Oncogene (1998) 16, 1625 ± 1631  1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00

The EWS/ATF1 fusion contains a dispersed activation domain that functions directly

Shu Pan, Koh Yee Ming, Theresa A Dunn, Kim KC Li and Kevin AW Lee

Department of Biology, Hong Kong University of Science & Technology, Clear Water Bay, Kowloon, Hong Kong, P.R.C.

Naturally occurring chromosomal fusion of the Ewings 1994). For all of the above malignancies, the EWS Sarcoma Oncogene (EWS) to distinct cellular transcrip- fusion function as potent transcriptional tion factors, produces aberrant transcriptional activators activators (May et al., 1993b; Ohno et al., 1993; that function as dominant oncogenes. In Malignant Bailly et al., 1994; Brown et al., 1995; Lessnick et al., Melanoma of Soft Parts the N-terminal region of 1995; Fujimura et al., 1996) in a manner that is EWS is fused to C-terminal region of the cAMP- dependent on the EWS N-terminal region, hereafter inducible ATF1. The EWS/ATF1 referred to as the EWS Activation Domain (EAD). It is fusion protein binds to ATF sites present in cAMP- envisioned that distinct tumors arise via de-regulation responsive promoters via the ATF1 bZIP domain and of di€erent , depending on the fusion partner for activates transcription constitutively in a manner that is EWS. In cases where it has been examined, agents that dependent on an activation domain (EAD) present in antagonise EWS-fusion proteins also inhibit cellular EWS. To further de®ne the requirements for trans- proliferation (Ouchida et al., 1995; Kovar et al., 1996; activation we have performed mutational analysis of Yi et al., 1997; Tanaka et al., 1997), indicating that EWS/ATF1 in mammalian cells and report several new EWS fusions can play a role in both tumor formation ®ndings. First, trans-activation by EWS/ATF1 does not and maintenance. The molecular mechanism by which require dimerisation with other ATF family members EWS activates transcription is therefore of signi®cance, present in mammalian cells. Second, in contrast to the both for understanding tumorigenesis and for develop- earlier suggestion of an allosteric role, the EAD can act ment of potential therapeutic agents that target EWS. directly. Third, determinants of trans-activation are The mechanism(s) by which the EAD alters the dispersed throughout the EAD and cooperate synergis- activity of oncogenic fusion proteins is not well tically to produce potent trans-activation. We also report characterized at the molecular level. To date there is that the region of EWS containing the EAD can activate evidence that the EAD can increase transcriptional transcription in Yeast. This latter ®nding might enable a activity via both allosteric and direct mechanisms. genetic approach to understanding the mechanism of Studies of the native oncogenic fusion proteins EWS/ transcriptional activation by EWS and development of FLI1 (Ohno et al., 1993) and EWS/ATF1 (Fujimura et high-throughput screens for EWS inhibitors. al., 1996) have suggested an allosteric role, while fusion of EWS or the EAD (or the related protein TLS/FUS) Keywords: EWS/ATF1; oncogene; MMSP; transcrip- to GAL4 has indicated that the EAD contains an tional activation; yeast activation domain that functions directly (May et al., 1993b; Bailly et al., 1994; Sanchez-Garcia and Rabbitts, 1994; Lessnick et al., 1995; Kim et al., 1997). EWS/ATF1 is a much more potent activator Introduction than EWS/FLI1 when compared with their corre- sponding non-tumorigenic counterparts, ATF and FLI Chromosomal translocations involving fusion of the N- 1 respectively. EWS/ATF1 is *200-fold more active terminal region of the Ewings Sarcoma Oncogene than ATF1 (Brown et al., 1995) whereas EWS/FLI1 is (EWS) to a variety of cellular transcription factors, only 5 ± 10-fold more active than FLI1. The potency of produce dominant oncogenes that cause distinct trans-activation by EWS/ATF1 therefore o€ers advan- sarcomas (reviewed by Rabbitts, 1994; Ladanyi, tages for transcriptional studies of native oncogenic 1995). In Malignant Melanoma of Soft Parts fusion proteins containing EWS and possibly for (MMSP) a causative t(12;22) chromosomal transloca- examining the function of the normal EWS protein. tion gives rise to a fusion protein (EWS/ATF1, Figure Important functions of EWS/ATF1 are contributed 1) in which EWS is fused to the C-terminal region of by both fusion partners (Figure 1). ATF1 binds the cellular transcription factor ATF1 (Zucman et al., directly to cAMP-inducible promoters via the bZIP 1993a). In Ewings sarcoma, EWS is fused to the ETS domain (Hurst et al., 1991; Lee and Masson, 1993) and domain family members FLI1 (Delattre et al., 1992; activates transcription upon phosphorylation by PKA May et al., 1993a; Zucman et al., 1993b) or ERG1 (Ribeiro et al., 1994). The function of the normal EWS (Sorenson et al., 1994) and in Desmoplastic Small protein is not well characterised but it contains an Round Cell Tumor (DSRCT) EWS is fused to the RNA-binding domain at its C-terminus (Burd and Wilms tumor oncogene WT1 (Ladanyi and Gerald, Dreyfuss, 1994) suggesting that it plays a role in some aspect of RNA metabolism. However, recent ®ndings have revealed a high degree of homology between EWS and the human TBP-associated factor (hTAF Correspondence: KAW Lee ll68, Received 26 August 1997; revised 24 October 1997; accepted 24 Bertolotti et al., 1996) suggesting that EWS may October 1997 function directly as a transcription factor. In contrast Transcriptional activation by the EWS/ATF1 oncogene SPanet al 1626 to ATF1, EWS/ATF1 functions as a potent constitu- ATF1 does not activate promoters that do not contain tive activator of several PKA-inducible promoters ATF binding sites (Brown et al., 1995). (Brown et al., 1995; Fujimura et al., 1996) and Similar to the trans-activation domains present in activation is strictly dependent on the EAD (Brown many transcriptional activators, the EAD is rich in et al., 1995; Fujimura et al., 1996). With respect to proline, glutamine and serine/threonine residues and promoter speci®city, it has been shown that (1) EWS/ has thirty one copies of a consensus repeat SYGQQS ATF1 can activate a broad range of ATF-dependent (Delattre et al., 1992). The important structural promoters; (2) activation is dependent on the ATF features of the EAD and its mechanism of action binding site in the promoter; (3) EWS/ATF1 does not have not been thoroughly investigated, although it has activate all ATF-dependent promoters and (4) EWS/ been suggested that, for EWS/ATF1, the EAD plays a

Figure 1 Functional regions of EWS/ATF1 and summary of EWS/ATF1 mutants. EWS/ATF1 contains the N-terminal region of EWS (residues 1 ± 325) and the C-terminal region of ATF1 (residues 66 ± 271). The EWS region present has a repetitive primary structure, with several prevalent amino acids (serine and threonine (*25%), proline (*10%), glutamine (*15%) and tyrosine (*10%)) dispersed evenly throughout. In addition, there are 31 copies of a repeat sequence, with the consensus SYGQQS, dispersed evenly throughout. EWS/ATF1 contains the C-terminal 75% ATF1 (residues 66 ± 271) but lacks the PKA phospho-acceptor site and does not function as a PKA-inducible activator. The bZIP domain (aa 214 ± 271) is necessary and sucient for dimerisation and DNA-binding and consists of the basic region (b) that directly contacts DNA and the (ZIP) that allows dimerisation. Q2 represents a glutamine-rich activation domain that has constitutive transcriptional activity (Brindle et al., 1993). For EWS/ATF1 mutants, the residues present are aligned with the intact protein at the top of the Figure and deletions are indicated with a dashed line. N-terminal deletions are named according to the number of EWS residues deleted and C-terminal deletions according to the number of EWS residues remaining. The EWS is represented by the striped box and the N-terminal 86 residues in D87C, D87CD and E(1 ± 87) by a shaded box. The black box in 4R and 8R represents one copy of the SYGQQS repeat sequence. Trans-activation assays in JEG3 cells were as described in Materials and methods. Reporter activity is CAT speci®c activity as determined by trans- activation of D(771)SomCat in the linear range for trans-activation and corrected for protein expression levels as determined by Western blot analysis of epitope-tagged proteins in transfected cells. For quantitation as %WT, the background activity (bkg) corresponds to the activity of D325 Transcriptional activation by the EWS/ATF1 oncogene SPanet all 1627 regulatory as opposed to a direct role in trans- Previous mutational analysis of EWS/ATF1 (Brown activation (Fujimura et al., 1996). Here we provide et al., 1995; Fujimura et al., 1996) demonstrated that evidence that the EAD can act directly but that the deletion of the N-terminal 78 residues of the EAD determinants of trans-activation are dispersed. Many strongly reduced activity (Figure 1) suggesting that this regions of the EAD have low activity by themselves region might be sucient for EAD function. We but cooperate synergistically to produce ecient trans- therefore tested this region fused to ATF1 residues activation. In addition, trans-activation by EWS/ATF1 66 ± 271 (D87C contains the N-terminal 86 residues of does not require dimerisation with endogenous EWS) for trans-activation and found that D87C gives partners of ATF1. We also show that the EAD can 80-fold higher activity than D325, while the N-terminal activate transcription in yeast. This latter ®nding might 86 residues of EWS by itself (E(1 ± 87)) has no activity enable a genetic approach to study the mechanism of (Figure 2). Thus the N-terminal 86 residues of the transcriptional activation by EWS and allow a high EAD fused to ATF1 has signi®cant activity. Examina- throughput screen for inhibitors of EWS. tion of protein expression levels however revealed that D87C is expressed at much higher levels than EWS/ ATF1 (data not shown). We therefore performed Mutational analysis of EWS/ATF1 titration experiments by varying the amount of DNA To test the ability of a series of EWS/ATF1 mutants transfected to achieve equal protein expression and to (Figure 1) to activate transcription we used a ensure that trans-activation was in the linear range previously described transient assay (Brown et al., (Figure 2). This revealed that D87C is *20-fold less 1995). Brie¯y, an expression vector for EWS/ATF1 (or active than intact EWS/ATF1. Thus in the context of mutants thereof) and a CAT reporter linked to a EWS/ATF1, the N-terminal 86 residues of the EAD is truncated somatostatin promoter (D(771)SomCAT, not sucient for strong trans-activation and other which contains a single ATF binding site) are regions of EWS must play an important role. transiently introduced into cells lacking EWS/ATF1 To test for contributions from other regions of (JEG3 cells) and transcriptional activity is monitored EWS/ATF1 we examined a number of additional by CAT assay. In this assay, intact EWS/ATF1 is a deletion mutants. The data are summarised in Figure very strong activator, dependent on the EAD and (at 1 and N-terminal deletions are as previously published least) the DNA-binding (bZIP) domain of ATF1 (Brown et al., 1995). First, progressive internal (Brown et al., 1995; Fujimura et al., 1996). deletions from the C-terminus of the EAD (D287C ±

∆87C 325 ∆ ∆87C ∆247C WT A-cm — A-cm —

cm — cm —

1 2 3 4 5 6 7 8 9 10

kDa

80 — kDa

49 — 80 — 49 — 32 —

32 —

2 3 4 1 5 6 7 8 9 10 Figure 2 Transcriptional activation by the N-terminal region of EWS/ATF1 in JEG3 Cells. Representative CAT assays are shown at the top and Western blot analysis of epitope-tagged proteins at the bottom. Each track (indicated numerically) for the CAT assay has a corresponding track for the Western analysis performed on the same transfected cells. Trans-activation assays in JEG3 cells were as described in Materials and methods except that activator plasmids were included at di€erent levels so as to achieve equal protein expression for the di€erent mutants. For tracks 1 ± 4, cells were transfected with 0.3, 1 and 5 mgofpD87C (indicated above the autoradiogram) and 5 mgofpD325. For tracks 5 ± 10, cells were transfected with 0.1, 0.5 and 1.5 mgofpD87C and 1.5 and 5 mg of pD247C (indicated above the autoradiogram) and 5 mg of pSVEWS/ATF1 (WT). For Western blots, nuclear extracts were prepared from the same cells used to measure CAT activity and epitope-tagged proteins (Masson et al., 1993) detected by Western blotting using the KT3 monoclonal antibody (MacArthur and Walter, 1984). Molecular weight standards (Biorad pre-stained low size range) are indicated to the left Transcriptional activation by the EWS/ATF1 oncogene SPanet al 1628 D87C) resulted in a gradual decrease in activity. ATF1 residues 66 ± 271 (Figure 3). Proteins 4R, 8R and Second, internal deletion of residues 87 ± 167 (D87 ± 12R gave a characteristic pattern of two discrete bands 167) resulted in only a modest *2.5-fold reduction in on SDS gels. Two considerations indicate that these activity. Third, deletion of the entire ATF1 sequence bands do not arise by proteolysis. First, the lower band except the bZIP domain resulted in an *fourfold in each case is consistent with the expected apparent decrease in activity (compare EbZ and D247C). Fourth, molecular weight for an intact polypeptide and second, two other regions of EWS (residues 174 ± 243) and the mobility of both bands increases in parallel, (residues 88 ± 173) expressed only low levels of protein indicating that there is no common proteolytic and had no detectable activity (data not shown). In fragment being produced. None of the arti®cial light of the overall sequence similarity between the N- proteins described above had signi®cant activity above terminal 86 residues of the EAD and the remainder of the levels given by ATF1 residues 66 ± 271 alone the EAD, we tested the e€ect of duplicating this region. (D325). Thus, in the context of EWS/ATF1, the D87CD contains two copies of the above region but consensus SYGQQS repeat does not have intrinsic has *sixfold more activity than a single copy (D87C) transcriptional activity. and is only *threefold less active than intact EWS/ ATF1. Thus there is a synergistic e€ect of duplicating EWS/ATF1 can activate transcription as a homodimer EAD residues 1 ± 86. The following conclusions can be made from the above results. First, the EAD is Over-expression of EWS/ATF1 in JEG3 cells results in constituted by dispersed elements throughout the transcriptional activation. However JEG3 cells contain EWS region of EWS/ATF1 with an additional CREB and ATF1 (Hurst et al., 1990) which have the contribution coming from the ATF1 portion. Second, potential to dimerise with EWS/ATF1 and contribute no single region of the EAD has high activity but to trans-activation. To test the activity of essentially di€erent regions function synergistically to produce homodimeric EWS/ATF1 we used a previously ecient activation. Third, the EAD acts directly as a described assay (Ribeiro et al., 1994) that prevents transcriptional activation domain. The latter conclu- interference from endogenous dimerisation partners sion is based on the ®nding that a protein containing (Figure 4). ZIP3 and ZIP12 proteins have a highly the N-terminal 247 amino acids of the EAD fused to similar overall structure to EWS/ATF1 except for (1) the ATF1 bZIP domain (EbZ) has *100-fold more replacement of the basic domain of ATF1 with the activity than D325 (Figure 1). basic domain of the Epstein Barr virus Zta protein and (2) compensatory mutations within the ATF1 leucine zipper that disrupt homodimerisation but allow The consensus SYGQQS repeat is not sucient for selective dimerisation of the mutated partners trans-activation (Ribeiro et al., 1994; Loriaux et al., 1993). Thus, The EAD contains 31 copies of a repeat sequence with neither ZIP3 nor ZIP12 will form homodimers or bind the consensus SYGQQS and the repeats contribute to a reporter (pZ7E4TCAT) containing Zta binding *60% of the residues present in the EAD. It therefore sites. However, when present together, ZIP3 and ZIP12 seemed probable, a priori, that the repeat would can form heterodimers and bind to the reporter, thus contribute to (and might even be sucient for) enabling trans-activation to be monitored. Co-transfec- transcriptional activation. We directly tested the tion of pZ7E4TCAT with expression vectors for either ability of a multimerised consensus repeat to activate ZIP3 or ZIP12 alone did not result in trans-activation, transcription, by making constructs that encode 4 (4R), as expected (Figure 4). However when ZIP3 and ZIP12 8 (8R) and 12 (12R) copies of the repeat fused to were co-expressed, strong trans-activation was ob-

a 325 210 87C

∆ ∆ ∆ 4R 8R 12R

A-cm — b ∆87C 325 210

∆ ∆ 4R 8R 12R kDa

49 —

32 — 27 —

cm —

Figure 3 Transcriptional activity of synthetic activators containing multiple copies of the SYGQQS repeat. (a) CAT assays. Trans- activation assays in JEG3 cells were under standard conditions as described in the Materials and methods. A representative autoradiogram is shown. (b) Protein expression. For Western blot analysis, di€erent amounts of cell extract were loaded onto the gel for D87C as indicated at the top. Molecular weight standards (Biorad pre-stained low size range) are indicated to the left Transcriptional activation by the EWS/ATF1 oncogene SPanet all 1629 served. This shows that a ZIP3/ZIP12 heterodimer GAL4 driven reporter) were transformed with a (essentially representing homodimeric EWS/ATF1) is a plasmid (pG4/E285) expressing the GAL4 DNA- transcriptional activator and indicates that trans- binding domain fused to EAD residues 1 ± 285. Several activation by EWS/ATF1 does not require dimerisa- independent transformants were stained on ®lters for tion with either CREB or ATF1. b-Gal activity or cell extracts were prepared for quantitative estimates. pG4/E285 gave readily detect- able transcriptional activation in yeast as determined The EAD can activate transcription in yeast by X-gal staining, although the level of activation was Many natural mammalian transcriptional activators weak compared with a positive control (pCL1) that function in yeast which o€ers advantages for studying expresses the intact GAL4 protein. pG4/E285 gave transcription factors. In particular, yeast are amenable *3% of the activity of intact GAL4. As controls, the to genetic analysis and can also be used for high GAL4-DNA binding domain alone (pGBT9) failed to throughput screens for small molecule inhibitors. We activate and a protein containing EAD residues 1 ± 285 therefore decided to test EWS for the ability to activate but lacking the GAL4-DNA binding domain (pDN6) transcription in yeast (Figure 5). We could not test was also inactive. A protein containing EAD residues EWS/ATF1 because reporters containing ATF-binding 1 ± 125 gave no activity in the solution assay and sites are constitutively activated by endogenous yeast stained weakly in the ®lter assay. Western blot analysis ATF (Lin and Green, 1989; Jones and Jones, 1989). indicated that pG4/E285 and pG4/E125 express similar Instead, we tested a GAL4-EAD fusion lacking any levels of GAL4/EAD fusion proteins (data not shown). ATF1 sequences (Figure 5). Yeast strain Y190 (GAL4 We conclude that EAD signi®cantly but weakly minus and containing a chromosomal b-Galactosidase activates transcription in yeast.

A-cm —

cm —

ZIP3 – + – + ZIP12 –– ++ Figure 4 Transcriptional activity of EWS/ATF1 homodimers. JEG3 cells were transfected with 5 mg pZ7E4TCAT reporter plasmid in the absence of activator, with 5 mg of pZIP3 or 5 mg of pZIP12 alone or with a combination of 2.5 mg of each of pZIP3 and pZIP12 as indicated at the bottom

Figure 5 Transcriptional activation by GAL4/EWS in yeast. Expression plasmids for activators are named to the left. Regions of GAL4 are shown in white and regions of EWS in black. GAL4(1 ± 147) contains the DNA-binding domain of GAL4. For reporter activity, yeast strain Y190 containing a chromosomal b-Galactosidase GAL4 driven reporter was transformed with activator plasmids and several distinct colonies were stained with X-gal on ®lters for b-Gal activity (®lter assay) or cell extracts were prepared for quantitative estimates by solution assay (units b-Galactosidase) as described in Materials and methods. The average result from three independent experiments is shown Transcriptional activation by the EWS/ATF1 oncogene SPanet al 1630 Discussion speci®c role in trans-activation by EWS/ATF1. Our ®ndings demonstrate that although this region has Previous studies (Brown et al., 1995; Fujimura et al., signi®cant activity (particularly when duplicated) it is 1996) have addressed the mechanism of trans-activa- not by itself a strong activation domain. We note tion by EWS/ATF1. These studies have established however that compared with other regions of similar that the EWS portion of EWS/ATF1 is required for size, the N-terminal 86 amino acids of the EAD has trans-activation and have suggested that the N- signi®cantly more activity. For example D210 and D87C terminal region of EWS (the EAD) acts allosterically both contain a similar number of EWS residues (86 and to regulate an activation domain present in ATF1 115 respectively) but D87C is *tenfold more active than (Fujimura et al., 1996). An allosteric role for EWS has D210 (see Figures 1 and 3). These ®ndings suggest that also been suggested in the case of EWS/FLI1 fusion the extreme N-terminal region of the EAD (residues 1 ± protein (Ohno et al., 1993). In contrast to the above 86) may have a speci®c function in trans-activation by ®ndings, the data presented here demonstrate a direct EWS/ATF1 and that this region represents a useful role for the EAD in trans-activation by EWS/ATF1. starting point for further molecular analysis. This conclusion is based on the ability of EWS residues In the case of EWS/FLI1 and EWS/ERG1 (but not 1 ± 247 to strongly activate transcription (*100-fold) yet examined for EWS/ATF1 and EWS/WT1) agents when fused to the ATF1 bZIP domain and on the that antagonise the function of these EWS-fusion ®nding that, for a large number of mutants, there is a proteins also inhibit cellular proliferation (Ouchida et positive correlation between trans-activation and the al., 1995; Kovar et al., 1996; Yi et al., 1997; Tanaka et number of EWS residues present. The proposition that al., 1997). This indicates that EWS fusion proteins play the EAD functions directly in the transcription a role in both tumor formation and tumor maintenance complex is also consistent with the ®nding that the and that it might be feasible to utilize small molecule

human TBP-associated factor (hTAFll68) has a high inhibitors of EWS as anti-tumor drugs. The ability to degree of homology with EWS (Bertolotti et al., 1996). measure trans-activation by EWS in yeast, allows for The potency of trans-activation by EWS/ATF1 development of a high throughput screen for inhibitors together with ®nding that the EAD acts directly, of transcriptional activation that would serve as lead suggests that further studies of EWS/ATF1 will help compounds for drug development. understand the role of the EAD in both oncogenesis and the putative function of the normal EWS protein in the transcription complex. Materials and methods The important structural features of the EAD await resolution. Structural data is not yet available and Plasmids and constructions theoretical analysis indicates that the EAD is likely to pD(771)SomCAT contains the somatostatin promoter to have an extended ¯exible shape with little secondary position 771, fused to the chloramphenicol acetyl structure. Given that the EAD has a highly repetitive transferase (CAT) coding sequences (Montminy et al., primary structure and the ®nding that several regions 1986). pSVEWS/ATF1 (which expresses intact EWS/ of the EAD have low activity themselves but synergise ATF1), pD78, pD166, pD210, pD325 and pDQ2 are as together, it is likely that a repetitive primary sequence described previously (Brown et al., 1995). All other EWS/ element is a critical component of the EAD. ATF1 mutants (Figure 1) have the same background as pSVEWS/ATF1 and contain a C-terminal 7 amino acid Furthermore, since the SYGQQS repeat motif makes peptide from SV40 T-antigen that is recognized by the up *60% of the residues present in the EAD, it monoclonal antibody KT3 (MacArthur and Walter, 1984) remains likely that this motif represents the core of the and the sequence GAGAAAATGGCGTCC incorporated repetitive structure. Lack of trans-activation by the at the N-terminus to provide an ecient translation repeat sequence in our experiments might be explained initiation site. C-terminal deletion mutants were obtained by an inappropriate con®guration of the repeats (i.e. as follows. HindIII/BglII ended PCR products encoding the spacing and/or ¯anking residues selected) or by the the designated EWS residues, were inserted into a vector consensus repeat sequence used. These suggestions are (pEAvec) containing ATF1 residues 66 ± 271, a unique supported by the observations that the N-terminal BglII site at residue 66 and a multiple cloning site including region of the EAD (residues 1 ± 86) has the highest HindIII adjacent to the BglII site. pD86CD contains two copies of the N-terminal 86 residues of EWS and was activity (compared with other regions) but contains obtained by insertion of a BglII ended PCR product divergent repeats that are spaced further apart. An encoding residues 1 ± 86 into BglII digested pN1vec. additional possibility is that the serine and tyrosine pKBvec was obtained by insertion of a double stranded residues present in the repeat, allow phosphorylation oligonucleotide with BglII overhangs and internal KpnI events that are important for trans-activation. In and BglII sites into BglII digested pD87C. pE(88 ± 173) was relation to our data, it is quite likely that the distinct obtained by insertion of a KpnI/BamHI restriction species observed on SDS gels for the arti®cial proteins fragment from pSVEWS/ATF1 (containing EWS residues 4R, 8R and 12R, arise due to phosphorylation. Future 88 ± 173) into KpnI/BglII digested pKBvec. pE(174 ± 243) experiments will be required to de®ne the role of the was obtained by ligation of KpnI/BglII ended PCR product repeat motif in trans-activation and the potential role (containing EWS residues 174 ± 243) into KpnI/BglII digested pKBvec. p4R contains four copies of the repeat of phosphorylation. sequence SYGQQS and was obtained by insertion of The extreme N-terminal region of the EAD is critical oligonucleotides containing BglII overhangs into BglII for trans-activation by EWS/ATF1 (Brown et al., 1995; digested pN1vec. The oligonucleotides encode a transla- Fujimura et al., 1996) but is not important for tion initiation codon and two copies of EWS residues 219 ± activation by a GAL4/EWS fusion (Lessnick et al., 230 separated by an NheI restriction site which creates an 1995). We were therefore prompted to ask whether this alanine residue between the second and third repeats. Thus region contains an activation domain that might play a the amino acid sequence of the four repeat unit is as Transcriptional activation by the EWS/ATF1 oncogene SPanet all 1631 follows: SYGQQSSYGQQSASYGQQSSYGQQS. p8R containing 10% FCS. All transfections were carried out by contains eight copies of the SYGQQS repeat and was calcium phosphate co-precipitation and CAT assays were obtained by insertion of the same oligonucleotides used to performed as previously described (Gorman et al., 1982). make p4R into BglII digested p4R. p12R contains 12 Precipitates contained 5 mg of reporter plasmid, the copies of the SYGQQS repeat and was derived from p8R indicated amount of activator plasmid(s) and 20 mgtotal in the same manner. pZIP3 and pZIP12 were obtained by DNA made up with pGEM3 as carrier. One third of the replacing the ATF1 bZIP domain in EWS/ATF1 with the precipitate was added to *50% con¯uent JEG3 cells in a corresponding region from pZA3 and pZA12 (Ribeiro et 60 mm culture dish. For quantitation of results, % al., 1994) respectively. An Nde1/Xba1 fragment from conversion of unacetylated to acetylated 14C-chloramphe- pZA3 or pZA12 was inserted into Nde1/Xba1 digested nicol under linear assay conditions was determined by pSVEWS/ATF1. pZ7E4TCAT is described elsewhere excision of spots from the TLC plate and quantitation of (Carey et al., 1992). Plasmids for yeast experiments were radioactivity using a liquid scintillation counter. as follows. pCL-1 and pGBT9 encode intact GAL4 and the GAL4 DNA-binding domain respectively and were from Western blotting Clonetech. pG4 vector was obtained by insertion of an oligonucleotide into the multiple cloning site of pGBT9 to Western blotting was performed as previously described create in frame restriction sites for subsequent insertion of (Masson et al., 1993) except that ®lters were incubated EWS sequences. pG4/E285 and pG4/E125 were obtained throughout with PBS containing 3% dried milk and 10% by inserting SalI/BglII fragments from pD287 and pD127 fetal calf serum. respectively into newly created SalIandBamHI sites in pG4 vector. pDN6 was obtained by removal of the GAL4 Yeast experiments DNA-binding domain by digestion of pG4/E285 with HindIII and EcoRI and insertion of an oligonucleotide to Yeast plasmid vectors (pCL-1 and pGBT9) and yeast strain provide an in-frame translation initiation codon. Y190 were from Clonetech.

Acknowledgements Cell culture, transfections and CAT assays This work was supported by a Hong Kong Government JEG-3 cells (Kohler and Bridson, 1971) were maintained as Research Grants Council grant (award #HKUST 645/ monolayers in Dulbecco's modi®cation of Eagles medium 96M) to KAWL.

References

Bailly RA, Bosselut R, Zucman J, Cormier, F, Delattre O, Lin YS and Green MR (1989). Proc. Natl. Acad. Sci., 86, Roussel M, Thomas G and Ghysdael J. (1994). Mol. Cell. 109 ± 113. Biol., 14, 3230 ± 3241. Loriaux MM, Rehfuss RR, Brennan RG and Goodman RH. Bertolotti A, Lutz Y, Heard DJ, Chambon P and Tora L. (1993). Proc. Natl. Acad. Sci., 90, 9046 ± 9050. (1996). EMBO J., 15, 5022 ± 5031. MacArthur H and Walter G. (1984). J. Virol., 52, 483 ± 491. Brindle P, Linke S and Montminy M. (1993). Nature, 364, Masson N, John J and Lee KAW. (1993). Nucleic Acids Res., 821 ± 824. 21, 4166 ± 4173. Brown AD, Lopez-Terrada D, Denny CT and Lee KAW. May WA, Gishizky ML, Lessnick SL, Lunsford LB, Lewis (1995). Oncogene, 10, 1749 ± 1756. BC,DelattreO,ZucmanJ,ThomasGandDennyCT. Burd CG and Dreyfuss G. (1994). Science, 265, 615 ± 621. (1993a). Proc. Natl. Acad. Sci., 90, 5752 ± 5756. Carey M, Kolman J, Katz DA, Gradoille L, Barberis L and May WA, Lessnick SL, Braun BS, Klemsz M, Lewis BC, Miller G. (1992). J. Virol., 66, 4803 ± 4813. Lunsford LB, Hromas R and Denny CT. (1993b). Mol. Delattre O, Zucman J, Plougastel B, Desmaze C, Melot Cell. Biol., 13, 7393 ± 7398. T, Peter M, Kovar H, Joubert I, de Jong P, Rouleau Montminy MR, Sevarino KA, Wagner JA, Mandel G and G, Aurias A and Thomas G. (1992). Nature, 359, 162 ± Goodman RH. (1986). Proc. Natl. Acad. Sci., 83, 6682 ± 165. 6686. Fujimura Y, Ohno T, Siddique H, Lee L, Rao VN and Reddy Ohno T, Rao VN and Reddy ES. (1993). Cancer Res., 53, ESP. (1996). Oncogene, 12, 159 ± 167. 5859 ± 5863. Gorman CM, Merlino GT, Willingham MC, Pastan I and Ouchida M, Ohno T, Fujimura Y, Rao VN and Reddy ESP. Howard BH. (1982). Proc. Natl. Acad. Sci., 79, 6777 ± (1995). Oncogene, 11, 1049 ± 1054. 6781. Rabbitts TH. (1994). Nature, 372, 143 ± 149. Hurst HC, Masson N, Jones NC and Lee KAW. (1990). Mol. Ribeiro A, Brown AD and Lee KAW. (1994). J. Biol. Chem., Cell. Biol., 10, 6192 ± 6203. 269, 31124 ± 31128. Hurst HC, Totty NF and Jones NC. (1991). Nucl. Acids Res., Sanchez-Garcia I and Rabbitts, TH. (1994). Proc. Natl. 19, 4601 ± 4609. Acad. Sci., 91, 7869 ± 7873. Jones RH and Jones NC. (1989). Proc. Natl. Acad., Sci., 86, Sorenson PHB, Lessnick SL, Lopez-Terrada D, Liu XF, 2176 ± 2180. Triche TJ and Denny CT. (1994). Nature Genetics, 6, 146 ± Kim J, Lee KAW and Pelletier J. (1997). Oncogene. In press. 151. Kohler PO and Bridson WE. (1971). J. Clin. Endocrinology, Tanaka K, Iwakuma T, Harimaya K, Sato H and Iwamoto 32, 683 ± 687. Y. (1997). Journal of Clinical Investigation, 99, 239 ± 247. Kovar H, Aryee DNT, Jug G, Henockl C, Schemper M, Yi,H-K,FujimuraY,OuchidaM,PrasadDDK,RaoVN Delattre O, Thomas G and Gadner H. (1996). Cell Growth and Reddy ESP. (1997). Oncogene, 14, 1259 ± 1268. & Di€erentiation, 7, 429 ± 437. Zucman J, Delattre O, Desmaze C, Epstein AL, Stenman G, Ladanyi M. (1995). Diagn. Mol. Pathol., 4, 162 ± 173. Speleman F, Fletchers CDM, Aurias A and Thomas G. Ladanyi M and Gerald W. (1994). Cancer Research, 54, (1993a) Nature Genetics, 4, 341 ± 345. 2837 ± 2840. Zucman J, Melot T, Desmaze C, Ghysdael J, Plougastel B, Lee KAW and Masson N. (1993). B.B.A., 1174, 221 ± 233. Peter M, Zucker JM, Triche TJ, Sheer D, Turc-Carel C, Lessnick SL, Braun BS, Denny CT and May WA (1995). Ambros P, Combaret V, Lenoir G, Aurias A, Thomas G Oncogene, 10, 423 ± 431. and Delattre O. (1993b). EMBO. J., 12, 4481 ± 4487.