Transcription Factor ATF2 Cooperates with v-Jun To Promote Growth Factor-Independent Proliferation In Vitro and Tumor Formation In Vivo

Abstract

ATF2 belongs to the bZIP family of transcription factors and controls gene expression via 8-bp ATF/CREB motifs either as a homodimer or as a heterodimer—for instance, with Jun—but has never been shown to be directly involved in oncogenesis. Experiments were designed to evaluate a possible role of ATF2 in oncogenesis in chick embryo fibroblasts (CEFs) in the presence or absence of v-Jun. We found that (i) forced expression of ATF2 cannot alone cause transformation, (ii) overexpression of ATF2 plus v-Jun specifically stimulates v-Jun-induced growth in medium with a reduced amount of serum, and (iii) the efficiency of low-serum growth correlates with the activity of a Jun-ATF2-dependent model promoter in stably transformed CEFs. Analysis of ATF2 and Jun dimerization mutants showed that the growth-stimulatory effect of ATF2 is likely to be mediated by v-Jun–ATF2 heterodimers since (i) v-Jun-m1, a mutant with enhanced affinity for ATF2, induces growth in low-serum medium much more efficiently than v-Jun, when expressed alone or in combination with ATF2; and (ii) ATF2/fos, a mutant that efficiently binds to v-Jun but is unable to form stable homodimers, shows enhanced oncogenic cooperation with v-Jun. In addition, we examined the role of ATF2 in tumor formation by subcutaneous injection of CEFs into chickens. In contrast to v-Jun, v-Jun-m1 gave rise to numerous fibrosarcomas while coexpression of ATF2 and v-Jun-m1 led to a dramatic development of fibrosarcomas visible within 1 week. Together these data demonstrate that overexpressed ATF2 potentiates the ability of v-Jun-transformed CEFs to grow in low-serum medium in vitro and contributes to the formation of tumors in vivo.

Activating transcription factor 2 (ATF2; also known as mXBP and CRE-BP1) is a member of the ATF/CREB bZip family of transcription factors (31, 45). ATF2 can act as a transcription factor either as a homodimer or as a heterodimer with certain other bZip proteins, including the c-Jun component of activator protein 1 (AP1) (6, 22, 32). AP1 consists of a collection of dimers of members of the Jun, Fos, and ATF/CREB bZip families. Each dimer is thought to be functionally unique as defined by its capacity to activate or repress transcription and to target a particular subset of AP1-regulated genes (2, 38). AP1 regulates transcription in response to a multitude of extracellular signals, and it plays a decisive role in embryonal development (27, 33), in cell proliferation and tumorigenesis (68), in the response to cellular stress (14, 55), and in apoptosis (10, 23).

The biological role of ATF2 is poorly understood. Results from the study of knockout mice show that ATF2 is required for the development of the central nervous system and the skeleton (53). ATF2 mRNA is expressed in many cell types and is particularly abundant in the brain (45, 61). The mode of regulation of its promoter is not known; however, ATF2 mRNA accumulates after partial hepactectomy in rats, suggesting a role for this protein in tissue regeneration and cell proliferation (61). The level of ATF2 mRNA is also higher in some clinical samples of human tumors than in normal tissues (61). The transactivating activity of ATF2 is regulated posttranslationally by phosphorylation, particularly by the JNK/SAPK and p38 groups of mitogen-activated protein kinases, after exposure to cellular stress (20, 44, 54, 66). ATF2 has also been implicated in mediating a transcriptional response to the transforming adenovirus protein E1A (21, 42, 43, 63). It is also known that overexpression of c-jun or of its mutated viral counterpart, v-jun, triggers transformation in chicken or rat primary embryo fibroblasts (9, 12, 57, 68). Although ATF2 has never been directly implicated in oncogenesis, these data suggested a role for this protein in cell proliferation and transformation.

Recently, chick embryo fibroblast (CEF) transformation studies with Jun mutants that preferentially heterodimerize with Fos or ATF family members allowed us to hypothesize that Jun-Fos-like dimers might control anchorage-independent growth in agar, whereas Jun-ATF2-like dimers might regulate growth factor-independent proliferation, i.e., proliferation in low-serum medium (64). To test this hypothesis, we asked in the present study whether ATF2 participates in the oncogenic process induced by v-Jun in primary chicken cells. Therefore, after having isolated the avian ATF2 gene, currently the only known member of the ATF family in birds, CEF cultures that stably overexpress virally introduced ATF2, v-Jun, or both ATF2 and v-Jun were generated. These primary cultures were analyzed for their transformed phenotype in vitro and for their capacity to induce tumors in chickens.

MATERIALS AND METHODS

DNA constructs.

The various v-Jun and ATF2 coding sequences were inserted into the pE plasmid for DNA sequencing, in vitro protein synthesis, and subsequent introduction into retroviral vector RCAS envA or RCAS envD (denoted R and RD, respectively, in this paper) (30). pE is a pBluescript II SK(+) derivative (Stratagene) in which the SacI-to-SalI fragment from the original polylinker has been modified; the new polylinker contains the following primer/promoter sequences, restriction sites, and poly(A) sequence (in order): RP-BssHII-T3-SP6-ClaI-EcoRI-SacI-SmaI-BamHI-XbaI-SalI/AccI-PstI-HindIII-ClaI-poly(A)₃₀-NotI-XhoI-ApaI-KpnI-T7-BssHII-M13. The ClaI-ClaI polylinker fragment is from the CLA12 adapter plasmid (30). The v-Jun (36) and v-Jun-m1 (64) coding sequences were inserted into pE as SacI-XbaI and SacI-BamHI fragments, respectively. The full-length cDNA of chicken ATF2 was isolated by screening a v-Src-transformed CEF cDNA library (constructed by F. Piu in phagemid vector Uni-Zap XR [Stratagene]) with a ³²P-labelled human ATF2 probe. The ATF2 cDNA was excised from the phagemid vector as a pBluescript plasmid, recloned into pE as a BamHI-SalI fragment, and sequenced. Avian ATF2/fos was constructed in two steps. In the first step, a classical PCR-directed, silent mutagenesis strategy using primer oligonucleotides nested at NcoI and PpuMI or at PvuII and SalI sites (downstream of the stop codon) was used to generate ATF2 HMB, which carries the additional HindIII, MfeI, and BstEII unique restriction sites; the new HindIII site (5′-CAAAGCTTG, covering amino acids Gln360-Ser361-Leu362) is located at the hinge between the basic domain and the leucine zipper, the MfeI site (5′-CAATTG, covering amino acids Gln392-Leu393) is inside the zipper, and the BstEII site (5′-CCGGTAACC, covering amino acids Pro401-Val402-Thr403) is downstream of the zipper. In the second step, the c-Fos zipper sequence was inserted in place of the natural zipper into ATF2 HMB between the HindIII and BstEII sites, thus removing MfeI; PCR sense and antisense primers were 5′-CTGGGTACAAAGCTTGCAGGCGGAGACGGACCAGCTGG and 5′-CTGCATGGCGGTTACCGGGCAATCCCGGTGCGCCGCCAGGATGAACTCC, respectively (underlined sequences are c-fos specific, and the template was the coding sequence from avian c-fos (pCKFos plasmid [46]). Changes in ATF2 HMB and ATF2/fos were confirmed by DNA sequencing. In the ATF2/fos protein, the fragment Glu363 to Lys398 is replaced by the following sequence from c-Fos: Gln363-Ala-Glu - Thr - Asp - Gln - Leu - Glu - Glu - Glu - Lys - Ser - Ala - Leu - Gln - Ala - Glu - Iso - Ala - Asn-Leu-Leu-Lys-Glu-Lys-Glu-Lys-Leu-Glu-Phe-Iso-Leu-Ala-Ala-His-Arg398.

Cell culture.

Primary CEF cultures were routinely prepared every week from 8-day-old C/E SPAFAS chicken embryos (Merial, Lyon, France) and grown in regular medium supplemented with 6% serum as previously described (12). v-Jun- and ATF2-expressing cultures were obtained by chronic infection with the replication-competent retrovirus RCAS (30). Coinfections were performed with RCAS vectors R and RD. Routinely, transfections with R (no insert) and with R–v-Jun, R–v-Jun-m1, R-ATF2, and R-ATF2/fos plasmid DNAs were performed after the first passage, using the dimethyl sulfoxide-Polybrene technique (39), and viruses were allowed to spread through the entire population over the following week. Doubly infected cultures were then generated by superinfection with culture supernatant from CEFs chronically infected by RD derivatives and allowed to grow one more week. Colony formation in agar and growth in low-serum medium were performed as previously described (12). However, the amount of serum in the low-serum medium ranged from 0.6% to 0.2%, depending on the batch of serum and the experiment. For the measurement of thymidine uptake, cells in low-serum (0.6%) medium were seeded at a density of 3 × 10³/well in a 96-well plate. After overnight incubation, 0.5 μCi of [³H]thymidine (2.0 Ci/mmol; Amersham) was added per well, and uptake was measured daily for 5 days. To generate cultures from tumor cells, tumoral tissues were sliced into small pieces and incubated overnight in regular medium supplemented with collagenase H (1 mg/ml final concentration; Boehringer). Cells were subsequently passaged in a medium routinely used to grow CEFs.

Protein cross-linking, immunoprecipitation, and electrophoretic mobility shift assay (EMSA).

[¹⁴C]leucine-labeled proteins were synthesized in vitro, using a rabbit reticulocyte lysate translation system (Promega). Equal amounts of proteins were incubated in binding buffer (20 mM HEPES-KOH [pH 7.9], 50 mM KCl, 2.5 mM MgCl₂, 1 mM dithiothreitol 10% glycerol) in a final volume of 50 μl for 30 min at room temperature. The cross-linking agent, dithiobis(succinimidyl propionate) (DSP; Pierce, Rockford, Ill.), was added to a final concentration of 2 μM. After 15 min, the reaction was stopped by addition of Tris buffer (pH 7.5) to a final concentration of 50 mM. After 15 min, the protein mixture was diluted with RIPA buffer (10 mM Tris-HCl [pH 7.5], 2 mM EDTA, 150 mM NaCl, 0.5% Nonidet P-40, 0.5% sodium deoxycholate, antiprotease mixture) to 200 μl and precleared for 1 h with protein A-Sepharose beads (Pharmacia). Complexes obtained by immune precipitation were resolved by sodium dodecyl sulfate (SDS)–10% polyacrylamide gel electrophoresis (PAGE) (7). Gel shift assays were performed essentially as described elsewhere (11, 63). Protein-DNA complexes were separated on 5% acrylamide-bisacrylamide minigels (0.075 by 5 by 9 cm; Bio-Rad) in a Tris-glycine-EDTA buffer supplemented with 2.5% glycerol. Gels were run for 5 h at room temperature at a current of 3 mA/gel.

Antibodies, Western blotting, and isoelectric focusing.

To generate avian ATF2 antibodies, full-length ATF2 cDNA was cloned in the BamHI site of pGEX2T (Pharmacia). The glutathione S-transferase (GST)-ATF2 fusion protein was prepared as described in the literature accompanying the GST gene fusion kit (Pharmacia). Rabbits were immunized by repeated intradermal injection of the purified protein in accordance with a standard technique (Covalab, Lyon, France). Preparation of cell extracts for Western blotting and isoelectric focusing was performed as described elsewhere (63). For Western blotting, 10 μg of total cell extract was resolved by SDS–10% PAGE and blotted onto nitrocellulose membranes, which were further incubated with specific anti-Jun antibody (Santa Cruz; catalog no. sc#44) or anti-ATF2 antibody. The peroxidase-coupled secondary anti-rabbit antibody was purchased from Amersham (ECL detection system).

Isoelectric focusing was conducted in a Multiphor II system (Pharmacia). The gel consisted of 7.5% acrylamide in a solution containing 7 M urea, 0.2% Triton X-100, 7.6% Ampholine (preblended Ampholine, pH 3.5 to 9.5; Pharmacia), 20% sorbitol, and 10 mM dithiothreitol. The cathode strip was soaked with 1 N NaOH, and the anode strip was soaked with 0.5 M H₃PO₄. The gel was prefocused at 10°C at a constant voltage of 200 V, followed by 30 min at 300 V and an additional 30 min at 400 V. Cellular extracts were dialyzed against 20 mM Tris-HCl, pH 8.0. Fifteen microliters of dialyzed extract (2 μg/μl) was diluted twice to obtain a final mixture containing 6.5 M urea, 20% sorbitol, and 2.3% preblended Ampholine (pH 3.5 to 9.5; Pharmacia). Each sample (30 μg/30 μl) was further supplemented with 1 μl of a mixture containing 15% β-mercaptoethanol, 6% Triton X-100, 0.045 μM aprotinin, 30 μM pepstatin, 30 μM leupeptin, and 30 mM Pefablock (Boehringer); solubilized for 30 min at room temperature; and centrifuged for 30 min at 15,000 × g just before loading. Samples were loaded at the anode, using a siliconized applicator, and submitted to isoelectric focusing (150 V for 30 min, then 300 V for 60 min, and then 2,000 V for 5 h). Proteins were electrophoretically transferred onto an Immobilon membrane (Millipore) placed at the cathode side of a Multiphor II Novablot apparatus (Pharmacia) at room temperature for 1 h in 0.7% acetic acid at a current of 2 mA/cm². The membrane was immunostained as described for Western blotting.

Transactivation on test promoters in stably infected CEFs.

Cells were seeded at a density of 3 × 10⁵ per 60-mm-diameter plate in normal medium and transfected 24 h later with 2 μg of either the 5×TRE-TK-, the 5×jun2-TK-, or the TK-luciferase reporter plasmid (64), along with 0.5 μg of pUC18 and 15 μl of Superfect transfection reagent (Qiagen). Cell lysates were prepared 40 h after transfection, and the luciferase activity was measured by using luciferase assay system (Promega). Fold activation represents luciferase activity of the 5×jun2-TK- or the 5×TRE-TK-luciferase reporter in the R-Jun and/or R-ATF2-infected CEFs relative to their basal activity in R-infected CEFs and normalized to the luciferase activity of the TK-luciferase reporter in the different cultures.

Nucleotide sequence accession number.

The nucleotide sequence of the ATF2 cDNA was submitted to the EMBL nucleotide sequence database and given accession no. Y17724.

RESULTS

Isolation of the avian ATF2 gene.

To study the involvement of ATF2 in transformation of chicken cells, we isolated the gene encoding the avian homolog of that protein. A cDNA clone containing the complete coding sequence for chicken ATF2 was obtained by screening a chicken cDNA library with a human cDNA probe (see Materials and Methods). Alignment of the chicken protein sequence with the sequences from the rat (37), Xenopus laevis (67), and humans (45) proteins was performed (data not shown). Like the rat and Xenopus proteins, chicken ATF2 lacks the first 18 amino acids present in human ATF2; it exhibits 94% identity to the human sequence throughout the rest of the protein. Notably, the first 100 amino acids, including the zinc finger and the major known phosphorylation sites (44), are completely identical. Moreover, only one substitution over the entire bZip domain was detected (Asn379 in the chicken protein versus Ser in the proteins of the rat and human). This position is located outside of the dimerization interface (position f of the α helix [18, 51]) and therefore is unlikely to affect dimerization specificity. The fact that the different ATF2 proteins have the same structural organization and highly conserved sequences suggests that these proteins have common physiological functions in eucaryotic cells.

Overexpression of avian ATF2 does not result in transformation of CEFs.

To examine the effect of overexpression of ATF2 on CEF growth, uninfected CEFs and CEFs chronically infected with retrovirus R or R-ATF2 were generated (see Materials and Methods). The R-ATF2-infected cultures were shown to accumulate the ATF2 protein to about 3.5 times the level of the endogenous product present in R-infected cultures (see Fig. 3A). We found no obvious differences between these cultures with respect to cell morphology (as judged by light microscopy) and growth rate in normal medium (doubling time, about 30 h) (data not shown). Moreover, these cultures did not grow in low-serum medium (Fig. 1), nor were they able to form colonies in agar (Table 1). We conclude that overexpressed avian ATF2 by itself cannot transform CEFs.

FIG. 3.

Accumulation of v-Jun, v-Jun-m1 (denoted m1), and ATF2 proteins in extracts from singly or doubly infected CEF cultures. Western blotting was followed by detection of the accumulated proteins with an anti-Jun or an anti-ATF2 antiserum, as indicated. (A) Extracts from singly infected CEFs; (B) extracts from doubly infected CEFs expressing both Jun and ATF2. The positions of molecular size markers are indicated on the right.

FIG. 1.

Growth in low-serum medium of fully infected CEF cultures expressing either v-Jun, v-Jun-m1, or ATF2 or combinations of Jun and ATF2. Control cultures (denoted R) were infected with the empty retroviral vector. (A and C) Growth curves of cultures plated at 1.5 × 10⁵ cells per 100-mm-diameter plate at day zero; (B) thymidine uptake of cultures shown in panel A. As indicated on the panels, the serum concentration was 0.6% for panels A and B and 0.3% for panel C. These experiments were repeated at least five times with similar results and with series of infected cultures generated independently from different, freshly prepared primary cultures.

TABLE 1.

Colony formation for CEF cultures coexpressing v-Jun and ATF2^a

CEFs chronically infected with:	Expt. no.	No. of colonies for CEFs chronically infected withb:
		Uninfected	R	R-ATF2
RD	I	—c	—	—
	II	NDd	—	—
RD–v-Jun	I	72/58/74	81/79/76	28/26/34
	II	ND	330/414	144/188

^aA total of 5 × 10³ infected cells were seeded per 60-mm-diameter plate. Colonies were scored after 2 weeks. These experiments were performed with two series of independently infected cultures. 

^bEach number in a series represents the number of colonies formed in a given plate. 

^c—, no colonies present. 

^dND, not done. 

Overexpression of ATF2 potentiates low-serum growth induced by v-Jun.

Since R-ATF2 did not transform by itself, we asked whether this retrovirus would affect transformation by v-Jun. For this purpose, CEF cultures overexpressing ATF2, v-Jun, or ATF2 plus v-Jun were generated (see the Western blotting analysis depicted in Fig. 3B). When grown in normal medium supplemented with 6% serum, virally expressed ATF2 did not significantly change the growth rate of v-Jun-transformed CEFs (doubling time, 15 h with v-Jun or v-Jun plus ATF2), nor did it alter the typical fusiform shape of cells transformed by v-Jun (8, 13) (data not shown). However, when grown in medium with a reduced serum concentration (0.6%), cells expressing v-Jun alone displayed a doubling time of 4 days whereas the ATF2-plus-v-Jun-expressing cultures showed a significantly enhanced growth rate, with the doubling time reduced to 2 days (Fig. 1A). This increase in proliferative capacity in the ATF2-plus-v-Jun-expressing cells correlates with an increased uptake of [³H]thymidine (Fig. 1B).

In contrast, the virally expressed ATF2 could not stimulate another major feature of in vitro cell transformation by viral oncogenes, anchorage-independent growth in agar (8, 35). Rather, ATF2 significantly reduced the formation of colonies in agar by cells expressing v-Jun alone (2.2- to 2.4-fold [Table 1]; see Discussion). Thus, although ATF2 does not shown any growth-stimulatory activity on its own, it specifically stimulates growth factor independence of v-Jun-transformed CEFs.

A v-Jun mutant with a higher affinity for ATF2 shows enhanced proliferation in low-serum medium.

The stimulatory effect of ATF2 on proliferation of v-Jun-transformed CEFs in low-serum medium might either be due to enhanced accumulation of v-Jun–ATF2 heterodimers or be established independently of v-Jun—for instance, by accumulation of ATF2 homodimers. If overexpressed ATF2 stimulates the formation of v-Jun–ATF2 heterodimers, one would expect a mutant with an enhanced affinity for ATF2 to recruit more ATF2 and, consequently, display an enhanced growth capacity in low-serum medium, both alone and with overexpressed ATF2. An example of such a mutant is the previously described Jun dimerization mutant Jun-m1 (64).

Since the Jun-m1 mutant has thus far been characterized only on the basis of its dimerization specificity in a human c-Jun background with a human ATF2 partner (64), we first wanted to confirm the binding preference of v-Jun-m1 for avian ATF2. For this purpose, in vitro-translated v-Jun, v-Jun-m1, and avian ATF2 were analyzed by immunoprecipitation and gel shift analysis. As a control we included v-Jun/gcn4, a mutant carrying the dimerization domain from the yeast transcription factor GCN4, which only forms homodimers (36). As shown in Fig. 2A, v-Jun-m1 protein preferentially immunoprecipitated with ATF2 by a 2.3-fold factor compared to v-Jun; under the same conditions, v-Jun/gcn4 could not be precipitated. An enhanced affinity of the v-Jun-m1 mutant for ATF2 was further demonstrated by an EMSA using a high-affinity ATF2-Jun binding site, jun2 (63). As shown in Fig. 2B, the amount of ATF2-Jun heterodimer-containing complexes was 1.8-fold larger with v-Jun-m1 than with v-Jun, thus demonstrating a higher affinity of the v-Jun mutant for ATF2 in the DNA-bound state.

FIG. 2.

In vitro characterization of the dimerization and DNA binding properties of v-Jun and v-Jun-m1 in the absence and presence of avian ATF2. (A) Coimmunoprecipitation of v-Jun or v-Jun-m1 with ATF2 from a mixture of Jun plus ATF2 after cross-linking. Similar amounts of v-Jun, v-Jun-m1, and v-Jun/gcn4 were incubated with ATF2 in the presence of the reversible cross-linker DSP. After immunoprecipitation with an anti-ATF2 antibody, the immune complexes were dissociated and analyzed by SDS-PAGE. The relative amounts of immunoprecipitated v-Jun and v-Jun-m1 are indicated below the gel. (B) Gel shift analysis of the Jun-ATF2 binding site jun2. Similar amounts of v-Jun and v-Jun-m1 were incubated in the presence or absence of excess ATF2, and the retarded bands were resolved by EMSA to separate DNA complexes containing ATF2 homodimers, ATF2–v-Jun heterodimers, and v-Jun homodimers. The relative amounts of the retarded bands containing ATF2 plus v-Jun on ATF2 plus v-Jun-m1 are indicated below the gel.

In agreement with the model that overexpression of ATF2 potentiates v-Jun-induced low-serum growth by elevating the levels of v-Jun–ATF2 heterodimer, v-Jun-m1- and, particularly, v-Jun-m1-plus-ATF2-expressing cultures showed clearly enhanced low-serum (0.3%) growth compared to v-Jun- and v-Jun-plus-ATF2-expressing cultures, respectively (Fig. 1C). When the experiment was performed under more-stringent conditions, the difference was even more pronounced: at a serum concentration of 0.2%, v-Jun-m1-expressing CEFs proliferated for 3 days, whereas v-Jun-expressing CEFs proliferated for only 1 day and then stopped growing (data not shown). The gain in growth potential observed with v-Jun-m1 was unlikely to be due to enhanced Jun or ATF2 expression, since the expression levels in all cultures were identical (Fig. 3B).

Enhanced Jun-ATF2-dependent transcription correlates with enhanced growth in low-serum medium in stably transformed CEFs.

We next examined whether the enhanced expression of ATF2 and/or v-Jun in the stably infected CEF cultures is paralleled by enhanced levels of Jun-ATF2-dependent transcription. The reporter plasmids used were the TK-luciferase control construct and the 5×jun2-TK-luciferase construct (63), containing the multimerized distal Jun-ATF2 binding site (the jun2 element from the c-jun promoter) in front of the thymidine kinase promoter (TK). To demonstrate the specificity of the activation at the jun2 element, the 5×TRE-TK-luciferase reporter, containing the classical tetradecanoyl phorbol acetate-responsive element (TRE; the cJun-cFos site) from the human collagenase promoter (2, 34), was examined in parallel. Three series of independently generated cultures were analyzed (Fig. 4), with the following results: (i) cultures infected with ATF2 alone did not show significantly enhanced 5×jun2 transcriptional activity (1.3-, 1.4-, and 1.5-fold in figure 4A, B, and C, respectively); (ii) transcription in the v-Jun cultures was only slightly elevated (1.9- and 2.4-fold); (iii) the activity of the 5×jun2 element in the v-Jun-m1 cultures was clearly enhanced (4.8- and 5.0-fold); and (iv) the highest transcription levels were obtained in cultures coinfected with v-Jun (or v-Jun-m1) and ATF2 (3.5- and 4.8-fold for vJun; 6.2- and 6.4-fold for vJun-m1). Importantly, the 5×TRE-TK promoter showed only weak activity in the cultures stably expressing Jun, which is in line with earlier, independent studies on the collagenase promoter in CEFs (25, 36). The absence of significant 5×jun2 activity in the cultures infected by R-ATF2 alone is in agreement with transfection studies in mammals (15, 41, 66).

FIG. 4.

Transactivating activity on the 5×jun2-TK-luciferase, Jun-ATF-dependent reporter in cultures stably overexpressing the v-Jun, v-Jun-m1, or ATF2 protein as indicated. Control cultures were infected with the empty retroviral vector R. Experiments presented in all panels were performed with independently generated primary cultures. In panels A and B, transactivations were done on a 5×TRE-TK-luciferase reporter (5×TRE) as a control.

In conclusion, these data show that although overexpressed ATF2 by itself cannot efficiently transactivate on a jun2 motif, it stimulates transactivation when combined with v-Jun and, better still, with v-Jun-m1. In the stably infected cultures, the transactivation capacity on the 5×jun2 motif follows the order ATF2 < v-Jun < v-Jun-m1 = v-Jun plus ATF2 < v-Jun-m1 plus ATF2. By comparison with the data in Fig. 1C, we determined the same order for the efficiency with which the infected CEF cultures grow in low-serum medium. These results establish a clear and direct correlation between transactivation on a high-affinity Jun-ATF2 binding site and growth in low-serum medium in the transformed CEFs (Fig. 5).

FIG. 5.

Direct correlation between the capacity to grow in low-serum medium and the ability to transactivate the 5×jun2-TK-luciferase, Jun-ATF-dependent reporter of CEF cultures stably overexpressing combinations of ATF2 and Jun. The relative increase in cell number in low-serum medium after 10 days was plotted against the relative transactivation activity. Reference values from R-infected CEFs (denoted CEF) were set to 1.0. The data were from the experiments shown in Fig. 1C and 4C, in which the various combinations of ATF and Jun cultures were generated from the same primary culture.

ATF2/fos, a chimeric protein with the zipper domain from c-Fos, also cooperates with v-Jun.

The data presented above strongly suggest that ATF2 cooperates with v-Jun as an ATF2–v-Jun heterodimer acting on Jun-ATF2 target genes. In an attempt to reinforce this hypothesis, we designed a chimeric derivative of ATF2 that should preferentially (if not exclusively) accumulate as a heterodimer with Jun in the cell. To this end, the leucine zipper dimerization domain of ATF2 was replaced by the zipper domain of avian c-Fos. It is well documented that the zipper domain of Fos cannot form stable homodimers but is sufficient to mediate preferential heterodimerization with the zipper from Jun, thus promoting the formation of a highly stable Jun-fos complex (51, 60). The 36-amino-acid sequence from the c-Fos zipper was introduced into ATF2 to generate ATF2/fos (see Materials and Methods). This artificial construct retained the entire basic DNA binding domain of ATF2. As expected, ATF2/fos did not form stable homodimers, in contrast to wild-type ATF2, although both proteins bound successfully to the jun2 binding site as ATF2-vJun heterodimers. In fact, the binding of ATF2/fos–v-Jun was about fourfold stronger than that of ATF2–v-Jun (Fig. 6A).

FIG. 6.

Characterization of the chimeric ATF2/fos protein (see Materials and Methods) in vitro and in vivo. (A) Retarded bands from a gel shift assay with in vitro-made proteins and a jun2 probe. (B) Western blotting and immunodetection of ATF2 and ATF2/fos in extracts from R-, R-ATF2-, and R-ATF2/fos-infected CEFs, using an anti-ATF2 antibody. The arrow indicates the position of the virally expressed ATF. The positions of molecular size markers are indicated on the left. (C and D) Properties of CEF cultures stably expressing ATF2/fos. (C) Transactivating activity on the 5×jun2-TK-luciferase reporter in CEFs expressing ATF2, v-Jun, ATF2/fos, or a combination of Jun and ATF; (D) growth capacities in low-serum medium as measured by thymidine incorporation. The various combinations of ATF and Jun cultures presented in panels C and D were generated from the same primary culture.

Upon introduction into CEFs by stable retroviral infection, the ATF2 and ATF/fos proteins accumulated to the same extent (Fig. 6B). Like ATF2, ATF/fos did not significantly enhance transactivation on the 5×jun2-TK-luciferase reporter in these cultures (Fig. 6C), did not induce any abnormal cellular morphology (data not shown), and did not induce growth in low-serum medium (Fig. 6D). Thus, ATF2/fos can neither transactivate nor transform on its own. In contrast, like wild-type ATF2, ATF2/fos was able to stimulate transactivation on the 5×jun2-TK promoter (Fig. 6C) and growth in low-serum medium (Fig. 6D) when coexpressed with v-Jun. Indeed, the ATF2/fos chimera was slightly more potent than wild-type ATF2. We conclude from these results that the abundance of v-Jun–ATF2 is important for Jun-induced, growth factor-independent proliferation in vitro. Moreover, since ATF2/fos cannot form stable homodimers, an excess of ATF2 homodimers is not required.

ATF2 stimulates v-Jun-m1-induced tumorigenesis in chickens.

Having shown that ATF2 cooperates with v-Jun in cell transformation in vitro, we next examined its role in tumorigenesis in chickens. CEFs overexpressing various combinations of ATF2, v-Jun, and v-Jun-m1 were subcutaneously injected into the wing web of 1-day-old chicks (2 × 10⁶ infected cells per bird). The formation of local tumors was monitored over a 6-week period. As shown in Table 2, overexpression of ATF2 did not induce the formation of detectable tumors, while v-Jun was weakly tumorigenic; in experiments I and IV, v-Jun induced tumors after 6 weeks in 1 of 13 injected birds, whereas in experiments II and III, some tumors were detected after 3 weeks (1 of 6 and 2 of 5 injected birds, respectively). Poor tumor induction by v-Jun was also reported previously (17, 36, 69). Coexpression of ATF2 and v-Jun did not significantly modify either the number of tumors or the lag before their appearance.

TABLE 2.

Tumor formation in chickens^a

Expt no.	CEF expressing:	No. of animals injected	No. of animals with tumors after wk:
			1	2	3	4	5	6
I	-	4			0	0	0
	v-Jun	6			0	0	0
	v-Jun-m1	7			4	5	6
II	-	5	0		0
	v-Jun	6	0		1
	v-Jun-m1	6	6		6
III	ATF2	4	0	0	0			0
	v-Jun	5	0	0	2			4
	v-Jun + ATF2	6	0	0	2			5
	v-Jun-m1	5	3	5	5
	v-Jun-m1 + ATF2	6	6	6	6
IV	ATF2	4		0	0	0		0
	v-Jun	7		0	0	0		1
	v-Jun + ATF2	7		0	0	0		1
	v-Jun-m1	7		1	1	4
	v-Jun-m1 + ATF2	7		5	5	5

^aInfected CEFs (2 × 10⁶ in 0.1 ml of medium) were injected subcutaneously in the wing web of 1-day-old SPAFAS chickens. Tumor formation was scrutinized every week after injection. All of the animals were dissected; no obvious internal tumors were detected. 

In striking contrast, v-Jun-m1 induced the formation of local tumors after 3 weeks in 65% of the chickens on average (and even in 100% of the injected birds in experiments II and III). The combination ATF2 plus vJun-m1 was even more efficient than v-Jun-m1 alone in the sense that the latency period was reduced (100% of chicks bearing tumors after 1 week; experiment III) and in that the tumors were about two times larger in diameter (data not shown). To document this difference more accurately, smaller numbers of infected cells were injected per animal (2 × 10⁶, 1 × 10⁶, or 0.5 × 10⁶). At 0.5 × 10⁶ cells/animal, v-Jun-m1 alone could barely induce the formation of tumors after 2 weeks (four small tumors in a total of six birds) (Fig. 7), while the additional presence of excess ATF2 caused extensive tumor development (five large tumors in a total of five birds) (Fig. 7). Histological analyses of several tumors induced by v-Jun-m1 and v-Jun-m1 plus ATF2 did not reveal differences; they were all diagnosed as typical fibrosarcomas, with an abundant collagen matrix (data not shown). Despite the presence of these tumors at the site of injection, the animals were otherwise healthy, and no obvious internal tumors could be detected at autopsy.

FIG. 7.

Tumor formation in 38 young chicks after injection of CEFs overexpressing v-Jun-m1 or v-Jun-m1 plus ATF2. Cells (2 × 10⁶, 1 × 10⁶, or 0.5 × 10⁶) were injected subcutaneously into the wing web of 1-day-old chicks. Data represent tumor sizes at 2 weeks after injection. Only one animal injected with 10⁶ cells expressing v-Jun-m1 (chick no. 20) and two animals injected with 0.5 × 10⁶ cells expressing v-Jun-m1 (chick no. 32 and 33) did not develop visible tumors.

To verify the expression levels of v-Jun and ATF2 in the resulting tumor cells, these cells were expanded in vitro. Western blotting followed by immunodetection indicated that v-Jun-m1 and ATF2 accumulated to approximately the same extent in the tumor-derived cultures as in the original cultures (Fig. 8A and C). In one tumor cell line, a degradation product of v-Jun-m1 was also detected, possibly because in that case v-Jun-m1 accumulated to a particularly high level (Fig. 8C, lower panel; chicken no. 13). Isoelectric focusing followed by immunodetection also confirmed that the v-Jun-m1 mutant was expressed in the tumors (Fig. 8B) (note that the m1 mutation is characterized by four negatively charged amino acids in place of four positively charged residues in the wild-type leucine zipper [64], thus allowing a clear-cut separation of the v-Jun and v-Jun-m1 proteins).

FIG. 8.

Accumulation of the v-Jun-m1 and ATF2 proteins in tumor-derived cultures. (A) SDS-PAGE, followed by Western blotting and immunodetection of extracts from cultures corresponding to chick no. 2, 3, and 6 (see Fig. 7). Molecular size marker positions are shown on the right. (B) Isoelectric focusing followed by Western blotting and immunodetection reveals the presence of the v-Jun-m1 protein in the cultures shown in panel A; note that the endogenous c-Jun is not detectable in the R-infected, control culture because of the low efficiency of transfer onto the membrane after isoelectric focusing. (C) SDS-PAGE, followed by Western blotting and immunodetection, of extracts from cultures corresponding to chick no. 8, 9, and 13 in Fig. 7.

These in vivo experiments show that (i) in contrast to wild-type v-Jun, v-Jun-m1 is a very effective tumor inductor; and (ii) vJun-m1 and ATF2 dramatically increases the incidence as well as the size of the tumors. These data strongly suggest that ATF2 participates in v-Jun-induced carcinogenesis in vivo.

DISCUSSION

We previously reported an analysis of two dimerization mutants of Jun, Jun-m0 and Jun-m1, that preferentially heterodimerize as either Jun-Fos-like complexes, binding to 7-mer AP1 consensus motifs, or as Jun-ATF2-like complexes, targeting 8-mer CREB/ATF motifs, respectively (64). Excess Jun-m0 or Jun-m1 in CEFs was found to induce only one aspect of transformation by wild-type Jun: anchorage-independent growth in the case of Jun-m0, and growth factor-independent proliferation in the case of Jun-m1. Coexpression of Jun-m0 and Jun-m1 restored the wild-type Jun phenotype. According to these data, the putative role of the ATF2-like protein in transformation is restricted to the regulation of growth in low-serum medium. In support of this model, we showed in the present study that overexpression of ATF2 efficiently enhances growth of v-Jun-transformed CEFs in low-serum medium whereas it does not enhance anchorage-independent growth. In fact, we consistently found that ATF2 significantly reduced the number of colonies induced by v-Jun (Table 1). This concomitant ATF2-mediated enhancement of low-serum growth and reduced colony formation in agar medium appears to further emphasize the role of the balance between the two types of v-Jun-containing heterodimers in Jun-induced transformation: excess ATF2 would result in sequestration of v-Jun as a v-Jun–ATF2 complex, thus indirectly lowering the level of v-Jun–Fos-like dimers responsible for growth in agar.

We also found that in CEF cultures stably transformed by various combinations of v-Jun (or v-Jun-m1) and ATF2 there is a direct correlation between the capacity to proliferate in low-serum medium and the capacity to activate the model promoter 5×jun2-TK (Fig. 5). Previous attempts to document a relationship between Jun-dependent transformation and transactivation by using multiple types of Jun mutants in avian cells failed to reveal a direct correlation (24, 25, 58). However, these studies only addressed Jun-induced growth in agar medium and transactivation via Jun-Fos-type binding sites. The relationship between autocrine growth and Jun-ATF2-dependent transactivation was not addressed; furthermore, the endogenous transactivation capacity of the transformed cultures themselves was not examined. It would be interesting to directly measure the transactivating capacity on Jun-Fos-dependent model promoters in CEF cultures with distinct potentials for growth in agar medium.

The observed correlation between growth in low-serum medium and transactivation via Jun-ATF2 binding sites provides further evidence supporting the idea that ATF2-controlled target genes encoding growth factors or growth factor receptors (19, 47, 48), or cell cycle-dependent proteins such as cyclin A (59), are part of a Jun-ATF2-dependent oncogenic pathway. The characterization of oncogenically relevant target genes of this Jun-ATF2 pathway that are actually responsible for the lower serum requirement in CEFs transformed by v-Jun constitutes an interesting future challenge (3, 5). One would expect the expression levels of such target genes to mirror the response of the 5×jun2-TK model promoter depicted in Fig. 5. Furthermore, the direct correlation between transactivation and growth in low-serum medium strongly favors the notion that the abundance of v-Jun–ATF2 heterodimer is responsible for the observed cooperation between the two transcription factors. Both the results obtained with the mutant v-Jun-m1 (exhibiting enhanced affinity for ATF2 and reduced affinity for Fos) and those obtained with the chimeric protein ATF2/fos (which is only able to form Jun-ATF2 heterodimers) support this hypothesis. Moreover, the cooperation between ATF2/fos and v-Jun demonstrates that excess ATF2 homodimer is not necessary for growth in low-serum medium.

A further intriguing result of our study is the finding that ATF2 can contribute to the outgrowth of tumors induced by Jun in vivo. It should be noted, however, that this effect was observed only in cells overexpressing v-Jun-m1 and not in those expressing wild-type v-Jun. A possible explanation is that particularly high levels of Jun and ATF2 are required for tumor development in vivo. We noticed that some of the large tumors gave rise to cell cultures in which ATF2 and v-Jun-m1 accumulated at abnormally high levels (see, for instance, the data for chicken no. 13 in Fig. 7 and 8C). A comparative analysis of the activity of AP1 between CEFs prior to injection and several tumor-derived cultures might thus be informative. Whatever the exact explanation for the higher efficiency of v-Jun-m1, the data herein clearly show that ATF2 can cooperate with a Jun product in vivo during oncogenesis.

When taken together, the in vitro and in vivo data on the cooperation between vJun-m1 and ATF2 presented above strongly suggest that a reduced growth factor requirement via the postulated v-Jun–ATF2 pathway, rather than anchorage independence, is critical for tumoral outgrowth in our experimental model, i.e., for development of primary, local fibrosarcomas in the wing web of young chicks. This view is further supported by recent results in our laboratory which show that (i) v-Jun-m0, the dimerization mutant that induces growth in agar but not in low-serum medium in vitro, cannot induce this kind of tumor; and (ii) coexpression of v-Jun-m0 with v-Jun-m1 does not significantly modify tumorigenesis in comparison to expression of v-Jun-m1 alone when assessed in terms of the number of tumors and the period of latency (30a). These observations are of interest since it is generally assumed that anchorage independence correlates well with tumorigenicity (1, 4, 8). Studies of rodent cells transformed by human adenoviruses also suggest a correlation between primary tumor formation and in vitro growth factor independence, rather than anchorage independence (56, 62, 66a). In analogy to v-Jun-m1-transformed CEFs, these adenovirus-transformed cells showed increased levels of c-Jun–ATF2 and even decreased levels of c-Jun–Fos transcriptional activity (49, 63, 65). Interestingly, inhibition of c-Jun-Fos activity by the adenovirus E1A oncogene product leads to a strongly reduced expression of the secreted proteases collagenase and stromelysin and offers an explanation for the negative effects of E1A on cancer development during later stages, i.e., the inhibition of metastasis by E1A (16, 28, 52). It will therefore be important to develop other in vivo tests in the avian model to try to reveal a critical contribution of the anchorage-independent Jun-Fos-like controlled pathway to tumor development, for instance, by assaying for secreted protease activity, extracellular matrix composition, cell migration, and angiogenesis (26, 29, 40, 50).