Synovial sarcoma: recent discoveries as a roadmap to new avenues for therapy

Functional studies of SS18-SSX

Initial functional studies of SS18-SSX used yeast two hybrids and GAL4 fusion constructs, where the relevant protein domains are fused to the DNA binding domain of GAL4. These studies showed that SS18 is a transcriptional co-activator and that C-terminal SSX domains mediate repression (12, 13). The fusion oncoprotein thus contains both activating and repressing domains, although neither partner has a DNA-binding domain. Several other putative binding partners were also identified as shown in Figure 1, but the mechanisms targeting the fusion protein to specific DNA refions were elusive until recently.

SS18 and SS18-SSX incorporate into the SWI/SNF complex

Thaete and colleagues showed association of both SS18 and SS18-SSX with the DNA-dependent ATPase BRM, the catalytic subunit of SWI/SNF chromatin remodeling complexes (14). Subsequently Kato et al showed that SS18 is a stable and integral component of SWI/SNF complexes, using co-immunoprecipitation and mass spectroscopy in nuclear extracts of HeLa cells (15).

Midelljans and colleagues extended these results by showing that the fusion oncoprotein is similarly incorporated into stable SWI/SNF complexes (16). All commonly observed subunits were recovered in reciprocal purifications between tandem affinity purification-tagged SS18-SSX1 and other subunits, indicating minimal perturbation of the core complex when the fusion oncogene is stably expressed in HEK293 cells. Kadoch and Crabtree observed high affinity binding of both SS18 and SS18-SSX to the core subunits of SWI/SNF, and immunodepletion of nuclear extracts showed undetectable levels outside of this association (17). In contrast to the results of Middlejans, these authors observed that expression of the fusion oncogene induced depletion of the BAF47 (SMARCB1) subunit from the SWI-SNF complex, in experiments using transiently expressed GFP-tagged SS18-SSX in a HEK293 background. The authors also reported SMARCB1 loss in synovial sarcoma cell lines, noting that siRNA knockdown of SS18-SSX restored SMARCB1 inclusion in SWI/SNF complexes.

Both studies suggest that SMARCB1's association with SWI/SNF may be more labile than for other subunits, and it is conceivable that SMARCB1 is displaced from SWI/SNF by aberrant protein interactions involving SS18-SSX. Alternately SMARCB1 may be stabilized in SWI/SNF complexes by unknown binding partners in different experimental conditions (e.g. transduced vs transfected HEK293 cells, or differences in the parent HEK293 cell lines, which have complex genomes and known clonal heterogeneity). HEK293 cells, while convenient for transfection and at least partially permissive for (short term) SS18-SSX expression, may not represent the most relevant cell line model as discussed in section 3, below. Resolution of these conflicting results will have important implications since SMARCB1 (aka SNF5, INI1) is a known tumor suppressor, with homozygous loss in 98% of rhabdoid tumors (18), 90% of epithelioid sarcomas and more than half of myoepithelial carcinomas (19). If SMARCB1 loss contributes to oncogenesis in synovial sarcoma, advances in the study of SWI/SNF directed therapies in rhabdoid tumors may have direct translational implications. Of potential importance, synthetic lethalities may be explored by analogy to rhabdoid tumors, where for example tumorigenesis was found to depend on functional BRG1 in the residual SWI/SNF complex (20). While caution is currently indicated in drawing simple parallels with rhabdoid tumors, these proteomic studies imply that the mechanisms of SS18-SSX mediated oncogenesis are intimately related to dysregulation of SWI/SNF chromatin remodeling.

SWI/SNF functions to reposition nucleosomes on genomic DNA, and can also promote nucleosome disassembly and histone exchange (reviewed in (21)). The canonical activity of SWI/SNF is to create nucleosome depleted regions at core promoters and regulatory regions, facilitating transcription factor access to DNA. SWI/SNF activity is broadly recruited across the genome to effect switches in chromatin state and has been associated with the induction of a wide range of transcriptional programs, including cell cycle control, stem cell maintenance, and differentiation. Importantly, recent exome sequencing studies have found mutations in SWI/SNF complex members in 20% of cases across a broad spectrum of tumor types, surveyed by Kadoch et al (22), who review the importance of this complex in tumor suppression. However the specific mechanisms of tumor suppression remain elusive, due to the diverse and tissue-specific roles of the complex in transcriptional regulation.

SMARCB1 and BRG1 inactivation lead to increased proliferation (23), at least in part due to impaired transcriptional activation of the CDKN2A/B tumor suppressors (24). This result has been related to the requirement of SWI/SNF activity for the eviction of repressive polycomb complexes from chromatin, and highlights the general principle of antagonism between SWI/SNF and polycomb activities. In this regard, the dominance of polycomb repression would have profound implications for the regulation of stemness and differentiation, and may prove to be a central aspect of oncogenesis in SWI/SNF mutated tumors. Other mechanisms of SWI/SNF oncogenesis have been elaborated, including mutations in ARID1A where it has been suggested that functional SWI/SNF is required to execute the p53 transcriptional response (25). Although not shown in synovial sarcoma, such an effect may explain the lack of p53 mutation in the majority of these lesions. SWI/SNF dysfunction also compromises DNA double strand break repair, and may confer sensitivity to DNA-damaging agents (26); this may contribute the sensitivity of synovial sarcom to radiation and anthracycline therapies.

SS18-SSX and epigenetic transcriptional repression

Whereas SS18, through its interactions with the SWI/SNF complex, might be expected to have a role in transcriptional activation, its fusion partner SSX associates with the Polycomb repressor complex which has opposing effects. An early observation was that SS18-SSX localizes at discrete nuclear foci within BMI-labeled Polycomb bodies (27). More recently ChIP-Seq results from HA-FLAG-tagged SS18-SSX, expressed in transfected C2C12 mouse myoblasts (28), correlated SS18-SSX binding with Polycomb-marked nucleosomes (trimethylated Histone H3K27) at a subset of genomic H3K27me3 sites.

In studies investigating possible target genes for SS18/SSX-mediated repression, expression profiling (29) showed the tumor suppressor EGR1 to be among the most consistently downregulated genes in synovial sarcoma. Chromatin immunoprecipitation studies revealed the presence of SS18-SSX at a cyclic AMP response element (CRE) in the EGR1 promoter (30). Furthermore, inhibitors of histone deacetylases (HDAC) – whose anticancer activity had been shown previously in synovial sarcoma xenografts (31) – were found to reverse the repression of EGR1, accompanied by loss of Polycomb residency at this locus.

Subsequent studies detailed proteomic and biochemical characterizations of SS18-SSX binding partners, and identified ATF2 as the transcription factor responsible for targeting the fusion oncoprotein to CRE sites (32). Moreover, these studies identified a robust interaction of the transcriptional corepressor TLE1 with the SSX portion of the fusion oncoprotein in human and mouse model synovial sarcoma cells, confirmed in surgically-excised patient tumor tissue. SS18-SSX serves as a bridge between ATF2 and TLE1, mediating repression of ATF2 targets including amongst others EGR1, ATF3 and CDKN2A. Both the SWI/SNF component BRM and the Polycomb component EZH2 were observed to co-elute with the SS18-SSX oncoprotein. Reciprocal immunoprecipitations showed co-purification of core PRC2 subunits EZH2, SUZ12, and EED with TLE1, SS18-SSX, and ATF2, indicating the presence of a stable repressive complex nucleated by the fusion oncoprotein. Chromatin immunoprecipitation and electrophoretic mobility shift assays confirmed the presence of this repressive complex at conserved CRE elements in ATF2 target gene promoters. Treatment with HDAC inhibitors disrupts this complex (32).

TLE family proteins are corepressors of Wnt target genes (33) and effectors of HES-mediated Notch repression (34), and are highly expressed in a number of embryonic progenitor fields where these pathways are known to regulate self-renewal and stemness. The identification of the corepressor TLE1 as an SS18-SSX interacting protein is also significant because expression profiling experiments have identified TLE1 among the most consistently highly expressed genes in primary synovial sarcoma specimens (29, 35). Incidentally, nuclear expression of TLE1 is readily detectable by immunohistochemistry on formalin-fixed paraffin-embedded tissues, and its diagnostic value for synovial sarcoma has been confirmed (36, 37), leading to its adoption into clinical use.

Chromosomal alterations in synovial sarcoma

Synovial sarcoma differs from most common forms of cancer by virtue of having much lower genetic complexity. Almost half of primary tumors have no chromosomal aberrations other than t(X;18), and the remainder display only small numbers of changes (median 2 – 4) (44). Notably, tumors with a balanced t(X;18) and no other genomic changes can show a completely “flat” genomic copy number profile. Additional genomic copy number changes, which are more common in adult (>18 years) than pediatric patients, are strongly associated with metastatic spread and poor outcomes. Not surprisingly, metastatic and recurrent tumors display increased chromosomal complexity (45).

Secondary mutations in synovial sarcoma

In keeping with the observed chromosomal stability, the most commonly mutated gene in human cancer, TP53, is rarely mutated in synovial sarcoma, occurring in 11 of 92 cases across three published studies (46-48); in these studies copy number gains in the p53 suppressing oncogene MDM2 were somewhat more frequent. Overall, wild-type p53 appears to be retained in most synovial sarcomas, although its function may be impaired through upstream regulatory events, as in several well-described pro-survival interactions within the AKT/PTEN pathway.

Otherwise, targeted sequencing approaches have highlighted mutations in PTEN, CTNNB1 and APC in 8-14% of cases (49-51), providing some support for oncogenic activation of the Akt/mTOR and Wnt signaling pathways in this sarcoma (see below).

As of this writing, only one whole exome sequencing study had been published on synovial sarcoma, on tumors from 7 patients (9). This work identified an average of 8 somatic mutations per tumor exome, and no genomic losses in the majority of cases. Solitary mutations in cancer pathway genes were identified for TP53, SETD2, and FBXW7.

On the whole, these results suggest an encouraging setting for targeted therapy of synovial sarcoma – if treated early in its natural history, when the majority of tumors do not exhibit genomic instability or extensive genomic rearrangements. Cell death pathways and growth controls may be largely intact and responsive to effective targeting of SS18-SSX. Compared to most common cancer types there may be fewer escape mechanisms for emergence of therapy-resistant subclones in synovial sarcoma.

Gene and protein expression in synovial sarcoma

A major theme from gene expression profiling studies of synovial sarcoma relates to high expression of mediators involved in the patterning systems of early embryogenesis, including Wnt (LEF1, AXIN2,WIF1, WNT5A, FZD10), Notch (HES1, JAG1, JAG2, NRARP), Hedgehog (PTCH1, GLI1, GLI2), FGF (FGFR2, FGFR3, FGF18, FGF9), and BMP pathways (BMP7, BMP5, BMPR2, SOSTDC1). These results and expression of other markers of embryonic primordia (SALL2, TLE1, SIX1, SIX4, DLX2) support the idea that synovial sarcoma cells have a stem-like or early progenitor phenotype. Of note, primary synovial sarcomas show a remarkable correspondence with the expression profiles of the conditional Myf5-Cre mouse model (11) and SS18-SSX transduced embryonic stem cells (41).

Synovial sarcomas commonly express relatively high levels of mRNA for Cancer-Testis Antigens, a group of potentially immunogenic proteins expressed in many tumor types, but not in normal adult tissues outside the germline: PRAME, CTAG1A (encoding NY-ESO-1), and MAGEE1 are prominent examples. The possible relevance of this aberrant antigen expression in terms of response in recently developed immunotherapy approaches remains to be determined.

High expression of neural (NPTX2, NEFH, NTNG2) and chondrocyte (COL2A1, COL9A3, SOX9, TRPS1) lineage markers has been interpreted as consistent with differentiation from neural crest progenitors. High mRNA expression of SOX2, an important determinant of neural progenitors and embryonic stem cells, is common in synovial sarcoma and has been shown critical for growth of some cell lines in vitro (15, 39). However its expression is absent in other cell lines and is not a universal feature of primary tumor specimens.

Other notable features of the synovial sarcoma profile include differential high expression of BCL2, as well as of the receptor tyrosine kinases PDGFRA, EGFR, ERBB2. The latter group may be important for the stimulation of PI3K-Akt signaling, which has been reported as critical to synovial sarcoma viability in several studies (52, 53), and is discussed further below.

Many of the above findings have been confirmed at the protein level in patient specimens, including FZD10 (54), HES1 (55), NY-ESO-1 (56), BMPR2 (57), SALL2 and EGFR (58), BCL2 (59), SOX9 (60), FGFs and their receptors (61), PDGFRA (52), and Wnt pathway mediators (discussed further in the next section).

Wnt/beta-Catenin signaling pathway

There has been longstanding evidence for Wnt activation in a subset of synovial sarcomas. Sanger sequencing studies found canonical activating mutations in CTNNB1, at frequencies of 4% (N=24) (62), 8%(N=49) (63) and 12% (N=16) (50), and APC mutations in 8% (N=49) of cases (51). At the protein level, Wnt pathway activation, as evidenced by nuclear accumulation of β-catenin, has been studied using immunohistochemistry. By this method, nuclear accumulation of β-catenin is detected in 30% to 60% of synovial sarcomas (64, 65) primarily in monophasic cases or in the spindle cell component of biphasic cases. The SYO-1 synovial sarcoma cell line harbors a codon 34 mutation in CTNNB1 (G34L) with concomitant protein accumulation and shows reduced proliferation and impairment in standard invasion and migration assays when transfected with an inhibitory dominant negative LEF1 construct (Saito T, Motoi T, Ladanyi M, unpublished data).

Recently, two independent studies have provided important functional evidence for a critical role of this signaling pathway in synovial sarcoma. Barham and colleagues (66) showed that genetic loss of β-catenin blocks tumor formation in the Myf5-CRE SYT-SSX2 transgenic model described above. To confirm this observation in the human setting, they showed that pharmacologic activation of CSNK1A by pyrvinium, known to inhibit β-catenin signaling, or RNAi knockdown of LRP6 could both impair growth of human synovial sarcoma cell lines. These investigators also showed that SYT-SSX can induce nuclear β-catenin accumulation, apparently by inducing autocrine signaling through its aberrant transcriptional effects. These data may explain the fact that the proportion of human synovial sarcoma tumors with nuclear β-catenin accumulation is several fold higher than the rate of CTNNB1 or APC mutations.

In a separate recent study, Trautmann and colleagues found that introduction of SYT-SSX into HEK293 induced activation of Wnt/β-catenin signaling (67). Conversely, small molecule TCF/β-catenin complex inhibitors reduced cell viability in vitro in five human synovial sarcoma cell lines and reduced in vivo growth of xenografted SYO-1 synovial sarcoma cells.

Together, these studies nominate the Wnt/β-catenin signaling pathway as an important new candidate therapeutic target in synovial sarcoma, with putative mechanisms as outlined in Figure 2. As new agents efficiently targeting this pathway are in clinical development, these studies provide a critical pre-clinical rationale for their evaluation in clinical trials (68,69). Beyond the direct targeting of Wnt/β-catenin signaling pathway components by small molecules, targeting of key interacting proteins also offers therapeutic opportunities. For instance, β-catenin forms a complex with the transcriptional modulator YAP1 and the YES1 tyrosine kinase, which may explain why dasatinib, which blocks the latter, also inhibits Wnt/β-catenin-dependent proliferation (69). Interestingly, a recent study found that dasatinib caused apoptosis and inhibition of cellular proliferation in synovial sarcoma cells; some data support that this effect is mediated by inhibition of activated SRC (70), but it is tempting to speculate that the potency of the drug may also have been due to concurrent effects on Wnt/β-catenin-dependent targets.

In light of the recent finding that SYT-SSX may, under certain circumstances, displace SMARCB1 from SWI/SNF chromatin remodeling complexes (17), it is also notable that loss of SMARCB1 results in activation of Wnt/β-catenin signaling in mouse and cell culture models, reflecting a role for the SWI/SNF complex in regulating this pathway (71).

AKT/MTOR signaling pathway

Various lines of evidence also support an important role for the AKT/MTOR pathway in synovial sarcoma, including genetic, protein-level, pre-clinical, and clinical trial data. At the genetic level, PTEN mutations are only found in a minority of cases (49, 62, 72) and canonical PIK3CA mutations are even more uncommon (72). Nonetheless, several groups have demonstrated frequent activation of AKT in synovial sarcoma, and its dependency on upstream receptor tyrosine kinases (RTKs) (53, 73, 74). As noted above, apoptosis induction in synovial sarcoma cell lines by HDAC inhibitors is at least partially dependent on derepression of EGR1 and subsequent induction of PTEN (30), a phenomenon which may help to identify strategic combinations involving these promising therapeutic agents.

Notably, fundamental interactions of PI3K/AKT/MTOR and RAS/MEK/ERK pathways have been identified downstream of RTK signaling, and multiple levels of feedback regulation have been identified within and between these pathways, summarized in figure 3. These interactions underlie intrinsic resistance to PI3K/AKT/mTORC1 targeted monotherapies in synovial sarcoma, for example involving feedback activation of AKT following MTOR inhibition, which can be mediated by both IFG1R and PDGFRA (43, 52). These data support the combination of mTOR inhibitors with inhibitors of particular RTKs that potentiate feedback AKT activation in a given tumor. Strategies which may intercept these phenomena at critical downstream nodes are also under investigation, including the combination of PI3K and MEK inhibitors, and the use of pan-mTOR kinase inhibitors targeting both mTORC1 and mTORC2 (which are not associated with reactivation of AKT (75)).

Antiapoptotic pathways

Histologically, synovial sarcomas typically display few mitotic figures, while also displaying few apoptotic bodies. The tumor's propensity in many cases for slow growth, late recurrences and chemotherapy resistance is consistent with an antiapoptotic phenotype. A striking feature of synovial sarcoma is its consistently very high expression of BCL2 (59, 76, 77), but no evidence has been found that the BCL2 gene is a direct target for transcriptional upregulation by SS18-SSX; therefore, its high level of expression may be a secondary oncogenic effect, or may reflect a prerequisite cellular background to be permissive for SS18-SSX. In contrast, the BCL2 family members BCL2A1 and MCL1 are direct targets for SS18-SSX-mediated repression. Resistance to new BH3 mimetic Bcl2 inhibitors in lymphomas is typically induced by activation of MCL1, but this escape mechanism is not available to synovial sarcomas, which appear to be highly sensitive to such drugs in human cells and an SS18-SSX2 mouse tumor model (78). These findings suggest that BCL2 inhibitors may be worth evaluating in synovial sarcoma, perhaps in combination with apoptosis-inducing cytotoxic chemotherapy, but as yet no such clinical trials have been opened.