Imprinted CDKN1C Is a Tumor Suppressor in Rhabdoid Tumor and Activated by Restoration of SMARCB1 and Histone Deacetylase Inhibitors

SMARCB1 is deleted in rhabdoid tumor, an aggressive paediatric malignancy affecting the kidney and CNS. We hypothesized that the oncogenic pathway in rhabdoid tumors involved epigenetic silencing of key cell cycle regulators as a consequence of altered chromatin-remodelling, attributable to loss of SMARCB1, and that this hypothesis if proven could provide a biological rationale for testing epigenetic therapies in this disease. We used an inducible expression system to show that the imprinted cell cycle inhibitor CDKN1C is a downstream target for SMARCB1 and is transcriptionally activated by increased histone H3 and H4 acetylation at the promoter. We also show that CDKN1C expression induces cell cycle arrest, CDKN1C knockdown with siRNA is associated with increased proliferation, and is able to compete against the anti-proliferative effect of restored SMARCB1 expression. The histone deacetylase inhibitor (HDACi), Romidepsin, specifically restored CDKN1C expression in rhabdoid tumor cells through promoter histone H3 and H4 acetylation, recapitulating the effect of SMARCB1 on CDKNIC allelic expression, and induced cell cycle arrest in G401 and STM91-01 rhabdoid tumor cell lines. CDKN1C expression was also shown to be generally absent in clinical specimens of rhabdoid tumor, however CDKN1A and CDKN1B expression persisted. Our observations suggest that maintenance of CDKN1C expression plays a critical role in preventing rhabdoid tumor growth. Significantly, we report for the first time, parallels between the molecular pathways of SMARCB1 restoration and Romidepsin treatment, and demonstrate a biological basis for the further exploration of histone deacetylase inhibitors as relevant therapeutic reagents in the treatment of rhabdoid tumor.


Introduction
Rhabdoid tumor (RT) is an aggressive although rare tumor of infancy and early childhood resistant to conventional chemotherapies and radiotherapy. The majority of afflicted children succumb to their disease within several months of diagnosis. Rhabdoid tumors mainly arise in the kidney where they are known as rhabdoid tumours and in the central nervous system where they are referred to as Atypical Teratoid Rhabdoid Tumor (AT/RT). They are characterized genetically by deletion or allelic loss of chromosome 22q, and associated inactivating mutations or deletion of the tumor suppressor gene SMARCB1 (OMIM 601607) [1,2,3,4,5,6].
Homozygous deletion of smarcb1 in mice is embryonic lethal, however, heterozygous mice develop tumors that are histologically similar to their human counterparts [7,8,9]. Tumor formation in mice is accelerated by coincident p53 mutation [10] and it has been recently proposed that tumor formation associated with loss of SMARCB1 may arise due to permissive defects in cellular DNA damage response pathways [11]. Although SMARCB1 deletion is predominantly associated with RT, recently SMARCB1 inactivation and mutation has been described in epitheloid sarcoma and familial schwannomatosis [12,13].
One suggested mechanism by which loss of SMARCB1 facilitates oncogenesis is through defective cell cycle regulation. Reexpression of SMARCB1 in human rhabdoid tumor cell lines causes G0/G1 arrest showing that restoration of SMARCB1 expression is sufficient to suppress proliferation [14,15]. This is associated with activation of p16INK4a and CDKN1A and down regulation of E2F target genes including cyclin A, E2F1 and CDC6. SMARCB1 arrest is critically dependent on the presence of functional pRB [16]. SMARCB1 is not able to arrest cells lacking functional pRB and arrest can also be reversed by disruption of pRB repressor complexes through restoration of cyclin D1 and cyclin E expression. Further, constituitively active pRB1 can induce arrest in RT cell lines lacking SMARCB1. SMARCB1 is part of an ATP dependent multiprotein SWI/ SNF chromatin remodelling complex [17]. It associates with ATPase subunits Brg1 (for Brahma-related gene 1, or SNF2b and Brm (for Brahma or SNF2a). In contrast to SMARCB1, Brg1 and Brm are required for cell cycle arrest mediated by pRB. Versteege et al [16] hypothesize that Brg1 and Brm are necessary for the chromatin remodelling associated with pRB repression of E2F and that SMARCB1 has a promoting but not a primary role in this remodelling. Deletion of Brg1 and Brm occurs in many cancer cell lines and is associated with gene specific changes in promoter methylation at CD44 and E-Cadherin leading to hyper-methylation and gene silencing [18]. Brg1 and Brm associate directly with the promoters of these genes and a more widespread role in epigenetic regulation of gene expression during tumor progression has been proposed. The direct role of SMARCB1 in chromatin remodelling has not been extensively explored. Pan et al [19] have shown that SMARCB1 represses the c-Fos promoter via histone deacetylation in 293T cells and that this occurs via direct interactions between HDAC4 and SMARCB1, and Zhang et al [20] showed that interactions between HDAC1 and hSNF5/INI1 (SMARCB1) were required to repress Cyclin D.
We hypothesized that the oncogenic pathway induced by SMARCB1 inactivation in RT may involve epigenetic silencing of key cell cycle target genes. This premise, if established, may reveal opportunities for treatment of RT with epigenetic therapies that restore the expression of key growth-regulating genes. In this work we demonstrate that the imprinted cell cycle inhibitor CDKN1C (OMIM 600856) is a downstream target for SMARCB1 epigenetic regulation. SMARCB1 consistently activated CDKN1C expression via histone H3 and H4 acetylation at the CDKN1C promoter and the histone deacetylase inhibitor (HDACi), romidepsin, restored imprinted CDKN1C expression in RT cells through promoter histone H3 and H4 acetylation. Significantly, CDKN1C expression was absent or negligible in clinical specimens, enforced expression of CDKN1C in G401 RT cells induced cell cycle arrest and knockdown of endogenous CDKN1C increased proliferation in G401 RT cells as well as attenuating the effects of SMARCB1 re-expression on cell proliferation. Our findings show that CDKN1C silencing is common in RT, suggest that CDKN1C is an important regulator of RT cell growth, and that RT cell proliferation is mediated in part via suppression of CDKN1C expression attributable to deletion of SMARCB1.

CDKN1C is up-regulated in RT cells with reconstituted expression of SMARCB1
CDKN1C (p57 Kip2 ) is commonly epigenetically silenced in cancer and is the only imprinted member of the CIP KIP family of cell cycle inhibitors. We hypothesized therefore that CDKN1C (p57 Kip2 ) may be a potential downstream target for SMARCB1. To examine this we expressed SMARCB1 under the control of an inducible promoter in two renal RT cell lines, G401 and STM91-01. G401 is an established renal RT cell line homozygously deleted for SMARCB1 [1]. STM91-01 is derived from a lung metastasis with 22q loss of heterozygosity and deletion of SMARCB1 exons 1 to 5 on the retained allele. Deletion of SMARCB1 protein in all RT cell lines used in the study was confirmed by Western blotting and further evidence for SMARCB1 deletion or rearrangement was obtained by southern blotting (results not shown).
G401 clones expressing SMARCB1 under the control of a tamoxifen inducible promoter system (pcDNA3 UAS SMARCB1 and pEF puro/GEV) were selected in geneticin and puromycin. Two G401 clones, F13 and F22, showing inducible SMARCB1 expression, were selected for further investigation. A pool of STM91-01 cells (STMpc), transduced with a lentiviral vector system (pF 56UAS Sv40 puro/ini1 and pF GEV16) was also examined. SMARCB1 protein was readily detectable 24 hours after exposure to 1 uM 4HT in all cultures, ( Figure 1A). Cell cycle analysis was performed on cultures of F22 cells 72 hours after induction of SMARCB1 protein expression. In SMARCB1 expressing cells 41.4+/25% of cells were arrested in G 0 compared with 7.0+/21.4% in un-induced cells (Supplementary Figure  S1A), consistent with observations made previously that SMARCB1 can induce cell cycle arrest [14][15][16]. In contrast, STMpc cells induced to express SMARCB1 showed little change in the cell cycle (Supplementary Figure S1A). One explanation for this is that the parental line, STM91-01, is derived from a RT lung metastasis and may have acquired additional oncogenic mutations and hence greater resistance to growth inhibitory signals. Reincke et al [21] also found this cell line difficult to transduce with recombinant adenovirus containing SMARCB1 and reportedly found inconsistent cell cycle changes following infection. CDKN1C mRNA was elevated in all lines (F22, F13, STMpc) following induction of SMARCB1 expression with 1 uM 4HT whereas other CDK family members including CDKN1A (p21/ WAF1) and CDKN1B (p27) did not show uniform responses following 4HT treatment ( Figure 1B). Quantitative analysis of CDKN1C expression by real-time PCR in F22 cells showed fourfold increases in CDKN1C expression 24 hours following SMARCB1 induction, and STMpc showed six-fold increases in CDKN1C expression ( Figure 1C). 1 uM 4HT alone had a negligible effect on CDKN1C expression in parental G401 and STM91-01 cultures demonstrating that the observed effect on CDKN1C expression was directly attributable to the induction of SMARCB1 protein (results not shown). Endogenous levels of CDKN1C protein were low in un-induced F22 cells however following SMARCB1 induction, were increased to levels consistent with the proportional increase in CDKN1C transcript levels ( Figure 1D).

Enforced expression of CDKN1C arrests proliferation in G401 cells and CDKN1C knock-down increases proliferation and attenuates SMARCB1-induced proliferation arrest
To examine the direct effect of induced CDKN1C expression in rhabdoid tumor cells, G401 clones over-expressing full-length human CDKN1C under the control of the tamoxifen inducible promoter (pcDNA3 UAS CDKN1C and pEFpuro/GEV), were selected in geneticin and puromycin. Induction of CDKN1C protein in G401 cells induced cell cycle arrest (Figure 2A and Supplementary Figure S1B). Endogenous CDKN1C protein was not detectable by Western blotting at the short exposure timesetting used to detect over-expression of CDKN1C. In clone G401 C1, expressing the highest level of CDKN1C protein, 24.7+/ 21.67% of cells were arrested in G 0 , compared with 2.3+/ 20.14% in un-induced cultures (Supplementary Figure S1B). To further ascertain the functional significance of CDKN1C induction in RT cells, a CDKN1C siRNA pool was used to knock-down the low level of endogenous CDKN1C in F22 cells, and to knockdown CDKN1C mRNA induced following SMARCB1 reexpression. CDKN1C siRNA transfection in F22 cells eliminated CDKN1C expression and led to increased proliferation, as measured by a greater positive change in MTS absorbance over the first 48 hours following transfection, as compared with F22 cells treated with a non-targeting control siRNA ( Figure 2B). CDKN1C siRNA also attenuated the effect of SMARCB1 induction on restricting cell proliferation. Cells in which SMARCB1 expression was induced with 1 uM 4HT by 24 hours and transfected with CDKN1C siRNA, exhibited a significant and reproducible positive increase in MTS absorbance between 24 and 48 hours, reflecting an increase in the number of proliferating cells in the culture when compared with cells transfected with nontargeting siRNA +4HT. This correlated with reduced levels of CDKN1C mRNA after 48 hours in the +4HT CDKN1C siRNAtreated cultures. Between 48 and 72 hours, cells in the 4HT and non-targeting siRNA, CDKN1C siRNA, and CDKN1C siRNA and 4HT cultures, showed a reduction in the number of viable cells present. In the CDKN1C siRNA culture this likely reflects a reduction in the rate of cell proliferation as the cells approach senescence. In the cultures treated with non-targeting siRNA and  4HT, and CDKN1CsiRNA and 4HT, the decrease in cell viability probably reflects the long-term effect of sustained SMARCB1 expression on cell viability. This may suggest that SMARCB1 modifies the expression of other genes in addition to CDKN1C that affect cell proliferation and viability. Nevertheless these results clearly show that CDKN1C is an important modulator of G401 growth as small changes in CDKN1C expression affect proliferation, and furthermore provide evidence for parallels between the actions of SMARCB1 and CDKN1C in regulating RT proliferation.

SMARCB1 expression causes increased histone acetylation at the CDKN1C promoter
To investigate the mechanism associated with the SMARCB1induced upregulation of CDKN1C in RT cells, a region within the 59CpG island spanning the CDKN1C transcription start site, from 241 to +114 was analysed for changes in methylation, pre and post SMARCB1 induction. Methylation of this region has previously been shown to correlate with silencing CDKN1C expression in cancer cell lines [22]. Bisulphite sequencing on 10 clones pre and post induction with 4HT revealed that the CDKN1C promoter was largely unmethylated in un-induced F22 and STMpc cells and remained unmethylated following SMARCB1 induction, (Supplementary Figure S2). Furthermore, treatment of parental G401 and STM91-01 cells with 2 uM 5aza2dC had no significant effect on CDKN1C expression consistent with a lack of promoter methylation.
We then performed CDKN1C promoter chromatin immunoprecipitation (ChIP) with antibodies to acetylated histones H3 and H4, 48 hours post SMARCB1 induction. Quantitative PCR was performed on recovered DNA spanning the CDKN1C transcription start site from 272 to +89 and normalized to GAPDH [22]. Three independent ChIP analyses each, on both F22 and STMpc rhabdoid tumor cell lines, showed that the CDKN1C promoter is acetylated in un-induced F22 and STMpc cells, however histone H3 acetylation increased 2 fold in both cell lines following SMARCB1 induction with 4HT. A less marked increase in histone H4 acetylation was observed in both cell lines cells ( Figure 3A and B). Histone acetylation was also examined at the CDKN1A and CDKN1B promoters in F22 cells. In contrast to acetylation of the CDKN1C promoter, histone H3 acetylation was either unchanged (CDKN1A promoter) or slightly decreased (CDKN1B promoter) following SMARCB1 induction, and histone H4 acetylation was not significantly changed at either promoter ( Figure 3C and D). These results show that SMARCB1 specifically acetylates histones H3 and H4 at the CDKN1C promoter either directly, or indirectly through the recruitment of histone acetyltransferases (HATs) or through inhibition of histone deacetylases (HDACs).

CDKN1C is upregulated in RT cells by HDACi and promoter histone acetylation is increased
Romidepsin is a potent broad-spectrum inhibitor of class 1 and class II histone decetylases (reviewed in [23]). We reasoned that HDACi might mimic the effect of SMARCB1 on CDKN1C. CDKN1C expression was significantly increased in all RT cell lines including G401, STM91-01, SJSC and BT16 lines following treatment with 10 nM Romidepsin for 24 hours ( Figure 4A to D). Levels of CDKN1C mRNA induction quantified by real-time PCR varied from six-fold in BT16 cells to twelve-fold in STM91-01 cells. After 48 hours culture with 10 nM Romidepsin, cells showed signs of apoptosis. This coincided with a decrease in the levels of the GUSB house-keeping gene therefore the levels of CDKN1C induction at 24 hours most reliably reflect the true effect of Romidepsin on CDKN1C expression. The levels of CDKN1C induction at 24 hours in G401 and STM91-01 cells treated with Romidepsin slightly exceeded those induced with SMARCB1. Cell cycle analysis was performed on G401 and STM91-01 cells treated with 1 nM Romidepsin after 72 hours. In both cell lines, this treatment led to a significant increase in the percentage of cells in G 0 (Supplementary Table S1). This was more pronounced in G401 cells compared with STM91-01 cells. In control G401 cells treated with 0.01% DMSO, 4.1+/20.35% of cells were in G 0 compared with 19.3+/21.6% in cultures CDKN1C signal was normalized to GAPDH. Acetylation at histone H3 and H4 was increased at the CDKN1C promoter in both F22 and STMpc cells. (C) Acetylation of histone H3 and histone H4 at the CDKN1A promoter. (D) Acetylation of histone H3 and histone H4 at the CDKN1B promoter. CDKN1A and CDKN1B signal was normalized to GAPDH. Error bars represent the standard error of the mean and were derived from three independent experiments. doi:10.1371/journal.pone.0004482.g003 treated with 1 nM Romidepsin. In STM91-01, 9.3+/21.9% of cells treated with 0.01% DMSO were in G 0 compared with 16.7+/22.0% treated with 1 nM Romidepsin. CDKN1C promoter ChIP analysis was performed in G401 and STM91-01 post treatment with 10 nM Romidepsin for 48 hours. Both histone H3 and H4 acetylation was significantly increased at the CDKN1C promoter in treated cells, consistent with the increase in CDKN1C expression, ( Figure 5A). Increases in promoter histone acetylation were also observed for CDKN1A however, these were not as pronounced as the histone acetylation observed for CDKN1C ( Figure 5B). In contrast, CDKN1B did not show significant increases in promoter histone acetylation following Romidepsin treatment ( Figure 5C).

CDKN1C allelic expression is modulated in RT cells with reconstituted expression of SMARCB1 and with HDACi
To examine CDKN1C imprinting, we took advantage of deletion polymorphisms within CDKN1C exon 1 in G401 and in STM91-01 cells, [24,25]. CDKN1C alleles differing in size by 24 bp and 12 bp were identified in G401 and STM91-01 cells respectively (Supplementary Figure S3). Both cell lines exhibited biallelic expression, suggesting loss of imprinting. This is in contrast to CDKN1C imprinting in the normal kidney where the gene shows monoallelic expression [26]. The effect of SMARCB1 expression on CDKN1C imprinting in G401 clones F13 and F22 and in the STM91-01 pooled cells, STMpc, was examined. Induction of SMARCB1 protein in F22 cells led to the restoration of monoallelic CDKN1C expression with expression predominantly from the short allele 24 hours post induction, and a trend towards expression from the short CDKN1C allele in the weaker SMARCB1 expressor, F13 ( Figure 6A). 4HT alone had no sustained effect on the pattern of allelic expression in parental G401 cells, demonstrating specificity for the SMARCB1 effect. A variable dominant expression from the long allele was sometimes observed in different wild-type and recombinant un-induced G401 cells reflecting the unstable nature of CDKN1C imprinting in these cultures. The short allele was however always reproducibly expressed in replicate cultures following SMARCB1 induction in the G401 background suggesting that SMARCB1 expression leads to stabilization of the CDKN1C imprint and promotes monoallelic expression. In contrast, induction of SMARCB1 expression in STMpc cells did not alter biallelic expression of CDKN1C ( Figure 6A). We then examined allelic expression in response to treatment with 0.3 uM TSA and 10 nM Romidepsin and observed a shift to the short allele mimicking the effect of SMARCB1 induction. This was most marked with 10 nM Romidepsin ( Figure 6B). The parental G401 line was also sensitive to TSA and Romidepsin, and the direction of monoallelic expression following treatment was identical to that observed in F22 and F13 cells ( Figure 6B). Consistent with the response to SMARCB1 induction, TSA and Romidepsin had no effect on the pattern of allelic expression in STM91-01 cells ( Figure 6B). Taken together these results show that SMARCB1 over-expression, and HDACi drugs, have identical affects on the allelic expression of CDKN1C.
Imprinting at CDKN1C is typically, although not exclusively, regulated via maintenance of allele specific patterns of methylation at the adjacent 11p15.5 Imprinting Centre 2 (IC2) involving antisense transcription of the non-coding RNA, LIT1 [27,28]. Mosaic loss of methylation at IC2 may lead to low levels of biallelic CDKN1C expression. However, in both G401 and STM91-01 cells, normal methylation at IC2 was observed suggesting that an alternative mechanism is associated with loss of CDKN1C imprinting in RT cell lines ( Supplementary Figure S4). Other RT cell lines including SJSC and BT16 (not informative for CDKN1C imprint analysis) also maintained normal methylation at IC2 (results not shown).

CDKN1C expression is absent in rhabdoid tumor specimens
The relevance of impaired CDKN1C expression in RT cell lines to the aetiology of RT was further ascertained by examination of clinical rhabdoid tumor specimens. Tumors classified as rhabdoid or ATRT on the basis of identified SMARCB1 mutations, absence of SMARCB1 protein expression or histological criteria, were examined for CDKN1C mRNA expression by RT-PCR and for CDKN1C protein by immunohistochemistry. 6/6 tumors with viable mRNA showed negligible or absent CDKN1C expression after 45 cycles of RT-PCR, although CDKN1A and CDKN1B expression persisted in these tumors, (Supplementary Figure S5A). All tumors examined by immunohistochemistry (n = 12) lacked nuclear staining for CDKN1C and showed negligible cytoplasmic staining (Table 1 and Supple-mentary Figure 5B). A subset of RT show loss of CDKN1C protein even in the presence of detectable wild-type SMARCB1 suggesting that additional genetic factors may contribute to CDKN1C silencing in rhabdoid and related tumors. Four tumor specimens, with confirmed SMARCB1 mutations, and lacking CDKN1C expression, exhibited methylation indices at IC2 within the normal range (results not shown) suggesting that epigenetic change at IC2 is not responsible for CDKN1C silencing in primary rhabdoid tumors, consistent with observations in RT cell lines. There was insufficient DNA available from SMARCB1 expressing tumors for IC2 methylation analysis by southern blotting, and it is unclear whether imprinting defects at IC2 contribute to the absence of CDKN1C expression within this sub-set.

Discussion
In this study we identify the imprinted cell cycle regulator, CDKN1C, as a new and important SMARCB1 target in RT and furthermore demonstrate that SMARCB1 has the capacity to restore the expression of genes that are epigenetically silenced, consistent with its proposed role in chromatin remodelling. We show evidence that the mechanism by which CDKN1C expression is induced by SMARCB1 is via increased promoter histone H3 and H4 acetylation and that the HDACi, Romidepsin, activates CDKN1C expression in a related manner, leading to cell cycle arrest in RT cell lines. The parallel effects of SMARCB1 and HDACi, in association with an absence of CDKN1C expression in clinical specimens of rhabdoid tumor, and the demonstrated sensitivity of RT cell lines to small changes in CDKN1C expression mediated via siRNA, suggest that loss of CDKN1C is integral to the aetiology of RT, and furthermore that RT may be responsive to specific HDACi that increase endogenous levels of CDKN1C. This study reveals new insight into the molecular aetiology of malignant RT.
Cell cycle regulators have been previously studied in RT cell lines. SMARCB1 is a transcriptional co-activator of p16 INK4a and is required to bring the BRG1 chromatin remodelling complex to the p16 INK4a promoter. Activation of p16 INK4a inhibits the Cyclin D-CDK4 complex with the result that pRB is retained in its antiproliferative hypo-phosphorylated state and cells arrest in G1 [15,16,29]. pRB is the critical endpoint for G1 arrest in RT. The CIP KIP family member CDKN1A is a critical effector of signal transduction pathways and has also previously been implicated as a downstream target for SMARCB1 in RT cell lines [14]. CIP KIP inhibitors bind and inhibit the kinase activity of CDK-cyclin complexes including Cyclin A-CDK2 and Cyclin E-CDK2 through a common shared N-terminal domain. CDKN1A is upregulated following re-expression of SMARCB1, and expression is coincident with G1 arrest in these cells [14]. Cyclin E can overcome SMARCB1 induced proliferation arrest in RT cell lines also implying involvement of CIP KIP inhibitors in rhabdoid tumor, however in other studies CDKN1A and CDKN1B expression were unchanged by SMARCB1 expression [16]. We identified minor transcriptional upregulation of CDKN1A or CDKN1B in two different RT cell backgrounds following transfection with SMARCB1 however, expression of these genes persisted in clinical specimens of ATRT with inactivated SMARCB1 arguing against a major role in rhabdoid tumor aetiology.
CDKN1C is a third member of the CIP KIP family. CDKN1C is regulated by p73 during mitotic exit and re-entry into G1 in T98G glioma cells and in myogenic differentiation, CDKN1C expression is induced by p73 and p63 to maintain pRB in an active hypophophorylated state [30]. CDKN1C is not transactivated by p53 [31]. Our results therefore suggest that SMARCB1 could be part of a pathway involving p73 or alternatively that SMARCB1 independently regulates CDKN1C. Due to a lack of suitable SMARCB1 antibodies for immunoprecipitation, we were unable to perform ChIP experiments to show whether SMARCB1 transactivated CDKN1C by direct promoter binding.
CDKN1C is expressed in human fetal brain and has been implicated in regulating the migration of neurons entering the neural plate and in the later stages of neurogenesis in driving the differentiation of oligodendrocyte precursor cells [32,33]. CDKN1C is also not expressed in human astrocytomas however re-expression leads to a G1 block associated with hypophosphorylation of pRB, consistent with a tumor suppressor role [34]. Significantly and perhaps not surprisingly we found that CDKN1C expression was absent from the clinical tumor specimens we examined, of which the majority were CNS AT/ RT, and our evidence from RT cell lines also supports an important role for CDKN1C in RT growth suppression.
Silencing of CDKN1C in cancer is usually epigenetic and associated with promoter methylation or histone deacetylation [22]. In gastric tumors both mechanisms silence CDKN1C and expression can be restored by demethylating reagents in cases where the promoter is hypermethylated or by HDACi in cases where promoter histones are deacetylated [35]. Growth suppression in CDKN1C-transduced leukaemia cell lines is dependent on the extent of endogenous CDKN1C promoter methylation within the recipient cell background [36]. Furthermore aberrant methylation of the cell cycle control pathway including p73, p15 and CDKN1C defines a subset of ALL patients with poor prognosis [37]. Evidence also exists for CDKN1C silencing in lung and breast cancer by promoter methylation [38]. Our results, showing increases in histone acetylation at the CDKN1C promoter in RT cell lines, following treatment of the cells with Romidepsin and following SMARCB1 induction, suggest that CDKN1C is epigenetically silenced in RT by histone deacetylation at the promoter, and that SMARCB1 is complexed with histone acetyl transferases or that it may have instrinsic acetylation properties. The promoters of CDKN1A and CDKN1B were not significantly affected by SMARCB1 induction and these genes exhibited minor changes in expression. In contrast, CDKN1C expression was markedly increased and this coincided with a specific increase in histone H3 acetylation and less marked increase in histone H4 acetylation. The predominant acetylation at histone H3 is consistent with studies showing that the transcriptionally active murine cdkn1c promoter is associated with acetylated histone H3 [39].
CDKN1C is a paternally imprinted gene and in most normal tissues is expressed from the maternal allele although frequently leaky expression also occurs from the paternal allele. SMARCB1 over-expression and HDACi did not universally induce monoallelic transcription of CDKN1C in all cell lines suggesting that SMARCB1 cannot be involved in establishing imprinting at CDKN1C, but that it is able to recognize and induce transcription from appropriately imprinted alleles. Significantly LIT1 methylation was normal in RT cell lines and in RT with loss of SMARCB1 function, demonstrating that loss of CDKN1C imprinting must involve mechanisms independent of IC2 on 11p15.5.
Cdkn1c has been recently shown to be a critical regulator of embryonic growth and is the only CIP KIP protein that is imprinted and essential for normal development [40]. 50% of patients with the overgrowth disorder Beckwith Wiedemann syndrome commonly have epigenetic silencing of CDKN1C caused by a methylation defect within IC2 on 11p15.5. These patients are at risk for cancer in early childhood. The spectrum of tumors in IC2 imprinting defect cases includes hepatoblastoma, rhabdomyosarcoma and gonadoblastoma [41] although these affect fewer than 5% of these patients. Jackson et al [42] recently reported a case with BWS and an IC2 defect with ATRT. However the ATRT also bore somatic loss of 22q and an inactivating SMARCB1 mutation, demonstrating that germ-line inactivation of CDKN1C was not sufficient to cause ATRT. This is consistent with the widespread inactivation of CDKN1C in many different cancers with additional oncogenic mutations and suggests that CDKN1C silencing contributes, but is alone insufficient, to cause tumors. Germline SMARCB1 mutations have been recently described in rare families with predisposition to RT and to schwannomatosis [5,12,13]. It might be expected that individuals carrying inactivating mutations in SMARCB1 in the germline might display phenotypic features in line with impaired CDKN1C expression such as high birth weight. However this has not been reported in the literature and too few cases have been identified to date to directly examine these associations.
The fact that epigenetic silencing of CDKN1C is so widespread in human cancer makes it an attractive target for epigenetic therapies. To our knowledge it has not been previously identified as a target for Romidepsin. The promoter of CDKN1A is however frequently acetylated by HDACi and expression is induced (reviewed in [43]). The fact that CDKN1C is an imprinted gene makes it a more attractive target as its imprinted status provides an additional layer of developmental regulation and implies that comparatively small changes in expression in embryonic-like cells are likely to have potent effects on cell proliferation and differentiation [40]. In addition it has recently been shown to be a critical regulator of embryonic growth regulating through IGF1, a growth factor frequently abnormally expressed in human cancer.
In summary we have demonstrated that CDKN1C is an important target for SMARCB1 and that RT cells are sensitive to small changes in CDKN1C expression. We show for the first time that SMARCB1 has a role in CDKN1C promoter histone acetylation suggesting opportunities for further investigation of signalling pathways that may be epigenetically regulated by SMARCB1. Furthermore as Romidepsin restores CDKN1C pro- moter histone acetylation, leading to increased CDKN1C expression, we propose that Romidepsin and related compounds could be further explored as therapeutic reagents in the treatment of RT through restoration of gene expression patterns lost as a result of SMARCB1 inactivation.

Cell lines and cell culture
The G401 RT cell line was obtained from the American Type Culture Collection (ATCC). The establishment of STM91-01, BT16 and SJSC were previously reported [44,45]. These RT cell lines were maintained in RPMI 1640 supplemented with 10% fetal calf serum (FCS) and grown at 37uC in 5%CO 2 . SMARCB1 recombinant STM91-01 cells were grown in the same media supplemented with 100 mg/mL hygromycin (Roche, Mannheim, Germany) and 1 mg/mL puromycin (Sigma Chemical Company, St. Louis, MO). G401 cells were maintained in McCoy's 5A modified media supplemented with 10% FCS, and both SMARCB1 and CDKN1C recombinant G401 clones were grown in the same media supplemented with 600 mg/mL geneticin (Invitrogen, Groningen, Netherlands) and 5 mg/mL puromycin (Sigma).
Induction of SMARCB1 and CDKN1C in transformed cell lines was achieved by the addition of 1 mM 4-hydroxytamoxifen (4HT: Sigma Aldrich Castle Hill Australia). 4HT (1 mM) and Trichostatin A (TSA; Sigma Aldrich) (3.0 mM) were prepared in 100% ethanol. Romidepsin (10 mM)(Gloucester Pharmaceuticals) and 5-Aza-29-deoxycytidine (5aza2dC; Sigma Aldrich) (2 mM) were prepared in 100% DMSO and stored in aliquots at 220uC. Appropriate dilutions were made into culture medium for experimental analyses. Relevant solvent controls were included in all experiments. Culture media and reagents were replaced every 24 hrs.
For imprinting and gene expression assays, cells were plated as replicate cultures at 2610 5 cells/well in 2 ml culture volumes. Cells were trypsinized at 24 hourly time points, over a total timecourse of 72 hours, washed in PBS, and pellets were frozen at 280uC for future RNA and DNA extraction.

Southern blotting
To confirm SMARCB1 deletion or rearrangement in RT cell lines and in clinical specimens of rhabdoid tumor, 10 ug of RT cell line genomic DNA was digested with a panel of restriction enzymes and run overnight on 0.8% agarose gels alongside normal human genomic DNA as a control. DNA was transferred overnight in 206 SSC buffer onto Zeta Probe R GT Genomic tested blotting membrane (Biorad, Hercules CA 94547). Membranes were fixed under UV light, prehybridized at 60uC in ExpressHyb buffer (Clontech, Mountain View CA 94043) and hybridized with a 32 P-labelled full length SMARCB1 PCR product, prepared by the random priming method, for 18 hours. Membranes were washed at room temperature with 26 SSC / 0.05% SDS for 40 mins and then with 0.16 SSC / 0.1% SDS for a further 40 mins at 55uC and exposed to phosphoimager screens for 24 to 48 hours. Bands were visualized with ImageQuant TL v2003.02 software following digital capture on a scanning phosphoimager (Molecular Dynamics).
Cell cycle analysis 5610 5 cells were seeded and treated for 72 hrs. Media was replaced each 24 hrs. 5610 5 cells were stained for cell cycle analysis using a modified flow cytometry assay [46]. Briefly, 5610 5 cells were collected, washed in PBS/0.1% FCS, fixed with 1% formaldehyde (Polysciences, Warrington, PA) on ice for 30 mins, then permeabilized on ice with 0.5% Triton X-100 in PBS (Roche) for a further 15 mins. Cells were then incubated on ice in the dark for 45 mins with either PE-conjugated mouse anti-human Ki67 monoclonal antibody or PE-conjugated mouse IgG1 k monoclonal isotype control (BD Biosciences). After a final wash with PBS/ 0.1% FCS, cells were resuspended in 0.3 mL of washing buffer containing 1 mg/mL of 7-aminoactinomycin D (7AAD; Molecular Probes, Eugene, USA). Flow cytometric analysis was performed on a LSRII analyser (Becton Dickson) and data analysed using the FACS Diva software program. For each sample, electronic gating was performed on the intact, single cell population. Ki67 expression and 7AAD signal intensity were used to determine the cell cycle status of viable cells.

Inducible expression system
The human SMARCB1 coding sequence was PCR amplified from plasmid (PCR4-TOPO-INI1) containing full-length human SMARCB1, using forward primer 59-GGATCCACCATGATGAT-GATGGCGCTGAGC-39 and reverse primer 59-GCCTCGAGT-TACCAGGCCGGGGCCGTGTTGG-39 containing BamH1 and Xho1 restriction sites respectively. The PCR product was digested with BamH1 and Xho1 and inserted into the vector pcDNA3 UAS/ neo (kindly provided by Dr.J.Silke, La Trobe University, Victoria, Australia ). The integrity of the construct was verified by direct automated sequencing. The vector was linearized and then cotransfected with a second linearized vector, pEF puro/GEV, into G401 cells. Transfection was achieved using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells containing both pcDNA3 UAS SMARCB1 and pEF puro/GEV were doubly selected in geneticin and puromycin. Single cells were selected and grown and clones were tested for SMARCB1 expression by Western blotting following the addition of 1 uM 4HT. The full length human CDKN1C coding sequence was PCR amplified from plasmid (pCMV6-XL4 CDKN1C) (Origene Technologies, Rockville, MD), using forward primer 59 GCGGTACCACCATGTCC-GACGCGTCCCTCC-39 and reverse primer 59-GCGGATCCT-CACCGCAGCCTCTTGCG-39 containing Kpn1 and BamH1 restriction sites respectively. The PCR product was digested with Kpn1 and BamH1 and inserted into the vector pcDNA3 UAS/neo. Transfection and selection in G401 cells was the same as for SMARCB1 and clones were tested for CDKN1C expression by western blotting following the addition of 1 uM 4HT.

Lentiviral production and transfection
The STM91-01 cell line was found to be refractory to transfection using the above vector system and a lentiviral system was used to produce pooled recombinant cell lines [47].
The human SMARCB1 coding sequence was amplified and cloned into the BamHI and XbaI sites of the vector, pF 56UAS Sv40 puro, using the same forward primer as above and a modified reverse primer, 59-GCTCTAGATTACCAGGCCGGGGCCGTGTTGG-39. HEK 293T cells were transiently transfected with pF 56UAS Sv40 puro containing SMARCB1or pF GEV16 hygro vector with the packaging vectors pCMVdR8.2 and VSVG, using Effectene (Qiagen, Hilden Germany) as described by the manufacturer. The media was changed after 16 hrs and virus-containing supernatant was collected after 48 hours, filtered through 0.45 mm Millipore filters, and frozen at 280uC.
For transduction in STM91-01 cells, 1610 6 of the target cells were washed in PBS and then incubated in virus-containing media containing 5 mg/mL polybrene (Sigma Aldrich) in 6-well plates. Plates were then spun at 2500 rpm for 90 mins at 30uC, incubated for a further 90 mins at 30uC, then returned to 37uC in 5%CO 2 . Media was changed the next day and 2 days after transduction the appropriate selection was added. First, cells were infected with pF GEV16 hygro virus and selected in hygromycin. Hygromycin resistant cells were then infected with pF 56UAS Sv40 puro/ini1 virus followed by selection in puromycin. A pool of resistant cells was then tested for SMARCB1 expression by western blotting following the addition of 1 uM 4HT.

siRNA transfection
Optimal transfection conditions for G401 F22 cells were determined using BLOCK-iT fluorescent oligo (Invitrogen). Cells were plated at optimal density for the culture plate used and allowed to adhere overnight. ON-TARGETplus SMARTpool CDKN1C siRNA (Dharmacon, CO, USA) was transfected at a final concentration of 10 nM in the presence of 3 mg/mL DharmaFECT2 transfection reagent (Dharmacon) according to the manufacturer's instructions. Additionally, ON-TARGETplus siCONTROL Non-targeting pool (Dharmacon) was transfected in the same manner as a control (mock) transfection. 6 hrs after transfection, 1 mM 4HT or ethanol control was added to induce SMARCB1 expression. Media was refreshed every 24 hrs, maintaining 1 uM 4HT or ETOH, as required, with 1 nM oligo. Cells were harvested at 24 hr intervals for RNA extraction and for determination of cell proliferation by the MTS assay. CDKN1C expression was examined as described under ''Gene expression and imprinting''. Cell proliferation was measured by the MTSbased CellTiter 96 Aqueous non-radioactive cell proliferation assay (Promega, WI, USA). Briefly, triplicate wells of siRNA transfected cells, induced or un-induced for SMARCB1, were incubated with [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt] (MTS)/ phenazine methosulfate (PMS) solution for 2 hrs at 37uC. Well absorbance was measured at 492 nm and a reference wavelength of 620 nm using a Multiskan ES spectrophotometer (Thermo Electron, ON, Canada) at 24 hourly time intervals. The absorbance at each time interval was normalized in each culture to the t 0 absorbance value, which was set to 1.0. Absorbance differences in each 24 hour interval, reflecting the rate of proliferation in individual cultures, were calculated by subtraction of normalized absorbance values.

Gene expression and imprinting
Genomic DNA was extracted from RT cell pellets, comprising at least 2610 5 cells, using the DNeasy Blood and Tissue kit (Qiagen) according to the manufacturer's instructions. CDKN1C primers F4 (59 GCGTTCTACCGCGAGACG 39) and R5 (59 CGAGTGCAGCTGGTCAGCGAG 39) were used to amplify across the region of deletion polymorphisms in exon 1 using the Enhancer PCR system (Invitrogen Corporation CA 92006) and Hi Fi Taq polymerase (Roche, Mannheim, Germany). PCR products were digested with PvuII and separated on 6% non-denaturing polyacryamide gels containing 10% glycerol to identify polymorphisms required for imprinting analysis as previously described [26]. Bands were visualised by staining with ethidium bromide and digital images were captured following UV exposure on a Biorad Gel Doc 1000 system. PCR products were sequenced using Big Dye terminator technology to determine allele sizes.
RNA was extracted from RT cell pellets comprising at least 2610 5 cells, using the RNeasy Mini kit (Qiagen) according to the manufacturer's instructions. 1 ug of RNA was reverse transcribed with random primers (pd(N)6) (Amersham Pharmacia Biotech) and M-MLV Reverse Transcriptase RNaseH minus (Promega Corporation). CDKN1C was amplified with primers F2 (59GCAGCTGCCTAGTGTCCCGGTC 39) in exon 1 and R7 (59 TAAAT TGGCTCACCGCAGCC 39) spanning exons 2 and 3 within the 39UTR using the Enhancer PCR system with Hi Fi Taq polymerase. Cycling was 95uC for 3 mins, 45 cycles of 95uC for 45 secs, 56uC for 30 secs, 68uC for 1 min 30 secs and a final extensions step at 68uC for 10 mins. PCR products were digested with PvuII and electrophoresed as described above.
Real-time Q-PCR for CDKN1C expression was performed using the primer and Taqman probe combination described in Niemitz et al [48]. These were p57 1044F 59 GCGGCGATCAA-GAAGCTG 39 and p57 1124R 59 CGACGACTTCT-CAGGCGC 39 with probe p57 1069T FAM 59 CTCTGATCTCCGATTTCTTCGCCAAGC 39 TAMRA. Reactions were performed using a Quantitect Probe PCR probe mastermix (Qiagen) and run in triplicate on a Rotorgene 3000 real-time thermal cycler (Corbett Technologies, Sydney Australia). Reaction tubes were set up using the CAS1200 robotic system of Corbett Technologies. Data was analysed using the Rotorgene's comparative quantitation software using selected untreated samples as the calibrator. Expression of the housekeeping gene, glucuronidase beta (GUSB), was examined by Q-PCR using the Quantitect Probe PCR kit (Qiagen) according to the manufacturer's instructions. Reactions were set up and performed in triplicate as for CDKN1C. Data was analysed using comparative quantitation software and untreated samples were used as the calibrator as for CDKN1C. CDKN1C data was normalized to GUSB to derive relative changes in CDKN1C expression. Levels of GUSB remained stable in all reactions, however after 48 hours of treatment with 10 nM Romidepsin, GUSB levels were noted to decline in the RT cell lines.

Methylation analysis and Chromatin Immunoprecipitation
Methylation at LIT1 (OMIM 604115) was examined by southern blotting following digestion of 5 ug of DNA with EcoRI and NotI. Southern blots were probed with the probe DMRP recognizing a 4.2 kb methylated and 2.7 kb unmethylated band [27]. Prehybridization, hybridization and washing steps were as described for southern blotting above. Bands were visualized and quantified with ImageQuant TL v2003.02 software following digital capture on a scanning phosphoimager (Molecular Dynamics). The methylation index was calculated as the peak volume of the methylated band divided by the sum of the volumes of the methylated and unmethylated bands after background subtraction.
Methylation within the CDKN1C promoter was analysed by sequencing bisulphite-modified DNA.1 ug of DNA extracted from RT cells was bisulphite modified using the MethyEasy TM DNA Bisulphite Modification Kit (Human Genetic Signatures Pty Ltd NSW Australia). Modified DNA easy was amplified using primers p57-CF 59 GGTTGGGYGTTTTATAGGTTA 39 and p57-CR 59 ACCTAACTATCCGATAATAAACTCTTC 39 generating a 155 bp product derived from the region across the CDKN1C transcription start site from 241 to +114 as described in Kikuchi et al [22]. The amplified product was then cloned into the PCR-Script Amp cloning vector (Stratagene, La Jolla CA). Cloning and transformation were performed as recommended in the PCR-Script TM Amp Cloning kit (Stratagene). Selected blue colonies were grown in liquid media and plasmid DNA was extracted and purified using the QIAprep R Spin Miniprep Kit (Qiagen). Plasmid DNA was sequenced using Big Dye terminator technology (Applied Biosystems) according to the manufacturer's instructions using vector primer T7. At least ten plasmids each were sequenced from un-induced and 4HT-induced RT cultures.
Chromatin immunoprecipitation was performed exactly as described in the protocols provided with the Acetyl-Histone H4 and H3 Immunoprecipitation (ChIP) assay kits (Upstate cell signalling solutions Temecula CA 92590). Sonication was performed by pulsing three times for 10 seconds at 30% power, using a Vibra Cell TM ultrasonic processor (SONICS). Sheared DNA was examined on 1% agarose gels. DNA bound to immunoprecipitated histone was recovered by phenol/chloroform extraction and ethanol precipitation.PCR on ChIP'd DNA was performed across the CDKN1C transcription start site from 272 to +89 as described in Kikuchi et al [22]. Primers were p57 ChIP F 59 GTA-TAAAGGGGGCGCAGGCGGGCT 39 and p57 ChIP R 59 TGGTGGACTCTTCTGCGTCGGGTTC 39 (product 161 bp). 2 uL of DNA was amplified in a reaction volume of 25 uL containing 16Sybr Green, 16Sensimix, 0.25 mM MgCl 2 , and 10 picomoles of each primer as per the recommendations in the 26 Sensimix DNA kit (Quantace, London). Cycling was 95uC for 10 mins, 45 cycles of 95uC for 15 secs, 62uC for 15 secs, 72uC for 20 secs followed by a melt step ramping from 72 to 95uC. GAPDH was amplified using primers GAPDH ChIPF 59 TCGGTGCGTGCCCAGTTGAACC 39 and GAPDH ChIPR 59 ATGCGGCTGACTGTCGAACAGGAG 39 (product 246 bp) under the same conditions. Reactions were set up in triplicate for each sample on a CAS1200 robotic system and run on a Rotorgene 3000 real-time thermal cycler (Corbett Technologies). Reaction conditions were designed to eliminate primer dimer formation. The comparative quantitation option in the Rotorgene software was used to analyse the data. A normal control DNA sample run in triplicate was used as the calibrator for both CDKN1C and GAPDH. CDKN1C/GAPDH ratios were derived for each run. CDKN1A and CDKN1B promoters in ChIP fractions were examined in a similar manner using CDKN1A promoter primers described in Takai et al [49] and CDKN1B promoter primers described in Li et al [50]. These were p21F PROM 5 9GGTGTCTAGGTGCTC-CAGGT 39 and p21R PROM 59 GCACTCTCCAGGAGGA-CACA 39 (product 255 bp), p27F PROM 59 GTCCCTTC-CAGCTGTCACAT 39 and p27R PROM 59 GGAAAC-CAACCTTCCGTTCT 39 (product 162 bp). Sensimix reaction mixes were as described for CDKN1C and GAPDH however additional MgCl 2 was not added. CDKN1A cycling was 95uC for 10 mins, 45 cycles of 95uC for 15 secs, 61uC for 15 secs, 72uC for 20 secs followed by a melt step ramping from 72uC to 95uC. CDKN1B cycling was 95uC for 10 mins, 45 cycles of 95uC for 15 secs, 57uC for 15 secs, 72uC for 20 secs followed by a melt step ramping from 72uC to 95uC. Reaction conditions were designed to eliminate primer dimer formation.

Tissues and mutation screening in ATRT
Specimens with a clinical diagnosis of RT were submitted to the laboratory for molecular diagnostic testing for SMARCB1 mutation, or absence of SMARCB1 protein [51], after obtaining informed consent. Analysis of clinical samples was approved by the institutional ethics committee under approval number RCH HREC 24073A. RNA was extracted from frozen sections using the RNeasy mini kit (Qiagen) and DNA was extracted from either frozen sections or formalin-fixed tissue using protocols described in the DNeasy Blood and Tissue kit (Qiagen). cDNA was prepared in a total volume of 40 uL from 1 ug RNA as described above for RT cell lines. Allele loss studies were performed using microsatellite markers mapping to 22q, on both blood and tumor DNA. Markers analysed included D22S303, D22S257, D22S301, D22S345, D22S1685 and TOPIP2. Alleles were examined by PCR amplification of 50-100 ng DNA using primers end-labelled with Fam, Hex or Tet. PCR products were electrophoresed on 4.5% 0.2 mm denaturing polyacrylamide gels on an ABI 377 DNA sequencer employing TAMRA 500 (Red) size standard. Sample lanes were tracked with Genescan software Version 3.1.2 and analysed using Genotyper Version 2.1. Mutation screening was performed by direct DNA sequencing of PCR products. Primers were located in introns flanking individual exons and were designed to amplify exonic and flanking splice site sequences. Exon 1 was divided into two separate overlapping fragments for PCR. Primer sequences and reaction conditions are available on request. Sequence analysis was performed by both manual inspection and with Mutation Surveyor (Softgenetics PA 16803). BLAST analysis was performed against genome databases.