Tumor-Specific PAX3-FKHR Transcription Factor, but Not PAX3, Activates the Platelet-Derived Growth Factor Alpha Receptor

Abstract

The t(2;13) chromosomal translocation occurs at a high frequency in alveolar rhabdomyosarcoma, a common pediatric tumor of muscle. This translocation results in the production of a chimeric fusion protein derived from two developmentally regulated transcription factors, PAX3 and FKHR. The two DNA binding modules, the paired domain and the homeodomain, of PAX3 are fused in frame to the transactivation domain of FKHR. Previously, tumor-specific PAX3-FKHR has been shown to bind to DNA sequences normally recognized by wild-type PAX3 and to exhibit relatively enhanced transcriptional activity. The DNA binding sites used to demonstrate that PAX3-FKHR is a more potent transcriptional activator than PAX3 have included recognition sequences for the paired domain of PAX3. In this report, we demonstrate the ability of PAX3-FKHR to activate the product of a growth control gene, platelet-derived growth factor alpha receptor (PDGFαR), by recognizing a paired-type homeodomain binding site located in the PDGFαR promoter. PAX3 alone cannot mediate transcriptional activation of this promoter under the conditions tested. This provides the first evidence that chromosomal translocation results in altered target gene specificity of PAX3-FKHR and suggests a transcriptional target that may play a significant role in oncogenic activity and rhabdomyosarcoma development.

Rhabdomyosarcomas (RMS) are the most common soft tissue tumors in children and young adolescents (5, 42, 66). These tumors often display a wide morphological spectrum ranging from poorly differentiated cells to well-differentiated rhabdomyoblasts that exhibit organized sarcomeric components and cross striations characteristic of striated muscles. Based on their histopathology, RMS can be classified as embryonal (subtype botryoid), alveolar, pleomorphic, or (for a small number) having a miscellaneous phenotype. In general, the less differentiated the subtype, the poorer the patient prognosis. The most prevalent subtypes of RMS are embryonal RMS and alveolar RMS (ARMS) (16, 30, 39), with ARMS being the most malignant. Recent genetic analysis has revealed that both subtypes are characterized by specific genetic abnormalities.

A unique chromosomal translocation involving chromosomes 2 and 13 is detected in most ARMS (14, 55, 60). The ARMS-specific t(2;13)(q35;q14) chromosomal translocation generates a novel chimeric protein of two transcription factors, PAX3 (chromosome 2) and FKHR (chromosome 13). As the result of the translocation, the chimeric transcription factor, PAX3-FKHR, contains an intact N-terminal PAX3 DNA binding domain but with the COOH-terminal region replaced by a truncated FKHR DNA binding domain and FKHR transcriptional activation domain (1, 23, 47). A variant t(1;13) translocation has also been detected in a small population of ARMS. In these cases, a closely related fusion protein is observed, with the PAX3 portion replaced by PAX7 (13).

Both the PAX3 and PAX7 genes belong to the nine-member gene family known as the Pax gene family (50). Members of the Pax protein family have a common 128-amino-acid DNA binding motif termed the paired domain (PD). Some members of the Pax family, including PAX3 and PAX7, contain a second DNA binding region of the paired-type homeodomain (HD) class. A proline-rich acidic region at the COOH terminus is identified as the transactivation domain for Pax proteins (24). PAX3 and PAX7 genes have been implicated in the development of myogenic cell lineage (6, 12, 28, 35, 51, 52). During early embryogenesis, they are expressed in the condensing somites, and PAX3 expression becomes restricted to the lateral dermomyotome which gives rise to the limb musculature (21, 26, 27, 62). Further evidence that PAX3 is important in muscle development comes from studies with Splotch mice. These mice, which have a mutated PAX3 gene, fail to develop limb muscle (17, 25, 57). In humans, PAX3 mutations have been identified in patients with Waardenburg’s syndrome, an autosomal dominant condition sometimes associated with limb muscle hypoplasia (53).

The FKHR gene is a member of the HNF-3/forkhead family of transcription factor genes (1, 23, 47). This protein family, like the Pax family, has also been implicated in developmental regulation. Family members share a conserved DNA binding motif referred to as the winged-helix (WH) motif (4, 31). Members of the HNF-3/forkhead family have been shown to function in regulating inflammatory responses of the liver and in the development of blood cell lineage (31, 61). Two members of this family, FKHR and qin, have been linked to the induction of neoplasia.

Although the ARMS translocation breakpoint has recently been defined and the resulting chimeric gene product has been identified, little information about the mechanism of PAX3-FKHR-induced oncogenesis is known. Since both DNA binding regions, the paired domain and the homeodomain, of the PAX3 protein are intact in the chimeric protein, it seems likely that PAX3-responsive genes would be targets for the transcriptional activation of PAX3-FKHR. Indeed, many PAX3-responsive sequences thus far identified also respond to PAX3-FKHR as determined by electrophoretic mobility shift assays (EMSA) and transient-transfection experiments. These experiments have suggested that PAX3-FKHR is a more potent transcriptional activator than the wild-type PAX3 (22). Therefore, one mechanism by which PAX3-FKHR is thought to function as an oncoprotein is by abnormally up regulating genes that are normally targets of PAX3. Although the chimeric protein also retains a portion of the FKHR-derived WH DNA binding domain, binding of PAX3-FKHR to WH-specific binding sequences has not been demonstrated and would seem unlikely, since the chimera lacks the first alpha helix of the WH domain, which is critical for the DNA binding activity of FKHR (4, 11).

In this report, we demonstrate for the first time that PAX3-FKHR can directly activate the transcription of the platelet-derived growth factor alpha receptor (PDGFαR) and that PAX3 cannot do so under identical conditions. PDGFαR is one of the two known cell surface transmembrane receptors specific for the mitogenic factor PDGF. We suggest that PDGFαR could be a specific target in PAX3-FKHR oncogenesis, providing a possible mechanism for PAX3-FKHR-mediated cell transformation and uncontrolled cell growth associated with ARMS.

MATERIALS AND METHODS

DNA constructs.

pcDNA3 expression vector constructs containing the wild-type and mutant mouse-human PAX3-FKHR hybrid cDNAs (Un-1, Bu35, PD-NH2, and S268A) and the pCMV expression vector containing the mouse PAX3 cDNA have been previously described (19, 22). The PD-HD and WH mutants were religated DNA products of an EcoRI partial digest of the pcDNA3-PAX3-FKHR expression vector, removing either the paired domain and homeodomain or the bisected FKHR binding domain, respectively. PAX3-specific rabbit polyclonal antibody was a kind gift from Frank Rauscher (Wistar Institute, Philadelphia, Pa.) (22). Unless stated otherwise, the reporter chloramphenicol acetyltransferase (CAT) gene constructs under the control of mouse PDGFαR promoter sequences have been previously described (58, 59).

Cell cultures.

All cell cultures were maintained in Dulbecco’s modified Eagle’s high-glucose medium supplemented with 200 U of penicillin per ml, 50 μg of streptomycin per ml, 1 mM glutamine, and 10% (vol/vol) calf serum. For P19 cells, monolayer cultures were maintained on tissue culture dishes that were pretreated with 0.3% gelatin.

Transient-transfection and CAT assay.

Cells were plated at a density of 10⁶ cells per 100-mm-diameter tissue culture dish 24 h prior to transient transfection. Transient transfection was carried out with a total of 20 μg of DNA that included 3 μg of β-galactosidase DNA (LacZ) driven by the β-actin promoter as an internal standard for monitoring transfection efficiency. For P19 cells, the transfection was carried out by the calcium phosphate method. P19 cells were exposed to DNA-CaPO₄ precipitate for 17 h, rinsed, and refed with growth medium for an additional 48 h before harvest. For COS cells, transfection was carried out by the DEAE-dextran method. Cells were exposed to DEAE-dextran-DNA complex for 3 h followed by incubation with serum-free medium containing 100 μM chloroquin for an additional 3 h. Cells were then rinsed and refed with growth medium for an additional 48 h before harvest. Cell lysates for LacZ and CAT assays were prepared as described previously (59). Deacetylase activity in the lysate was inactivated by heating the lysates to 60°C for 7 min (37). A typical CAT assay reaction mixture consisted of 0.7 μg of acetyl-coenzyme A, 0.2 μCi of [¹⁴C]chloramphenicol (1 Ci = 37 GBq), and cell lysates in a final volume of 150 μl. The amount of cell lysate used in each CAT reaction was standardized by its β-galactosidase activity. Unless stated otherwise, the routine CAT assays were carried out at 37°C for 1 to 3 h and terminated by extraction with 1 ml of ice-cold ethyl acetate. Quantitative analysis of CAT activity was carried out by measuring the radioactivity of each radioactive spot in a beta-scintillation counter.

Preparation of labeled DNA probe for EMSA.

Most of the synthetic versions of different PFαR regions and the e5 sequence were made as complementary pairs of unphosphorylated, single-stranded oligonucleotides with different but compatible restriction ends (a BamHI site at the 5′ end and a BglII site at the 3′ end). These oligonucleotide duplexes were phosphorylated by T4 polynucleotide kinase before they were cloned into the BamHI-linearized pKS⁺-Bluescript vector (Strategene). Oligonucleotides that do not contain compatible restriction ends were cloned into pKS⁺-Bluescript vector at the SmaI site. For EMSA studies, DNA fragments containing the oligonucleotide sequences were first released from the vectors by EcoRI-XbaI double digestion and then labeled by nucleotide fill-in reaction with [α-³²P]dATP or [α-³²P]dCTP by using Klenow polymerase. The sequences of the synthetic oligonucleotides of the wild-type PDGFαR promoter sequence (top strand shown) were 5′ gatccGCCTCACAATCCAGCCTTTCAAAAACCCATCATCTa 3′ (PFαR₁), 5′ gatcCCATCTTCCTATTAGACTCCACAGTTTCCTAATCCCa (PFαR₂), 5′ gatcCATCCCATTAAAGGATTAGCAACTACACGGCACTTa 3′ (PFαR₃), and 5′ gatccCAGTTTCCTAATCCCATTAAAGGATTAGCAACTACa 3′ (PFαR₄). The sequences of the synthetic mutant oligonucleotides of PFαR₄ (top strand shown) were 5′ gatccCAGTTTCCGCCGCCCATTAAAGGATTAGCAACTACa 3′ (M₂), 5′ gatccCAGTTTCCTAATCCCCGGCAAGGATTAGCAACTACa 3′ (M₃), and 5′ gatccGTTTCCTAATCCCATTAAAGGCGGCGCAACTACa 3′ (M₄). The lowercase nucleotides represent the restriction ends designed to facilitate the determination of the orientations of the oligonucleotide duplexes after cloning into plasmid vectors and to allow additional copies of oligonucleotide duplexes be added in sequential cloning steps.

EMSA.

Nuclear extracts were prepared as previously described, with minor changes (46). In brief, cells were rinsed with ice-cold phosphate-buffered saline twice and allowed to swell in hypotonic buffer (10 mM HEPES [pH 7.9], 0.75 mM spermidine, 0.15 mM spermine, 10 mM KCl, 0.1 mM EGTA, 0.1 mM EDTA, 1 mM dithiothreitol [DTT], 0.5 mM phenylmethylsulfonyl fluoride [PMSF], 0.2 mM NaF, 0.2 mM sodium vanadate) before lysis with a Dounce homogenizer. After homogenization, the lysed cells were mixed with 0.1 volume of sucrose restore buffer (67.5% sucrose, 50 mM HEPES [pH 7.9], 0.75 mM spermidine, 0.15 mM spermine, 10 mM KCl, 0.2 mM EDTA, 1 mM DTT, 0.5 mM PMSF, 0.2 mM NaF, 0.2 mM sodium vanadate) and subjected to 30 s of centrifugation at 10,000 × g to pellet nuclei. The nuclei were then resuspended in nuclear resuspension buffer (20 mM HEPES [pH 7.9], 0.75 mM spermidine, 0.15 mM spermine, 0.2 mM EDTA, 2 mM EGTA, 2 mM DTT, 0.5 mM PMSF, 0.2 mM NaF, 0.2 mM sodium vanadate, 25% glycerol), from which proteins were extracted by the addition at 1/10 (vol/vol) of 4 M ammonium sulfate. The extraction was done at 4°C for 30 min with continuous rocking. The debris was sedimented by centrifugation at 100,000 × g for 1.5 h. The clarified supernatant was then collected and stored in aliquots in liquid nitrogen until use.

EMSA was performed by incubating nuclear extract or glutathione-S-transferase (GST) fusion proteins with nonspecific carrier DNA, such as poly(dI-dC) and poly(dA-dT) in binding buffer containing 20 mM HEPES (pH 7.9), 1 mM β-mercaptoethanol, 0.5 mM EDTA, 50 mM KCl, and 0.2% bovine serum albumin for 5 min on ice. Routinely, 0.2 ng of a ³²P-labeled DNA probe prepared by Klenow labeling was added to the EMSA reaction mixture (final volume of 20 μl) and allowed to form DNA-protein complexes during a 20-min incubation at room temperature. The complex was analyzed on a 5% nondenaturing polyacrylamide gel. Electrophoresis was carried out in 0.5× Tris-borate-EDTA buffer at 200 V at 4°C until the bromophenol blue dye reached the bottom, and the gel was dried and autoradiographed. For antibody competition assays, 1 μl of a 1:3 dilution of the PAX3 antibody was first mixed with the extract after the initial 5 min and allowed to incubate at room temperature for 10 min before further addition of the probe. The antibody specific for PAX3 was generated against the region of the mouse PAX3 corresponding to amino acids 280 to 479 as described previously (22).

The GST fusion protein constructs expressing the DNA binding domains of the wild-type and mutant PAX3-FKHR proteins were constructed by inserting the EcoRI-EcoRI fragment encompassing the entire DNA binding domains into the EcoRI site of the pGEX-2T vector (Pharmacia). The e5 sequence used in the CAT assay and EMSA is TCGGGCAGCACCGACGATTAGCACCGTTCCGCTCAGGCTCGG. This e5 sequence contains recognition sites for the paired domain (GTTCC) and for the homeodomain (ATTA) of the PAX3 and PAX3-FKHR proteins, and it has been shown to respond to transcriptional activation by PAX3 and PAX3-FKHR in P19 cells (9, 10, 26, 54).

Footprinting analysis.

DNase I footprinting was performed as described previously (18). Briefly, the −912αRCAT plasmid was digested with BspE1, labeled with [α-³²P]dCTP by using Klenow polymerase, and digested with EcoRI. A 240-bp fragment including the PAX3-FKHR-responsive element (PFRE) was gel purified and incubated with GST-PAX3 paired-domain plus homeodomain proteins (19) in the presence of 50 μg of poly(dI) · (dC) per ml. The DNA binding motifs of the GST-PAX3 paired-domain plus homeodomain are the same as those found in mouse-human PAX3-FKHR hybrid used in the present study. After brief digestion with DNase I, samples were run on a 6% polyacrylamide–urea gel in parallel with a guanine reaction sequencing ladder prepared from the original end-labeled probe by the method of Maxam and Gilbert (36).

Site-directed mutagenesis.

To introduce mutant sequences into the PDGFαR promoter, we used the Strategene site-directed mutagenesis PCR kit under the conditions recommended by the manufacturer. All of the constructs made by this method were verified by restriction enzyme digestion and DNA sequence analysis. The sequences of the oligonucleotide primers used to generate mutations in the ATTA sites (only the top strand is shown) were 5′ GTTTCCTCCGCCCATTAAAGG 3′ (αRM₂), 5′ CTAATCCCCGGAAAGGAATTAGC 3′ (αRM₃), and 5′ CCATTAAAGGAGGCGCAACTAC 3′ (αRM₄).

RNase protection analysis.

Total RNA was prepared from cultured cells by the guanidinium HCl-phenol extraction method (Trizol reagent [GIBCO-BRL]) according to the manufacturer’s recommendations. For the RNase protection assay, 30 μg of total RNA isolated from transfected cells was hybridized to a PAX3-, PAX3-FKHR-, PDGFαR-, or cyclophillin-specific RNA probe overnight at 45°C before the sample was further subjected to RNase A and T₁ digestion. The protected RNA bands were size fractionated on a 6% polyacrylamide–urea sequencing gel. After electrophoresis, the gel was fixed in 10% acetic acid–10% methanol for 30 min, dried, and exposed for autoradiography at −80°C.

Western blot analysis.

To verify that PAX3-FKHR protein was synthesized by transfected COS cells or by bacterial GST fusion proteins, a COS cell nuclear extract or partially purified GST fusion extract was first size fractionated on sodium dodecyl sulfate–10% polyacrylamide gels. The proteins were transblotted onto nitrocellulose membranes by electrophoresis in transblot buffer (20 mM Tris [pH 8.0], 192 mM glycine, 10% methanol, 0.1% sodium dodecyl sulfate). Detection of PAX3-FKHR or derivative proteins was carried out by using the chemiluminescent-antibody detection kit (NEN) under the conditions recommended by the manufacturer. The primary rabbit polyclonal PAX3 antibody was used at a dilution of 1:500.

RESULTS

Sequence analysis identifies a putative PAX3 binding site.

The overlapping expression patterns of several of the Pax gene family members and PDGFαR in embryonic tissues, coupled with evidence of altered PDGF signaling in oncogenesis, led us to examine the possibility that Pax genes might directly regulate the expression of the PDGFαR gene. We examined the proximal 6 kb of the PDGFαR promoter and identified a potential PAX3 paired-domain binding site, GTCACGCCT, at position −812 (Fig. 1A). This sequence conforms well to the optimal PAX3 paired-domain binding sequence (GTCACGC/ATT) previously identified by in vitro binding site selection (8, 20). In addition, this sequence is similar to a functional PAX3 paired-domain binding site found in the c-Met promoter that mediates PAX3-dependent gene induction during limb muscle development (20).

FIG. 1.

Transactivation of the PDGFαR promoter by PAX3 and PAX3-FKHR in P19 cells. (A) Schematic illustration of the PDGFαR promoter-containing CAT reporter gene constructs. The black bar indicates the PAX3 consensus binding sequences identified within the PDGFαR receptor promoter. (B) P19 cells were transiently transfected with a total of 20 μg DNA containing 2.5 μg of the CAT reporter construct, 1 μg of the PAX3 or PAX3-FKHR expression vector (or, in the inset, various amounts of PAX3 or PAX3-FKHR in the presence of −912αRCAT), 3 μg of the β-galactosidase gene under the control of the β-actin promoter, and nonspecific pGEM3 plasmid DNA under the conditions described in Materials and Methods. Whole-cell lysates were prepared 48 h after transfection and assayed for both β-galactosidase activity and CAT activity. The amounts of lysates used for CAT assays were normalized to the β-galactosidase activity. p6e5CAT contains six tandem copies of the e5 sequence inserted upstream of the E1B-TATA basal promoter that drives the CAT reporter gene. Fold induction for the PDGFαR promoter-containing constructs was calculated as that over the CAT activity measured in cells transfected with the promoterless pGEMCAT. Fold induction for the e5 sequence was calculated as that over the CAT activity measured in cells transfected with the basal pG5ECAT promoter construct. The percentage of conversion in CAT activity of pGEMCAT or pG5ECAT was then assigned an arbitrary value of 100%. The values represent means and standard deviations for 5 (±2) determinations. The standard deviation is defined as the root mean square deviation of n − 1 determinations.

The PDGFαR promoter is selectively activated by the tumor-specific PAX3-FKHR fusion transcription factor.

To determine if this sequence could respond to PAX3-induced transcription activation, we cotransfected a pcDNA3 expression vector encoding either PAX3 or PAX3-FKHR and the PDGFαR promoter-driven CAT reporter gene construct −912- or −100αRCAT (Fig. 1A) into murine P19 teratocarcinoma cells. P19 cells have been shown to be capable of hosting PAX3- and PAX3-FKHR-mediated transactivation (19). The −912αRCAT promoter contained the putative paired-domain binding site, whereas the −100αRCAT promoter contained the basal promoter sequence, as previously determined (58). As a positive control for transactivation, we used the p6e5CAT construct, which contains a basal promoter and six copies of an e5 sequence (see Materials and Methods). The e5 sequence was first identified as part of a response element from the Drosophila even skipped promoter for the Eve transcription factor (26). The e5 sequence includes adjacent paired-domain (GTTCC) and homeodomain (ATTA) binding sites that are recognizable by PAX3 and PAX3-FKHR. The p6e5CAT construct has previously been shown to respond to transcriptional activation by PAX3 and PAX3-FKHR.

In agreement with prior reports, the e5 sequence was transactivated by both PAX3 and PAX3-FKHR, with PAX3-FKHR being the more potent transactivator (Fig. 1B). In contrast, the PDGFαR promoter was transactivated only by PAX3-FKHR and not by PAX3. We conducted dose-response experiments using the −912αRCAT reporter plasmid and either PAX3 or PAX3-FKHR (Fig. 1B, inset). No PAX3-dependent induction of the PDGFαR promoter was detected despite a wide range of PAX3/−912αRCAT reporter ratios being tested. We tested ratios up to 100-fold higher than that necessary to elicit unambiguous PAX3-FKHR-dependent activation and 10-fold higher than that required to elicit maximal PAX3-FKHR-dependent activation. At even higher doses, we began to see inhibition of the cotransfected β-galactosidase activity, perhaps secondary to squelching or to the inhibitory effects of PAX3 that have previously been reported to be independent of DNA binding (7). In contrast, when the p6e5CAT reporter was used, we consistently saw an approximately twofold transactivation potential of PAX3-FKHR compared to PAX3 (Fig. 1B and data not shown). The selective response of the PDGFαR reporter gene to the tumor-specific PAX3-FKHR suggests that (i) PAX3 was too weak a transactivator for the PDGFαR promoter, and stimulation by the more potent transactivator PAX3-FKHR could still be detected, or (ii) PAX3-FKHR might recognize a DNA sequence in the promoter that was not capable of mediating functional transactivation by PAX3. To differentiate between the two possibilities, we examined whether deletion of the putative paired-domain binding sequence would eliminate PAX3-FKHR-dependent induction. Using both Erase-a-Base and restriction enzyme methods, we generated several serial deletion constructs from the −912αRCAT reporter (Fig. 2A). Each of these constructs was transfected into P19 cells in the presence and absence of PAX3-FKHR. As shown in Fig. 2B, removal of the putative paired-domain binding sequence in −799αRCAT did not prevent the promoter from responding to PAX3-FKHR stimulation. This result suggests that a different DNA target sequence is involved. Serial truncations and internal deletions localized the sequence responsible for mediating PAX3-FKHR activation to within the −646 to −559 region of the promoter. Since deletion of other DNA sequences surrounding the −646 to −559 fragment of the promoter did not significantly impair the induction by PAX3-FKHR, we believe that the PAX-FKHR-responsive element (PFRE) was contained within this 87-bp fragment. PAX3 had no effect on this 87-bp fragment in cotransfection assays (data not shown).

FIG. 2.

Transactivation of PDGFαR promoter-CAT deletion constructs by PAX3-FKHR. (A) Schematic illustration of the serial deletion constructs of the PDGFαR promoter. (B) Transient-transfection and CAT assays were performed under the same conditions as described in the legend to Fig. 1B. Fold induction was calculated as that over the CAT activity measured in cells transfected with the promoterless pGEMCAT. The percentage of conversion in CAT activity of pGEMCAT was then assigned an arbitrary value of 100%. The values represent means and standard deviations for 5 (±2) determinations. The standard deviation is defined as the root mean square deviation of n − 1 determinations.

Induction of the endogenous PDGFαR gene by PAX3-FKHR.

To examine whether the promoter analysis reflected how the endogenous PDGFαR gene responds to PAX3-FKHR stimulation at a cellular level, we measured endogenous PDGFαR mRNA levels in P19 cells that were transfected with PAX3-FKHR (Fig. 3). We transiently transfected cells with 10 μg of the expression vector alone or with various amounts of the vector expressing PAX3 or PAX3-FKHR. Total RNA was collected 24 h after transfection and analyzed for PDGFαR, PAX3, and PAX3-FKHR expression. As shown in Fig. 3, we did not detect PDGFαR mRNA in cells transfected with the expression vector alone or in cells transfected with PAX3. In the panel indicating the PAX3 signal, the band observed in the lane with vector alone represents residual undigested PAX3 RNA probe, which is also observed above the PAX3-specific signals in all of the PAX3-expressing lanes. In contrast, the PDGFαR mRNA signal was clearly detectable in cells transfected with PAX3-FKHR. An RNA probe specific for cyclophillin was used as an internal control to normalize for the amount of RNA used in each reaction mixture.

FIG. 3.

Induction of endogenous PDGFαR mRNA by transiently expressed PAX3-FKHR. P19 cells were transiently transfected with a total of 10 μg of DNA containing either pcDNA3 vector (10 μg) or pcDNA3 vector expressing PAX3 or PAX3-FKHR (2.5, 5, or 10 μg [increasing doses are indicated by black triangles]) by the CaPO4 method as described in Materials and Methods. After 24 h of transfection, cells were harvested for extraction of total cellular RNA with Trizol reagent. Total cellular RNA (30 μg) of each sample was assayed for the expression of PAX3-FKHR, PDGFαR, and cyclophillin by the RNase protection assay. RNA (10 μg) prepared from the ARMS cell line RH4, which expresses all three messages, was used as the positive control. Yeast tRNA (10 μg) was included in the assay as the negative control. The signals for PAX3-FKHR and PDGFαR in the left panels were from a 3-day exposure of the X-ray film. The signals for PAX3 and PDGFαR in the right panels were from a 4-day exposure of the X-ray film. The signal for cyclophillin was obtained from a 1-day exposure of the X-ray film. Data presented in the PAX3-FKHR panels are from the same film at the exposure times indicated above. The gel on the right was run slightly longer, such that the lower cyclophillin band has run off the bottom of the gel.

The activity of the PFRE is dependent on the homeodomain, not the paired domain, of PAX3-FKHR.

Since there was not any notable similarity between the 87-bp PFRE and the paired-domain consensus binding site for PAX3, we tested the involvement of different DNA binding domains (i.e., the PAX3 paired domain, homeodomain, and the FKHR bisected WH domain) present in PAX3-FKHR by cotransfecting cells with PDGFαR promoter-CAT constructs and expression vectors encoding mutant PAX3-FKHR proteins (Fig. 4A). Both Un-1 and Bu35 encode single-amino-acid replacement mutants with mutations in paired domain of PAX3-FKHR that have been shown to abolish DNA binding activity. PD-NH2 is a deletion mutant that is missing the 5′ half of the paired-domain sequence and is predicted to abolish DNA binding by the paired domain. S268A, HD-C, and HD encode mutants with mutations in the homeodomain. S268A encodes a single amino acid replacement of the serine residue by alanine. The serine residue is at position 9 of the third recognition helix of the homeodomain, which has been shown to impair the ability of the paired-type homeodomains to form homodimers (19). HD-C encodes a protein missing the third alpha helix of the homeodomain, and HD encodes a protein missing the entire homeodomain. PD-HD encodes a deletion mutant that does not contain any of the PAX3 DNA binding domains. Presumably, the only possible DNA binding activity of the mutant would be mediated by the bisected WH DNA binding domain of FKHR. The WH mutant of PAX3-FKHR is missing the bisected WH DNA binding domain of FKHR but retains intact PAX3 DNA binding domains and the FKHR transactivation domain.

FIG. 4.

Transactivation of PDGFαR promoter-CAT constructs by wild-type and mutant PAX3-FKHR. (A) Schematic illustration of the wild-type and mutant PAX3-FKHR expression vectors. (B) Effect of wild-type (WT) and mutant PAX3-FKHR on the transcriptional activation of the PDGFαR promoter. The PFRE was contained within the −912αRCAT construct and was absent from the −549αRCAT construct. (C) Effect of wild-type and mutant PAX3-FKHR on the activity of e5-controlled CAT activity. Conditions used for cotransfection for panels B and C were the same as those described in the legend to Fig. 1B. Fold induction was calculated as that over the CAT activity measured in cells transfected with the expression vector alone. The percentage of conversion in CAT activity measured under these conditions was then assigned an arbitrary value of 100%. The values represent means and standard deviations for three determinations. The standard deviation is defined as the root mean square deviation of n − 1 determinations.

As shown in Fig. 4B, PAX3-FKHR mutations that disrupted paired-domain-mediated DNA binding (Bu35, Un-1, and PD-NH2) did not affect the ability of PAX3-FKHR to transactivate the PDGFαR promoter. On the other hand, mutations that disrupted the homeodomain, such as those in HD and HD-C, totally abolished the stimulatory ability of PAX3-FKHR. The S268A mutant, which can bind to homeodomain consensus sequences but exhibits altered dimerization potential, could still induce transactivation activity of the PDGFαR promoter, although at a much-reduced level. The WH mutant, which contains intact PAX3 DNA binding domains, had no effect on PAX3-FKHR transactivation activity, supporting the notion that the bisected WH domain does not have a direct role in DNA binding and transactivation. Neither the wild-type nor the mutants were able to transactivate a PDGFαR promoter construct (−549) that did not contain the PFRE. In contrast to transactivation of the PFRE, transactivation of reporter constructs containing the reiterated e5 sequence required both the paired domain and homeodomain, consistent with previous reports (54) (Fig. 4C). These results clearly demonstrate that an intact homeodomain is the only DNA binding domain of PAX3-FKHR required for transactivation of the PDGFαR promoter constructs containing the PFRE.

The PFRE contains a paired-type homeodomain binding site.

Many homeodomain-containing transcription factors recognize DNA sequences containing a core ATTA (or TAAT on the opposite strand) consensus (15). The finding that the homeodomain, rather than the paired domain, of PAX3-FKHR was responsible for the transactivation of the PFRE led us to identify four ATTA sequences (ATTA₁, TAAT₂, ATTA₃, and ATTA₄) within the 87-bp PFRE (see Fig. 6A). We tested the ability of the PAX3 homeodomain to bind to these ATTA sequences by performing DNase I footprinting and EMSA experiments. As shown in Fig. 5, the purified PAX3 homeodomain was able to protect a 22- to 24-bp fragment of DNA from DNase I digestion. This region correlated exactly with three of the ATTA sequences identified within the PFRE. The footprinted region included two adjacent ATTA sequences arranged as inverted repeats with a spacing of 3 bp (TAATCCCATTA). This motif is similar to a previously reported P3 site that is recognized by the Drosophila Prd homeodomain (63).

FIG. 6.

Gel shift analysis of PAX3-FKHR binding to regions of the PFRE. (A) Schematic presentation of the four synthetic oligonucleotides that correspond to sequences retained within the 87-bp PFRE sequence (−646 to −559) of the PDGF promoter. (B) Western blot analysis of wild-type and mutant PAX3-FKHR DNA binding domain proteins expressed in COS cells. A total of 10 μg from each nuclear extract was used for the analysis under the conditions described in Materials and Methods. (C) The four oligonucleotides were released from pKS⁺-Bluescript plasmid by EcoRI and XbaI double digestion and gel purified before they were end labeled with [α-³²P]dCTP by use of Klenow polymerase. For EMSA, the COS nuclear extract first was incubated with 0.05 μg of poly(dI) · (dC) per μl and 0.0375 μg of poly(dA) · (dT) per μl as nonspecific DNA carriers in binding buffer supplemented with 3.75 mM MgCl₂ and 9.5 mM spermidine and then was incubated with the radiolabeled PFRE probes. To affirm specific binding, preimmune and PAX3-specific antibodies (Ab) were included during the initial incubation with the nonspecific DNA carrier. Under these conditions, the antibody remained capable of supershifting the PAX3-FKHR–DNA complex (see asterisk). The arrow and arrowhead indicate the positions of intact and degraded PAX3-FKHR DNA binding domains, respectively. (D) Comparison of protein binding activity between the e5 and PFαR₄ sequences by the DNA binding domains of the wild-type (PF), paired-domain mutant (PD-NH2) and homeodomain mutant (HD-C) PAX3-FKHR. The conditions for the gel shift analysis are the same as those described for panel C. It should be noted that data presented panel C are extracted from the same gel and autoradiogram at a single exposure time.

FIG. 5.

DNase I footprint analysis of the PDGFαR promoter by the DNA binding domains of PAX3-FKHR. The DNase I-digested PDGFαR promoter fragment in the presence of increasing amounts of bacterially expressed PAX3 paired-plus-homeodomain protein (GST-PdHd, lanes 2 to 5) reveals a protected region of approximately 22 to 24 bp (solid vertical bar, protected sequence on right) corresponding to the PFRE. Three ATTA sequences (TAAT₂, ATTA₃, and ATTA₄) are contained within the footprinted region. Lane 1 contains digested probe alone without GST-PdHd protein. Lane 6 (G) is a guanine Maxam-Gilbert sequence ladder.

To evaluate the role of the PDGFαR P3 site as well as the other single ATTAs in the PDGFαR promoter, we designed four overlapping oligonucleotides covering most of the 87-bp PFRE sequence (Fig. 6A). These oligonucleotides were designed to contain (i) no ATTA sequence (PFαR₁), (ii) ATTA₁ and TAAT₂ (PFαR₂), (iii) ATTA₃ and ATTA₄ (PFαR₃), or (iv) TAAT₂, ATTA₃, and ATTA₄ (PFαR₄). Each of these four oligonucleotides was tested for PAX3-FKHR protein binding (Fig. 6) and transactivation (see Fig. 7). As shown in Fig. 6C, the PFαR₄ oligonucleotide was bound far more efficiently than any of the other oligonucleotides by COS cell-expressed wild-type PAX3-FKHR. Despite adequate HD-C protein expression as determined by Western blot analysis (Fig. 6B), the PFαR₄ oligonucleotide was also bound by the paired-domain mutant PD-NH2 but not by the homeodomain mutant HD-C (Fig. 6D). The control e5 sequence behaved as expected in that it interacted only with the wild-type protein and not with either mutant protein. We believe that the more rapidly migrating complex (indicated by the arrowhead in Fig. 6D) represents the probe bound to degraded PAX3-FKHR protein. The amount of this degradation product gradually increases with increasing age and freeze-thaw cycles of the nuclear extract preparation.

FIG. 7.

Effect of transactivation of the PFRE by PAX3 and PAX3-FKHR. Transactivation of PFRE-driven TKCAT by PAX3 and by wild-type or mutant PAX3-FKHR is shown. P19 cells were transfected with pTKCAT reporter constructs containing double copies of PFRE oligonucleotides (PFαR₁ to PFαR₄) in the presence of wild-type or mutant PAX3-FKHR expression vectors. Fold induction of CAT activity was calculated as the level of change in the PFRE-driven TKCAT activities in response to PAX3-FKHR stimulation over that of TKCAT. The percentage of CAT activity obtained from the TKCAT construct in the absence of stimulation was assigned an arbitrary value of 100%. The values represent means and standard deviations for 5 (±2) determinations. The standard deviation is defined as the root mean square deviation of n − 1 determinations.

In support of the above-described gel shift analysis, the PFαR₄ oligonucleotide in the context of the pTKCAT reporter construct (CAT gene under the control of the thymidine kinase promoter) is the only one tested that could respond to PAX3-FKHR activation (Fig. 7). The PFαR₄ oligonucleotide responded to the PAX3-FKHR mutants in the same fashion as the intact PDGFαR promoter; that is, it could be transactivated by the PD-NH2 mutant but not by the HD-C mutant. Collectively, these results implicate the P3 site as the critical region within the 87-bp PFRE mediating transactivation by PAX3-FKHR.

Mutation of the P3 site completely abolishes the activity of the PFRE.

To further assess the role of the P3 site, we generated several mutant PFαR₄ oligonucleotides that contained nucleotide substitutions within each of the three ATTA sequences (TAAT₂, ATTA₃, and ATTA₄) (Fig. 8A) and determined if any of these mutations would disrupt the ability of PFαR4 to interact with PAX3-FKHR in gel shift or transactivation assays.

FIG. 8.

Effect of ATTA mutation on the protein binding activity of the PFRE. (A) Schematic illustration of the different ATTA mutant oligonucleotides of PFαR₄. (B) Western blot analysis of bacterially expressed wild-type and mutant GST fusion proteins. A total of 10 μg of bacterial lysate containing the fusion protein was used for the analysis under the conditions described in Materials and Methods. (C) The binding activity of the mutant oligonucleotides was tested by EMSA analysis in the presence of 35 ng of different GST fusion proteins that contained the DNA binding domains of wild-type PAX3-FKHR and mutants PD-NH2 and HD-C. The EMSA conditions used in these experiments were similar to the conditions described in Materials and Methods, except in the presence of 0.05 μg of poly(dI) · (dC) per μl.

We tested the abilities of the PAX3-FKHR DNA binding domains (expressed as a GST-fusion protein and here designated the GST-PF) to bind to wild-type and mutant PFαR₄ sequences (Fig. 8C). For verification of the binding specificity, we also included two PAX3-FKHR mutants, PD-NH2 and HD-C, in parallel assays. All three GST fusion proteins were adequately expressed as demonstrated by Western blot analysis (Fig. 8B). As predicted, the e5 probe was efficiently bound by the GST-PF but not by GST–PD-NH2 and GST–HD-C. The wild-type PFαR₄ was bound efficiently by the GST-PF and GST–PD-NH2 but not by GST–HD-C. The mutant oligonucleotides, however, showed binding patterns that can be summarized as follows. Mutation of TAAT₂ (M₂) significantly reduced the binding by GST–PAX3-FKHR and severely affected the binding by GST–PD-NH2. Mutation of ATTA₃ (M₃) completely disrupted the ability of the GST-PF and GST–PD-NH2 to bind. On the other hand, mutation of ATTA₄ (M₄) had little effect on the binding by the GST-PF and GST–PD-NH2. Collectively, these results indicate a critical role of TAAT₂ and ATTA₃ in the binding specificity of the PFRE.

In transactivation assays, both M₂ and M₃ mutants that had reduced or lost binding activity also failed to exhibit transcriptional activation upon PAX3-FKHR cotransfection (Fig. 9A). Similar findings were obtained when we mutated each of the three ATTA sequences within the context of the promoter (−912αRCAT construct). As shown in Fig. 9B, mutation of the TAAT₂ and ATTA₃ sites completely abolished the responsiveness of the native PDGFαR promoter to PAX3-FKHR stimulation, whereas mutation of the ATTA₄ site had little effect on the responsiveness.

FIG. 9.

Effect of ATTA mutation on the ability of the PFRE to respond to PAX3-FKHR transactivation. (A) Double copies of the ATTA mutant oligonucleotides were subcloned into the TKCAT reporter construct and tested in P19 cells for their ability to respond to PAX3-FKHR stimulation. Cotransfection was performed with wild-type and mutant PAX3-FKHR expression vectors. Fold induction of CAT activity was calculated as the level of change in the oligonucleotide TKCAT activities in response to PAX3-FKHR stimulation over the level of change in the TKCAT activity in response to PAX3-FKHR stimulation. The percentage of TKCAT activity obtained from the TKCAT construct in the absence of stimulation was assigned an arbitrary value of 100%. Results are averages from 4 (±1) transfections. (B) Transactivation of wild-type and mutant PDGFαR promoter constructs by PAX3-FKHR. The three ATTA sequences of the PFRE within the −912αRCAT construct were individually mutated by PCR-directed site mutagenesis. The wild-type and mutant −912αRCAT constructs were transiently transfected into P19 cells and tested for β-galactosidase and CAT activities. Fold induction was calculated as that over the CAT activity measured in cells transfected with the promoterless pGEMCAT. The percentage of conversion in CAT activity of pGEMCAT was then assigned an arbitrary value of 100%. The values represent means and standard deviations for 4 (±1) determinations. The standard deviation is defined as the root mean square deviation of the n − 1 determinations.

DISCUSSION

Recent elucidation of the molecular structures of several tumor-related chromosomal translocations has revealed that many result in the production of novel tumor-specific chimeric transcription factors. In some cases, the DNA binding domain of one transcription factor is fused to the transactivation domain of a second transcription factor. The altered properties of these chimeric transcription factors lead to aberrant patterns of gene expression that are likely to play significant roles in oncogenesis. The PAX3-FKHR chimera has been shown to have more potent transactivation properties than PAX3, although the DNA binding specificity appears to be identical to that of PAX3. In this report, we describe a novel mechanism for oncogenesis by this chimeric protein. We show that PAX3-FKHR has acquired a unique target gene specificity that enables it to activate a gene that is not normally activated by PAX3.

The novel specificity observed in the transactivation of the PDGFαR promoter by PAX3-FKHR is a consequence of a preferential selection for the homeodomain-mediated protein-DNA contacts. Although the PDGFαR promoter contains both a PAX3 paired-domain consensus sequence (−812) and several homeodomain consensus sequences (−646 to −549), our data convincingly demonstrate that only the homeodomain consensus sequences are required for transcriptional activation by PAX3-FKHR. This region of the PDGFαR promoter, designated the PFRE, contains a cluster of four ATTA core sequences. Specifically, the second and third ATTA repeats, which are inverted and separated by 3 bp, resemble a P3 paired-type homeodomain binding site (TAAT PyNPu ATTA). The P3 sites have been previously identified by random-site selection as sequences frequently recognized by the isolated Drosophila Prd homeodomain protein (63). Schafer et al. have demonstrated that PAX3 can bind inverted ATTA repeats in vitro (44), and recently, the crystal structure of the isolated Prd homeodomain bound to the P3 site has also been solved (64). However it is important to note that neither P3 nor its related sequences have been shown to play a functional role in PAX3-dependent gene regulation. The only known functional P3 site is that identified in the rhodopsin 1 promoter which mediates transactivation by the PAX6 homolog eyeless in Drosophila (48). Our data presented here indicate for the first time that this P3-like element in the PDGFαR promoter mediates both PAX3-FKHR DNA binding and PAX-FKHR transactivation. Mutation of either of the inverted ATTA sequences in the P3 site completely abolishes the ability of the PDGFαR promoter to respond to PAX3-FKHR activation (Fig. 9). Interestingly, although neither ATTA half site is by itself sufficient for mediating the PFRE response, mutation of one half site (ATTA₃) appears to disrupts PAX3-FKHR binding more than mutation of the other half site (TAAT₂) (Fig. 8). Since the configuration of the two ATTA half sites is palindromic, the binding data suggest that perhaps sequences flanking the P3 site might also affect binding affinity and/or stability.

Our data also clearly demonstrate that the paired domain of PAX3-FKHR is dispensable for transactivating the PDGFαR promoter. While the interaction between the paired domain and the homeodomain in the transcriptional activity of PAX3 and PAX3-FKHR remains undefined, the analysis of transcriptional activators that contain only a paired domain (such as PAX1 and PAX9) or only a paired-type homeodomain (such as gsc and lune) supports the notion that these two types of DNA binding motifs can function independently of each other. In our studies, we show that PAX3-FKHR mutations that are known to abolish DNA binding by the paired domain (e.g., Un-1 and Bu35) do not affect transactivation of the PDGFαR promoter. In fact, deletion of the entire amino-terminal region of the paired domain (PD-NH2) also has no effect. Likewise, deletion of the bisected FKHR WH domain does not affect transactivation of the PDGFαR promoter. In contrast, mutations that abolish the ability of the homeodomain to bind to DNA (e.g., HD and HD-C) completely abrogate the transactivation. Hence, these results show that the homeodomain of PAX3-FKHR alone not only is required but also is sufficient for the transcriptional activation of the PDGFαR promoter. Interestingly, a mutation in the homeodomain (S268A) that impairs the ability of the isolated PAX3 homeodomain to form homodimers (19) significantly reduces the ability of PAX3-FKHR to activate the PDGFαR promoter. This suggests that protein-protein interaction might be an important step in the transactivation of the PDGFαR promoter by PAX3-FKHR. Further analysis will be needed to determine whether full-length PAX3-FKHR forms homo- or heterodimers when binding to the PDGFαR promoter in vivo.

The discovery of a homeodomain-mediated PAX3-FKHR function prompts an intriguing question; that is, if both PAX3 and PAX3-FKHR proteins contain the same DNA binding modules, why is PAX3-FKHR able to transactivate the PFRE while PAX3 is not? This difference could result from two possible mechanisms: a difference in their DNA binding abilities or a difference in their abilities to transactivate once bound. We have preliminary evidence indicating that in vitro-translated full-length wild-type PAX3 appears to be capable of binding the PFRE (57a). If this was indeed the case, the result would imply that DNA binding alone is not sufficient to induce transactivation. Potentially, PAX3 and PAX3-FKHR bind to PFRE by different mechanisms and by adopting critically different conformations. This possibility may be relevant, since Underhill and Gros (56) have recently demonstrated that the PAX3 paired domain has an inhibitory effect on the homeodomain binding to a P3 site. Deletion of the paired domain appears to abrogate this negative effect, allowing efficient homeodomain binding to the P3 site. It is therefore possible that in the case of PAX3-FKHR, a conformational change is introduced during the fusion process that allows an independent homeodomain binding to the PFRE despite the presence of an intact paired domain. It is also possible that the dimerization properties of PAX3-FKHR also differ from those of PAX3. In addition, PAX3 might complex with an unknown cofactor(s) that prevents it from activating transcription of PDGFαR, an interaction that has been overcome by the PAX3-FKHR fusion. In support of the notion that PAX3 does not regulate PDGFαR expression, we have been unable to detect any difference in PDGFαR expression in wild-type versus homozygous mutant Splotch mouse embryos lacking PAX3 as determined by in situ hybridization (17a).

Finally, the identification of the PDGFαR promoter as a target for PAX3-FKHR action may provide new insights into processes of muscle cell transformation and tumorogenesis. Developmentally regulated PDGFαR expression is known to be involved in the development of several mesenchymal cell lineages, including the dermomyotome (38, 40, 41, 43, 45). PDGF has been shown to affect muscle development by activating proliferation of muscle precursor cells prior to their differentiation to form muscle fibers (32–34, 65). PDGF receptor expression is highest in proliferating muscle precursors and declines gradually to undetectable levels in differentiated muscle cells (32–34). Thus, down regulation of PDGF receptor expression may be an important transition step leading to muscle cell differentiation. Moreover, a knockout mouse that contains a disrupted PDGFαR gene develops a deficiency in myotome formation, suggesting that PDGFαR plays a critical role in muscle development (49). Since PDGF is also synthesized in normal muscle cells and muscle-supporting fibroblasts, constitutive activation of PDGF receptors could potentially cause uncontrolled muscle cell growth, leading to tumor formation by autocrine and/or paracrine regulatory pathways. A similar mechanism, i.e., establishment of autocrine and paracrine loops for PDGF, has been indicated in the pathogenesis of glioblastoma (29). Our observation that PAX3-FKHR can induce PDGFαR promoter activity leads us to speculate that aberrant PDGFαR function could play an important role in the establishment and/or maintenance of transformed RMS phenotypes. In the future, it will be of interest to determine if the expression level of PDGFαR correlates with the presence of PAX3-FKHR translocation and the phenotypes of RMS.

PAX3 and PAX3-FKHR have both been shown previously to be capable of activating expression of another cell surface tyrosine kinase receptor, c-Met. In that case, transcriptional activation is dependent on binding of the PAX3 paired domain to the c-Met promoter, and this activation is required for normal migration of limb muscle progenitor cells (2, 3). c-Met has also been shown to be up regulated in many, but not all, cell lines derived from PAX3-FKHR-expressing RMS cell lines (20). Since muscle also produces the ligand for c-Met, i.e., scatter factor/hepatocyte growth factor, inappropriate expression of c-Met in RMS expressing PAX3-FKHR has been postulated to lead to an autocrine loop enhancing tumorogenicity. The present results suggest that there may be multiple pathways whereby PAX3-FKHR promotes tumor formation, including at least two growth factor/receptor activation cascades. It will be of interest to test whether PDGFαR is activated in those cell lines in which c-Met is not overexpressed. It is also possible that these various pathways affect different stages of cancer progression; for instance, inappropriate PDGFαR expression could lead to uncontrolled cell proliferation, while c-Met activation could enhance migration and metastasis formation.

In conclusion, the PAX3-FKHR chimera protein is thought to play a crucial role in muscle cell transformation and the formation of RMS. Presently, little is known about the molecular mechanism(s) underlying this transformation process. We have presented several lines of evidence to demonstrate that the tumor-specific PAX3-FKHR transcription factor has the ability to activate the PDGFαR gene, which is not normally regulated by PAX3 under the same testing conditions. This finding provides new insights into the molecular basis for PAX3-FKHR oncogenesis as well as new avenues for designing approaches to specifically inactivate the PAX3-FKHR transforming activity without affecting normal PAX3 function.