Monochromosome Transfer and Microarray Analysis Identify a Critical Tumor-Suppressive Region Mapping to Chromosome 13q14 and THSD1 in Esophageal Carcinoma

Esophageal cancer is geographically diverse, with only a 10.7% 5-year survival rate (1). Esophageal squamous cell carcinoma (ESCC) is the major histologic form. ESCC molecular pathogenesis still remains poorly understood.

Chromosome 13q deletions are frequent events in several human cancers, including ESCC (2, 3), nasopharyngeal (4), and lung (5) cancers. This current study is the first functional study of the tumor-suppressive role of chromosome 13 in ESCC and is initiated by the high-frequency ESCC chromosome 13q losses detected by comparative genomic hybridization (3) and loss of heterozygosity (LOH) studies (2, 6-9). Comparative genomic hybridization analysis showed 100% losses on chromosome 13q in 17 ESCC cases (3). The extremely high 13q loss was independently verified in genome-wide LOH studies (2), with 95% of 77 13q markers showing LOH. Thus, functional inactivation of TSGs on chromosome 13q is likely key to ESCC development.

Monochromosome transfer into tumorigenic cell lines allows functional complementation of existing defects and study of tumor-suppressive effects driven by native endogenous regulatory environments, with control of single gene copy number gains more closely mimicking normal physiologic levels. Our previous studies identified tumor-suppressive regions localized to 3p14 (10), 9q33-34 (11), and 14q32 (12) in the ESCC cell line SLMT-1. This present study examines the tumor-suppressive role of chromosome 13 in SLMT-1. Comparative differential gene expression observed after competitive hybridization in cDNA oligonucleotide microarrays of the tumorigenic parental cell line and three tumor-suppressing chromosome 13 microcell hybrids (MCH) identified novel candidate TSGs. Thrombospondin type I domain-containing 1 (THSD1), at 13q14.3, showed 100% down-regulation in a panel of 18 ESCC and other cancer cell lines. Transfection of wild-type THSD1 into SLMT-1 significantly reduced colony formation, providing functional evidence for growth-suppressive activity. The current data indicate that the mechanism for THSD1 down-regulation in ESCC involved both LOH and epigenetic silencing. Demethylation treatment restored THSD1 expression in THSD1 down-regulated ESCC cell lines; results from methylation-specific PCR (MSP) analysis of ESCC cell lines and bisulfite sequencing of the promoter region of THSD1 in both cell lines and primary tissues showed that loss of THSD1 expression could be partially attributed to hypermethylation in ESCC.

Fifty markers were used for allelotyping ESCC SLMT-1 and human chromosome 13 donor cell line, MCH204.3 (Fig. 1). SLMT-1 contained two regions of contiguous homozygosity, one with 10 markers at 13q12.11-13q12.3 and another with 27 markers at the 13q13.3 region to the telomeric end of chromosome 13q. The two consecutive homozygous regions only contain a single allele in each of the 37 loci studied. The random chance for this occurring in a diploid genome is small. These results suggest nearly a complete deletion of a single copy of chromosome 13 in SLMT-1 cells. The loss of 13q was not only observed molecularly by allelotyping but also by comparative genomic hybridization analysis (data not shown). Loss of one copy of nearly the entire chromosome 13 in SLMT-1 strongly suggested the presence of TSGs, based on the assumption that the relevant genes in the remaining allele would also be inactivated.

Fluorescence in situ hybridization (FISH) of SLMT-1 using human chromosome 13 WCP probe showed one intact signal of chromosome 13 and six signals representing translocations to other chromosomes (Supplementary Fig. S1). None of the morphologies of chromosome 13 detected in the metaphase spreads appeared normal. Using microcell-mediated chromosome transfer, eight chromosome 13 MCH cell lines were obtained. Screening by DNA slot blot hybridization confirmed that five MCHs were mouse DNA-free (data not shown).

The distinctive acrocentric pattern of the transferred exogenous chromosome 13 in the metaphases of chromosome 13 MCHs enabled its precise identification. Whole chromosome 13 FISH analysis verified successful exogenous chromosome 13 transfer into MCH13-111, MCH13-113, and MCH13-117 (Supplementary Fig. S1).

PCR microsatellite typing confirmed the presence of donor and recipient alleles, validating the successful transfer of chromosome 13 into SLMT-1 in three MCH13 cell lines. Of 50 microsatellite markers, 37 were informative. An ideogram summarizes the genotyping results (Fig. 1).

Statistically significant delays in tumor growth kinetics and reduced tumor sizes were observed with all three chromosome 13 hybrids compared with the parental cell line (Fig. 2A). The tumorigenic potentials of MCH13-111 and MCH13-113 differed significantly from the SLMT-1 recipient cell line (P < 0.001). For MCH13-117, only three of six sites formed tumors with significantly longer latency periods of 2 to 8 weeks (P = 0.001). MCH13-111 formed tumors in four of six sites with latency periods of 3 to 5 weeks. MCH13-113, with deletion of 13q22.1-qter, showed uniformly smaller tumors and latency periods of 7 to 18 weeks. This clone showed that transfer of chromosome 13q, missing the 13q22.1 region around D13S156, 13q31.1 region around D13S170, and 13q32.1-qter region, retains its tumor-suppressive effects.

Reconstitution of mouse tumors into cell culture established eight tumor segregant (TS) cell lines. Thirty-seven microsatellite markers spanning the whole chromosome 13 detected TS deletions in four critical regions (CR; Fig. 1). CR1 displayed 87% and 75% exogenous and endogenous allelic loss, respectively. This 2.34-Mb CR is bounded by D13S217 and D13S1287. The 0.373-Mb CR2 bordered by D13S893 and D13S260 is lost in half of the TSs. CR3 is ∼5.99 Mb mapping between D13S218 and D13S757; specific exogenous loss at D13S263 was detected in 50% of TS cell lines. Representative loss of D13S133 in CR4 in MCH13-111/TS2R is shown in Fig. 2B. CR4 is ∼0.979 Mb between D13S1325 and D13S137; specific exogenous loss at D13S133 occurs in 50% of TSs.

Nude mouse assays of two TS cell lines carrying different deletions were carried out to assess their functional consequences on tumorigenicity. MCH13-111/TS2R, which exhibited the most extensive loss of the transferred chromosome 13, displayed the highest tumorigenicity (Fig. 2A). Both of the tumor segregants, MCH13-111/TS2R and MCH13-117/TS1L, showed a higher tumorigenicity compared with their matched MCHs (Fig. 2A), implying that multiple TSGs may reside on chromosome 13q, which contribute to tumorigenesis.

Successful chromosome 13 transfer into SLMT-1 resulted in tumor suppression; presumably intact chromosome 13 candidate TSG(s) are restored in those MCHs. Thus, whole genome differential gene expression of the three tumor-suppressing MCHs was compared with the tumorigenic SLMT-1 by gene profiling on 19k oligonucleotide microarrays. Implicated candidate TSGs show down-regulated gene expression in cancer cells compared with normal cells. Hence, the gene expression profile of SLMT-1 versus NE1, a reference immortalized esophageal epithelial cell line, was also examined by competitive microarray hybridization. Genes down-regulated in SLMT-1 compared with the immortalized normal esophageal cell line, as well as up-regulated in the three-suppressing hybrids, were shortlisted as chromosome 13 candidate TSGs. The 19k chip contains 248 chromosome 13 oligonucleotide probes representing ∼44.6% of the total possible 556 chromosome 13 genes (National Center for Biotechnology Information Map Viewer, Homo sapiens Build 36.1).⁸

A list of the remaining 40 genes on other chromosomes, which may be involved in the downstream cascades of tumor suppression in the nude mouse model, is arranged according to their gene ontology terms (Table 1B). The categories of genes involved include cell proliferation (CDKN1A, NFKBIA, RB1, RFP2), cell death regulator (TNFAIP3), transcription factor (JUN, FOXD1, RUNX2), signal transduction (ST7, CHRNE, TORC3), extracellular matrix (TFPI2), cell adhesion (LAMC2, ITGA5, L1CAM), cytoskeleton (GAF1, ENC1, PLEKHC1), immune response (DAF, IL1B, IFIT5), heat shock proteins (DNAJB9, HSPA5, HSPCAL3), calcium and metal ion binding (FSTL1, CAMLG, ZSWIM6, PJA2), and others.

The differential gene expression of eight 13q14 candidate TSGs in NE1, SLMT-1, and the three tumor-suppressing chromosome 13 MCHs was verified by reverse transcription-PCR (RT-PCR; Fig. 3A). An additional 15 ESCC, cervical cancer (HeLa), and lung cancer (A549) cell lines show the general importance of these candidate TSGs (Fig. 3A). All eight genes were up-regulated in chromosome 13 MCHs compared with SLMT-1 and down-regulated in SLMT-1 compared with NE1, as observed by the microarray analysis. The most promising candidate was THSD1, which has two transcript variants. All 18 cell lines examined showed either moderate to severe down-regulation or absence of gene expression compared with NE1. PHF11 was another important gene detected with 100% down-regulation in 18 cell lines with severe down-regulation or absence of gene expression in 10 of 18 (55.6%) and slight to moderate down-regulation in 8 of 18 (44.4%). For KIAA0853 and CHC1L, down-regulation or absence of gene expression was observed in 11 of 18 (61.1%) and 9 of 18 (50.0%), respectively. A low frequency of severe loss of expression was detected for the remaining candidates ESD (0 of 18), RFP2 (2 of 18), FLJ11712 (0 of 18), and C13orf9 (0 of 18).

The down-regulation fold changes in THSD1 and PHF11 were most significant and were further quantified by real-time PCR; both showed 100% down-regulation compared with NE1. Results were concordant with RT-PCR results (Fig. 3B).

The consistent down-regulation of THSD1 in all 18 cancer cell lines and up-regulation in three tumor-suppressive hybrids strongly suggested THSD1 as a good candidate TSG. The full-length wild-type cDNAs of both THSD1 variants were transfected into SLMT-1 to obtain functional evidence of their effect on growth suppression. RT-PCR showed that the THSD1 was overexpressed 1 day after transient transfection (Fig. 4A). Compared with the pCR3.1 vector-alone transfectants, a significant reduction of colony numbers was observed in THSD1 transfected cells (Fig. 4B). There was a 94% and 93% reduction of colonies with THSD1 v1 and v2, respectively (P < 0.001).

Based on the LOH and comparative genomic hybridization data from ESCC studies, one copy of chromosome 13 may be lost in ESCC cell lines. To investigate this possibility, microsatellite typing of NE1, EC18, and KYSE180 with 11 markers spanning 24.715 Mb from D13S757 (43.921 Mb) to D13S275 (68.636 Mb) on chromosome 13 was done. NE1, which served as the control cell line, was homozygous in 2 (D13S887 and D13S1320) of 11 markers (Fig. 1). The result was typical for cell lines carrying two copies of chromosome 13, because the microsatellite markers are highly polymorphic. However, all 11 markers (∼24.715 Mb) were consecutively homozygous for EC18. Except for the telomeric D13S275 marker, 10 of 11 markers (∼18.781 Mb) were also consecutively homozygous for KYSE180 (Fig. 1). These results imply that at least a large portion of one of the two copies of chromosome 13, which spans the THSD1 region at 51.849 Mb, was lost in EC18 and KYSE180.

The THSD1 gene expression levels in ESCC cell lines were down-regulated up to 600-fold (Fig. 3B). In addition to LOH, the reduced THSD1 gene expression was hypothesized to be due to epigenetic modification of the THSD1 gene by promoter hypermethylation and/or histone deacetylation. To ascertain whether promoter hypermethylation is one of the mechanisms for THSD1 silencing, MSP analysis of the THSD1 promoter region was done in six ESCC cell lines versus an immortalized esophageal normal epithelial cell line, NE1. Only the unmethylated allele was detected in the control NE1, whereas both the methylated and unmethylated alleles were detected in all six ESCC cell lines (Fig. 4C). Additionally, clonal bisulfite sequencing of NE1, KYSE140, KYSE180, and two pairs of matched nontumor and ESCC tumor tissues (2N/T and 80N/T) was done. The methylation status of 12 CpG sites in the THSD1 promoter obtained by bisulfite sequencing is summarized in Fig. 4D. The results show that NE1 does not contain methylated CpG sequences, whereas both KYSE140 and KYSE180 cell lines contain aberrant methylated CpG sites. In the ESCC 80N nontumor tissue, no methylated sites were detected; the ESCC tumor 80T tissue showed heterogeneous methylation. Heterogeneous methylation was also observed with both the 2N nontumorous tissue and the 2T tumor specimens, but the level of methylation in the tumor tissue was very extensive compared with the normal tissue. The two paired normal and ESCC tumor specimens (80N/T and 2N/T), which showed down-regulation of THSD1 by RT-PCR (data not shown), were included for bisulfite sequencing analysis to rule out any possibility of cell line artifacts. The MSP and bisulfite sequencing data show that the THSD1 promoter is partially and heterogeneously hypermethylated in ESCC cell lines and primary tumor tissues.

Further evidence of promoter methylation regulation of THSD1 gene expression was shown by a substantial reactivation of THSD1 expression in EC18 and KYSE180 with a demethylation drug, 5-aza-2′-deoxycytidine (Fig. 4E). With the treatment of both 5-aza-2′-deoxycytidine and trichostatin A, the THSD1 (variant 1) gene expression was fully reactivated to the level of NE1, as shown by quantitative RT-PCR (Fig. 4F). The significance of other indirect transcriptional activation and epigenetic modifications such as histone deacetylation in silencing of THSD1 was shown by the synergistic effect of restoration of THSD1 expression in the presence of both 5-aza-2′-deoxycytidine and trichostatin A, an inhibitor of histone deacetylase (Fig. 4E and F).

Introduction of a single intact human chromosome 13q via microcell-mediated chromosome transfer resulted in a marked decrease in tumorigenicity of SLMT-1. Small tumors arose in the mice only after a delayed latency period. Cell lines established from these tumor segregants, when reinjected into mice, showed an increased tumorigenic potential (Fig. 2A). To localize the tumor-suppressive activities, microsatellite deletion mapping was done to identify critical regions of nonrandom losses in the TSs derived from the suppressive MCHs. CRs were localized at 13q12 and 13q14.

The TSs showing endogenous loss at 13q12.3 at D13S1299, D13S1229, and D13S1226; 13q13.1 at D13S260; and 13q13.2 at D13S267 are colocalized with the nonhomozygosity regions of chromosome 13q in SLMT-1 at 13q12.3-13.3 (Fig. 1). The reason for favoring endogenous loss at 13q12.3-13q13.2 is unknown and may be related to a gene dosage effect (13). Of interest, a 0.373 Mb CR2, mapped to 13q12.3, was colocalized to the same CR identified in a NPC model system in a chromosome 13 transfer study (14). CR1 and CR2 are both located at 13q12. LOH at 13q12 occurs frequently in ESCC (6-8) and is associated with lymph node metastasis (15). BRCA2 is a candidate TSG at 13q12-13, but both the microarray gene expression profile analysis and TS deletion analysis did not support a role for BRCA2 in SLMT-1 tumorigenesis. Unknown TSG(s) within 13q12.3 remain to be identified.

RB1 is a well-known TSG mapped to 13q14, but no studies of RB1 cDNA transfer in ESCC and mutation of RB1 in primary ESCC have been reported. In contrast to the high LOH reported at the RB1 locus (16), RB1 protein expression was absent in only a small subset (6.4%, 11 of 172) of ESCC patients by immunohistochemical staining (17). RB1 is outside the 13q14 LOH region mapped by Li et al. (9). Thus, these studies suggest that unknown novel TSG(s), other than RB1, are responsible for the high allelic loss at 13q14 and remain to be identified for ESCC tumorigenesis.

The present study used the SLMT-1 cell line, which has a reduced level of the 110 kDa RB1 protein expression (Supplementary Fig. S2). From the microarray analysis, the three tumor-suppressing MCH cell lines showed an average of 1.89-fold increase in RB1 expression (Table 1). Western blot analysis of the three MCHs showed slight increases in total RB protein. However, when compared with three hTERT immortalized esophageal epithelial cell lines, the normal active form of RB is not increased in the MCHs (Supplementary Fig. S2). In addition to RB1 protein expression analysis in SLMT-1, the RB1 marker, D13S153, located at exon 2 of RB1, was not deleted in any of the eight TSs. The complete loss of p16^INK4a in SLMT-1 (18), together with the current deletion analysis of TSs by PCR-based microsatellite typing and RB1 Western blotting of chromosome 13 MCHs, provide evidence that RB1 is not the target TSG at 13q14 responsible for the tumor suppression observed in SLMT-1, the current ESCC model system.

The current functional studies implicate the importance of chromosome 13q14 in ESCC tumor suppression. TS microsatellite deletion mapping analysis localized two CRs (CR3 at D13S263 and CR4 at D13S133) at chromosomal region 13q14, which are frequently eliminated. Microarray analysis of differential gene expression identified nine 13q14 candidates genes up-regulated in the three tumor-suppressing chromosome 13 MCHs. RT-PCR was done for KIAA0853, ESD, RB1, CHC1L, PHF11, RFP2, FLJ11712, THSD1, and C13orf9. Results validated the frequent down-regulation of THSD1 and PHF11 in a panel of 18 cancer cell lines. THSD1 is located between FLJ11712 and C13orf9 (Fig. 1) within 13q14.3, but only specific loss of THSD1 expression in all cancer cell lines was detected. Because THSD1 showed a more prominent loss than that of PHF11, it was chosen for further functional characterization in this current study. PHF11 remains another interesting target for further investigations in a future study. The wild-type THSD1 transfection in SLMT-1 resulted in significant reduction of colony formation ability, providing evidence for a growth-suppressive role of THSD1 in ESCC tumorigenesis. In six ESCC cell lines showing 50-fold to 600-fold down-regulated THSD1 expression compared with NE1, both methylated and unmethylated alleles were observed (Fig. 4C). Additionally, demethylation treatment increased THSD1 expression in EC18 and KYSE180 (Fig. 4E and F), further indicating that hypermethylation is involved in THSD1 down-regulation. Promoter methylation may be one of the mechanisms responsible, at least in part, for the loss of THSD1 expression in ESCC tumorigenesis. For KYSE140 and 2T, this seems to be an important inactivation mechanism. For KYSE180 and 80T, which do not show such extensive methylation, both epigenetic silencing and LOH are likely involved in the inactivation of THSD1. Clonal bisulfite sequencing results with cell lines (Fig. 4D) indicate that the normal NE1 does not contain methylated sequences, whereas KYSE180 contains 10% to 50% partial methylated sequences and KYSE140 contains more extensive, 80% to 100%, methylated sequences. The partial and extensive methylation in KYSE180 and KYSE140, respectively, detected by bisulfite sequencing (Fig. 4D), were concordant with the observation of a much more intense band of the M allele detected by MSP analysis for KYSE140 versus KYSE180 (Fig. 4C). The presence of both the methylated and unmethylated alleles in KYSE 180 detected in MSP analysis and partial methylated CpG sites observed by bisulfite sequencing reflect heterogeneous methylation or a mix of cells carrying methylated or unmethylated alleles, with one allele lost by LOH (Fig. 1). Comparative examination of tumor tissues and their adjacent nontumor margins showed greater hypermethylation in tumors. None of the CpG sites were methylated in the 80N specimen, but its tumor tissue showed 10% to 30% partial methylation. In addition to partial methylation, THSD1 inactivation in 80N/T likely involves LOH, because LOH was detected in the DNA from the same pair of tissues (80N/T) using two microsatellite markers in close proximity to THSD1 (D13S284 and D13S133, data not shown). Although 10% to 30% methylation was observed with the 2N specimen obtained from the nontumorous margin of the esophageal tissue resection, the tumor tissue 2T showed an increased methylation intensity of 20% to 60%. Previous investigators have also reported a lower extent of methylation of nontumorous tissues compared with tumor specimens (19). The results of this study indicate that the THSD1 promoter is partially and heterogeneously hypermethylated. Further studies are required to determine whether full restoration of THSD1 expression to the immortalized normal epithelial NE1 levels after trichostatin A treatment can be attributed to other epigenetic events such as histone deacetylation. Such a synergism between 5-aza-2′-deoxycytidine and trichostatin A has been shown for many genes. Methylated promoters may recruit methyl CpG-binding proteins (MeCP2), which, in turn, recruit histone deacetylases and other corepressors, to form large protein complexes. The binding of the large protein complexes with DNA leads to an inactive chromatin structure and blocks gene transcription (20). Hence, the maximal reactivation of genes silenced by methylation may require simultaneous blockage of DNA methylation and histone deacetylation.

The function of THSD1 is unknown. It encodes a transmembrane molecule containing a thrombospondin type 1 repeat, which may be involved in cell adhesion and angiogenesis. A review on thrombospondins and tumor angiogenesis commented that proteins possessing the thrombospondin type 1 repeat, including thrombospondin-1, ADAM metallopeptidase with thrombospondin type 1 motif 1 (ADAMTS1) and ADAMTS8, and brain-specific angiogenesis inhibitor 1 (BAI1), regulate angiogenesis (21). Recently, Myc-induced enhanced neovascularization in mouse colonocytes correlated with down-regulation of antiangiogenic thrombospondin-1 and other proteins with thrombospondin type 1 repeat, such as connective tissue growth factor and THSD1 (22), suggested that THSD1 may be involved in tumor angiogenesis. An interaction network for THSD1 v1 predicted by “SMART”⁹

Differential gene expression profiles of three tumor-suppressing MCHs and their tumorigenic parental SLMT-1 cell line were revealed by cDNA oligonucleotide microarrays. A list of 40 genes (Table 1B), showing down-regulation in SLMT-1 compared with NE1, the immortalized normal esophageal epithelial cell line, in regions outside the critical 13q regions identified on chromosome 13q, might be involved in global downstream cascades of tumor suppression in the nude mouse model. The identified genes had a wide spectrum of functions, including cell proliferation, cell death regulator, transcription factor, signal transduction, extracellular matrix, cell adhesion, immune response, and heat shock proteins. Of these, CDKN1A is a cyclin-dependent kinase inhibitor that negatively regulates cell proliferation. Their importance in ESCC tumorigenesis and usefulness as biomarkers remains to be further evaluated.

Human ESCC recipient cell line, SLMT-1, and mouse donor cell line containing the intact human chromosome 13, MCH204.3, were used for microcell-mediated chromosome transfer. Culture conditions for the SLMT-1, donor, MCH, and TS cell lines, and details for a panel of 16 ESCC cell lines were described previously (10, 12). NE1, a human papillomavirus E6/E7/telomerase–immortalized normal esophageal epithelial cell line (23), is the reference cell line for gene expression profile analysis in the microarray and RT-PCR studies.

The intact chromosome 13 was introduced into the ESCC SLMT-1 cell line using microcell-mediated chromosome transfer techniques, as described previously (12). Chromosome 13 MCHs were obtained after selection for 4 weeks.

Genomic DNAs from the donor, recipient, MCHs, and TSs were extracted, as previously described (12). The MCHs were tested for the absence of mouse DNA contamination by DNA slot blot hybridization, as previously described (12). The presence of the extra chromosome 13 in the MCHs was verified by FISH using WCP 13 SpectrumGreen chromosome 13–specific probes (Vysis). FISH and image capture were done, as previously described (11). A minimum of 20 metaphase spreads was analyzed.

Fifty microsatellite markers spanning the entire chromosome 13 were used. PCR amplification and capillary electrophoresis of PCR products by the semiautomated ABI PRISM 3100 genetic analyzer were done as described (12). The mapping information (Fig. 1) and primer sequences were obtained from the University of California Santa Cruz Genome Bioinformatics¹⁰

Each of the six injection sites was inoculated with 2 × 10⁶ cells in the nude mouse assay for three chromosome 13 MCHs (MCH13-111, MCH13-113, MCH13-117) and two TS cell lines (MCH13-111/TS2R and MCH13-117/TS1L). Eight TS cell lines were established. Details of the tumorigenicity assay and criteria for TS deletion mapping analyses were previously described (12).

Total RNA was extracted from the cell lines with RNeasy Midi Kit (Qiagen). Twenty micrograms of total RNA were reverse transcribed with SuperScript II reverse transcriptase (Invitrogen) into cDNA. The cDNAs of SLMT-1 and MCHs and NE1 were labeled with Cy5-dUTP and Cy3-dUTP (Amersham Biosciences), respectively. Reciprocal dye swap labeling was done in each case. For SLMT-1/NE1 comparison, the whole experiment was repeated twice. The high-density microarrays used for our experiments were spotted with 18,912 60-mer oligonucleotides from a human oligonucleotide library (Sigma-Genosys) onto poly-l-lysine–coated glass slides at the Genome Institute of Singapore (24). The labeled probes were hybridized to the 19k chips in a Maui hybridization chamber (BioMicro Systems) for 16 h at 42°C. The slides were washed sequentially in 2× SSC and 0.1% SDS for 1 min, 1× SSC for 30 s, 0.2× SSC for 30 s, 0.05× SSC for 8 s, and then spin dried. Hybridized arrays were scanned on a GenePix 4100A scanner (Axon Instruments) at 635 and 532 nm for Cy5- and Cy3-labeled cDNAs. The images were analyzed by GenePix Pro 4.0 and the resulting data was processed by the mAdb microarray database of the Genome Institute of Singapore. Candidates were shortlisted, when intensities of hybridization signals were ≥2 in four of four slides for SLMT-1/NE1 and ≥1.2 in at least five of six slides in the three cases of SLMT/chromosome 13 MCHs.

For RT-PCR, total RNAs were reverse transcribed with Moloney murine leukemia virus (GE Healthcare) into cDNA, and PCR was done with AmpliTaq Gold DNA polymerase (Applied Biosystems) as previously described (11). Details of primer sequences are shown in Table 2. Real-time quantitative PCR was done using Mx3000P real-time PCR system (Stratagene; ref. 10). The THSD1 (transcript variant 1), PHF11, and GAPDH Taqman probes were purchased from Applied Biosystems.

A fragment of 910 bp at location −808 to +101 (chr13:51878221-51879130) was identified as the putative THSD1 promoter by Gene2Promoter. One CpG island of 338 bp was located within the putative promoter of THSD1 with an observed/expected CpG ratio, 0.60; a total of 35 CpG sites were found. THSD1 methylation status in NE1 and ESCC cell lines was determined by chemical treatment with sodium metabisulfite, MSP analysis, and bisulfite sequencing. Briefly, 2 μg genomic DNA were modified by sodium metabisulfite at 55°C for 15 h. Bisulfite-treated DNA was purified with QIAquick PCR Purification Kit according to the manufacturer's protocol and was ethanol precipitated. Primer sequences for MSP analysis and bisulfite sequencing (Table 2) were designed by MethPrimer v1.1 β (25). Bisulfite-treated genomic DNAs were amplified by PCR, gel purified, and subcloned into pMD18-T vectors (TaKaRa Biotech). Ten clones from each cell line or tissue were sequenced. The PCR amplicon encompassed 12 CpG sites in the −121 to −331 region of the CpG island around THSD1 exon 1.

EC18 and KYSE180 were treated with 5 μmol/L 5-aza-2′-deoxycytidine (Sigma), as previously described (10). On the 5th day of 5-aza-2′-deoxycytidine treatment, 300 nmol/L trichostatin A (Sigma) was added for 24 h before cell harvesting.

The full-length wild-type cDNAs of both THSD1 variants were cloned from NE2, an immortalized epithelial cell line, using the following primers: THSD1-NheI-start-2 5′-TACGGCTAGCATGAAACCAATGTTGAAAGAC-3′ (forward) and THSD1-Stop-NotI-2 5′-ATAGTTTAGCGGCCGCCTAGATCACCAGCTTCTCC-3′ (reverse). The complete coding sequence of THSD1 v1 and v2 contains 2,559 and 2,400 bp, respectively. The coding sequence of THSD1 v1 and v2 was sequence-verified with reference to the Genbank accession no. NM_018676 (GI:40805850) and NM_199263 (GI:40805851), respectively. Subsequently, both variants were cloned into pCR3.1 neomycin-resistant plasmid with NheI and NotI. One microgram of plasmid DNA was transfected into SLMT-1 with Lipofectamine 2000 reagent (Invitrogen) for 4.5 h. After 4 weeks of selection in DMEM/5% FCS containing 500 μg/mL neomycin, colonies were fixed and stained with Giemsa for colony scoring. The assays were repeated independently in quadruplicate.

The independent t test was used for nude mouse assay and colony forming assay.