The N-terminal domain of human TAFII68 displays ... - Nature

ã

Oncogene (1999) 18, 8000 ± 8010 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc

The N-terminal domain of human TAFII68 displays transactivation and oncogenic properties Anne Bertolotti1,2, Brendan Bell1 and LaÁszloÁ Tora*,1 1

Institut de GeÂneÂtique et de Biologie MoleÂculaire et Cellulaire, CNRS/INSERM/ULP, BP 163 - 67404 Illkirch Cedex, CU de Strasbourg, France

In Ewing tumor, the (11;22) chromosomal translocation produces a chimeric molecule composed of the aminoterminal domain of EWS fused to the carboxyl-terminal DNA-binding domain of FLI-1. Previously, we have identi®ed a novel protein TAFII68, which is highly similar to EWS and another closely related protein TLS (also called FUS). We demonstrate that the N-terminus of TAFII68 eciently stimulates transcription when fused to two di€erent DNA binding domains and that overexpression of TAFII68-FLI-1 chimeras in NIH3T3 cells leads to oncogenic transformation. We have also investigated the molecular mechanisms which could account for the transcriptional activation and the oncogenic transformation potential of the N-termini of TAFII68 and EWS. Thus, we have tested whether the arti®cial recruitment of components of the preinitiation complex (PIC) or a histone acetyltransferase (HAT) could bypass the requirement for the activation domains of either EWS or TAFII68. Recruitment of individual components of the transcription machinery or the GCN5 HAT is not sucient to promote activation from FLI-1 responsive genes either in transfection experiments or in oncogenic transformation assays. These results suggest that the TAFII68 or EWS activation domains enhance a step after PIC formation in the transcriptional activation process. Keywords: oncogene; sarcoma; transcription factors II; histone acetylation; RNA polymerase II; recruitment

Introduction The genes encoding EWS and TLS are often the targets of chromosomal translocation in human sarcomas. EWS was originally discovered as being rearranged in Ewing sarcoma (Delattre et al., 1992). It has since been found to be implicated in several di€erent translocation events. These various somatic genetic rearrangements involving the gene encoding EWS result in the fusion of its N-terminal coding region to the C-terminal DNA binding domain (DBD) of one of several transcription factors including members of the ETS family like FLI1, ERG, ETV1, E1AF and FEV or other transcription factors including ATF-1, WT1 and the orphan receptor *Correspondence: L Tora 2 Current address: Department of Medicine, Cell Biology, The Skirball Institute of Biomolecular Medicine, NYU Medical Center, New York, NY 10016, USA Received 13 May 1999; revised 27 August 1999; accepted 9 September 1999

TEC1 (for review see Rauscher, 1997; May and Denny, 1997). In human myxoid liposarcoma, a chromosomal translocation event fuses the N-terminal domain of TLS to the transcription factor CHOP (Crozat et al., 1993; Rabbitts et al., 1993) and in acute myeloid leukemia to ERG (Panagopoulos et al., 1994). EWS and TLS are homologous proteins which are involved in strikingly similar translocation events giving rise to the production of structurally similar oncoproteins. EWS- and TLS-derived fusion proteins contain the amino-terminus provided by one of the two homologous proteins fused to the DBD of one amongst several transcription factors. Cellular transformation by EWS-derived and TLS-derived fusion oncogenes is dependent on both the N-terminal sequence contributed by TLS or EWS and on the integrity of the DNA binding domain of the translocation partner (May et al., 1993a; Zinszner et al., 1994). The N-terminal domains of the oncoproteins provided by TLS or EWS are potent transactivators when fused to heterologous DBDs (May et al., 1993b; Sanchez-Garcia and Rabbitts, 1994; Bailly et al., 1994; Zinszner et al., 1994; Lessnick et al., 1995). It is thus believed that deregulation of target genes contacted by the DNA binding component of the oncogenic chimeras is a common mechanism involved in the transformation process (Brown et al., 1995; Braun et al., 1995; Bailly et al., 1994; Karnieli et al., 1996; Fujimura et al., 1996; Zinszner et al., 1994). We have previously identi®ed a novel protein, TAFII68 which shares extensive sequence similarity with the human sarcoma-associated proteins EWS and TLS (also known as FUS) (Bertolotti et al., 1996). Like EWS and TLS, hTAFII68 contains a consensus RNA binding domain (RNP-CS) which allows it to bind not only RNA, but also single stranded (ss) DNA (Bertolotti et al., 1996). The three closely related human proteins have a particularly conserved RNAbinding motif that deviates from the organization of such domains commonly found in other RNA-binding proteins. The most striking particularity of the RNPCS motifs of hTAFII68, EWS, and TLS is the presence of an unusually long predicted loop immediately after the ®rst a helix (Nagai et al., 1990; Burd and Dreyfuss, 1994; Bertolotti et al., 1996). Thus, hTAFII68, EWS, and TLS all belong to a new subfamily of RNP-CS containing proteins that we have called the TET family (Bertolotti et al., 1996). The common structural features which are limited to the TET family members suggest that they bind RNA and/or ssDNA in a unique way. TAFII68 was identi®ed as a subunit of a speci®c population of the TFIID complex. TFIID is a multiprotein complex composed of the TATA binding

TAFII68 displays oncogenic properties A Bertolotti et al

protein (TBP) and TBP associated factors (TAFIIs) and its binding to the promoter is the ®rst step of preinitiation complex assembly on protein coding genes (Roeder, 1996; Bell and Tora, 1999). We have shown that di€erent TFIID complexes exist in human HeLa cells and that they exhibit functional speci®city (Brou et al., 1993a,b; Jacq et al., 1994; Bell and Tora, 1999). TAFII68 is also associated with another multiprotein complex, the human RNA polymerase II (Pol II) and is able to enter into the preinitiation complex together with Pol II (Bertolotti et al., 1996). This suggests TAFII68 may be a bridging factor between transcription initiation and elongation. We have recently demonstrated that the structural resemblance of the TET family members can be extended to certain functional similarities. Indeed, TLS, EWS are also associated with TFIID, but interestingly, they are each present in a distinct TFIID population (Bertolotti et al., 1996, 1998). Moreover, like TAFII68, EWS is also associated with Pol II (Bertolotti et al., 1998). Together, these results suggests that the TET proteins family members share common functional properties but they exhibit subtle di€erences. Since the TET family members are associated with two complexes involved in transcription of proteins coding genes, TFIID and RNA Pol II, they may participate in transcriptional regulation of class II genes. This hypothesis is reinforced by the observation that a Drosophila protein, homologous to the TET family members [termed Cabeza (Stolow and Haynes, 1995) or SARFH (Immanuel et al., 1995)], was found to be associated with the majority of active transcription units in preparations of polythene chromosomes from salivary gland nuclei (Immanuel et al., 1995). TLS seems to be involved in regulation of more than one aspect of gene expression. TLS is identical to hnRNP-P2, a subunit of a complex involved in maturation of RNA precursors (Calvio et al., 1995) and like many hnRNP proteins, TLS can shuttle between the nucleus and the cytoplasm (Zinszner et al., 1994, 1997). Thus, TLS could enter into the preinitiation complex (PIC) with TFIID, later associate with the elongating RNA polymerase II and play a role in mRNA processing. These properties are similar to those of the cleavage-polyadenylation speci®city factor (CPSF), which was recently shown to be brought to the PIC by TFIID (Dantonel et al., 1997). Moreover, CPSF is also associated with the CTD of RNA Pol II (McCracken et al., 1997; Hirose and Manley, 1998) a feature shared with certain splicing factors (Mortillaro et al., 1996; Yuryev et al., 1996). Consistent with these observations, TLS, and probably also TAFII68 and EWS could play a role in coupling transcription and mRNA processing. Since TAFII68, EWS and TLS share many common characteristics, we hypothesized that the N-terminus of hTAFII68 could also have transactivation and transforming properties. To address this question we have analysed the ability of hTAFII68 N-terminus to activate transcription, as well as its ability to transform NIH3T3 cells when fused to the DBD of FLI-1. During the transcriptional activation process, activators are thought to stimulate and/or facilitate the formation of PIC. Numerous biochemical analyses have suggested that activation domains can directly or indirectly contact components of the basal transcrip-

tion machinery (Tjian and Maniatis, 1994; Ranish and Hahn, 1996). As a consequence, these interactions recruit, stabilize and also induce conformational changes of the PIC (Goodrich et al., 1996; Roberts and Green, 1994; Barberis and Gaudreau, 1998). Transcriptional regulation by activators also involves modi®cation of chromatin by recruiting chromatin remodelling complexes to promoters to de-compact the chromatin structure around the genes which will be transcribed (Kuo and Allis, 1998; Workman and Kingston, 1998). In agreement with these results, it has been shown that activation domains can interact with histone acetyl transferases (HATs) (Workman and Kingston, 1998). As a consequence, these interactions open the chromatin structure around the promoter and therefore, facilitate the recruitment of the preinitiation complex to the initiation site (Roberts and Green, 1994; Chi and Carey, 1996; Wang et al., 1992; Chrivia et al., 1993; Workman and Kingston, 1998). Recently, it was shown that arti®cial recruitment of one of various components of the basal transcription machinery or HAT is sucient to stimulate transcription (Chatterjee and Struhl, 1995; Xiao et al., 1995; Klages and Strubin, 1995; Apone et al., 1996; Barberis et al., 1995; Gonzalez-Couto et al., 1997; Himmelfarb et al., 1990; Farrell et al., 1996; Xiao et al., 1997; MartinezBalbas et al., 1998; Barberis and Gaudreau, 1998). These ®ndings suggest that overcoming the rate limiting step of preinitiation complex assembly can be sucient to promote gene activation. Here we tested whether the arti®cial recruitment of components of the transcription machinery or a HAT could bypass the requirement for the activation domains of either EWS or TAFII68 on FLI-1 responsive promoters.

Results The amino-terminus of TAFII68 contains a transactivation domain To analyse whether the amino-terminus of TAFII68 contains a transactivation domain, we constructed two chimeric proteins: GAL4-TAFII68DC198 and GAL4TAFII68DC328 (Figure 1a). GAL4-TAFII68DC198 is composed of the N-terminal 198 amino acids of TAFII68 fused to the heterologous DBD of the yeast GAL4 activator, and the longer chimera (GAL4TAFII68DC328) contains the N-terminal 328 amino acids of TAFII68, including the ®rst RGG repeat containing region and the RNP-CS motif, fused to the GAL4 DBD (Figure 1a). As a positive control, we also constructed a GAL4-EWS chimera (Figure 1a). These constructs were transfected into HeLa cells together with the 17M2-G.CAT reporter gene (Webster et al., 1988b) containing two GAL4 binding sites upstream of the rabbit b-globin promoter fused to the CAT reporter gene (schematized in Figure 2a). As expected transcription of the CAT gene was eciently stimulated by GAL4-EWS when compared to the activation obtained by the transfection of the empty vector (Figure 2a, compare 12 to lane 1). GAL4TAFII68DC198 activated expression of the CAT reporter gene in a similar dose-dependant manner as GAL4-EWS however, with a lower magnitude (Figure 2a, compare lanes 16 ± 22 to lanes 9 ± 15). In contrast,

8001

TAFII68 displays oncogenic properties A Bertolotti et al

8002

surprisingly the longer TAFII68 chimera (GAL4TAFII68DC328) did not activate transcription, probably due to the presence of the RNP-CS motif in the this fusion protein, which may mask the activation domain (Bertolotti et al., 1998). Overexpression of the negative control GAL4 DBD does not stimulate transcription of the same reporter gene. Moreover, a maximum stimulation was obtained by transfecting 100 ng of each expression vector which gave rise to approximately 50-fold activation by GAL4-EWS (Figure 2a, lane 12) and 24-fold GAL4TAFII68DC198 (Figure 2a, lane 19). When more than 100 ng of the di€erent expression vectors have been transfected, transcriptional activation decreased (Figure 2a, lanes 13 ± 15 and 20 ± 22). This e€ect can be interpreted as transcriptional interference or squelching that could re¯ect the titration of a limiting factor(s) required in the transactivation process by an activator (Gill and Ptashne, 1988; Tasset et al., 1990). Thus, the N-terminal 198 amino acids region of TAFII68 contains a transcriptional activation domain, similarly to the homologous domains of EWS and TLS. EWS-FLI-1 has been shown to activate transcription from various reporter genes under the control of di€erent ETS-responsive promoters (Bailly et al., 1994; Lessnick et al., 1995; Ohno et al., 1993). We

next tested whether the amino-terminus of TAFII68 can also activate transcription from ETS responsive promoters. Therefore, we generated two di€erent TAFII68-FLI-1 fusion proteins. By alignment of TAFII68 and EWS proteins and their genomic sequences (data not shown), we found that the ®rst 159 amino acids of TAFII68 correspond to the domain of EWS present in the type II EWS-FLI-1 fusion proteins (Delattre et al., 1992). Therefore, we generated the TAFII68DC159-FLI-1 fusion (Figure 1). As the ®rst 198 amino acids of TAFII68 were found to activate transcription eciently when fused to GAL4 on the 17M2-G.CAT reporter gene we generated the TAFII68DC198-FLI-1 chimera as well (Figure 1). In contrast, as the the longer TAFII68 chimera did not activate transcription when fused to the GAL4 DBD, we did not further analyse this chimera. ETS proteins have been shown to activate transcription of the HTLV-1 long terminal repeat (LTR) through binding to two ETS binding sites (EBS) contained in this promoter region (Gegonne et al., 1993; Bosselut et al., 1990). Moreover, it has been shown that EWS-FLI-1 can activate transcription from a reporter gene containing two tandem copies of the ETS responsive element of the HTLV-1 LTR (Bailly et al., 1994). Similarly, transcription from a reporter gene composed of the HTLV-1 LTR (schematized in Figure 2b) fused to the CAT gene (Brady et al., 1987) is activated by overexpression of EWS-FLI-1 (Figure 2b, lane 4). Both TAFII68-FLI-1 chimeras are able to stimulate expression of the CAT protein. The TAFII68DC198-FLI-1 chimera gives rise to approximately the same stimulation level as EWS-FLI-1 (Figure 2b, compare lane 8 to 4) whereas the shorter fusion, TAFII68DC159-FLI-1, is a less potent transactivator (Figure 2b, lane 6). The full length FLI-1 protein stimulates weakly if at all, transcription of the reporter gene (Figure 2b). Importantly, an inhibition of transcription was observed when more than 20 ng of expression vector was transfected in this system (Figure 2b, compare lanes 3, 5, 7, 9 to lanes 2, 4, 6, 8, respectively). Together, these results demonstrate that TAFII68-FLI1 fusion proteins are functionally similar to EWS-FLI1 in their ability to stimulate the expression of an ETS responsive gene. Moreover, similarly to EWS-FLI-1, TAFII68-FLI-1 fusion proteins are more potent transcriptional activators than FLI-1. Since the amino-terminus of TAFII68, without the RGG repeats and the RNP-CS motif, activates transcription when fused to two di€erent DBDs [independently of its position (Figures 1 and 2)], we conclude that this domain contains an autonomous transcription activation domain. TAFII68-FLI-1 chimeras transform NIH3T3 cells

Figure 1 Schematic representation of EWS, TAFII68, FLI-1 and the fusion proteins: (a) The di€erent domains of EWS and TAFII68 are indicated. The amino-terminal domains of each of the two proteins was fused to the DNA binding domain of GAL4. (b) FLI-1 and FLI-1 derived fusion proteins containing the ETS DNA binding domain of FLI-1 (DBD) are represented. The numbers refer to amino acid position in each protein

EWS-FLI-1 is thought to exert its oncogenic activity by deregulating expression of FLI-1 target genes (Brown et al., 1995; Braun et al., 1995; Bailly et al., 1994; Karnieli et al., 1996; Fujimura et al., 1996; Zinszner et al., 1994). Moreover, it has previously been shown that strong activation domains, like the acidic activation domain of VP16, when fused to FLI-1 can transform NIH3T3 cells (Lessnick et al., 1995). Thus, a strong correlation exists between the strength of an activation domain and its oncogenic

TAFII68 displays oncogenic properties A Bertolotti et al

8003

Figure 2 The aminoterminus of TAFII68 contains a transcriptional activation domain. (a) HeLa cells were transfected with increasing amount (5, 20, 50, 100, 200, 500 ng and 1 mg) of the indicated expression vectors or an empty vector pSG5 (7) together with 1 mg of the 17 M2-G-CAT reporter plasmid schematized at the top of the ®gure, containing two GAL4 binding sites (17M2). The phosphoimager quanti®cation of the CAT assays are presented as histograms in term of fold stimulation. (b) TAFII68-FLI-1 is a transcriptional activator on an EBS-containing promoter. HeLa cells were transfected with 20 or 100 ng of the indicated expression vectors together with 1 mg of the 10-1 reporter plasmid (Brady et al., 1987), containing the HTLV-1 LTR, schematized at the top of the ®gure. Hatched boxes represent the two ETS binding sites (EBS) of the HTLV-1 promoter. Histogram depicting the quantitation of CAT activity is presented. In (a) and (b) each histogram represents the mean of at least three independent transfections after normalization for the internal control luciferase activity

activity. Since the two shorter TAFII68 chimeras (containing the ®rst 159 or 198 amino acids of TAFII68) can activate transcription, in contrast to the the longer GAL4-TAFII68DC328 chimera (Figure 2a, b), we next investigated whether the two shorter TAFII68-FLI-1 fusions display oncogenic properties like EWS-FLI-1. Therefore, we constructed retroviral expression vectors for chimeric proteins in which the C-terminal domain of FLI-1, present in the oncoproteins from Ewing tumor, was fused to the N-terminal transactivation domain of TAFII68 (Figure 1b). As positive and negative controls, EWS-FLI-1 and the full length FLI-1 were inserted into the same retroviral vector. Nuclear extracts were prepared from the infected and polyclonal selected NIH3T3 cells and analysed by Western blot using the anti-FLI-1 monoclonal antibody mAb 7.3 (Melot et

al., 1997). FLI-1, EWS-FLI-1 and the two TAFII68FLI-1 chimeras are expressed and localized in the nucleus (Figure 3, lanes 1 ± 4 and data not shown). Note that all fusion proteins migrate much slower than their predicted molecular weight suggesting that they are the targets of posttranslational modi®cations. When plated in soft agar, both TAFII68DC159FLI-1 and TAFII68DC198-FLI-1 expressing cells were able to form macroscopic colonies (Figure 4). As previously reported, EWS-FLI-1 transforms NIH3T3 cells whereas FLI-1 does not (May et al., 1993b), nor does the empty vector (Figure 4) The two di€erent TAFII68DC159-FLI-1 and TAFII68DC198FLI-1 fusion proteins transform NIH3T3 cells with about the same eciency (135 and 128 colonies formed respectively) but are about three times less ecient than EWS-FLI-1 (430 colonies). These

TAFII68 displays oncogenic properties A Bertolotti et al

8004

results demonstrate that TAFII68-FLI-1 chimeras can induce oncogenic transformation, being however less potent oncoproteins than EWS-FLI-1. In conclusion, the amino-terminus of TAFII68 displays not only transactivation but also transformation properties.

Figure 3 Stable expression of chimeric proteins in NIH3T3 cells. Nuclear extract of retrovirally transduced and polyclonally selected NIH3T3 cell population was analysed by Western blot using a monoclonal anti-FLI-1 antibody. The fusion proteins are indicated by an arrow and the asterix indicates an non speci®c band also present in the non infected cells (lane 10). The position of the molecular weight markers (in kDa) are indicated

The recruitment of various components of the preinitiation complex or the GCN5 HAT to ETS-responsive promoters is not sucient to activate transcription nor to promote cellular transformation Since transcriptional stimulation of the FLI-1 target genes by EWS-FLI-1 or TAFII68-FLI-1 leads to oncogenic transformation, we next investigated the molecular mechanism of activation of these genes in vivo. We have previously shown that both EWS and TAFII68 are associated with TFIID and the RNA Pol II complexes (Bertolotti et al., 1996, 1998). Moreover, the amino-termini of either EWS or TAFII68 directly interact with certain subunits of the TFIID and the RNA Pol II complexes (Bertolotti et al., 1998). It has been shown that arti®cial recruitment of components of the PIC or a HAT to promoters could lead to transcriptional activation (see Introduction). To analyse whether the transcriptional activation and transforming properties of EWS-FLI-1 or TAFII68FLI-1 could be achieved by the facilitation or the stimulation of PIC formation, we investigated whether activation of ETS responsive genes and transformation could be obtained by arti®cial recruitment of di€erent PIC components or a HAT. For this purpose, we generated fusion proteins composed of either one of the TFIID subunits (TBP and TAFII28), one of the Pol II subunits (RPB3 and RPB7) or the GCN5 HAT fused to the FLI-1 C-terminal domain (Figure 1b). The ®rst 96 amino acids of TBP are composed of 38

Figure 4 The amino-terminus of TAFII68 has oncogenic properties when fused to the carboxy-terminal domain of FLI-1. (a) Polyclonal selected NIH3T3 cells infected with FLI-1, EWS-FLI-1, or TAFII68-FLI-1-expressing retrovirus were seeded in soft agar at a density of 8000 cells per plate in duplicates. EWS-FLI-1 and the two TAFII68-FLI-1 fusion proteins promote the formation of colonies. (b) Colonies were detectable after 1 week and the number of colonies was scored after 3 weeks of growth. +: standard deviation

TAFII68 displays oncogenic properties A Bertolotti et al

glutamines, and this glutamine rich content is characteristic of some transcription activation domains (Mitchell and Tjian, 1989; Seipel et al., 1993). When this region of TBP is fused to a heterologous DBD it is able to activate transcription (Seipel et al., 1993). Therefore, in our TBP-FLI-1 fusion protein this glutamine rich N-terminal end of TBP has been removed (Figure 1b). The full length coding sequences of hTAFII28, hRPB3 and hRPB7 where also fused to FLI-1 C-terminal domain (Figure 1b). To create the GCN5-FLI-1 chimera, the GCN5 cDNA which encodes the short form of human GCN5 protein (also called hGCN5-S) with a similar size to yeast GCN5 (Candau et al., 1996; Smith et al., 1998; Xu et al., 1998), were fused to FLI-1 C-terminal domain (Figure 1b). The di€erent TAFII-FLI-1 and TBP-FLI-1 chimeras were named TFIID-FLI-1 fusions, the chimeras containing the di€erent Pol II subunits were named `Pol II-FLI-1' fusions and the chimera with the GCN5 was named `HAT-FLI-1'. We tested whether the TFIID-, Pol II- and HATFLI-1 chimeras could activate transcription in mammalian cells similarly to EWS-FLI-1 or TAFII68-FLI-1. When the various TFIID-FLI-1, Pol II-FLI-1 or GCN5-FLI-1 expression vectors were transfected in HeLa cells together with a reporter plasmid responsive to EWS-FLI-1, none of them activated transcription eciently when compared to EWS-FLI-1 (Figure 5). However, similarly to the EWS- and TAFII68-FLI-1 chimeras (Figures 2b and 5) an inhibition of transcription was observed from the HTLV-1 LTR when more than 20 ng of expression vector was transfected for TFIID-FLI-1, Pol II-FLI-1 and the GCN5-FLI-1 chimeras (Figure 5; and data not shown) probably by competing for the binding to the ETS-binding sites (EBSs) with the endogenous ETS proteins. These results together with Western blot analyses (Figure 3 and data not

shown) indicate that the di€erent chimeras are expressed and that they are able to bind to their binding sites on the promoter tested. Importantly, when more than 1 mg of TBP-FLI-1 (or GAL-TBP) expression vector was transfected into HeLa cells, the luciferase expression that was used to normalize transfection eciencies was always strongly inhibited (data not shown), indicating that high fusion protein concentration in the cells impair transcription in general and thus cannot be used in our transfection system. In conclusion, the fusion of a core subunit of the preinitiation complex or the GCN5 HAT to the FLI-1 DBD is not sucient to mediate transcriptional activation of the tested ETS responsive gene. To test the di€erent TFIID-, Pol II- or GCN5-FLI1 chimeras on natural endogenous promoters, the various fusion proteins were retrovirally transduced into NIH3T3 cells. Nuclear extracts were prepared from the di€erent cell lines and protein expression was analysed by Western blot using an anti-FLI-1 antibody. All fusions were expressed and localized in the nucleus (Figure 3, and data not shown). Polyclonal cell populations were next analysed in soft agar assays in order to determine their oncogenic properties. The TFIID-, Pol II- or GCN5-FLI-1 fusion proteins were unable to promote the formation of colonies in soft agar even after more than 6 weeks of growth (Figure 6). These results demonstrate that the TFIID or Pol II subunit derived fusion proteins and GCN5-FLI-1 chimera, which have no transactivation properties on the EWS-FLI-1 responsive gene tested, have no oncogenic properties when overexpressed in NIH3T3 cells. In conclusion, there is a good correlation between the transactivation potential and the oncogenic properties of the FLI-1 fusion proteins further suggesting that deregulation of FLI-1 target genes is a key mechanism of oncogenic transformation. The fact

Figure 5 `TFIID-FLI-1', `Pol II-FLI-1' or `HAT-FLI-1' type fusion proteins do not transactivate transcription from an ETS binding site (EBS) containing reporter. HeLa cells were transfected with 20 or 100 ng of the indicated expression vector and 1 mg of the 10-1 reporter, containing the HTLV-1 LTR, schematized at the top of the ®gure. The histograms depict the fold stimulation of the reporter gene obtained with the di€erent chimeras. Each histogram represents the mean of at least three independent transfections after normalization for the internal control luciferase activity

8005

TAFII68 displays oncogenic properties A Bertolotti et al

8006

Figure 6 `TFIID-FLI-1', `Pol II-FLI-1' or `HAT-FLI-1' type fusion proteins are not oncogenic when stably expressed in NIH3T3 cells. (a) Soft agar assays were performed with polyclonal selected NIH3T3 cells infected with the indicated retroviruses. Eight thousand cells were inoculated (for each fusion protein in duplicates) per plate. (b) After 3 weeks of growth, colonies were counted and the results are summarized. +: standard deviation

that only EWS-FLI-1 and TAFII68-FLI-1, but none of the TFIID-, Pol II- or HAT-FLI-1 fusion proteins promote activation or transformation suggests that the arti®cial recruitment of di€erent components of the PIC or the GCN5 HAT to ETS responsive promoters is not sucient for either transactivation or cellular transformation.

Discussion TAFII68 is a proto-oncogene Here we show that the N-terminal domain of TAFII68 has oncogenic properties when it is fused to the FLI-1 DBD since TAFII68-FLI-1 chimeras lead to oncogenic

TAFII68 displays oncogenic properties A Bertolotti et al

transformation in NIH3T3 cells. Thus, the N-terminal domain of TAFII68 is not only structurally but also functionally very similar to those of EWS and TLS.FUS. EWS-FLI-1 is a more potent transcriptional activator than FLI-1, and in contrast to FLI-1, can transform NIH3T3 cells (May et al., 1993b). TLS and EWS are the targets of various translocation events in human sarcoma. To date, eight translocation partners have been found for EWS and two for TLS which are strikingly common to TLS (see Introduction). In each case, the translocation event fuses the N-terminal domain of one of the two related proteins to the DBD of one of several transcription factors, removing the C-terminal RNA binding domain of the germline EWS or TLS protein. Like the other TET family members, we found that the N-terminus of TAFII68 contains a transcriptional activation domain which can activate transcription when fused to two heterologous DBDs (Figure 2). However, GALTAFII68 fusions exhibit about a twofold weaker activation potential than GAL-EWS on a GAL4 responsive promoter. When fused to the FLI-1 Cterminal domain present in the oncogenic protein expressed in Ewing tumor, the N-terminal domain of TAFII68 has similar transactivation properties as the N-terminal domain of EWS on a EWS-FLI-1 responsive reporter gene. Since the transformation by EWS-FLI-1 is thought to be a consequence of deregulation of FLI-1 target genes, and having shown that TAFII68-FLI-1 is a more potent transactivator than FLI-1 (Figure 2b), we tested the oncogenic properties of TAFII68-FLI-1. NIH3T3 cells overexpressing the N-terminal 198 amino acids of TAFII68-FLI-1 form colonies in soft agar similarly to cells expressing EWS-FLI-1 in contrast to FLI-1 (Figure 4). However, TAFII68-FLI-1 is about three times less oncogenic in this essay than EWS-FLI-1 even though the two proteins were expressed at approximately the same levels (Figure 3). The di€erences in the oncogenic eciency of the two fusion proteins can be correlated with the observation that GAL-TAFII68 is about twofold less potent activator than GAL4-EWS (Figure 2a). This difference is undetectable in the case of the TAFII68-FLI-1 or EWS-FLI-1 fusion proteins on the ETS responsive reporter may be because the maximum stimulation observed is much lower than the one obtained with the GAL4 responsive reporter (Figure 2a,b). Nevertheless, TAFII68 N-terminus is a strong transactivation domain and thus, there is a good correlation between the transactivation eciency and the oncogenic capacity of TAFII68. Our results further support the model in which the oncogenic transformation by EWS-FLI-1 in Ewing tumors is a direct consequence of its ability to stimulate eciently FLI-1 responsive genes. Consistent with these results, it has previously been shown that strong activation domains, like the acidic activation domain of VP16, when fused to FLI1 can transform NIH3T3 cells (Lessnick et al., 1995). Thus, the strength of a transcriptional activation domain may determine its oncogenic transformation capacity in the context of a fusion transcription factor created by chromosomal translocations (see also below). Although the gene encoding TAFII68 has not yet been implicated in human cancers, our results

suggest that the TAFII68 gene is a proto-oncogene and that the hTAFII68 gene could potentially be involved in chromosomal rearrangement(s) in human tumors. Arti®cial recruitment of components of the PIC or the GCN5 HAT does not substitute for transcriptional activation on ETS responsive genes Having demonstrated that TAFII68 N-terminus can activate transcription when fused to FLI-1 DBD and that the TAFII68-FLI-1 chimeric proteins can transform NIH3T3 cells similarly to EWS-FLI-1 we investigated the molecular mechanism by which these activators stimulate transcription. We have previously shown that TAFII68 and EWS N-termini can establish contacts with several TFIID and Pol II subunits (Bertolotti et al., 1998). It has been proposed that a single interaction between a transcriptional activation domain and a component of the holoenzyme can be sucient to promote gene activation (for review see Barberis and Gaudreau, 1998). To test whether the transcriptional stimulation by TAFII68 or EWS Nterminal activation domains could be bypassed by arti®cial recruitment of the PIC or a HAT we generated several TFIID-, Pol II- or HAT-FLI-1 fusion proteins. None of these chimeras mediate transcriptional activation on the FLI-1 responsive HTLV-1 LTR reporter gene in HeLa cells (Figure 4) and from other natural FLI-1 responsive promoters in NIH3T3 cells (Figure 6). Recruitment of TFIID to the promoter is amongst the early events that leads to the assembly of an active PIC. On the FLI-1 responsive promoter, recruitment of TBP or di€erent TAFIIs by the FLI-1 DBD is not sucient to stimulate transcription. Note that in our study the same region of TBP was fused to the FLI-1 C-terminal domain as the one used by Xiao et al. (1997) to generate a GAL4TBP fusion protein which is able to activate transcription from certain promoters. Our results obtained in the soft agar assays suggest that the recruitment of the TFIID complex is not the rate limiting step in the transcriptional activation for at least some of the FLI-1 responsive genes in vivo. Moreover, tethering RNA Pol II to FLI-1 promoters does not lead to transcriptional activation of the FLI-1 responsive genes. This result suggests that arti®cial recruitment of Pol II to FLI-1 responsive promoter is not sucient to activate transcription. A similar result has been obtained fusing ten di€erent Pol III subunits to the GAL4 DBD (Marsolier et al., 1994), suggesting that RNA polymerase recruitment is not sucient to initiate transcription. Furthermore, since the GCN5FLI-1 chimera did not eciently activate transcription or result in oncogenic transformation, it is conceivable that the recruitment of a histone acetylase activity is not sucient to overcome the physical need for an activation domain. To date, the only HAT which is able to activate transcription from a promoterproximal position, when fused to a DBD, is the CBP/p300 HAT which may have the ability to interact with TFIID and the RNA polymerase II holoenzyme simultaneously (Abraham et al., 1993; Nakajima et al., 1997). Nevertheless, since the arti®cial recruitment of HAT-, TFIID- or Pol II-FLI-1 chimeras did not bypass the physical need for TAFII68 and EWS

8007

TAFII68 displays oncogenic properties A Bertolotti et al

8008

activation domains on ETS-responsive promoters we would rather speculate that the FLI-1 responsive genes are activated by TAFII68-FLI-1 or EWS-FLI-1 at a post-recruitment step. In vivo at certain promoters of transcriptionally inactive genes the PIC can be found prebound but inactive (Chen et al., 1994; Barberis and Gaudreau, 1998) and therefore transcription cannot be activated by the additional recruitment of components of the PIC. In agreement with our study, others have shown that certain promoters such as the human immunode®ciency virus 1 (HIV-1) LTR cannot be activated by arti®cial recruitment (Xiao et al., 1997). The fact that the TFIID-, Pol II- or GCN5-FLI-1 fusions have an inhibitory e€ect on transcription (see Figure 5), show that these fusions can bind to their binding sites and suggest that they are unable to in¯uence a step after PIC formation in NIH3T3. Consequently, GAL-TBP induces mainly short transcripts in the absence of TAT from the HIV LTR, containing a GAL4 binding site (Majello et al., 1998), suggesting that GAL-TBP can enhance recruitment of the PIC, but in contrast to other activators it is unable to a€ect the processivity of Pol II. Certain transcriptional activators (i.e. VP16 or HSF) have been shown to act at events after PIC formation, at steps like promoter clearance or stimulation of elongation (Blair et al., 1996; Blau et al., 1996; Yankulov et al., 1994; Lis and Wu, 1993). Our preliminary results indicating that TAFII68 is able to activate transcription elongation in vitro (A Bertolotti and L Tora unpublished results) strongly suggest that the TAFII68 or EWS activation domains play an important role after PIC formation. This interpretation is consistent with the fact that the activation domain of VP16, which stimulates elongation, when fused to FLI-1 is able to induce oncogenic transformation in NIH3T3 cells (Lessnick et al., 1995), similarly to TAFII68 and EWS. Thus, in order to activate transcription eciently TAFII68-FLI-1 and EWS-FLI-1 may stimulate the transcription elongation process in vivo and may belong in a class of activators which can enhance both initiation and elongation. In conclusion, we have demonstrated here that the N-terminal domain of hTAFII68 has transcriptional and transforming activities. The present data are the ®rst to directly show that genes encoding TFIID subunits can be proto-oncogenes. Taken together with our previous work showing that the other TET family members, TLS and EWS, can enter TFIID complexes (Bertolotti et al., 1996, 1998), our data strengthen the link between oncogenic transformation by derivatives of TET family members and the deregulation of Pol II transcription. Moreover, our investigation into the mechanism of transformation by TAFII68-FLI-1 and EWS-FLI-1 suggest that neither the recruitment of TAFII68 and EWS targets TFIID or Pol II subunits, nor the recruitment of a HAT activity can recapitulate the transactivation and/or transforming functions of the N-termini of TAFII68 and EWS. These results are consistent with recent data suggesting that a bona ®de transcriptional activation domain performs functions in addition to the recruitment of individual components of the PIC (He and Weintraub, 1998; Nevado et al., 1999).

Materials and methods Construction of GAL4 and FLI-1 fusion proteins cDNAs encoding EWS and hTAFII68 were PCR ampli®ed with the appropriate restrictions sites, digested and inserted in frame with the GAL4 (1 ± 147) DNA binding domain into the pG4M polyII vector (Webster et al., 1988a) (Figure 1a). The FLI-1 cDNA in the pCMV Neo expression vector was described in (Petermann et al., 1998). The cDNA encoding FLI-1 was PCR ampli®ed with SnaBI and SalI sites and inserted in the corresponding sites of the pBabe Puro vector (Morgenstern and Land, 1990). The EWS-FLI-1, TAFII68FLI-1 and TBP-FLI-1 fusions were generated by PCR with the appropriate restriction sites. EWS-FLI-1 ¯anked by SnaBI and SalI sites was inserted in the corresponding sites of the of the pBabe Puro (Morgenstern and Land, 1990) and in the SmaI and XhoI sites of the pXJ41 (Xiao et al., 1991). TAFII68-FLI-1 and TBP-FLI-1 ¯anked by BamHI and SalI sites were inserted into the corresponding sites of the pBabe Puro vector and in BamHI and XhoI sites of the pXJ41 (Xiao et al., 1991). To generate the TAFII28-FLI-1, RPB3-FLI-1, RPB7-FLI1 and GCN5-FLI-1 the cDNA encoding the C-terminal 233 amino acids FLI-1 was ®rst PCR ampli®ed with HindIII and SalI sites and inserted into the corresponding sites of the BSK+ vector (PharMingene). The cDNA corresponding to hTAFII28 (Mengus et al., 1995), hRPB3 and hRPB7 (Acker et al., 1997) or hGCN5-S (Candau et al., 1996) were PCR ampli®ed (see Figure 1) with BamHI and HindIII sites and inserted in frame into the same sites of the BSK-FLI-1 plasmid described above. Then the di€erent cDNAs encoding the various FLI-1 fusion proteins were excised from the pBSK+ using a BamHI-SalI digestion. The inserts were then recloned into the BamHI and SalI sites of the pBabe Puro vector and into the BamHI and XhoI sites of the pXJ41 vector (Xiao et al., 1991). All constructs were veri®ed by sequencing. Transfections and CAT assays HeLa cells were grown in Dulbecco's modi®ed Eagle's medium (DMEM) supplemented with 5% fetal calf serum. Approximately 106 cells were seeded and transfected 1 h after by the calcium phosphate co-precipitation method. In addition to the expression vectors or reporter plasmids described in each ®gure, all transfections contain 1 mg of the luciferase reporter pRSV-Luc (de Wet et al., 1987) as an internal standard and pBSK- as a DNA carrier up to 20 mg. Cells were incubated with the DNA coprecipitate for 12 h, washed, further grown and harvested 40 h after transfection. Luciferase assays were performed as described in (de Wet et al., 1987) and quanti®ed using a Berthold MicroLumat luminometer. CAT assays were performed by standard procedure (Sambrook et al., 1989) using amount of protein extracts normalized to the luciferase activity. Quantitative phosphoimager analysis was performed on a Fujix BAS 2000 apparatus. Similar results (+5%) were obtained in at least three independent transfections. The results of typical experiments are shown in the ®gures. Fold stimulation by the di€erent activators is measured with respect to the reporter gene cotransfected with the empty expression vector. Nuclear extract, Western blot and antibodies Nuclear extract was performed as described in (Bertolotti et al., 1998) modi®ed for small scale preparation. Approximately 26107 cells were washed with PBS, collected and resuspend in 200 ml bu€er I [50 mM Tris-HCl pH 7.9, 1 mM EDTA, 1 mM DTT], containing 2.5 mg/ml of leupeptin, pepstatin, aprotinin, antipain, and chymostatin and gently vortexed for 10 min on ice. Cell lysates were centrifuged

TAFII68 displays oncogenic properties A Bertolotti et al

5 min at 3000 g and nuclear pellets were resuspended in 150 ml bu€er II [50 mM Tris-HCl pH 7.9, 25% (v/v) glycerol, 0.5 M Nacl, 1 mM EDTA, 1 mM DTT], containing 2.5 mg/ml of leupeptin, pepstatin, aprotinin, antipain, and chymostatin. Cell nuclei were broken by ten passages through a 27G needle, followed by three cycles of freeze-thaw. Insoluble material was removed by a 10 min centrifugation at 13000 g and the supernatant (nuclear extract) was collected. Western blots were performed exactly as described in (Bertolotti et al., 1998). The chimeric FLI-1 fusion proteins were detected by Western blots using the mAb7.3 described in (Melot et al., 1997). Transformation assay Replication de®cient retroviral stocks were created as described in (Pear et al., 1993). Brie¯y, BOSC 23 packaging cells were transiently transfected (Pear et al., 1993) with FLI1 or the various fusions described above cloned in the pBabe Puro vector. Medium containing the di€erent viruses was harvested and used to infect NIH3T3 cells. Following infection, cells were cultured 5 days under selection with puromycin (2,5 mg/ml). The transduction eciency was approximately the same for each construct (more than 90%). Nuclear extract were prepared from these polyclonal selected cell populations and analysed by Western blot to control the level of expression of the di€erent FLI-1-fusion proteins. Soft agar assays were performed as follows. 86103 cells were seeded per 3 cm diameter plate in 1.5 ml DMEM containing 10% calf serum (CS) and 0.3% noble agar (Difco) on a 3 ml layer of 0.6% noble agar in the same media. The following day, a top layer of 0.3% agar was added and the culture were fed with DMEM containing 10% CS every 4 days. Macroscopically visible colonies were observed after 1 week of growth and scored after 3 weeks. Plates were photographed at a 3.56 magni®cation after 3 weeks of growth. Each experiment was performed in duplicate at least

three times and the result of a typical experiment is presented in the ®gures. Note added in proof While this manuscript has been processed for publication SjoÏrgen et al. [Cancer Research, 59, 5054 ± 5067, (1999)] identi®ed the gene encoding hTAFII68 (TAF2N) as a fusion partner to the TEC1 gene (see Introduction) in extraskeletal myxoid chodrosarcomas. In these sarcomas the chromosomal translocation created an oncoprotein in which the ®rst 159 amino acids of TAFII68 (the same domain fused to FLI-1 in our study) are fused in frame to the entire TEC1 protein. Thus, the TAFII68 is a protooncogene. Acknowledgements We thank P Chambon for his continuous support. We are grateful to A Boureux, J Ghysdael and S Casarosa for the kind gift of reagents and very useful advice; I Davidson, M Vigneron, O Delattre, H Kovar, R Petermann and S Berger for cDNAs encoding hTAFII28, hRPB3, hRPB7, EWS, FLI-1 and hGCN5-S and the anti-FLI-1 antibody; C Werle and B Boulay for illustrations and photography. A Bertolotti was supported by a fellowship from Association pour la Recherche contre le Cancer, B Bell was supported by fellowships from Fondation pour la Recherche Me dicale and Institut National de la Sante et de la Recherche MeÂdicale (INSERM). This work was supported by funds from the INSERM, the Centre National de la Recherche Scienti®que (CNRS), the HoÃpital Universitaire de Strasbourg, the Association pour la Recherche sur le Cancer (ARC), the Fondation pour la Recherche MeÂdicale (FRM), the Ligue Nationale Contre le Cancer and the Comite DeÂpartemental du Haut-Rhin de la Ligue National Contre le Cancer.

References Abraham SE, Lobo S, Yaciuk P, Wang HG and Moran E. (1993). Oncogene, 8, 1639 ± 1647. Acker J, de Graa€ M, Cheynel I, Khazak V, Kedinger C and Vigneron M. (1997). J. Biol. Chem., 272, 16815 ± 16821. Apone LM, Virbasius CM, Reese JC and Green MR. (1996). Genes Dev., 10, 2368 ± 2380. Bailly RA, Bosselut R, Zucman J, Cormier F, Delattre O, Roussel M, Thomas G and Ghysdael J. (1994). Mol. Cell. Biol., 14, 3230 ± 3241. Barberis A, Pearlberg J, Simkovich N, Farrell S, Reinagel P, Bamdad C, Sigal G and Ptashne M. (1995). Cell, 81, 359 ± 368. Barberis A and Gaudreau L. (1998). Biol. Chem., 379, 1397 ± 1405. Bell B and Tora L. (1999). Exp. Cell. Res., 246, 11 ± 19. Bertolotti A, Lutz Y, Heard DJ, Chambon P and Tora L. (1996). EMBO J., 15, 5022 ± 5031. Bertolotti A, Melot T, Acker J, Vigneron M, Delattre O and Tora L. (1998). Mol. Cell. Biol., 18, 1489 ± 1497. Blair WS, Fridell RA and Cullen BR. (1996). EMBO J., 15, 1658 ± 1665. Blau J, Xiao H, McCracken S, O'Hare P, Greenblatt J and Bentley D. (1996). Mol. Cell. Biol., 16, 2044 ± 2055. Bosselut R, Duvall JF, Gegonne A, Bailly M, Hemar A, Brady J and Ghysdael J. (1990). EMBO J., 9, 3137 ± 3144. Brady J, Jeang KT, Duvall J and Khoury G. (1987). J. Virol., 61, 2175 ± 2181. Braun BS, Frieden R, Lessnick SL, May WA and Denny CT. (1995). Mol. Cell. Biol., 15, 4623 ± 4630. Brou C, Chaudhary S, Davidson I, Lutz Y, Wu J, Egly JM, Tora L and Chambon P. (1993a). EMBO J., 12, 489 ± 499.

Brou C, Wu J, Ali S, Scheer E, Lang C, Davidson I, Chambon P and Tora L. (1993b). Nucleic Acids Res., 21, 5 ± 12. Brown AD, Lopez-Terrada D, Denny C and Lee KA. (1995). Oncogene, 10, 1749 ± 1756. Burd CG and Dreyfuss G. (1994). Science, 265, 615 ± 621. Calvio C, Neubauer G, Mann M and Lamond AI. (1995). RNA., 1, 724 ± 733. Candau R, Moore PA, Wang L, Barlev N, Ying CY, Rosen CA and Berger SL. (1996). Mol. Cell. Biol., 16, 593 ± 602. Chatterjee S and Struhl K. (1995). Nature, 374, 820 ± 822. Chen J, Ding M and Pederson DS. (1994). Proc. Natl. Acad. Sci. USA, 91, 11909 ± 11913. Chi T and Carey M. (1996). Genes Dev., 10, 2540 ± 2550. Chrivia JC, Kwok RP, Lamb N, Hagiwara M, Montminy MR and Goodman RH. (1993). Nature, 365, 855 ± 859. Crozat A, Aman P, Mandahl N and Ron D. (1993). Nature, 363, 640 ± 644. Dantonel JC, Murthy KG, Manley JL and Tora L. (1997). Nature, 389, 399 ± 402. de Wet JR, Wood KV, De Luca M, Helinski DR and Subramani S. (1987). Mol. Cell. Biol., 7, 725 ± 737. Delattre O, Zucman J, Plougastel B, Desmaze C, Melot T, Peter M, Kovar H, Joubert I, de Jong P and Rouleau G et al. (1992). Nature, 359, 162 ± 165. Farrell S, Simkovich N, Wu Y, Barberis A and Ptashne M. (1996). Genes Dev., 10, 2359 ± 2367. Fujimura Y, Ohno T, Siddique H, Lee L, Rao VN and Reddy ES. (1996). Oncogene, 12, 159 ± 167. Gegonne A, Bosselut R, Bailly RA and Ghysdael J. (1993). EMBO J., 12, 1169 ± 1178.

8009

TAFII68 displays oncogenic properties A Bertolotti et al

8010

Gill G and Ptashne M. (1988). Nature, 334, 721 ± 724. Gonzalez-Couto E, Klages N and Strubin M. (1997). Proc. Natl. Acad. Sci. USA, 94, 8036 ± 8041. Goodrich JA, Cutler G and Tjian R. (1996). Cell, 84, 825 ± 830. He S and Weintraub SJ. (1998). Mol. Cell. Biol., 18, 2876 ± 2883. Himmelfarb HJ, Pearlberg J, Last DH and Ptashne M. (1990). Cell, 63, 1299 ± 1309. Hirose Y and Manley JL. (1998). Nature, 395, 93 ± 96. Immanuel D, Zinszner H and Ron D. (1995). Mol. Cell. Biol., 15, 4562 ± 4571. Jacq X, Brou C, Lutz Y, Davidson I, Chambon P and Tora L. (1994). Cell, 79, 107 ± 117. Karnieli E, Werner H, Rauscher FJ, Benjamin LE and Le Roith D. (1996). J. Biol. Chem., 271, 19304 ± 19309. Klages N and Strubin M. (1995). Nature, 374, 822 ± 823. Kuo MH and Allis CD. (1998). Bioessays, 20, 615 ± 626. Lessnick SL, Braun BS, Denny CT and May WA. (1995). Oncogene, 10, 423 ± 431. Lis J and Wu C. (1993). Cell, 74, 1 ± 4. Majello B, Napolitano G, De Luca P and Lania L. (1998). J. Biol. Chem., 273, 16509 ± 16516. Marsolier MC, Chaussivert N, Lefebvre O, Conesa C, Werner M and Sentenac A. (1994). Proc. Natl. Acad. Sci. USA, 91, 138 ± 142. Martinez-Balbas MA, Bannister AJ, Martin K, HausSeu€ert P, Meisterernst M and Kouzarides T. (1998). EMBO J., 17, 2886 ± 2893. May WA, Gishizky ML, Lessnick SL, Lunsford LB, Lewis BC, Delattre O, Zucman J, Thomas G and Denny CT. (1993a). Proc. Natl. Acad. Sci. USA, 90, 5752 ± 5756. May WA, Lessnick SL, Braun BS, Klemsz M, Lewis BC, Lunsford LB, Hromas R and Denny CT. (1993b). Mol. Cell. Biol., 13, 7393 ± 7398. May WA and Denny CT. (1997). Curr. Top. Microbiol. Immunol., 220, 143 ± 150. McCracken S, Fong N, Yankulov K, Ballantyne S, Pan G, Greenblatt J, Patterson SD, Wickens M and Bentley DL. (1997). Nature, 385, 357 ± 361. Melot T, Gruel N, Doubeikovski A, Sevenet N, Teillaud JL and Delattre O. (1997). Hybridoma, 16, 457 ± 464. Mengus G, May M, Jacq X, Staub A, Tora L, Chambon P and Davidson I. (1995). EMBO J., 14, 1520 ± 1531. Mitchell PJ and Tjian R. (1989). Science, 245, 371 ± 378. Morgenstern JP and Land H. (1990). Nucleic Acids Res., 18, 3587 ± 3596. Mortillaro MJ, Blencowe BJ, Wei X, Nakayasu H, Du L, Warren SL, Sharp PA and Berezney R. (1996). Proc. Natl. Acad. Sci. USA, 93, 8253 ± 8257. Nagai K, Oubridge C, Jessen TH, Li J and Evans PR. (1990). Nature, 348, 515 ± 520. Nakajima T, Uchida C, Anderson SF, Lee CG, Hurwitz J, Parvin JD and Montminy M. (1997). Cell, 90, 1107 ± 1112. Nevado J, Gaudreau L, Adam M and Ptashne M. (1999). Proc. Natl. Acad. Sci. USA, 96, 2674 ± 2677. Ohno T, Rao VN and Reddy ES. (1993). Cancer Res., 53, 5859 ± 5863.

Panagopoulos I, Aman P, Fioretos T, Hoglund M, Johansson B, Mandahl N, Heim S, Behrendtz M and Mitelman F. (1994). Genes Chromosomes Cancer, 11, 256 ± 262. Pear WS, Nolan GP, Scott ML and Baltimore D. (1993). Proc. Natl. Acad. Sci. USA, 90, 8392 ± 8396. Petermann R, Mossier BM, Aryee DN, Khazak V, Golemis EA and Kovar H. (1998). Oncogene, 17, 603 ± 610. Rabbitts TH, Forster A, Larson R and Nathan P. (1993). Nat. Genet., 4, 175 ± 180. Ranish JA and Hahn S. (1996). Curr. Opin. Genet. Dev., 6, 151 ± 158. Rauscher FJ. (1997). Curr. Top. Microbiol. Immunol., 220, 151 ± 162. Roberts SG and Green MR. (1994). Nature, 371, 717 ± 720. Roeder RG. (1996). Methods Enzymol., 273, 165 ± 171. Sambrook J, Fritsch EF and Maniatis T. (1989). Molecular cloning (A laboratory manual), 2nd edition. Cold Spring Harbor Laboratory. Sanchez-Garcia I and Rabbitts TH. (1994). Proc. Natl. Acad. Sci. USA, 91, 7869 ± 7873. Seipel K, Georgiev O, Gerber HP and Scha€ner W. (1993). Nucleic Acids Res., 21, 5609 ± 5615. Smith ER, Belote JM, Schiltz RL, Yang XJ, Moore PA, Berger SL, Nakatani Y and Allis CD. (1998). Nucleic Acids Res., 26, 2948 ± 2954. Stolow DT and Haynes SR. (1995). Nucleic Acids Res., 23, 835 ± 843. Tasset D, Tora L, Fromental C, Scheer E and Chambon P. (1990). Cell, 62, 1177 ± 1187. Tjian R and Maniatis T. (1994). Cell, 77, 5 ± 8. Wang W, Gralla JD and Carey M. (1992). Genes Dev., 6, 1716 ± 1727. Webster N, Jin JR, Green S, Hollis M and Chambon P. (1988a). Cell, 52, 169 ± 178. Webster NJ, Green S, Jin JR and Chambon P. (1988b). Cell, 54, 199 ± 207. Workman JL and Kingston RE. (1998). Ann. Rev. Biochem., 67, 545 ± 579. Xiao H, Friesen JD and Lis JT. (1995). Mol. Cell. Biol., 15, 5757 ± 5761. Xiao H, Lis JT and Jeang KT. (1997). Mol. Cell. Biol., 17, 6898 ± 6905. Xiao JH, Davidson I, Matthes H, Garnier JM and Chambon P. (1991). Cell, 65, 551 ± 568. Xu W, Edmondson DG and Roth SY. (1998). Mol. Cell. Biol., 18, 5659 ± 5669. Yankulov K, Blau J, Purton T, Roberts S and Bentley DL. (1994). Cell, 77, 749 ± 759. Yuryev A, Patturajan M, Litingtung Y, Joshi RV, Gentile C, Gebara M and Corden JL. (1996). Proc. Natl. Acad. Sci. USA, 93, 6975 ± 6980. Zinszner H, Albalat R and Ron D. (1994). Genes Dev., 8, 2513 ± 2526. Zinszner H, Sok J, Immanuel D, Yin Y and Ron D. (1997). J. Cell. Sci., 110, 1741 ± 1750.