in this region in 8/23 (35%) blood group A/B patients. In 7 of these 8 patients, the changes correlated with loss of blood group A antigen expression although only one of them showed A allele loss by genotype analysis. Loss of antigen expression in these cases may be explained by the hypothesis that LOH in the vicinity of the ABO gene could comprise upstream genes involved in regulation of ABO expression.18 In the remaining case, retention of A antigen expression in the tumor could be explained by retention of the A allele and loss of the O allele, as determined by genotype analysis.
Aberrant promoter hypermethylation of normally unmethylated CpG islands has been documented to be associated with transcriptional inactivation of many cancer-related genes.26, 37, 38 Studies of the regulatory mechanism of ABO gene transcription have demonstrated the presence of 2 promoter regions.19, 28, 39 Expression of the ABO gene in epithelial and erythroid cells lines was shown to be dependent on the methylation status of the proximal constitutive promoter encoding most of the ABO transcripts, as an inverse relationship was found between promoter hypermethylation and ABO gene expression.19, 20 It was further demonstrated that treatment of cells with the demethylating agent, 5-aza-2′-deoxycytidine, can result in demethylation of the ABO promoter region and restore transcriptional activity.19 The activity of the distal promoter is less well described but seems to be dependent on cell type. In our present study, hypermethylation of the distal promoter was found in both oral carcinomas and normal tissues. These data are in agreement with previous studies showing cell specificity of distal promoter methylation, and suggest that the methylation status of this promoter is not related to expression of blood group antigens during the malignant progress. Previous data have suggested that negative regulatory mechanisms other than DNA methylation might play a role in downregulation of transcription from the distal promoter in some cells.28, 39
We found hypermethylation of the ABO gene constitutive proximal promoter region in 10 of 30 cases of oral squamous carcinoma by MS-PCR and MS-MCA analyses. The latter technique allowed us to distinguish between methylated and unmethylated alleles, and between fully and heterogeneously methylated alleles.29 Two samples showed high amounts of heterogeneously methylated DNA. In these 2 patients, a low amount of heterogeneously methylated DNA was also found in dysplastic and hyperplastic epithelia adjacent to the tumors. In another patient, 2 populations of tumor cells corresponding to different histological pictures were identified. One was a well-differentiated carcinoma that showed a small amount of methylated ABO alleles, whereas the other was poorly differentiated and contained high amounts of fully methylated alleles. These data suggest a direct correlation between ABO hypermethylation and the degree of malignancy. The levels of DNA methylation have been shown to increase during aging.40, 41, 42 Considering that aberrant ABO hypermethylation could be found in oral carcinomas but not in adjacent normal tissue, however, it is unlikely that this epigenetic event is an age-related phenomenon. When hypermethylation in hyperplastic or dysplastic epithelium is found, it may therefore be an early sign of malignant transformation.
In normal epithelium, A/B antigen expression is known to correlate with the epithelial differentiation pattern (keratinized versus non-keratinized epithelium).43, 44 This has led to the suggestion that changes in ABO gene expression through cellular differentiation may be influenced by methylation of the promoter region.19 Hypermethylation of the proximal ABO promoter region was unrelated to the differentiation-dependent expression seen in normal tissue such as skin and keratinized gingiva; these tissues showed very little or no expression of blood group antigen, but a normal methylation pattern.
It is possible that hypermethylation of the proximal ABO promoter is merely a manifestation of regional hypermethylation and that another gene in this region is the critical target for this epigenetic change. Hypermethylation of the death-associated protein kinase gene (DAPK1), which is located proximal to the ABO gene at 9q34,45 was found in 5 of the 30 oral carcinomas. There was no association, however, between ABO and DAPK1 hypermethylation. Concomitant ABO allelic loss/LOH and hypermethylation were found in 6 patients (Table I), which is in agreement with previous findings showing that known hot-spots for LOH in cancer are also frequent targets for aberrant hypermethylation.46
Identification of allelic loss, LOH, and hypermethylation targeting the ABO gene may substantiate previous experimental and clinical findings suggesting that loss of A/B antigen is a causal event in the progression of oral squamous cell carcinoma.6, 8, 47 Hypermethylation, but not LOH, was also found in hyperplastic and dysplastic epithelia adjacent to tumors, indicating that aberrant methylation of the proximal ABO promoter may represent an early event in the development of oral carcinoma. Additional finding of different methylation patterns in well- and poor-differentiated tumor cells indicates that ABO hypermethylation may be related to malignant degree.
As the decreased expression of blood group A/B antigens in oral carcinoma could not in all cases be attributed to allelic loss or hypermethylation of ABO, it is possible that other negative regulatory factor(s) outside of the ABO promoter is functional in transcriptional regulation of the ABO gene.19 Furthermore, aberrant glycosylation in oral cancer may not only be controlled by changes of A/B glycosyltransferase gene but also by other factors. For example, nucleotide sugar transport in the Golgi apparatus should be taken into consideration. The alteration in nucleotide sugar transport activity is expected to greatly affect the synthesis and expression of carbohydrate antigens on cancerous cells. The Golgi transport of UDP-Galactose has in this way been shown as one of the important factors influencing expression of cancer-associated carbohydrate determinants involved in cell adhesion in colon cancer.48 Such dysfunction may be reflected in the altered and diffused distribution of certain GalNAc-transferase seen in oral carcinomas.49 Finally, the activity of other enzymes such as glycosidase and glycosylhydrolases50, 51 or the competition among multiple glycosyltransferases for the limited amount of transported donor within Golgi compartments should be considered as well.52
We would like to thank Dr. E. P. Bennett for his supervision of the genotyping study, Professor K.W. Chang for provision of samples, Ms. D. Nielsen for her daily assistance, Associate Professor U. Mandel for providing antibodies HH6 and 2F7, and Ms. H.L. Hansen, Ms. V. Ahrenkiel, Ms. A. Mikkelsen, Ms. L. Rasmussen and Ms. W. Wang for their expert technical assistance.