TEL is a novel member of the ETS family of transcriptional regulators which is frequently involved in human leukemias as the result of specific chromosomal translocations. We show here by co‐immunoprecipitation and GST chromatography analyses that TEL and TEL‐derived fusion proteins form homotypic oligomers in vitro and in vivo. Deletion mutagenesis identifies the TEL oligomerization domain as a 65 amino acid region which is conserved in a subset of the ETS proteins including ETS‐1, ETS‐2, FLI‐1, ERG‐2 and GABPα in vertebrates and PNTP2, YAN and ELG in Drosophila. TEL‐induced oligomerization is shown to be essential for the constitutive activation of the protein kinase activity and mitogenic properties of TEL–platelet derived growth factor receptor β (PDGFRβ), a fusion oncoprotein characteristic of the leukemic cells of chronic myelomonocytic leukemia harboring a t(5;12) chromosomal translocation. Swapping experiments in which the TEL oligomerization domain was exchanged by the homologous domains of representative vertebrate ETS proteins including ETS‐1, ERG‐2 and GABPα show that oligomerization is a specific property of the TEL amino‐terminal conserved domain. These results indicate that the amino‐terminal domain conserved in a subset of the ETS proteins has evolved to generate a specialized protein–protein interaction interface which is likely to be an important determinant of their specificity as transcriptional regulators.
ETS proteins form a class of >20 different sequence‐specific transcriptional regulators involved in the response of cells to a variety of developmental and environmental cues (for review, see Scott et al., 1994; Bories et al., 1995; Hill and Treisman, 1995; Janknecht et al., 1995; Muthusamy et al., 1995; Wassarman et al., 1995). Six members of this family are implicated to date in a variety of oncogenic processes in both animal species and man. c‐ets1 is the cellular homolog of the v‐ets oncogene of avian leukemia virus E26 which, together with a co‐transduced v‐myb oncogene and a partial gag gene, encode the Gag‐Myb‐Ets transforming protein (Leprince et al., 1983; Nunn et al., 1983). Both v‐myb and v‐etscontribute in an interdependent and synergistic way to the transforming and leukemogenic properties of E26 (Metz and Graf, 1991a,b). Spi‐1/PU1 and FLI‐1 are overexpressed in the leukemic cells of the erythroleukemias induced in mice by the Friend virus complex and the Friend helper virus, respectively (Moreau‐Gachelin et al., 1988; Ben‐David et al., 1991). Overexpression of Spi‐1 has been shown to affect various aspects of the differentiation and proliferation control of primary erythroid progenitors and of Friend leukemic cell lines (Schuetze et al., 1993; Delgado et al., 1994; Tran Quang et al., 1995). In man, the carboxy‐terminal domains of FLI‐1, ERG‐2 or ETV‐1 including their ETS domain are fused to the amino‐terminal half of EWS, an RNA binding protein encoded by the EWS gene on chromosome 22, in Ewing sarcomas and related tumors of childhood as the result of specific chromosomal translocations (Delattre et al., 1992; Zucman et al., 1993; Jeon et al., 1995). Similarly, the ERG‐2 DNA binding domain is fused to TLS, a protein related to EWS, in several forms of acute myelogenous leukemias harboring a t(16;21) translocation (Shimizu et al., 1993; Ichikawa et al., 1994). In these cases, substitution of the amino‐terminal moiety of FLI‐1, ERG‐2 or ETV‐1 by either EWS or TLS sequences confers aberrant transcriptional activation properties and activates the transforming properties of these proteins (May et al., 1993a,b; Bailly et al., 1994; Prasad et al., 1994). Finally, a series of leukemias are associated with specific translocations which result in the fusion of the amino‐terminal domain of TEL to a variety of partners. These include the carboxy‐terminal domain of the platelet‐derived growth factor receptor β (PDGFRβ) and c‐ABL tyrosine‐specific protein kinases in chronic myelomonocytic leukemia (CMML) with t(5;12) and acute lymphoid leukemia with t(9;12), respectively (Golub et al., 1994; Papadopoulos et al., 1995); and the AML1/CBFα subunit of core binding factor (CBF), a tissue‐specific transcriptional regulator, in childhood B‐cell acute lymphoid leukemia with t(12;21) (Golub et al., 1995; Romana et al., 1995). TEL is also involved in reciprocal t(12;22) chromosomal translocations involving MN1, a gene of unknown function (Buijs et al., 1995).
The common feature of ETS proteins is a domain of 85 amino acid residues (ETS domain) which is structured as a winged helix–turn–helix motif responsible for their binding to specific sequences centered over a conserved GGAA/T core motif (Donaldson et al., 1994, 1996; Liang et al., 1994; Kodandapani et al., 1996). Moreover, a subset of ETS proteins, including ETS‐1, ETS‐2, ERG‐2, FLI‐1, GABPα and TEL in vertebrates and PNTP2, ELG and YAN in Drosophila also share a region of homology of ∼65 amino acid residues which has been referred to diversely as the B domain (Boulukos et al., 1989), the pointed domain (Klambt, 1993) or the HLH domain because of a loose resemblance to the helix–loop–helix motif of other transcriptional regulators (Seth and Papas, 1990). Although deletion analyses indicate that the amino‐terminal conserved region of ETS‐1, ETS‐2 and FLI‐1 modulates the transcriptional properties of these proteins (Gegonne et al., 1992; Schneikert et al., 1992; Rao et al., 1993), its precise mode of action remains unclear.
The TEL–PDGFRβ and TEL–ABL fusion proteins are formed by the fusion of the 154 amino‐terminal residues of TEL, including the conserved amino‐terminal domain (B domain), to either the transmembrane and catalytic domain of PDGFRβ or the SH2, SH3 and catalytic domains of c‐ABL, respectively (Golub et al., 1994; Papadopoulos et al., 1995). Since normal ligand‐induced activation of PDGFRβ depends upon ligand‐induced dimerization (for review, see Heldin, 1995), it has been suggested that the TEL‐derived sequence might contribute to the oncogenic properties of TEL–PDGFRβ by favoring constitutive dimerization and activation of the protein kinase activity of the PDGFRβ moiety of the fusion protein (Golub et al., 1994). Likewise, TEL‐induced oligomerization might activate the protein kinase activity (Papadopoulos et al., 1995) and transforming properties of TEL–ABL in a way reminiscent of the BCR‐induced oligomerization of the BCR–ABL fusion protein of Philadelphia‐positive human leukemias (McWhirter et al., 1993).
Here, we show that the amino‐terminal conserved domain of TEL forms a homotypic interaction interface which mediates oligomerization of TEL and TEL‐derived fusion proteins. This property appears to be specific to TEL, since the amino‐terminal conserved domain of other vertebrate proteins fails to mediate oligomerization. TEL‐induced oligomerization is shown to play an essential role in the activation of the protein kinase activity and mitogenic property of TEL–PDGFRβ.
To determine whether the 154 amino‐terminal residues of TEL encode an oligomerization interface, SV40 early promoter‐based expression plasmids encoding intact TEL–PDGFRβ (TP‐0) or a hemagglutinin (HA) epitope‐tagged version of a mutant derived thereof by deletion of the 448 carboxy‐terminal residues, HA‐TP‐0 (tt), were transfected either alone or co‐transfected in HeLa cells (see Figure 1 for a schematic of the proteins used in this study). Transfected cells were metabolically labeled with l‐[35S]methionine and l‐[35S]cysteine, and the respective proteins were immunoprecipitated from cell lysates using specific antibodies. The results of Figure 2 show that a monoclonal antibody to the HA epitope specifically precipitated HA‐TP‐0 (tt) (Figure 2A, compare lanes 5 and 6) whereas a rabbit antiserum directed to the carboxy‐terminal end of PDGFRβ only precipitated TP‐0 (Figure 2A, compare lanes 3 and 4). In contrast, immunoprecipitation of lysates of cells co‐expressing both proteins with either of these antibodies resulted in the co‐precipitation of TP‐0 and HA‐TP‐0 (tt) (Figure 2A, lanes 7 and 8). Similar results were obtained when these proteins were co‐expressed following in vitro translation of their respective mRNAs in rabbit reticulocyte lysates and immunoprecipitated with the same set of antibodies (Figure 2B, lanes 4–9). We conclude from these experiments that TP‐0 and truncated forms derived thereof form a complex in vivo and in vitro.
To analyze whether the TEL moiety is required for this association, a series of truncated polypeptides that contained TEL amino‐ and carboxy‐terminal deletions was generated (see Figure 1A). The ability of these polypeptides to interact with intact TP‐0 was analyzed by co‐immunoprecipitation assays using antibodies specific of each of them (Figure 3A and B, lanes 1–6). TEL amino acids 1–53 were found to be dispensable for complex formation since their deletion in HA‐TP‐3 (t) did not affect the ability of this protein to associate with TP‐0 (Figure 3A, lanes 7 and 8). In contrast, further deletion of 65 amino acid residues in HA‐TP‐4 (t) or complete deletion of TEL‐derived sequences in HA‐TP‐5 (t) abolished the ability of these proteins to co‐precipitate with TP‐0 (Figure 3A, lanes 9–12), indicating that complex formation is dependent upon specific sequences in the TEL domain. Analysis of TEL–PDGFRβ mutants carrying carboxy‐terminal deletions in the TEL moiety showed that removal of TEL amino acid residues 119–154 (TP‐11) did not affect complex formation with HA‐TP‐0 (t) (Figure 3B, lanes 7 and 8), whereas further deletion into the central, conserved domain of TEL in TP‐13 abolished it (Figure 3B, lanes 9 and 10). The essential role of the conserved domain of TEL in complex formation is borne out further by the inability of a mutant TEL–PDGFRβ deleted in this domain (TP‐10) to associate with HA‐TP‐0 (t) (Figure 3B, lanes 11 and 12). Moreover, fusion of the TEL conserved domain (amino acid residues 54–119 of TEL) to 132 residues of PDGFRβ sequences, HA‐TP‐12 (t), is sufficient to engage these proteins in complex formation with TP‐0 in vitro (Figure 3C, lanes 3 and 4; compare also with Figure 3A, lanes 9 and 10). Complex formation between TP‐0 and HA‐TP‐12 (tt) was also observed in vivo in HeLa cells transfected with the respective expression plasmids (data not shown).
The studies reported so far were carried out using TEL deletion mutants made in the context of the TEL–PDGFRβ fusion protein. To determine whether the amino‐terminal conserved domain of TEL also mediates complex formation of TEL and TEL‐derived nuclear proteins, we analyzed the ability of intact TEL and TEL derivatives to associate in vivo and in vitro. Expression plasmids encoding an HA‐tagged version of TEL or an ETS‐1 chimeric protein in which the 129 amino‐terminal residues of ETS‐1 are replaced by the corresponding residues of TEL, TEL(1–119)/ETS‐1(130–441) (see Figure 1C for a schematic of these constructs), were transfected either alone or co‐transfected in HeLa cells. Cells were metabolically labeled with l‐[35S]methionine and l‐[35S]cysteine and the respective proteins were immunoprecipitated using antibodies specific for each of them (Figure 4A, lanes 3–6). The results of Figure 4A show that an antibody specific for ETS‐1 was able to precipitate TEL only when it was co‐expressed with the chimeric TEL(1–119)/ETS‐1(130–441) protein (Figure 4A, compare lanes 4, 6 and 8). Conversely, in anti‐HA immunoprecipitates, the TEL(1–119)/ETS‐1(130–441) protein was detected only when it was co‐expressed with HA‐TEL (Figure 4A, compare lanes 3, 5 and 7). Complex formation between TEL(1–119)/ETS‐1(130–441) and TEL is dependent upon the presence of TEL sequences since HA‐ETS‐1(130–441) was found to be unable to co‐precipitate with HA‐TEL (Figure 4B, compare lanes 2, 4 and 6). To analyze whether the conserved domain of TEL is sufficient to mediate complex formation, we co‐expressed HA‐TEL with a TEL–ETS‐1 chimera limited to this domain of TEL, HA‐TEL(54–119)/ETS‐1(130–441), and analyzed the association of these proteins using antibodies specific for their respective carboxy‐terminal domains (ETS domains). The results in Figure 4B show that immunoprecipitation of cell lysates co‐expressing both proteins with either the ETS‐1‐specific antibody or the TEL‐specific antibody resulted in the co‐precipitation of HA‐TEL and HA‐TEL(54–119)/ETS‐1(130–441) (Figure 4B, compare lanes 1 and 7–10). The same results were obtained using in vitro translated proteins in reticulocyte lysates (data not shown). We conclude from these experiments that the TEL amino‐terminal conserved domain is essential to engage TEL in complex formation in vivo and in vitro independently of its protein context.
To analyze whether this domain alone is by itself sufficient for complex formation, we generated an HA‐TEL miniprotein restricted to this domain, HA‐TEL(54–119), and analyzed its ability to co‐precipitate with either TP‐0 (t) or TEL(1–119)/ETS‐1(130–441) in vitro. As shown in Figure 4C, immunoprecipitation of TP‐0 (t) or TEL(1–119)/ETS‐1(130–441) with the anti‐HA specific antibody was only observed when co‐expressed with HA‐TEL(54–119) (Figure 4C, compare lanes 1–5). To determine whether the TEL amino‐terminal conserved domain is able to interact with itself, a GST fusion protein containing amino acids 54–119 of TEL was immobilized on glutathione–agarose beads and incubated with in vitro translated l‐[35S]methionine‐labeled TP‐0 or HA‐TEL (54–119). As shown in Figure 5, both TP‐0 (Figure 5A, lanes 6–10) and HA‐TEL(54–119) (Figure 5B, lanes 6–10) bound to the GST–TEL(54–119) affinity matrix. In contrast, neither GST alone nor a GST fusion protein containing the amino‐terminal conserved domain of ETS‐1 (data not shown) was found to be able to bind TP‐0 (Figure 5A, lanes 1–5) and HA‐TEL(54–119) (Figure 5B, lanes 1–5). We conclude from these experiments that the conserved amino‐terminal of TEL is sufficient for oligomerization.
The domain of TEL responsible for oligomerization is phylogenetically conserved in a subset of the ETS family members which include, in vertebrates, the highly related ETS‐1 and ETS‐2, the related FLI‐1 and ERG‐2 and the α subunit of GABP and, in Drosophila, PNTP2, ELG and YAN. To analyze whether the corresponding domains of other vertebrate ETS proteins are also involved in complex formation, TEL–PDGFRβ fusion proteins were created in which the conserved domain of TEL was substituted for those of ETS‐1, ERG‐2 or GABPα to generate TP‐7, TP‐8 and TP‐9 (see Figure 1B for a schematic of these mutants). The properties of a TEL–PDGFRβ fusion protein (TP‐6) in which the 119 amino‐terminal residues of TEL were replaced by the first 130 amino acid residues of ETS‐1 were also analyzed. These fusion proteins were co‐translated in vitro with ETS‐1, ERG‐2 and GABPα, respectively, and complex formation analyzed by co‐immunoprecipitation analyses using specific antisera (Figure 6A, lanes 1–4 and 7–8; B, lanes 1–4 and 7–10). The results of Figure 6 show that in conditions in which intact TEL–PDGFRβ co‐immunoprecipitated with TEL (Figure 6C), none of the fusion proteins in which the TEL amino‐terminal conserved region was exchanged for that of ETS‐1 (Figure 6A, lanes 5–6 and 9–10), ERG‐2 (Figure 6B, lanes 5–6) or GABPα (Figure 6B, lanes 11–12) were able to form homotypic complexes as evidenced by their failure to co‐precipitate with ETS‐1, ERG‐2 or GABPα, respectively. Also, unlike intact TP‐0, neither TP‐6, TP‐7, TP‐8 nor TP‐9 was found to be able to associate with TEL (data not shown). We conclude from these experiments that the ability to oligomerize is a specific property of the amino‐terminal conserved domain of TEL and that TEL‐induced oligomerization is homotypic. Intact TEL–PDGFRβ (TP‐0) was found to localize to the cytoplasm of transfected cells (Figure 7C, compare panels 1 and 2); this property is independent of the presence of the oligomerization domain since no difference in subcellular localization was observed between TP‐0 and TP‐10 (Figure 7C, compare panels 2 and 3).
The mitogenic and chemotactic properties of PDGFRβ are dependent upon PDGF‐induced dimerization and subsequent activation of its tyrosine‐specific protein kinase activity (for review, see Heldin, 1995). To analyze whether TEL‐mediated oligomerization is involved in the activation of the tyrosine‐specific protein kinase activity of TEL–PDGFRβ, we compared the autophosphorylation properties of TP‐0 and of a series of deletion or substitution mutants derived thereof which either maintain or abolish the ability to oligomerize. Proteins were translated in reticulocyte lysates and aliquots of each lysate were electrophoresed on polyacrylamide gels and, after transfer to nitrocellulose, the blots were revealed with either an antibody to the carboxy‐terminal end of PDGFRβ to compare for expression levels or with a phosphotyrosine‐specific antibody to analyze for autophosphorylation of the PDGFRβ‐derived moiety. The results of Figure 7B show that all fusion proteins tested were expressed at a similar level. Analysis of the same blot with an anti‐phosphotyrosine antibody revealed a high level of tyrosine phosphorylation associated with TP‐0 (Figure 7A, lane 1). This reflects TEL–PDGFRβ autophosphorylation since a TEL–PDGFRβ mutant carrying a K634A substitution in the ATP binding site, a mutation known to inactivate the intrinsic protein kinase activity of PDGFRβ (Williams, 1989), failed to react with the anti‐phosphotyrosine antibody when tested in the same conditions (data not shown). All mutants of TEL–PDGFRβ which scored as oligomerization‐defective in our co‐precipitation assays, including the deletion mutants HA‐TP‐4, HA‐TP‐5, TP‐10 and TP‐13 (Figure 7A, lanes 2 and 4–6) and the substitution mutants TP‐6 (Figure 7A, lane 3), TP‐7, TP‐8 and TP‐9 (data not shown), were also found to lack protein kinase activity. In contrast, all TEL–PDGFRβ mutants which maintain their ability to oligomerize, including TP‐11, HA‐TP‐3 and HA‐TP‐12, also maintain a high level of protein kinase activity (Figure 7A, lanes 7–9). We conclude from these experiments that oligomerization mediated by the TEL conserved amino‐terminal domain is instrumental in the constitutive activation of the TEL–PDGFRβ protein tyrosine kinase activity.
To analyze whether the TEL‐dependent oligomerization of TEL–PDGFRβ is important to its mitogenic properties, we made use of the ability of the interleukin 3 (IL‐3)‐dependent Ba/F3 murine hematopoietic cell line (Palacios and Steinmetz, 1985) to become independent of IL‐3 for survival and proliferation upon expression of several receptor and non‐receptor protein tyrosine kinases or cytokine receptors (Mathey‐Prevot et al., 1986; Daley and Baltimore, 1988; Daley et al., 1992; McWhirter and Wang, 1993; Damen, 1995; Gobert et al., 1995). The cDNAs encoding TP‐0 as well as oligomerization‐defective mutants derived thereof by either deletion of TEL sequences, including HA‐TP‐5 and TP‐10, or by substitution of TEL sequences by the corresponding sequence of ETS‐1 (TP‐6) were subcloned into the pBabeNeo retroviral expression vector (Morgenstern and Land, 1990). These constructs as well as the pBabeNeo control were electroporated into Ba/F3 cells and transfected cells were selected in IL‐3‐containing growth medium in the presence of G418. Immunoprecipitation analyses of l‐[35S]methionine‐ and l‐[35S]cysteine‐labeled cells showed that all transfectants expressed the expected proteins (Figure 8A). As expected from our previous analyses using in vitro translated proteins, only the intact TEL–PDGFRβ (TP‐0) was active as a protein tyrosine kinase (Figure 8B, lane 2) whereas oligomerization‐defective forms were inactive (Figure 8B, lanes 3–5). To assay for IL‐3 independence, transfectants were switched to growth medium without IL‐3. Whereas the control transfectants and those expressing the oligomerization‐defective TEL–PDGFRβ mutants were found to be unable to proliferate in these conditions, transfectants expressing TP‐0 could proliferate in an IL‐3‐independent fashion (Figure 8C). Ba/F3 cells expressing TP‐0 were found to proliferate in the absence of IL‐3 for a period of >1 month at a rate which is, however, reduced as compared with that observed in the presence of IL‐3 (Figure 8C). The control Ba/F3 cell line as well as all transfectants were found to proliferate at the same rate in the presence of IL‐3 (data not shown). Supernatants from TP‐0 cells failed to support the growth factor‐independent proliferation of Ba/F3 cells, indicating that TP‐0‐induced proliferation does not result from induction of growth stimulatory cytokines but is a direct consequence of TEL–PDGFRβ signaling. We conclude from these experiments that TEL‐induced oligomerization of TEL–PDGFRβ is essential for its activity as a protein tyrosine kinase and its mitogenic properties in vivo.
The results presented in this study show that the amino‐terminal conserved domain of TEL (amino acid residues 54–119) is a homotypic oligomerization domain as assessed by co‐immunoprecipitation analyses and by affinity chromatography using a specific GST matrix. TEL‐induced oligomerization is an autonomous property of this domain since a miniprotein corresponding to the TEL oligomerization domain was found to engage in complex formation with itself, with TEL and with TEL fusion proteins maintaining an intact amino‐terminal conserved domain.
The primary amino acid sequence of the oligomerization domain of TEL is highly conserved in a subset of the ETS family members, implying a conserved structural fold (Figure 9). Previous alignment studies have proposed this region to display similarity to the HLH motif of the basic‐HLH (bHLH) proteins (Seth and Papas, 1990). To gain further insight into the possible secondary structure of this domain, the primary sequence of the amino‐terminal conserved domains of ETS‐1, ETS‐2, ERG‐2, FLI‐1, GABPα and TEL, together with those of Drosophila melanogaster PNTP2, YAN and ELG and a sea urchin ETS protein, were aligned using the CLUSTAL W multiple alignment program (Thompson et al., 1994). Secondary structure predictions were performed using the SIMPA 95 program (Levin and Garnier, 1988; J.Levin, in preparation). As shown in Figure 9, two helices are strongly predicted at each end of the domain, separated by a region with a low β‐strand folding probability. The helix predicted at the carboxy‐terminal end of the domain co‐localizes with helix 2 of the presumed HLH. In contrast, the proposed region of similarity with helix 1 of bona fide bHLH proteins is not predicted to fold as an α‐helix in the multiple alignment analysis. Furthermore, this region shows additional features which make it unlikely to resemble helix 1 of bHLH proteins. First, a phenylalanine residue (F129 in MyoD1; Ma et al., 1994) which is conserved in the middle of helix 1 of bona fide bHLH proteins is substituted by glycine in all ETS proteins (Seth and Papas, 1990; see Figure 9). Finally, FLI‐1, ERG‐2 and PNTP2 show an extra amino acid residue in the middle of this region when compared with other ETS proteins or bHLH proteins (Seth and Papas, 1990; see Figure 9).
The non‐conservation in ETS proteins of the hydrophobic residues necessary for stabilization of the putative helix 1 as well as the insertion observed in that region for some ETS proteins, together with our secondary structure prediction, suggest that this region is unlikely to fold as an α‐helix. In the absence of experimental data about the structure adopted by this domain in either a monomeric or oligomeric state, its description as an HLH domain clearly appears to be premature.
The conserved fold of the amino‐terminal domain of ETS proteins is likely to underlie a conserved function. Our results show, however, that the conserved amino‐terminal domains of ETS‐1, ERG‐2 and GABPα are not homotypic oligomerization domains since they failed to replace the conserved amino‐terminal domain of TEL in inducing oligomerization when analyzed in an identical setting (Figure 6). These data are in accordance with other experimental approaches which have led to the conclusion that ETS‐1 and ETS‐2 are monomeric in solution (for review, see Wasylyk et al., 1993; our unpublished observations). We interpret these observations as suggesting that the amino‐terminal conserved domain has evolved in different ETS proteins to generate a specialized protein–protein interaction interface which is likely to make an important contribution to their specificity. In line with this notion, the amino‐terminal conserved domains of the closely related FLI‐1 and ERG‐2 proteins and of the ETS protein subclass formed by ETS‐1, ETS‐2 and PNTP2 are included in a more extended, subclass‐specific, region of homology of ∼200 and 100 amino acid residues, respectively. In the case of PNTP2, ETS‐1 and ETS‐2, this extended region of homology includes a conserved threonine residue which is the target for phosphorylation by members of the MAP kinase family, a modification which recently has been shown to play an essential role in the physiological and transcriptional response of these proteins to specific environmental signals involving MAP kinases as intracellular signaling components (Brunner et al., 1994; O'Neil et al., 1994; Rabault et al., 1996; Yang et al., 1996). Specificity and diversity in transcriptional regulation by ETS proteins is highly dependent upon their assembly into multiprotein complexes and in their synergy with other cis‐linked transcriptional regulators (Wasylyk et al., 1990; Hipskind et al., 1991; Dalton and Treisman, 1992; Pongubala et al., 1992; Gegonne et al., 1993; Wotton et al., 1993; Giese et al., 1995; John et al., 1995; Sieweke et al., 1996). In many of these situations, the domains involved in the multiple contacts made by ETS proteins with other transcriptional regulators have not been identified. As ETS‐1, ETS‐2, FLI‐1, ERG‐2 and GABPα display a similar DNA binding specificity (Brown and McKnight, 1992; Nye et al., 1992; Woods et al., 1992), the specialized protein–protein interaction interface provided by their amino‐terminal conserved domain could play an important role in governing either the specificity of the assembly or the operation of ETS‐containing multiprotein complexes.
Like other ETS family members, TEL displays sequence‐specific DNA binding properties towards consensus ETS binding sites in vitro (Poirel et al., 1997). Unlike other ETS proteins, however, TEL is a homodimer or higher order oligomer, a property which might favor its binding to promoters/enhancers displaying tandemly arranged ETS binding sites and contribute, therefore, to its specificity with respect to other members of the family. This situation is reminiscent of the DNA binding properties of GABP and its human E4TF1/NRF‐2 homolog (LaMarco et al., 1991; Watanabe et al., 1993) in which an ETS family α subunit is complexed to either of two isoforms of a β subunit. One of these isoforms include a carboxy‐terminal dimerization domain which allows the formation of α2β2 tetramers which display preferred DNA binding activity to tandem ETS binding sites (Thompson et al., 1991; Brown and McKnight, 1992; Virbasius et al., 1993; Gugneja et al., 1995). Experiments are in progress to determine whether oligomerization of TEL is involved in the regulation of its DNA binding activity and DNA binding specificity.
Leukemic cells of childhood B cell acute lymphoblastic leukemias with t(12;21) express a specific TEL–AML1 fusion protein in which the 333 amino‐terminal residues of TEL are fused to the N‐terminus of AML1 (Golub et al., 1995; Romana et al., 1995). TEL–AML1 has been shown recently to interfere with the AML1B‐dependent activation of the T‐cell receptor β enhancer in transient transfection assays. Furthermore, a region including the TEL oligomerization domain identified here was found to be essential for TEL–AML1‐mediated repression (Hiebert et al., 1996). We found that substitution of the amino‐terminal conserved domain of ETS‐1 by that of TEL in TEL(1–119)/ETS‐1(130–441) and HA‐TEL(54–119)/ETS‐1(130–441) resulted in the conversion of ETS‐1 from an activator into a repressor of ETS‐1‐responsive model reporter genes (data not shown). These observations suggest that the TEL conserved amino‐terminal domain either encodes a bona fide repression domain or acts by masking the AML1‐B and ETS‐1 activation domains through oligomerization. TEL is more closely related to D.melanogaster YAN/POK than to any other member of the ETS family in both its DNA binding and amino‐terminal conserved domains. Genetically, YAN has been shown to be a negative regulator of R7 photoreceptor cell development, acting antagonistically to the Sevenless/Ras/MAP kinase proneural differentiation pathway (O‘Neil et al., 1994). Biochemical evidence indicates that YAN can function as a repressor of ETS‐responsive reporter genes and that phosphorylation of YAN by Rolled/MAP kinase on several residues has been proposed to down‐regulate its activity by decreasing its stability and by altering its subcellular localization (O'Neil et al., 1994; Rebay and Rubin, 1995). In line with its close structural relatedness to the TEL oligomerization domain, we found that the amino‐terminal conserved domain of YAN also appears to be able to induce homotypic oligomerization, although the stability of YAN oligomers is significantly weaker than that of TEL oligomers (data not shown). It would be interesting to analyze whether the conserved amino‐terminal domain of YAN is important for its function in the Sevenless signaling pathway and, if so, whether it can be substituted by that of TEL. Like YAN, TEL contains several potential MAP kinase phosphorylation sites and is a phosphoprotein in vivo (Poirel et al., 1997). It is possible, therefore, that both YAN and TEL normally function as signal‐regulated transcriptional repressors and that this activity depends upon their oligomerization properties.
The result of the t(5;12) chromosomal translocation in the leukemic cells of CMML is the expression of a fusion protein in which the extracellular ligand binding domain of PDGFRβ is substituted by the first 154 amino acid residues of TEL. Our results show that deletion of the PDGFRβ extracellular domain (mutant TP‐5) is not sufficient to explain the constitutive autophosphorylation and the mitogenic properties of TEL–PDGFRβ. They support instead the notion that these properties are critically dependent upon TEL‐mediated homotypic oligomerization of TEL–PDGFRβ. TEL–PDGFRβ forms oligomers in stably transfected Ba/F3 cells (data not shown), and transformation by TEL–PDGFRβ is likely to be linked to the intermolecular cross‐phosphorylation of each subunit in the oligomer on specific tyrosine residues, resulting in further activation of the catalytic activity of TEL–PDGFRβ and in the generation of specific binding sites for critical substrates containing SH2 domains. The protein kinase activity of TEL–PDGFRβ is essential to its mitogenic properties since a tyrphostin inhibitor specific for the PDGF receptor kinase specifically inhibits the IL‐3‐independent proliferation in Ba/F3 cells (A.Ostman, J.Ghysdael and C.H.Heldin, in preparation). In that respect, activation of TEL–PDGFRβ appears to recapitulate in a constitutive manner the activation of PDGFRβ which follows ligand binding (Bishayee et al., 1989; Heldin et al., 1989). It is also reminiscent of the oncogenic activation of other receptor‐type protein kinases as the result of constitutive oligomerization either by mutation, as in the case of c‐ErbB‐2/Neu (Bargmann et al., 1986; Weiner et al., 1989), or by fusion to unrelated oligomerization domains including the TPR sequences of TPR–MET and TPR–TRK (Greco et al., 1992; Rodrigues and Park, 1993) or the regulatory subunit of cAMP‐dependent protein kinase in PTC‐RET (Grieco et al., 1990; Bongarzone et al., 1993).
Several observations suggest that PDGFRβ signaling could play a role in hematopoietic cell proliferation. First, low level expression of PDGFRβ has been observed in murine myeloblastic, mast and multipotential cell lines (de Parseval et al., 1993) and during the differentiation of primary monocytes and of cell lines THP1 and HL60 (Pantazis et al., 1991; Inaba et al., 1993). Second, PDGF treatment in several of these situations has been shown to induce a mitogenic response in both immature and mature cells (de Parseval et al., 1993; Inaba et al., 1993). It is possible, therefore, that the mitogenic properties of TEL–PDGFRβ might be mimicking, in a constitutive manner, normal PDGFRβ signaling. The situation is probably more complex since our immunofluorescence studies (Figure 7C) and subcellular fractionation experiments (data not shown) show that TEL–PDGFRβ is essentially cytoplasmic and is therefore likely to access or phosphorylate protein substrates other than those normally involved in PDGFRβ signaling.
Recent evidence has shown that optimal proliferation of Ba/F3 cells in response to cytokine requires at least three distinct and independent signaling pathways (Miyazaki et al., 1995). Constitutive activation of any of two of these pathways is sufficient to induce cytokine‐independent proliferation of Ba/F3 cells, albeit at a reduced rate. Expression of TEL–PDGFRβ in Ba/F3 cells was found to induce their proliferation at a lower rate as compared with Ba/F3 cells maintained in IL‐3, suggesting that TEL–PDGFRβ only activates a subset of the signaling pathways required for maximal proliferation. TEL–PDGFRβ‐expressing Ba/F3 cells could, therefore, provide a useful system to identify other pathways which can cooperate with TEL–PDGFRβ‐dependent signaling to induce proliferation. Identification of the components of these pathways could help to identify the molecular events which cooperate with TEL–PDGFRβ in human leukemias.
Materials and methods
Construction of recombinants
To generate the HA epitope‐tagged version of TP‐0, the cDNA encoding TEL–PDGFRβ (Golub et al., 1994) was inserted into EcoRI‐ and HindIII‐restricted M13mp18. The single‐stranded M13 recombinant DNA was used as substrate for oligonucleotide site‐directed mutagenesis to create a SalI restriction enzyme site immediately downstream of the initiator methionine codon (mutagenic primer M1). The insert was recovered from the corresponding replicative form (RF) by digestion with SalI and HindIII and subcloned between the XhoI and HindIII sites of the previously described SV40 early promoter‐based expression plasmid ΔEB‐HA (Rabault and Ghysdael, 1994). A pBluescript KS(+) allowing the T3‐driven expression of the 3′‐truncated HA‐TP‐0 (t) was obtained by digestion of ΔEB‐HA‐TP‐0 with EcoRI and ApaI and subcloning into the EcoRI‐ and ApaI‐restricted pBluescript KS(+). This truncated form of TP‐0 only included the 133 amino‐terminal residues of the PDGFRβ‐derived domain of TEL–PDGFRβ. Mutant TP‐1 was constructed from the same M13 DNA template using the mutagenic primer M2 to create a BglII restriction enzyme site at nucleotide position 382 of TEL. These mutations resulted in a His→Gln substitution at codon 119 of TEL. The RF containing the mutagenized insert was digested with EcoRI and ApaI and subcloned into EcoRI‐ and ApaI‐restricted pBluescript KS(+) to generate TP‐1 (t). Mutant TP‐2 was created by site‐directed mutagenesis of the M13 DNA template containing TP‐1 to create a BamHI restriction enzyme site at nucleotide position 179 using the mutagenic primer M3. These mutations resulted in Ala→Gly and His→Ile substitutions at codons 52 and 53 of TEL, respectively. The substitution in TP‐2 did not affect its oligomerization properties as compared with TP‐0 (data not shown). The RF containing the mutagenized insert was digested with EcoRI and ApaI and subcloned into EcoRI‐ and ApaI‐restricted pBluescript KS(+) to generate TP‐2 (t). Mutant HA‐TP‐3 (t) was obtained by substitution of the EcoRI–BamHI fragment of pBS‐TP‐2 (t) with the EcoRI–BglII fragment of ΔEB‐HA. Mutant HA‐TP‐4 (t) was obtained by substitution of the EcoRI–BglII fragment of pBS‐TP‐2 (t) with the EcoRI–BglII fragment of ΔEB‐HA. Mutant HA‐TP‐5 (t) was obtained by insertion of a BglII‐bordered insert obtained by PCR amplification of PDGFRβ sequences encoding amino acid residues 527–660 into BglII‐restricted ΔEB‐HA. The amplimers used for PCR amplification were 5′‐GAAGATTCTTCGGGCCCAAGGTGACTCATGATCTCT‐3′ (5′ amplimer) and 5′‐GGAGATCTTCTCCTTGCCCTTTAAGGTGGTGGTG‐3′ (3′ amplimer). The HA‐tagged TP‐5 (t) was subcloned via EcoRI and ApaI into EcoRI‐ and ApaI‐restricted pBluescript SK(+). To generate TP‐6 (t), the EcoRI–BglII fragment obtained from the 5′ part of a chicken c‐ets1 cDNA (Boulukos et al., 1988) was inserted into the EcoRI‐ and BglII‐restricted pBS‐TP‐1 (t). Mutants TP‐7 (t), TP‐8 (t) and TP‐9 (t) were obtained by substitution of the BamHI–BglII fragment of TP‐2 (t) by fragments bordered by in‐frame BamHI (5′ side) and BglII (3′ side) generated by PCR amplification of the chicken ETS‐1 cDNA (Boulukos et al., 1988), human ERG‐2 cDNA (Reddy et al., 1987 a generous gift of Dr E.Reddy) and mouse GABPα cDNA (LaMarco et al., 1991 a generous gift of Drs Brown and S.McKnight), respectively. The amplimers used for PCR amplification were: ETS‐1 5′, 5′‐CAGGATCCTCCCCAAAGATCCCCAGCAGTG‐3′; ETS‐1 3′, 5′‐GGAGATCTTCTCCAGGTGTTCCCAAAGGATATC‐3′; ERG 5′, 5′‐CAGGATCCTGCCAGCAGATCCTACGCTATGGAGT‐3′; ERG 3′, 5′‐GGAGATCTGGTGGAGATGTGAG AGAAGGATGTCG‐3′; GABPα 5′, 5′‐CAGGATCCTCCCCTATGATCCTATACGCTGG‐3′; GABPα 3′, 5′‐GGAGATCTGCTCCAGATGACTCCAAAGAATTTCTCC‐3′. Mutant TP‐10 (t) was obtained by BamHI and BglII digestion of pBS‐TP‐2 (t) followed by religation. To generate mutant TP‐11 (t), the BglII–ApaI fragment of pBS‐TP‐2 (t) was substituted by the corresponding fragment of pBS‐HA‐TP‐5 (t). Mutant HA‐TP‐12 (t) was constructed by substitution of the BglII–ApaI fragment of pBS‐HA‐TP‐3 (t) by the BglII–ApaI fragment of HA‐TP‐5 (t). Mutant TP‐13 (t) was constructed by substitution of the BglII–ApaI fragment of pBS–TP–2 (t) with the BglII–ApaI fragment of pBS‐HA‐TP‐5 (t). All mutants were also constructed so as to contain the full‐length PDGFRβ‐derived region of TEL–PDGFRβ by insertion of the EcoRI–SacII fragment obtained from each mutant into the EcoRI‐ and SacII‐restricted HA‐TP‐0.
The HA‐TP‐0 and HA‐TP‐12 were digested with EcoRI and SacI, and subcloned into EcoRI‐ and SacI‐restricted ΔEB‐HA to generate ΔEB‐HA‐TP‐0 (tt) and ΔEB‐HA‐TP‐12 (tt) respectively. pBS SK+ allowing the T3‐driven expression of chicken ETS‐1 was constructed by inserting the chicken c‐ets1 cDNA insert (Boulukos et al., 1988) at the EcoRI of pBS SK+. pBS SK+ allowing expression of HA‐TEL was obtained by subcloning the EcoRI–HindIII HA‐TEL fragment of ΔEB‐HA‐TEL encoding the full‐length human TEL fused to the HA epitope at its amino‐terminus will be described elsewhere. To generate TEL(1–119)/ETS‐1(130–441) and HA‐TEL(54–119)/ETS‐1(130–441), the XbaI and BglII fragments of TP‐2 (t) and HA‐TP‐12 (t) were subcloned into a XbaI‐ and BglII‐restricted pBS SK+ ETS‐1, respectively. HA‐ETS‐1(130–441) was obtained by substitution of the XbaI and BglII fragment of pBS SK+ ETS‐1 with the XbaI and BglII HA fragment of HA‐TP‐0 (t). The eukaryotic expression vectors for these mutants were obtained by EcoRI digestion of the corresponding pBS SK+ plasmids and subcloning at the EcoRI site of the SV40 early promoter‐based expression plasmid ΔEB (Boulukos et al., 1989). To construct the pBabeNeo derivatives, pBS SK+ encoding TP‐0, HA‐TP‐5, TP‐6 and TP‐10 were digested with EcoRI and SalI and the resulting fragments were subcloned into the EcoRI‐ and SalI‐restricted pBabeNeo (Morgenstern and Land, 1990). To obtain the GST–TEL (54–119) fusion protein, the cDNA corresponding to the BamHI–BglII fragment of the TP‐2 (t) construct was subcloned at the BamHI site of the pGEX‐3X vector (Pharmacia Biotech). The resulting plasmid encoded a protein containing a GST sequence fused to amino acid residues 54–119 of TEL. The sequences of fragments obtained by oligonucleotide site‐directed mutagenesis and PCR amplification were checked by sequencing using the dideoxy sequencing method. The sequences of the mutagenic primers used in site‐directed mutagenesis were as follows: M1, 5′‐GGAGTCTCAGACATGGTCGACAGCGAGAGAGATCAGG‐3′; M2, 5′‐CGAGGTTTCCTCTGCTTCAAGATCTGCTGAAGGAGTTCATAGAGC‐3′; M3, 5′‐CCAGTAAATTGGC TGCAAGCGCAGGAT CCCAGGCAGGCGGATCGAGTCTTCC‐3′.
Transient transfection and immunoprecipitation analyses
HeLa cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum (FCS). For transient expression studies, 7×105 cells were seeded in 100 mm Petri dishes and transfected 24 h after with 5 μg of the SV40 early promoter‐based ΔEB expression plasmid encoding the indicated proteins, using the calcium phosphate co‐precipitation method. After overnight incubation, culture media were changed and cells incubated for an additional 36 h before being processed for metabolic labeling.
Ba/F3 cells were maintained in RPMI‐1640 medium containing 10% FCS and 5% conditionned medium from the WEHI‐3B cell line as a source of IL‐3. Cells were transfected by electroporation as described previously (Dusanter‐Fourt et al., 1994).
For metabolic labeling with radioactive amino acids, transiently transfected HeLa and stably transfected Ba/F3 cells (107 cells) were rinsed once in methionine‐ and cysteine‐free Eagle's medium (Flow Laboratories) and incubated for the indicated times in the same medium supplemented with 1% dialyzed FCS and 500 μCi/ml of a l‐[35S]methionine and l‐[35S]cysteine mixture (Promix, Amersham). Cells were washed twice in phosphate‐buffered saline, lysed in 3 ml of RIPA buffer [Tris–HCl 10 mM, pH 7.4; NaCl 0.1 M; EDTA 0.001 M; Triton X‐100 1%; sodium deoxycholate 0.5%; sodium dodecylsulfate 0.1%; aprotinin (Sigma) 1%; phenylmethylsulfonyl fluoride (PMSF) 100 μg/ml and leupeptin (Sigma) 0.1%] and centrifuged at 100 000 g for 30 min.
Immunoprecipitation analyses were performed using a rabbit antiserum to the DNA binding domain of ETS‐1 (Ghysdael et al., 1986), a rabbit antiserum to the DNA binding domain of TEL (Poirel et al., 1997), a monoclonal antibody specific for the influenza HA (12CA5, Boehringer Mannheim) and a rabbit antiserum specific for amino acids 1013–1025 of human PDGFRβ receptor (Upstate Biotechnology Inc.; Claesson‐Welsh et al., 1988).
Immunoprecipitates were analyzed by denaturing polyacrylamide gel electrophoresis in the presence of SDS followed by fluorography of the dried gel (Amplify, Amersham).
In vitro translation
The cDNAs encoding TEL, ETS‐1, ERG‐2, GABPα and TEL–PDGFRβ (Golub et al., 1994) and their derivatives were subcloned into pBluescript SK (+) (Stratagene). In vitro T3‐driven transcription and translation were performed in micrococcal nuclease‐treated reticulocyte lysates using the TNT‐coupled transcription/translation kit (Promega) using l‐[35S]methionine as radioactive tracer amino acid. An aliquot (2 μl) of the programed lysate was analyzed directly by polyacrylamide gel electrophoresis, the remainder of the lysate was diluted in 10 volumes of RIPA buffer, centrifuged at 15 000 g for 10 min and aliquots of the supernatant were subjected to immunoprecipitation analyses.
Immunoblotting analyses were performed after electrophoretic transfer of proteins separated by SDS–PAGE on nitrocellulose membrane and incubation in the presence of 1 μg/ml of the 4G10 anti‐phosphotyrosine monoclonal antibody (Upstate Biotechnology Inc.) or a 1/100 dilution of the rabbit anti‐human PDGFRβ. Bands were visualized using the ECL chemiluminescence kit (Amersham).
In vitro binding experiments on glutathione–agarose beads
For GST fusion protein overexpression, Escherichia coli Dh5α cells transformed with the indicated plasmid were induced with 0.1 mM IPTG for 1 h at 37°C. Proteins were extracted as described by Frangioni and Neel (1993). Protein association experiments were performed on glutathione–agarose beads using GST fusion proteins.
For binding experiments, 10 μl of a standard in vitro translation reaction was incubated overnight at 4°C with 10 μl of beads in the presence of 1 mg/ml bovine serum albumin and 1 mM PMSF. After centrifugation, the supernatant (unbound fraction) was removed and the beads were washed three times with RIPA buffer. The bound proteins were eluted from the beads by boiling in SDS–PAGE sample buffer. Equivalent amounts of the unbound, wash and elution fractions were analyzed by polyacrylamide gels electrophoresis. After staining with Coomassie blue to detect GST fusion proteins, gels were treated with Amplify (Amersham Corp.), dried and analyzed by fluorography.
Immunofluorescence analyses were performed on HeLa cells transfected with 1 μg of either the ΔEB‐TP‐0 or ΔEB‐TP‐10 expression plasmids, or ΔEB as control. Cells were fixed in 4% paraformaldehyde, permeabilized using 0.2% Triton X‐100 and analyzed by indirect immunonofluorescence as previously described (Bailly et al., 1994).
The authors thank D.Bohmann, T.A.Brown, G.Gilliland, C.H.Heldin, G.Mavrothalassitis, S.L.McKnight, P.Mayeux and E.Reddy for reagents; M.Yaniv and C.H.Heldin for helpful discussions; M.Pironin for expert technical assistance; F.Arnouilh for patient typing; and G.Calothy and F.Lamballe for critical reading of the manuscript. C.J., C.C., A.B. and C.T.Q. are predoctoral fellows of the French MESR; C.O. holds a post‐doctoral fellowship from the EEC. This work was supported by funds from the Centre National de la Recherche Scientifique, Institut Curie, Ligue Nationale contre le Cancer (Axe Oncogénèse et Réseaux de Signalisation), Association pour la Recherche sur le Cancer, Fondation contre la Leucémie and EEC (Biomed, Contract BMH4‐CT96‐1355).
- Copyright © 1997 European Molecular Biology Organization