SYT-SSX 1 and SYT-SSX 2 Interfere with Repression of E-Cadherin by Snail and Slug : A Potential Mechanism for Aberrant Mesenchymal to Epithelial Transition in Human Synovial Sarcoma

Synovial sarcoma is a primitive mesenchymal neoplasm characterized in almost all cases by a t(X;18) fusing the SYT transcriptional coactivator gene with either SSX1 or SSX2 , with the resulting fusion gene encoding an aberrant transcriptional regulator. A subset of synovial sarcoma, predominantly cases with the SYT-SSX1 fusion, shows foci of morphologic epithelial differentiation in the form of nests of glandular epithelium. The striking spontaneous mesenchymal to epithelial differentiation in this cancer is reminiscent of a developmental switch, but the only clue to its mechanistic basis has been the observation that most cases of synovial sarcoma with glandular epithelial differentiation (GED) contain SYT-SSX1 instead of SYT-SSX2 . We report here that SYT-SSX1 and SYT-SSX2 interact preferentially with Snail or Slug, respectively, and prevent these transcriptional repressors from binding to the proximal E-cadherin promoter as shown by coimmunoprecipitation and chromatin immunoprecipitation. Luciferase reporter assays reveal that SYT-SSX1 and SYT-SSX2 can respectively overcome the Snailor Slugmediated repression of E-cadherin transcription. This provides a mechanism by which E-cadherin expression, a prerequisite of epithelial differentiation, is aberrantly derepressed in synovial sarcoma and may also explain the association of GED with the SYT-SSX1 fusion because it interferes with Snail, the stronger repressor of the E-cadherin promoter. Thus, our data provide a mechanistic basis for the observed heterogeneity in the acquisition of epithelial characteristics in synovial sarcoma and highlight the potential role of differential interactions with Snail or Slug in modulating this phenotypic transition. (Cancer Res 2006; 66(14): 6919-27)


Introduction
Synovial sarcomas are primitive mesenchymal tumors that account for f10% of all soft tissue sarcomas.Synovial sarcomas contain a t(X;18)(p11.2;q11.2),representing the fusion of SYT (also known as SS18) with either SSX1 or SSX2 or rarely with SSX4 (1, 2).There are two major morphologic forms of synovial sarcoma: monophasic, entirely composed of spindle cells with or without solid epithelial areas, and biphasic, containing epithelial cells lining open spaces [i.e., glandular epithelial differentiation (GED)] in a background of spindle cells (Supplementary Fig. S1).Both the spindle and epithelial elements of synovial sarcoma contain the t(X;18) and are thus clonally related (2,3).The GED in synovial sarcoma has the hallmarks of a genuine mesenchymal to epithelial transition (MET) akin to those seen in embryonic development (e.g., in developing kidney).Thus, the epithelial cells in synovial sarcoma express E-cadherin, keratins, a-catenin, h-catenin, and g-catenin, whereas the spindle cells express vimentin and, focally, N-cadherin (4,5).Epithelial differentiation in synovial sarcoma is also well documented at the ultrastructural level (6,7).
An intriguing observation in synovial sarcoma is that the type of fusion gene (SYT-SSX1 versus SYT-SSX2) is strongly correlated with tumor phenotype (monophasic versus biphasic histology as defined by the presence of GED with lumen formation; refs.8,9).An analysis of aggregate data from the three largest studies, including a total of 471 synovial sarcoma tumors (10)(11)(12), indicates that tumors with the SYT-SSX1 fusion are at least five times as likely to show GED compared with SYT-SSX2-bearing tumors (P < 0.0001).This points to a possible role for the SYT-SSX proteins in regulating GED, but how these alternative forms of the SYT-SSX chimeric transcriptional protein might exert such effects has been unknown.Understanding the mechanistic basis of the strong association between t(X;18) fusion type and biphasic histology should provide insights into the basic histogenetic processes of architectural epithelial differentiation and developmental switching between mesenchymal and epithelial phenotypes.
E-cadherin (CDH1) is a protein, whose extracellular domain acts as a molecular zipper mediating cell-cell adhesion whereas its cytoplasmic tail is linked to the actin cytoskeleton via catenins.Multiple lines of evidence have defined it as a critical determinant of the epithelial phenotype, and its loss seems central to the process of epithelial to mesenchymal transition (EMT; refs.[13][14][15].EMT, like its reverse counterpart, MET, is not only a key process in embryonic development but has also been implicated in tumor progression.Absence of E-cadherin expression is seen in synovial sarcoma lacking GED (i.e., monophasic synovial sarcoma), and this is associated with either overexpression of the Snail transcriptional repressor or inactivating mutations in the E-cadherin gene (16,17).Interestingly, inactivating mutations in E-cadherin are observed only in monophasic synovial sarcoma with SYT-SSX1 (16).This is intriguing because, as described above, it is known that synovial sarcoma with SYT-SSX1 is much more likely to display GED than SYT-SSX2-positive synovial sarcoma.Together, these data suggested a model, in which the capacity for GED may be intrinsically greater in synovial sarcoma cases with SYT-SSX1 than those with SYT-SSX2 but is abrogated in many SYT-SSX1 cases by mutational or transcriptional down-regulation of E-cadherin.In contrast to the findings in synovial sarcoma, E-cadherin is only rarely expressed in other spindle cell sarcomas (4,18).
Based on these data, we hypothesized that GED in this mesenchymal neoplasm is modulated by SYT-SSX and that SYT-SSX1 and SYT-SSX2 might differentially regulate the E-cadherin gene possibly by interactions with specific DNA-binding transcription factors.Because E-cadherin expression seems to be regulated primarily by transcriptional repressors (Snail, Slug, SIP1, and Twist; refs.13,14,19,20), we examined whether its transcriptional repression by Snail and Slug in synovial sarcoma is altered by SYT-SSX.We report here that SYT-SSX1 and SYT-SSX2 interact preferentially with either Snail or Slug, respectively, and prevent these repressors from binding to the proximal E-cadherin promoter, thereby interfering with transcriptional repression of the E-cadherin gene to different degrees and potentially triggering an epithelial differentiation program leading to a MET with GED in synovial sarcoma.
Luciferase reporter gene constructs containing wild-type human E-cadherin promoter sequences ( from À76 to À490) were generated by amplifying the promoter region and ligating the fragment into the KpnI and BglII sites of the pGL3-enhancer vector (Promega, Madison, WI).Mutated reporter constructs, in which three E-box sequences in the promoter were changed from CANNTG to AANNTA (13), were generated using the Quick Mutagenesis kit (Stratagene, La Jolla, CA).
HEK293 cells or HeLa cells were seeded in six-well plates at density of 2 Â 10 5 /mL and allowed to grow for 24 hours before transfection and done using Fugene 6 (Roche, Alameda, CA).For experiments assessing activation of E-cadherin reporter gene constructs by endogenous transcription factors, 1.0 Ag of empty vector or appropriate expression plasmids, 0.35 Ag of reporter gene, and 20 ng of pRL-SV40 were transfected per well.To examine the effect of Snail and Slug in transactivation assays, 1.0 Ag of empty vector, 3xFLAG/Snail, or 3xFLAG/Slug was cotransfected with the above set of plasmids.
Western blot.Western blotting was done after preparation of cell lysates with radioimmunoprecipitation assay buffer.Nitrocellulose membranes were preincubated with 5% nonfat dry milk in TBS-Tween 20 before incubation with specific primary antibodies for 2 hours.Specific molecules were visualized with horseradish peroxidase-labeled anti-mouse or antigoat secondary antibodies and enhanced chemiluminescence (Amersham Biosciences).
Production of polyclonal antibody against SSX COOH-terminal region.The sequence of the synthetic peptide used as immunogen was the following: C-LVIYEEISDPEEDD-NH 2 , representing a conserved sequence in the SSXRD shared by SSX1 and SSX2 and included in both SYT-SSX1 and SYT-SSX2.This antibody also recognizes native SSX1 and SSX2 proteins (Supplementary Figs.S2 and S3).Rabbit polyclonal antibody production was done by AnaSpec (San Jose, CA).
Establishment of SYT-SSX1 and SYT-SSX2-inducible cell lines.T-Rex 293 cells and T-Rex HeLa cells (Invitrogen) were cultured in 10-cm dishes with DMEM supplemented with 10% fetal bovine serum (FBS; Tet systemapproved FBS; BD Biosciences, Franklin Lakes, NJ).At 70% confluence, cells were transfected with 2 Ag DNA of SYT-SSX1/TO or SYT-SSX2/TO, respectively, using Fugene 6 transfection reagent.At 48 hours after transfection, drug selection was started in fresh medium supplemented with 400 Ag/mL Zeocin (Invitrogen) for 6 weeks and drug-resistant colonies were isolated.Cells from isolated colonies were treated with 1 Ag/mL tetracycline (Invitrogen) to induce SYT-SSX expression.
Chromatin immunoprecipitation assays.Chromatin immunoprecipitation (ChIP) assays were done with a ChIP assay kit (Upstate Biotechnology, Inc., Charlottesville, VA), with minor modifications.Briefly, 293T cells were transiently transfected with either SYT-SSX1/HisMyc or SYT-SSX2/HisMyc expression plasmid for 48 hours.To improve the efficiency of the ChIP assay, we treated the cells first with 10 mmol/L dimethyl adipimidate, a protein cross-linking agent, and 0.25% DMSO in PBS for 45 minutes (24,25).Formaldehyde was then added at a final concentration of 1% for 10 minutes at 37jC to cross-link myc-tagged SYT-SSX or endogenous SYT-SSX to its DNA target sites in vivo.The cells were harvested, suspended with SDS lysis buffer [1% SDS, 10 mmol/L EDTA, 50 mmol/L Tris-HCl (pH 8.3)], and incubated on ice for 10 minutes.Lysates were sonicated, and debris was removed by centrifugation for 10 minutes at 14,000 Â g.An aliquot of each chromatin solution (200 AL) was set aside and designated as the input fraction.Supernatants were diluted 3-fold in immunoprecipitation buffer and precleared with Sepharose A/G plus agarose beads that had been preabsorbed with salmon sperm DNA.The precleared chromatin solution was incubated with either anti-myc mouse monoclonal antibody ( for 293T cells; Invitrogen) or original rabbit SSX or Snail (H-130, Santa Cruz Biotechnology) or Slug (H-140, Santa Cruz Biotechnology) antibodies ( for SYO-1 and HS-SY-II cells) or normal rabbit or mouse IgG for 16 hours at 4jC.The immune complexes were then collected with the addition of Sepharose A/G plus agarose beads followed by several washes with appropriate buffers according to the manufacturer's protocol.Each sample was eluted with freshly prepared 1% SDS and 0.1 mol/L NaHCO 3 , and then cross-links were reversed with the addition of 5 mol/L NaCl.Chromatin-associated proteins were digested with proteinase K (10 mg/mL), and the immunoprecipitated DNA was recovered by phenol/ chloroform extraction and ethanol precipitation and analyzed by PCR.The primer pairs for the detection of the E-cadherin promoter were as follows: E-cadherin-ChIP, 5 ¶-CGAACCCAGTGGAATCAGAA-3 ¶ ( forward 1), 5 ¶-GCGG-GCTGGAGTCTGAACTG-3 ¶ (reverse 1; amplicon from À359 to À63), 5 ¶-TG-GTGGTGTGCACCTGTACT-3 ¶ ( forward 2), and 5 ¶-GACCTGCACGGTTCT-GATTC-3 ¶ (reverse 2; amplicon from À600 to À329).The primer pair for the amplification of the 3 ¶-untranslated region (UTR) of the E-cadherin gene was as follows: E-cadherin-ChIP-3 ¶-UTR, 5 ¶-CAAGTGCCTGCTTTTGATGA-3 ¶ ( forward) and 5 ¶-GCTTGAACTGCCGAAAAATC-3 ¶ (reverse).PCR products were resolved and visualized with ethidium bromide.

Cancer Research
Cancer Res 2006; 66: (14).July 15, 2006 To investigate changes in histone modifications at the E-cadherin promoter after induction of the SYT-SSX, ChIP assays were done in HeLa T-Rex SYT-SSX-inducible cell lines with or without tetracycline treatment using anti-acetyl histone H3 and anti-acetyl histone H4 antibodies (Upstate Biotechnology) for immunoprecipitation and the same primer pair for PCR detection.The bound fraction DNA was extracted from the fraction bound to protein A-agarose/salmon sperm DNA, and the unbound fraction DNA was extracted from cell lysate (supernatant) in the same tube.
To see whether Snail or Slug still remains bound to the proximal E-cadherin promoter in the presence of SYT-SSX1 or SYT-SSX2, respectively, mock, SYT-SSX1, or SYT-SSX2 expression vector was cotransfected with Snail or Slug expression vector into 293T cells, and ChIP assay was done by immunoprecipitation with anti-FLAG antibody ( for Snail or Slug).
RNA interference.To investigate the effect of endogenous SYT-SSX on E-cadherin expression, we used RNA interference by a small interfering RNA (siRNA) duplex against SYT-SSX.In brief, 24 hours before transfection, 80% confluent cells were trypsinized and diluted with fresh medium without antibiotics at 3 Â 10 5 /mL and transferred to six-well plates (2 ml/well).Transfection of siRNAs was carried out using Oligofectamine reagent (Invitrogen) and 0.84 Ag of siRNA duplex.The sense and antisense siRNA sequences were 5 ¶-CCAGAUCAUGCCCAAGAAGdTdT-3 ¶ and 5 ¶-CUU-CUUGGGCAUGAUCUGGdTdT, respectively (Dharmacon, Chicago, IL).

Results
SYT-SSX1 and SYT-SSX2 interact with either Snail or Slug as a gain of function.SYT-SSX expression vectors were cotransfected with Snail expression plasmid into 293T cells.Coimmunoprecipi-tation experiments revealed that SYT-SSX1 interacts strongly with Snail whereas SYT-SSX2 does not (Fig. 1A).Because transcriptional repression by Snail may be modulated in a competitive fashion by Slug through lower affinity binding of the same target sequences (26), we also examined the interaction of SYT-SSX1 and SYT-SSX2 with Slug.In contrast to Snail, Slug appeared to be bound more efficiently by SYT-SSX2 than SYT-SSX1 (Fig. 1A).Because SYT-SSX1 and SYT-SSX2 differ only within their SSX-derived portions, this raised the question if native SSX can also interact with Snail/Slug.However, coimmunoprecipitation experiments revealed that native SSX proteins do not interact with either Snail or Slug (data not shown).Furthermore, full-length SYT also did not interact with either Snail or Slug (data not shown).These results suggested that this interaction between SYT-SSX and Snail/Slug is a gain of function of the fusion proteins relative to their native counterparts.
SYT-SSX proteins interfere with transcriptional repression of E-cadherin by Snail/Slug.We first confirmed the reported transcriptional repression of E-cadherin by Snail and Slug in HeLa cells (Supplementary Fig. S2A and B; refs.13,14,26).The levels of repression of the E-cadherin promoter in this system differ somewhat from Snail/Slug-dependent repression of this promoter in Madin-Darby canine kidney (MDCK) cells (Supplementary Fig. S2C; ref. 26) but are consistent with findings in breast carcinoma cells (27).We then examined whether SYT-SSX can transactivate the E-cadherin promoter.Transfection of SYT-SSX expression vectors into HeLa cells produced f20-fold (SYT-SSX1) to 30-fold (SYT-SSX2) transactivation of the E-cadherin reporter (Fig. 1B).However, the same experiment done in 293 cells resulted in only 2-fold (SYT-SSX1) to 3-fold (SYT-SSX2) transactivation (Fig. 1C).These contrasting results in different cell lines may reflect differences in the available interacting proteins in these two cell lines.Specifically, endogenous Snail and Slug are inactivated by phosphorylation in E-cadherin-positive cell lines, such as 293 cells, but are strongly expressed and active in HeLa cells (28).Thus, transactivation of the E-cadherin reporter by SYT-SSX proteins may be, at least partly, dependent on the presence of Snail/Slug, raising the possibility that SYT-SSX proteins exert their effect on this reporter in part by overcoming transcriptional repression by endogenous Snail/Slug.To test this, Snail/Slug expression vectors were cotransfected with either SYT-SSX1 or SYT-SSX2 and E-cadherin reporter vector in HeLa cells.Cotransfection of Snail or Slug expression vectors into HeLa cells along with SYT-SSX vectors showed that the latter can overcome the transcriptional repression of an E-cadherin reporter construct by Snail or Slug (Fig. 1B).Interestingly, the derepressive effect was stronger for SYT-SSX1 on Snail cotransfection and for SYT-SSX2 on Slug cotransfection, in line with the preferential interactions of these respective proteins shown by the above coimmunoprecipitation studies.Moreover, when the assays were repeated with an E-cadherin reporter vector in which the three E-box sequences known to be bound by Snail and Slug (13) were mutated, the derepressive effect of SYT-SSX was abolished (Fig. 1D).SYT-SSX also exerted a modest transactivating effect on the mutated E-cadherin promoter probably through a mechanism other than Snail/Slug interaction (Fig. 1D).In contrast, this derepressive effect could not be observed for full-length native SYT, SSX1, or SSX2 (Supplementary Fig. S3A and B).Taken together, these data suggest that SYT-SSX1 and SYT-SSX2 interact preferentially with Snail or Slug, respectively, and interfere with the normal functions of these transcriptional repressors at least in these model settings.
SSX repression domain is necessary for interaction of SYT-SSX with Snail/Slug.SYT-SSX1 and SYT-SSX2 contain both transcriptional activation and repressor domains (Fig. 2A).Our observation of contrasting interactions of SYT-SSX1 and SYT-SSX2 with Snail or Slug prompted us to ask if the SSX repression domain, the most divergent portion of SYT-SSX1 and SYT-SSX2, might be responsible for this derepressive effect.Coimmunoprecipitation following the cotransfection of either SYT-SSX1RD(À) or SYT-SSX2RD(À) with Snail/Slug revealed that these COOH-terminal deletion mutants (see diagram in Fig. 2A) could not interact with Snail/Slug (data not shown).Luciferase reporter assays showed that these COOH-terminal deletion mutants could transactivate E-cadherin promoter only very weakly (<3-fold) in HeLa cells (with high endogenous Snail/Slug; Fig. 2B), reminiscent of the results in 293 cells (with low endogenous Snail/Slug) transfected with fulllength SYT-SSX (Fig. 1C).Furthermore, the aforementioned derepressive effect of full-length SYT-SSX in the setting of Snail/ Slug cotransfection was disrupted in these COOH-terminal deletion mutants (Fig. 2B).These results suggest that the SSX repression domain of SYT-SSX is necessary for the interaction between SYT-SSX and Snail/Slug and the resulting derepressive effect on Snail/ Slug target promoters.
SYT-SSX is also present at the E-cadherin promoter independent of Snail or Slug.Next, to understand how SYT-SSX proteins exert their transactivating effects on the mutated E-cadherin promoter lacking canonical E-boxes in HeLa cells and on the intact E-cadherin promoter in 293 cells (with low endogenous Snail/Slug), we asked if SYT-SSX is present at the endogenous E-cadherin promoter in the absence of Snail/Slug.HEK293T cells were transiently transfected with myc-tagged SYT-SSX expression plasmids.ChIP was done with anti-myc antibody.This revealed that both forms of SYT-SSX are present in vivo at the E-cadherin promoter in the absence of Snail/Slug, suggesting that this minor component of SYT-SSX-associated transactivation is through a mechanism other than Snail/Slug interaction (Fig. 3B).Next, we confirmed that endogenous Snail or Slug is present in vivo at the E-cadherin promoter in synovial sarcoma cell lines (Fig. 3C, top) and that endogenous SYT-SSX is also present in vivo at the E-cadherin promoter in synovial sarcoma cell lines using a custom polyclonal SSX antibody (Fig. 3C, bottom).The suitability of this antibody for immunoprecipitation of SYT-SSX was confirmed by transient transfection of SYT-SSX expression plasmid into HEK293T cells (Supplementary Figs.S4 and S5).Although these experiments confirmed the presence of Snail/Slug and SYT-SSX at the E-cadherin promoter in synovial sarcoma cell lines, it remained unclear whether Snail or Slug remains localized to the proximal E-cadherin promoter when bound by SYT-SSX proteins.To address this, we did several ChIP experiments.Snail or Slug expression vector was transfected into 293T cells with either mock or SYT-SSX1 or SYT-SSX2 expression vector.These experiments revealed that Snail binding to the E-cadherin promoter appears lower in the presence of SYT-SSX1 than in the presence of SYT-SSX2 (Fig. 3D, top) and that, conversely, Slug binding to the E-cadherin promoter is lower in the presence of SYT-SSX2 than in the presence of SYT-SSX1 (Fig. 3D, bottom).Taken together with the data from coimmunoprecipitation and luciferase reporter assays, these results support a model in which SYT-SSX1 and SYT-SSX2 interact with either Snail or Slug and prevent Snail or Slug from binding to the proximal E-cadherin promoter, resulting in a derepressive effect on the E-cadherin promoter.The relatively weak signals in the SYT-SSX lanes in Fig. 3C may also be consistent with this interpretation, if indeed most of the functional interaction of SYT-SSX with this promoter occurs through the physical interaction with Snail or Slug that reduces their binding to the E-boxes and thereby also removes SYT-SSX proteins from the vicinity of this promoter.
Induction of endogenous E-cadherin expression by SYT-SSX.To gauge the effect of exogenous SYT-SSX on endogenous E-cadherin expression, total RNA was extracted from 293 T-Rex cells and HeLa T-Rex cells stably transfected with an inducible SYT-SSX expression plasmid.E-cadherin transcript levels were measured with or without induction of SYT-SSX by tetracycline (Fig. 4A) at five time points (24, 48, 72, 96, and 120 hours after induction).Quantitative RT-PCR revealed only a minimal increase (f2-fold) in E-cadherin expression on induction of SYT-SSX in 293 T-Rex cells (Fig. 4B).In contrast, in HeLa T-Rex cells, E-cadherin transcript levels were significantly up-regulated in a time-dependent fashion after SYT-SSX induction for both SYT-SSX1 (Fig. 4C) and SYT-SSX2 (Fig. 4D).These results parallel those of the luciferase reporter assays described above and suggest that the effects of SYT-SSX proteins on the E-cadherin promoter are enhanced in cells that contain high levels of Snail/Slug, such as HeLa.
Knockdown of SYT-SSX decreases E-cadherin expression in synovial sarcoma cell lines.Knockdown of SYT-SSX was achieved by transient transfection of a siRNA targeting the SYT-SSX fusion transcript junction (the sequence at the junction is identical in SYT-SSX1 and SYT-SSX2).This siRNA caused up to 75% knockdown of the endogenous SYT-SSX transcript at 48 hours after transfection in both synovial sarcoma cell lines compared with a control nonspecific siRNA.Quantitative RT-PCR showed that E-cadherin mRNA levels in both synovial sarcoma cell lines were significantly decreased at 8 hours after transfection of the SYT-SSX siRNA duplex (Fig. 5).The E-cadherin quantitative RT-PCR was done at an early time point to minimize secondary effects mediated by other possible interactions of SYT-SSX proteins.
Transactivation of the E-cadherin gene by SYT-SSX1 is associated with histone hyperacetylation caused by dissociation of Snail from the E-cadherin promoter.Next, we investigated how SYT-SSX proteins exert their derepressive effect on E-cadherin transcription.Because Snail requires HDAC activity to repress the E-cadherin promoter, as manifested by deacetylation of histones H3 and H4 (25), we studied the effect of SYT-SSX1mediated suppression of Snail function on histone modifications at the E-cadherin promoter.We did ChIP assays with antibodies against acetyl histones H3 and H4 in SYT-SSX1-inducible HeLa T-Rex cell lines.The precipitated DNA was subjected to PCR with primers for the E-cadherin proximal promoter region (E-cadherin-ChIP forward primer 1 pair).This showed that SYT-SSX1 induction is associated with a significant increase in the levels of acetylated histones H3 and H4 at the E-cadherin promoter (Supplementary Fig. S6A).SYT-SSX2 induction in HeLa T-Rex cells did not change the acetylation status of histones H3 and H4 at the E-cadherin promoter (data not shown), consistent with the notion that its preferential interaction with the weaker transcriptional repressor Slug would result in little or no alteration of Snail-mediated repression.Although we have shown that some SYT-SSX is present at the E-cadherin promoter independent of the interaction with either Snail or Slug, it remained unclear whether this increased acetylation of histones H3 and H4 is caused by the direct interaction of SYT-SSX1 protein and HDACs to cause acetylation of histones as opposed to histone acetylation resulting from activation of transcription by the SYT-SSX1-induced dissociation of Snail from E-cadherin promoter.To address this issue, we did a coimmunoprecipitation experiment to examine the interaction between SYT-SSX and HDAC1, HDAC2, and HDAC3.This revealed no such interaction (Supplementary Fig. S6B).Taken together with the ChIP data, these results indicate that the increased histone acetylation observed after SYT-SSX1 induction results from the transcriptional activation (derepression) of the E-cadherin gene caused by the dissociation of Snail from the E-cadherin promoter and, therefore, that histone acetylation is an indirect effect of SYT-SSX.

Discussion
The strong correlation of SYT-SSX fusion type and GED in synovial sarcoma has long suggested a role for this fusion oncoprotein in the control of epithelial differentiation in this sarcoma, but the underlying mechanisms have remained obscure.This is due in part to the fact that the regulatory targets of the SYT-SSX proteins have been difficult to define because these fusion proteins lack a DNA-binding domain and few interacting transcription factors have been identified (29).In parallel with these observations, E-cadherin down-regulation has been reported to be associated with the loss or absence of GED in synovial sarcoma and this loss of E-cadherin expression is associated either with inactivating mutations of E-cadherin or with higher expression levels of its transcriptional repressor Snail (16).Our present data link these two sets of observations and suggest that the interaction of SYT-SSX with members of the Snail family of transcription factors may determine the propensity for GED in synovial sarcoma.
Snail family transcription factors have been implicated in embryonic development, mesoderm differentiation, neural crest formation, neural development, apoptosis, and neoplastic EMT (30).In epithelial cells, the induction of EMT by Snail or Slug is mediated by the direct transcriptional repression of E-cadherin, reversing its intercellular adhesion functions (13,14,26).In embryonic development, the transcriptional repression of E-cadherin is reversible (31)(32)(33).Indeed, EMT as a whole can be a reversible process in different tissues during embryonic development (31,32).Therefore, we hypothesized that, in some synovial sarcoma tumor cells, the transcriptional repression of E-cadherin is abrogated due to interactions of SYT-SSX1 and SYT-SSX2 with Snail and other E-cadherin transcriptional repressors, leading to their functional inactivation and resulting in the acquisition of epithelial characteristics.
Our data provide multiple lines of evidence supporting an interaction of SYT-SSX with these transcriptional repressors of E-cadherin.Moreover, these interactions have a reciprocal aspect: SYT-SSX1 interacts with Snail and SYT-SSX2 with Slug.Coimmunoprecipitation experiments to detect potential interactions between either full-length SYT or full-length SSX and Snail/Slug were negative.Thus, these interactions seem specific to  SYT-SSX fusion gene products, indicating that this is a gain of function of SYT-SSX relative to the corresponding native proteins.Further experiments to evaluate the transactivation of the E-cadherin promoter by SYT-SSX proteins in the presence of the high levels of Snail/Slug in HeLa cells showed that these alternative interactions could overcome repression of this promoter by Snail and Slug.The finding that E-cadherin transcription was increased in the presence of Snail/Slug in SYT-SSX-inducible cell lines supports a positive role of SYT-SSX fusion proteins in the upregulation of the E-cadherin gene in E-cadherin-negative cell lines.Furthermore, this differential interaction of SYT-SSX1 and SYT-SSX2 with either Snail or Slug may provide a mechanistic basis for the finding that SYT-SSX2 is at least 5-fold less likely than SYT-SSX1 to be associated with GED in synovial sarcoma tumors (see Introduction) because it has been suggested that Slug functions as a weaker repressor than Snail at the mouse E-cadherin promoter (26), an observation that we have further confirmed using the human E-cadherin promoter reporter vector in MDCK cells (Supplementary Fig. S2C).In addition, it has been also reported that, in MDCK cells, Snail is a stronger repressor than Slug at the promoter of human Claudin-1, which has two E-box sequences and encodes a tight junction protein (34).Thus, Snail-mediated repression of the E-cadherin promoter may be modulated in a competitive fashion by Slug through lower affinity binding of the same E-box sequences.Taken together, these findings suggest a model in which the alternative interactions of SYT-SSX1 and SYT-SSX2 with either Snail or Slug, respectively, alter the availability of one or the other of these repressors to bind the three E-boxes in the human E-cadherin promoter.In synovial sarcoma cells with SYT-SSX1, the fusion protein interacts with Snail and prevents it from binding to the E-cadherin promoter, preventing it from repressing E-cadherin transcription but not impairing the repressive function of Slug.Because Slug was shown to have a lower affinity to E-box sequences than Snail (26), this may explain the greater prevalence of E-cadherin expression and GED in synovial sarcoma with the SYT-SSX1 fusion.Conversely, in synovial sarcoma cells with SYT-SSX2, the fusion protein interacts with Slug and prevents it from binding to the E-cadherin promoter but does not alter the function of Snail.This results in stronger transcriptional repression of E-cadherin and hence a lower likelihood of GED in this subset of synovial sarcoma (Fig. 6A).This model is also consistent with the recent data showing an inverse correlation between E-cadherin and Snail expression levels in synovial sarcoma (16).Figure 6B integrates the present functional data with previous data on E-cadherin mutations and Snail expression levels in synovial sarcoma into a model of the determinants of epithelial differentiation in this mesenchymal cancer.
Although SYT-SSX1 caused a stronger induction of E-cadherin at early time points after SYT-SSX induction, we noted that, at later time points, SYT-SSX2 caused a stronger up-regulation of E-cadherin than SYT-SSX1 did.This may seem unexpected given that Slug (bound by SYT-SSX2) is a weaker repressor of E-cadherin than Snail (bound by SYT-SSX1) at least in MDCK cells (26).However, this is otherwise consistent with our finding that, in HeLa cells, Slug represses the E-cadherin promoter equally or more strongly than Snail (Supplementary Fig. S4A and  B) as also reported in a subset of breast cancer cell lines (27).Alternatively, this result in HeLa cells may in part reflect functional differences between SYT-SSX1 and SYT-SSX2 unrelated to their interactions with Snail or Slug because we have observed a stronger intrinsic transactivation by SYT-SSX2 than SYT-SSX1 when either is fused to the GAL4 DNA-binding domain as assayed in a GAL4 reporter system. 1 Figure 6.A, proposed model for the role of alternative interactions of SYT-SSX1 and SYT-SSX2 with either Snail or Slug in derepression of the E-cadherin transcription in synovial sarcoma.a, in the absence of SYT-SSX, E-cadherin expression is transcriptionally silenced in cells of mesenchymal origin by the binding of Snail or Slug to three E-boxes in the E-cadherin promoter; b, in synovial sarcoma cells with SYT-SSX1, SYT-SSX1 protein interacts with Snail and keeps it from binding to the E-cadherin promoter, preventing it from repressing E-cadherin transcription without impairing Slug, the weaker repressor at this promoter; c, conversely, in synovial sarcoma cells with SYT-SSX2, SYT-SSX2 protein binds Slug, preventing it from repressing E-cadherin transcription without impairing Snail, the stronger repressor at this promoter, and hence, this subset of synovial sarcoma shows a lower likelihood of E-cadherin expression and subsequent GED.B, proposed model of determinants of epithelial differentiation pathways in synovial sarcoma.The differences in the degree of transcriptional derepression of E-cadherin by SYT-SSX1 and SYT-SSX2 result in different propensities for epithelial differentiation.A subset of SYT-SSX1-bearing synovial sarcoma cases escapes epithelial differentiation by down-regulating E-cadherin either by transcriptional repression by high levels of Snail or through E-cadherin-inactivating mutations (16).Arrows, thickness is roughly proportional to the percentage of cases in each pathway.Overall, f25% of all synovial sarcomas are biphasic (almost all SYT-SSX1 positive) and 75% are monophasic (half SYT-SSX1 positive and half SYT-SSX2 positive).
Because the SSX repression domain SSXRD (35), which is included in the SYT-SSX fusion proteins, is the portion that is most divergent between SYT-SSX1 and SYT-SSX2, we examined the role of this domain in these interactions, which differ so sharply between the two forms of the fusion protein.Indeed, SYT-SSX constructs in which the SSXRD was deleted could not interact with Snail/Slug and could not derepress the effect of excess amount of Snail/Slug on the E-cadherin promoter in HeLa cells, indicating that the SSXRD is necessary for these interactions.
It has been shown that Snail mediates E-cadherin repression by the recruitment of the mSin3A/HDAC1/HDAC2 complex (25).mSin3A, a component of the HDAC complex, can interact with native SYT and SYT-SSX (36).SYT-SSX may also interact with the chromatin remodeling factor hBRM/hSNF2a (37).Histone deacetylation may also play role in transcriptional repression by Slug (38).Along these lines, we found that the acetylation of histones H3 and H4 at the E-cadherin promoter increased on induction of SYT-SSX1 in HeLa cells.Furthermore, we showed that this increased histone acetylation observed after SYT-SSX1 induction was dependent on the dissociation of Snail from the E-cadherin promoter and, therefore, that histone acetylation is an indirect effect of SYT-SSX.The minimal alterations in acetylated histones at the E-cadherin promoter following induction of SYT-SSX2 in HeLa cells suggested that its preferential interaction with Slug results in little or no alteration of histone modifications, possibly indicating that these are mainly triggered by Snail at this promoter.
However, it should be also noted that SYT-SSX proteins had some activity on the E-cadherin reporter plasmid that seemed independent of Snail/Slug based on luciferase reporter assays in HeLa cells using an E-cadherin reporter plasmid, in which all three E-box sequences, and based on ChIP assays that showed specific binding of SYT-SSX to the E-cadherin promoter without cotransfection of Snail or Slug even in cells expressing little or no endogenous Snail or Slug, such as HEK293T cells.Together, these results suggest that SYT-SSX proteins can also transactivate the E-cadherin gene through a mechanism independent of interactions with Snail or Slug, but the in vivo relevance of this phenomenon is unclear because the endogenous E-cadherin transcript is only weakly increased on induction of SYT-SSX in 293 T-Rex cell lines.
In summary, differential interactions of SYT-SSX1 and SYT-SSX2 with either Snail or Slug may provide a mechanistic basis for the observed heterogeneity in the acquisition of epithelial characteristics in this unique mesenchymal cancer (Fig. 6B).However, it is likely that this model is an oversimplification for at least two reasons.Firstly, the relative concentrations of Snail or Slug and SYT-SSX, as well as that of other E-cadherin repressors (E12/E47, ZEB1, SIP1, and Twist; refs.19,[39][40][41] and potential coregulators, may also be important in regulating the role of E-cadherin in epithelial differentiation in synovial sarcoma cells.Secondly, Snail may repress many other genes involved in epithelial differentiation, including among others the ELF3 transcription factor gene (42), previously shown by us to be overexpressed in biphasic synovial sarcoma (16,43).SYT-SSX fusion proteins may interfere with the repression of these other significant Snail targets as well.Nonetheless, our findings strengthen the concept that the protein interactions with Snail or Slug are important in the modulation of MET and EMT in mesenchymal differentiation and neoplasia.

Figure 1 .
Figure 1.A, alternative interactions of SYT-SSX1 and SYT-SSX2 with either Snail or Slug.293T Cells were cotransfected with pcDNA4-HMB-SYT-SSX (À1 or À2) and p3xFLAG-CMV-10-Snail or Slug.Total cell lysate was extracted 48 hours after transfection.After immunoprecipitation (IP ) and Western blotting, membranes were probed with anti-FLAG antibody (M2) to detect Snail or Slug.B, transactivation of the E-cadherin promoter by SYT-SSX with or without cotransfection of Snail or Slug expression vector in HeLa cells.Both forms of SYT-SSX can overcome transcriptional repression by Snail or Slug, but their efficiencies differ in parallel with the coimmunoprecipitation findings.Note that each of the relative luciferase values of the three experiments (mock, Snail, or Slug cotransfection) was normalized separately to that of the corresponding mock cotransfection (light blue ).Thus, the data cannot be directly compared between the three experiments.C, transactivation of the E-cadherin promoter by SYT-SSX without cotransfection of Snail or Slug expression vector in 293 cells.D, transactivation of the E-cadherin promoter with three mutated E-boxes by SYT-SSX with or without cotransfection of Snail or Slug expression vector in HeLa cells.The derepressive effect of SYT-SSX on Snail/Slug-mediated repression of the E-cadherin promoter reporter is abolished.However, SYT-SSX1 and SYT-SSX2 nonetheless exert a modest activating effect on this mutant reporter.

Figure 2 .
Figure 2. A, schematic diagram of domain structure of the SYT, SSX, and SYT-SSX proteins.Amino acid residues representing the boundaries of selected domains.The scale is approximate.SNH, SYT NH 2 -terminal domain; QPGY, SYT glutamine-, proline-, glycine-, and tyrosinerich domain; KRAB, Kru ¨ppel-associated box; DD, SSX divergent domain; SSXRD, SSX repressor domain.Bottom, COOH-terminal deletion mutant.B, the SSX repression domain of the SYT-SSX is necessary for the derepressive effect.COOH-terminal deletion mutants of SYT-SSX can transactivate the E-cadherin promoter only very weakly (<3-fold) in HeLa cells.Moreover, the aforementioned derepressive effect on Snail/Slug was disrupted in these COOH-terminal deletion mutants.

Figure 3 .
Figure 3. ChIP assays at the E-cadherin promoter.A, map of the E-cadherin promoter showing the location of the three E-boxes and the sites of amplification for ChIP analyses.B, ChIP in 293T cells transfected with SYT-SSX1 shows that exogenously expressed SYT-SSX is present at the E-cadherin promoter.C, top, endogenous Snail or Slug is present at the proximal E-cadherin promoter in synovial sarcoma cell lines; bottom, ChIP in SYO-1 and HS-SY-II cell lines shows that endogenous SYT-SSX protein is also present at the E-cadherin promoter in synovial sarcoma cells.Primer pairs 1 and 2 amplify the E-cadherin promoter region from À359 to À63 and from À600 to À329, respectively.D, top, snail binds to the E-cadherin promoter to a lesser extent in the presence of SYT-SSX1; bottom, Slug binds to the E-cadherin promoter to a lesser extent in the presence of SYT-SSX2.

Figure 4 .
Figure 4. Effect of SYT-SSX induction on endogenous E-cadherin expression in heterologous cell lines.A, robust induction of SYT-SSX in HeLa T-Rex cells.B, E-cadherin transcript levels increased slightly after the induction of SYT-SSX in 293 T-Rex cells.C and D, E-cadherin transcript increased markedly after SYT-SSX1 (C ) and SYT-SSX2 (D ) induction in HeLa T-Rex cells.Data at each time point were normalized to the data from corresponding noninduced cells.These results are consistent with those of luciferase reporter assays and suggest that E-cadherin promoter activity is strongly enhanced by SYT-SSX only in cell lines that express Snail/Slug (such as HeLa).

Figure 5 .
Figure 5. Knockdown of endogenous SYT-SSX in synovial sarcoma cell lines results in decreased endogenous E-cadherin expression.Real-time quantitative RT-PCR revealed that E-cadherin transcript levels decreased significantly in the SYT-SSX siRNA-transfected synovial sarcoma cell lines HS-SY-II (A) and SYO-1 (B).