X-linked inhibitor of apoptosis (XIAP) is the most potent member of the IAP family that exerts antiapoptotic effects by interfering with the activities of caspases. Recently, XIAP-associated factor 1 (XAF1) and two mitochondrial proteins, Smac/DIABLO and HtrA2, have been identified to negatively regulate the caspase-inhibiting activity of XIAP. To explore the candidacy of XAF1, Smac/DIABLO, and HtrA2 as a tumor suppressor in gastric tumorigenesis, we investigated the expression and mutation status of the genes in 123 gastric tissues and 15 cancer cell lines. Whereas Smac/DIABLO and HtrA2 transcripts were normally expressed in all cancer specimens we examined, XAF1 transcript was not expressed or present at extremely low levels in 40% (6 of 15) of cancer cell lines and in 23% (20 of 87) of primary carcinomas. Abnormal reduction of XAF1 expression showed a strong correlation with stage and grade of tumors, and a tumor-specific down-regulation of XAF1 was observed in 45% (9 of 20) of matched sets. Unlike XAF1, XIAP expression exhibited no detectable alteration in cancers. Whereas loss of heterozygosity within the XAF1 region or somatic mutations of the gene was not detected, expression of XAF1 transcript was reactivated in all nonexpressor cell lines after 5-aza-2-deoxycytidine treatment. The 5′ upstream region of the XAF1 gene encompasses no gastric cell-rich region that rigorously satisfies the formal criteria for CpG islands. However, bisulfite DNA sequencing analysis for 34 CpG sites in the promoter region revealed a strong association between hypermethylation and gene silencing. Moreover, transcriptional silencing of XAF1 was tightly associated with hypermethylation of seven CpGs located in the 5′ proximal region (nucleotides −23 to −234). Additionally, loss or abnormal reduction of XAF1 expression was found to inversely correlate with p53 mutations, suggesting that epigenetic inactivation of XAF1 and mutational alteration of p53 might be mutually exclusive events in gastric tumorigenesis. Collectively, our study suggests that epigenetic silencing of XAF1 by aberrant promoter methylation may contribute to the malignant progression of human gastric tumors.
Apoptosis is essential for elimination of defective or potentially dangerous cells and provides a defense against malignant transformation and autoimmunity (1). Several genes critical in the regulation of apoptosis have been identified, including the IAP3 family (2). The IAP proteins are a new class of intrinsic cellular regulators of apoptosis that are structurally defined by the presence of the evolutionary conserved BIR domain (2, 3). Deregulation of IAPs has been suggested to play a key role in the aberrantly increased cell viability and resistance to the anticancer therapy in human cancers, whereas overexpression seems to suppress apoptosis against a large variety of triggers (4, 5). The human IAP family includes cIAP-1, cIAP-2, XIAP, NAIP, survivin, apollon, ILP2, and livin, and several members of the human IAP family including XIAP, c-IAP-1, and c-IAP-2 have been shown to be potent inhibitors of caspase-3, -7, and -9 (2, 6, 7).
Of the eight known human IAP proteins, XIAP is the most potent and versatile inhibitor of caspases and apoptosis. XIAP demonstrates significant inhibition of apoptosis induced by many triggers, including serum withdrawal and etoposide exposure (8, 9). XIAP possesses three BIR domains and a COOH-terminal RING zinc finger and inhibits both the initiator caspase-9 and the effectors caspase-3 and -7. XIAP mRNA levels are relatively high in the majority of cancer cell lines, and high levels of XIAP protein are generally found in renal cancer and melanoma cell lines (10, 11).
The caspase-inhibiting effects of IAPs are antagonized by apoptosis-promoting proteins. Two mitochondrial proteins, termed Smac/DIABLO and HtrA2, have been identified to promote caspase activation by antagonizing the caspase-inhibitory activity of XIAP (12, 13, 14). Smac/DIABLO is released from mitochondria together with cytochrome c during mitochondria-induced apoptosis. The binding of Smac/DIABLO to XIAP is proposed to destabilize the XIAP-caspase interaction by steric hindrance, resulting in disruption of the XIAP-caspase complex (15, 16). HtrA2 is a mammalian serine protease homologous to the bacterial HtrA endoprotease and directly binds to the BIR3 domain of XIAP and activates caspases. In a similar manner to Smac/DIABLO, HtrA2 is released from mitochondria into the cytosol and forms complexes with XIAP at high stoichiometry as a result of apoptosis stimulus. These observations demonstrate that Smac/DIABLO and HtrA2 sensitize cells to apoptosis in response to a number of stimuli by binding to and antagonizing XIAP.
Very recently, a novel negative regulator of XIAP termed XAF1 was isolated based on its ability to bind XIAP (17). Transient overexpression of XAF1 sensitizes tumor cells to the proapototic effects of etoposide and reverses the XIAP-mediated inhibition of caspase-3 activity. In contrast to Smac/DIABLO and HtrA2, XAF1 resides in the nucleus and can effect a marked relocalization of XIAP protein from the cytoplasm to the nucleus. The XAF1 gene is located at 17p13.2, approximately 3 cM telomeric to the p53 tumor suppressor gene, and encodes Mr 33,100 protein with seven zinc fingers with high amino acid similarity with the zinc finger domains of both FLN29 and TRAF3 (10, 17). XAF1 mRNA is expressed ubiquitously in all normal adult and fetal tissues but is present at very low or undetectable levels in various cancer cell lines. Expression analysis using the NCI 60 cell line panel also revealed that cancer cell lines exhibit very low levels of XAF1 mRNA, whereas the majority of cancer cell lines express relatively high levels of XIAP mRNA, suggesting that deregulation of apoptosis through the loss of XAF1 expression might be important for malignant cell survival, and a high level of XIAP to XAF1 expression in cancer cells might provide a survival advantage through the relative increase of XIAP antiapoptotic function (10).
Gastric cancer is one of the most commonly diagnosed malignancies worldwide and a leading cause of cancer mortality in certain areas, such as Korea, Japan, South America, and Eastern Europe (18). Although evidence has accumulated indicating the involvement of the alterations of multiple genes such as p53, K-ras, c-erbB2, K-sam, and E-cadherin, the underlying molecular events that drive the neoplastic process in gastric cancer are largely undefined (19). In the present study, we investigated the expression and mutation status of XAF1, Smac/DIABLO, and HtrA2 in a series of primary gastric adenocarcinomas and cell lines to explore the candidacy of these genes as a suppressor in gastric carcinogenesis. Our data demonstrate that XAF1 mRNA expression is lost or significantly down-regulated in a considerable fraction of gastric cell lines and primary tumors by aberrant promoter CpG hypermethylation, whereas none of tumors show detectable abnormality of Smac/DIABLO and HtrA2. Moreover, loss or down-regulation of XAF1 expression because of hypermethylation was strongly associated with advanced stage and higher grade of tumor, suggesting that epigenetic inactivation of XAF1 may play a critical role in the malignant progression of gastric cancers.
Materials and Methods
Tissue Specimens and Human Cell Lines.
A total of 138 gastric tissues including 87 primary adenocarcinomas, 3 adenomas, 6 hamartomas, 7 hyperplastic polyps, and 20 normal gastric tissues were obtained from 87 gastric cancer patients and 36 noncancer patients by surgical resection in the Kyung Hee University Medical Center (Seoul, Korea). Tissue specimens were snap-frozen in liquid N2 and stored at −70°C until used. Tissue slices were subjected to histopathological review, and tumor specimens composed of at least 80% carcinoma cells were chosen for molecular analysis. Fifteen human gastric cancer cell lines (SNU1, SNU5, SNU16, SNU216, SNU484, SNU601, SNU620, SNU638, SNU719, MKN1, MKN28, MKN45, MKN74, AGS, and KATO-III) were obtained from the Korea Cell Line Bank (Seoul National University, Seoul, Korea) or American Type Culture Collection (Manassas, VA).
Quantitative PCR Analysis.
One microgram of DNase1-treated RNA was converted to cDNA by reverse transcription using random hexamer primers and MoMuLV reverse transcriptase (Life Technologies, Inc., Gaithersburg, MD). PCR was performed initially over a range of cycles (24, 26, 28, 30, 32, 34, 36, and 38 cycles), and 2 μl of 1:4 diluted cDNA (12.5 ng/50 μl PCR reaction) undergoing 28–36 cycles was observed to be within the logarithmic phase of amplification with primers Xaf1–3 (sense, 5′-ATGGAAGGAGACTTCTCGGT-3′) and Xaf1–4 (antisense, 5′-TTGCTGAGCTGCATGTCCAG-3′) for XAF1, Smac-1 (sense, 5′-GAGCAGTGTCTTTGGTAACA-3′) and Smac-2 (antisense, 5′-ACCTGCAGTTTCACCAGCTG-3′) for Smac/DIABLO, HtrA2–1 (sense, 5′-AGATCCTGGACCGGCACCCT-3′) and HtrA2–2 (antisense, 5′-TCCAGAGTTTCCAAAATCAA-3′) for HtrA2, Xiap-3 (sense, 5′-GATTATGAAGCACGGATC-3′) and Xiap-4 (sense, 5′-GACTTGACTCATCTTGCA-3′) for XIAP, and G2 (sense, 5′-CATGTGGGCCATGAGGTCCACCAC-3′) and G3 (antisense, 5′-AACCATGAGAAGTATGACAACAGC-3′) for an endogenous expression standard gene, GAPDH. PCR primers for Smac/DIABLO and HtrA2 were designed to amplify the common exonic regions of the transcripts (exons 3–6 of Smac/DIABLO and exons 1–4 of HtrA2) to cover all known splicing variants, and primers for XIAP RT-PCR were designed to specifically amplify XIAP transcript but not the XIAP pseudogene. PCR was performed for 32–36 cycles at 95°C (1 min), 56–60°C (0.5 min), and 72°C (1 min) in 1.5 mm MgCl2-containing reaction buffer (PCR buffer II; Perkin-Elmer). For quantitative genomic PCR, the exon 6 region of XAF1, exon 8 region of p53, and intron 5 region of GAPDH were amplified separately with intron-specific primers Xaf1-I1 (sense, 5′-TCACAATTGCAGGGTAAATG-3′) and Xaf1-I2 (antisense, 5′-TAAGCAGAGAACAGAGGCTG-3′), SG85 (sense, 5′-TCCTTACTGCCTCTTGCTTCTCTTT-3′) and SG83 (antisense, 5′-TCTCCTCCACCGCTTCTTGT-3′), and G3 (see above) and G5 (antisense, 5′-GAGTCCTTCCACGATACCAAAG-3′), respectively. Ten microliters of PCR products were resolved on 2% agarose gels. Quantitation was achieved by densitometric scanning of the ethidium bromide-stained gels. Absolute area integrations of the curves representing each specimen were then compared after adjustment for GAPDH. Integration and analysis were performed using the Molecular Analyst software program (Bio-Rad, Hercules, CA). Quantitative PCR was repeated at least three times for each specimen, and the mean was obtained.
Western Blot Assay.
Cells were lysed in a lysis buffer containing 20 mm Tris (pH 7.4), 150 mm NaCl, 1 mm EDTA, 1 mm EGTA, 1% Triton, 2.5 mm sodium phosphate, 1 mm β-glycerolphosphate, 1 mm Na3VO4, 1 μg/ml leupeptin, and 1 mm phenylmethylsulfonyl fluoride. The cell lysate was clarified by centrifugation, and 20 μg of total protein were supplemented with Laemmli buffer and loaded on a 10% SDS-polyacrylamide gel for electrophoresis. XIAP was detected by immunoblot using an anti-XIAP monoclonal antibody (BD Biosciences, San Diego, CA). Antibody binding was detected by enhanced chemiluminescence (Amersham Biosciences, Piscataway, NJ) using a secondary antibody conjugated to horseradish peroxidase. For stripping, the blots were incubated in a stripping buffer [0.2 m glycine (pH 2.2), 0.1% SDS, and 1% Tween 20] at room temperature for 60 min.
To assess reactivation of XAF1 expression, cells were plated in 6-well tissue plates 24 h before treatment. 5-Aza-2′-deoxycytidine (Sigma Chemical Co., St. Louis, MO) was added to the fresh medium at concentrations of 5 μm in duplicate, and cells were harvested after 4 days.
Bisulfite DNA Sequencing.
One microgram of genomic DNA was incubated with 3 m sodium bisulfite (pH 5.0), and 50 ng of bisulfite-modified DNA were subjected to PCR amplification of the XAF1 promoter region using primer sets: MS12 (sense, 5′-GTTTAGGTTGGAGTGTAGTGG-3′) and MS2 (antisense, 5′-CATATTCTACTCTCTACAAAC-3) for nucleotides +3 to −1852 (1855 bp) and MS3 (sense, 5′-TGTTAGTTTTAGGGAGGTAGA-3′) and MS2 (antisense; see above) for nucleotides +3 to −257 (260 bp). The PCR products were cloned into pCRII vectors (Invitrogen, Carlsbad, CA), and 10 clones of each specimen were sequenced by automated fluorescence-based DNA sequencing to determine the methylation status.
LOH was determined using three polymorphic CA markers (D17S796, D17S1832, and D17S1828) localized at chromosome 17p13.1–13.2. PCR amplification was performed on each tumor and normal DNA sample pair obtained from 87 patients using primers 796-S (sense, 5′-CAATGGAACCAAATGTGGTC-3′) and 796-AS (antisense, 5′-AACAACCATTTACTTACTAG-3′) for D17S796, 1832-S (sense, 5′-CTTGACATAGTTGCCCACAG-3′) and 1832-AS (antisense, 5′-CTTTAGGTTTGGATCCAGCC-3′) for D17S1832, and 1828-S (sense, 5′-GCAGGTATACAGCCACACAC-3′) and 1828-AS (antisense, 5′-TGGATTCAGCCATACCTGAA-3′) for D17S1828. Ten microliters of the PCR products were electrophoresed on standard denaturing 8% polyacrylamide gels. If the two alleles appeared in the normal tissue DNA, the patient was considered an informative case for the particular marker. Signal intensity of fragments and the relative ratio of both tumor and normal allele intensities were determined by scanning densitometry. Because certain numbers of noncancerous cells might be present in tumor tissues, LOH was assigned when the intensity ratio of the two tumor alleles differed by at least 50% from that observed on its corresponding normal DNA. The same PCR products were also subjected to nonisotopic SSCP analysis for verification of LOH.
Nonisotopic RT-PCR–SSCP Analysis.
To screen the presence of somatic mutations, RT-PCR–SSCP analysis of XAF1 was performed. The XAF1 transcript was amplified with seven sets of primers that were designed to cover the entire coding region of the gene. Sequences of the primers used will be obtained on request. The PCR products of over 300 bp in lengths were digested with endonuclease(s) to increase the sensitivity of SSCP analysis. Twenty microliters of PCR products were mixed with 10 μl of 0.5 N NaOH, 10 mm EDTA, and 15 μl of denaturing loading buffer (95% formamide, 20 mm EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol). After heating at 95°C for 5 min, samples were loaded in wells precooled to 4°C and run using 8% nondenaturating acrylamide gels containing 10% glycerol at 4–8°C and 18–22°C.
Expression of XAF1, Smac/DIABLO, and HtrA2 in Normal and Benign Tumor Tissues.
To investigate the candidacy of XAF1, Smac/DIABLO, and HtrA2 as a tumor suppressor in gastric tumorigenesis, we initially characterized mRNA expression status of the genes in 20 normal gastric tissues and 16 benign tumors including 3 adenomas, 6 hamartomas, and 7 hyperplastic polyps. For validation of our quantitative PCR approach, serially diluted cDNA was subjected to PCR amplification of XAF1, Smac/DIABLO, HtrA2, and GAPDH over a range of cycles. Linearity of the cDNA dilution experiments demonstrated the ability of our PCR procedure to discriminate the various levels of transcripts (data not shown). As shown in Fig. 1 A, mRNA expression of XAF1, Smac/DIABLO, and HtrA2 was easily detectable in all normal and benign tumor tissues examined, and no significant variations in expression levels were observed among specimens (XAF1/GAPDH, 0.92–1.37; Smac/GAPDH, 0.98–1.42; HtrA2/GAPDH, 0.74–1.06). Quantitative RT-PCR was repeated at least three times for each specimen, and the mean expression level of XAF1, Smac/DIABLO, and HtrA2 in noncancerous tissues was determined as 1.16, 1.22, 0.92, respectively.
Loss of XAF1 mRNA Expression in Gastric Cancer Cell Lines.
We next investigated mRNA expression of XAF1, Smac/DIABLO, and HtrA2 in 15 human gastric carcinoma cell lines. As shown in Fig. 1,B, expressions of Smac/DIABLO and HtrA2 transcripts were easily detectable in all cell lines, and its levels (Smac/DIABLO, 1.00–1.44; HtrA2, 0.76–1.12) were comparable with those of noncancerous tissues. In contrast, the XAF1 transcript was not detected in four cell lines (SNU1, SNU5, SNU484, and SNU719) and was extremely low in two cell lines (MKN45 and MKN74). The mRNA expression status of XAF1 in cancer cell lines was further verified by Northern blot assay (Fig. 1 B). The mean expression level of XAF1 mRNA in gastric cancer cell lines was determined as 0.70, which is significantly low compared with noncancerous tissues (P < 0.0001). Collectively, 40% (6 of 15) of human gastric carcinoma cell lines exhibited loss or significantly reduced levels of XAF1 mRNA expression.
Frequent Reduction of XAF1 mRNA Expression in Primary Gastric Cancers.
Next, we evaluated the expression of XAF1, Smac/DIABLO, and HtrA2 in 87 primary gastric carcinomas including 20 matched sets. Whereas Smac/DIABLO and HtrA2 transcripts were expressed at similar levels in all tumor tissues examined, marked reductions of XAF1 expression were identified in a substantial fraction of tumors (Fig. 1,C). Moreover, tumor-specific reduction of XAF1 was detected in 9 of 20 (45%) matched sets from the same patients. As shown in Fig. 2,A, XAF1, Smac/DIABLO, and HtrA2 transcript levels in primary carcinomas were determined in the ranges of 0.30–1.39, 0.94–1.48, and 0.66–1.17, respectively. Unlike Smac/DIABLO and HtrA2, expression levels of XAF1 in primary carcinomas were significantly low compared with noncancerous tissues (P < 0.0001). We arbitrarily set expression levels less than a half (XAF1, <0.58; Smac/DIABLO, <0.61; HtrA2, <0.46) of normal means as abnormally low. On this basis, 23% (20 of 87) of primary tumors were classified as abnormally low XAF1 expressors whereas none of the tumors were identified as abnormal expressors of Smac/DIABLO and HtrA2. Furthermore, loss or abnormal reduction of XAF1 was significantly high in advanced tumors (17 of 62, 27%) compared with early stage tumors (3 of 25, 12%; P = 0.005) and more frequent in poorly differentiated tumors (14 of 50, 28%) than well or moderately differentiated tumors [16% (5 of 31) and 17% (1 of 6), respectively; P = 0.043; Fig. 2 B]. However, XAF1 alteration was not associated with histological types of tumors [diffused, 25% (13 of 52); intestinal, 20% (7 of 35)]. Expression levels of Smac/DIABLO and HtrA2 showed no correlation with histopathological characteristics of tumors (data not shown). Collectively, our results indicate that loss or abnormal reduction of XAF1 is a frequent event in gastric tumorigenesis and may contribute to the malignant progression of human gastric cancers.
No Alteration of XIAP Expression in Primary Gastric Carcinomas and Cell Lines.
It was reported previously that XIAP mRNA levels are relatively high in the majority of cancer cell lines, suggesting that high expression of XIAP mRNA coupled with low expression of XAF1 mRNA is a very common characteristic in cancer cells and may provide a survival advantage through the relative increase of XIAP antiapoptotic function (10). This prompted us to investigate whether abnormal overexpression of XIAP is implicated in gastric tumorigenesis. We first analyzed mRNA levels of XIAP in 15 gastric cancer cell lines, but no detectable variation was recognized among specimens (Fig. 3). Moreover, XIAP mRNA levels in cancer cell lines (1.20–1.66) and primary carcinomas (1.18–1.55) were not significantly high compared with normal and benign tumor tissues (1.12–1.48). To explore the possibility that XIAP is regulated at the posttranscriptional level, XIAP protein levels were analyzed in cancer cell lines using Western blot assay. Consistent with mRNA expression, all of the 15 gastric cell lines exhibited similar levels of XIAP protein, suggesting that overexpression of XIAP might not be a predominant mechanism for the relative increase of XIAP activity in human gastric cancers.
Absence of Allelic Deletion of the XAF1 Gene in Gastric Carcinomas.
The 17p13 region, where the XAF1 gene is located, undergoes frequent allelic losses in a variety of human malignancies, including gastric cancer (20). Recent microsatellite analysis using the NCI 60 cell line panel revealed significantly decreased heterozygosity at the XAF1 locus, suggesting that allelic loss of XAF1 is prevalent in cancer cell lines (10). To elicit whether loss or abnormal reduction of XAF1 mRNA expression in gastric cancers is associated with allelic deletion of the gene, we examined the XAF1 gene level using quantitative genomic PCR. XAF1 is located at the 17p13.2, approximately 3 cM to the p53 tumor suppressor. Consistent with the previously reported allelic status of p53, absence or low gene levels of p53 were detected in nine cell lines harboring homozygous deletion or LOH of p53 (Fig. 4 A; Refs. 21 and 22). However, in contrast to p53, none of the 15 cell lines showed detectable reduction of the XAF1 gene level. Moreover, whereas 22 of 87 (25.3%) primary tumors showed marked reduction of the p53 gene level, none of the tumors, including 20 abnormal mRNA expressors of XAF1, exhibited reduction of the XAF1 gene. These results, thus, suggest that genomic deletion of XAF1 might be infrequent and not associated with abnormal down-regulation of XAF1 mRNA in human gastric cancers.
To define further the allelic status of the XAF1 gene, we surveyed 48 tumors for LOH of the 17p13.1–13.2 region using three polymorphic markers: D17S796, D17S1832, and D17S1828. Among 48 matched sets tested, 12 (25.0%) were informative at the centromeric marker D17S796. LOH at D17S796, which is located approximately 2.5 cM telomeric of the p53 locus, was observed in 4 of 12 (33.3%) informative cases (Fig. 4,B). In contrast, all of the 11 tumors that are informative for at least one of two telomeric markers (D17S1832 and D17S1828) were found to retain heterozygosity. Consistent with these results, our quantitative PCR analysis of three marker DNAs demonstrated that the four LOH tumors have relatively low genomic levels of D17S796, compared with the eight retention of heterozygosity tumors, and all tumor specimens examined have normal genomic levels of D17S1832 and D17S1828 (Fig. 4 B). Interestingly, however, all four LOH tumors were identified to have normal genomic and expression levels of XAF1, whereas two of four tumors exhibited low genomic and mRNA levels of p53, suggesting that LOH at D17S796 might be associated with allelic deletion of p53 but not with XAF1. Collectively, these observations showed that allelic deletion at the centromeric region of the XAF1 locus occurs in a subset of gastric cancers but might rarely extend into the XAF1 gene.
Aberrant Hypermethylation at the CpG Sites in the XAF1 Promoter.
To investigate whether altered expression of XAF1 is associated with promoter hypermethylation of the gene, 15 gastric cell lines were treated with the demethylating agent 5-aza-2′-deoxycytidine. As shown in Fig. 5,A, XAF1 mRNA expression was reactivated in all four nonexpressor cell lines (SNU1, SNU5, SNU484, and SNU719) and significantly increased in two low expressor cell lines (MKN45 and MKN74), indicating that XAF1 is transcriptionally silenced in these cells by DNA hypermethylation. To explore the relationship between aberrant CpG methylation and gene silencing, we performed bisulfite DNA sequencing analysis of the XAF1 gene promoter. The 5′ upstream region of the XAF1 gene is not highly enriched in CpGs, and only two short regions (regions I and II in Fig. 5,B), which encompass only 200 bp (region I, nucleotides −1248 to −1447; obs/exp CpG, 0.60; 51.9% G+C) and 380 bp (region II, nucleotides −1558 to −1937; obs/exp CpG, 0.68; 50.3% G+C), satisfy the formal criteria for CpG islands (23). Thus, we analyzed the methylation status of 34 CpGs including 19 CpGs located in regions I and II and 14 CpGs located in the 5′ proximal region (nucleotides −23 to −692). The sequence region (nucleotides −23 to −1831) spanning these 34 CpG sites was amplified by PCR using sodium bisulfite-modified DNA as templates, and 10 PCR clones were sequenced to determine methylation frequency at individual CpG sites (complete methylation, 70–100%; partial methylation, 10–60%; unmethylation, 0%; Fig. 5,C). As shown in Fig. 5,D, approximately 82–88% (28–30 sites) and 71–77% (24–26 sites) of the 34 CpGs were partially or completely methylated in four nonexpressor and two low expressor cell lines, respectively, whereas only 18–44% (6–15 sites) of the CpGs were methylated in nine normal expressor cell lines. In addition, completely methylated CpG sites were more frequently observed in four nonexpressors (66 of 136, 48%) than two low expressors (10 of 68, 15%), indicating that methylation extent correlates with mRNA expression status. The methylation rate of the 5′ proximal region (nucleotides −23 to −692) was approximately 4-fold higher in abnormal expressors (88.9%, 80 of 90) compared with normal expressors (23.0%, 31 of 135), whereas the regions I and II in abnormal expressors showed only 1.8-fold higher methylation compared with normal expressors (75.4%, 86 of 114 versus 41.5%, 71 of 171). Intriguingly, the methylation status of seven CpGs (numbers 1–7 in Fig. 5,B) within nucleotides −23 to −234 was most tightly associated with mRNA expression. Whereas all of the seven CpGs were methylated in six abnormal expressor cell lines, none of these CpGs was methylated in nine normal expressor cell lines, suggesting that hypermethylation of CpG sites located in the 5′ proximal region might be critical for the transcriptional silencing of XAF1. A tight correlation of gene silencing with methylation of the seven CpGs was also recognized in primary tumors. Eighteen of 20 (90%) carcinoma tissues with abnormal expression showed methylation of the seven CpGs, whereas none of the adjacent normal tissues exhibited methylation (Fig. 6). The seven CpGs were unmethylated in 15 primary tumors with normal expression and five noncancerous tissues included for comparison (data not shown).
Absence of XAF1 Mutations in Gastric Cancers.
To evaluate the mutational status of XAF1, we performed RT-PCR–SSCP analysis for 11 XAF1-expressing cell lines and 45 primary carcinomas. The entire coding region of XAF1 transcript was amplified using seven sets of exon-specific primers. To improve mutation detection sensitivity, the same RT-PCR products were digested using a different restriction endonuclease(s) and SSCP was performed using two different running conditions. However, we failed to find any types of mutation leading to amino acid substitutions or frameshifts in the XAF1 transcripts expressed, although 33% (15 of 45) of the same set of primary tumors and 67% (10 of 15) of cancer cell lines were found to carry homozygous deletions or mutations of p53, indicating that somatic mutation of XAF1 is infrequent in gastric cancers.
Inverse Correlation between Down-Regulation of XAF1 and p53 Mutations.
Recent studies demonstrated that methylation of p14ARF is associated with the absence of p53 mutations in human colon cancer and p53 acts as an upstream controller of the DNA methylation apparatus, raising the possibility that hypermethylation in cancer cells could be associated with the mutational status of p53 (24, 25). In this context, we explored whether epigenetic silencing of XAF1 is associated with p53 alterations in gastric cancer. Mutation analysis of p53 revealed that 10 (SNU5, SNU16, SNU216, SNU484, SNU601, SNU620, SNU638, MKN1, MKN28, and KATOIII) of 15 cell lines and 15 of 45 primary cancers harbor homozygous deletions or mutations of p53. Interestingly, loss or down-regulation of XAF1 expression was found in 4 (SNU1, SNU719, MKN45, and MKN74) of 5 (80%) cell lines with wild-type p53 but only in 2 (SNU5 and SNU484) of 10 (20%) cell lines harboring homozygous deletions or mutations of p53. Similarly, whereas 7 of 9 (78%) primary tumors with altered expression of XAF1 carried wild-type p53, 13 of 15 (87%) primary tumors with mutant p53 showed normal expression of XAF1, suggesting that epigenetic silencing of XAF1 by hypermethylation and mutational alteration of p53 might be mutually exclusive events in human gastric tumorigenesis.
In the present study, we demonstrate first that a substantial fraction of gastric cancer cell lines and primary carcinomas express no or extremely low levels of XAF1 transcript whereas two other IAP antagonists, Smac/DIABLO and HtrA2, are normally expressed in all cancer specimens examined. Loss or down-regulation of XAF1 expression was strongly associated with aberrant CpG methylation in the promoter region. Moreover, abnormal expression of XAF1 correlated with advanced stage and high grade of tumors, suggesting that epigenetic silencing of XAF1 because of promoter hypermethylation might contribute to the malignant progression of human gastric cancers.
Despite the possible role for XAF1 in the suppression of malignancy, the mechanism by which XAF1 expression is down-regulated in human cancers remains to be characterized. Recent microsatellite analysis using the NCI 60 cell line panel revealed significantly decreased heterozygosity within the XAF1 region at 17p13.2 (10). Of the 58 cell lines tested, 22 were shown to be homozygous at all three polymorphic markers tested, suggesting that allelic loss of the XAF1 gene might be prevalent in human cancers. However, no matched controls for these cell lines were available for LOH assay and none of the cell lines tested showed any gross rearrangements of the XAF1 gene by Southern blot analysis. Therefore, the question remains as to whether loss or reduction of XAF1 mRNA expression in human cancer cells is caused by other transcriptional regulatory mechanisms rather than homozygous or allelic deletion of the gene. In the present study, our quantitative genomic PCR and LOH analyses indicated that allelic deletion at the XAF locus rarely extends into the XAF1 gene. Interestingly, we found that XAF1 expression is reactivated in all nonexpressor cell lines and significantly elevated in low expressor cell lines after 5′-aza-2-deoxycytine treatment, suggesting that loss or down-regulation of XAF1 expression in cancer cells might be caused by an epigenetic mechanism such as aberrant DNA methylation. Our bisulfite DNA sequencing analysis revealed that methylation extent in the promoter region is closely associated with the mRNA expression status of the XAF1 gene in both cancer cell lines and tumor tissues. Aberrant CpG methylation was also found in many different types of human cell lines including lung, bladder, kidney, and prostate cancers, indicating that transcriptional silencing of XAF1 by promoter hypermethylation is not a gastric cancer-specific phenomenon (data not shown).
It has been demonstrated extensively that hypermethylation in CpG-rich promoter or exonic region is strongly associated with transcriptional silencing. CpG islands are more methylated in cancers compared with the non-CpG island region, and hypermethylation at CpG islands in the transcription regulatory region is a critical event leading to the epigenetic inactivation of tumor suppressor genes in human tumorigenesis (26). In this context, it should be noted that the 5′ upstream region of the XAF1 gene is not highly enriched in CpGs. Within the 5-kb upstream region, only two short regions (regions I and II in Fig. 5 B; 200 bp; obs/exp CpG, 0.60; 51.9% G+C; and 380 bp; obs/exp CpG, 0.68; 50.3% G+C, respectively) were identified to satisfy the formal criteria for CpG islands (23). However, these regions (−1867 to −1343; 54.4% G+C; obs/exp CpG, 0.60) do not fulfill the more rigorous description of a CpG island (≥500 bp; G+C, >55%; obs/exp CpG, >0.65) proposed by Takai and Jones (27). In addition, in cancer cells lacking XAF1 expression, the 5′ upstream region (nucleotides −23 to −692) are more significantly methylated compared with regions I and II, and aberrant methylation of the seven CpG sites located in the 5′ proximal region (nucleotides −23 to −234) most tightly correlate with gene silencing. This observation leads to the conjecture that hypermethylation of the regions I and II does not act so critical for gene silencing as CpG island hypermethylation but has an inhibitory effect on XAF1 transcription. This is in line with a recent study demonstrating that silencing of caspase 8 expression in cancer cells is related to hypermethylation of the 5′ non-CpG island region of the gene (28). Collectively, our findings suggest that hypermethylation of the 5′ non-CpG island region of the gene is critical for the transcriptional silencing of the XAF1 gene in human gastric cancers.
It was reported previously that XIAP mRNA levels are relatively high in the majority of cancer cell lines, raising the hypothesis that the relative increase of XIAP to XAF1 expression in cancer cells may contribute to the development of the transformed phenotype by suppressing apoptotic signaling (10). In this study, however, we could not observe any detectable increase of XIAP mRNA expression in gastric cancer cell lines and primary tumors compared with normal tissues. Western blot analysis of XIAP in 15 gastric cancer cell lines also showed that protein levels of XIAP are not variable among specimens and are comparable with mRNA expression patterns. This observation, thus, suggest that overexpression of XIAP mRNA is not a frequent event and may not be implicated in the relative increase of XIAP to XAF1 in human gastric cancers.
Very recently, XAF1 was identified as a novel IFN-stimulated gene that contributes to IFN-β-dependent sensitization of cells to tumor necrosis factor-related apoptosis-inducing ligand-induced apoptosis (29). XAF1 mRNA was up-regulated by IFN-α and -β, and high levels of XAF1 protein were induced predominantly in cell lines sensitive to the proapoptotic effects of IFN-β. It was also observed that A375 melanoma cells expressing XAF1 constitutively are more sensitive to tumor necrosis factor-related apoptosis-inducing ligand-induced apoptosis compared with empty vector-transfected cells and the degree of sensitization by XAF1 is correlated with the level of XAF1 expressed. Previous studies demonstrated that IFN-α, -β, and -γ induce apoptosis and/or growth arrest of gastric cancer cells and a combination of IFN-α or -β with 5-fluorouracil resulted in an additive or synergistic effect against clinical gastric carcinomas (30, 31). However, some gastric cancer cells are resistant to IFN-induced growth arrest and apoptosis, leading to the conjecture that a defective response to growth suppression effect of IFNs may give the tumors a selective advantage and abet escape from T-cell antitumor response. In this context, it could be suspected that gastric cancers with loss or abnormal reduction of XAF1 might be more resistant to IFN therapy than cancers with normal XAF1 expression. On this basis, it will be valuable to examine that expression status of XAF1 could be a clinically useful marker for cancer treatment including IFN therapy.
In conclusion, our data presented here clearly demonstrate that XAF1 undergoes epigenetic silencing in a considerable proportion of gastric cancer cell lines and primary carcinomas by aberrant CpG site hypermethylation of the gene promoter. Loss or abnormal reduction of XAF1 mRNA expression showed a strong correlation with tumor stage and grade, suggesting that XAF1 inactivation might contribute to the malignant progression of human gastric cancers. Although additional studies are required to characterize the biological significance of XAF1 inactivation in gastric tumorigenesis, our study suggests that aberrant hypermethylation of the XAF1 gene could be a molecular marker for detection and treatment of human gastric cancers.
Supported by a grant (R02-2001-000-00052-0) from the Basic Research Program of the Korea Science and Engineering Foundation and an intramural grant-in-aid from the Kyung Hee University (2002), Republic of Korea.
The abbreviations used are: IAP, inhibitor of apoptosis; BIR, baculovirus IAP repeat; XIAP, X-linked IAP; LOH, loss of heterozygosity; SSCP, single-strand conformation polymorphism; obs/exp CpG, observed/expected CpG; XAF1, XIAP-associated factor 1; GAPDH, glyceraldehyde-3-phosphate dehydrogenase.