Genome-Wide Transcriptional Response to 5-Aza-2′-Deoxycytidine and Trichostatin A in Multiple Myeloma Cells

Introduction

Previous molecular studies have shown that multiple myeloma (MM) cells from the majority of patients harbor cytogenetic aberrations, particularly illegitimate rearrangements of the immunoglobulin heavy chain gene (IGH) on 14q32, monosomy 13, and deletions of 13q ( 1– 3). In addition to genetic aberrations, there is increasing evidence that epigenetic changes play an important role in the pathogenesis of MM ( 4, 5). Aberrant methylation (referred to as methylation) of gene promoter regions leading to gene silencing is to date the most widely studied epigenetic abnormality in human malignancies ( 6). Costello et al. ( 7) reported that, on average, 600 CpG islands are targets for methylation in malignant diseases. To date, ∼20 cancer-related genes have been identified that are frequently silenced by methylation in MM ( 5, 8, 9). Thus, identification of unknown epigenetically inactivated cancer-related genes in MM is of tremendous importance for a better understanding of the pathogenesis of this disease. In addition to DNA methylation, histone hypoacetylation plays a critical role in epigenetic gene silencing ( 10). Chromatin inactivated by histone H3 and/or H4 deacetylation is associated with silencing of several cancer-related genes in cancer cells ( 11– 13).

In contrast to genetic alterations, epigenetic changes are potentially reversible. DNA methyltransferase inhibitors 5-azacytidine and 5-aza-2′-deoxycytidine (Aza-dC) are analogues of deoxycytidine, which is the target nucleoside for methylation; these drugs have been extensively studied in reactivating by methylation-silenced genes ( 6, 14). Trichostatin A (TSA) is a potent histone deacetylase inhibitor. Moreover, Xiong et al. ( 15) reported that besides HDAC activity, TSA has DNMT3b down-regulating properties, and several studies have shown that TSA is capable of activating the expression of methylation-silenced genes ( 16– 18) and of genes that are silenced by aberrant histone deacetylation in the absence of DNA methylation ( 19). Aza-dC and TSA have been shown to act synergistically in the reexpression of methylated cancer-related genes ( 16). In addition, several studies have shown that Aza-dC and TSA inhibit cancer cell growth ( 20– 22).

Here, we describe the results of a microarray-based genome-wide screen for genes responding to DNA methyltransferase inhibition and HDAC inhibition in MM cell lines. Our study presents a large number of genes whose epigenetic silencing in MM was unknown thus far. Based on the results of our microarray assay, we selected several cancer-related genes whose epigenetic silencing in monoclonal gammopathies has been unknown thus far and analyzed their methylation status in a large number of monoclonal gammopathies of undetermined significance (MGUS) and MM samples. Moreover, the methylation results were compared with certain clinicopathologic variables of the patients. Some of them may be of clinical relevance for MM patients.

Materials and Methods

Clinical specimens. Bone marrow aspirates and clinical data were collected from 24 patients with MGUS and from 111 patients with MM as reported recently ( 5). For control experiments, bone marrow specimens from a healthy bone marrow donor and from nine patients with localized non–Hodgkin's lymphoma without bone marrow infiltration were analyzed. Mononuclear cells were isolated using Histopaque-1077 (Sigma) according to the manufacturer's instructions and stored at −80°C.

Cell culture and treatment. The human MM cell lines MM1, U266, and NCI-H929 were grown in RPMI 1640 supplemented with 10% fetal bovine serum. Cells (2 × 10⁵/mL) were treated either with 0.5 μmol/L Aza-dC for 7 days or with 100 ng/mL TSA for 24 h or with the combination of 0.5 μmol/L Aza-dC for 7 days and 100 ng/mL TSA for additional 24 h. Control cells received no drug treatment.

RNA extraction and cRNA preparation. Total cellular RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instruction. First- and second-strand cDNA synthesis was performed using SuperScript double-stranded cDNA synthesis kit (Invitrogen) and 10 μg of purified RNA according to the manufacturer's protocol with the use of an oligo-dT primer containing a T7 RNA polymerase promoter site. IVT was performed using the BioArray High Yield RNA Transcript Labeling kit (Enzo Life Sciences). Fifteen micrograms of cRNA were fragmented at 94°C for 35 min in a fragmentation buffer (Affymetrix).

Array hybridization and scanning. The microarray assay was carried out using the Affymetrix GeneChip HG-U133A. The hybridization, scanning, washing, and staining procedure was performed as recommended by the manufacturer. The readings from the quantitative scanning were analyzed by the Affymetrix Gene Expression Analysis Software. Hybridization quality control variables were attended according to Affymetrix guidelines for efficient hybridization. All microarray analyses were done in duplicates starting from the same independently treated cell population.

Statistical analysis of microarray data. Affymetrix Microarray Analysis Suite version 5 (MAS5) was used to process the scanned chip images and to generate a cell intensity file for each chip. Statistical analysis was performed using Bioconductor's affy package. ⁶ The significance of changes in the expression level of each gene was evaluated with the CyberT algorithm. ⁷ Differentially expressed genes between treated and untreated groups were selected using following criteria: (a) results derived from drug-treated cells had to be at least 4-fold higher than those derived from control cells in each pairwise comparison; (b) differences in gene expression between drug-treated and control cells had to be statistically significant (P < 0.001); and (c) the detection call had to be present (P) in minimum 1 sample.

Gene Ontology analysis. To classify up-regulated genes into statistically significant overrepresented functional categories Gene Ontology analysis was performed using GoMiner ( 23) comparing the total set of genes represented on the HG-U133A GeneChip and the subset of genes that was up-regulated after drug treatment.

Real-time reverse transcription-PCR. For validation of microarray results, real-time reverse transcription-PCR (RT-PCR) was performed using Taqman Gene Expression Assays (Applied Biosystems) as recommended by the manufacturer.

Nucleic acid isolation and methylation-specific PCR. Genomic DNA was isolated from MM cell lines and from mononuclear cells by digestion with Proteinase K, followed by standard phenol-chloroform extraction and ethanol precipitation ( 24). One microgram of genomic DNA was modified by treatment with sodium bisulfite as reported previously ( 25). Using methylation-specific PCR (MSP) analysis, the methylation status of the genes CPEB1, CD9, GJA1, BCL7c, GADD45G, AKAP12, TFPI2, CCNA1, SPARC, and BNIP3 was analyzed. Primer sequences and PCR conditions of these genes are available upon request.

Immunostaining. Paraffin-embedded bone marrow biopsies were cut in 5-μm-thick sections, dewaxed, and immunostained for SPARC using rabbit anti-SPARC polyclonal antibody at a 1:1,000 dilution (Abcam). Labeling was detected with the Envision Plus Detection Kit (DAKO) as recommended by the manufacturer, and all sections were counterstained with hematoxylin.

Statistical analysis of methylation results. The methylation status of the genes CPEB1, CD9, GJA1, BCL7c, GADD45G, AKAP12, TFPI2, CCNA1, SPARC, and BNIP3 was compared with certain clinicopathologic characteristics from the MM patients including age, gender, β₂-microglobulin, lactate dehydrogenase (LDH), hemoglobin, serum creatinine and calcium levels, type of paraprotein, type of light chain, tumor stage, tumor grade, and deletion of chromosome 13q14. Statistical analysis was performed using χ² test and Fisher's exact test for differences between groups and t tests between means. Overall survival was calculated using Kaplan-Meier log rank testing.

Results

Analysis of Global Gene Expression Pattern of MM Cell Lines after Drug Treatment

Using the DNA microarray technique, we analyzed global changes in gene expression after treatment of three MM cell lines (MM1, U266, and NCI-H929) with Aza-dC, TSA, and a combination of Aza-dC and TSA. The gene expression profile was compared before and after treatment using Affymetrix HG-U133A GeneChip, which contains 22,283 probe sets. Replicate array analysis using RNA from independently treated cell lines revealed a mean correlation coefficient of 0.96 (range 0.93–0.98), indicating high comparability between the arrays.

Identification of genes induced by Aza-dC in MM cell lines. Aza-dC treatment resulted in up-regulation of 92 probe sets (0.4% of 22,283 transcripts analyzed) in the cell line MM1, 65 probe sets (0.2%) in U266, and 220 probe sets (0.9%) in NCI-H929, respectively. The level of induced expression varied from 4.2-fold to a maximum of 147.03-fold in MM1, from 5.04-fold to 131.42-fold in U266 and from 4.28-fold to 137.32-fold in NCI-H929 Aza-dC–treated cells compared with untreated cells. As multiple probe sets for the same gene are present on Affymetrix microarrays, up-regulated probe sets represent 78 unique genes in the case of MM1, 60 unique genes in the case of U266 and 194 unique genes in case of NCI-H929. Overall, 284 unique genes were up-regulated in at least one of the three MM cell lines after exposure to Aza-dC (Supplementary Table S1). ENSEMBL database was used to obtain the genomic sequence, including 5′ region, exons and introns of each of the 284 up-regulated genes. The MethPrimer program ( 26) and the CpG Island Searcher ( 27) were used to determine whether the sequences contain CpG islands. The data showed that 73% of up-regulated genes contained CpG islands within the analyzed regions.

Identification of genes induced by TSA in MM cell lines. We also determined the gene expression profile of the three MM cell lines after exposure to the HDAC inhibitor TSA. TSA treatment resulted in up-regulation of 87 probe sets (0.4% of 22,283 transcripts analyzed) in the cell line MM1, 174 probe sets (0.8%) in U266, and 195 probe sets (0.9%) in NCI-H929, respectively. The level of induced expression varied from 4.39-fold to a maximum of 52.38-fold in MM1, from 4.3-fold to 199.33-fold in U266 and from 4.36-fold to 112.51-fold in NCI-H929 TSA-treated cells compared with untreated cells. Up-regulated probe sets represent 70 unique genes in the case of MM1, 146 unique genes in the case of U266, and 169 unique genes in case of NCI-H929. Overall, 324 unique genes of which 94% contain CpG islands were induced in one or more of the three MM cell lines after exposure to TSA (Supplementary Table S2).

Identification of genes induced by treatment of MM cell lines with both drugs. At last, we determined the gene expression profile of the three MM cell lines after exposure to the combination of Aza-dC/TSA, which resulted in up-regulation of 215 probe sets (1% of 22,283 transcripts analyzed) in the cell line MM1, 276 probe sets (1.2%) in U266, and 254 probe sets (1.1%) in NCI-H929, respectively. The level of induced expression varied from 4.26-fold to a maximum of 147.44-fold in MM1, from 4.35-fold to 295.04-fold in U266, and from 4.43-fold to 209.97-fold in NCI-H929 Aza-dC/TSA-treated cells compared with untreated cells. Up-regulated probe sets represent 180 unique genes in the case of MM1, 242 unique genes in the case of U266, and 216 unique genes in case of NCI-H929. Overall, 470 unique genes of which 88% contain CpG islands were induced in one or more of the three MM cell lines after exposure to the combination of Aza-dC and TSA (Supplementary Table S3).

In addition, the distribution of up-regulated genes was relatively even; however, the highest proportions of induced genes were located at chromosome 6 where a 7 Mb cluster with high density of up-regulated genes was identified (Supplementary Fig. S1A and S1B). This cluster contains genes encoding for MHC class II proteins that are normally constitutively expressed in antigen-presenting cells.

Overlap of up-regulated genes between MM cell lines and drug treatment. To evaluate overlaps of Aza-dC and TSA up-regulated genes between MM cell lines, Venn diagrams were generated which show that there is only a limited overlap of Aza-dC and/or TSA-induced genes in the three MM cell lines ( Fig. 1A ). Four genes were found to be up-regulated in all 3 MM cell lines after Aza-dC treatment, 11 genes after TSA treatment, and 35 genes after Aza-dC/TSA treatment, respectively. By comparing genes up-regulated in two cell lines after Aza-dC, TSA, or Aza-dC/TSA treatment, it is noteworthy that in MM1 and NCI-H929 more commonly up-regulated genes were found than in U266 when compared with MM1 and NCI-H929. Next, we identified genes whose expression was up-regulated as response to a particular drug in each cell line individually ( Fig. 1B). Interestingly, only 2, 1, and 2 genes were commonly up-regulated after Aza-dC and TSA treatment in MM1, U266, and NCI-H929, respectively, indicating that Aza-dC and TSA affect different groups of genes and pathways. In addition, a high percentage (57% in MM1, 67% in U266, and 41% in NCI-H929, respectively) of genes induced by TSA were also found to be up-regulated after Aza-dC/TSA treatment in each cell line. However, the percentage of commonly Aza-dC and Aza-dC/TSA up-regulated genes was much lower (14% in MM1, 28% in U266, and 19% in NCI-H929, respectively; Fig. 1B).

Identification of genes down-regulated after drug treatment. Although at a lower number we also identified genes whose expression was down-regulated in MM cell lines. Treatment of the cell lines with Aza-dC resulted in down-regulation of 76 genes in MM1, of 20 genes in U266, and of 15 genes in NCI-H929, respectively. After TSA treatment, 79 genes were found to be down-regulated in MM1, 107 in U266, and 148 in NCI-H929, respectively. Aza-dC/TSA treatment resulted in down-regulated expression of 149 genes in MM1, 70 genes in U266, and 141 genes in NCI-H929, respectively. Interestingly, by Aza-dC and TSA, down-regulated genes included several proto-oncogenes that have been found to be substantially overexpressed in MM (i.e., cyclin D1, cyclin D2, MAF, MAFF, MAFG, FGFR3, MMSET; members of the myc protein family and PIM2; Table 1 ). Of note, as shown in Table 1, down-regulation of these genes was more a response to TSA and Aza-dC/TSA than to Aza-dC alone. Results from microarray analysis from certain cancer-related genes were validated by real-time RT-PCR ( Tables 1 and 3). Detailed results are available online (Supplementary Table S4).

Functional Analysis by GoMiner

To identify biological processes statistically significantly affected by Aza-dC and TSA genes whose expression was up-regulated >4-fold in at least one MM cell line after treatment with Aza-dC, TSA or Aza-dC/TSA were analyzed using the GoMiner program ( 23). GoMiner dereplicates total and changed input files so that only one instance of a gene name is processed. This resulted in 13,018 reference genes to which 284 genes whose expression was up-regulated after Aza-dC treatment, 324 genes whose expression was up-regulated after TSA treatment, and 470 genes whose expression was up-regulated after Aza-dC/TSA treatment were compared. Detailed results are shown in Table 2 . Aza-dC and TSA were found to up-regulate different groups of genes. Although Aza-dC treatment resulted in statistically highly significant up-regulation of genes involved in response to stimulus and immune response TSA treatment up-regulated genes involved in cell organization and biogenesis. In addition, statistically significant up-regulation of genes involved in apoptosis, cell cycle, cell adhesion, proliferation, and cell migration was observed ( Table 2).

Aza-dC and TSA Affected Cancer-Related Genes

Our approach identified several Aza-dC and/or TSA up-regulated genes that are involved in important cancer-related pathways, including cell cycle, proliferation, apoptosis, and cell adhesion ( Table 3 ). Aza-dC up-regulated genes previously reported to be methylated in MM include CDH1 ( 4, 5), DAPK ( 5), TIMP3 ( 5), and SOCS1 ( 5). More important, cancer-related genes, hitherto unknown to be subject to epigenetic alterations in MM, have been identified (i.e., JUP, BIK, CD9, BCL7c, AKAP12, TFPI2, CCNA1, SPARC, and BNIP3). A summary of these genes is shown in Table 3. Additionally, we found several cancer-related genes induced after combined treatment of MM cell lines with Aza-dC/TSA (i.e., CDKN1A/p21, WIG1, BIK, CGREF1, JUP, and IGFBP3). Interestingly, treatment of MM cell lines with TSA alone also resulted in up-regulation of important cancer-related genes including ING1^p33, TIMP3, CDKN2D, PTENP1, CDKN1C/p57^KIP2, DAPK, and WIG1, suggesting that aberrant histone deacetylation is an important mechanism for inactivation of cancer-related genes in MM.

Methylation Analysis of 10 Cancer-Related Genes in MM Cell Lines by MSP

Based on our microarray data, we selected 10 genes whose expression was up-regulated after drug treatment of MM cell lines and analyzed their methylation status in six MM cell lines. The genes analyzed were CPEB1, CD9, GJA1, BCL7c, GADD45G, AKAP12, TFPI2, CCNA1, SPARC, and BNIP3. All of these genes were identified as having a CpG island in their 5′ regions. We found methylation of all of these genes in the six MM cell lines at varying frequencies ( Fig. 1C). The most frequently methylated gene was BCL7c (methylated in 100%) followed by AKAP12, SPARC, and TFPI2 (83%, respectively); CPEB1 (67%); GJA1, CCNA1, and BNIP3 (50%, respectively); CD9 (33%); and GADD45G (17%). Aberrant methylation of up to 9 of the 10 genes per cell line was observed.

Methylation Analysis of 10 Cancer-Related Genes in Samples from Patients with MGUS and MM

To investigate if methylation of the 10 genes was not only a phenomenon in MM cell lines, we performed MSP analysis of the 10 genes also in samples from 24 MGUS and 111 MM patients. As shown in Fig. 1E, the frequency of methylation of the 10 genes was lower in MGUS samples (17% for CPEB1, 16% for CD9, 13% for GJA1, 8% for BCL7c, 5% for AKAP12, and 4% for BNIP3; no methylation was found for GADD45G, TFPI2, CCNA1, and SPARC) compared with MM samples (50% for CPEB1, 28% for CD9, 23% for GJA1, 21% for BCL7c, 19% for GADD45G, 13% for AKAP12, 10% for TFPI2, 8% for CCNA1, 8% for SPARC, and 5% for BNIP3). This difference was statistically significant in the cases of CPEB1 (P = 0.003, χ² test) and GADD45G (P = 0.025, Fisher's exact test). Examples of MSP results are shown in Fig. 1D. In addition, bone marrow specimens from 10 control persons were analyzed as negative controls for methylation of the 10 genes. Methylation of at least one gene was detected in 73% of MM samples including methylation of one gene in 31% of MM samples, methylation of two genes in 16% of MM samples, methylation of three genes in 6% of MM samples, methylation of four genes in 8% of MM samples, methylation of five genes in 5% of MM samples, methylation of six genes in 2% of MM samples, methylation of seven genes in 3% of MM samples, and methylation of eight genes in 2% of MM samples.

Comparison of Clinicopathologic Characteristics from MM Patients with Results of Methylation Analysis

We next analyzed possible correlations between methylation data and clinicopathologic variables of the MM patients. Factors analyzed included age and gender of the patients, β₂-microglobulin, LDH, hemoglobin, serum creatinine and calcium levels, type of paraprotein, type of light chain, tumor stage, tumor grade, and chromosome 13q14 deletion at the time of diagnosis. Factors that have been found to statistically significantly correlate with methylation of a certain gene include age, type of paraprotein and light chain, and deletion of chromosome 13q14 ( Table 4 ). In addition, 105 patients with MM were used for overall survival analysis. We observed that methylation of BNIP3 and SPARC correlated with a poor overall survival of the patients (P = 0.003, P = 0.017, respectively; log-rank test; Fig. 2A ). Moreover, BNIP3 and/or SPARC methylation that has been observed in 12% of MM patients statistically significantly correlated with a poor overall survival of these patients (P = 0.0004). Additionally, immunostaining of bone marrow biopsies of MM patients whose bone marrow specimen were SPARC methylated (n = 4) and of patients whose bone marrow specimen were SPARC unmethylated (n = 3) was performed. Interestingly, in all samples that were found to be methylated myeloma cells did not express SPARC. In two of three samples that were found to be SPARC unmethylated, myeloma cells showed homogeneous SPARC expression. In the remaining sample, myeloma cells did not express SPARC ( Fig. 2B). The methylation status of any of the other 10 genes had no prognostic effect. As reported previously, deletion of chromosome 13q is associated with a poor prognosis of MM patients ( 1, 3). This finding was also observed in our cohort of patients (P = 0.002).

Discussion

Epigenetic silencing of cancer-related genes is a frequently occurring event in carcinogenesis ( 6). To date, only a few genes are known to be aberrantly methylated in MM. In an effort to identify additional epigenetically silenced genes in MM cells and to better understand the effects of Aza-dC and TSA on gene expression of MM cells, we examined expression profiles of three MM cell lines treated with Aza-dC and TSA, either alone or in combination using the microarray technique. Moreover, we compared these results with expression results from untreated cells. Additionally, using MSP, we investigated aberrant methylation patterns of 10 selected genes in MM cell lines and samples from patients with MGUS and MM.

Our microarray data revealed that expression of many genes was up-regulated after Aza-dC and/or TSA treatment of MM cells. The number of induced genes was very similar as reported for other cell types ( 28– 30). Using the online tools MethPrimer ( 26) and CpG Island Searcher ( 27), we found that 73% of Aza-dC up-regulated genes contained a CpG island. This is in agreement to other reports that showed that inhibiting DNA methylation has both direct and indirect effects on gene expression ( 30, 31). Interestingly, we observed that 94% of genes whose expression was induced by TSA contained a CpG island, suggesting that TSA more specifically induces expression of CpG island containing genes than Aza-dC. The difference between Aza-dC and TSA up-regulated CpG island positive genes was statistically significant (P = 0.0000000007, χ² test).

We observed that a few genes show an overlap in the up-regulation between the three cell lines. In agreement to previous reports demonstrating that each tumor contains up to 600 methylated genes, we also observed a high number of epigenetically regulated genes in each cell line ( 7). We hypothesize that this pattern represents the individual character of each cell line; however, we believe that the commonly altered genes may be “key players” in the pathogenesis of MM.

Increased expression of proto-oncogenes has been described for many different malignancies. One of the novel findings of our present study is that several proto-oncogenes that are involved in the pathogenesis of MM were down-regulated after treatment of MM cells with Aza-dC and TSA ( Table 1). Down-regulation may be explained either by a direct inhibitory effect of Aza-dC and TSA or by an indirect down-regulation by Aza-dC and TSA affected genes. Proto-oncogenes that are frequently overexpressed in MM include CCND1, MAF, FGFR, and MMSET ( 32– 34). In our microarray assay, MAF is substantially down-regulated by TSA and Aza-dC/TSA in all three MM cell lines analyzed. In addition, expression of target genes (CCND2, CCR1, and ITGB7) of the MAF transcription factor was down-regulated in MM cell lines after TSA and Aza-dC/TSA treatment, suggesting that TSA primarily inhibits MAF and secondarily down-regulates CCND2, CCR1, and ITGB7. Similar results were found for FGFR3, which promotes MM cell proliferation and antiapoptosis and MMSET. We found expression of both genes strongly down-regulated in response to TSA and Aza-dC/TSA in all three MM cell lines analyzed. In addition, we observed that components of the IL-6 signaling pathway are affected by Aza-dC and/or TSA providing an additional mechanism to inhibit proliferation and induce apoptosis of MM cells. The IL-6/IL-6R complex associates with gp130 (IL-6ST), which was found to be up-regulated in myeloma cells ( 35), resulting in promotion of cell proliferation and survival. We found that TSA and Aza-dC/TSA treatment of U266 cells, which are known to produce IL-6 ( 36), resulted in marked down-regulation of IL-6R. Moreover, although down-regulation did not pass our filtering criteria, gp130 down-regulation was observed. Interestingly, Mitsiades et al. ( 37) reported that treatment of MM cells using the HDAC inhibitor SAHA resulted in down-regulation of both IL-6R and gp130, suggesting that inhibition of IL-6 pathway using HDAC inhibitors might be a new strategy in the future treatment of MM patients.

By comparing expression profiles of Aza-dC– and/or TSA-treated cells and untreated cells, we identified a large number of up-regulated cancer-related genes that are involved in several cancer-associated pathways, including cell cycle, cell growth, and apoptosis ( Tables 2 and 3). Thus far, epigenetic silencing was not known for most of these genes. We selected 10 genes (CPEB1, CD9, GJA1, BCL7c, GADD45G, AKAP12, TFPI2, CCNA1, SPARC, and BNIP3) whose expression was up-regulated after drug treatment and determined their methylation status in 6 MM cell lines, 24 samples from patients with MGUS, and 111 samples from patients with MM by MSP. Our methylation results support previous studies that showed that methylation of multiple cancer-related genes is a frequently occurring event in the pathogenesis of MM. The most frequently methylated gene in MM samples was CBEP1, a gene whose epigenetic silencing in malignant diseases has not been reported thus far. CPEB1 belongs to the cytoplasmic polyadenylation element binding protein family. Besides its function as translational activator, it has been reported that CPEB1 has also translational repressor properties by inducing stress granules ( 38). By comparing our methylation data with clinicopathologic characteristics from the MM patients, we observed that CPEB1 methylation was already present in MGUS samples but at a statistically significantly higher frequency in MM samples (P = 0.003), suggesting that CPEB1 methylation may be an early event in the development of monoclonal gammopathies. In addition, although the difference in overall survival between patients with and without CPEB1 methylation was statistically not significant, we observed a trend toward a poor prognosis for MM patients with CPEB1 methylation. Additional studies are necessary to elucidate the role of CPEB1 in the pathogenesis of MM.

Interestingly, we observed that methylation of two genes, SPARC and BNIP3, was associated with a poor overall survival of MM patients ( Fig. 2A), suggesting that methylation of these genes may be a potential biomarker for prognosis in patients with MM. SPARC is a matrix-associated protein that modulates cellular interaction with the extracellular matrix and inhibits cellular proliferation ( 39). Its role in tumorigenesis is complex, and expression is often being down-regulated in tumor cells accompanied by up-regulation in juxtatumoral stromal cells. It has been suggested that acting of SPARC either as tumor suppressor or oncogene may depend on the tumor stage ( 40). Recent reports showed that SPARC methylation is associated with loss of SPARC expression in lung, pancreatic, and colon cancer ( 40– 42). We investigated SPARC protein expression in bone marrow biopsies of a small group of MM patients by immunostaining. Whereas all SPARC methylated samples did not express SPARC in the myeloma cells, all except one unmethylated sample showed homogeneous SPARC protein expression in the myeloma cells. Although the number of samples investigated was very small, our results show that SPARC methylation is associated with loss of SPARC protein expression, suggesting the hypothesis that SPARC acts as a tumor suppressor in MM. BNIP3 is a proapoptotic member of the Bcl-2 protein family involved in hypoxia-induced cell death. Loss of BNIP3 expression due to DNA methylation has been reported recently in colorectal and pancreatic cancer ( 43, 44). In addition, Murai et al. ( 45) found BNIP3 methylation in 21% (3 of 14) of MM samples. However, the frequency of BNIP3 methylation in our study was much lower (5%). This discrepancy may be explained by the different sample sizes analyzed. Our results suggest that MM cells of some patients escape hypoxia-induced apoptosis through down-regulating BNIP3 by methylation and that these cells are more aggressive than BNIP3 unmethylated MM cells.

In conclusion, we show that expression of a large number of genes that are involved in important cancer-related pathways is affected by Aza-dC and TSA. Methylation of certain genes whose epigenetic silencing in MM has not been reported thus far might be important in the development and progression of MM. In addition, our results suggest that inhibition of DNA methyltransferases and HDACs might be new therapeutic strategies in the future treatment of MM patients.