Identifying Targets for the Restoration and Reactivation of BRM

ORIGINAL ARTICLE Identifying targets for the restoration and reactivation of BRM

B Kahali1, SJB Gramling1, SB Marquez1, K Thompson1,LLu2 and D Reisman1

Brahma (BRM) is a novel anticancer gene, which is frequently inactivated in a variety of tumor types. Unlike many anticancer genes, BRM is not mutated, but rather epigenetically silenced. In addition, histone deacetylase complex (HDAC) inhibitors are known to reverse BRM silencing, but they also inactivate it via acetylation of its C-terminus. High-throughput screening has uncovered many compounds that are effective at pharmacologically restoring BRM and thereby inhibit cancer cell growth. As we do not know which speciﬁc proteins, if any, regulate BRM, we sought to identify the proteins, which underlie the epigenetic suppression of BRM. By selectively knocking down each HDAC, we found that HDAC3 and HDAC9 regulate BRM expression, whereas HDAC2 controls its acetylation. Similarly, we ectopically overexpressed 21 different histone acetyltransferases and found that KAT6A, KAT6B and KAT7 induce BRM expression, whereas KAT2B and KAT8 induce its acetylation. We also investigated the role of two transcription factors (TFs) linked to either BRM (GATA3) or HDAC9 (MEF2D) expression. Knockdown of either GATA3 and/or MEF2D downregulated HDAC9 and induced BRM. As targets for molecular biotherapy are typically uniquely, or simply differentially expressed in cancer cells, we also determined if any of these proteins are dysregulated. However, by sequencing, no mutations were found in any of these BRM-regulating HDACs, HATs or TFs. We selectively knocked down GATA3, MEF2D, HDAC3 and HDAC9, and found that each gene-speciﬁc knockdown induced growth inhibition. We observed that both GATA3 and HDAC9 were greatly overexpressed only in BRM-negative cell lines indicating that HDAC9 may be a good target for therapy. We also found that the mitogen-activated protein (MAP) kinase pathway regulates both BRM acetylation and BRM silencing as MAP kinase pathway inhibitors both induced BRM as well as caused BRM deacetylation. Together, these data identify a cadre of key proteins, which underlie the epigenetic regulation of BRM.

Oncogene (2014) 33, 653–664; doi:10.1038/onc.2012.613; published online 25 March 2013 Keywords: epigenetic; histone transferase; SWI/SNF; histone deacetylase complex; tumor suppressor; brahma

INTRODUCTION movement of histones within chromatin.11,12 These subunits also SWI/SNF is a multimeric chromatin-remodeling complex ﬁrst have distinct subdomains (Bromodomains) that are important for discovered in yeast as regulators of the SWItch and Sucrose histone interactions, as well as an Rb-binding domain (containing NonFermenting phenotypes.1,2 This complex has since been the LXCXE sequence required to bind to Rb), and a helicase 3 shown to have a role in human gene expression, where it domain that is tied to the mechanistic function of the complex. 13–16 associates with and is required for the function of a diverse array Both BRM and BRG1 bind Rb and facilitate Rb function. The of key cellular proteins and transcription factors (TFs), many of loss of BRM and BRG1 renders Rb nonfunctional and represents which have important roles in cancer development.3,4 The SWI/ another way that the Rb pathway can be potentially inactivated in SNF complex remodels chromatin such that regions and domains cancer. Like Rb, p53 has also been functionally linked to BRG1 and 17–19 of DNA are made more accessible to cellular proteins that possibly BRM. In murine model systems, BRG1 is a low- promote gene expression.5–7 SWI/SNF therefore serves a catalytic penetrance tumor suppressor while BRM loss does not induce role in facilitating and promoting gene expression. As a result of tumors, but rather potentiates tumor development when 20–22 the large number of cellular proteins that are functionally linked to combined with a carcinogen. It has been suggested that this complex, it is not surprising that this complex is frequently each ATPase subunit can compensate for the loss of the other and targeted and disrupted during cancer development. SWI/SNF has lessen the impact of inactivation of one of these subunits. If true, been linked to growth control, DNA repair, differentiation, this could potentially explain why BRM and BRG1 are often development and cellular adhesion, and when SWI/SNF is concomitantly lost in cancer cell lines and primary tumors. To rendered nonfunctional or impaired, these processes are also date, however, the proteins that regulate BRM and BRG1 have not negatively impacted3,4 potentially promoting cancer development been established, and this lack of knowledge is an obstacle to and progression. understanding how to restore the expression of these genes in a The SWI/SNF complex is composed of one of two ATPase/ manner that is clinically effective. catalytic subunits, Brahma (BRM) or Brahma-related gene 1 (BRG1), BRM and BRG1 are inactivated in 30–40% of lung cancer cell along with 8–10 additional subunits.8–10 Although the exact lines and 15–20% of primary lung tumors.23–28 Although BRG1 is functions of the non-ATPase subunits are not yet completely mutated in 70–80% in cancer cell lines,24,28 the mechanism of understood, BRM and BRG1 function as the catalytic subunits that BRG1 loss in primary tumors is not yet known and it appears not convert ATP energy into the mechanical motion necessary for the to involve genetic mutation.29,30 Like BRG1, BRM is silenced in

1Division of Hematology/Oncology, Department of Medicine, University of Florida, Gainesville, FL, USA and 2Department of Pathology, University of Florida, Gainesville, FL, USA. Correspondence: Dr D Reisman, Division of Hematology/Oncology, Department of Medicine, University of Florida, 2033 Mowry Road, Cancer/Genetic Building, Office 294, Gainesville, FL 32611, USA. E-mail: dnreisman@ufl.edu Received 26 March 2012; revised 12 November 2012; accepted 14 November 2012; published online 25 March 2013 Restoration and reactivation of BRM B Kahali et al 654 10–25% of a broad spectrum of solid tumor types,25 similar to the BRM has been shown by several laboratories to be readily frequency of KRAS, HER2, EGFR and ALK alterations in breast induced as well as inactivated via acetylation following treatment and lung cancers. However, unlike tumor-suppressor genes that with pan-HDAC inhibitors such as: suberoylanilide hydroxamic acid are most often irreversibly mutated, BRM is not mutated, but (SAHA), Trichostatin A, Entinostat (MS-275) and butyrate.25,31,32,34 As rather reversibly suppressed.25 Broad-acting histone deacetylase these HDAC inhibitors nonspecifically inhibit both class 1 (HDACs 1, complex (HDAC) inhibitors such as suberoylanilide hydroxamic 2,3and8)andclass2(HDACs4,5,6,7,9,10and11)HDACs,these acid, Trichostatin A and butyrate readily induce BRM expression in data indicate that one or more HDACs regulate BRM—but the BRM-deficient cell lines.25,31,32 However, despite the ability of findings do not implicate a particular enzyme. To identify which these compounds to induce BRM expression, they were also HDACs control BRM expression, we selectively knocked down each discovered to inactivate BRM via C-terminal acetylation31 thereby of the class 1 HDACs (HDAC1, HDAC2, HDAC3, HDAC8 and HDAC11; precluding the use of these compounds for clinical restoration of Supplementary Figure 1) as compared with control shRNA to define BRM function. Hence, either epigenetic silencing or direct the role of these proteins, particularly HDAC3, in the induction of acetylation can disrupt BRM function. BRM. A comparison of BRM expression levels following the shRNA- Although the ability of HDAC inhibitors to reactivate BRM mediated knockdown of each of these class 1 HDACs demonstrates expression implicates HDACs in the regulation of BRM, few, if any, that BRM expression was only induced when HDAC3 was specific HDAC proteins have yet been linked to BRM regulation. suppressed (Figure 1a: BRM protein; Supplementary Figure 2 for Similarly, as histone acetyltransferases (HATs) are known to work BRM mRNA data). Although inhibition of HDAC3 restored BRM in close concert with HDACs to regulate protein acetylation, what expression in deficient cell lines, it remained unclear whether this role this family of proteins has in the regulation of BRM is likewise induced BRM was functional. Thus, as a marker of BRM activity as unknown. We recently conducted a high-throughput screen and function of HDAC3 knockdown, we measured the expression of identified a number of different compounds that can restore BRM three BRM-dependent genes in both the SW13 and C33A cell lines expression,33 although the mechanism of BRM restoration still that we have previously shown to be upregulated when BRM is remains unclear. In order to identify any potential clinically functionally induced.35 We conducted quantitative PCR (qPCR) for effective drugs that could be used to restore BRM, it is necessary these genes in parental C33A and SW13 cell lines and daughter cell to understand the endogenous proteins involved in BRM lines where anti-BRM shRNAs were introduced to suppress BRM silencing, as well as its inactivation. To accomplish this, we used induction. We observed induction of these BRM-dependent genes a combination of short hairpin RNA (shRNA) knockdown, and only in the parental cell lines where BRM was induced B7- to 15- transfection experiments in both BRM-positive and BRM-negative fold (Figure 1b) and not in the C33A and SW13 BRM-knockdown cell lines to determine the HDACs and HATs that control BRM daughter cell lines. These data indicate that suppressing HDAC3 expression; we used similar experiments to see which of these restores both BRM expression and function. By using a class 1 HDAC proteins control BRM acetylation. We uncovered two TFs, and inhibitor, we found comparable results validating this shRNA data using this RNA interference technology and transfection (Supplementary Figure 3). experimentation, we found that they sequentially regulate Along with others, we have previously shown that many of the HDAC9 and then BRM expression. Although any gene that nonspecific HDAC inhibitors that induce BRM expression (for regulates BRM, could in theory, be targeted pharmacologically example, MS-275, butyrate, CI994 and Trichostatin A) also induce to clinically induce BRM, genes that are either specifically mutated BRM acetylation, resulting in BRM inactivation.25,31 To determine or overexpressed in tumor specimens tend to make the best which of the class 1 HDACs are functionally linked to BRM targets for therapy. Hence, we sequenced the identified BRM- acetylation, we knocked down each of the class 1 HDACs (HDAC1, regulating HDACs, HATs and two TFs for any mutations or HDAC2, HDAC3, HDAC8 and HDAC11) with gene specific shRNAs alterations that could explain the loss of BRM seen in cancer. We or scrambled shRNA (control) in the H441 and H661 cell lines. We knocked down each of these BRM-linked genes to determine the noted that the induction of BRM acetylation was only observed impact on cellular growth when each of them was inhibited. when HDAC2 expression was inhibited (Figure 1c and Supplemen- Finally, as inhibitors to mitogen-activated protein kinase kinase 1 tary Figure 4). As class 2 HDACs are much more tissue specific in (MAP2K1; alias: MEK1) and mitogen-activated protein kinase 1 their expression compared with broadly expressed class 1 HDACs, (MAPK1; alias: ERK) both induced BRM expression and caused we performed an initial screen for expression of each class 2 HDAC BRM deacetylation, this indicates that BRM silencing and BRM in four BRM-deficient cell lines (A427, SW13, C33A and H522). We acetylation are not in fact regulated by separate pathways, but found that HDAC4 and HDAC9, but not HDAC5, HDAC6, HDAC7 or instead are jointly regulated by the MAP kinase pathway. The HDAC10, were expressed in all four BRM-deficient cell lines results of these experiments illustrate the potential anticancer (Supplementary Figure 5). To determine if the class 2 HDACs impact of pharmacologically targeting each of these genes and (HDAC9 and/or HDAC4) affect BRM expression, we knocked down the complexity of the mechanism underlying BRM regulation. HDAC4 and HDAC9 with gene-specific shRNA (Supplementary Figure 1) and compared these results with scrambled shRNA in both SW13 and C33A cell lines. We observed that suppression of RESULTS either HDAC4 or HDAC9 induced BRM protein (Figure 1d) and mRNA HDAC3 and HDAC9 regulate BRM function and HDAC2 controls expression (Supplementary Figure 6: BRM mRNA (B6- to 14-fold BRM acetylation induction)). Using anti-HDAC4 or anti-HDAC9 shRNAs showed that Our initial goal was to determine which specific HDACs regulate the suppression of either HDAC4 or HDAC9 does not induce BRM BRM. We used both a specific HDAC-class 1 inhibitor and a pan- acetylation (Figure 1d). This finding is similar to knockdown with HDAC-class 2 inhibitor in order to help determine which HDACs anti-HDAC3 (Figure 1a), which also does not induce BRM acetylation. are involved in BRM regulation and acetylation. We then used Of note, HDAC9 suppression, as compared with HDAC4, resulted in a selective shRNAs to knock down and suppress each of the known significantly more robust induction of BRM protein (B4-fold more). 11 HDACs to specify which HDACs control BRM expression and Thesedatawereconfirmedbyusingapan-class2HDACinhibitor which control BRM acetylation, because it is critical to determine (Supplementary Figures 7 and 8). the specific proteins that regulate each of these processes. To determine if a given HDAC controls BRM inactivation or BRM function, we evaluated, by western blot, the presence or absence MEF2D and GATA3 regulate BRM and HDAC9 of BRM acetylation and also demonstrated BRM function by The GATA and MEF2 families of TFs are known to be linked to the showing the induction of BRM-dependent genes. function and/or expression of HDACs 3, 4 and 9.36–39 Specifically,

Oncogene (2014) 653 – 664 & 2014 Macmillan Publishers Limited Restoration and reactivation of BRM B Kahali et al 655

Figure 1. (a) HDAC1, 2, 3, 8 or 11 genes were virally suppressed with gene-specific shRNAs in the BRM-deficient SW13 and C33A cells. After puromycin selection, protein lysates were collected. Western blotting showed that only the knockdown of HDAC3 induced BRM protein. (b) HDAC3 knockdown induced BRM-dependent genes, LGAL, DDX58, P8 and XAF1 between 7- and 15-fold in the parental cell lines, C33A and SW13, but not in the daughter cell lines where BRM expression has been suppressed by anti-BRM shRNA. (c) The expression of HDAC1, HDAC2, HDAC3, HDAC8 and HDAC11 was virally knocked down with the gene-specific shRNA in both H441 and H661 cell lines. BRM acetylation was observed only when HDAC2 expression was suppressed. (d) The suppression of HDAC4 yielded induction of BRM protein by western blotting, whereas silencing of HDAC9 by gene-specific shRNAs produced a qualitatively stronger induction of BRM. Glyceraldehyde 3- phosphate dehydrogenase (GAPDH) was used as loading control; H460 was the positive control; the negative controls were the untreated parental cell lines, SW13 or C33A.

HDAC9 expression is known to be controlled by MEF2D, whereas (Figure 2a), induced BRM expression in both BRM-deficient cell GATA2 and GATA3 are known to be involved in the regulation of lines, C33A and SW13. Moreover, the knockdown of MEF2D was BRM.40 We therefore reasoned that these TFs might regulate BRM also associated with a B5-fold decrease in HDAC9 expression expression via their impact on either HDAC3 and/or HDAC9 (Figure 2b). This is consistent with studies of the HDAC9 promoter expression. To investigate this, we used two complementary that have shown that MEF2D binds to the HDAC9 promoter and approaches: we first used shRNA knockdown of these genes in drives its transcription.39 In BRM-deficient C33A and SW13 cell BRM-negative cell lines and second, we overexpressed these lines, we next silenced the expression of GATA2 and GATA3 using genes, via transfection, in BRM-positive cell lines. In addition, shRNA and found that GATA3 knockdown as compared with cells nuclear run-on experiments were used to determine if the levels transduced with scrambled shRNA (control) not only induced BRM of BRM and HDAC9 change because of an alteration in their mRNA expression about B10-fold, but also downregulated transcription. HDAC9 (B9-fold; Figures 2a and b); in comparison, HDAC9 We conducted western blotting and qPCR to determine which expression was not changed appreciably when MEF2D or GATA3 of the MEF2 family members (MEF2A-D) are consistently was knocked down in the BRM-positive cell lines H157 and H441 expressed in BRM-deficient cell lines. We found that MEF2B and (Figure 2b). GATA2 and MEF2B knockdown had no effect on BRM MEF2D were expressed in four BRM-negative and four BRM- expression in either the BRM-deficient cell lines (SW13 and C33A) positive lung cancer cell lines (Supplementary Figure 9). Knocking or in the BRM-positive cell lines (H441 and H157). Similarly, shRNA down MEF2B and MEF2D using gene-specific shRNA as compared knockdown of GATA2, GATA3, MEF2B or MEF2D had no effect on with control shRNAs revealed that only the silencing of MEF2D HDAC3.

& 2014 Macmillan Publishers Limited Oncogene (2014) 653 – 664 Restoration and reactivation of BRM B Kahali et al 656

Figure 2. (a) MEF2D, GATA3 or both (MEF2D plus GATA3) were knocked down using gene-speciﬁc shRNA in the SW13 and C33A cell lines. After 72 h, protein lysates were harvested from both cell lines. By western blotting, we observed that the knockdown of MEF2D, GATA3 or both resulted in the induction of BRM. (b) HDAC9 mRNA expression was examined by qPCR as after MEF2D or GATA3 shRNA knockdown in both BRM-positive cell lines, H157 and H441, and BRM-negative cell lines, C33A and SW13. MEF2D and GATA3 knockdown caused B3- and B6.5- fold downregulation, respectively, in HDAC9 mRNA expression in the BRM-negative cell lines while no change in HDAC9 was observed in the BRM-positive cell lines. (c) Two BRM-positive cell lines, H441 and H157 were virally infected with MEF2D, GATA3 or HDAC9 gene. After 72 h of puromycin selection, we examined BRM expression by qPCR and found that forced expression of HDAC9, GATA3 or MEF2D decreased BRM mRNA expression by B7-, B5- and B3-fold in these cell lines, respectively. (d) We transfected in HDAC9, GATA3, MEF2D or control (empty vector) expression vectors into the BRM-positive cell lines H441 and H157. By qPCR, ectopic expression of HDAC9, MEF2D or GATA3 induced HDAC9 mRNA by 15- to 20-fold, 5- to 6-fold or 20- to 22-fold, respectively, as compared empty vector transfected controls in the H157 and H441 cell lines. (e) To determine if GATA3 and MEF2D can regulate BRM independently of HDAC9 expression, the BRM-deﬁcient cell lines SW13 and C33A were transduced with either the HDAC9 gene (test) or an empty cassette (control) to achieve a relatively stable expression of HDAC9. qPCR of BRM expression in these cell lines showed that exogenous HDAC9-driven expression blocked BRM induction, when either GATA3 or MEF2D were also knocked down. In comparison, in the absence of exogenous HDAC9-driven expression (in cell lines infected with empty cassette), shRNA silencing of either GATA3 or MEF2D induced BRM mRNA approximately 7- to 10-fold. (f) To determine how BRM is regulated by HDAC9, GATA3 and MEF2D, we knocked down each of these genes and then measured BRM transcription by nuclear run-on. Knockdown of HDAC9, GATA3 and MEF2D induced BRM transcription B20-, B14- and B12-fold, respectively, in both the SW13 and C33A cell lines.

To further confirm a role for these genes in BRM regulation, the significantly affect each other’s expression in these cell lines. GATA3, MEF2D or HDAC9 genes were each transduced into the These data indicate that GATA3 and MEF2D can drive HDAC9 BRM-positive cell lines: H441 and H157. Regardless of whether overexpression, consequently resulting in the suppression of BRM. GATA3, MEF2D or HDAC9 was introduced, the overexpression of This finding, however, does not rule out the possibility that one or each of these genes caused the downregulation of BRM mRNA both of these TFs can impact BRM expression independently of in BRM-positive cell lines (Figures 2c; B5-, B2- and B6.5-fold HDAC9. decrease, respectively). In addition, we examined if GATA3, MEF2D Hence, to determine if GATA3 or MEF2D can regulate BRM and HDAC9 were able to affect the levels of each other’s dependently or independently of HDAC9, we knocked down expression. Using qPCR, we observed that overexpression of GATA3 or MEF2D while driving the overexpression of HDAC9. After either GATA3 or MEF2D induced HDAC9 expression (Figure 2d) an shRNA knockdown of GATA3 and MEF2D, we found that GATA3 average of B17.5- and B4-fold, respectively, in H441 and H157 and MEF2D mRNA expression was decreased B6- to 7-fold by cells. In comparison, transfection of HDAC9 into these cells yielded qPCR, and their respective proteins were undetectable by western B16-fold higher levels of HDAC9 mRNA by qPCR (Figure 2d). In blot in either SW13 or C33A cell lines (Supplementary Figure 10). contrast, the overexpression of MEF2D and GATA3 did not In the presence of HDAC9 overexpression, we did not observe an

Oncogene (2014) 653 – 664 & 2014 Macmillan Publishers Limited Restoration and reactivation of BRM B Kahali et al 657 induction in BRM expression when either MEF2D or GATA3 were knocked down (Figure 2e: data columns D and E). This is in contrast to the induction in BRM expression observed when either GATA3 or MEF2D was knocked down in the absence of HDAC9 ectopic overexpression. (Figure 2e: data columns B and C; B7- to 11-fold). These data support the hypothesis that GATA3 and MEF2D primarily regulate BRM expression by regulating HDAC9 expression. Finally, we conducted nuclear run-on experiments to examine BRM transcription rates after MEF2D, GATA3 or HDAC9 was knocked down in either SW13 or C33A cell lines. We found BRM transcription was increased between 12- and 20-fold when these three genes were individually suppressed (Figure 2f), Figure 3. (a) Twenty-one different lysine acetyltransferases (KATs) indicating that BRM transcription is indeed regulated by these were introduced into SW13 and C33A cells by transient transfection, proteins. and after 72-h total protein was harvested from these cell lines. Transfection of only the KAT6A, KAT6B or KAT7 plasmids were observed to induced BRM protein expression by western blotting in HATs can restore or inactivate BRM SW13 and C33A cell lines. (b) Each of these 21 different KATs were also transfected into the BRM-nonacetylated cell lines H661 and Our next question was to resolve the role of HATs in BRM H441. After 72 h, protein lysates were obtained from both cell lines. regulation, because HATs have an antagonistic function to HDACs. By western blotting, we observed that by transient transfection of We examined each HAT to determine which specific HAT(s) either KAT2B or KAT8 into these cell lines induced acetylation of the control BRM expression. This was accomplished by using BRM protein. complementary experimental approaches: we first transfected (overexpressed) each HAT in BRM-negative cell lines and second, in BRM-positive cell lines, we performed shRNA knockdown of those HATs linked to BRM regulation. Similar to the methods used MEF2D splicing generally occurs in all cell lines or only in BRM- for studying the HDACs, in order to determine which HATs control deficient cell lines, we compared these findings with those BRM acetylation, we transfected each HAT into BRM-positive cells obtained from the sequencing of several BRM-positive cell lines and western blotted for the induction of BRM acetylation. and found no such splicing variant in any of the BRM-positive cell The balance between HDACs and HATs determines the final lines. We next examined the expression levels of GATA3 and acetylation status of a protein, typically histones.41 As inhibition of MEF2D mRNA by qPCR in BRM-positive and BRM-negative cell HDACs can restore BRM expression,25 we reasoned that ectopic lines to determine if changes in GATA3 or MEF2D mRNA expression correlate with BRM silencing. We found that GATA3, expression of one or more HATs might similarly reactivate B BRM expression in BRM-deficient cell lines. To determine which but not MEF2D, mRNA levels were overexpressed 10- to 12-fold HATs induce BRM, we transiently transfected 21 (Supplementary (average) in BRM-deficient cell lines, as compared with BRM- Table 1, includes a list of HATs tested and their sources) different positive cell lines (Figure 4A). We similarly investigated HDAC3 and HDAC9 mRNA levels in BRM-negative and BRM-positive cell lines HATs into both SW13 and C33A cell lines (Supplementary Table 2 B shows the average fold increases for different HATs after transient and found that HDAC9, but not HDAC3, mRNA levels were 500- transfection). Of these HATs, we found that ectopic expression of fold higher (on average) in BRM-deficient cell lines (Figure 4B) as KAT6A, KAT6B or KAT7 specifically induced BRM mRNA (B15- to compared with BRM-positive cell lines. 25-fold; Figure 3a). Conversely, we knocked down KAT6A, KAT6B To determine if this overexpression of GATA3 and HDAC9 or KAT7 in BRM-positive cell lines (H157, H460 or H661) and occurs in both primary lung tumors and lung cancer cell lines and observed a decrease in B6- to 7-fold for each of these HAT correlates with BRM expression, we conducted qPCR and western mRNAs, as determined by qPCR (Supplementary Table 2 shows blots for HDAC9 and GATA3 in BRM-negative versus BRM-positive fold decreases of individual HATs mRNA after shRNA-mediated non-small cell lung cancers. We found GATA3 and HDAC9 mRNA were expressed at higher levels in BRM-negative lung cancers as knockdown). We examined the impact of each HAT knockdown on B B BRM expression by qPCR and found a significant downregulation compared with BRM-positive tumors, 12- and 25-fold, of BRM expression (Supplementary Figure 11). We also introduced respectively (Figure 4C). Similarly, we also found higher GATA3 these same 21 HATs into the BRM-positive cell lines H661 and and HDAC9 protein expression in BRM-negative tumors as H441, which lack BRM acetylation, and immunoblotted for BRM opposed to that of BRM-positive tumors (Supplementary acetylation. In both cell lines, we found that only KAT2B and KAT8 Figure 12). To ensure that this differential expression of GATA3 transfections led to the induction of BRM acetylation at levels and HDAC9 was derived from the tumor and not the normal tissue comparable to BRM acetylation seen in the H460 cell line within the tumors samples, we immunohistochemically stained (Figure 3b). both BRM-positive and BRM-negative tumors for GATA3 and HDAC9 expression. We found that qualitatively, HDAC9 and GATA3 are expressed at higher levels in BRM-negative tumors as Alterations of TFs, HATs and HDACs linked to BRM regulation.To compared with BRM-positive tumors (Figure 4D). Of particular extend our findings, we next explored the extent to which any of note, HDAC9 overexpression appears to occur in a variety of the HATs, HDACs or TFs identified as having a role in BRM different cancer types,42 indicating that HDAC9 might be a good regulation are dysregulated, potentially leading to inappropriate target for therapy, similar to HER2 in breast cancer43,44 or BRM silencing or acetylation. To study this, we sequenced the epidermal growth factor receptor (EGFR) in lung cancer.45 mRNAs for each of the genes. We first sequenced HDAC3 and HDAC9 mRNA in four BRM-deficient cell lines and found no mutations or alterations in either HDAC. We also sequenced Targeting GATA3, MEF2D, HDAC3 or HDAC9 leads to BRM-dependent MEF2D, and although we did not find any mutations, we identified growth inhibition. BRM and the SWI/SNF complex are known to a unique splice variant involving exon 9 in the four BRM-negative be essential co-factors for Rb-, p107- and p130-mediated growth cell lines sequenced that has not yet been reported. This alternate inhibition.46,47 Multiple laboratories have shown that Rb is splice variant lacks the small 18-bp exon 9 and thus eliminates six ineffective in arresting growth unless the SWI/SNF complex is amino acids from the primary protein sequence. The impact of this active.13,15,16,48 Furthermore, either ectopic or pharmacologic change is currently unknown. To determine if this alternative restoration of BRM expression is sufficient to induce growth

& 2014 Macmillan Publishers Limited Oncogene (2014) 653 – 664 Restoration and reactivation of BRM B Kahali et al 658

Figure 4. (A) We analyzed six BRM-positive cell lines: H460, H441, H157, H1355, H827 and H358, as well as ﬁve BRM-negative cell lines: C33A, SW13, A427, H522 and H125 for GATA3 expression by qPCR. We observed that GATA3 mRNA expression is 8- to 16-fold higher in these BRM- negative cell lines as compared with these BRM-positive cells. (B) In six BRM-positive and ﬁve BRM-negative cell lines, we examined HDAC9 mRNA expression as function of BRM expression by qPCR. We observed that HDAC9 on average was approximately 500-fold higher (range: 32- to 1024-fold) in the BRM-negative cell lines as compared with the BRM-positive cell lines. (C) Four BRM-positive and four BRM-negative tumors were analyzed for HDAC9 and GATA3 mRNA expression by qPCR. We observed B12 and B25-fold higher for GATA3 and HDAC9 mRNA, respectively, in BRM-negative tumors. (D) BRM-positive and BRM-negative non-small cell lung cancers were immunohistochemically stained for BRM, HDAC9, and GATA3. We observed appreciably higher HDAC9 and GATA3 expression in the BRM-negative tumors as compared with the BRM-positive tumors.

inhibition.35 Hence, we next determined the impact on cell growth had been silenced by BRM shRNA (Figure 5b). In contrast to the when we restored BRM expression by knocking down GATA3, class 2 HDACs, we identified a class 1 HDAC inhibitor (RGFP966) MEF2D, HDAC3 and/or HDAC9. that was relatively specific for HDAC3. Treatment with this We transduced the two BRM-deficient cell lines, C33A and compound induced both BRM mRNA and BRM protein at SW13, with either: anti-GATA3, anti-MEF2D, anti-HDAC3 or anti- approximately B5 mM (Supplementary Figure 13). Similar to our HDAC9 shRNA or with control (scrambled) shRNA and compared results with HDAC3 shRNA-mediated knockdown (Figure 5a), we the growth of each of these cell lines. In both cell lines (C33A and observed marked growth inhibition in both BRM-negative cell SW13) we observed growth inhibition (B70% and B83%, lines (C33A and SW13 transduced with scrambled shRNA), when respectively) when any of these genes was suppressed as treated with this HDAC inhibitor (Figure 5b). However, when BRM compared with the scrambled shRNA (Figure 5a). To induction is blocked by anti-BRM shRNA, the observed RGFP966- demonstrate that the observed growth inhibition was BRM- mediated growth inhibition was blunted by B75% in these dependent in these cell lines, we performed essentially the same cell lines. Similarly, treatment with the pan-class 2 HDAC inhibitor, experiment except we also silenced BRM using anti-BRM shRNAs. MC1568 also resulted in growth inhibition in the parental In each of the BRM-suppressed cell lines, the growth inhibition BRM-negative cell lines transduced with control shRNA, but not was blunted, and their growth approached that of the control cell in the daughter cell lines harboring anti-BRM shRNA. Taken lines (Figure 5a). Although compounds do not yet exist that together, the selective targeting of HDAC3, HDAC9 or both, specifically inhibit HDAC4 or HDAC9, the class 2 inhibitor MC1568, appears to be an effective means of inhibiting growth in BRM- which targets multiple class 2 HDACs, was used in BRM-deficient deficient cancer cells. cell lines that do not express HDAC 5, 6, 7, 8 or 10, to determine the impact of inhibiting HDAC4 and/or HDAC9. From our studies MAP kinase pathway controls BRM induction and BRM acetylation. using this compound, we observed growth inhibition of 490% To begin to understand the signal transduction pathways only in parental cell lines infected with control (scrambled) shRNA, that control BRM silencing and/or BRM acetylation, we used our compared with the daughter cell lines in which BRM expression high-throughput screening assay33 and tested the ability of a

Oncogene (2014) 653 – 664 & 2014 Macmillan Publishers Limited Restoration and reactivation of BRM B Kahali et al 659

Figure 6. (a) The two BRM-deficient cell lines, SW13 and C33A, were treated with PD98059 (MAP2K1 inhibitor) or PD169316 (p38 MAPK1 inhibitor) at a concentration of 10 mM. After 72 h of treatment with either compound, protein lysates were harvested from each cell line. Western blotting of these protein samples showed that BRM protein is readily induced in treated cell lines. (b) The BRM acetylated cell lines, H460 and H157, were treated for 72 h with either PD98059 or PD169316 at concentration of 10 mM. Western blotting revealed that these compounds each resulted in BRM deacetylation. (c) The two Figure 5. (a) Expression of HDAC9, GATA3 or MEF2D were each BRM-deficient cell lines, SW13 and C33A, were treated with PD98059 knocked down using gene-specific shRNA in both parental and (MAP2K1 inhibitor) or PD169316 (p38 MAPK1 inhibitor) at a daughter SW13 and C33A cell lines, whereas BRM was knocked concentration of 10 mM. After 72 h of treatment, protein lysates were down using anti-BRM shRNA in only the daughter cell lines. The collected from each treated cell line. Western blot analysis of these knockdown of HDAC9, GATA3 or MEF2D in the parental SW13 and protein samples showed an almost complete downregulation of C33A resulted in cellular growth inhibition (65–90% inhibition). In HDAC9 protein and B13-fold downregulation of GATA3 protein. comparison, these same knockdowns of either HDAC9, GATA3 or MEF2D in these daughter cell lines that had BRM suppressed with shRNA showed appreciably less inhibition (percentage inhibition various MAP kinases; however, the activation of BRM caused these 5–12%) in comparison with the parental cell lines. (b) The parental cells to undergo growth arrest, and no cells could be propagated, BRM-negative cell lines, C33A and SW13, or the daughter cell lines which limited our analysis. We further tested the impact of using which harbor anti-BRM shRNA (to block BRM induction) were these MAP kinase inhibitors on GATA3 and HDAC9 expression. We treated with either the HDAC3-specific inhibitor (RGFP966) at 5 mM or found a complete downregulation of both proteins by western the pan-class 2 HDAC inhibitor (MC1568) at 5 mM. Whereas in the blot (Figure 6c) in the BRM-deficient cell lines, C33A and SW13. parental cell lines the treatment with either these compounds Similarly, we found GATA3 and HDAC9 mRNAs were reduced by resulted in appreciable growth inhibition (70–80%) over 5 days, this B44- and B16 000-fold (Supplementary Figure 15), respectively, observed growth inhibition was blunted in both daughter cell lines (5–10% inhibition). by these compounds. These data clearly indicate a connection between BRM regulation and the MAP kinase pathway. Interest- ingly in lung cancer, the MAP kinase pathway is known to be frequently activated by EGFR mutations (9%),50 B-Raf (B3%)51 or number of different kinase inhibitors to induce BRM. Using this K-ras activation (25–30%),52,53 but BRM silencing occurs much less luciferase reporter assay, we found that the compounds PD98059 frequently, (B16–17%). As such, there must be other factors, and PD169316, which are known to be highly selective MAP2K1 49 which work in conjunction with the activated MAP kinase pathway and p38 MAPK1 inhibitors, respectively, readily induced to facilitate the downregulation of BRM. luciferase activity by B4-fold, which is an indirect measure of BRM expression. To verify BRM is induced, we treated two BRM- deficient cell lines with each of these compounds and found that BRM protein was induced by B11- and B6.5-fold by PD98059 DISCUSSION and PD169316, respectively (Figure 6a and mRNA data in We have shown that HDAC3 and HDAC9 regulate BRM expression, Supplementary Figure 14). In addition, we tested two cell lines whereas HDAC2 controls BRM acetylation. As the balance between (H460 and H157) that have BRM acetylation (Figure 6b). Applica- HATs and HDACs together controls the acetylation status of tion of these compounds not only induced BRM, but we observed proteins, we also identified that HATs are involved in BRM that both of these compounds also reversed BRM acetylation regulation—in particular, we found that KAT6A, KAT6B and KAT7 (Figure 6b). These data indicate that BRM acetylation and BRM control BRM expression, whereas KAT2B and KAT8 control BRM silencing pathways are linked at the level of the MAP kinase acetylation. We also wanted to know which of these proteins pathway. We also attempted to validate these data using shRNA to linked to BRM regulation might be mutated, overexpressed or

& 2014 Macmillan Publishers Limited Oncogene (2014) 653 – 664 Restoration and reactivation of BRM B Kahali et al 660 targeted genes are usually specifically mutated or overexpressed, making them unique to cancer cells. Having a uniquely or differentially expressed target within a given cancer is key to the success of these molecular-based therapies. The targeting of kinases, especially cell-surface kinases such as EGFR, HER2 and vascular endothelial growth factor receptor (VEGFR), is well- established as a potent clinical therapy. However, cancer is not only driven by acquisition of oncogenic mutations that act as cellular ‘gas pedals’: it is also activated by the loss of cellular ‘brakes’—tumor suppressors. Reactivation of suppressor genes as a form of targeted therapy has not yet been widely recognized because most of the tumor suppressors studied to date, such as Rb, p53, BRCA1 and phosphatase and tensin homolog deleted on chromosome 10, are also irreversibly inactivated in cancers by mutations. Unfortunately, gene therapy aimed at re-inserting normal tumor-suppressor genes back into deficient cancer cells has had relatively limited success.55,56 However, not all anticancer genes are mutated; there exists a group of genes that are Figure 7. Proposed model of BRM regulation. Transcriptional/ epigenetically—and therefore reversibly—silenced in human 57 translational regulation: as shown on the left side of the figure. cancer. The prototype gene has been p16, which is frequently MAP kinase drives GATA3 expression, which in turn, drives HDAC9 turned off via DNA methylation.58 As such, inhibitors of chromatin- expression in conjunction with MEF2D. The overexpressed HDAC9 remodeling proteins such as HDACs have proven effective in likely forms a complex with HDAC3 that blocks BRM transcription. treating certain tumor types.59,60 However, despite the potential The lysine acetyl transferases, KAT6A, -6B and -7 work in the effectiveness of HDAC inhibitors in certain tumors, the anticancer opposite manner whereby their activity results in inducing BRM genes targeted by HDACs have not yet been elucidated. In protein. Post-translational regulation: BRM protein can be post- translationally modified and inactivated by acetylation. As shown on contrast, our research has led us to the discovery that BRM is, in the right side of this figure, BRM protein can either be acetylated fact, one of these genes that is not mutated, but reversibly (non-functional) or non-acetylated (functional). HDAC2 serves to silenced in cancer cells and that restoring its function in BRM- 33,35,61 remove these acetyl groups from BRM thus activating the protein, deficient cells potently inhibits growth. In this work, we whereas KATs 2B and 8 acetylate BRM and inactivate it. The MAP have identified several proteins that appear to be key regulators of kinase pathway also functions to induce BRM acetylation by acting BRM expression, specifically: GATA3, MEF2D, HDAC3 and HDAC9. on KAT2B/8 and/or HDAC2. Therefore, the MAP kinase pathway Furthermore, the targeted inhibition of each of these proteins controls both the induction and acetylation of BRM. restores BRM expression and its potential anticancer effects. These findings have important implications for the targeted therapy of BRM-deficient cancers. HDAC3 emerges as a particularly otherwise dysregulated. We found that GATA3 and HDAC9 are intriguing target, as it is appears to be involved in a number of significantly overexpressed in cells that demonstrate BRM silen- human cancers,62 and specific inhibitors already exist for this cing. HDAC9 is a class 2 HDAC that is not ubiquitously expressed HDAC. However, as class 1 HDACs are more ubiquitously like the class 1 HDACs are, and it is overexpressed in BRM-negative expressed,63 targeting HDAC3 may have more toxic side effects. cells. Based on these data, HDAC9 may represent a good target for In comparison, HDAC9 appears to have several advantages if it therapy because its overexpression appears to be relatively was targeted for therapy. First, as a class 2 HDAC, HDAC9 specific for BRM-negative tumor cells. Finally, we show that the expression is much more restricted.63 Second, we and others have inhibition of HDAC9, either pharmacologically or by shRNA, results found HDAC9 to be overexpressed in many different cancer in growth inhibition. Together, these findings define the proteins types,64–67 so its dysregulation may be a common theme in underlying BRM suppression and illustrate that at least one of cancer. To date, there are no specific HDAC9 inhibitors that we these proteins, HDAC9, is a promising target for therapy. could utilize to explore whether the abrogation of HDAC9 function In addition, to define which HATs and HDACs control BRM would be clinically beneficial to patients. However, in this work, silencing and BRM acetylation, we also demonstrated that the using a pan-HDAC class 2 inhibitor (MC1568) that likely inhibits MAP kinase pathway is also part of the regulatory system and only HDAC9 and HDAC4, BRM was induced, which in turn led to appears to function upstream of the HATs and HDACs. This is not the arrest of cell growth. However, exposing patients to a the first report that the MAP kinase pathway controls BRM pan-HDAC inhibitor would also likely have unintended side expression, as Muchardt et al.54 showed that K-ras expression effects. In addition, as the apparent effective concentration for causes the downregulation of BRM, and BRM appears to this compound to induce BRM is in the micromolar range, this conversely suppress K-ras when ectopically overexpressed. would necessarily preclude the use of such drugs for the purpose Furthermore, activated B-Raf in melanoma cell lines has been of inhibiting HDAC9 as it is difficult to achieve such high drug observed to downregulate BRM as well (Dr Del La Serna, University levels in humans.68 As such, our findings demonstrate the of Toledo, personal communication). Although the MAP kinase potential utility of developing potent inhibitors (in the low pathway is known to have a critical role in transducing growth nanomolar range) specific to HDAC9 and testing their clinical signals, from our data, this pathway also appears to abrogate application in BRM-deficient cancers. growth inhibition in part by facilitating the downregulation of GATA3 and MEF2D also appear to be restricted in their BRM. We have illustrated our data in a diagram showing the distribution of tissue expression and not ubiquitously expressed, complex interplay of HATs, HDACs, TFs and the MAP kinase like the class 1 HDACs. Although MEF2D is not mutated or pathway, all of which act to regulate BRM (Figure 7). Just how this overexpressed in BRM-deficient cell lines, we found a yet- occurs and what other proteins are necessary to link the MAP unreported splice variant that might impact its function. In kinase pathway with HDACs/HATs still remains to be defined. addition, MEF2D could be aberrantly functioning because of With its proven therapeutic efficacy and relatively low toxicity, defective post-translational modifications, which are known to targeted therapy is an invaluable approach in the treatment of regulate its function.38,69,70 Although MEF2D appears to be tissue cancer. Most effective molecular therapies target kinases; these restricted in its expression, targeting MEF2D therapeutically might

Oncogene (2014) 653 – 664 & 2014 Macmillan Publishers Limited Restoration and reactivation of BRM B Kahali et al 661 be difficult to accomplish without producing significant side cells that feature aberrant growth control and in which mice effects because it is neither overexpressed nor mutated like other develop a far greater number of tumors when exposed to currently used targeted therapies. In contrast to MEF2D, GATA3 is carcinogens.25,77 However, it is important to remember that overexpressed in a variety of tumor types71,72 including BRM- reactivation of BRM represents the reversal of an epigenetic deficient lung cancers. Our work shows that elevated GATA3 mechanism that likely regulates many other genes, and not only expression drives HDAC9 overexpression such that GATA3 is in BRM. Hence, silencing via this mechanism may be a much more large part indirectly linked to the suppression of BRM via HDAC9 effective way to silence a cadre of genes rather than simply overexpression. Sequencing GATA3 failed to reveal any mutations, disrupting each gene individually. Although BRM silencing clearly although like MEF2D, GATA3 function could be impacted by can contribute to cancer development, it may not be the most aberrant post-translational modifications.73–75 As such, either important anticancer gene targeted by this overall mechanism. GATA3 or HDAC9 could be targeted for therapy as a means to By using a combination of specific HDAC inhibitors and an RNA restore the anticancer effect mediated by BRM. The next step will interference approach (shRNA), we determined that HDAC3, be to develop specific GATA3 and HDAC9 screening assays and HDAC4 and HDAC9 all regulate BRM. This is not surprising, as use them to determine which BRM-inducing ‘hits’, uncovered class 1 HDACs are known to recruit class 2 HDACs,78 and HDAC3 previously by high-throughput screening,33 specifically target is known to specifically interact with HDAC9.63 We also showed either GATA3 and/or HDAC9. Unlike HDAC9, which has a that a different HDAC (HDAC2) underlies BRM acetylation catalytic site that can be easily targeted and blocked, blocking and inactivation. In addition, as the function of HDACs are the function of a TF such as GATA3 (specifically, from binding) has counterbalanced by HATs, we also investigated which HATs not yet been accomplished. contribute to BRM regulation and found that KAT6A, KAT6B and As BRM is inactivated through both silencing and acetylation, KAT7 can induce BRM when ectopically expressed in BRM- identifying compounds that reverse both of these processes deficient cell lines. Although some HATs are known to be would be advantageous. Our data suggest that known inhibitors frequently mutated or form hybrid transcripts in leukemias and of the MAP kinase pathway are likely capable of activating BRM other cancers,79–82 we did not find any such alterations in BRM- when it has been silenced or acetylated. Hence, these data raise deficient cancer cells that might explain why BRM is silenced. questions regarding whether inhibitors of the upper MAP kinase Although it has been long appreciated that the HDAC inhibitors pathway, such as clinically available Raf inhibitors (for example, can restore BRM expression, but not necessarily BRM function,25 it Sorafenib) or perhaps EGFR inhibitors (Erlotinib (Tarceva) or was not until the work of Bourachot and Muchardt31 that it was Gefitinib (Iressa)), require restoration of BRM as a prerequisite for understood why. They elegantly demonstrated most HDAC their anticancer functions, or perhaps their failure to restore BRM inhibitors cause acetylation of the C-terminus of BRM protein might underlie development of clinical resistance over time. resulting in inactivation of BRM. Unlike BRM silencing where two Mechanisms of growth inhibition by kinase inhibitors are complex, HDACs can be linked to BRM expression, we found that only and such an approach would depend not only on Rb, but also on HDAC2 controlled BRM acetylation. Similarly, we found that KAT2B the BRM homolog, BRG1, as well as other proteins. However, and KAT8 can drive BRM acetylation. Again, although HDAC2, defining the role of BRM status as a function of these inhibitors KAT2B and KAT8 can be altered and thus become dysfunctional in will give us a better understanding of how and when these drugs cancer,79,81–83 sequencing the mRNAs of these three genes did should be used for ideal efficacy. not reveal any mutations in our BRM-deficient cancer cell lines. BRM is not only regulated by silencing and post-translational Although our unpublished work has revealed that BRM acetylation inactivation.15,25,31 Its expression is also modulated by occurs in both cell lines and primary tumors, further investigation microRNA.76 This complex regulation of BRM likely stems from is required to determine whether this form of BRM inactivation the multitude of cellular processes with which BRM and SWI/SNF occurs with any significant frequency. are involved, particularly growth control, differentiation and The study described is limited by the current understanding development.3,4 It is unclear why BRM is sometimes inactivated within the field of cancer biology as to how both HATs and HDACs, by acetylation while in other instances it is silenced by especially class 2 HDACs, are regulated and in turn regulate other suppression. Interestingly, the MAP kinase pathway is able to genes. Only now are we beginning to understand how these induce BRM silencing and in other cases, BRM acetylation. As the proteins are recruited to DNA regions where their modifications MAP kinase pathway can control both BRM acetylation and result in changes to the imprinting of chromatin. Such chromatin silencing, this suggests that the BRM regulation pathway likely modifications are the underlying mechanisms to the histone code, diverges below the level of the MAP kinase pathway. Although we which in turn underlies proper differentiation, development and believe that BRM acetylation completely inactivates BRM, it may normal cellular function. By first identifying some of the major only partially inhibit BRM, thereby restricting its function (it may proteins that impact the regulation of BRM, we are now in a bind to certain proteins and not others), whereas total BRM position to begin investigating exactly how they contribute to the silencing ensures complete loss of BRM activity. Further, BRM mechanism regulating BRM. Moreover, as epigenetic mechanisms acetylation and BRM sequencing might act sequentially, where if are unlikely to evolve to regulate to just one gene, many one mechanism can’t silence BRM the other mechanism can. additional genes may, in fact, be regulated by this same Furthermore, it is unclear whether BRM is important for cancer mechanism. Therefore, understanding the mechanism has broad development and why it is not simply mutated and eliminated like implications in the field of epigenetics, and cancer biology. other anticancer genes. Perhaps BRM is not a target of cancer The clinical reversal of the epigenetic regulation of an development, but instead is suppressed or acetylated because it anticancer gene is a novel, emerging, therapeutic strategy that happens to be collaterally impacted. Yet, arguing against this, SWI/ is potentially both effective and beneficial. Specifically, we show in SNF and BRM have been tied to many cellular functions that the article that restoring BRM is a novel target for therapy. oppose cancer development. This is most clearly demonstrated by Currently, there is little known about how BRM is epigenetically the fact that re-expression of BRM ectopically, virally or silenced or inactivated via acetylation, hence in this research, we pharmacologically, inhibits growth and induces a differentiated- defined which HDACs and HATs underlie the regulation of BRM. like morphology in essentially every cell line examined to date More importantly, we identified proteins that could make a good (B15).35,46,47 Moreover, BRM is known to bind to Rb and the Rb target for therapy by examining which of these proteins might be family members, p107 and p130, to facilitate growth inhibition selectively overexpressed in cancer cells. Before the best clinical when it is co-expressed with Rb.15,16,46,47 This is further supported targets for this purpose are clearly defined, it will likely require us by BRM-null animal models, which show the loss of BRM produces to understand most of if not the entire mechanism of BRM

& 2014 Macmillan Publishers Limited Oncogene (2014) 653 – 664 Restoration and reactivation of BRM B Kahali et al 662 regulation. Nevertheless, our work points toward HDAC9 being a Generation of RNA interference knockdowns good, potential, clinical target for therapy because it is: (1) pLKO.1 plasmids containing gene-specific or scrambled shRNA (control) selectively expressed in only a small subset of normal tissues were used in the analysis of BRM, HAT and HDAC (source: Open thereby reducing the chance of toxicity if inhibited and (2) highly Biosystems, Thermo Scientific, Lafayette, CO, USA). All shRNAs were overexpressed in cancer cells. For these reasons, HDAC9 appears obtained from Open Biosystems and the catalog number for each shRNA is to be a viable and ideal target for therapy. As such, it is anticipated listed in the Supplementary Table 4. For infection, cells were plated at that clinically targeting HDAC9 will lead to the effective restoration about 30–40% confluency in 24-well plates and infected 2–3 times at 20 Â concentration for 16 h each time. Subsequently, cells were puromy- of BRM and more importantly, pave the way for a novel avenue of cin-selected for 3 days and knockdown efficiency was determined by qPCR targeted therapy, gene reactivation. and western blot.

HDAC9 overexpression study MATERIALS AND METHODS For the analysis, cells were plated at B70–75% confluency in T75 flasks. Cell culture shRNA-mediated knockdowns of GATA3 or MEF2D were conducted (as All cells lines were obtained by American Type Culture Collection described above) in C33A and SW13 cell lines. HDAC9 was also transfected (Manassas, VA, USA) and were grown in RPMI media supplemented with into these cells, and after 72-h RNA collected and processed as described 5% fetal bovine serum, 1% glutamax and 1% pen/strep. Cells were tested above. QPCR was performed to determine the level of BRM, GATA3, MEF2D to be mycoplasma free, and the authenticity of the cell lines was confirmed and HDAC9 mRNA expression after and before GATA3 and/or MEF2D by sequencing p53 in each cell line and matched to published p53 data. shRNA knockdowns. For all pharmacological treatments, cells were plated in six-well plates at about 60% confluency and were treated with appropriate compounds with Nuclear run-on assay selected doses. Compounds used for the analysis were MGCD0103 (Selleck 61 Chemicals, Houston, TX, USA), MC1568 (Selleck Chemicals, Houston, TX, Nuclear run-ons were conducted as described previously. USA), RGFP963 (Repligen, Waltham, MA, USA), PD98059, PD169316 (Cayman Chemical, Ann Arbor, MI, USA). For RNA, cells were collected Immunohistochemical staining after 48 h, whereas for protein, cells were collected after 72 h of treatment Immunohistochemical staining was conducted as previously and processed accordingly. For the transfection experiments, cells were described.25,28,85 Briefly, formalin-fixed paraffin-embedded tissue blocks plated in a T75 flask at 60% confluency and transfected with appropriate (Providence Hospital, Southfield, MI, USA) were analyzed by immuno- plasmid(s) using Polyplus Jetprime Reagent (VWR, Radnor, PA, USA), as per histochemical staining. Antigen retrieval for BRM, HDAC9 and GATA3 were the manufacturer’s instructions. Cells from the transfections were performed using 10 mM Tris (pH 10), 10 mM sodium citrate buffer (pH 6.0), processed for both RNA and proteins after 72 h. and 10 mM Tris with 1 mM EDTA and 0.05% Tween 20 (pH 8), respectively following the manufacturer’s guidelines. Anti-BRM rabbit Ab is described in Glaros et al.25 and used at dilution of 1:500; rabbit polyclonal anti-HDAC9 Growth inhibition assay antibody (ab70954, Abcam, Cambridge, MA, USA) was used at a dilution of Cell lines were grown in six-well plates from a cell density of B10–15% 1:50; the mouse monoclonal GATA3 Ab was used (sc-268, Santa Cruz and allowed to grow up to B95% confluent. Cell lines under different Biotechnology) at a dilution of 1:50. A goat anti-rabbit or goat anti-mouse treatment regimens were counted daily for 5–7 days using an Accuri C6 biotinylated secondary was used with these primary antibodies at flow cytometer (BD Biosciences, San Jose, CA, USA). Cell viability was 1:200 (GE Healthcare). Sections were incubated with primary antibodies determined by Trypan blue staining from parallel wells. Manual counts for 2 h at room temperature, and with secondary antibodies for 1 h. We using a hemocytometer were done on day 1 and on the last day of each used an ABC staining kit with DAB/nickel detection reagent (BD experiment to verify the end point cell numbers. Pharmingen, San Diego, CA, USA). Slides were counterstained with Harris hematoxylin for 2 min. Western blots and qPCR Cell pellets were lysed using Trizol reagent84 and total mRNA isolated using CONFLICT OF INTEREST the Sigma RNA extraction kit (Sigma-Aldrich, St Louis, MO, USA), as per the D Reisman is the major shareholder in the company Zenagene. The other authors manufacturer’s instructions. Complementary DNA was generated from 1 mg declare no conflict of interest. of RNA using Bio-Rad iScript kit (Bio-Rad, Hercules, CA, USA), following the manufacturer’s instructions. QPCR was performed using RT Fast SYBR Green/ ROX qPCR Master Mix (SA Biosciences/Qiagen, Valencia, CA, USA) according ACKNOWLEDGEMENTS to the following conditions: 95 1C for 10 min, and 40 cycles of 95 1C for 10 s and 60 1C for 30 s. Primers used for qPCR are listed in Supplementary We thank Rachel Dresbeck for her time and effort in editing the manuscript. This Table 3. For western blot analysis, whole-cell lysates were extracted using research was supported by 5R01CA136683-03 Identifying Agents which Restored BRM expression. Urea buffer (8.8 M urea, 5 M NaH2PO4,1M Tris, pH 8.0) as described in Reyes et al.77 In all, 80 mg proteins were mixed with 6 Â Lamelli buffer and boiled for 10 min and then loaded in 4–15% Bio-Rad precast gel (Bio-Rad). Gels were run for 1 h at constant voltage of 150 V. Subsequently, proteins were REFERENCES transferred to a Millipore Immobilon-P membrane (Millipore, Billerica, MA, 1 Laurent BC, Treich I, Carlson M. Role of yeast SNF and SWI proteins in transcrip- USA). Proteins were transferred for 1 h at constant current of 350 mA. For tional activation. Cold Spring Harb Symp Quant Biol 1993; 58: 257–263. detection of BRM, a polyclonal anti-BRM antibody (1:500) was used.25 Anti- 2 Carlson M, Laurent BC. The SNF/SWI family of global transcriptional activators. rabbit acetylated-BRM antibody was a gift from Christian Muchardt, and was Curr Opin Cell Biol 1994; 6(3): 396–402. used at 1:500 dilution.31 Polyclonal anti-HDAC antibodies were obtained 3 Reisman D, Glaros S, Thompson EA. The SWI/SNF complex and cancer. Oncogene from Biovision (Milpitas, CA, USA) and were used at 1:500 dilutions. A 2009; 28(14): 1653–1668. monoclonal anti-GATA3 antibody was obtained from Santa Cruz 4 Gordon V, Rogers C, Reisman D. Alteration to the SWI/SNF complex in human Biotechnology (Santa Cruz, CA, USA) and was used at 1:300 dilutions. cancers. Oncol Rev 2010; 4:2. Appropriate secondary antibody (GE Healthcare, Little Chalfont, UK) was 5 Muchardt C, Yaniv M. ATP-dependent chromatin remodelling: SWI/SNF and Co. used at 1:2000 dilutions. Glyceraldehyde 3-phosphate dehydrogenase are on the job. J Mol Biol 1999; 293(2): 187–198. antibody (GeneTex Inc., Irvine, CA, USA) was used as the loading control. 6 Muchardt C, Yaniv M. The mammalian SWI/SNF complex and the control of cell Western blots were developed using ECL Prime western blot detection kit growth. Semin Cell Dev Biol 1999; 10(2): 189–195. (GE Healthcare). Western blots were analyzed using ImageJ (National 7 Muchardt C, Yaniv M. When the SWI/SNF complex remodels the cell cycle. Institutes of Health (NIH), Bethesda, MD, USA) software. H460 and SW13 Oncogene 2001; 20(24): 3067–3075. were used as positive and negative BRM expression controls, respectively, 8 Wang W, Chi T, Xue Y, Zhou S, Kuo A, Crabtree GR. Architectural DNA binding by a whereas the cell lines H661 and H441 (BRM-nonacetylated), and H460 and high-mobility-group/kinesin-like subunit in mammalian SWI/SNF-related H157 (-acetylated) were used as controls for BRM acetylation studies. complexes. Proc Natl Acad Sci USA 1998; 95(2): 492–498.

Oncogene (2014) 653 – 664 & 2014 Macmillan Publishers Limited Restoration and reactivation of BRM B Kahali et al 663 9 Wang W, Cote J, Xue Y, Zhou S, Khavari PA, Biggar SR et al. Purification and 34 Mizutani T, Ito T, Nishina M, Yamamichi N, Watanabe A, Iba H. Maintenance of biochemical heterogeneity of the mammalian SWI-SNF complex. EMBO J 1996; integrated proviral gene expression requires Brm, a catalytic subunit of SWI/SNF 15(19): 5370–5382. complex. J Biol Chem 2002; 277(18): 15859–15864. 10 Wang W, Xue Y, Zhou S, Kuo A, Cairns BR, Crabtree GR. Diversity and specialization 35 Gramling S, Rogers C, Liu G, Reisman D. Pharmacologic reversal of epigenetic of mammalian SWI/SNF complexes. Genes Dev 1996; 10(17): 2117–2130. silencing of the anticancer protein BRM: a novel targeted treatment strategy. 11 Muchardt C, Yaniv M. A human homologue of Saccharomyces cerevisiae SNF2/ Oncogene 2011; 30(29): 3289–3294. SWI2 and Drosophila brm genes potentiates transcriptional activation by the 36 Potthoff MJ, Olson EN.. MEF2: a central regulator of diverse developmental pro- glucocorticoid receptor. EMBO J 1993; 12(11): 4279–4290. grams. Development 2007; 134(23): 4131–4140. 12 Khavari PA, Peterson CL, Tamkun JW, Mendel DB, Crabtree GR. BRG1 contains a 37 Ozawa Y, Towatari M, Tsuzuki S, Hayakawa F, Maeda T, Miyata Y et al. Histone conserved domain of the SWI2/SNF2 family necessary for normal mitotic growth deacetylase 3 associates with and represses the transcription factor GATA-2. Blood and transcription. Nature 1993; 366(6451): 170–174. 2001; 98(7): 2116–2123. 13 Strobeck MW, Knudsen KE, Fribourg AF, DeCristofaro MF, Weissman BE, Imbalzano 38 Gregoire S, Xiao L, Nie J, Zhang X, Xu M, Li J et al. Histone deacetylase 3 interacts AN et al. BRG-1 is required for RB-mediated cell cycle arrest. Proc Natl Acad Sci USA with and deacetylates myocyte enhancer factor 2. Mol Cell Biol 2007; 27(4): 2000; 97(14): 7748–7753. 1280–1295. 14 Strobeck MW, DeCristofaro MF, Banine F, Weissman BE, Sherman LS, Knudsen ES. 39 Haberland M, Arnold MA, McAnally J, Phan D, Kim Y, Olson EN. Regulation of The BRG-1 Subunit of the SWI/SNF Complex Regulates CD44 Expression. J Biol HDAC9 gene expression by MEF2 establishes a negative-feedback loop in the Chem 2001; 276(12): 9273–9278. transcriptional circuitry of muscle differentiation. Mol Cell Biol 2007; 27(2): 15 Reisman DN, Strobeck MW, Betz BL, Sciariotta J, Funkhouser Jr. W, Murchardt C 518–525. et al. Concomitant down-regulation of BRM and BRG1 in human tumor cell lines: 40 Itoh T, Miyake K, Iijima S. Differentiation-specific expression of chromatin differential effects on RB-mediated growth arrest vs. CD44 expression. Oncogene remodeling factor BRM. Biochem Biophys Res Commun 2008; 366(3): 827–833. 2002; 21(8): 1196–1207. 41 Dwarakanath BS, Verma A, Bhatt AN, Parmar VS, Raj HG. Targeting protein 16 Strobeck MW, Reisman DN, Gunawardena RW, Betz BL, Angus SP, Knudsen KE et acetylation for improving cancer therapy. Indian J Med Res 2008; 128(1): 13–21. al. Compensation of BRG-1 function by Brm: insight into the role of the core SWI- 42 DelBove J, Rosson G, Strobeck M, Chen J, Archer TK, Wang W et al. Identification of SNF subunits in retinoblastoma tumor suppressor signaling. J Biol Chem 2002; a core member of the SWI/SNF complex, BAF155/SMARCC1, as a human tumor 277(7): 4782–4789. suppressor gene. Epigenetics 2011; 6(12): 1444–1453. 17 Lee JH, Lee JY, Chang SH, Kang MJ, Kwon H. Effects of Ser2 and Tyr6 mutants of 43 Marijon H, Andre F. Her2 a paradigm for targeted therapy. Bull Cancer 2011; 98(9): BAF53 on cell growth and p53-dependent transcription. Mol Cells 2005; 19(2): 1011–1017. 44 Tagliabue E, Balsari A, Campiglio M, Pupa SM. HER2 as a target for breast cancer 289–293. therapy. Expert Opin Biol Ther 2010; 10(5): 711–724. 18 Xu Y, Zhang J, Chen X. The activity of p53 is differentially regulated by Brm- and 45 de Castro-Carpeno J, Belda-Iniesta C, Casado Saenz E, Hernandez Agudo E, Feliu Brg1-containing SWI/SNF chromatin remodeling complexes. J Biol Chem 2007; Batlle J, Gonzalez Baron M. EGFR and colon cancer: a clinical view. Clin Transl 282(52): 37429–37435. Oncol 2008; 10(1): 6–13. 19 Oh J, Sohn DH, Ko M, Chung H, Jeon SH, Seong RH. BAF60a interacts with p53 to 46 Strober BE, Dunaief JL, Guha S, Goff SP. Functional interactions between the recruit the SWI/SNF complex. J Biol Chem 2008; 283(18): 11924–11934. hBRM/hBRG1 transcriptional activators and the pRB family of proteins. Mol Cell 20 Bultman S, Gebuhr T, Yee D, La Mantia C, Nicholson J, Gilliam A et al. A Brg1 null Biol 1996; 16(4): 1576–1583. mutation in the mouse reveals functional differences among mammalian 47 Dunaief JL, Strober BE, Guha S, Khavari PA, Alin K, Luban J et al. The retino- SWI/SNF complexes. Mol Cell 2000; 6(6): 1287–1295. blastoma protein and BRG1 form a complex and cooperate to induce cell cycle 21 Bultman SJ, Herschkowitz JI, Godfrey V, Gebuhr TC, Yaniv M, Perou CM et al. arrest. Cell 1994; 79(1): 119–130. Characterization of mammary tumors from Brg1 heterozygous mice. Oncogene 48 Zhang HS, Gavin M, Dahiya A, Postigo AA, Ma D, Luo RX et al. Exit from G1 and S 2008; 27(4): 460–468. phase of the cell cycle is regulated by repressor complexes containing HDAC-Rb- 22 Glaros S, Cirrincione GM, Palanca A, Metzger D, Reisman D. Targeted knockout of hSWI/SNF and Rb-hSWI/SNF. Cell 2000; 101(1): 79–89. BRG1 potentiates lung cancer development. Cancer Res 2008; 68(10): 3689–3696. 49 Nakada S, Kawano T, Saito-akita S, Iwase S, Horiguchi-Yamada J, Ohno T et al. MEK 23 Rodriguez-Nieto S, Canada A, Pros E, Pinto AI, Torres-Lanzas J, Lopez-Rios F et al. and p38MAPK inhibitors potentiate TNF-alpha induced apoptosis in U937 cells. Massive parallel DNA pyrosequencing analysis of the tumor suppressor BRG1/ Anticancer Res 2001; 21(1A): 167–171. SMARCA4 in lung primary tumors. Hum Mutat 2011; 32(2): E1999–E2017. 50 Haber DA, Bell DW, Sordella R, Kwak EL, Godin-Heymann N, Sharma SV et al. 24 Medina PP, Romero OA, Kohno T, Montuenga LM, Pio R, Yokota J et al. Frequent Molecular targeted therapy of lung cancer: EGFR mutations and response to EGFR BRG1/SMARCA4-inactivating mutations in human lung cancer cell lines. Hum inhibitors. Cold Spring Harb Symp Quant Biol 2005; 70: 419–426. Mutat 2008; 29(5): 617–622. 51 Brose MS, Volpe P, Feldman M, Kumar M, Rishi I, Gerrero R et al. BRAF and 25 Glaros S, Cirrincione GM, Muchardt C, Kleer CG, Michael CW, Reisman D. The RAS mutations in human lung cancer and melanoma. Cancer Res 2002; 62(23): reversible epigenetic silencing of BRM: implications for clinical targeted therapy. 6997–7000. Oncogene 2007; 26(49): 7058–7066. 52 Califano R, Landi L, Cappuzzo F. Prognostic and predictive value of K-RAS 26 Medina PP, Carretero J, Fraga MF, Esteller M, Sidransky D, Sanchez-Cespedes M. mutations in non-small cell lung cancer. Drugs 2012; 72(Suppl 1): 28–36. Genetic and epigenetic screening for gene alterations of the chromatin-remo- 53 Zhang X, Wharton W, Yuan Z, Tsai SC, Olashaw N, Seto E. Activation of the growth- deling factor, SMARCA4/BRG1, in lung tumors. Genes Chromosomes Cancer 2004; differentiation factor 11 gene by the histone deacetylase (HDAC) inhibitor tri- 41(2): 170–177. chostatin A and repression by HDAC3. Mol Cell Biol 2004; 24(12): 5106–5118. 27 Fukuoka J, Fujii T, Shih JH, Dracheva T, Meerzaman D, Player A et al. Chromatin 54 Muchardt C, Bourachot B, Reyes JC, Yaniv M. Ras transformation is associated with remodeling factors and BRM/BRG1 expression as prognostic indicators in decreased expression of the brm/SNF2alpha ATPase from the mammalian non-small cell lung cancer. Clin Cancer Res 2004; 10(13): 4314–4324. SWI-SNF complex. EMBO J 1998; 17(1): 223–231. 28 Reisman DN, Sciarrotta J, Wang W, Funkhouser WK, Weissman BE. Loss of BRG1/ 55 Berger AH, Knudson AG, Pandolfi PP. A continuum model for tumour suppression. BRM in human lung cancer cell lines and primary lung cancers: correlation with Nature 2011; 476(7359): 163–169. poor prognosis. Cancer Res 2003; 63(3): 560–566. 56 Bilanges B, Stokoe D. Mechanisms of translational deregulation in human tumors 29 Wilson BG, Roberts CW. SWI/SNF nucleosome remodellers and cancer. Nat Rev and therapeutic intervention strategies. Oncogene 2007; 26(41): 5973–5990. Cancer 2011; 11(7): 481–492. 57 Balch C, Montgomery JS, Paik HI, Kim S, Huang TH, Nephew KP. New anti-cancer 30 Bock VL, Lyons JG, Huang XX, Jones AM, McDonald LA, Scolyer RA et al. BRM and strategies: epigenetic therapies and biomarkers. Front Biosci 2005; 10: 1897–1931. BRG1 subunits of the SWI/SNF chromatin remodelling complex are down- 58 Tsou JA, Hagen JA, Carpenter CL, Laird-Offringa IA. DNA methylation analysis: a regulated upon progression of benign skin lesions into invasive tumours. Br J powerful new tool for lung cancer diagnosis. Oncogene 2002; 21(35): 5450–5461. Dermatol 2011; 164(6): 1221–1227. 59 Hess-Stumpp H. Histone deacetylase inhibitors and cancer: from cell biology to 31 Bourachot B, Yaniv M, Muchardt C. Growth inhibition by the mammalian SWI-SNF the clinic. Eur J Cell Biol 2005; 84(2-3): 109–121. subunit Brm is regulated by acetylation. EMBO J 2003; 22(24): 6505–6515. 60 Konofaos P, Kontzoglou K, Georgoulakis J, Megalopoulou T, Zoumpouli C, 32 Yamamichi N, Yamamichi-Nishina M, Mizutani T, Watanabe H, Minoguchi S, Christoni Z et al. The role of ThinPrep cytology in the evaluation of estrogen Kobayashi N et al. The Brm gene suppressed at the post-transcriptional level in and progesterone receptor content of breast tumors. Surg Oncol 2006; 15(4): various human cell lines is inducible by transient HDAC inhibitor treatment, which 257–266. exhibits antioncogenic potential. Oncogene 2005; 24(35): 5471–5481. 61 Liu G, Gramling S, Munoz D, Cheng D, Azad AK, Mirshams M et al. Two novel BRM 33 Gramling S, Reisman D. Discovery of BRM targeted therapies: novel reactivation of insertion promoter sequence variants are associated with loss of BRM expression an anticancer gene. Lett Drug Des Discov 2011; 8(1): 93–99. and lung cancer risk. Oncogene 2011; 30(29): 3295–3304.

& 2014 Macmillan Publishers Limited Oncogene (2014) 653 – 664 Restoration and reactivation of BRM B Kahali et al 664 62 Weichert W. HDAC expression and clinical prognosis in human malignancies. 74 Maneechotesuwan K, Xin Y, Ito K, Jazrawi E, Lee KY, Usmani OS et al. Regulation of Cancer Lett 2009; 280(2): 168–176. Th2 cytokine genes by p38 MAPK-mediated phosphorylation of GATA-3 63 Bertos NR, Wang AH, Yang XJ. Class II histone deacetylases: structure, function, J Immunol 2007; 178(4): 2491–2498. and regulation. Biochem cell Biol ¼ Biochimie et biologie cellulaire 2001; 79(3): 75 Ho IC, Pai SY. GATA-3—not just for Th2 cells anymore. Cell Mol Immunol 2007; 243–252. 4(1): 15–29. 64 Milde T, Oehme I, Korshunov A, Kopp-Schneider A, Remke M, Northcott P et al. 76 Sakurai K, Furukawa C, Haraguchi T, Inada K, Shiogama K, Tagawa T et al. HDAC5 and HDAC9 in medulloblastoma: novel markers for risk stratification and MicroRNAs miR-199a-5p and -3p target the Brm subunit of SWI/SNF to generate a role in tumor cell growth. Clin Cancer Res 2010; 16(12): 3240–3252. double-negative feedback loop in a variety of human cancers. Cancer Res 2011; 65 Skov V, Larsen TS, Thomassen M, Riley CH, Jensen MK, Bjerrum OW et al. Increased 71(5): 1680–1689. gene expression of histone deacetylases in patients with Philadelphia-negative 77 Reyes JC, Muchardt C, Yaniv M. Components of the human SWI/SNF complex are chronic myeloproliferative neoplasms. Leuk Lymphoma 2012; 53(1): 123–129. enriched in active chromatin and are associated with the nuclear matrix. J Cell Biol 66 Petrie K, Guidez F, Howell L, Healy L, Waxman S, Greaves M et al. The histone 1997; 137(2): 263–274. deacetylase 9 gene encodes multiple protein isoforms. J Biol Chem 2003; 278(18): 78 Mai A, Massa S, Rotili D, Cerbara I, Valente S, Pezzi R et al. Histone deacetylation in 16059–16072. epigenetics: an attractive target for anticancer therapy. Med Res Rev 2005; 25(3): 67 Moreno DA, Scrideli CA, Cortez MA, de Paula Queiroz R, Valera ET, da Silva Silveira 261–309. V et al. Differential expression of HDAC3, HDAC7 and HDAC9 is associated with 79 Moore SD, Herrick SR, Ince TA, Kleinman MS, Dal Cin P, Morton CC et al. Uterine prognosis and survival in childhood acute lymphoblastic leukaemia. Br J Haematol leiomyomata with t(10;17) disrupt the histone acetyltransferase MORF. Cancer Res 2010; 150(6): 665–673. 2004; 64(16): 5570–5577. 68 Duong V, Bret C, Altucci L, Mai A, Duraffourd C, Loubersac J et al. Specific activity 80 Avvakumov N, Cote J. The MYST family of histone acetyltransferases and their of class II histone deacetylases in human breast cancer cells. Mol Cancer Res 2008; intimate links to cancer. Oncogene 2007; 26(37): 5395–5407. 6(12): 1908–1919. 81 Glozak MA, Seto E. Histone deacetylases and cancer. Oncogene 2007; 26(37): 69 Gregoire S, Tremblay AM, Xiao L, Yang Q, Ma K, Nie J et al. Control of MEF2 5420–5432. transcriptional activity by coordinated phosphorylation and sumoylation. J Biol 82 Gervais C, Murati A, Helias C, Struski S, Eischen A, Lippert E et al. Acute myeloid Chem 2006; 281(7): 4423–4433. leukaemia with 8p11 (MYST3) rearrangement: an integrated cytologic, cytoge- 70 Ma K, Chan JK, Zhu G, Wu Z. Myocyte enhancer factor 2 acetylation by p300 netic and molecular study by the groupe francophone de cytogenetique hema- enhances its DNA binding activity, transcriptional activity, and myogenic tologique. Leukemia 2008; 22(8): 1567–1575. differentiation. Mol Cell Biol 2005; 25(9): 3575–3582. 83 Avvakumov N, Cote J. Functions of myst family histone acetyltransferases and 71 Albergaria A, Paredes J, Sousa B, Milanezi F, Carneiro V, Bastos J et al. Expression their link to disease. Subcell Biochem 2007; 41: 295–317. of FOXA1 and GATA-3 in breast cancer: the prognostic significance in hormone 84 Chomczynski P, Sacchi N. Single-step method of RNA isolation by acid receptor-negative tumours. Breast Cancer Res 2009; 11(3): R40. guanidinium thiocyanate-phenol-chloroform extraction. Anal Biochem 1987; 162(1): 72 Zheng R, Blobel GA. GATA transcription factors and cancer. Genes Cancer 2010; 156–159. 1(12): 1178–1188. 85 Reisman DN, Sciarrotta J, Bouldin TW, Weissman BE, Funkhouser WK. The 73 Lamonica JM, Vakoc CR, Blobel GA. Acetylation of GATA-1 is required for chro- expression of the SWI/SNF ATPase subunits BRG1 and BRM in normal human matin occupancy. Blood 2006; 108(12): 3736–3738. tissues. Appl Immunohistochem Mol Morphol 2005; 13(1): 66–74.

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

Oncogene (2014) 653 – 664 & 2014 Macmillan Publishers Limited