Identification of Aberrantly Methylated Genes in Association with Adult T-Cell Leukemia

Human T-cell leukemia virus type I (HTLV-I) is the causative retrovirus of a neoplastic disease, adult T-cell leukemia (ATL) and an inflammatory disease, HTLV-I-associated myelopathy/tropical spastic paraparesis (1, 2, 3, 4). After infection with HTLV-I, a small proportion of carriers (about 2–5%) develop ATL after a long latent period (5). In this virus-induced leukemia, viral proteins encoded by HTLV-I play an important role in the proliferation of infected cells and leukemogenesis. Among them, Tax is considered to play a central role in leukemogenesis because of its pleiotropic actions (6, 7), such as transcriptional activation of cellular genes (8, 9, 10), trans-repression of cellular genes transcription (11, 12), and functional inactivation of p53 and MAD1 (12, 13). These pleiotropic functions render HTLV-I-infected cells able to proliferate, and confer resistance to apoptotic signals, resulting in clonal expansion.

In the late stage of leukemogenesis, tax is frequently inactivated through several mechanisms (14) such as loss of 5′-long terminal repeat (LTR) (15), genetic alterations of tax gene (16), and DNA hypermethylation in 5′-LTR (17), indicating that Tax is not always necessary for leukemogenesis. Because Tax is the major target molecule of CTLs in vivo (18), the expression of Tax confers a growth advantage to infected cells, but on the other hand, it renders infected cells susceptible to CTLs. Fully transformed ATL cells are considered to acquire the ability to proliferate in vivo in the absence of Tax expression. Such a transformation process is thought to include alterations of host genome: genetic and epigenetic changes. Although the genetic changes, such as mutation of p53 (19) and deletion of p16 (20, 21), in ATL cells were reported, they are not frequent and are observed predominantly in the late stage of the disease.

In addition to genetic alterations, DNA hypermethylation of promoter region CpG islands has been analyzed in the context of oncogenesis because this process silences gene transcription of tumor-suppressor genes. This epigenetic alteration is observed commonly in various cancer cells. Although methylation “profiling” studies have shown that some genes are frequently methylated in various tumor cells, other genes are methylated in a tumor-type-specific manner (22, 23). To date, several methods have been developed to isolate differentially methylated DNA regions in cancers (24, 25, 26, 27, 28, 29). Recently, with methylated CpG island amplification/representational difference analysis (MCA/RDA) method, we isolated hypomethylated DNA regions and demonstrated that MEL1S gene was hypomethylated and aberrantly transcribed in ATL cells (30).

The present study was designed to isolate hypermethylated DNA regions in ATL cells compared with cells in the carrier state using the MCA/RDA method and to identify those genes that have an expression associated with DNA hypermethylation. On the basis of our results, we discuss the association between aberrant DNA methylation and leukemogenesis of ATL.

Peripheral blood mononuclear cells (PBMCs) were isolated from 10 patients with ATL (five acute type cases and five chronic type cases), five asymptomatic carriers, and five uninfected individuals using Ficoll-Paque density centrifugation method. We also used the cell lines ED, ATL-43T, ATL-48T, ATL-55T, MT-1, and TL-Om1, which were derived from leukemic clones, and MT-2, which is derived from nonleukemic cells. To study the effect of demethylation, ATL-43T was cultured in media supplemented with 10 μmol/l 5-aza-2′-deoxycytidine (5-aza-dC; Sigma, St. Louis, MO) for 3 days, 10 μmol/l 5-aza-dC and 1 μmol/l trichostatin A (TSA; Sigma) for the last 24 hours, or 1 μmol/l TSA for 24 hours alone. The human embryonic kidney cell line, HEK 293, was used for the packaging of recombinant adenovirus vectors.

Viral FLICE-inhibitory protein (FLIP) derived from the equine herpes virus type 2, E8 (31), and the long form of murine cellular FLIP (mCasper_L; Ref. 32) expression vectors were transfected into ATL-43T cells by electroporation with a Gene Pulser II (Bio-Rad, Hercules, CA). Stable transfectants were selected and maintained in culture medium containing G418 (500 μg/ml; Nacalai tesque, Kyoto, Japan). The transfected cell lines by each of the vectors were designated as ATL-43T-E8 and ATL-43T-mCas, respectively.

To identify aberrantly hypermethylated DNA regions in ATL cells, we used the MCA/RDA method, as reported previously (28). Five micrograms of genomic DNA were digested with 100 units of SmaI (New England Biolabs, Beverly, MA) twice and then digested once with 20 units of XmaI (New England Biolabs). RMCA adaptor was prepared by annealing of the oligonucleotides RMCA24 (5′-CCACCGCCATCCGAGCCTTTCTGC-3′) and RMCA12 (5′-CCGGGCAGAAAG-3′), and ligated to the digested DNA fragments using T4 DNA ligase (New England Biolabs). To amplify the hypermethylated DNA fragments, which were ligated adaptors in both ends, PCR was performed using the RMCA24 oligonucleotides as primers. The amplicons were synthesized using samples from an ATL patient and a HTLV-I carrier. For detection of ATL cell-specific hypermethylated DNA sequences, MCA products from a carrier were used as a driver of RDA, and those products from an acute ATL patient as a tester. We used GeneFisher Basic Reagent Set (TaKaRa, Shiga, Japan) for RDA. In RDA step, 500 and 100 ng of ligation mixture were used for the first and second selective PCR, respectively. Oligonucleotides used for RDA were JMCA24 (5′-GTGAGGGTCGGATCTGGATGGCTC-3′), JMCA12 (5′CCGGGAGCCAGC-3′), NMCA24 (5′-GTTAGCGGACACAGGGCGGGTCAC-3′), and NMCA12 (5′-CCGGGTGACCCG-3′). Subcloning of the MCA/RDA products was carried out using pCR-XL-TOPO (Invitrogen, Carlsbad, CA) or pGEM-T Easy (Promega, Madison, WI) as vectors, and then the sequences of each fragment were determined by PCR using M13 primers. Sequence homologies and localization in chromosomes were identified using the BLAST program of the National Center for Biotechnology Information.³

To confirm that the isolated DNA regions are specifically hypermethylated in ATL cells, Southern blot method was used. MCA products from an ATL patient and a HTLV-I carrier (500-ng each) were separated by electrophoresis in 1.5% agarose gels and then transferred to Hybond-N + (Amersham Biosciences, Piscataway, NJ). All of the isolated DNA fragments were labeled with ³²P, and hybridized to these filters.

For nine DNA regions identified by MCA/RDA, the methylation status of the DNA regions was determined by Combined Bisulfite Restriction Analysis or bisulfite sequencing as described previously (33). First, genomic DNAs were treated with sodium bisulfite (34) and then amplified by nested PCR using the specific primers listed in Supplementary Table 1. The PCR products of these regions were digested with TaqI (New England Biolabs) or AccII (TaKaRa), subjected to electrophoresis in 3% agarose gels, and visualized by ethidium bromide staining. The percentage of DNA methylation was calculated by the intensities of methylation and unmethylation signal determined by ATTO densitometry software (ATTO, Tokyo, Japan).

For detailed analysis of DNA methylation in Kruppel-like factor 4 (KLF4) and early growth response 3 (EGR3) genes, we performed bisulfite sequencing. The PCR products of the isolated regions and promoter regions of these genes were subcloned into pGEM-T Easy, thereafter, the sequences of each of 10 clones were determined. Because the promoter sequence of KLF4 has not been determined, we predicted its sequence using the program⁴ supported by the BioInformatics & Molecular Analysis Section, Computational Bioscience and Engineering Lab, Center for Information Technology, and NIH. Primers for bisulfite sequencing are also listed in Supplementary Table 1.

Total RNA was extracted from the PBMCs and cell lines using Trizol reagent (Invitrogen) and then treated DNaseI (Invitrogen). cDNAs were synthesized from 0.5 μg of total RNA with the Superscript First-Strand Synthesis System for reverse transcription (RT)-PCR (Invitrogen) and used for semi-quantitative RT-PCR as template. The primers used for RT-PCR and their annealing temperatures are summarized in Supplementary Table 2. The number of PCR cycles was appropriately determined for each quantification (Supplementary Table 2). We used 1.25 units of ExTaq polymerase (TaKaRa) for each reaction. All experiments were performed including samples of whole brain (Clontech, Palo Alto, CA) and skeletal muscle (Stratagene, La Jolla, CA) as positive control of PCR reaction.

The recombinant adenovirus vectors containing KLF4 and EGR3 gene (KLF4-AD and EGR3-AD, respectively) were generated using Adeno-X Expression System (Clontech) according to the manufacturer’s protocol. These adenovirus vectors were concentrated and purified by Virakit for adenovirus 5 and recombinant derivatives (Virapur, San Diego, CA), and then the viral titers were determined using Adeno-X Rapid Titer Kit (Clontech). The lacZ-containing adenovirus vector (lacZ-AD) was also prepared as a negative control. All adenovirus vectors were used to infect an ATL cell line, ATL-43T, at 1,000 infectious units/cell.

The flow cytometry (model EPICS XL flow cytometer, Beckman Coulter, Miami Lakes, FL) was used for analyses of apoptosis. Annexin V-FITC/PI double staining and terminal-deoxynucleotidyl transferase-mediated dUTP-FITC nick-end labeling (TUNEL) assay were performed for detection of apoptosis, using MEBCYTO apoptosis kit (MBL, Nagoya, Japan) and MEBSTEIN apoptosis kit direct (MBL), respectively.

To identify hypermethylated regions in the genome of ATL cells, we carried out MCA/RDA, which was used previously to isolate a number of methylated CpG islands in colon cancer cell line (28). MCA products were generated from the genomic DNA of a carrier (driver) and an acute ATL patient (tester). After the second round of RDA, the PCR products were subcloned, and their sequences were determined. To confirm that identified DNA fragments were amplified in tester amplicon, we examined whether isolated DNA fragment specifically hybridized to the tester amplicon using Southern blot method (MCA-Southern). Specific hybridization to the tester amplicon implied that isolated DNA regions were hypermethylated in ATL cells compared with peripheral blood mononuclear cell (PBMC) from a carrier. Finally, we identified 53 differentially hypermethylated DNA fragments in ATL cells. The chromosomal locations of all of the fragments were analyzed by NCBI BLAST program. We tested whether these identified regions satisfied the criteria for CpG islands proposed by Takai and Jones (35). The results revealed that the majority of clones (48 of 53 clones) were located in CpG islands. Information of isolated sequences is described in Table 1.

Chronic ATL is characterized as an indolent form, which later progresses to aggressive forms (i.e., acute or lymphoma-type ATL). To confirm that DNA hypermethylation identified in this study is associated with disease progression, we analyzed the extent of DNA methylation of nine DNA regions at different stages by Combined Bisulfite Restriction Analysis. Fig. 1,A shows the profiles of the methylation status in these DNA fragments. In cell lines, CpG sites in identified DNA fragments were highly methylated, which was consistent with the finding of DNA methylation in the established cell lines. This confirmed that the isolated DNA regions were hypermethylated in ATL cells and that MCA-Southern could identify the hypermethylated DNA regions. In the carrier state, most DNA fragments were unmethylated. On the other hand, they were frequently methylated in chronic ATL, and the level of methylation increased in acute ATL, indicating that DNA methylation in the isolated DNA regions tends to accumulate according to disease progression. This was also confirmed in the sequential samples from a HTLV-I carrier, who developed acute ATL (Fig. 1 B).

To analyze the influence of identified DNA hypermethylation upon gene transcription, the neighboring genes were searched using NCBI BLAST program as described in Materials and Methods. We found that 31 of 53 (58%) clones were located within the exon or intron of the gene, and 41 of 53 (77%) loci were located within 10 kb from the transcriptional start site of the nearest gene (Table 1). Because the aberrant methylation of some identified genes, such as PAX5 (clone 10; ref. 36) and CSPG2 (clone 27; ref. 28), have been reported in various types of cancer cells, it confirmed that MCA/RDA method in this study isolated the hypermethylated DNA regions. Then, we analyzed the transcription of genes in which the transcriptional start sites existed within 2 kb from the identified hypermethylated DNA regions (Fig. 2). On the basis of their expression profiles, we could divide the genes identified into two groups; group I contained genes with an expression that was observed in activated T lymphocytes but suppressed in HTLV-I-transformed and ATL cell lines (Fig. 2,A). On the other hand, the transcription of genes in group II was not detected in activated T cells and HTLV-I-associated cell lines whereas their expression was found in the brain and/or skeletal muscle (Fig. 2 B). Group II genes were hypermethylated only in ATL cells but not in normal T lymphocytes. These results suggest that the suppressed expression of group I genes is implicated in leukemogenesis.

We studied the detailed DNA methylation status in the promoter and isolated regions of KLF4 and EGR3 genes, which belong to group I, using the bisulfite sequencing method. The sequences of each of the 10 clones are summarized in Fig. 3. In both ATL-43T and an acute ATL, the CpGs in the isolated region of KLF4 gene, which existed in exon 3, were heavily methylated (Fig. 3,A) whereas there was little methylation in normal PBMCs. In the predicted KLF4 promoter sequence, there was dense DNA methylation in ATL-43T and mild methylation in fresh ATL cells whereas the CpGs in normal PBMC were little methylated (Fig. 3,A). DNA methylation in the promoter region of KLF4 has been studied in primary cells with different stage of ATL to analyze the association with disease progression (Fig. 3,B). DNA methylation increased according to disease progression from carrier to leukemia although there was little difference between chronic and acute ATL. In the case of the EGR3 gene, the isolated region, which was in exon 2, was hypermethylated in the ATL cell line and fresh ATL cells but hypomethylated in normal PBMC (Fig. 3 C). Although the promoter region of EGR3 was hypermethylated in the ATL cell line, it was not methylated in fresh ATL cells and normal PBMCs.

Next, we analyzed whether the transcriptions of these silenced genes could be recovered by the demethylating agent, 5-aza-dC, and/or a histone deacetylase inhibitor, TSA in ATL cells. The combination of 5-aza-dC and TSA is known to induce a synergistic effect on DNA demethylation (37). As shown in Fig. 2 C, the transcripts of KLF4 gene were re-expressed by 5-aza-dC alone or by combination treatment but not by TSA alone, suggesting that DNA methylation of KLF4 gene is associated with its silencing in ATL cells. On the other hand, the transcript of EGR3 gene was detected more clearly when ATL cells were treated with TSA alone or their combination than with 5-aza-dC alone, indicating that the EGR3 gene was silenced by histone deacetylation rather than by DNA methylation. However, because demethylation by 5-aza-dC partially recovered EGR3 gene transcription, we consider that DNA methylation is associated in part with suppressed expression of EGR3 gene. In addition, the transcriptions of other group I genes, such as CSPG2, MAFB, and ADRB2, were also recovered by 5-aza-dC treatment. Because the transcription of ADAMTS2 and PTGDR could not be recovered by either 5-aza-dC or TSA, the silencing of these genes might be because of other mechanism(s).

To investigate the function of KLF4 and EGR3 genes in ATL cells, adenovirus vectors expressing KLF4 (KLF4-AD) or EGR3 (EGR3-AD), were transfected into ATL-43T, in which the transcription of both genes were completely suppressed. Transfection of KLF4-expressing adenovirus vector induced the transcription (Fig. 4,A) and resulted in accumulation of apoptotic cells as demonstrated by Annexin V-PI staining (Fig. 4,B). The apoptosis was also confirmed by TUNEL assay (data not shown). The number of apoptotic cells reached maximum 48 hours later (30.1%, Fig. 4 C). This percentage was similar to that of X-gal-stained cells (49.0%) when lacZ-AD was transfected into ATL-43T. Taken together, these results suggest that KLF4 expression induced apoptosis in most transduced cells.

Because the EGR3 gene is reported to be critical for Fas ligand (FasL) gene transcription (38), we studied whether enforced expression of EGR3 could induce apoptosis of ATL cells. After transfection, both EGR3 and FasL were transcribed 48 hours later (Fig. 5,A), which coincided with increased apoptotic cells in ATL-43T infected EGR3-AD, in contrast to control (Fig. 5,B and C). In addition, increased apoptotic cells were also confirmed by TUNEL assay (data not shown). Thus, enforced expression of EGR3 was considered to result in Fas-FasL-mediated apoptosis. To clarify whether this apoptosis is actually mediated by Fas signaling, we transfected vectors expressing mCasper_L or E8. mCasper_L is a mouse c-FLIP that inhibits the activation of procaspase 8 at the death-inducing signaling complex, whereas E8 is a viral FLIP derived from the equine herpes virus type 2. Transfection of EGR3-AD did not increase apoptosis of ATL-43T cells that expressed mCasper_L and E8 (Fig. 5 A and C), confirming that EGR3-induced apoptosis in ATL cell line is mediated by Fas-mediated signal.

In the present study, we identified hypermethylated DNA regions by MCA/RDA method. Consistent with the previous study (28), this method could identify hypermethylated CpG islands; 48 of 53 (91%) DNA clones identified in this study satisfied the criteria of CpG islands, and 41 of 53 (77%) DNA clones were located within 10 kb from the transcriptional start site of the nearest gene. Identified genes in the vicinity of isolated hypermethylated DNA regions could be divided into two groups; genes of group II are not expressed and not methylated in normal T lymphocytes, however, are hypermethylated in ATL cells. On the other hand, genes of group I are expressed but not methylated in normal T lymphocytes whereas their expression is suppressed in ATL cells in association with DNA methylation. It is possible that the mechanism of de novo methylation is dysregulated, resulting in aberrant methylation of genes despite their transcription as observed in group II genes. In this regard, Toyota et al. (28) isolated 33 hypermethylated DNA sequences in a colon cancer cell line using the MCA/RDA method and named these clones MINT1–33. Among DNA regions identified in this study, four clones were identical to MINT clones (Table 1, clone 2, clone 15, clone 30 and clone 48). These findings indicate that such DNA regions in the genome are prone to be methylated in cancer cells, which is consistent with an earlier report (22), although the factors that determine such susceptibility to methylation remain unresolved. In addition to such DNA methylation observed among different types of cancer cells, there are hypermethylated genes specifically observed in ATL cells. Analysis of DNA methylation of such genes in non-ATL T-cell lines showed that they were also methylated (data not shown), suggesting that DNA methylation of such genes is T-cell specific.

In addition to hypermethylation, we also reported hypomethylated genes in ATL cells, which included MEL1S, CACNA1H, and Nogo receptor genes as identified by the MCA/RDA method (30). Among them, the aberrant expression of MEL1S was frequently observed in ATL cells and has been shown to confer resistance against transforming growth factor-β. Thus, the MCA/RDA method indicates the involvement of both hyper- and hypomethylation in leukemogenesis of ATL although in a different manner.

In the case of EGR3 gene, only the coding regions were methylated with little DNA methylation of the promoter region in fresh ATL cells although its expression was suppressed. TSA has more profound effect than 5-aza-dC, suggesting that histone modification, rather than DNA methylation, in the promoter region might silence the transcription of the EGR3 gene. However, because DNA methylation in the coding region was associated with such silencing, detection of DNA methylation in non-promoter regions is also capable of identifying such silenced genes as observed in the EGR3 gene.

EGR3 is a transcriptional factor containing zinc finger domain as well as KLF4. It has been reported that enforced expression of EGR3 gene resulted in expression of FasL in HeLa cells (38), indicating that EGR3 is a critical transcriptional factor for FasL transcription. In agreement with these results, we also showed that expression of EGR3 induced FasL transcription, resulting in apoptosis of ATL cells. Although ATL cells possess a phenotype of activated T cells and highly express Fas antigens on their surfaces, they do not produce FasL. On the other hand, normal T lymphocytes can express both Fas antigens and FasL after activation, and the number of activated T lymphocytes is regulated by Fas-FasL system-mediated apoptosis, which is designated as activation-induced cell death (39). Activation-induced cell death controls the number of activated T lymphocytes, consequently suppressing the immune response. Suppression of EGR3 gene in ATL cells could account for lack of expression of FasL, which enables ATL cells to escape from activation-induced cell death. In the present study, we demonstrated that enforced expression of EGR3 gene-induced FasL expression and apoptosis. Thus, because both KLF4 and EGR3 are accelerators of ATL cell apoptosis, KLF4 and EGR3 genes are considered new tumor-suppressor gene candidates in ATL.

KLF4 is a member of the Kruppel-like factor family, which is highly expressed in epithelial tissues such as the gut and skin, especially in the terminally differentiated cells (40, 41). Previous studies reported that KLF4 plays important roles in the regulation of G₁-S and G₂-M cell cycle checkpoint in colon cancer cells and that these functions are likely to be p53-dependent (42, 43, 44). According to these findings, KLF4 is thought to be associated with tumorigenicity of colon cancer cells. However, there is no report regarding the functional role of KLF4 gene in lymphoid cells. Our study demonstrated that KLF4 expression induced apoptosis of ATL cells. Although the mechanism of apoptosis needs additional study, its silencing by DNA methylation facilitates the survival of ATL cells.

The present study showed that the densities of CpG methylation in identified DNA regions tend to increase with disease progression. Moreover, analysis of sequential samples from a patient who was followed from carrier state until the onset of acute ATL revealed that DNA methylation accumulated at the onset of ATL (Fig. 1 B). These data indicate that serial analysis of the methylation status in identified hypermethylated regions might be a useful tool in the diagnosis and staging of ATL.

HTLV-I-infected clones have been shown to persist over seven years in the same HTLV-I carrier (45), suggesting that HTLV-I-infected cells survive for a long time through the action of viral proteins. On the other hand, it has been reported that aging is closely related to alterations of DNA methylation. Progressive loss of 5-methylcytosine content is observed in normal aging cells, primarily within DNA-repeated sequences. In contrast, some genes show progressive, age-related increases of DNA methylation, resulting in silencing their expressions (46). Taken together, the prolonged life span of HTLV-I-infected T cells might be a predisposing factor for aberrant DNA methylation. During the long latent period, it is possible that HTLV-I-infected cells that are adapted for survival are selected in vivo. In such evolution of infected cells, genetic and epigenetic changes are thought to play critical roles by suppressing the transcription of genes with tumor suppressor functions or activating the expression of genes, which exerts positive effects on survival of tumor cells.

In conclusion, we have demonstrated in the present study that the MCA/RDA method could identify the differentially methylated DNA regions and genes according to disease progression of ATL. Such identification of aberrantly methylated genes allows for the diagnosis and staging of ATL and clarifies the molecular mechanism of leukemogenesis.

We are grateful to Michiyuki Maeda for providing us valuable cell lines and to Yoshihiro Koya for valuable help. The authors also thank Dr. F. G. Issa (word-medex.com.au) for careful reading and editing of the manuscript.