The chromatin remodelling factor BRG1 is a novel binding partner of the tumor suppressor p16INK4a

Background CDKN2A/p16INK4a is frequently altered in human cancers and it is the most important melanoma susceptibility gene identified to date. p16INK4a inhibits pRb phosphorylation and induces cell cycle arrest, which is considered its main tumour suppressor function. Nevertheless, additional activities may contribute to the tumour suppressor role of p16INK4a and could help explain its specific association with melanoma predisposition. To identify such functions we conducted a yeast-two-hybrid screen for novel p16INK4a binding partners. Results We now report that p16INK4a interacts with the chromatin remodelling factor BRG1. We investigated the cooperative roles of p16INK4a and BRG1 using a panel of cell lines and a melanoma cell model with inducible p16INK4a expression and BRG1 silencing. We found evidence that BRG1 is not required for p16INK4a-induced cell cycle inhibition and propose that the p16INK4a-BRG1 complex regulates BRG1 chromatin remodelling activity. Importantly, we found frequent loss of BRG1 expression in primary and metastatic melanomas, implicating this novel p16INK4a binding partner as an important tumour suppressor in melanoma. Conclusion This data adds to the increasing evidence implicating the SWI/SNF chromatin remodelling complex in tumour development and the association of p16INK4a with chromatin remodelling highlights potentially new functions that may be important in melanoma predisposition and chemoresistance.


Background
The cyclin dependent kinase inhibitor p16 INK4a is frequently inactivated in human cancers and is a highly pen-etrant melanoma susceptibility gene; more than 50 germline mutations have been identified in high-risk melanoma-prone families [1]. The principal function of p16 INK4a is to inhibit cell cycle progression by preventing the cyclin dependent kinases CDK4 and CDK6 from phosphorylating the retinoblastoma protein, pRb. In the presence of p16 INK4a , pRb remains hypophosphorylated and forms active pRb-E2F transcriptional repressor complexes that silence genes required for S-phase entry. Consequently, ectopic expression of p16 INK4a promotes pRbdependent G1 cell cycle arrest and senescence. Moreover, functional p16 INK4a is commonly maintained in pRb-deficient tumors (reviewed by Sherr & Roberts [2]), and this underscores the dependency of p16 INK4a on the pRb pathway.
Hypophosphorylated pRb can repress gene transcription at least partly by remodelling chromatin structure through its interactions with proteins such as HDAC1, BRM and BRG1 [3][4][5]. As the catalytic core of the SWI/SNF chromatin remodelling complex, the interaction between BRG1 and pRb was proposed to recruit the complex to E2F responsive promoters and enhance pRb transcriptional repressor activity. [5] There is also evidence that BRG1 acts upstream of pRb by repressing cyclin D1 expression [6] and upregulating the expression of the CDK inhibitors p21 Waf1 , p15 INK4b and p16 INK4a [7][8][9] to maintain pRb in an active, hypophosphorylated state. Not surprisingly, BRG1 may function as a tumor suppressor; BRG1 hemizygous mice are susceptible to tumors [10], while complete loss of BRG1 potentiates lung cancer development [11] and BRG1 is silenced or mutated in human tumor cell lines derived from breast, ovarian, lung, brain and colon cancers [4,12]. BRG1 is also lost in established neuroendocrine carcinomas and adenocarcinomas of the cervix [13], and the loss of BRG1 expression in lung cancer is associated with a poor prognosis [14,15].
In this study, it is identified for the first time that BRG1 specifically associates with p16 INK4a in vivo, and that both proteins are frequently lost in human melanomas. Although both BRG1 and p16 INK4a regulate pRb activity we found no evidence that p16 INK4a and BRG1 co-operate in cell cycle regulation. Targeted silencing of BRG1 did not diminish pRb-dependent p16 INK4a activities; p16 INK4a retained potent cell cycle inhibitory activity and induced senescence in the presence and absence of BRG1. Contrary to previous reports, that BRG1-deficient cells are relatively resistant to p16 INK4a -induced cell cycle arrest [16], we show that pRb activity is BRG1-independent and thus, BRG1 does not influence p16 INK4a -mediated cell cycle inhibition. Together with the frequent loss in primary melanomas the novel BRG1 interaction with the melanoma associated tumor suppressor p16 INK4a implies an important role for BRG1 in melanoma.

BRG1 binds p16 INK4a
From a yeast two-hybrid screen using full-length human p16 INK4a as bait, we isolated the C-terminal 530 amino acids of the chromatin remodelling factor BRG1 as a potential binding partner ( Figure 1A). This segment of BRG1 incorporates the ATPase domain, which facilitates ATP hydrolysis, and the bromodomain, which enables binding to acetylated histones [17]. To confirm that fulllength BRG1 also binds p16 INK4a in human cells, both pro-Identification of BRG1 as p16 INK4a binding partner Figure 1 Identification of BRG1 as p16 INK4a binding partner. A Schematic illustration of BRG1 highlighting the domains isolated in the yeast 2-hybrid screen (Y2H clone) B U2OS cells were transfected with MYC-p16 INK4a and FLAG-BRG1 or control vector and immunoprecipitations were performed with a mouse-anti-FLAG antibody or a matched mouse IgG as indicated. BRG1 and p16 INK4a were detected on immunoblots with anti-FLAG and anti-MYC antibodies. C Fluorescent microscopy images (FM) and confocal microscopy images (CF) of SW-13 cells grown on cover slips and transfected with MYC-p16 INK4a and FLAG-BRG1 and probed with anti-FLAG and anti-MYC antibodies.
teins were co-expressed transiently in U2OS osteosarcoma cells and MYC-tagged p16 INK4a was specifically co-purified with FLAG-tagged BRG1 in immunoprecipitation assays using a FLAG-specific antibody ( Figure 1B). Further, when both proteins were co-expressed in the SW-13 adrenocortical carcinoma cell line, they co-localized in the nucleus in distinct nuclear speckles ( Figure 1C).
To verify that endogenous BRG1 also interacts with p16 INK4a , we initially utilized the WMM1175_p16 INK4a inducible melanoma cell model, which we have previously described [18]. p16 INK4a expression was induced with IPTG to reach physiologically relevant levels comparable to those seen in the WS-1 normal human dermal fibroblasts at passage 20 ( Figure 2A). Using a p16 INK4aspecific antibody we isolated BRG1 from nuclear WMM1175_p16 INK4a lysates ( Figure 2B). Importantly, the interaction between BRG1 and p16 INK4a was also confirmed in WS-1 normal human dermal fibroblasts at passage 20, using a p16 INK4a -specific antibody ( Figure 2C).

pRb pathway in human cell lines
To establish the role of BRG1 on p16 INK4a function we selected six cancer cell lines, varying in their p16 INK4a , pRb and BRG1 status [12,16]. As shown in Figure 3 and Table  1, p16 INK4a expression was inversely related to pRb expression and only detected in the pRb-negative SAOS-2 osteosarcoma and C33A cervical cancer cells. All other cell lines had detectable pRb and no p16 INK4a . (Note, there is a slight leakage of the ectopically introduced p16 INK4a in the p16-inducible WMM1175_p16 INK4a cells without IPTG upon long exposure.) The BRG1 homologue, BRM was expressed in all but the C33A cells and SW-13 adrenocortical carcinoma cells. Importantly, SW-13 and C33A cells were also negative or extremely low for BRG1 expression levels. The H1299 lung cancer cells were deficient for BRG1 expression, and all remaining cell lines had detectable levels of BRG1. It is also worth noting that the HCT116 cells carry only a mutated, functionally impaired BRG1 allele (BRG1 Leu1163Pro ) [12]. CDK4 was expressed strongly in all cell lines, while its homologue, CDK6 was either absent or poorly expressed in the pRb negative SAOS-2 and C33A cells and present in the remaining cells.

p16 INK4a requires pRB to induce cell cycle arrest
To define the impact of BRG1 on p16 INK4a function we transiently expressed either BRG1, p16 INK4a or both proteins in this panel of six cell lines. The short-term expression of BRG1 alone had no effect on the cell cycle distribution of the cell lines tested. As expected, neither p16 INK4a alone nor p16 INK4a in combination with BRG1 promoted cell cycle arrest in cells deficient for pRb (SAOS-2 and C33A). In contrast, introduction of p16 INK4a induced potent cell cycle arrest in all cell lines expressing pRb (U2OS, H1299, HCT116, SW-13) even when the cells lacked BRG1 (H1299) or carried a reported mutant form of BRG1 (HCT116) [12]. Further, co-expression of BRG1 did not significantly enhance the p16 INK4a induced cell cycle arrest in the U20S, H1299 or HCT116 cells. Importantly even in SW13 cells, which lack both BRG1 and its homologue BRM, p16 INK4a expression alone induced a significant cell cycle arrest and this was enhanced to some extent by over-expressing BRG1 ( Figure 4). These data Immunoprecipitations were performed using a mouse anti-p16 INK4a antibody or a matched mouse IgG from nuclear cell lysate, as indicated. Immunoblots were probed for endogenous BRG1 and induced p16 INK4a using a mouse anti-BRG1 and rabbit anti-p16 INK4a , respectively. C Endogenous BRG1 was co-immunoprecipitated with p16 INK4a from WS-1 normal dermal human fibroblasts grown to passage 20 as detailed above.
confirm that p16 INK4a -induced cell cycle arrest requires intact pRb, but not BRG1.

p16 INK4a does not require BRG1 to promote cell cycle arrest or induce cell senescence
To thoroughly evaluate any functional interaction between p16 INK4a and BRG1, we stably silenced BRG1 in the inducible WMM1175_p16 INK4a cell model. These cells were transfected with a silencing molecule specifically targeting BRG1 or a non-specific (NS) silencing molecule directed to the luciferase transcript. Two BRG1-silenced clones, WMM1175_p16 INK4a _siBRG1 W9 and X1, with > 95% reduction in BRG1 accumulation and two control clones WMM1175_p16 INK4a _sicontrol E1 and X2, with unaltered BRG1 expression, were selected for analysis. All clones remained inducible for p16 INK4a expression ( Figure  5A, B).
Silencing of BRG1 had no significant impact on the proliferation rate or cell cycle distribution of the WMM1175_p16 INK4a cell line. In the absence of BRG1, p16 INK4a retained the ability to inhibit the proliferation of the WMM1175 cells ( Figure 5C), and this was associated with arrest in the G1-phase of the cell cycle with a concomitant S-phase inhibition ( Figure 5D) that was maintained over the five-day induction period (data not shown). Moreover, the silencing of BRG1 had no impact on the ability of p16 INK4a to totally prevent outgrowth of colonies upon low seeding density ( Figure 5E). BRG1 has been reported to induce senescence in SW-13 cells and in mesenchymal stem cells [7,19] and the role of p16 INK4a in initiating and maintaining senescence is widely acknowledged (reviewed by Huschtscha & Reddel [20]). We investigated the role of BRG1 in p16 INK4ainduced senescence. The long term induction of p16 INK4a in WMM1175_p16 INK4a cells was not influenced by the BRG1 status, caused pRb hypophosphorylation ( Figure  6A) and induced senescence-like features in the pRb pathway proteins in cell lines Figure 3 pRb pathway proteins in cell lines. Expression of BRG1 and BRM was analyzed using 50 μg of nuclear cell lysates. All other proteins were analyzed from 50 μg of total cell lysates.
Expression of the indicated proteins is summarized with + or -. Mutant status of BRG1 in HCT-116 cells has been reported [12].
WMM1175 cells as reported previously [18,21], ( Figure  6B). These features included increased cell size and granularity, positive senescence-associated β-galactosidase activity and the appearance of senescence-associated heterochromatin foci. Formation of foci coincides with the recruitment of pRb to E2F-responsive promoters and is associated with the stable repression of E2F-target genes [22]. This important marker of pRb activity was not affected by BRG1 silencing. Similarly, BRG1 silencing did not alter the build up of SA-β-galactosidase induced by p16 INK4a (Figure 6B) or p16 INK4a induced changes in cell size and granularity in the WMM1175 cells ( Figure 6C), the latter corresponds to senescence associated vacuolisation. This data confirms that cell cycle regulation and induction of cell senescence by p16 INK4a does not require BRG1.

BRG1 is lost in melanoma
To evaluate the role of BRG1 in melanomas, we examined immunohistochemically stained paraffin sections from archival paraffin-embedded tissue blocks of a series of primary and metastatic melanomas for expression of the chromatin remodelling factor BRG1 and p16 INK4a ( Figure  7). As presented in Table 2 was found to localise to the nucleus and cytoplasm. The proportion of BRG1 expression was slightly higher in the metastatic melanomas than in the primary melanomas, but, this did not reach significance using a Mann-Whitney Wilcoxon test. BRG1 and p16 INK4a were readily detectable in cultured, normal, primary human melanocytes (data not shown) and therefore our data imply that BRG1-loss has an important role in melanoma development.

Discussion
The p16 INK4a tumor suppressor has a critical influence on melanoma tumorigenesis. We have now shown that the chromatin remodelling factor BRG1 is a novel binding partner of p16 INK4a and confirm this interaction in vivo.
More importantly, we show that loss of BRG1 occurs frequently in primary and metastatic melanomas and propose that BRG1 may play an important role as a tumor suppressor in this cancer.
We have also shown that p16 INK4a requires pRb, but not BRG1 to promote cell cycle arrest. This differs from several previous findings in the literature but agrees with others: It has been suggested that the pRb-BRG1 interaction is required for the pRb repression of E2F-target genes such as cyclin E and cyclin A, and thereby cell cycle arrest. According to this hypothesis, cells lacking BRG1 would harbor only inactive pRb, thus conferring resistance to p16 INK4a induced growth arrest [5,16]. These findings differ from those of Bultman et al. [23] who did not observe a functional interaction between pRB and BRG1 in their murine models and Kang et al. [7], who showed that the BRG1-pRB interaction was not required for BRG1 induced cell cycle arrest in SW-13 cells. In contrast to our work, Kang et al. [7] used long-term BRG1 expression, which caused growth arrest in SW-13 cells, and showed that BRG1 bound the p21Waf1 promoter and upregulated its expression 3-7 days after BRG1 expression. This was sufficient to induce cell cycle arrest and senescence independent of the BRG1 ability to complex with pRb. In this study we have clearly demonstrated that p16 INK4a requires pRb, but not BRG1, to promote cell cycle arrest. Our data is mainly based on the thorough analysis of a well-defined melanoma cell model, with inducible physiological relevant expression levels of p16 INK4a and the use of highly specific BRG1-silencing molecules. In this model, p16 INK4a induction promotes rapid G1-cell cycle arrest followed by cellular senescence, and these functions were not affected by silencing of BRG1. Chromatin changes, which involve chromatin remodelling, are an important step in p16 INK4 /pRb dependent senescence [24]. It was recently shown that the BRG1 homologue, BRM, forms an initiating component of heterochromatin complexes during the senescence of melanocytes [25]. BRG1 has also been implicated in senescence of melanocytes, as it has been identified in the SWI/SNF complex facilitating transcription in response to IGFBP7, the latter itself being an important player in oncogenic BRAF-induced senescence [26]. However, our data show that p16 INK4a is able to promote senescence in WMM1175 melanoma cells in the absence of BRG1 indi-cating that the p16 INK4a /pRb senescence pathway does not require BRG1.
As the catalytic component of the SWI/SNF chromatin remodelling complex, BRG1 facilitates unwinding of DNA helices bound to and wrapped around histone structures. The SWI/SNF chromatin remodelling complex can be recruited by specific DNA binding molecules such as transcriptional activators or repressors and directed to specific DNA targets. For instance, BRG1 promotes p53 dependent transcription by interacting with this tumor suppressor [27,28], while it functions as a co-repressor of E2F dependent transcription by associating with the E2F transcriptional repressor pRb [5]. Furthermore, BRG1 has recently been reported to promote transcriptional activity of the melanocyte specific transcription factor MITF-M [29]. MITF-M plays an important role in melanocyte proliferation and survival (reviewed by Goding) [30] and activates the expression of p16 INK4a [31]. It is possible that the p16 INK4a interaction with BRG1 modulates any one or more of these functions. Regardless of the function of the BRG1-p16 INK4a complex, it is evident that BRG1 expression can be lost relatively early in melanoma development, with a significant proportion (> 70%) of primary melanomas showing no detectable BRG1 expression, while BRM expression was usually maintained in these tumors (< 20% loss). Overall, the rate of BRG1 loss was high in melanomas and comparable to that of p16 INK4a [32], which suggests that selection against BRG1 expression arises relatively early in melanoma genesis. The fact that, additionally to the frequent loss of either tumor suppressor, a high proportion of melanomas show loss of both proteins correlates with our data showing BRG1-independence of the p16 INK4a cell cycle regulatory functions and this suggests BRG1 independent and dependent functions of p16 INK4a . BRG1 is proposed to be an important modulator of chromatin in melanocytic cells. In particular, BRG1 promotes transcriptional activity of the melanocyte specific transcription factor MITF-M [29], reduction of BRG1 expression in zebrafish embryos leads to a reduction in neural crest derived cells including melanocytes [33] and thirdly we found BRG1 expression in normal, primary human melanocytes. Therefore we propose that BRG1 is a vital melanoma associated tumour suppressor, which is frequently lost in the initial stages of the disease.  The identification of BRG1 as a potential tumor suppressor in melanoma adds to the increasing evidence implicating the SWI/SNF chromatin remodelling complex in tumor development. BRG1 mutations have been identified in small cell lung carcinomas [34] and loss of BRG1 expression or mislocalisation of BRG1 to the cytoplasm has been associated with poor prognosis in this malignancy [14,15]. Another study showed that 71% of neuroendocrine carcinomas of the cervix had lost BRG1 expression [13] and BRG1 has been implicated in breast cancer through its role in estrogen receptor dependent transcription [35], its interaction with the breast cancer susceptibility gene BRCA1 [27] and because BRG1 haploinsufficient mice are prone to mammary tumors [23]. Furthermore, BRG1 is often lost or mutated in various tumor cell lines including cells derived from pancreatic, ovarian, lung, brain and colon cancer [12]. In primary melanoma, the chromosomal region of BRG1 (19p13.2) is not deleted at high frequency [36], nevertheless, translocations in this chromosomal region have been associated with the disease in three cases [37].

Conclusion
We have identified BRG1 as a novel binding partner of the tumor suppressor p16 INK4a and confirmed this interaction in normal cells. Together with our immunohistologic data confirming frequent BRG1 loss in primary melanomas, this implicates BRG1 as an important tumor suppressor in melanoma.

Yeast two-hybrid screen
The Matchmaker2 Gal4 yeast two-hybrid system (Clontech, Mountain View, CA, USA) was used to screen for p16 INK4a binding partners in the Y190 yeast strain with p16 INK4a cloned into the pAS2 vector in frame to the Gal4 binding domain and a human brain cDNA library cloned into the pACT2 vector in frame with the Gal4 transactivation domain (Clontech, Mountain View, CA, USA) according to the manufacturers instructions.

Stable BRG1 silenced p16
WMM1175_wtp16 cells expressing a siRNA targeting BRG1 or a control siRNA molecule targeting luciferase were induced for 1, 3 or 5 days with 4 mM IPTG or mock treated. For each time point the cell cycle distribution was determined as described above.
Only tumor samples with enough tissue for staining of at least two of the proteins were included in the study.