Mammalian SWI/SNF-related complexes are ATPase-powered nucleosome remodeling assemblies crucial for proper development and tissue-specific gene expression. The ATPase activity of the complexes is also critical for tumor suppression. The complexes contain seven or more noncatalytic subunits; only one of which, hSNF5/Ini1/BAF47, has been individually identified as a tumor suppressor thus far. The noncatalytic subunits include p270/ARID1A, which is of particular interest because tissue array analysis corroborated by screening of tumor cell lines indicates that p270 may be deficient in as many as 30% of renal carcinomas and 10% of breast carcinomas. The complexes can also include an alternative ARID1B subunit, which is closely related to p270, but the product of an independent gene. The respective importance of p270 and ARID1B in the control of cell proliferation was explored here using a short interfering RNA approach and a cell system that permits analysis of differentiation-associated cell cycle arrest. The p270-depleted cells fail to undergo normal cell cycle arrest on induction, as evidenced by continued synthesis of DNA. These lines fail to show other characteristics typical of arrested cells, including up-regulation of p21 and down-regulation of cyclins. The requirement for p270 is evident separately in both the up-regulation of p21 and the down-regulation of E2F-responsive products. In contrast, the ARID1B-depleted lines behaved like the parental cells in these assays. Thus, p270-containing complexes are functionally distinct from ARID1B-containing complexes. These results provide a direct biological basis to support the implication from tumor tissue screens that deficiency of p270 plays a causative role in carcinogenesis.
The ATPase-powered SWI/SNF chromatin remodeling complex in yeast regulates the mating type switch and other areas of specialized gene expression (reviewed in ref. 1). Mammalian SWI/SNF-related complexes likewise contain an ATPase-powered nucleosome remodeling activity associated with transcriptional regulation. The activity of the complexes is crucial for proper tissue-specific gene expression, development, and hormone responsiveness (reviewed in ref. 1). More recently, it has become apparent that these complexes also play critical roles in suppression of tumorigenesis in mice and humans (reviewed in ref. 2).
The complexes contain seven or more noncatalytic subunits that presumably help to modulate the targeting and activity of the ATPase. Mammalian complexes have variable compositions because some subunits occur as sets of related proteins. For example, there are two alternative ATPase subunits: mammalian BRM and BRG1. These are closely related proteins, but in mouse knockout studies only BRG1 proved essential for embryonic development and tumor suppression (3, 4). The ATPases are mutated in multiple human tumor cell lines and their loss correlates with poor prognosis of non–small cell lung cancers (5–7). Noncatalytic components of the complex may be important for tumor suppression as well; however, their individual roles are less well understood.
Among the noncatalytic subunits, hSNF5 (syns: INI1; BAF47) is recognized as a tumor suppressor in mice (8–10). In humans, hSNF5 is deficient in a high proportion of pediatric malignant rhabdoid tumors (e.g., refs. 11–13). Germ-line mutations have been identified, and carriers are predisposed to malignant rhabdoid tumors and tumors of the central nervous system (14–16).
Expression of functional BRG1 or hSNF5 is associated with specific aspects of cell cycle regulation. Expression of the cell cycle inhibitor p21CIP1/WAF1 has been repeatedly identified as BRG1 responsive, and several studies indicate that BRG1-dependent or hSNF5-dependent cell cycle arrest is enacted through a pRb-dependent or overlapping pathway (e.g., refs. 17–25). However, these effects have only been seen in the context of reintroduction of exogenous complex components into tumor cell lines where they were deficient. The significance of the complexes in the expression of these biological targets has yet to be shown during differentiation-associated cell cycle arrest, when the effects of complex dysfunction on carcinogenesis would be most significant.
Subunits required for the tumor suppression activity of the complexes have great potential as diagnostic and prognostic markers and as targets for drug therapy. Thus, a major question now is the distinction of which additional noncatalytic subunits are required for the cell cycle arrest functions of the complexes. The noncatalytic components of the complex include the p270 subunit [syns.: ARID1A, SMARCF1, BAF250a, hOSA1; refs. 26–28], which is a member of the ARID family of DNA binding proteins (reviewed in refs. 29, 30). The role of p270 in cell cycle regulation is of particular interest because recent results from a cDNA tissue array analysis and corroborating screens of panels of tumor cell lines indicate that p270 may be deficient in as many as 30% of renal carcinomas and 10% of breast carcinomas (28, 31, 32). A mutually exclusive alternative to p270 in the complexes is the ARID1B (syns: hOSA2, BAF250b) subunit, which is ∼50% identical to p270 across its entire length, but is the product of an independent gene. The ARID family proteins are determinants that distinguish key divisions among the multiple, distinct SWI/SNF-related complexes that exist in mammalian cells. A major distinction is between the complex first identified as the BRG1-associated factors (BAF) complex (also called the human SWI/SNF or hSWI/SNF complex) and a distinct complex designated PBAF. The BAF complex contains at least BRG1 (or BRM), p270 (or ARID1B), BAF170, BAF155, BAF60, BAF57, BAF53, actin, and hSNF5. The PBAF complex is characterized by the absence of both p270 and ARID1B and the presence of a 180 kDa protein designated Polybromo (syn: BAF180). Thus, p270 and ARID1B distinguish between the BAF and PBAF complexes, whereas BRG1 and hSNF5 do not (subunit composition of SWI/SNF-related complexes is reviewed in ref. 1). In addition to the BAF and PBAF division, the BAF series of complexes itself encompasses at least four different entities because p270 and ARID1B can each associate with mammalian BRM and BRG in all four possible combinations (28, 33), so that p270 and ARID1B each define a specific limited set among the various combinational permutations of SWI/SNF-related complexes that exist in mammalian cells.
The importance of p270 and ARID1B in proliferation control was explored here using a short interfering RNA (siRNA) approach. The knockdowns were constructed in the MC3T3-E1 preosteoblast line because these nontransformed cells undergo a tightly regulated and well-characterized progression through cell cycle arrest and into tissue-specific gene expression (e.g., refs. 34–39). This is an important model system because it permits an examination of the normal roles of the complex subunits during differentiation-associated cell cycle arrest. Parental MC3T3-E1 cells arrest by day 4 postinduction with the differentiation signal. The results described here show in parallel conditions that p270-depleted cells fail to arrest normally. This is evidenced by continued synthesis of DNA and by a lack of other characteristics typical of arrested cells, including up-regulation of p21, down-regulation of cyclins, and decreases in histone H4 expression and cdc2-specific kinase activity. The ARID1B-depleted lines behaved like the parental cells in each of these assays. The analysis of the respective roles of p270/ARID1A and ARID1B establishes a new paradigm that the choice of ARID-containing subunits confers specificity of function on the complexes. The specific requirement for p270 is evident separately in both the up-regulation of p21 and the down-regulation of E2F-responsive products. The clinical findings suggested indirectly that deficiency of p270 plays a causative role in carcinogenesis. The identification of specific proliferation control steps dependent on the presence of p270 now provides a direct molecular basis to support the clinical findings. Moreover, the demonstration that the complexes are required separately for regulation of at least two distinct steps in proliferation control underscores the carcinogenic potential of cells that have lost function of a required subunit.
Materials and Methods
Materials. Fetal bovine serum (FBS) was purchased from Summit Biotech (Fort Collins, CO), α-MEM from Irvine Scientific (Santa Ana, CA), and penicillin and streptomycin from Mediatech (Herndon, VA). Histone H1, ascorbic acid, β-glycerol phosphate, and protease inhibitors were obtained from Sigma Chemical Co. (St. Louis, MO), and G418 from Life Technologies, Inc. (Grand Island, NY). Radiochemicals were obtained from NEN (Boston, MA).
Short interfering RNA and isolation of stable p270 knockdown lines. The siRNA sequences were tested in a pSUPER vector constructed as described in ref. 40. Test oligonucleotides were synthesized as complementary pairs, each 64 bases long, containing two inversely repeated copies of a 19 bp target sequence separated by a 9 bp spacer region. Six different target sequences were tested in transient expression assays in 293T cells against an exogenously introduced p270 partial expression construct. In a similar manner, four different target sequences were tested against an ARID1B partial expression construct. Expression was monitored by Western blot. The most effective sequences were chosen for construction of the stable knockdown lines. In the p270-targeted construct, the 64 bp forward sequence is GATCCCCCTCATTGGTTTCACAAGTCTTCAAGAGAGACTTGTGAAACCAATGAGTTTTTGGAAA (the 19 bp target sequence is underlined). The respective 19 bp ARID1B target sequence is CTCTCTGGTTGCATCTGTC. The pSUPER-derived vectors containing the respective knockdown sequences (pSUPER.p270.7182 and pSUPER.ARID1B.5400) were introduced into MC3T3-E1 cells by lipofection together with a selectable neo marker. G418-resistant clones were amplified and screened by Western blot for p270 expression. Aliquots of low passage depleted lines were frozen as stocks. The target sequences for each construct were designed against nucleotide stretches that are identical between the mouse and human genes so that they can be used in cells of either species origin.
Cell culture. Low passage MC3T3-E1 cells were a gift from Roland Baron (Yale University, New Haven, CT). Cells were maintained in α-MEM plus 10% FBS supplemented with 50 units/mL penicillin and 50 μg/mL streptomycin. For differentiation assays, cells were plated at an approximate initial density of 5 × 104 cells/cm2. Differentiation was induced by addition of 50 μg/mL final concentration ascorbic acid and 10 mmol/L final concentration β-glycerol phosphate to standard growth medium. The medium was changed every 3 to 4 days and the inducing agents were replaced with each media change.
Immunoblotting. Cells were washed and harvested in PBS and lysed in p300 lysis buffer [0.1% Nonidet P-40, 250 mmol/L sodium chloride, 20 mmol/L sodium phosphate (pH 7.0), 30 mmol/L sodium pyrophosphate, 5 mmol/L DTT, and protease and phosphatase inhibitors: 0.1 mmol/L sodium vanadate, 1 mmol/L phenylmethylsulfonyl flouride (PMSF), 100 kIU aprotinin, 1 μg/mL leupeptin, and 1 μg/mL pepstatin]. Proteins were separated by PAGE, transferred to Immobilon-P membrane (Millipore, Billerica, MA), and visualized as described previously (41).
Radiolabeling and immunoprecipitation. Cells were washed with methionine-free and serum-free α-MEM and incubated with this medium for 1 hour. [35S]Methionine [200 μCi; Perkin-Elmer (Boston, MA) or Amersham (Piscataway, NJ)] was added to each 10 cm monolayer, and the plates were incubated for a further 3 hours. Cells were washed and harvested in PBS and lysed in p300 lysis buffer. Three milligrams of total cell lysate were precleared with 3% protein A-sepharose beads and immunoprecipitated as described previously (41).
Alkaline phosphatase assay. Cell monolayers were rinsed in PBS, fixed in 100% methanol, rinsed with PBS, then overlaid with 1.5 mL of 0.15 mg/mL 5-bromo-4-chloro-3-indolyl phosphate (Sigma) plus 0.3 mg/mL nitroblue tetrazolium (Promega, Madison, WI) for 30 minutes and rinsed again with PBS.
RNA analysis and probes. Total cell RNA was prepared at designated times postinduction using Trizol reagent (Life Technologies) according to the recommendations of the manufacturer. RNA in serial 10-fold dilutions (10, 1, and 0.1 μg) was applied to nitrocellulose BA85 in a slot-blotting apparatus (Schleicher & Schuell, Keene, NH) and cross-linked by UV irradiation. 32P-labeled probes were prepared using a random primed labeling kit (Boehringer-Mannheim, Indianapolis, IN). The histone H4 probe was described previously (35). Plasmid pGB.glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was constructed from MC3T3-E1 cell RNA by generating a reverse transcription-PCR fragment using primers (5′-ACTTTGTCAAGCTCATTTCC-3′ and 5′-TGCAGCGAACTTTATTGATG-3′) corresponding to the murine GAPDH cDNA sequence, and subcloning the resulting PCR fragment into the TA cloning vector, pCR2.1 (Invitrogen, Carlsbad, CA).
Antibodies. The p270-specific monoclonal antibody (mAb) PSG3 and the ARID1B-specific mAb KMN1 have been described previously (33). A peptide used to generate a BAF155-specific mAb, DXD7 (33), also gave rise to a distinct BAF155-reactive mAb, designated DXD12, which cross-reacts with BAF170, and was used here. An SV40 tag-specific mAb, mAb 419 (obtained from Ed Harlow, Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Cambridge, MA), was used as a negative control. Commercially purchased antibodies include the p21CIP1/WAF1/SDI-specific antibody (BD Biosciences, San Jose, CA) and the hsc70-specific antibody (Stressgen, San Diego, CA), as well as antibodies of the following specificities obtained from Santa Cruz Biotechnology (Santa Cruz, CA): cyclin A (C-19), cyclin B2 (N-20), and cyclin C (H-184). Rabbit polyclonal serum was raised against the cdc2-G6 peptide sequence CDNQIKKM.
Kinase assays. The cdc2-dependent kinase assays were done as described previously (42) using cdc2-specific immunoprecipitation complexes from 1 mg of total cell lysate and histone H1 as exogenous substrate.
DNA synthesis assay. Induced cells were labeled with [3H]thymidine (5 μCi/mL of culture medium; Perkin-Elmer) in 1-hour pulses at the times postinduction indicated in the text, lysed in 0.3 mol/L NaOH, and assayed for trichloracetic acid–precipitable counts as described previously (43).
Virus infection. The generation and culture of the E1A-inactivated 9S adenovirus (used here as a negative control) has been described previously (44). A stock of p21 expression virus (Ad5CMVp21; ref. 45) was provided by Judit Garriga (Fels Institute, Temple University School of Medicine, Philadelphia, PA). MC3T3-E1 cells were infected at a multiplicity of infection of 25 plaque-forming units per cell.
Generation of p270-deficient and ARID1B-deficient MC3T3-E1–derived cell lines. Potential interfering oligonucleotide sequences were tested in a pSUPER-derived system by standard protocols. Effective sequences were identified and introduced by stable integration from the plasmid vector into the MC3T3-E1 line. Depletion was monitored by Western blotting with p270-specific and ARID1B-specific mAbs. Ten independent lines with reduced expression of each target were selected, amplified, and stored. As a control, vector-only lines were selected and amplified in parallel. In each transfection, colonies appeared at similar frequencies and showed essentially the same doubling time in normal growth medium as parental cells. A representative p270 knockdown line (MC.p270.KD.AA2) and an ARID1B knockdown line (MC.1B.KD.CA6B) are each shown in Fig. 1A. The depleted lines (lanes 3 and 6) show weak p270 or ARID1B signals, respectively, in comparison with the parental line or a clonal line isolated after transfection with the empty vector (lanes 1, 2, 4, and 5). The blots were additionally probed with a mAb that recognizes the closely related BAF155 and BAF170 mammalian SWI/SNF complex subunits. Expression of these subunits is similar in each line. The overall integrity of the complexes in the p270-depleted cells was verified by immunoprecipitation of 35S-labeled cell lysates with the BAF155-reactive antibody (Fig. 1B). Maintenance of expression of the alternative ARID family protein in the conversely depleted cells was confirmed by immunoprecipitation with p270-specific and ARID1B-specific mAbs (Fig. 1C). All selected lines showed a similar degree of depletion. [35S]Methionine pulse labeling indicates that new synthesis of each protein is reduced about 10-fold in the knockdown lines. As a further probe for the overall integrity of the complexes in the knockdown lines, the remaining ARID family product was removed by immune depletion of the cell lysates before immunoprecipitation with the BAF155-reactive antibody (Fig. 1D). The results confirm that complex assembly is stable in the absence of either subunit.
p270-depleted and ARID1B-depleted cells both show impaired induction of the tissue-specific marker alkaline phosphatase. MC3T3-E1 cells continue to proliferate for several days after induction with ascorbic acid, then undergo cell cycle arrest at about day 3 postinduction. Expression of the earliest differentiation marker, alkaline phosphatase, can be detected at this time. The knockdown lines were tested for induction of alkaline phosphatase in an in situ enzyme assay scored by color development (positive cells stain purple-black). Three independent knockdown lines from each series were tested, and all showed severe impairment of alkaline phosphatase induction; vector-only lines behaved like the parental line. In multiple independent experiments, both series of knockdown lines showed severely reduced induction of alkaline phosphatase at every point tested, both early and late. Representative experiments at day 3 and day 14 are shown in Fig. 2. These results indicate that the level of depletion achieved for each target is functionally significant, and that both ARID-containing products are required for normal onset of differentiation. We have used these knockdown lines to explore the role of each protein specifically in cell cycle arrest functions.
p270-deficient cells fail to undergo normal cell cycle arrest. The effect of p270 deficiency versus ARID1B deficiency was tested on specific cell cycle arrest functions. We have previously used a gene array approach to identify many of the changes in gene expression that occur in MC3T3-E1 cells as they proceed through the differentiation program (35). Expression was assayed on the arrays at days 0, 3, 7, and later times postinduction. Between day 0 and day 7, a number of changes occurred that corresponded with the shutdown of cell cycle activity. Among the most prominent was induction of the cell cycle inhibitor p21Cip1/Waf1. Severalfold induction of p21 was apparent by day 3 postinduction. When this response was tested here at the protein level, a similar pattern of induction was apparent in the parental line, but p21 expression was not induced in p270-depleted cells. A representative Western blot depicting results from the MC.p270.KD.CA6 line is shown in Fig. 3A. The same pattern was seen with the MC.p270.KD.AA2 and MC.p270.KD.DD2 lines (not shown). In contrast to the p270-depleted cell lines, the ARID1B-depleted lines showed no impairment of p21 induction. Results from a representative line (MC.1B.KD.FD2) are shown in Fig. 3A; the same result was observed with the MC.1B.KD.CA6B and MC.1B.KD.JD6 lines. As a loading control, the blots were also probed with an antibody reactive against the constitutive form of the 70 kDa heat shock protein, hsc70. The hsc70 signal was similar in all lanes.
The gene array also indicated down-regulation of several cyclins as the MC3T3-E1 cells enter growth arrest. These responses were probed here with the MC.p270.KD.CA6 line and the MC.1B.KD.FD2 line (Fig. 3B). The array showed down-regulation of B-type cyclins, particularly cyclin B2, by day 3 postinduction. Consistent with the RNA signals, a decreased level of cyclin B2 was apparent in the parental cells by day 4 on the Western blot. However, p270-depleted cells again failed to show the parental response; levels of cyclin B2 remained high. Cyclin C was also sharply down-regulated by day 3 in the gene array probe. Consistent with the RNA signal, down-regulation of cyclin C in the parental cells was clear by day 2 in the Western blot. However, cyclin C levels were unaffected by the induction protocol in the p270-depleted cells. The gene array also indicated down-regulation of cyclin A, but this response was delayed relative to the response patterns of cyclins B and C. On the array, a decreased cyclin A signal was apparent at day 7 postinduction, but not at day 3. The protein probe shows a slight decrease in the cyclin A level in the parental cells by day 6, and no detectable decrease in the p270-depleted cells. The cdc2 kinase gene was not represented on the array, but expression of this gene is of interest as a known E2F target subject to pRb-mediated repression (reviewed in ref. 46). The cdc2 protein product was assayed by Western blotting (Fig. 3B), which shows a sharp decrease in protein levels in the parental cells by day 4 postinduction, and no detectable decrease in p270-depleted cells by day 6. In each of these assays, the ARID1B-depleted cells behaved indistinguishably from the parental cells.
A well-characterized marker of proliferation state in differentiating osteoblasts is histone H4 expression (reviewed in ref. 39). Expression of this marker declines dramatically as the cells arrest after induction. This response was also compared in parental MC3T3-E1 and the knockdown cells. Consistent with other markers of cell cycle activity, the histone H4 signal decreased sharply in the parental cells and the ARID1B knockdown line, but remained high in the p270 knockdown line (Fig. 4).
The gene expression patterns that accompany induction to the differentiation phenotype in parental MC3T3-E1 cells imply that a sharp decline in cyclin-dependent kinase activity would occur by day 4. The activity of cdc2-associated complexes was assayed here directly. The kinase activity shows a sharp decline by day 4 postinduction in the parental cells and the ARID1B-depleted cells, but no decline was detectable in the p270-depleted cells, even at day 6. A representative kinase assay done with the MC.p270.KD.DD2 line and the MC.1B.KD.FD2 line is shown in Fig. 5A. Results from independent experiments with three different p270-depleted lines (MC.p270.KD.AA2, MC.p270.KD.CA6, and MC.p270.KD.DD2) and three different ARID1B-depleted lines (MC.1B.KD.FD2, MC.1B.KD.JD6, and MC.1B.KD.CA6B) were quantified on a phosphoimager and the averages are shown graphically in Fig. 5B.
The cell cycle status of the cells was probed directly by assessing the rate of [3H]thymidine incorporation over several days postinduction. Parental cells, p270-depleted cells, and ARID1B-depleted cells were plated and induced in parallel. At 24-hour intervals, [3H]thymidine was added to the culture medium for 1 hour, after which the labeled cells were harvested and assayed for incorporation of the isotope. The results shown for the parental cells are the averages of three independent platings. The results shown for the p270-depleted cells are the averages from three different knockdown lines (MC.p270.KD.AA2, MC.p270.KD.CA6, and MC.p270.KD.DD2), as are the results shown for the ARID1B-depleted lines (MC.1B.KD.FD2, MC.1B.KD.JD6, and MC.1B.KD.CA6B). The parental cells show a sharp decline in the rate of [3H]thymidine incorporation by day 4 postinduction, indicating a shutdown of DNA synthesis consistent with cell cycle arrest. The same pattern is seen in the ARID1B-depleted lines. In contrast, the rate of [3H]thymidine incorporation decreases only slightly in the p270-depleted cells, indicating continued DNA synthesis and failure to undergo normal cell cycle arrest (Fig. 5C).
These results identify the p270 subunit as critical for normal cell cycle arrest. In each of the cell cycle assays described here, cloned cell lines containing a functional ARID1B-targeted siRNA sequence behaved exactly like the parental line, indicating that the failure of p270-depleted cells to undergo a normal cell cycle arrest response is a specific effect of p270 deficiency.
Induction of p21 and repression of E2F-responsive promoters are independent events that each require p270. Previous studies have used differential expression of BRG1 and hSNF5 to probe the role of SWI/SNF-related complexes in cell cycle arrest. Several studies, often using cloned rather than endogenous promoters, found that BRG1 enhances pRb-mediated repression of E2F-responsive genes, and suggested that SWI/SNF subunits are associated with pRb in repressor complexes (23, 25, 47, 48). Decreased expression of endogenous E2F-responsive gene products such as cyclin A and cdc2 was generally apparent at the protein level when BRG1 expression was restored to naturally deficient tumor cell lines (20, 23), but in another study the effects were modest or cell line specific, at least at the RNA level (19). An upstream effect with the potential to activate pRb was seen consistently; exogenous expression of BRG1 in BRG1-deficient tumor cell lines results in a sharp increase in p21 expression with most other cell cycle inhibitors, including p16ink4a, remaining relatively unaffected (19, 20). Expression of hSNF5 in malignant rhabdoid tumor lines has likewise been linked with decreased levels of E2F-responsive gene products such as cyclin A (21, 22, 49). Restoring hSNF5 to malignant rhabdoid tumor cells does not alter p21 expression, but up-regulates the p16ink4a cell cycle inhibitor (17, 22, 46), in contrast to the effect of BRG1. However, the difference may be a function of cell type rather than subunit effect because siRNA-mediated depletion of hSNF5 in HeLa (cervical carcinoma) or MG63 (osteosarcoma) cells causes a sharp decrease in p21 levels, with no effect on p16 (20).
The failure of p270-depleted cells to induce p21 could theoretically account for all of the cell cycle arrest defects observed in these lines. Repression of E2F-responsive genes and the downstream effects of this repression might be impaired indirectly if cyclin-dependent kinase activity is not appropriately inhibited, leaving targets, such as pRb, phosphorylated, inactive, and unable to mediate repression of E2F-responsive genes. The effect of the complexes on the p21 promoter is apparently direct, as two studies have shown the presence of BRG1 at the p21 promoter, although a target element in the promoter could not be established (19, 20). A key mechanistic question that remains unclear is whether down-regulation of E2F-responsive genes requires the action of the complexes independently of their effects on p21 levels. Chromatin association assays have limited usefulness for these studies because the complexes associate with chromatin widely. This question was therefore addressed here genetically by introducing exogenous expression of p21 in the p270-depleted cells simultaneously with the differentiation signal.
Parental and p270-depleted cells were infected in parallel at the time of ascorbic acid induction with an adenovirus vector expressing p21 or with a negative control virus containing an inactivated E1A gene. The cells were monitored for DNA synthesis as described above. Parental cells that received the p21 expression construct underwent accelerated shutdown of DNA synthesis; [3H]thymidine incorporation was severely repressed by day 2 postinduction (Fig. 6A). Uninfected cells, or cells infected with the negative control virus, showed the same kinetics seen in Fig. 6 (i.e., without exogenous p21, DNA synthesis in the parental line remained high at day 2); a severe decrease was not seen until day 4. In the p270-depleted cells without exogenous expression of p21, DNA synthesis remained high at least until day 6. However, exogenous expression of p21 caused DNA synthesis to shut down with the same rapid kinetics seen in the parental cells (Fig. 6B; exogenous expression of p21 was verified by Western blotting, shown in the left-hand lanes of Fig. 6C; normal induction of p21 in the parental cells and the failure of p21 induction in the p270 depleted cells can be seen in the right-hand, 9S-infected, control lanes). The rapid down-regulation of DNA synthesis associated with exogenous p21 expression was expected even in the p270-depleted cells because p21-induced inhibition of cyclin dependent kinase activity is expected to result in the inactivation of essential DNA replication factors. What is of special interest here is the status of E2F-responsive products. This was monitored by Western blotting for the representative E2F-responsive gene products cdc2, cyclin A, and cyclin B2. Expression of cyclin C was also examined (Fig. 6C). The results show that expression levels of cdc2 and the cyclins remain high in the p270-depleted cells despite exogenous expression of p21, indicating that regulation at the p21 promoter and at the E2F-responsive promoters each independently requires the function of the chromatin remodeling complexes, and of p270 specifically, during differentiation-associated cell cycle arrest. The expected repression of the cdc2-associated kinase activity in p21-expressing p270-knockdown cells, despite the maintenance of a high level of cdc2 expression, was verified in a kinase assay (Fig. 6D).
The work described here identifies the p270 subunit of mammalian SWI/SNF-related complexes as critical for normal cell cycle arrest in differentiating cells exiting the cell cycle. The evidence of p270 deficiency in certain tumors and tumor cell lines (32) implied indirectly that p270 plays a required role in the tumor suppressor activity of the complex(es). The results presented here establish a specific biological basis for the clinical findings, demonstrating directly that p270 is essential for both the induction of p21 and the repression of E2F-responsive genes, such as cdc2, during differentiation-associated cell cycle arrest. Involvement in both activation and repression is a feature of SWI/SNF-related complexes generally, presumably determined by the spectrum of transactivators and repressors that recruit the complexes. The demonstration that the complexes are required separately for regulation of at least two distinct steps in proliferation control underscores the carcinogenic potential of cells that have lost function of a required subunit.
These results are particularly significant because previous studies about the roles of SWI/SNF complex components in expression of cell cycle markers have largely relied on reintroduction of BRG1 or hSNF5 into tumor cell lines where they were lacking (e.g., refs. 17, 19, 20, 22, 23, 25, 49), rather than monitoring the role of complex components in cells undergoing physiologic progression from a proliferative state to cell cycle arrest. The identification of p270 as a subunit required for cell cycle arrest in vivo is additionally significant because, unlike BRG1 and hSNF5, p270 is not among those subunits considered to form the “functional core” of the complex(es). The concept of a functional core was based on the observation that BRG1 has a relatively low level of enzyme activity when purified from other subunits, and that a level of remodeling and ATPase activity similar to that of the intact complex(es) can be reconstituted in vitro by assembly of a subset of components consisting of BRG1, BAF170, BAF155, and hSNF5 (50). The in vivo requirement for p270 shows that it plays an essential role in the physiologic functions of the complex(es), regardless of whether it contributes directly to the overall level of enzymatic activity.
A cell cycle arrest function has previously been ascribed to the BRG1 and hSNF5 components; however, those subunits do not distinguish between the BAF and PBAF complexes. The p270-depletion phenotype constitutes the first formal evidence of a requirement for the BAF complex series, as opposed to the PBAF complex, in cell cycle regulation. The data presented here shed further light on the question of specificity among the several distinct configurations of the BAF complexes. The alternative ATPase subunits, BRM and BRG1, may be partially redundant in their ability to support cell cycle arrest when exogenously expressed (51), but the most physiologic experiments (3, 4) suggest strongly that BRG1-specific complexes and not BRM-specific complexes are essential for this function. The present study indicates that p270-containing complexes, but not ARID1B-containing complexes, are required for cell cycle arrest. Logically, it seems that, of the four combinations made possible by these alternative subunits, it is the BRG1 and p270 combination that plays the major role in cell cycle arrest. hSNF5 is not a determinant of specificity between complexes, but its presence along with BRG1 and p270 is required for the activities that the complex contributes to cell cycle arrest. These findings will help to clarify targets for drug intervention therapies.
The essential biochemical activities contributed by the noncatalytic subunits have yet to be determined. The amino acid sequence of hSNF5 gives little clue to the function of the protein, which remains unknown, except that it can facilitate DNA end-joining in vitro (52). In p270, the most obvious structural motif is the ∼100-amino-acid-long ARID DNA binding domain, but this large protein (2,285 amino acids) also contains potential protein-protein interaction surfaces (26–28, 41) that may be more important for its specific function. ARID1B contains an ARID domain that is 80% identical with p270 (alignments can be seen in refs. 30, 53), and both proteins belong to a subclass of the ARID family that binds DNA without regard to sequence specificity (33, 53). Thus, a likely scenario is that DNA binding is a function common to both, whereas sequences outside the ARID determine specificity of function. Structure-function analysis of both these ARID-containing subunits is in progress.
The effects linked individually with BRG1, hSNF5, and p270 are generally similar but not identical. What is not clear is whether differences thus far reported are due more to methodology or cell type than to true differences in function between the subunits. Approaches based on reintroduction of specific components into deficient tumor cell lines preclude the ability to compare function within a single cell line or in nontransformed cells. The ability to knock down expression is freeing investigators to study the role of the complexes in nontransformed cell lines. BRG1 and BRM can be inhibited by dominant-negative forms with an inactivating mutation in the ATP binding site. Dominant-negative inhibition of BRG1/BRM in NIH 3T3 mouse fibroblasts inhibited MyoD-dependent differentiation, but surprisingly did not inhibit concomitant cell cycle arrest; p21 is induced in these conditions and its induction was likewise unaffected by expression of the dominant-negative construct (54). However, cell cycle arrest independent of SWI/SNF complex activity may be a phenomenon specific to the function of MyoD because a similar dominant-negative approach in BALB/c mouse fibroblasts was sufficient to inhibit CCAAT enhancer-binding protein α (C/EBPα)–induced cell cycle arrest severely (55). In the latter study, siRNA-mediated depletion of hSNF5 or BRM had the same effect on cell growth curves as expression of the dominant-negative construct, but expression of individual genes was not assessed.
The system used here maintains the benefits of probing function in nontransformed cells, and has the additional advantage of relying on an external induction signal to initiate differentiation and cell cycle arrest functions rather than engineered overexpression of a single gene such as MyoD or C/EBPα. The MC3T3-E1 cell system is amenable to knockdown studies targeting each of a succession of subunits and probing an array of extracellular signals. Further studies are in progress.
Note: Peter B. Dallas is currently at the Institute for Child Health Research and the Center for Child Health Research, University of Western Australia, Subiaco, Western Australia 6008. George R. Beck, Jr. is currently at the Division of Endocrinology, Metabolism, and Lipids, Emory University School of Medicine, Atlanta, GA 30322.
Grant support: Public Health Service grant RO1CA53592 (E. Moran), a Shared Resources grant to the Fels Institute (1R24CA88261), Department of Defense Breast Cancer Research Program fellowships DAMD-17-02-1-0577 (N.G. Nagl, Jr.) and DAMD-17-02-1-0578 (A. Patsialou), and a Daniel Swern fellowship from Temple University (A. Patsialou).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Peter Yaciuk, Walter Long, and Michael Van Scoy for help with hybridoma development and culture, and Xavier Graña-Amat, Judit Garriga, and E. Premkumar Reddy for gifts of reagents. We are also grateful to Barbara Hoffman, Scott Shore, and members of our lab for advice and critical comments.