sign in sign up
<<
Home , Amphiregulin, EPH receptor B1, FGF11, GDF6, GFRA2 (gene), HOXB5

Oncogene (2012) 31, 2323–2334 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE Reprogramming of mesenchymal stem cells by the -associated oncogene SYT–SSX2

CB Garcia1, CM Shaffer2, MP Alfaro3, AL Smith1, J Sun4, Z Zhao1,4, PP Young3,5, MN VanSaun1 and JE Eid1

1Department of Cancer Biology, Vanderbilt University School of Medicine, Nashville, TN, USA; 2Center for Human Genetics Research, Vanderbilt University, Nashville, TN, USA; 3Department of Pathology, Vanderbilt University School of Medicine, Nashville, TN, USA; 4Department of Biomedical Informatics, Vanderbilt University School of Medicine, Nashville, TN, USA and 5Department of Veterans Affairs Medical Center, Nashville, TN, USA

Cell identity is determined by its expression Introduction programs. The ability of a cell to change its identity and produce cell types outside its lineage is achieved by the Synovial sarcoma (SS) is an aggressive malignancy that activity of transcription controllers capable of reprogram- afflicts adolescents and young adults. It is characterized ming differentiation gene networks. The synovial sarcoma by a recurrent chromosomal translocation t(X;18) where (SS)-associated , SYT–SSX2, reprograms myo- the SYT gene (also known as SS18) on -18 genic progenitors and human bone marrow-derived me- is fused with the SSX1, SSX2 or (in a few cases) SSX4 senchymal stem cells (BMMSCs) by dictating their on the , forming the SYT–SSX commitment to a pro-neural lineage. It fulfills this function oncogene. The translocation protein is thought to by directly targeting an extensive array of neural-specific regulate , however its mechanism of genes as well as genes of developmental pathway action and the resulting phenotypic consequences are mediators. Concomitantly, the ability of both myoblasts still largely unclear (Ladanyi, 2001). and BMMSCs to differentiate into their normal myogenic Binding of the SYT portion to the SWI/SNF and adipogenic lineages was compromised. SS is believed chromatin-remodeling complex (Nagai et al., 2001) to arise in mesenchymal stem cells where formation of the and direct association of the SSX segment with t(X/18) translocation product, SYT–SSX, constitutes the components of Polycomb (Barco et al., 2009) are primary event in the cancer. SYT–SSX is therefore considered at least partially responsible for its role in believed to initiate tumorigenesis in its target stem cell. transcriptional regulation. SWI/SNF complexes may The data presented here allow a glimpse at the initial either activate or repress transcription, whereas Poly- events that likely occur when SYT–SSX2 is first comb complexes silence genes through chromatin expressed, and its dominant function in subverting the modification and compaction (Martens and Winston, nuclear program of the stem cell, leading to its aberrant 2003; Schuettengruber et al., 2007). As the SWI/SNF differentiation, as a first step toward transformation. In and Polycomb complexes are major regulators of addition, we identified the fibroblast differentiation-specific genes in development (Schuetten- gene, Fgfr2, as one occupied and upregulated by SYT– gruber et al., 2007), it can be anticipated that such genes SSX2. Knockdown of FGFR2 in both BMMSCs and SS would be targeted by SYT–SSX, resulting either in their cells abrogated their growth and attenuated their neural aberrant activation or repression, leading to tumorigen- phenotype. These results support the notion that the SYT– esis (Lubieniecka et al., 2008; Barco et al., 2009). Indeed SSX2 nuclear function and differentiation effects are the participation of developmental mediators in tumor conserved throughout sarcoma development and are formation and maintenance has been well-documented required for its maintenance beyond the initial phase. (Lobo et al., 2007). By nature of its interacting , They also provide the stem cell regulator, FGFR2, as a SYT–SSX is poised to mediate tumorigenesis through promising candidate target for future SS therapy. the transcriptional deregulation of developmental pro- Oncogene (2012) 31, 2323–2334; doi:10.1038/onc.2011.418; grams, however support for this hypothesis is lacking at published online 26 September 2011 the molecular level. The identities of SYT–SSX target genes are mostly unknown, and because SYT–SSX has Keywords: cancer; differentiation; gene expression no DNA-binding domain, the mechanism of its recruit- ment to specific loci is also unclear. Solving these questions will not only provide a deeper understanding of the nuclear function of SYT–SSX, but also illuminate Correspondence: Dr JE Eid, Department of Cancer Biology, potential therapeutic interventions for the treatment of SS. Vanderbilt University School of Medicine, 740 Preston Research Recently, a stem cell origin for SS was reported (Naka Building, 2220 Pierce Avenue, Nashville, TN 37232-0021, USA. E-mail: [email protected] et al., 2010). In those studies, SYT–SSX silencing in SS Received 23 February 2011;revised and accepted 15 August 2011; cells permitted their differentiation into multiple published online 26 September 2011 mesenchymal lineages, supporting the hypothesis that SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2324 SS arises in multipotent stem cells. Deregulation of program, while altering normal differentiation, in both normal differentiation driven by SYT–SSX is therefore murine myoblasts and human bone marrow-derived believed to be the basis for multipotent stem cell mesenchymal stem cells (BMMSCs). This programming transformation that leads to cancer development (Naka was due to the occupancy of an extensive number of et al., 2010). However, it remains to be determined how genes involved in neural differentiation and function by SYT–SSX expression in normal somatic stem cells the SYT–SSX2 nuclear complex. Moreover, SYT–SSX2 affects their differentiation. It will also be interesting appeared to target the network that controls stem cell to investigate if SYT–SSX expression results in the lineage commitment and differentiation, resulting in acquisition of additional features that influence tumor . One such direct target is the progression and behavior. fibroblast (Fgfr2) gene, whose Here we report that the SS-associated protein, SYT– activation, we believe, contributes to the differentiation SSX2, reprograms mesenchymal progenitor and stem effects driven by SYT–SSX2 in the mesenchymal stem/ cells by dictating their lineage commitment. Expression progenitor cells and which appear to be occurring in the of SYT–SSX2 results in the activation of a pro-neural SS tumor cells as well.

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2325 Stem cell differentiation is tightly controlled and its at http://www.ncbi.nlm.nih.gov/geo/) and can be ob- deregulation leads to disease formation. The SYT–SSX tained through accession numbers GSE26562 (C2C12 translocation is implicated in initiating the process of microarray), GSE26563 (BMMSC microarray) and sarcomagenesis. The present studies provide an impor- GSE26564 (C2C12 ChIPSeq)). Comparison of these tant insight into the primary events that unfold upon data to a microarray of eight human SSs (Nielsen et al., SYT–SSX expression in the undifferentiated stem cell 2002) identified nearly 100 upregulated genes shared from which the cancer is thought to arise. between SYT–SSX2-expressing myoblasts and human tumors (Supplementary Table S1). Strikingly, many of these shared genes were lineage determinants (Pth1r, Ngfr, Hoxb5 and Sox9) and mediators of developmental Results pathways such as fibroblast growth factor (FGF) (Fgfr2 and Fgfr3), Notch (Dll1), Hedgehog (Gli2) and Wnt SYT–SSX2 expression deregulates developmental (Tle4) (Figure 1a). Their expression was validated by programs and differentiation in myoblasts reverse transcription–PCR (RT–PCR) (Figure 1a, aster- SYT–SSX expression is sufficient to drive tumorigenesis isks). Furthermore, upregulation of various Wnt (Nagai et al., 2001; Haldar et al., 2007), and previous ligands, (Figure 1a, Wnt4 and Wnt11; Nielsen et al., studies showed that SYT–SSX fusions might alter the 2002) in both SYT–SSX2-expressing myoblasts and differentiation potential of SS cells (Ishibe et al., 2008; human SS tumors reflected a sustained activation of this Naka et al., 2010). What has been lacking, however, is a pathway. thorough analysis of the initial changes that occur in the An overall functional categorization of the upregu- mesenchymal precursor cell when SYT–SSX is ex- lated genes revealed 85 (12.2%) to be involved in pressed. We chose to conduct such analysis in C2C12 development (Figure 1b), the majority of which function myoblasts because they are a well-characterized, un- in lineage specification (Supplementary Table S2). transformed system of mesenchymal lineage capable of Similarly, 85 (12.6%) of the downregulated genes are differentiation into multiple cell types (Odelberg et al., involved in differentiation and development (Figure 1b). 2000). Additionally, in a transgenic SS model, SYT– Most striking, however, was the upregulation of 166 SSX2 expression in muscle progenitors formed tumors genes (23.9%) normally involved in neural differentia- that recapitulated the human disease (Haldar et al., tion and function (Figure 1b and Supplementary Table 2007), further indicating that myoblasts are a relevant S3). Notably, we observed a simultaneous downregula- model system. tion of 52 (7.7%) myogenic genes (Figure 1b), including After confirming their myogenic identity with a terminal differentiation markers (troponin and muscle marker profile and myotube formation (Supplementary myosins) as well as transcriptional controllers of Figure S1), we transduced C2C12 myoblasts with myogenesis (Myf5, MyoD and ; Supplemen- HA-Flag-SYT–SSX2 or a vector control, to define their tary Table S4 and Supplementary Figure S2A). These genetic programs. The microarray analysis generated data suggested a myogenic differentiation block that 700 upregulated and over 800 downregulated genes (all was confirmed when SYT–SSX2-expressing myoblasts microarray and chromatin immunoprecipitaton (ChIP)- failed to form multinucleated myotubes and continued Sequencing (ChIPSeq) data have been deposited in the to grow as mononucleated cells (Supplementary Figure Gene Expression Omnibus (Edgar et al., 2002; available S1). Altogether, SYT–SSX2 expression in myoblasts led

Figure 1 Alterations in cellular programs in myoblasts by SYT–SSX2. (a) RT–PCR in myoblasts expressing SYT–SSX2 (X) or vector control (V). Of 17 genes tested, 14 showed upregulated expression in SYT–SSX2 cells. The asterisks (*) denote genes shared by SYT– SSX2-expressing myoblasts and human SS tumors (Nielsen et al., 2002). GAPDH served as loading control. The X/V ratios represent fold increase in gene transcription, as measured by using the Fluorchem 8900 densitometer, with the AlphaEase FC program. ‘(b) Functional categorization of significantly (1.6-fold change) regulated genes in SYT–SSX2-positive cells. Of the total upregulated and downregulated genes, 695 upregulated (left) and 677 downregulated (right) genes had known functions and are represented in the pie charts. (c) Functional categorization of significantly upregulated genes that also have the SYT–SSX2 complex-binding sites between 0 and 10 kb upstream from their TSS. The functions of 159 genes were annotated and are represented in the pie chart. (d) Motif analysis of SYT–SSX2 ChIPSeq peaks. First column: A putative SYT–SSX2 complex-binding motif derived by MEME. Residue height is proportional to the probability of its presence at a given position. Second column: Number of peaks containing the motif. The E-score and P-value denote the statistical validity of the consensus sequence. Third column: Transcription factors that may bind to the given motif. (e) NEF formation in SYT–SSX2- and SXdel3-expressing myoblasts. Hemagglutinin (HA)-Flag-vector (top row), HA- Flag-SYT–SSX2 (middle row) and HA-Flag-SXdel3 cells were stained for HA (green) and NEF (red), and visualized by fluorescence microscopy. 40,6-diamidino-2-phenylindole (DAPI, blue) is the nuclear stain. The arrow indicates a cell with a long projection. The images were taken at  20 magnification. The diagram shows the deleted segment of SSX2 in SXdel3. A Flag immunoblot shows expression levels in V (vector), X (SYT–SSX2) and del3 (SXdel3) cellular lysates. The histogram represents the average ratio of NEF- positive cells to HA-positive cells in V control, X and del3 cells (n ¼ 3). The error bars denote the standard deviation. (f) FGFR2 RT– PCR in myoblasts expressing vector control (V), SYT–SSX2 (X) or SXdel3 (del3). The numbers represent the ratios of expression levels in X and del3 over V cells. Signal intensities were measured by using the Fluorchem 8900 densitometer, with the AlphaEase FC program. Lower panel: ChIP of SYT–SSX2 and SXdel3 at the peak located upstream from the Fgfr2 gene, using the anti-Flag antibody, in V control, X and del3 myoblasts. Whole-cell extract DNA (WCE) served as positive control. The histogram represents the average of SYT–SSX2, SXdel3 and control vector binding to the Fgfr2 peak, measured as percent input. Results from four ChIP–PCR experiments were used for the analysis. The error bars are the standard deviations. X:P denotes the significance of X values relative to V. Del3:P denotes the significance of SXdel3 values relative to X.

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2326 to the abrogation of myogenesis, with an apparent As a starting point, we selected genes with SYT–SSX2 concomitant reprogramming toward neural lineages. peaks situated up to 10 kb upstream from their TSS, in the event the oncogene, like other known transcription regulators (Farnham, 2009), binds beyond the tradi- Targeting of SYT–SSX2 to chromatin is required for tional promoter region. Moreover, by nature of their occupancy of neural genes and induction of the neural function, the Polycomb and SWI/SNF chromatin- phenotype modifying complexes SYT–SSX2 associates with, may To identify the specific subset of genes to which the direct its docking at sites farther from the traditional SYT–SSX2 complex is recruited, we conducted ChIPSeq 5-kb proximal regulatory region (Mateos-Langerak and analysis and determined the genome-wide occupancy of Cavalli, 2008). SYT–SSX2. We used myoblasts transduced with HA- Focusing on the window 10 kb upstream from gene Flag-SYT–SSX2 owing to lack of an appropriate TSSs, cross-validation of the microarray with the antibody that efficiently recognizes native SYT–SSX2 ChIPSeq data revealed that SYT–SSX2 was physically epitopes in ChIP experiments. ChIPSeq for SYT–SSX2 recruited to approximately 200 of the upregulated genes yielded over 19 million unique tags compared with over and to only 51 of the downregulated genes. Function- 16 million unique tags in the control (all microarray and ally, the downregulated 51 genes followed the general ChIPSeq data have been deposited in the Gene distribution of genes in the overall microarray (Supple- Expression Omnibus (Edgar et al., 2002; available at mentary Figure S2B). Strikingly, genes associated with http://www.ncbi.nlm.nih.gov/geo/) and can be obtained neural development and function were quite prevalent through accession numbers GSE26562 (C2C12 micro- (42.8%; 68 genes) among the upregulated 200 genes array), GSE26563 (BMMSC microarray) and bound by the SYT–SSX2 complex (Figure 1c). These GSE26564 (C2C12 ChIPSeq)). Putative SYT–SSX2 genes are active in different aspects of neural differ- target sites were determined by the Model-based entiation, including patterning, axon guidance, signaling Analysis of ChIPSeq program. This analysis validates and growth (Table 2). This is remarkable as C2C12 cells true peaks by calculating the significance of each are mesenchymal progenitors and do not naturally candidate peak relative to the control using a signifi- differentiate into neural lineages. cance threshold (Zhang et al., 2008). In our study, this Further analysis of the ChIPSeq peaks located method generated approximately 53 000 peaks with a upstream from the 200 genes (approximately 500 peaks maximum false discovery rate of 2.8% (Gene Expression total) derived a recruitment motif for SYT–SSX2, rich Omnibus accession GSM653014 provides direct access in C and T residues (Figure 1d, first column), and to SYT–SSX2 ChIPSeq peak data.). The ChIP peaks contained in 132peaks. This motif contains potential were categorized by their distance upstream from binding sites for a group of transcription factors transcription start sites (TSSs, þ 1) and the results are belonging to the homeodomain, and shown in Table 1. On the whole, the majority of peaks Sp1 families (Figure 1d, third column) known to be (approximately 60%) are located at distances greater involved in stem cell programming and differentiation. than 50 kb from TSSs. Closer to known genes, 20% of The extensive association of SYT–SSX2 with neural the peaks are located within 20 kb upstream from TSSs, genes led us to question if these myoblasts showed a with over 6000 sites (11.5% of the total peaks) between 0 matching phenotype. In a background of 80–90% and 10 kb (Table 1). A total of 3440 sites are located infection efficiency, 40% of SYT–SSX2-expressing within 5 kb upstream from TSSs corresponding to 1352 myoblasts expressed neurofilament (NEF; Figure 1e, genes (Table 1). Given that SYT–SSX2 associates with middle row, right panel and histogram) whereas control transcription regulators, we decided to analyze more cells showed minimal (o2%) NEF staining (Figure 1e, closely genes with SYT–SSX2 occupancy near their TSS. top row, right panel). The empty vector produces a short HA-Flag-peptide that allows the visualization of posi- tively infected control cells (Figure 1e, top row, middle panel). Moreover, oncogene-expressing cells showed Table 1 Distribution of SYT–SSX2 ChIP peaks relative to gene TSSs long projections (Figure 1e, arrow), similar to a and corresponding genes phenotype we observed in SYT–SSX2-expressing fibro- blasts (Barco et al., 2007) and consistent with the Distance Number Percentage of Total number (kb) of peaks total peaks of genes neurogenic features noted in SS cells (Ishibe et al., 2008). Overall, stimulation of a pro-neural program appears to 0–5 3440 6.5 1352 be a pronounced feature of SYT–SSX2. 5–10 2654 5 1076 Mediators of the Wnt, Hedgehog and FGF pathways, 10–15 2400 4.5 1016 15–20 2287 4.3 933 and differentiation formed an additional 11.9% (19 20–50 10 230 19.3 2026 genes) of the 200 genes occupied by SYT–SSX2 50–100 10 312 19.5 1693 (Figure 1c; Supplementary Table S2, asterisks). In 100–150 6035 11.4 973 particular, we noticed the presence of FGF mediators 150–200 3956 7.5 659 4200 11 678 22 984 throughout our analyses. By microarray, a number of FGF pathway members were upregulated (Supplemen- Abbreviations: ChIP, chromatin immunoprecipitaton; TSS, transcrip- tary Table S2), and increased expression of FGF tion start site. receptors, Fgfr2 and Fgfr3, was confirmed by RT–PCR

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2327 Table 2 Selected list of upregulated genes bound by the SYT–SSX2 complex involved in neural development and function Symbol Gene description Symbol Gene description

Development and differentiation Bhlhe23 Basic helix–loop–helix, e23 Fezf2 Fez family zinc finger L1cam Cell adhesion molecule Olig2 Oligodendrocyte lineage factor Prox1 Prospero Ptpru Protein tyrosine phosphatase Zcchc12 Zinc finger with CCHC domain Zic2 Zinc-finger protein of cerebellum

Patterning and axon guidance Crmp1 Collapsin response mediator Dpysl5 Dihydropyrimidinase-like-5 Epha8 Eph receptor-A8 Efnb1 -B1 Ephb1 Eph receptor-B1 Slit3 Slit homolog-3 Unc5a Netrin receptor

Neurotransmitter signaling and metabolism Abat Aminotransferase Adra2c Adrenergic receptor Cacna1h Calcium channel Cacng5 Calcium channel Chrna4 Cholinergic receptor Grm4 Glutamate receptor Kcnip3 Kv channel-interacting protein-3 Nptx1 Neuronal pentraxin-1 Slc6a1 GABA transporter Th Tyrosine hydroxylase

Neuropeptide, lipid and hormone signaling Cck Cholecystokinin Faah Fatty acid amide hydrolase Gpr50 G-protein-coupled receptor-50 Gal Galanin Mgll Monoglyceride lipase Ntsr1 Neurotensin receptor-1 Pdyn Prodynorphin Sst Somatostatin

Adhesion, growth and survival Amigo2 Adhesion molecule Bai1 Brain inhibitor-1 Bai2 Brain angiogenesis inhibitor-2 Cdh23 Cadherin-23 Gjb2 Gap junction protein Ngfr receptor

(Figure 1a). Importantly, the same FGFR2 and FGFR3 SYT–SSX2 causes aberrant differentiation in human were also upregulated in human SSs (Supplementary mesenchymal stem cells Table S1; Nielsen et al., 2002). ChIPSeq analysis Myoblast reprogramming by SYT–SSX2 prompted us indicated that the Fgfr2 gene is directly targeted by to question whether dictating lineage commitment in SYT–SSX2, and further ChIP experiments confirmed undifferentiated precursors is an intrinsic feature of the the presence of SYT–SSX2 at the peak located 4.3 kb oncogene. As SS is thought to arise in a mesenchymal upstream from the Fgfr2 gene (Figure 1f, ChIP–PCR stem cell (Mackall et al., 2004; Naka et al., 2010), we panel). Notably, the Fgfr2 peak contains a sequence questioned whether SYT–SSX2 expression could matching the SYT–SSX2 recruiting motif (Figure 1d). elicit similar effects in multipotent, human BMMSCs Thus, the FGF receptor appears to be a direct target of (Colter et al., 2001; Sekiya et al., 2002; Supplementary the oncogene. Figure S3; Supplementary Materials and methods). SYT–SSX2 associates with Polycomb complexes, Searching for a neural phenotype in SYT–SSX2-expres- modulators of chromatin and lineage determination. sing BMMSCs, we observed a robust and heterogeneous To determine whether the ability of SYT–SSX2 to target NEF expression in a significant population (Figure 2a, chromatin is required for the observed effects, we tested bottom row, right panel and arrowheads). Neither naı¨ve SXdel3, an SYT–SSX2 mutant with a 20-residue nor vector-expressing BMMSCs produced NEFs deletion in its SSX-targeting module (Figure 1e, dia- (Figure 2a, top and middle rows, right panels). gram). SXdel3 is unable to colocalize with Polycomb We next asked whether SYT–SSX2 influenced the and antagonize its Bmi1 component in U2OS cells ability of BMMSCs to differentiate into their normal (Barco et al., 2009). When assayed in C2C12 cells, lineages. We discovered that oncogene expression SXdel3 failed to induce NEF formation (Figure 1e, caused a marked inhibition of adipogenesis, whereas bottom row, right panel), indicating an inactive neural naı¨ve and vector-expressing cells differentiated normally program. Furthermore, we observed that the ability of (Figure 2b, top row). By contrast, SYT–SSX2 expres- SXdel3 to upregulate FGFR2 expression (Figure 1e, sion accelerated osteogenesis as evidenced by an intense RT–PCR panel), or bind upstream from the gene alkaline phosphatase staining 48 h after infection with- (Figure 1e, ChIP–PCR panel and histogram), was out addition of osteogenic factors (Figure 2b, bottom markedly diminished. To summarize, these studies row, right panel). As expected, in the absence of demonstrate that SYT–SSX2 activates a pro-neural inducing factors, naı¨ve and vector-expressing BMMSCs program and blocks normal myogenesis. Its ability to showed minimal alkaline phosphatase staining, bind chromatin is required for its transcriptional and (Figure 2b, bottom row, left and middle panels). phenotypic effects. Alkaline phosphatase positivity was heterogeneous

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2328

Figure 2 SYT–SSX2 deregulates differentiation in mesenchymal stem cells. (a) NEF expression in SYT–SSX2-expressing BMMSCs. Naı¨ve, vector control (V) and SYT–SSX2 (X)-expressing BMMSCs were stained for HA (green) and NEF (red). The images were taken at  20 magnification. The arrowheads indicate heterogenous expression of NEF. The Flag immunoblot shows SYT–SSX2 expression in X-expressing cells. The histogram represents the average ratio of NEF-positive cells to HA-positive V and X cells (n ¼ 3). The error bars represent the standard deviation. (b) SYT–SSX2’s effects on BMMSCs differentiation. Bright-field microscopy of naı¨ve, vector- and SYT–SSX2-expressing cells stained with Oil Red O after adipogenic stimulation (top row) or for alkaline phosphatase expression (bottom row) without osteogenic stimulation. The images were taken at  20 magnification. The arrow indicates heterogeneous expression of alkaline phosphatase. (c) Functional categorization of significantly (2.0-fold change) regulated genes in SYT–SSX2-expressing BMMSCs. The functions of 735 upregulated (upper chart) and 506 downregulated (lower chart) genes were documented, and these genes are represented in the pie charts.

(Figure 2b, arrow), indicating that the early osteogenesis www.ncbi.nlm.nih.gov/geo/) and can be obtained was activated at varying degrees across the cell popula- through accession numbers GSE26562 (C2C12 micro- tion. Inhibition of adipogenesis and acceleration of array), GSE26563 (BMMSC microarray) and osteogenesis by SYT–SSX2 were observed in two GSE26564 (C2C12 ChIPSeq)). Functional categoriza- additional BMMSCs lines, one human (Supplementary tion of the upregulated genes revealed nearly one-third Figure S4 and Materials and methods), and one murine (27.8%) to participate in neural differentiation and (Alfaro et al., 2008; data not shown). Altogether, these signaling (Figure 2c and Table 3). Notably, several of data suggest that SYT–SSX2 induces a neural and/or these genes were also shown to be occupied and osteogenic program(s) in BMMSCs, while inhibiting upregulated by SYT–SSX2 in myoblasts (Table 3, their adipogenic differentiation. footnote). This implies that SYT–SSX2 targets the same A full characterization of the gene expression profiles genes and promotes neural programs regardless of cell initiated by SYT–SSX2 in BMMSCs identified approxi- type. By contrast, promoters of osteoblast differentia- mately 750 significantly upregulated and more than 500 tion (BMP2, BMP6, FGFR3 and OSR2) represented significantly downregulated genes when compared with 1.9% of the upregulated genes (Figure 2c and Supple- vector-transduced BMMSCs (all microarray and ChIP- mentary Table S5). Taken together, these data suggest Seq data have been deposited in the Gene Expression that SYT–SSX2 indeed activates programs of neural Omnibus (Edgar et al., 2002; available at http:// and osteogenic differentiation in BMMSCs.

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2329 Table 3 Selected list of genes involved in neural development and function upregulated by SYT–SSX2 in human mesenchymal stem cells Symbol Gene description Symbol Gene description

Development and differentiation ENC1 Ectodermal-neural cortex EYA4 Eyes absent homolog FGF11 -11 GBX2 Brain homeobox-2 L1CAMa Cell adhesion molecule NEUROD1 Neurogenic differentiation NEUROG3 Neurogenin-3 PROX1a Prospero homeobox-1

Patterning and axon guidance CRMP1a Collapsin response mediator DPYSL5a Dihydropyrimidinase-like-5 EFNA4 Ephrin-A4 EFNB1a Ephrin-B1 EPHA3 EPH receptor-A3 EPHB1a EPH receptor-B1 GLDN Gliomedin NRP2 Neuropilin-2 ROBO1 Roundabout SEMA3D SLIT1 Slit homolog-1 UNC5Aa Netrin receptor

Neurotransmitter signaling and metabolism ABATa Aminotransferase ADRA1D Adrenergic, receptor CHRNA4a Cholinergic receptor DRD2 Dopamine receptor-D2 GABRE GABA receptor, epsilon GATM Glutamate decarboxylase GRIK3a Glutamate receptor GRM4a Glutamate receptor metabotropic SLC18A3a Acetylcholine transporter

Neuropeptide, lipid and hormone signaling CRHR1a Neuropeptide receptor GALa Galanin GPR50a G-protein-coupled receptor-50 NPY Neuropeptide-Y NTSR1a Neurotensin receptor-1 PCSK1 Convertase subtilisin/kexin PNOC Pronociceptin SSTa Somatostatin

Adhesion, growth and survival AREG BAI2a Brain angiogenesis inhibitor-2 GFRA2 GDNF receptor GPC4 Glypican-4 NCAM1 Neural cell adhesion molecule-1 NGFRa Nerve growth factor receptor PCDH10 Protocadherin-10 TPPP3 Tubulin polymerization aUpregulated genes in BMMSCs that are also occupied and upregulated by SYT–SSX2 in C2C12 myoblasts.

Stem cell controllers such as Wnt, Notch, transform- assess the contribution of FGFR2 to the visible effects ing growth factor-b/bone morphogenetic protein of SYT–SSX2, we decided to analyze the consequences (BMP), Shh and FGF (Figure 2c and Table 4) of its inhibition in the stem cells. NEF formation and constituted 16% of the upregulated genes. Three of cell growth were both used as read-outs to measure the these genes, AXIN2, Shh and FGFR2, were also dependence of SYT–SSX2-expressing BMMSCs on upregulated and occupied by SYT–SSX2 in myoblasts FGFR2. We started by inhibiting FGFR activity with (Table 4, footnote). PD173074, a small molecule with high selectivity for the Notably, the C2C12 and the BMMCs microarrays FGFR kinase (Pardo et al., 2009). A 2-day treatment overlapped with 248 differentially expressed genes, 85 with PD173074 led to a marked diminution of NEF (34%) of which belonged to the neurogenic program signal in SYT–SSX2-expressing BMMSCs (Figure 3a, and 54 (B22%) were developmental mediators and left histogram), reflecting the dependence of the neural transcription factors (Supplementary Table S6). marker on active FGFR. More specifically, infection of SYT–SSX2-expressing BMMSCs with two FGFR2- specific short-hairpin RNA (shRNA) vectors (833 and The role of FGFR2 in SYT–SSX2’s differentiation effects 703; Figure 3a, middle panel) showed significant growth Throughout our analyses Fgfr2 held our interest as it retardation when compared with a non-targeting vector was noticeably upregulated not only in BMMSCs and (2910; Figure 3a, right histogram, dark gray bars). myoblasts but also in human SS tumors (Nielsen et al., Apart from growth inhibition, FGFR2 depletion caused 2002). Moreover, our ChIPSeq analysis revealed Fgfr2 a specific attenuation of the NEF signal in the SYT– as a direct target of SYT–SSX2. Importantly, FGFR2 is SSX2-expressing BMMSCs cells (more pronounced with a major inducer of both osteogenesis and neurogenesis 703; Figure 3a, right histogram, light-gray bars). during development (Huang et al., 2007; Villegas et al., Importantly, the 2910, 833 and 703 vectors did not 2010) and could be contributing, in part, to the shift in affect the growth of vector control BMMSCs. These lineage commitment seen in human BMMSCs. FGFR2 findings suggest that FGFR2 signaling is required for was, therefore, our prime candidate for an upstream the proper growth of SYT–SSX2-expressing mesenchy- signaling pathway whose activation would explain the mal stem cells and the expression of neural differentia- induction of the neural cascade by SYT–SSX2. To tion markers.

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2330 Table 4 Selected list of developmental pathway mediators upregulated by SYT–SSX2 in human mesenchymal stem cells Symbol Gene description Symbol Gene description

WNT AXIN2a Conductin, axil SFRP1 Secreted -related protein DACT1 Antagonist of b-catenin TLE1 Transducin-like enhancer protein FZD3 Frizzled homolog TLE2 Transducin-like enhancer protein Frizzled-related protein TLE3 Transducin-like enhancer protein KREMEN1 Kringle-containing protein WNT4 WNT -4 LEF1 Lymphoid enhancer-binding factor WNT7B WNT ligand-7B PRICKLE1 Prickle homolog WNT11 WNT ligand-11 RSPO1 R-spondin homolog

NOTCH DLL1 Delta-like JAG1 Jagged-1 DTX1 Deltex homolog JAG2 Jagged-2 DTX4 Deltex-4 homolog LFNG Lunatic fringe HES1 Hairy and enhancer of split NOTCH1 Notch homolog HEY2 HES-related with YRPW motif SIX1 SIX homeobox

TGFb/BMP BAMBI BMP and activin inhibitor FAM46C Family with sequence similarity-46 BMP3 BMP ligand-3 FSTL4 -like BMP4 BMP ligand-4 GDF6 Growth differentiation factor BMP7 BMP ligand-7 SOST Sclerosteosis BMPER BMP-binding endothelial regulator TGFB2 Transforming growth factor, b

SHH PTCH1 Patched homolog-1 PTCHD2 Patched domain-2 PTCHD1 Patched domain-1 SHHa Sonic hedgehog homolog

FGF FGFBP2 FGF-binding protein-2 FGFR2a Fibroblast growth factor receptor-2

aUpregulated genes in BMMSCs that are also occupied and upregulated by SYT–SSX2 in C2C12 myoblasts.

Figure 3 Contribution of FGFR2 to SYT–SSX2’s differentiation effects and to cell growth. (a) Loss of neurite extensions and NEF signal intensity after inhibition of FGF signaling in SYT–SSX2 (HA-positive) BMMSCs. The top left image depicts a reference NEF (red)-positive, SYT–SSX2-expressing BMMSC. The left histogram represents the average ratio of NEF-positive to HA-positive cells 2 days after treatment with PD173074 at the indicated concentrations (n ¼ 4; approximately 1000 cells were included for each concentration). D is vehicle DMSO. The error bars denote the standard deviation. The P-values reflect the significance of the experimental values compared with the vehicle (D). Middle panel: Immunoblot of FGFR2 levels in SYT–SSX2-expressing BMMSCs infected with the indicated FGFR2-shRNAs. 2910 is the non-targeting vector and tubulin is loading control. The numbers indicate the ratio of FGFR2 signal in the cells expressing targeting shRNAs relative to non-targeting vector (value 1). Right histogram: The dark gray bars are the average of 833 and 703 cell number over 2910 (value 1). The 2910, 833 and 703 originated from the same SYT–SSX2- expressing BMMSCs pool (n ¼ 3). The light-gray bars are the average ratio of NEF-positive 833 and 703 cells over 2910 NEF-positive (value 1) cells. The error bars indicate the standard deviation. The P-values indicate the significance of the experimental values with the targeting shRNAs as compared with non-targeting vector (2910). (b) Decreased NEF expression and growth of SS SYO-1 cells after inhibition of FGF signaling. The left image panel depicts the NEF signal (red) with increasing concentrations of PD173074 in SYO-1 cells. Nuclear SYT–SSX2 (green) was visualized with the anti-SSX2 B56 antibody. DMSO was the vehicle control. The images were taken at  20 magnification. Middle upper histogram: The average ratio of NEF-positive cells exposed to DMSO (D) or PD173074 to untreated (U; value 1) SYO-1 cells (n ¼ 2; over 1000 cells were included in each category). The error bars indicate the standard deviation. The P-value reflects the significance of the experimental values as compared with the vehicle (D). The middle lower histogram shows growth inhibition of SYO-1 cells with increasing concentrations of PD173074 (n ¼ 2). Cell growth was estimated by using the SRB colorimetric assay. The error bars represent the standard deviation. The P-value reflects the significance of the experimental values as compared with the vehicle (D). The immunoblot shows the FGFR2 levels in shRNA-treated SYO-1 cells. Tubulin is the loading control. The numbers indicate the ratio of FGFR2 signal in targeting shRNA cells relative to non-targeting vector (2910). Upper right histogram: The effect of 2910, 833 and 703 FGFR2-shRNAs on NEF expression in SYO-1 cells, relative to NEF-positive naı¨ve (N; value 1) cells. The error bars represent the standard deviation (n ¼ 3; approximately 1000 cells were included for each category). The P-value indicates significance of the experimental values with the targeting shRNAs as compared with the non-targeting vector (2910). The lower right histogram demonstrates the effect of the three FGFR2-shRNAs on SYO-1 growth by using the SRB assay (n ¼ 2). The error bars represent the standard deviation. The P-value indicates the significance of the experimental values with the targeting shRNAs as compared with the non-targeting vector (2910). (c)Effect of SYT–SSX2 small interfering RNA in SYO-1 cells. Left immunoblot: SYT–SSX2 levels in INV control and two SSX2-targeting RNAs (Si-SSX2A and Si-SSX2B) in SYO-1 lysates detected with the antibodies B56 (anti-SSX2) and SV11 (anti-SYT). Tubulin is the loading control. Middle immunoblot: FGFR2 levels in the same lysates. The numbers indicate the ratio of FGFR2 signal with the targeting Si- SSX2 SiRNAs over a control RNA (INV). Histogram: The effect of SYT–SSX2 small interfering RNA on NEF formation in SYO-1 cells. The numbers indicate the average ratio of NEF-positive Si-SSX2A and Si-SSX2B cells to NEF-positive INV control cells (value 1). The error bars denote the standard deviation (n ¼ 3; over 1000 cells were counted for each category). The P-value indicates the significance of the experimental values with the targeting Si-SSX2 SiRNAs as compared with the control RNA (INV). Measurements of FGFR2 depletion by the targeting shRNAs, or by the SYT-SSX2 SiRNAs, were performed by using the Fluorchem 8900 densitometer, and analyzed with the AlphaEase FC software.

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2331 We then repeated these analyses in the human SYO-1 their growth (Figure 3b, left and middle panels). As in SS cells that carry the SYT–SSX2 translocation (Kawai the SYT–SSX2-expressing BMMSCs, FGFR2 depletion et al., 2004). We observed that approximately 15% of with the 833 and 703 shRNAs also led to a marked the SYO-1 cell population contained NEF, and decrease in the number of NEF-positive SYO-1 cells as PD173074 caused a graded disappearance of NEF- well as a delay in their growth (Figure 3b, right panel). positive SYO-1 cells and an incremental inhibition of We next asked whether these events in SYO-1 cells are

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2332 dependent on SYT–SSX2 expression. We found that Whether SYT–SSX2 acts in a similar manner remains to depletion of SYT–SSX2 in SYO-1 cells with specific be seen. Regardless, the dominant effect on cellular small interfering RNAs (Figure 3c, left panel) was identity is thought to be a part of oncogenesis initiation by accompanied by a concomitant decrease in FGFR2 sarcoma-associated translocations and a necessary step levels (Figure 3c, middle panel) and a marked decrease toward malignant transformation (Mackall et al., 2004). in the relative number of NEF-positive cells (Figure 3c, These observations allow us to speculate on the cell- histogram). We refrained from measuring the effect of of-origin for this malignancy. The capacity of SS cells to SYT–SSX2 depletion on SYO-1 growth, as the inherent be differentiated into mesenchymal and neural cell types cell toxicity of RNA interference assays would interfere (Ishibe et al., 2008; Naka et al., 2010) implied that the with its accuracy. disease originates in multipotent cells from either of In summary, these studies suggest that SYT–SSX2 these lineages. Our data indicate that the neural features recruitment to the Fgfr2 gene results in the activation of are primarily caused by SYT–SSX2 itself, irrespective of FGF signaling, thereby driving the neural phenotype in the cellular context, so the target cell may not necessarily BMMSCs and affecting their growth. This mechanism be of neural origin. Expression of SYT–SSX2 in appears to be occurring in the human SS cells as well. multiple lineages in mice recapitulates human SS in all cases, attesting to the dominant program established by the oncogene and its capacity to transform different cell types (Haldar et al., 2009). Additionally, expression of Discussion SYT–SSX2 in committed myogenic progenitor cells resulted in tumor formation in mice (Haldar et al., 2007) In the present report, we show that the SS oncogene suggesting that the cell-of-origin could be a more SYT–SSX2 reprograms mesenchymal stem/progenitor differentiated entity. However, in this model, genomic cells by activating a pro-neural gene network while plasticity was essential, as SYT–SSX2 was non-tumori- disrupting normal differentiation. This was most likely genic in differentiated muscle cells (Haldar et al., 2007). due to the recruitment of SYT–SSX2 to an extensive array The upregulation of several mediators representing of neural genes, resulting in their activation. This central pathways known to modulate stem cell behavior corroborates previous reports in which a neural phenotype was another striking result. It uncovered a propensity of was observed in SYT–SSX-expressing SS cell lines (Ishibe SYT–SSX2 for regulating developmental pathways. et al., 2008). Furthermore, knockdown of SYT–SSX in SS This may reflect an ability of SYT–SSX2 to create an cells led to a loss of neuronal features (the present study imbalance in the microenvironment of the cancer cell and Naka et al.,2010).Wewerealsoabletoshowthe in vivo, furthering malignancy. We have previously dependence of this neural phenotype on FGF signaling. reported that SYT–SSX2 mediates the nuclear translo- Genes with SYT–SSX2 binding within 10 kb from cation and activation of b-catenin (Pretto et al., 2006). their TSS were analyzed. An association between SYT– Consistent with this finding, upregulation of Wnt SSX2’s ability to target chromatin and its differentiation components was seen in our microarray studies. The effects was established. However, the majority of the crosstalk among Wnt, transforming growth factor-b/ binding sites are situated farther than 20 kb away from BMP, FGF, Hedgehog and Notch, and their impact on the TSS, pointing to a likely long-range function of tumor cell behavior, are the focus of future studies. SYT–SSX2. This is to be expected, given its association Our high-throughput analyses identified FGFR2 as a with Polycomb. Thus, the oncogene may regulate gene critical signaling node in the behavior of SYT–SSX2- expression at faraway distances, and/or is involved in expressing cells. Its enhanced signaling by SYT–SSX2 nuclear patterning (Mateos-Langerak and Cavalli, may explain the accelerated osteoblastogenesis as well as 2008). Further analysis of SYT–SSX2 occupancy to the dominance of the pro-neural gene profile. With clarify its long-range functions is underway. -activated (MAPK)/extracellular Only a subset (17%) of regulated genes was bound by signal-regulated kinase (ERK) and phosphatidylinosi- the SYT–SSX2 complex. This may be attributed to the tol-3-kinase (PI3K) activation, FGFR2 signaling pro- affinity of SYT–SSX2 for mediators of stem cell motes neurogenesis and skeletogenesis through crosstalk pathways, as each may be responsible for altering the with Wnt, Hedgehog, Notch and BMP signals (Ever and expression of numerous genes. Gaiano, 2005; Chadashvili and Peterson, 2006; Maric The remarkable number of neural and developmental et al., 2007; Zhao et al., 2008; Miraoui and Marie, 2010). genes shared by the myoblasts and the BMMSCs Furthermore, the benefit of FGF pathway attenuation showcases the dominant programming effect of SYT– to inhibit SS cell growth was reported previously (Ishibe SSX2. Imposing a lineage commitment on stem/pro- et al., 2005) and corroborated by our studies. Chemical genitor cells appears to be a recurrent feature of inhibition of FGFR2 signaling and its depletion using sarcoma-associated translocations (Mackall et al., shRNA caused loss of NEF expression and decreased 2004). One prominent example is PAX3-FKHR, the cell growth in both SYT–SSX2 BMMSCs and SS SYO-1 rhabdomyosarcoma fusion product that drove NIH-3T3 tumor cells. Significantly, upregulation of FGF ligands fibroblasts into a myogenic program (Khan et al., 1999). in the myoblast and BMMSC microarrays suggests that It is thought to induce tumorigenesis through stimulation SYT–SSX2 establishes an autocrine FGF signaling of lineage commitment and simultaneous prevention of loop. If this is the case, identification of FGFR2 as the terminal differentiation (Charytonowicz et al., 2009). mediator of these signals designates it as a candidate for

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2333 potential SS tumor reversal. Increased FGFR2 activity Materials and methods is already linked to advanced malignant phenotypes in endometrial, uterine, ovarian, breast, lung and gastric Cell culture and reagents cancers. Strategies designed to target FGFR2 in these Lists of the cell lines, antibodies and chemicals are provided in cancers (Katoh, 2008; Katoh and Katoh, 2009) are the Supplementary Materials and methods. ongoing. The deregulation of differentiation in our model Analysis of ChIP DNA by next-generation sequencing systems can also be explained by these findings. Preparation of DNA to be sequenced by the Illumina Genome FGFR2-induced osteoblast maturation (Miraoui and Analyzer II in the Vanderbilt Genome Technology Core is Marie, 2010) inhibits adipogenesis in mesenchymal stem detailed in the Supplementary Materials and methods. cells (Muruganandan et al., 2009). Similarly, the stimulation of a neural program by SYT–SSX2 may Analysis of SYT–SSX2 ChIPSeq have abrogated myogenesis. In C2C12 cells, the two The Illumina Analysis Pipeline was used for image analysis outcomes were shown to be mutually exclusive and base calling. Analysis of the Sequence Alignment/Map (Watanabe et al., 2004). Alternatively, direct silencing (SAM) files by the Model-based Analysis of ChIPSeq program of myogenic genes could have also contributed to this and calculations of nearest distance of peaks to annotated gene phenotype. The ChIPSeq analysis revealed a putative TSSs are described in the Supplementary Materials and methods. SYT–SSX2-binding site upstream from the downregu- lated MyoD gene. Additional studies are underway to test this possibility and identify potential recruitment Cross-validation of C2C12 microarray by ChIPSeq factors associated with transcriptional silencing by Annotation of peaks to regulated genes is described in the SYT–SSX2. Supplementary Materials and methods. While SYT–SSX2 transgenic myoblasts equivalent to C2C12 (Asp et al., 2011) developed tumors in mice Motif analysis (Haldar et al., 2007), neither the C2C12 cells nor the Use of the MEME program to derive the consensus motif for BMMSCs were transformed by SYT–SSX2 in vitro. One binding is detailed in the Supplementary Materials and interpretation is the requirement of a surrounding methods. stroma to provide a self-renewing, tumor-promoting environment. Alternatively, our observations may re- Additional methods present a priming step in cancer initiation by SYT– Retroviral infections, microarray, ChIP, RT–PCR and ChIP– SSX2, and a permissive epigenetic or genetic event is PCR, immunofluorescence, differentiation, SRB, small inter- required for full transformation. fering RNA and shRNA experiments were performed by using In summary, our studies in mesenchymal stem and standard protocols. Please refer to Supplementary Materials and methods for details. progenitor cells have uncovered a function of SYT– SSX2 in differentiation programming, and our genome- wide analyses provided a glimpse into the early events of tumor initiation by the oncogene. The question is how Conflict of interest SYT–SSX2 imposes a heritable gene-regulatory state that redirects cell fate (Gurdon and Melton, 2008). We The authors declare no conflict of interest. will acquire a better understanding once the defined factors are identified. Overall, we believe that the deregulation of differ- Acknowledgements entiation is a manifestation of the ability of SYT–SSX2 to target lineage-specific programs. FGFR2 was identi- Funding was provided by the Alex’s Lemonade Stand fied as a cardinal player in SYT–SSX2-associated Foundation (Innovation Award) and the Sarcoma Foundation phenotypes, but it is likely that additional pathway of America. Microarrays were performed at the Vanderbilt mediators also contribute to SYT–SSX2-induced char- Functional Genomics Shared Resource core supported by the acteristics. Future investigation of other targets identi- VICC (P30 CA68485), the VDCC (P30 DK58404) and the fied through this method will lead to a better Vanderbilt Vision Center (P30 EY08126). Some of the understanding of the interplay among these pathways materials used in this work were provided by the Texas A&M Health Science Center College of Medicine Institute for and SS pathology. This combination analysis also Regenerative Medicine at Scott & White through a grant from provides a powerful tool in the discovery of novel NCRR of the NIH (grant no. P40RR017447). We thank BV therapeutic targets and will be advantageous in under- Mary and H Trinity for assistance; T Ito and M Ladanyi for standing the biology of other oncogenic proteins directly SYO-1 cells and A Sandelin and S Wilhite for technical affecting transcriptional programs. support.

References

Alfaro MP, Pagni M, Vincent A, Atkinson J, Hill MF, Cates J et al. engraftment, granulation tissue formation and myocardial repair. (2008). The Wnt modulator sFRP2 enhances mesenchymal stem cell Proc Natl Acad Sci USA 105: 18366–18371.

Oncogene SYT–SSX2-mediated lineage reprogramming CB Garcia et al 2334 Asp P, Blum R, Vethantham V, Parisi F, Micsinai M, Cheng J et al. Lubieniecka JM, de Bruijn DR, Su L, van Dijk AHA, Subramanian S, (2011). Genome-wide remodeling of the epigenetic landscape during van de Rijn M et al. (2008). Histone deacetylase inhibitors reverse myogenic differentiation. Proc Natl Acad Sci USA 108: E149–E158. SS18–SSX2-mediated polycomb silencing of the tumor suppressor early Barco R, Hunt LB, Frump AL, Garcia CB, Benesh A, Caldwell RL growth response 1 in synovial sarcoma. Cancer Res 68: 4303–4310. et al. (2007). The synovial sarcoma SYT–SSX2 oncogene remodels Mackall CL, Meltzer PS, Helman LJ. (2004). Focus on sarcomas. the cytoskeleton through activation of the ephrin pathway. Mol Biol Cancer Cell 2: 175–178. Cell 18: 4003–4012. Maric D, Fiorio Pla A, Chang YH, Barker JL. (2007). Self-renewing Barco R, Garcia CB, Eid JE. (2009). The synovial sarcoma-associated and differentiating properties of cortical neural stem cells are SYT–SSX2 oncogene antagonizes the polycomb complex protein selectively regulated by basic fibroblast growth factor (FGF) Bmi1. PLoS One 4: e5060. signaling via specific FGF receptors. J Neurosci 27: 1836–1852. Chadashvili T, Peterson DA. (2006). Cytoarchitecture of fibroblast Martens JA, Winston F. (2003). Current advances in understanding growth factor receptor 2 (FGFR-2) immunoreactivity in chromatin remodeling by Swi/Snf complexes. Curr Opin Genet Dev of neurogenic and non-neurogenic regions of the young adult and 13: 136–142. aged rat brain. J Comp Neurol 498: 1–15. Mateos-Langerak J, Cavalli G. (2008). Polycomb group proteins and Charytonowicz E, Cordon-Cardo C, Matushansky I, Ziman M. long-range gene regulation. Adv Genet 61: 45–66. (2009). Alveolar rhabdomyosarcoma: is the cell of origin a Miraoui H, Marie PJ. (2010). Fibroblast growth factor receptor mesenchymal stem cell? Cancer Lett 279: 126–136. signaling crosstalk in skeletogenesis. Sci Signal 3: re9. Colter DC, Sekiya I, Prockop DJ. (2001). Identification of a Muruganandan S, Roman AA, Sinal CJ. (2009). Adipocyte differ- subpopulation of rapidly self-renewing and multipotential adult entiation of bone marrow-derived mesenchymal stem cells: crosstalk stem cells in colonies of human marrow stromal cells. Proc Natl with the osteoblastogenic program. Cell Mol Life Sci 66: 236–253. Acad Sci USA 98: 7841–7845. Nagai M, Tanaka S, Tsuda M, Endo S, Kato H, Sonobe H et al. Edgar R, Domrachev M, Lash AE. (2002). Gene expression omnibus: (2001). Analysis of transforming activity of human synovial NCBI gene expression and hybridization array data repository. sarcoma-associated chimeric protein SYT–SSX1 bound to chroma- Nucleic Acids Res 30: 207–210. tin remodeling factor hBRM/hSNF2a. Proc Natl Acad Sci USA 98: Ever L, Gaiano N. (2005). Radial ‘glial’ progenitors: neurogenesis and 3843–3848. signaling. Curr Opin Neurobiol 15: 29–33. Naka N, Takenaka S, Araki N, Miwa T, Nobuyuki H, Yoshioka K Farnham PJ. (2009). Insights from genomic profiling of transcription et al. (2010). Synovial sarcoma is a stem cell malignancy. Stem Cells factors. Nat Rev Genet 10: 605–616. 28: 1119–1131. Gurdon JB, Melton DA. (2008). Nuclear reprogramming in cells. Nielsen TO, West RB, Linn SC, Alter O, Knowling MA, O’Connell JX Science 322: 1811–1815. et al. (2002). Molecular characterisation of soft tissue tumours: a Haldar M, Hancock JD, Coffin CM, Lessnick SL, Capecchi MR. gene expression study. Lancet 359: 1301–1307. (2007). A conditional mouse model of synovial sarcoma: insights Odelberg SJ, Kohlhoff A, Keating MT. (2000). Dedifferentiation of into a myogenic origin. Cancer Cell 11: 375–388. mammalian myotubes induced by . Cell 103: 1099–1109. Haldar M, Hedberg ML, Hockin MF, Capecchi MR. (2009). A Pardo OE, Latigo J, Jeffery RE, Nye E, Poulsom R, Spencer-Dene B CreER-based random induction strategy for modeling transloca- et al. (2009). The fibroblast growth factor receptor inhibitor tion-associated sarcomas in mice. Cancer Res 69: 3657–3664. PD173074 blocks small cell lung cancer growth in vitro and Huang W, Yang S, Shao J, Li YP. (2007). Signaling and transcrip- in vivo. Cancer Res 69: 8645–8651. tional regulation in osteoblast commitment and differentiation. Pretto D, Barco R, Rivera J, Neel N, Gustavson MD, Eid JE. (2006). Front Biosci 12: 3068–3092. The synovial sarcoma translocation protein SYT–SSX2 recruits b- Ishibe T, Nakayama T, Aoyama T, Nakamura T, Toguchida J. (2008). catenin to the nucleus and associates with it in an active complex. Neuronal differentiation of synovial sarcoma and its therapeutic Oncogene 25: 3661–3669. application. Clin Orthop Relat Res 466: 2147–2155. Pruitt KD, Tatusova T, Maglott DR. (2007). NCBI Reference Sequence Ishibe T, Nakayama T, Okamoto T, Aoyama T, Nishijo K, Shibata KR (RefSeq): a curated non-redundant database of genomes, transcripts et al. (2005). Disruption of fibroblast growth factor signal and proteins. Nucleic Acids Res 35(Database issue): D61–D65. pathway inhibits the growth of synovial sarcomas: potential application Schuettengruber B, Chourrout D, Vervoort M, Leblanc B, Cavalli G. of signal inhibitors to molecular target therapy. Clin Cancer Res 11: (2007). Genome regulation by polycomb and trithorax proteins. Cell 2702–2712. 128: 735–745. Katoh M. (2008). Cancer genomics and genetics of FGFR2. Int J Sekiya I, Larson BL, Smith JR, Pochampally R, Cui JG, Prockop DJ. Oncol 33: 233–237. (2002). Expansion of human adult stem cells from bone marrow Katoh Y, Katoh M. (2009). FGFR2-related pathogenesis and stroma: conditions that maximize the yields of early progenitors and FGFR2-targeted therapeutics. Int J Mol Med 23: 307–311. evaluate their quality. Stem Cells 20: 530–541. Kawai A, Naito N, Yoshida A, Morimoto Y, Ouchida M, Shimizu K Villegas SN, Canham M, Brickman JM. (2010). FGF signaling as a et al. (2004). Establishment and characterization of a biphasic mediator of lineage transitions—evidence from embryonic stem cell synovial sarcoma cell line, SYO-1. Cancer Lett 204: 105–113. differentiation. J Cell Biochem 110: 10–20. Khan J, Bittner ML, Saal LH, Teichmann U, Azorsa DO, Gooden GC Watanabe Y, Kameoka S, Gopalakrishnan V, Aldape KD, Pan ZZ, et al. (1999). cDNA microarrays detect activation of a myogenic Lang FP et al. (2004). Conversion of myoblasts to physiologically transcription program by the PAX3-FKHR fusion oncogene. Proc active neuronal phenotype. Genes Dev 18: 889–900. Natl Acad Sci USA 96: 13264–13269. Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE Ladanyi M. (2001). Fusions of the SYT and SSX genes in synovial et al. (2008). Model-based analysis of ChIP-Seq (MACS). Genome sarcoma. Oncogene 20: 5755–5762. Biol 9: R137. Lobo NA, Shimono Y, Qian D, Clarke MF. (2007). The biology of Zhao C, Deng W, Gage FH. (2008). Mechanisms and functional cancer stem cells. Annu Rev Cell Dev Biol 23: 675–699. implications of . Cell 132: 645–660.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene