Fusion of the SS18 and either one of the SSX genes is a hallmark of human synovial sarcoma. The SS18 and SSX genes encode nuclear proteins that exhibit opposite transcriptional activities. The SS18 protein functions as a transcriptional coactivator and is associated with the SWI/SNF complex, whereas the SSX proteins function as transcriptional corepressors and are associated with the polycomb complex. The domains involved in these opposite transcriptional activities are retained in the SS18-SSX fusion proteins. Here, we set out to determine the direct transcriptional consequences of conditional SS18-SSX2 fusion protein expression using complementary DNA microarray-based profiling. By doing so, we identified several clusters of SS18-SSX2–responsive genes, including a group of genes involved in cholesterol synthesis, which is a general characteristic of malignancy. In addition, we identified a group of SS18-SSX2–responsive genes known to be specifically deregulated in primary synovial sarcomas, including IGF2 and CD44. Furthermore, we observed an uncoupling of EGR1, JUNB, and WNT signaling in response to SS18-SSX2 expression, suggesting that the SWI/SNF-associated coactivation functions of the SS18 moiety are impaired. Finally, we found that SS18-SSX2 expression affects histone modifications in the CD44 and IGF2 promoters and DNA methylation levels in the IGF2 imprinting control region. Together, we conclude that the SS18-SSX2 fusion protein may act as a so-called transcriptional “activator-repressor,” which induces downstream target gene deregulation through epigenetic mechanisms. Our results may have implications for both the development and clinical management of synovial sarcomas. (Cancer Res 2006; 66(19): 9474-82)
Synovial sarcomas are aggressive soft tissue tumors that account for up to 10% of all human sarcomas (1). Cytogenetically, synovial sarcomas are characterized by a recurring chromosomal translocation, t(X;18)(p11.2;q11.2), which is found in >95% of all cases (2). As a result of this translocation, the SS18 gene (previously called SYT or SSXT) on chromosome 18 is fused to one of three closely related SSX genes on the X chromosome, SSX1, SSX2, or SSX4 (3–6).
The human and mouse SS18 genes and their encoded proteins are highly homologous and are expressed widely, both in embryonic and in adult tissues (7, 8). The SS18 protein resides in the nucleus, displaying a punctate pattern, distinct from nuclear domains such as coiled bodies, splicing factor speckles, or PML oncogenic domains (9, 10). Within the SS18 protein, two functional domains have been identified, the SS18 NH2-terminal homology (SNH) domain, and the glutamine-, proline-, glycine-, and tyrosine-rich QPGY domain (11, 12). Previously, we found that the SS18 protein interacts directly with the AF10 protein, the fusion partner to MLL in t(10;11)-positive acute leukemias (12). Through this interaction, the SS18 protein associates with the glioma-amplified protein GAS41 (13) and the rhabdomyosarcoma-associated SWI/SNF protein INI1 (14). SWI/SNF proteins form multimeric complexes that activate gene transcription through ATP-dependent chromatin remodeling mechanisms (15). Interestingly, the SS18 protein also interacts and colocalizes with two SWI/SNF ATPase subunits, i.e., the Brahma (BRM) and Brahma-related gene 1 (BRG1) proteins (11, 16). In addition, the SS18 protein interacts directly with the histone acetyltransferase p300 (17) and the histone deacetylase–associated corepressor SIN3A (18). Both latter proteins are involved in epigenetic gene regulation through covalent chromatin modifications, in particular (de-)acetylation of histone tails. Because all these interactions are mediated by the SS18 SNH domain, this domain emerges as a versatile protein-protein interaction domain. The SS18 QPGY domain acts as a transcriptional activation domain and is able to mediate multimerization of the SS18 protein (10, 11, 16). In addition, this domain shares structural characteristics with two SWI/SNF proteins, i.e., ARID1A (BAF250a/SMARCF1) and ARID1B (ELD/OSA1/BAF250b; ref. 14). Interestingly, the SS18-SWI/SNF association is also found in the Caenorhabditis elegans protein-protein interaction map (19), indicating that its functional significance is conserved during evolution. The lack of obvious DNA-binding domains, combined with the QPGY-mediated transcriptional activation capacities, indicate that the SS18 protein acts as a transcriptional coactivator. The SNH domain–mediated protein-protein interactions may serve to tether SS18 to its DNA targets.
Currently, nine human SSX genes, encoding highly similar proteins, have been identified (20). The NH2-terminal moieties of the SSX proteins exhibit homology to the Krüppel-associated box (KRAB) domain (5), a domain that is known to be involved in transcriptional repression (21). Although SSX proteins are able to repress the transcription of reporter genes (5, 11), only part of this activity has been attributed to the KRAB-like domain. Considerably stronger transcriptional repression was exerted by the highly acidic COOH-terminal 34 amino acids of SSX (22). Therefore, this sequence was designated SSX repression domain. The SSX proteins are localized in the nucleus, being distributed both diffusely and in nuclear speckles (9, 10). These speckles were found to also harbor several polycomb group (PcG) proteins, i.e., HPC2, BMI1, and RING1 (23, 24). PcG proteins form multimeric protein complexes that induce the repression of target genes through modulation of chromatin structures (25).
Because the wild-type SS18 and SSX proteins exhibit apparently opposite transcriptional activities, and because the domains critical to these activities are retained in the synovial sarcoma–associated SS18-SSX fusion proteins, we questioned what the transcriptional consequences of SS18-SSX expression are. Therefore, we set out to assess the SS18-SSX2 downstream effects using an inducible cell system in conjunction with microarray-based gene expression profiling, chromatin immunoprecipitation (ChIP), and global and gene-specific histone modification and DNA methylation assays. Our data indicate that the SS18-SSX2 fusion protein induces the deregulation of downstream target genes through epigenetic mechanisms.
Materials and Methods
Cloning and sequencing procedures. All cloning procedures were essentially as described before (7, 8, 12). The sequence of all oligos used are available upon request. Sequence analyses were done at the DNA Sequencing Facility of the Radboud University Nijmegen Medical Centre. DNA and protein databases were searched using BLAST and/or BLAT search algorithms at the National Center for Biotechnology Information or University of California, Santa Cruz (UCSC), respectively (http://www.ncbi.nlm.nih.gov/; http://genome.ucsc.edu/). All SS18, SSX2, and SS18-SSX2 fragments used as probes on Northern blots and/or for the construction of the T-Rex tetracyclin-inducible cell system were generated by PCR from full-length complementary DNAs (cDNA) and sequence-verified prior to probe preparation and/or cloning procedures.
Conditional expression of the SS18-SSX2 fusion protein and real-time PCR analysis. For the identification of SS18-SSX downstream target genes, we generated a cell line exhibiting conditional expression of the SS18-SSX2 fusion protein using the T-Rex system (Invitrogen, Carlsbad, CA). Human HEK293-T-Rex cells were purchased from Invitrogen and cultured in T-Rex culture medium (DMEM with 10% fetal bovine serum and 1 μg/mL blasticidin). A full-length SS18-SSX2 cDNA was cloned into the pcDNA4-TO vector, thereby yielding a construct in which the T-Rex tetracyclin-inducible promoter controls the expression of a myc-tagged SS18-SSX2 protein. This construct, and the empty vector, were transfected into the HEK293-T-Rex cells using a gene-pulser electroporation apparatus (Bio-Rad, Hercules, CA). Transfected cells were selected in T-Rex culture medium supplemented with 1 μg/mL of zeocin (Invitrogen). After 2 weeks of selection, positive clones were isolated and pooled, such that each pool was a mix of 10 to 15 independent clones, thereby minimizing the transcriptional effects of independent integration sites. For induction, 1 ng/mL of tetracyclin (Sigma-Aldrich, St. Louis, MO) was added to the HEK293-T-Rex empty vector (HEK-VEC) and HEK293-T-Rex SS18-SSX2 (HEK-FUS) cell lines for 2, 6, or 24 hours. Cells were harvested by scraping in PBS, and RNA was isolated using a Trizol protocol (Boehringer, Ridgefield, CT) in conjunction with RNeasy midi columns (Qiagen, Valencia, CA). Northern blots were prepared as described previously (7). cDNA was prepared from 1 μg of RNA, using the Omniscript reverse transcription kit (Qiagen) in a 20 μL volume. Real-time quantitative reverse transcription-PCR (QRT-PCR) was done in duplicate with 0.2 μL of reverse transcriptase material on an ABI 7700 sequence detection system, using a SYBR green PCR master mix (Applied Biosystems, Foster City, CA). All QRT-PCR data were normalized to the expression of the β-tubulin gene.
Microarray-based expression profiling. RNA labeling, microarray preparation, hybridization, scanning, and image analysis were done as described before (26). For each hybridization, 100 μg of total RNA was fluorescently labeled with Cy3 (HEK-VEC) and Cy5 (HEK-FUS) using direct dye incorporation by oligo(dT)-primed reverse transcriptase. Annotations of the gene names for each cDNA clone are according to Build 170 of the UniGene human sequence collection (http://www.ncbi.nlm.nih.gov/UniGene/). The scanned fluorescent images (red and green channels) constituted the raw data from which differential gene expression ratio values were calculated. For all targets, the ratios of the red (Cy5) to green (Cy3) intensities (R/G) were determined, and ratio normalizations were done based on 88 preselected internal controls as previously described (26). Spots were included for further analyses when either the quality scores (assigned by the DeArray software) were >0.5, or the signal-to-background ratios in both channels were >3. In addition, data were excluded when the normalized ratios between duplicate experiments varied by more than a factor of 2. The ratio threshold values, indicating down-regulation or up-regulation of the genes represented by the cDNA clones on the array, were set to −0.6 and 0.6, respectively, which corresponds to a 50% difference in gene expression.
Western blotting and immunoprecipitation. Protein extracts were prepared from the HEK-FUS cells after 0, 2, 6, and 24 hours of tetracyclin induction in extraction buffer containing 150 mmol/L of NaCl, 0.5% NP40, 50 mmol/L of HEPES (pH 7), 5 mmol/L of EDTA, 1 mmol/L of phenylmethylsulfonyl fluoride, 0.5 mmol/L of DTT, 2 mg/mL of leupeptin, and 2.5 mg/mL of aprotinin. Protein extracts were analyzed on a Western blot, incubated with a mouse monoclonal anti-myc antibody (9E10; Santa Cruz Biotechnology, Santa Cruz, CA) and a horseradish peroxidase–coupled rabbit anti-mouse IgG antibody (Dako, Glostrup, Denmark) as a secondary antibody. Signals were detected with the enhanced chemiluminescence system (Amersham, Piscataway, NJ) and X-omat R X-ray films (Kodak, New Haven, CT). For the SS18-SSX2/BRG1 immunoprecipitations, fresh nuclear extracts were prepared from HEK-FUS cells, according to the Lamond lab protocol (http://www.lamondlab.com/). Approximately 400 μg of nuclear proteins were subjected to immunoprecipitations with 1.6 μg of 9E10 anti-myc antibody and a mouse preimmune serum in the presence of protein A/G agarose beads (Santa Cruz). After washing, the whole immunoprecipitates, and 40 μg of the nuclear extract as an input control, were loaded on a Western blot. This blot was incubated with an anti-BRG1 antibody (sc-10768, Santa Cruz), after which the signals were detected as described above. Next, the blot was stripped (5 minutes in 0.2 mol/L of glycine, pH 2.8, and 0.5 mol/L of NaCl), neutralized (5 minutes in 0.5 mol/L of Tris and 1.5 mol/L of NaCl), and incubated with the anti-SS18 antibody RA2009 (9). The signals were detected as described above.
ChIP analysis. Chromatin was isolated from HEK-FUS cells at 0 and 24 hours of tetracyclin induction as described by Boyd et al. (27). Protein-DNA complexes were immunoprecipitated with antibodies directed against BRG1 (see above), histone H4 acetylated at lysine 16 (H4K16-Ac, ab-1762; Abcam, Cambridge, MA), histone H4 methylated at lysine 20 (H4K20-Me, ab9051; Abcam), histone H3 trimethylated at lysine 4 (H3K4-triMe, ab8580; Abcam), and a rabbit preimmune serum. The recovered genomic DNAs were analyzed by Q-PCR with primer pairs spanning the IGF2 or CD44 promoters. A primer pair located 2 to 3 kb in a downstream intronic region was used as a negative control. All Q-PCR data are presented as promoter/intron ratios (P/I) to indicate the enrichment of IGF2 and/or CD44 promoter sequences in the ChIP.
Bisulfite sequencing of genomic DNA. Genomic DNAs from the HEK-VEC and the HEK-FUS cell lines (after 0 and 24 hours of tetracyclin induction) were treated with bisulfite as described by Frommer et al. (28). Briefly, 2 μg of total genomic DNA was diluted in 50 μL of H2O, after which 5.5 μL of 2 mol/L NaOH was added and incubated at 37°C for 10 minutes. Subsequently, 30 μL of 10 mmol/L hydroquinone (Sigma) and 520 μL of 3 mol/L sodium bisulfite (Sigma), both freshly prepared, were added and the mixture was incubated at 50°C for 16 hours. Next, this (sulfonated) DNA was isolated using a wizard cleanup kit (Promega, Madison, WI) and deaminated by the addition 1/10 volume of 3 mol/L NaOH and incubation for 5 minutes at 25°C. Finally, the bisulfite-adapted DNA was precipitated and resuspended in 20 μL of H2O. For each PCR reaction, 1 μL of the treated DNA was used. Primers for bisufite sequencing of genomic targets were selected using the MethPrimer program (29). Bisulfite PCR products were cloned into pGEM-T and at least 20 independent clones were sequenced for evaluation of the methylation status of the selected target.
Results and Discussion
Identification of SS18-SSX2 target genes. For the identification of SS18-SSX2 target genes, we generated a stable HEK293-T-Rex cell line displaying tetracyclin-inducible SS18-SSX2 expression (HEK-FUS, see Materials and Methods). At four different time points of tetracyclin induction (t = 0, 2, 6, and 24 hours), total RNA and protein were extracted from the HEK-FUS cells and the inducible SS18-SSX2 expression was monitored on Northern and Western blots. On Northern blots, using SS18- and SSX2-specific probes, the SS18-SSX2 mRNA was detected after 2 hours of induction (Fig. 1A, and B; closed arrowheads). In addition, we detected endogenous SS18 mRNA expression, which does not change during this induction series (Fig. 1A,, open arrowhead). Using real-time QRT-PCR on the same RNA samples, we already detected a low SS18-SSX2 mRNA expression level prior to tetracyclin induction (Fig. 2C), indicating that the HEK293-T-Rex cell system is somewhat leaky (confirmed by the manufacturer). On Western blots, using an anti-myc antibody, a 65 kDa protein, representing the induced myc-tagged SS18-SSX2 protein was detected in the induction series (Fig. 1C,; closed arrowhead), thereby validating the inducible SS18-SSX2 mRNA and protein expression in this cell system. Upon longer exposure, the SS18-SSX2 protein was detectable at low levels in the HEK-FUS t = 0 extract as well (data not shown). As expected, HEK-VEC displayed no SS18-SSX2 mRNA expression (Fig. 2C). As an independent variable of SS18-SSX2 expression, we found that the HEK-FUS cell line grows significantly faster than the HEK-VEC cell line (doubling time, 37.7 versus 50.5 hours, respectively). This increase in growth rate is in full conformity with the putative role of SS18-SSX in cellular transformation (30).
Total RNAs extracted from the HEK-VEC and HEK-FUS cells (both tetracyclin-induced for 0, 2, 6, and 24 hours) were labeled and used for expression profiling on NHGRI 6K cDNA microarrays (26). In a total of nine hybridizations, the Cy5-labeled HEK-FUS and Cy3-labeled HEK-VEC samples were matched for each time point, and hybridized in duplicate (t = 0A and B, t = 2A and B, t = 6A and B) or triplicate (t = 24A, B, and C). Of the 6,548 cDNA clones on the array, 4,085 (62%) complied with our inclusion criteria (see Materials and Methods) in at least six out of nine hybridizations done. The normalized, log2-transformed Cy5/Cy3 ratios of these cDNA clones were combined in one database and the mean ratios from duplicate hybridizations at each time point were used for clustering purposes. With the preset threshold levels of −0.6 and 0.6 (see Materials and Methods), 795 SS18-SSX2–responsive genes were identified, each displaying up-regulation or down-regulation at one or more time points (Supplementary Table S1). Within this group of genes, two major clusters could be discerned, each encompassing a set of genes displaying specific transcriptional responses upon SS18-SSX2 induction (Fig. 2A and B). The first cluster included 180 genes which were up-regulated or down-regulated at all time points. A limited subset of this cluster, comprising 64 genes, displayed mean log2 ratios meeting more stringent criteria at all time points (i.e., threshold levels of −0.8 and 0.8). In Fig. 2A, the ratios of all genes in this latter cluster are depicted in a red/green diagram. Given the low, but detectable, SS18-SSX2 mRNA and protein levels already present before induction (t = 0; see above), these gene expression changes illustrate the sensitivity of this cell system for the expression of exogenous genes, especially when these genes encode proteins involved in transcription regulation. The second cluster included 72 genes, all displaying mean log2 ratio changes of at least 0.4 between two consecutive time points (corresponding to 30% up-regulation or down-regulation). These changes were sustained during later time points. In Fig. 2B, the ratios of all genes in this latter cluster are depicted in a red/green diagram. Because this cluster, as expected, included the up-regulated SS18 and SSX genes (Fig. 2B,, closed arrowheads), an intrinsic validation of the microarray data was thus obtained. In addition, we did QRT-PCR on six genes, i.e., the induced SS18-SSX gene, and the LDLR, CD44, EGR1, CXXC5, and IGF2 genes (Fig. 2A, and B, arrows). All QRT-PCR data were in accordance with the microarray data, thereby providing a further validation of the microarray data (Fig. 2C). Validation for the correct annotation of the spotted cDNA clones came from the observation that sequencing of 30 selected clones yielded the expected sequence in 25 of these clones (83%).
SS18-SSX2 induced deregulation of genes involved in cholesterol synthesis. Within the cluster of genes that was up-regulated or down-regulated at all time points (first cluster, Fig. 2A), a set of genes that is involved in cholesterol metabolism could readily be identified. This set includes the main regulator (low-density lipoprotein receptor, LDLR) and the main rate-limiting enzyme (3-hydroxy-3-methylglutaryl coenzyme-A reductase, HMGCR), as well as two other genes which encode enzymes (IDI1 and FDFT1) of the de novo cholesterol synthesis pathway (Fig. 3). Based on this information, we re-queried the entire data set for additional genes involved in cholesterol metabolism and its regulation. By doing so, we found transcriptional changes that substantiate an up-regulation of cholesterol synthesis in response to SS18-SSX2 expression.
Normally, exogenous cholesterol activates the LDL receptor, which induces down-regulation of the HMGCR gene and, thereby, inhibition of de novo cholesterol synthesis (Fig. 3). The LDLR gene, in turn, is regulated through association of the early growth response 1 (EGR1) and CCAAT/enhancer binding protein-β (CEBPB) proteins (31). Another, distinct, function of the CEBPB protein is the regulation of amino acid synthesis through asparagine synthetase (ASNS; ref. 32). In the microarray data, we observed down-regulation of the CEBPB gene in response to SS18-SSX2 induction. This CEBPB down-regulation was accompanied by down-regulation of the LDLR and ASNS genes and by up-regulation of the HMGCR gene (Fig. 3), which is in conformity with literature data (31, 32). Additionally, we observed the up-regulation of five out of nine genes encoding enzymes of the cholesterol synthesis pathway (Fig. 3). Furthermore, the EGR1 gene displayed a 3-fold up-regulation at t = 0, but dropped down to threshold levels after SS18-SSX2 induction, thereby adding to the effect of LDLR down-regulation and, hence, HMGCR up-regulation. The microarray data were confirmed by QRT-PCR (see above, Fig. 2C). Taken together, our expression profiling data indicate that the regulation of various genes involved in cholesterol metabolism is altered as a direct consequence of SS18-SSX2 fusion gene expression. Because enhanced cholesterol synthesis has been denoted as a hallmark of tumorigenesis (33), these responses underscore a previous notion that SS18-SSX2 fusion gene expression may be directly related to the malignant transformation of cells (30). Interestingly, similar elevated expression of genes involved in cholesterol synthesis has also been associated with elevated IGF2 expression (34).
SS18-SSX2-induced deregulation of IGF signaling. Within the cluster of genes displaying up-regulation or down-regulation upon induction of SS18-SSX2 expression (second cluster, Fig. 2B), a set of genes was identified that has previously been implicated in tumor development, i.e., IGF2, NF2, JUNB, CCND1, CAV1, CD44, ENPP2, IGFBP5, and IGFBP7. Comparison of our data with data generated in two independent expression profiling studies of primary synovial sarcomas (26, 35) yielded a limited set of genes that was consistently up-regulated or down-regulated, both in our inducible cell system and in the primary tumors, and which included IGF2 and CD44. The up-regulation of the IGF2 gene in our system was accompanied by a down-regulation of the IGF binding protein 5 (IGFBP5) gene. Within the group of IGF-associated genes, this opposite regulation was restricted to the IGF2 and IGFBP5 genes, each represented by two independent cDNA clones on the array (Fig. 4). Complementary expression of these two genes has also been observed during rat pituitary development (36) and in early chicken embryogenesis, where IGFBP5 expression was observed in a pattern similar to Sonic Hedgehog expression (37). Similar transcriptional switch responses, i.e., ratio changes of at least 0.6, either between the 2- and 6-hour time points or between the 6- and 24-hour time points, were observed for 21 genes (Fig. 2B , open arrowheads). These genes included the JUNB gene and two JUN-inducible genes (38), autotaxin (ENPP2), and fatty acid synthase (FASN). Interestingly, we found that this activator-responder combination was uncoupled in response to SS18-SSX2 expression induction (JUNB up-regulated, and ENPP2 and FASN down-regulated). Because a similar effect was observed for the EGR1 gene (down-regulated) and two EGR1-inducible genes (39), IGF2 and CCND1 (both up-regulated), the activatory functions of JUNB and EGR1 seem to be impaired upon SS18-SSX2 induction. An explanation for this phenomenon may be that the transcriptional coactivation machinery, recruited to these transcription factors, functions differently in the presence of the SS18-SSX2 protein. Because the activator protein transcription factor JUNB may recruit members of the SWI/SNF complex for transcriptional coactivation (40), and because the SS18-SSX2 fusion protein retains the domain necessary for the interactions with the SWI/SNF complex, this impaired transcriptional activation may be caused by compromised SWI/SNF functions.
Epigenetic regulation of SS18-SSX2 target genes. From the genes displaying up-regulation or down-regulation after induction of SS18-SSX2 expression (second cluster, Fig. 2B), we selected the IGF2 and CD44 genes for further analysis. For the IGF2 gene, which displayed the most extreme up-regulation in our microarray data, we confirmed a strong up-regulation at t = 24 (Fig. 2C). For the CD44 gene, we confirmed that it is already down-regulated at t = 0, and is repressed even more in response to SS18-SSX2 expression induction at t = 24 (Fig. 2C). We also did semiquantitative RT-PCRs for the CD44 gene on a panel of primary synovial sarcomas, including 11 SS18-SSX1-positive and 11 SS18-SSX2-positive cases. By doing so, we found low to undetectable CD44 expression levels in three of the SS18-SSX1-positive tumors (27%) and in six of the SS18-SSX2-positive tumors (54%; data not shown). These results support the microarray data obtained in our inducible cell system, and substantiate the notion that the IGF2 and CD44 genes may represent genuine SS18-SSX target genes.
Previously, it was found that CD44 gene expression is up-regulated by the WNT signaling pathway (41) or through DNA demethylation after treatment of cells with 5-aza-2-deoxycytidine (42). Using this information, we again re-queried our microarray data set to search for other genes meeting these criteria. WNT target genes include Axin, CD44, CCND1, C-MYC, FN1, and TCF1, which are activated by β-catenin, in conjunction with TCF transcription factors and SWI/SNF proteins such as BRG1 (43). In our data set, the CCND1 gene showed up-regulation, whereas the CD44, C-MYC, and FN1 genes were either down-regulated, or not affected at all. Additionally, we observed up-regulation of the CXXC5 gene (Fig. 2), of which the encoded protein displays a high degree of similarity to the WNT-inhibiting protein IDAX, especially around the KTXXXI motif (44, 45). Activation of a putative WNT repressor such as the CXXC5 may be responsible for the observed SS18-SSX2–induced down-regulation of CD44. However, in line with the transcriptional uncoupling processes described above, the observed CD44 down-regulation may also result from compromised SWI/SNF functions. Previously, CD44 (up)regulation was found to rely specifically on BRG1 protein expression (43). An important premonition for any inhibitory effect exerted by the SS18-SSX2 protein on the SWI/SNF protein BRG1 would be that the SS18-SSX2 and BRG1 proteins interact directly in our cell system. To establish this, we did an immunoprecipitation of myc-tagged SS18-SSX2 protein from HEK-FUS cells at 24 hours of tetracyclin induction. After Western blot analysis with an anti-BRG1 antibody, the coimmunoprecipitated BRG1 protein was readily detected (Fig. 5A), thereby confirming that both proteins interact directly. After stripping and reprobing the same blot with the anti-SS18 antibody, RA2009 (9), the immunoprecipitated SS18-SSX protein was also readily detected (Fig. 5B). In order to establish a direct binding of the BRG1 protein to the putative IGF2 and CD44 target promoters, we did ChIP on HEK-FUS cells at 0 and 24 hours of tetracyclin induction. The recovered ChIP DNAs were analyzed by Q-PCR for enrichment of IGF2 and CD44 promoter sequences, expressed as P/I values (see Materials and Methods). By doing so, we found that all P/I values were >1, indicating that the BRG1 ChIP DNAs were indeed enriched for IGF2 and CD44 promoter sequences (Fig. 5C). In addition, we found that the IGF2 P/I values decreased significantly upon induction of the SS18-SSX2 protein, whereas the CD44 P/I values remained constant after the induction of the SS18-SSX2 protein (Fig. 5C). These ChIP data indicate that the presence SS18-SSX2 protein leads to dissociation of the BRG1 protein from the IGF2 promoter, but not from the CD44 promoter. Combined with the expression data (IGF2, up-regulation; CD44, down-regulation), this implies that the BRG1 protein may, in conjunction with the SS18-SSX2 protein, exert negative regulatory effects on the IGF2 and CD44 promoters, respectively. These data support our hypothesis that SWI/SNF-related functions may be impaired after SS18-SSX2 expression induction. Through additional ChIP analyses, using antibodies directed against specifically modified histones, we found that histone modification patterns at the IGF2 promoter change in response to the SS18-SSX2 induction, whereas they remained constant at the CD44 promoter. Specifically, we observed that the levels of H4K16-Ac and H3K4-triMe, which are associated with active genes (46), were increased significantly in the IGF2 promoter (Fig. 5D) upon SS18-SSX2 induction. Together, these data indicate that the SS18-SSX2 protein, possibly in conjunction with BRG1, affects gene expression in an epigenetic fashion.
SS18-SSX2 targets include methylation-sensitive genes. Methylation-sensitive genes (http://www.missouri.edu/~hypermet/list_of_promoters.htm), including IGF2, SPARC, BTG2, CSPG2, IGFBP7, and CD44, were also found to be deregulated in response to SS18-SSX2 expression (Fig. 2). Because the sensitivity of genes for DNA methylation generally relies on the presence of promoter-associated CpG islands, we checked for the presence of this feature in all genes listed in Fig. 2. Using the UCSC genome browser (http://genome.ucsc.edu/), we found that 112 of 136 (82%) of these genes indeed contain CpG islands in their promoter regions, thereby turning them into putative targets for transcriptional (de)regulation through DNA methylation. A genome-wide survey of the UCSC data indicated that the promoter regions of 16,983 of 96,377 (18%) singly mapped UniGene clusters are associated with CpG islands. Based on the striking enrichment of CpG island–positive genes within the SS18-SSX2–responsive genes, we hypothesize that the SS18-SSX2–mediated transcriptional effects may, in part, be induced by (epigenetic) changes in DNA methylation. Either through, or in parallel with, defects in the SWI/SNF complex, this may lead to changes in DNA (cytosine) methylation. These methylation changes, in turn, may induce changes in expression of downstream target genes. The observations that the SS18-SSX2 and BRG1 proteins interact physically, that the SWI/SNF subunits are associated with transcriptional repression through DNA methylation (47), and that BRG1-induced DNA methylation changes affect CD44 expression (48), are in full support of this hypothesis.
SS18-SSX2 expression affects DNA methylation at the IGF2 locus. The above hypothesis prompted us to measure genome-wide DNA methylation levels in the SS18-SSX2–expressing cells. To this end, genomic DNAs extracted from the HEK-VEC and HEK-FUS cell lines (t = 0 and t = 24), were digested to mononucleotides, and the total methyl-deoxycytosine (mdC) content in these samples was measured by high-performance liquid chromatography. In the past, this method has been used to detect genome-wide mdC levels in tumor samples, normal tissues, and knockout models (49, 50). Through this approach, we detected an increase (5%) in the overall mdC level in SS18-SSX2–expressing cells as compared with the empty vector control cells (data not shown). Literature data indicate that a knockout of the DNA methyltransferase 1 (DNMT1) gene induced a 20% decrease in overall mdC levels (50), whereas DNMT1 overexpression induced a 15% increase in overall mdC levels (51). Although the SS18-SSX2–induced effect on overall DNA methylation was limited, our data provides support for the proposed DNA methylation hypothesis.
In addition, we selected the IGF2 gene for a locus-specific DNA methylation analysis. The IGF2 gene, which we found to be the most extremely up-regulated gene in response to SS18-SSX2 expression, and which was found to be consistently up-regulated in primary synovial sarcomas (26, 35), is equipped with both a CpG island and a differentially methylated region, also referred to as H19 imprinting control region (H19-ICR). In its unmethylated state, this ICR forms a chromatin boundary that prevents IGF2 activation by distant enhancers. Binding of the zinc finger protein CTCF to the ICR is essential for this chromatin boundary function (52). Methylation of the CTCF binding sites within the ICR abolishes CTCF binding, thereby leading to IGF2 activation. Because we observed genome-wide DNA hypermethylation, as well as IGF2 up-regulation (see above), we decided to analyze the DNA methylation status of the H19-ICR in detail. To this end, we did bisulfite sequencing on genomic DNAs extracted from HEK-VEC and the HEK-FUS cells after 0 and 24 hours of tetracyclin induction. As a target, we selected one of the eight consensus CTCF binding sites located upstream of the H19 gene (Fig. 6). After sequencing of 20 to 22 independently cloned PCR products from each time point of both cell lines, we observed a significant increase of methylated cytosine residues in the HEK-FUS cell line as compared with the HEK-VEC cell line (Fig. 6). The most striking differences were observed in the first three CpG dinucleotides (including the CTCF binding site, Fig. 6). This enhanced ICR methylation is fully compatible with the observed IGF2 up-regulation and supports our hypothesis that SS18-SSX2 expression affects DNA methylation, thereby leading to transcriptional deregulation of downstream target genes.
In conclusion, we did cDNA microarray-based expression profiling in conjunction with a SS18-SSX2 inducible cell system. By doing so, we identified various SS18-SSX2–responsive genes, including a group of genes involved in cholesterol synthesis, a hallmark of malignancy. Integration of our data with literature data on primary synovial sarcomas led to the identification of a group of genes, including IGF2 and CD44, that seems to be controlled directly by the SS18-SSX2 fusion protein. The opposite transcriptional regulation observed for the IGF2 and IGFBP5 genes indicates that SS18-SSX2 expression may affect IGF signaling. In addition, SS18-SSX expression may lead to an uncoupling of JUN and EGR1-mediated transcription regulatory processes. The deregulation of the SWI/SNF-responsive gene CD44, prompted us to hypothesize that this uncoupling of transcription regulatory processes may be caused by an SS18-SSX–induced abrogation of the SWI/SNF complex. This hypothesis is in line with a proposed function of the SS18-SSX fusion protein as an “activator-repressor,” and is fully supported by our observations that the SS18-SSX and BRG1 proteins interact directly and that BRG1 binds directly to the IGF2 and CD44 promoters. In addition, we observed specific histone acetylation and methylation modifications within the IGF2 promoter in response to SS18-SSX2 protein expression, which are fully compatible with the enhanced activity of this gene. Finally, we observed SS18-SSX2–induced changes in DNA methylation, both at the genome-wide level and at the single-locus level in the IGF2 imprinting control region. Taken together, our data indicate that the SS18-SSX2–induced transcriptional responses may represent epigenetic phenomena hampering normal SWI/SNF functions. The recent finding that SWI/SNF subunits may act as tumor suppressors (53), and that the histone deacetylase–inhibitory drug FK228 exerts a positive effect on the expression of the SWI/SNF protein BRM (54) and a concomitant negative effect on the growth of synovial sarcoma cells (55), fully supports our hypothesis.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: Dutch Cancer Society (Koningin Wilhelmina Fonds).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The authors thank Paul Joosten (Department of Cell Biology, Radboud University Nijmegen) for providing the anti-BRG1 antibody.