Inactivation of CACNA1G, a T-Type Calcium Channel Gene, by Aberrant Methylation of Its 5′ CpG Island in Human Tumors

Introduction

CpG islands are GC-rich regions of DNA, which are coincident with the promoters of 60% of human genes (1) and are normally unmethylated, regardless of gene expression (2) . Methylation of CpG islands is detected in genes that are located on the inactive X chromosome (3) and genes that are inactivated by imprinting (4) . Aberrant cytosine methylation in the promoter region is also implicated as one mechanism of tumor suppressor gene inactivation in cancer (reviewed in Refs. 5 and 6 ). To date, various tumor suppressor genes, including Rb1, VHL, p16, BRCA1, hMLH1, and E-cad (5, 6, 7, 8, 9) , have been shown to be inactivated by hypermethylation in sporadic cancers. In a subset of genes, CpG island methylation is also associated with aging in normal tissues (10 , 11) .

Recently, a putative tumor suppressor gene, HIC-1, was isolated based on hypermethylation of a CpG island on human chromosome 17p13, where loss of heterozygosity is detected in various human tumors (12) . Because a HIC-1 allele is frequently lost by gene deletion (13) , homozygous inactivation of this gene may occasionally be caused by the combination of hypermethylation and deletion. Therefore, aberrant methylation of CpG islands in frequently deleted chromosomal regions could serve as a marker to identify targeted tumor suppressor genes. To date, several techniques have been used to identify genes differentially methylated in cancer (14, 15, 16, 17) , and several known and unknown gene fragments were identified. However, information about the genes inactivated by hypermethylation in cancer is still limited.

We have developed a new PCR-based technique termed methylated CpG island amplification, which allows us to selectively enrich for methylated, GC-rich, and short (300-bp to 2-kb) DNA fragments (18) . Using amplicons from cancer as a tester and amplicons from normal tissue as a driver, it is possible to identify CpG islands differentially methylated between cancer and normal through subtraction techniques, such as representational difference analysis. By using this strategy, we have cloned several CpG islands differentially methylated between normal colon tissue and colorectal cancer (18) . One of them, termed MINT31, mapped to chromosome 17q21. Loss of heterozygosity in this region is frequently detected in various cancer types, including lung, colon, stomach, prostate, esophagus, and head and neck cancers (19, 20, 21, 22, 23, 24) . Furthermore, loss of 17q may influence pathogenesis, progression, and survival in various tumors (19 , 23) . Whereas BRCA1 maps to 17q21, it is rarely inactivated in sporadic tumors, suggesting that another tumor suppressor gene exists in this locus (21 , 22) .

In this report, we have identified a T-type calcium channel gene on 17q21 as a novel target for the aberrantly methylated DNA cloned with the above approach. There is growing evidence that calcium signaling is involved in cell proliferation and cell death (25) . Inactivation of a calcium channel gene may, therefore, play an important role in cancer development and progression.

Materials and Methods

Tissue Samples and Cell Lines.

Forty-nine primary colorectal cancers, 28 colorectal adenomas, 16 primary gastric cancers, and 23 acute myelogenous leukemia samples were used for methylation analyses. DNA from eight colon cancer cell lines (Caco2, RKO, SW48, HCT116, DLD1, Lovo, SW837, and HT29), four lung cancer cell lines (OH3, H249, H157, and H209), four brain tumor cell lines (Dauy, D283, U87, and U373), nine breast cancer cell lines (MB-468, MCF7, MB-231, MB-474, MB-435, MB-453, BT20, CAMA1, and SKBR3), seven hematopoietic cell lines (CEM, Raji, KG1A, HL60, ML-1, Molt3, and K562), and four prostate cancer cell lines (DU145, DUPRO, LNCaP, and TSUPRL) were also investigated. DNA was extracted by standard procedures. RNA was isolated from cell lines and adenomas using Trizol (Life Technologies, Inc., Gaithersburg, MD). For reexpression analysis, cell lines were treated daily with 5-deoxyazacytidine (Sigma Chemical Co., St. Louis, MO) at a final concentration of 1 μm for 6 days. All tissue samples were obtained from patients who gave informed consent according to institutional guidelines.

RT 3 -PCR.

Six μg of total RNA were reverse-transcribed using the Superscript kit (Life Technologies, Inc.) for first strand cDNA synthesis. One hundred ng of cDNA were used as template for RT-PCRs. To design the RT-PCR primers, we performed a Blast 4 search (program of the National Center for Biological Information) using the rat Cacna1G cDNA sequence (GenBank accession no. AF027984) reported previously (26) , and exon-intron boundaries of the human CACNA1G were predicted by this analysis. Each primer set was designed to amplify the cDNA across several introns. 5 Glyceraldehyde-3-phosphate dehydrogenase was also amplified as a control using primers GAPDHF (5′-CGGAGTCAACGGATTGGTCGTAT-3′) and GAPDHR (5′-AGCCTTCTCCATGGTGGTGAAGAC-3′). All reactions were performed with RT-negative controls. PCR amplification was performed for 35 cycles of 95°C 30 s, 60–65°C for 30 s, and 72°C for 30 s, and the products were analyzed by agarose gel electrophoresis.

DNA Sequencing and Data Analysis.

PCR products were precipitated with ethanol, resuspended in diluted water, and cloned into the pCR2.1 vector using the TA cloning kit (Invitrogen, San Diego, CA), according to the manufacturer’s instruction. After transformation, plasmid DNA was purified using the Wizard Miniprep Kit (Promega, Madison, WI). DNA sequence analysis was carried out at the Johns Hopkins University Sequence Facility using automated DNA sequencers (Applied Biosystems). Sequence homology was identified by the Blast program. An IMAGE cDNA clone (GenBank accession no. H13333) was identified by Blast analysis using the sequence of BAC AC004590 (GenBank), which includes MINT31. Putative genes (G1 and G2) were identified by GENSCAN, 6 using the BAC sequence data. IMAGE cDNA clone H1333 was then obtained from the American Type Culture Collection (Manassas, VA) and completely sequenced. Potential transcription factor binding sites and promoter prediction were examined using the TESS and TSSG programs, respectively, available at the Baylor College of Medicine BCM Launcher. 7 The nucleotide sequence of part of the 5′ end of the cDNA of CACNA1G has been submitted to GenBank (accession no. AF124351).

Bisulfite-PCR Methylation Analysis.

Bisulfite treatment was performed as reported previously (27) . Briefly, 2 μg of genomic DNA were denatured with 2 m NaOH for 10 min, followed by incubation with 3 m sodium bisulfite (pH 5.0) for 16 h at 50°C. After treatment, DNA was purified using a Wizard Miniprep Column (Promega), precipitated with ethanol, and resuspended in 20 μl of diluted water. Two μl of the aliquot were used as template for each PCR. Semiquantitative bisulfite-PCR was performed essentially as described (28) . To avoid overestimation of the methylated alleles, we considered the following points. (a) Primers were designed to contain a minimum number of CpG dinucleotides in the sequence to avoid the biased amplification of methylated alleles. If primers did contain CpG sites, they were designed to amplify methylated and unmethylated alleles equally (using a mixture of C or T for sense and a mixture of G or A for antisense primers). (b) The primers were designed to contain a maximum number of thymidines converted from cytosines to avoid amplifying the nonconverted genomic sequence. (c) Restriction sites that only appear after bisulfite conversion (e.g., ACGC to ACGT) were used if possible (regions 1–8). PCR was performed as described previously (27) .⁵ Twenty % of the PCR products were digested with the appropriate restriction enzymes, precipitated with ethanol, and separated by 5% PAGE. Gels were stained with ethidium bromide, and the intensity of each allele was calculated by densitometry, using the ImageQuant software (Molecular Dynamics, Sunnyvale, CA).

5′ RACE.

RACE was performed using the Marathon cDNA Amplification Kit (Clontech, Palo Alto, CA). Single-stranded DNA was synthesized from human brain poly(A) RNA (Clontech) using Thermoscript (Life Technologies, Inc.) and using random hexamers. cDNA synthesis was performed at 60°C for 30 min to disrupt the secondary structure of RNA related to high GC content. Second strand DNA was synthesized, and the RACE adapter was ligated according to the manufacturer’s instructions. The first round of PCR was performed using RACE primer 1 and CAC-1 (5′-AGAGGGCGAACGAAAAGGGGGGCA-3′). A fraction (1/100) of the PCR product was then used as a template for a second round of (nested) PCR using RACE primer 2 and CAC-8 (5′-GAGGGGAGGAGGGATCTCTTTAGG-3′). Amplified fragments were cloned using the TOPO-TA cloning Kit (Invitrogen) and sequenced as above.

Results

Identification of CACNA1G as a Target for Hypermethylation on Human Chromosome 17q21.

To identify genes differentially methylated in colorectal cancer, we used methylated CpG island amplification followed by representational difference analysis (18) . One of the clones recovered (MINT31) mapped to human chromosome 17q21 using a radiation hybrid panel, and a Blast search revealed this fragment to be completely identical to part of a BAC clone (GenBank accession no. AC004590) sequenced by high-throughput genomic sequencing (Fig. 1A) ⇓ . The region surrounding MINT31 fulfills the criteria of a CpG island (29) : GC content, 0.67; CpG/GpC ratio, 0.78; and a total of 305 CpG sites in a 4-kb region (Fig. 1B) ⇓ . Using this CpG island and 10 kb of flanking sequences in a Blast analysis, we identified several regions highly homologous to the rat T-type calcium channel gene, Cacna1G (26) . Several ESTs were also identified in this region. Using GENSCAN, we identified two putative coding sequences (G1 and G2; Fig. 1A ⇓ ). Blastp analysis revealed that G1 has a high homology to the EH-domain-binding protein, epsin (30) , whereas G2 is homologous to a Caenorhabditis elegans hypothetical protein (GenBank accession no. 2496828).

The MINT31 CpG island corresponds to the 3′ regions of G1 and G2 (Fig. 1) ⇓ , based on the direction of the open reading frame and the presence of a poly(A) tail, and is unlikely to influence their transcription. The EST closest to MINT31 (H13333) was sequenced entirely and was found not to contain a continuous open reading frame, but a polyadenylation signal was identified on one end, along with a poly(A) tail. These data suggest that H13333 corresponds to the last two exons of an unidentified gene. MINT31 is in the intron of this gene and is unlikely to influence its transcription. However, based on both promoter prediction (TSSG) analysis of this region and homology to the rat Cacna1g sequence, the MINT31 CpG island is also in the 5′ region of the human CACNA1G gene and may play a role in its transcriptional activity.

The human CACNA1G sequence deposited in GenBank lacks the 5′ region of the gene, when compared with the rat homologue. To determine the 5′ region of human CACNA1G, we amplified cDNA by RT-PCR using primers based on the BAC sequence (AC004590; GenBank). The PCR products were cloned and sequenced, and the genomic organization of the gene was determined by comparing our new sequences as well as the known sequences to the BAC that covers this region. CACNA1G is composed of 34 exons that span a 70-kb area (Fig. 1A) ⇓ . On the basis of sequences deposited in GenBank, the gene has two possible 3′ ends caused by alternate splicing. CACNA1G is highly homologous to rat Cacna1g, with 93% identity at the protein level and 89% identity at the nucleotide level. The 5′ flanking region of CACNA1G lacks TATA and CAAT boxes, which is similar to many housekeeping genes. A putative TFIID-binding site was identified 547–556 bp upstream from the translation start site, and several other potential transcription factor binding sites such as AP1 (1 site), AP2 (2 sites), and SP1 (10 sites), were identified upstream of CACNA1G exon 1 using the promoter prediction program, TESS (data not shown).

Methylation Analysis of CACNA1G.

The CACNA1G CpG island is 4 kb and is larger than many typical CpG islands. MINT31 corresponds to the 5′ edge of the island, whereas CACNA1G is in the 3′ region. It is not known whether large CpG islands such as this are coordinately regulated with regards to protection from methylation and aberrant methylation in cancer. To address this issue, we studied the methylation status of the 5′ region of CACNA1G using bisulfite-PCR of DNA from normal tissues as well as 36 human cancer cell lines from colon, lung, prostate, breast, brain and hematopoietic neoplasms. The CpG island was divided into eight regions (Fig. 1B) ⇓ , and the methylation status of each was examined separately. The genomic DNA was treated with sodium bisulfite and PCR-amplified using primers containing no or a minimum number of CpG sites. Methylated alleles were detected by digesting the PCR products using restriction enzymes that specifically cleave sites created or retained due to the presence of methylated CpGs. Examples of these results are shown in Fig. 2 ⇓ . None of the regions was methylated in normal colon, consistent with a uniform protection against de novo methylation.

Regions 1 and 2 were frequently methylated in cancer cell lines and behaved in a concordant manner. These two regions were methylated in most cancer cell types except gliomas, and most cell lines where methylation was found methylated both regions simultaneously. Region 3, which is less GC-rich than any of the other regions, had either no methylation or very low levels of methylation in most cell lines. Regions 5, 6, and 7 behaved quite differently than did regions 1–3. Methylation of these regions was less frequent than in regions 1 and 2: 22 of 36 cell lines had no detectable methylation there, despite often showing methylation of regions 1 and 2. However, when methylation was present (in 13 of 36 cell lines), it affected all three regions simultaneously, although to varying extents. Finally, regions 4 and 8 behaved differentially again, being partially methylated primarily in colon and breast cell lines. Therefore, with regards to hypermethylation in cancer, the CpG-rich region upstream of CACNA1G appears to be composed of two CpG islands that behave independently. MINT31 corresponds to the upstream CpG island (island 1, regions 1 and 2), whereas the 5′ region of CACNA1G is contained in the downstream CpG island (island 2, regions 5–7). Regions 3, 4, and 8 correspond to the edge of these CpG islands and behave a little differently than the hearts of the CpG islands, as described previously for the E-cad gene (31) .

Overall, the methylation patterns fell into five distinct categories (examples in Fig. 2 ⇓ ): pattern 1, no methylation in any region (normal tissues); pattern 2, slight methylation of island 1 (6 cell lines; for example, see TSU-PRL in Fig. 2 ⇓ ); pattern 3, heavy methylation of island 1 but no methylation of island 2 (16 cell lines; for example, see Caco2 in Fig. 2 ⇓ ); pattern 4, heavy methylation of island 1 and moderate to heavy methylation of island 2 (6 cell lines; for example, see RKO and Raji in Fig. 2 ⇓ ); and pattern 5, high methylation of island 1 and low to moderate methylation of island 2 (7 cell lines; for example, see MB-231 in Fig. 2 ⇓ ).

Expression Analysis of CACNA1G.

In a previous study of rat Cacna1G, this gene was shown to be expressed most abundantly in the brain (26) . To determine the expression of CACNA1G in normal and neoplastic human cells, we performed RT-PCR using cDNA from various normal tissues and from a panel of 27 tumor cell lines (Fig. 3A) ⇓ . CACNA1G was expressed ubiquitously in a variety of tissues and cell lines. In normal tissues, expression was relatively low but easily detectable, whereas most cell lines had relatively high expression of CACNA1G. However, some cell lines had negligible or totally absent levels of CACNA1G expression. We next correlated the results of CACNA1G expression with the detailed methylation analysis described previously. The correlation between methylation and expression of CACNA1G is summarized in Fig. 4 ⇓ . In this analysis, a remarkable pattern emerged. Methylation of regions 1–4 and 8 had no effect on CACNA1G expression. However, there was a strong correlation between methylation of regions 5–7 and expression of the gene. In fact, all cell lines that lack methylation of this region strongly express the gene. All six cell lines with pattern 4 methylation studied had no detectable expression. Finally, the 7 cell lines with pattern 5 methylation (examples, DLD-1 and MB-453) had variable levels of expression ranging from very low to near normal. The fact that patterns 3 and 5 differ significantly with regards to expression but are almost identical with regards to methylation of all regions except 7 suggests that this area is important in the inactivation of CACNA1G.

To confirm whether methylation of the 5′ CpG island of CACNA1G is really associated with gene inactivation, three nonexpressing cell lines showing pattern 4 methylation (RKO, SW48, and Raji) and two weakly expressing cell lines showing pattern 5 methylation (MB-231 and MB-435) were treated with 1 μm of the methyltransferase inhibitor 5-deoxyazacytidine. After treatment, all these cell lines reexpressed CACNA1G mRNA (Fig. 3B) ⇓ . Consistent with reexpression, demethylation of region 7 was observed after 5-deoxyazacytidine treatment (Fig. 3C) ⇓ .

5′ RACE analysis of CACNA1G.

The data presented above suggested that regions 6 and 7 are close to the transcription start site of the gene. To verify this, we determined the 5′ end of CACNA1G by 5′ RACE. To overcome the problems associated with RACE analysis of very long transcripts with a GC-rich 5′ end, we used random hexamer-primed cDNA synthesized at 60°C. Using poly(A) mRNA from human brain, we recovered a RACE transcript that begins 527 bp upstream of the translation initiation site. This region is a few bp away from the putative TFIID-binding site. No product was recovered when RACE was attempted using an upstream primer suggesting that the 5′ end of CACNA1G is in this region. As predicted by our methylation/expression correlations, the 5′ end of CACNA1G is located in region 6 and is ∼100 bp upstream of region 7.

Methylation and Expression of the CACNA1G 5′ Region in Primary Human Tumors.

De novo cytosine methylation is thought to sometimes occur in vitro during cell propagation (32) . To determine whether the methylation of CACNA1G occurs in vivo, we examined primary human tumors for methylation of the 5′ region of CACNA1G. Representative results of the analysis of region 7 (which correlates best with expression) by bisulfite-PCR are shown in Fig. 5A ⇓ . Aberrant methylation was detected in 17 of 49 (35%) colorectal cancers, 7 of 23 colorectal adenomas (25%), 4 of 16 (25%) gastric cancers, and 3 of 17 (13%) acute myelogenous leukemia cases (Fig. 5A) ⇓ . In colorectal cancers, there was a significant correlation between methylation of CACNA1G and methylation of p16 (P < 0.005) and hMLH1 (P < 0.001) as well as a strong correlation with the presence of microsatellite instability and the recently identified (33) CpG island methylator phenotype.

To determine whether aberrant methylation of the 5′ region of CACNA1G affects the expression status of this gene in primary tumors, we performed RT-PCR using cDNA from a series of colorectal adenomas. Six of eight cases that showed no methylation of region 7 expressed CACNA1G. In sharp contrast, all five cases that showed methylation of region 7 had no detectable expression of this gene (Fig. 5B) ⇓ .

Discussion

In this study, we have identified a human T-type calcium channel gene as a target of aberrant methylation and silencing in human tumors. Detailed analysis of the CpG island upstream of CACNA1G revealed that methylation 300–800 bp upstream of the gene closely correlated with transcriptional inactivation. This region also contained the transcription start site of the gene, as determined by 5′ RACE analysis.

The CACNA1G promoter is contained in a large GC-rich area that is not coordinately methylated in cancer. The CpG island around MINT31 is much more frequently methylated in cancers compared with that just upstream of CACNA1G. This may simply be caused by differential susceptibility to de novo methylation between these two regions, with methylation of MINT31 serving as a trigger and eventually spreading to CACNA1G, as described in other genes (31) . However, we believe that these two regions are controlled by different mechanisms because: (a) cell lines kept in culture for countless generations do not, in fact, spread methylation from MINT31 to CACNA1G (for example, see Caco2); (b) region 3, which separates the two islands, is infrequently and sparsely methylated in cancer; and (c) we found two cases of primary colorectal cancer that are methylated at the CACNA1G promoter but not at MINT31 (data not shown). Therefore, methylation of MINT31 appears to be independent of methylation of CACNA1G, suggesting that they are two distinct CpG islands regulated by different mechanisms. These data leave open the possibility that MINT31 is the promoter for an unidentified gene, which may perhaps be transcribed opposite to CACNA1G.

Many CpG islands of silenced genes appear to be methylated uniformly and heavily throughout the island (7 , 31) . In contrast, the methylation patterns of the 5′ region of CACNA1G (regions 5–7) were heterogeneous in the cell lines that did not express this gene. These results are consistent with the methylation patterns of MGMT (34) and FHIT in nonexpressing cell lines (35) . Nevertheless, methylation does appear to play a role in CACNA1G repression because demethylation readily reactivates the gene. It is probable then that the amount of methylated cytosines detected is enough for the inactivation of the gene through chromatin condensation, as shown for several genes (36 , 37) . In the case of MGMT, methylation of the 5′ region of the gene is heterogeneous, although it has been proposed that the extensively methylated region (methylation “hot spot” within the island) is more important for transcriptional silencing than the rarely methylated region (34) . Therefore, because our studies were limited to a few cytosines in this CpG island, it is also possible that we have missed a methylation hot spot in the area. In fact, methylation in the nonexpressing group is quite high at region 7, which may then be close to this putative hot spot. To clarify which CpG sites are critical for CACNA1G gene silencing, further examination using bisulfite sequencing may be necessary.

The causes of CACNA1G methylation remain to be determined. Methylation was not detected in normal colon mucosa, placenta, normal breast epithelium, and normal bone marrow, including samples from aged patients, suggesting that methylation of this region is cancer specific. However, there was a significant correlation between methylation of CACNA1G and other tumor suppressor genes, such as p16 and hMLH1. Thus, CACNA1G probably is a target for the recently described CpG island methylator phenotype, which results in a form of epigenetic instability with simultaneous inactivation of multiple genes (33) . Therefore, methylation of CACNA1G may, in fact, result from mutations in a “master” gene that controls the methylation status of a subset of CpG islands.

T-type calcium channels are involved not only in electrophysiological rhythm generation but also in the control of cytosolic calcium during cell proliferation and cell death (reviewed in Ref. 25 ). Our results demonstrate that the expression of CACNA1G is not limited to brain and heart, suggesting that it may play a role in these other tissues. It has previously been shown that Ca²⁺ influx via T-type channels is an important factor during the initial stages of cell death, such as apoptosis (25) , ischemia (38) , and complement-induced cytotoxicity (39) . A high level of Ca²⁺ in mitochondria is essential to activate the genes associated with programmed cell death (40) . The proto-oncogene BCL2 inhibits cell death induced by mitochondrial Ca²⁺ uptake (41) . These results indicate that intracellular calcium signaling plays an important role in apoptosis. Furthermore, calcium channel antagonists, which specifically block T-type channels, inhibit cell death by decreasing the influx of Ca²⁺ through T-type calcium channels (39) . Finally, T-type calcium channels are down-regulated in fibroblasts transformed by platelet-derived growth factor (42) . Therefore, the impairment of voltage-gated calcium channels may play an important role in cancer development and progression through altering calcium signaling, and CACNA1G deserves further consideration as a candidate tumor suppressor gene in various human neoplasms. Further studies, including functional analyses of CACNA1G reexpression in cancer, are necessary to understand the roles this gene might play in calcium signaling in cell proliferation and cell death. Inactivation of the CACNA1G gene by other mechanisms, such as point mutations and chromosome deletions, will also need to be examined to further clarify the role of this gene in the development of human tumors.