TFII-I was initially identified as the general transcription factor that binds to initiator (Inr) elements in vitro. Subsequent studies have shown that TFII-I activates transcription of various genes either through Inr elements or through other upstream elements in vivo. Since, however, most studies so far on TFII-I have been limited to over-expression and reporter gene assays, we reevaluated the role of TFII-I in vivo by using stable knockdown with siRNA and by examining the expression of endogenous genes. Contrary to the widely accepted view, here we show that TFII-I is not important for cell viability in general but rather inhibits the growth of MCF-7 human breast cancer cells. MCF-7 cells are known to proliferate in an estrogen-dependent manner. Through analysis of TFII-I's cell-type specific growth inhibitory effect, we show evidence that TFII-I down-regulates a subset of estrogen-responsive genes, only those containing Inr elements, by recruiting estrogen receptor (ER) α and corepressors to these promoters. Thus, this study has revealed an unexpected new role of TFII-I as a negative regulator of transcription and cell proliferation
TFII-I was initially identified as one of the general transcription factors, the factor that binds to Inr elements in vitro (Roy et al. 1991). Inr elements are located on transcription initiation sites and according to an estimate, are present on 85% of all the human genes (Suzuki et al. 2001). Subsequent studies have shown that TFII-I activates transcription of various genes either through Inr elements or through other upstream elements in vivo (Grueneberg et al. 1997; Kim et al. 1998; Morikawa et al. 2000; Parker et al. 2001; Jackson et al. 2005; Ku et al. 2005). For example, TFII-I binds to the Inr element of the T cell receptor Vβ gene and activates its transcription in reporter gene assays (Cheriyath et al. 1998). TFII-I also activates transcription of c-fos and Goosecoid through binding to the serum response element and the distal element, respectively (Grueneberg et al. 1997; Ku et al. 2005). In addition, the human gene for TFII-I is located on 7q11.23, whose deletion is associated with Williams Syndrome, a rare, congenital disorder characterized by mental and physical problems including impulsive and outgoing personality, hyperacusis, developmental delay, heart and blood vessel problems and elfin-like face (Perez et al. 1998). Thus, TFII-I is thought to play very important roles in supporting transcription of a wide variety of genes (Roy 2000).
Estrogens, such as 17β-oestradiol (E2), are hormones that are synthesized predominantly in ovaries and play pivotal roles in the physiology of the female reproductive tract and in the homeostasis of other tissues (Nilsson et al. 2001). While estrogens are protective against pathologies such as osteoporosis, Alzheimer's disease, and cardiovascular disease, they are involved in the development of breast, ovarian, and endometrial cancers (Bian et al. 2001; Jordan et al. 2001; Honjo et al. 2003). The effects of estrogens are mediated through the estrogen receptors ERα and ERβ, which belong to the nuclear receptor superfamily and function as ligand-dependent transcriptional regulators (Robinson-Rechavi et al. 2003; NR committee 1999). Estrogen binding to the receptor's ligand binding domain induces a dynamic conformational change that leads to ER dimerization and association with coactivator or corepressor proteins, and ultimately causes transcriptional activation or repression of estrogen-responsive genes (Glass & Rosenfeld 2000; Klinge 2000; McKenna & O'Malley 2002a; Belandia & Parker 2003). Estrogen agonists, such as E2, induce sequential recruitment of coactivators including steroid receptor coactivator-1 (SRC-1), the histone acetyl transferase CBP/p300, SWI/SNF, and the Mediator, whereas antagonists, such as tamoxifene, trigger recruitment of corepressors including nuclear receptor corepressor (N-CoR)/silencing mediator of retinoid and thyroid receptor (SMRT) and histone deacetylases (HDACs) (Shang et al. 2000; Metivier et al. 2003). Unliganded ERα is generally considered to be transcriptionally inert (Metivier et al. 2004).
Most studies so far on TFII-I have been limited to over-expression and reporter gene assays. Therefore, we reevaluated the role of TFII-I in vivo by using stable and efficient knockdown with siRNA and by examining the expression of endogenous genes. Here we show that contrary to the widely accepted view, TFII-I is not important for cell viability in general but rather inhibits the growth of MCF-7 human breast cancer cells. Through analysis of its cell-type specific growth inhibitory effect, we show evidence that TFII-I down-regulates estrogen-responsive genes containing Inr elements by recruiting ERα and corepressors to these promoters. Thus, this study has revealed an unexpected new role of TFII-I as a negative regulator of transcription and cell proliferation.
TFII-I is a negative regulator of MCF-7 cell growth
To study the cellular function of TFII-I, we knocked down its expression in HeLa human cervical carcinoma cells and MCF-7 human breast cancer cells using siRNA. Western blot analysis shows that the protein level of TFII-I is reduced > 90% after siRNA delivery (Fig. 1A,B). Contrary to TFII-I being broadly important for transcription, its knockdown had negligible effect on HeLa cell growth (Fig. 1A). HeLa cells expressing a very low level of TFII-I continued to proliferate normally for several weeks (data not shown). In MCF-7 cells, which proliferate in an estrogen-dependent manner, TFII-I knockdown instead enhanced cell growth regardless of the addition of E2 (Fig. 1B). FACS analysis indicates that TFII-I knockdown increases the fraction of MCF-7 cells in S phase (Fig. 1C). As an alternative test, we over-expressed FLAG-tagged TFII-I together with the hygromycin resistance gene in MCF-7 cells and counted the number of hygromycin-resistant colonies after antibiotic selection. Compared to the control, TFII-I over-expression significantly reduced the number of hygromycin-resistant colonies (Fig. 1D). Taken together, these results demonstrate that TFII-I is not important for cell viability in general but rather inhibits the growth of MCF-7 cells.
TFII-I represses transcription of estrogen-responsive genes through Inr elements
Considering that MCF-7 cell growth is dependent on estrogen, TFII-I may exert its growth-related function through regulation of estrogen-responsive genes. To test this idea, we quantified, using real-time RT-PCR, mRNA levels of several endogenous genes in MCF-7 cells various times after addition of E2 or vehicle. Among the genes examined, pS2, cyclin D1, GREB1 and amphiregulin are estrogen-responsive, whereas β-actin and HSP70 are not. Each of the pS2, cyclin D1, amphiregulin and β-actin promoters contains a canonical Inr element on the transcription start site, whereas the GREB1 and HSP70 promoters do not. As shown in Fig. 2A, TFII-I knockdown more or less increased mRNA levels of the Inr-containing pS2, cyclin D1 and amphiregulin genes at both basal and E2-stimulated states, whereas it had negligible effect on transcription of the Inr-lacking GREB1 gene and the estrogen-independent β-actin and HSP70 genes. In HeLa cells, these genes were not significantly affected by TFII-I knockdown (Fig. 2B and data not shown). These results indicate that TFII-I plays a rather specific role in estrogen/ER-dependent transcription, negatively regulating estrogen-responsive genes containing Inr elements.
To test the specificity of the siRNA used, we designed two additional siRNAs against TFII-I. When introduced, these siRNAs effectively suppressed TFII-I expression on Western blot with a concomitant increase in growth rate and pS2 expression at the RNA level (data not shown). Thus, the observed effects are caused by no other than TFII-I deficit.
TFII-I physically interacts with ERα in an estrogen-independent manner
The above results prompted us to investigate whether TFII-I and ERα physically interact with each other. To this end, human embryonic kidney 293FT cells were co-transfected with FLAG-TFII-I and either vitamin D receptor (VDR), androgen receptor (AR), retinoid X receptor (RXR) α or ERα, and then FLAG-TFII-I was immunoprecipitated with anti-FLAG M2 antibody. Among the nuclear receptors used, only ERα was co-immunoprecipitated with FLAG-TFII-I, irrespective of the presence of its ligand (Fig. 3A,B). We tested ERα mutants lacking either the N-terminal A/B domain or the C-terminal E/F domain and found that the A/B domain is dispensable for ERα binding to TFII-I (Fig. 3B). We also attempted to map the TFII-I's region involved in the interaction. The N- and C-terminal parts of TFII-I are involved in DNA binding and transcriptional activation, respectively, and there are six 89 amino acid repeats over its entire length. Mutational analysis indicates that either the central region encompassing aa 300-780 or the C-terminal region encompassing aa 750-957 is sufficient for its binding to ERα (Fig. 3C). GST pull-down assays using recombinant proteins were also carried out, and essentially the same results were obtained (data not shown), indicating that the TFII-I-ERα interaction occurs directly.
TFII-I recruits ERα and corepressors to the pS2 promoter and inhibits Pol II's association with the promoter
To gain insight into the mechanism by which TFII-I down-regulates estrogen-responsive genes, we performed chromatin immunoprecipitation analysis for the Inr-containing pS2 promoter and the Inr-lacking GREB1 promoter in MCF-7 cells. The β-actin promoter was also examined as a control. As expected, a substantial amount of TFII-I was found associated with the pS2 promoter region, but not with the GREB1 promoter region, regardless of the presence of E2, and their association was abrogated by TFII-I knockdown (Fig. 4). In accordance with a previous report (Metivier et al. 2004), small amounts of ERα were associated with these promoters before induction, and more ERα was recruited after induction. As for pol II, there were only background levels of association with these promoters prior to induction, and the association increased significantly upon induction. Remarkably, TFII-I knockdown led to a significant decrease in ERα and to a concomitant increase in pol II associated with the pS2 promoter region before induction. After induction, TFII-I knockdown had a modest effect on the amount of ERα associated with the pS2 promoter region. On the other hand, the knockdown did not detectably affect ERα and pol II's association with the GREB1 promoter region. As for β-actin, consistent with the presence of Inr, TFII-I and pol II were constitutively associated with the promoter region. TFII-I knockdown significantly reduced the occupancy of TFII-I but had little effect on pol II, consistent with the observation that TFII-I knockdown had negligible effect on the β-actin mRNA level (Fig. 2). A possible explanation for the negative effect of TFII-I on pS2 transcription is that TFII-I may recruit ERα together with corepressors. Indeed, SMRT and HDAC1, components of the corepressor complex, were found associated with the pS2 promoter region before induction, and their association was significantly reduced after induction. As expected, TFII-I knockdown also decreased their association before and after induction. These results indicate that TFII-I down-regulates a subset of estrogen-responsive genes at least in part by recruiting ERα and associated corepressors to these promoters.
TFII-I has been considered to play important roles in supporting transcription of a number of genes (Johansson et al. 1995; Cheriyath et al. 1998; Kim et al. 1998; Wu & Patterson 1999; Morikawa et al. 2000; Parker et al. 2001; Ku et al. 2005). Contrary to the widely accepted view, here we showed that TFII-I is not important for cell viability in general but rather inhibits the growth of MCF-7 cells by down-regulating a subset of estrogen-responsive genes. Indeed, there are a few reports consistent with the idea that TFII-I represses transcription. For example, TFII-I has been identified as an interactor of HDAC1, 2, and 3 (Tussie-Luna et al. 2002; Hakimi et al. 2003; Wen et al. 2003). Such multiple functions of TFII-I may be exerted through its interactions with other transcriptional regulators. Here we showed that TFII-I directly interacts with ERα. In cases where TFII-I is involved in transcriptional activation, TFII-I is reported to interact with such transcriptional regulators as SRF (in the case of c-fos) and Smad2 (in the case of Goosecoid) (Grueneberg et al. 1997; Ku et al. 2005).
It has generally been considered that unliganded ERα is transcriptionally inert, becoming associated with coactivator and corepressor proteins only in the presence of estrogen agonists and antagonists, respectively (Jepsen & Rosenfeld 2002; McKenna & O'Malley 2002b). Recently, however, Metivier et al. (2004) reported that ERα46, an alternative splice variant of ERα lacking the N-terminal A/B domain, is capable of recruiting corepressor proteins to the pS2 promoter and causing transcriptional repression in the absence of any ligand. ERα46 is expressed in various cell types including MCF-7 cells (Flouriot et al. 2000; Denger et al. 2001). It remains to be determined whether ERα46 is associated with the pS2 promoter in MCF-7 cells, since antibody specific to the shorter isoform does not exist. In light of the finding that TFII-I interacts with an ERα mutant lacking the A/B domain (Fig. 3), however, it seems likely that ERα46 is recruited to the pS2 promoter via TFII-I prior to induction. This idea is consistent with the finding that SMRT and HDAC1 are associated with the pS2 promoter (Fig. 4).
We have shown that transcription of the Inr-containing pS2 and amphiregulin genes is regulated positively and negatively by estrogen and TFII-I. As shown in Fig. 2A, TFII-I knockdown and E2 treatment, independently and in combination, increased the pS2 and amphiregulin mRNA levels. Remarkably, the HDAC1 occupancy at the pS2 promoter region was inversely correlated with its mRNA level (Fig. 4). These findings indicate: (i) that TFII-I represses transcription of a subset of estrogen-responsive genes before induction; (ii) that HDAC1 is involved in this repression; and (iii) that this repression is partially, but not completely, relieved by E2. From these findings, we propose the following model (Fig. 5). Inr-bound TFII-I recruits ERα (possibly ERα46) and corepressors to the promoter and thereby represses transcription through histone deacetylation. This recruitment may involve TFII-I's interactions with ERα and possibly with HDACs. The multiprotein complex on the promoter may be partially persistent after E2 induction, as TFII-I-dependent transcriptional repression is observed both before and after induction (Fig. 2). An alternative, but not mutually exclusive, model posits that Inr-bound TFII-I physically prevents pol II's association with the promoter. Since TFIID and pol II have also been shown to interact with Inr (Carcamo et al. 1991; Kaufmann & Smale 1994), TFII-I may block preinitiation complex assembly through competitive binding to Inr. After induction, perhaps the combinatorial action of SRC-1, p300/CBP, SWI/SNF, and the Mediator among others may prevail over TFII-I, facilitating preinitiation complex assembly. Of these models, we prefer the first because it accounts for the cell-type specific role of TFII-I in repressing estrogen-responsive genes (Fig. 2); assuming only the second model, the TFII-I-mediated repression would be seen in HeLa cells.
Here we showed that ERα is recruited to the pS2 promoter in both TFII-I/Inr-dependent and -independent manners (Fig. 3). The latter may involve the transcription factor FoxA1, which is shown to be important for ERα recruitment to a number of estrogen-responsive gene promoters (Carroll et al. 2005). While ERα is capable of binding to an ERE-containing DNA fragment in a ligand-independent fashion in vitro (Medici et al. 1991), recent studies have established that it undergoes a characteristic dissociation association cycle on a target gene promoter in vivo (Shang et al. 2000; Metivier et al. 2003). Proteasome-mediated degradation of ERα has also been implicated in this process (Reid et al. 2003; Tateishi et al. 2004). TFII-I may constitute an element of the circuitry controlling the cyclic action of ERα.
Of note, there are some similarities between TFII-I and BRCA1, the product of the tumor suppressor gene that is frequently mutated in breast and ovarian cancers (Easton et al. 1993; Martin & Weber 2000). First, suppression of BRCA1 expression in breast or ovarian cancer cells leads to estrogen-independent overgrowth and to estrogen-independent transcriptional up-regulation of estrogen-responsive genes (Thompson et al. 1995; Fan et al. 1999; Zheng et al. 2001). Second, over-expression of BRCA1 causes growth retardation in various cell types including MCF-7 cells (Holt et al. 1996; Aprelikova et al. 1999). Third, BRCA1 directly interacts with ERα (Ma et al. 2005). Considering that BRCA1 is expressed normally in MCF-7 cells, BRCA1 and TFII-I may play similar but non-redundant roles in the regulation of ERα. In vivo, TFII-I may down-regulate basal level transcription of estrogen-responsive genes to allow their appropriate response to varying concentrations of estrogens, and possibly to prevent unwanted growth of estrogen-dependent cells or tissues.
To prepare the mammalian expression vector pHygTFII-I, the open reading frame for hexahistidine-tagged TFII-I was isolated from pET11d-II-I (Roy et al. 1997) and inserted into pCMV-Tag2A (Stratagene). Flag-TFII-I was then amplified by PCR and subcloned into pHygEF2, which carries the human EF1α promoter and the hygromycin resistance gene. For generation of TFII-I deletion mutants, nucleotide regions corresponding to amino acids (aa) 1-780, 1-355, 300-957, and 751-957 were PCR-amplified from pCMV-Tag2A-TFII-I, and the PCR products were cloned into pCMV-Tag2B. To construct pcDNA-ERα, the open reading frame for ERα was isolated from pSG5-ERα and inserted into pcDNA3.1(+) (Invitrogen). For its deletion mutants, nucleotide sequences encompassing aa 1-282 and 178-595 were PCR-amplified from pcDNA-ERα and cloned into pcDNA-Myc.
HeLa, 293FT (Invitrogen), and MCF-7 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS), 100 µg/mL penicillin, and 100 µg/mL streptomycin. For counting, cells were seeded at a density of 1 × 104 cells/mL in phenol-red free DMEM containing 5% charcoal-dextran-stripped FBS, with or without 100 nm E2 (Wako), and harvested every day. The cell numbers were counted using the live cell counting reagent SF (Nacalai Tesque).
Knockdown using siRNA
TFII-I knockdown has been done either by transfecting chemically synthesized siRNA or by infecting lentiviral vector expressing short-hairpin RNA, and essentially the same results have been obtained. Most of the data in this paper were obtained using lentiviral vectors, except those shown in Fig. 1B. For transfection, synthetic siRNA was introduced into cells using Lipofectamine 2000 (Invitrogen). For infection, recombinant lentiviruses were first produced in 293FT cells by co-transfecting ViraPower packaging mix (Invitrogen) and one of pLenti-based plasmid vectors carrying the mouse U6 promoter and an oligonucleotide cassette encoding short-hairpin RNA. Then, HeLa or MCF-7 cells were infected and selected as recommended by the manufacturer (Invitrogen). The following siRNA sequences were used: 5′-AAGTTACTCAGCCAAGAACGA-3′ for TFII-I and 5′-GAACGGCATCAAGGTGAAC-3′ for EGFP.
Total RNA was isolated using Sepazol (Nacalai Tesque) according to the manufacturer's instruction. cDNA was made with Superscript III reverse transcriptase (Invitrogen) and oligo-dT primer, and then real-time PCR was carried out using SYBR Green real-time PCR master mix (Toyobo) and ABI PRISM 5700 sequence detection system (Applied Biosystems). The following primers were used: 5′-GAATGGCCACCATGGAGAACAAGG-3′ and 5′-GCGGATCCACGAACGGTGTCGTCGAA-3′ for pS2, 5′-GCTGCTGTACCTCTGTGACTC-3′ and 5′-GTCTTCCTCAGATGTGTCGTC-3′ for GREB1, 5′-TGAAGGAGAAGCTGAGGAACG-3′ and 5′-CACTGGAAAGACCGACTC-3′ for amphiregulin, 5′-AGTGCAAGGCCTGAACCTGAGGAG-3′ and 5′-TCAGATGTCCACGTCCCGCACGTC-3′ for cyclin D1, 5′-TCCTTCCTGGGCATGGAGTCC-3′ and 5′-GAGGAGCAATGATCTTGATCTTC-3′ for β-actin, and 5′-ACAGTGGTGCCTACCAAGAAG-3′ and 5′-CTTGTCTTCAGCTGTCACTCG-3′ for HSP70.
Chromatin immunoprecipitation was carried out according to the protocol provided by Upstate Biotechnology. MCF-7 cells were plated on 15 cm plates and grown for at least five days in phenol red-free DMEM containing 5% charcoal-dextran-stripped FBS. Either 100 nm E2 or ethanol was added 45 min before harvest. Cells were crosslinked with 1% formaldehyde at 37 °C for 10 min. Immunoprecipitation was performed overnight using one of the following antibodies: anti-RNA polymerase II (pol II) (8WG16, Covance), anti-ERα (H-184, Santa Cruz), anti-TFII-I (Bethyl Laboratory), anti-HDAC1 (Upstate Biotechnology), or anti-SMRT (Abcam). Co-precipitated DNA was purified, and genomic sequences of interest were amplified by conventional PCR or real-time PCR. Conventional PCR was carried out using Taq DNA polymerase (Toyobo) for 30 cycles, empirically determined conditions that achieve semiquantitative measurement. The following primers were used: 5′-GGCCATCTCTCACTATGAATCACTTCTGC-3′ and 5′-GGCAGGCTCTGTTTGCTTAAAGAAGCG-3′ for the pS2 promoter region, 5′-TGAGCATTTGTGGATTTTGGCATC-3′ and 5′-CTGGAAACAGGGAAAGAAGGAAGG-3′ for the pS2 upstream region, 5′-GAGGGTGCAGTATGAGCAAAG-3′ and 5′-CAGAGGAGAGCCCTTCCTATG-3′ for the GREB1 promoter region, 5′-AGCAGTGAAAAAAAGTGTGGCAAC-3′ and 5′-CGACCCACAGAAATGAAAAGGCAG-3′ for the GREB1 estrogen-responsive element (ERE) region, and 5′-GCCAAAACTCTCCCTCCTCCT-3′ and 5′-GCTCGAGCCATAAAGGCAAC-3′ for the β-actin promoter region.
We thank Robert Roeder for the TFII-I cDNA and helpful discussions. We also thank Ko Sakabe for MCF-7 cells. This work was supported in part by a Grant-in-Aid for Scientific Research on Priority Areas and by the Grant of the 21st Century COE Program from the Ministry of Education, Culture, Sports, Science and Technology of Japan. This work was also supported by a grant from the New Energy and Industrial Technology Development Organization, Japan.