A Dominant-negative Receptor for Type f? Transforming Growth Factors Created by Deletion of the Kinase Domain*

To prove the postulated role of type P transforming growth factors (TGFP) in cardiac development and other events, specific inhibitors of TGFP signal transduction are needed. We truncated the type I1 TGFP receptor cDNA (AkTPRII), to delete the predicted serinekhreonine kinase cytoplasmic domain. AkTPRII was co-transfected into neonatal cardiac myocytes, together with reporter constructs for two cardiac-restricted genes that are regulated antithet-ically by TGFP. AkTPRII impaired activation of the skeletal a-actin promoter by TGFP1, -2, and -3 and, conversely, impaired TGFP inhibition of a-myosin heavy chain transcription. Thus, a kinase-defective TPRII blocks signaling by all three mammalian TGFP isoforms, and can disrupt both positive and negative control of transcription by TGFP. a-actin -39#+24SkALuc, constructed by subcloning nucleotides -394 to +24 of the chicken SkA gene as an RsaI-Hind111 fragment between the SmaI and HindIII sites of the fire- fly luciferase reporter expression vector The aMHC-lu- ciferase reporter, 5500aMHCLuc, by subcloning the in-tergenic region between the murine pMHC and aMHC gene loci as a 5.5-kilobase pair Kpn I-Hind111 fragment into pXPl fragment was selected because previous studies had proven its fidelity to tissue-restricted, stage-specific, and thyroid hormone-dependent expression of endogenous aMHC (33), and because TGFP-inhibited ele- ments cannot not yet be ascribed to smaller portions of the gene. The constitutive P-galactosidase expression vector, pCMVp, places the Es- cherichia lac2 gene under the transcriptional control of the cytome-galovirus immediate-early

(ll), and to control the expression of at least six cardiac-restricted genes (12,13). Unlike the global suppression of differentiated gene expression by TGFp in skeletal muscle (14-161, neonatal cardiac myocytes possess a continuum of responses to TGFp1: up-regulation of a gene ensemble, including skeletal a-actin (SkA), expressed preferentially in fetal myocardium, concurrent with down-regulation of genes including a-myosin heavy chain (aMHC) that are associated with adult ventricular muscle (12,13), dichotomous responses which correspond to the generalized "fetal" phenotype produced by mechanical load (4,17). Positive and negative control of developmentally regulated genes thus coexist i n this system, making the cardiac myocyte particularly intriguing as a model for studies of TGFO signal transduction. Investigations of pluripotent cell lines (18), amphibian cardiac progenitor cells (191, and avian cardiac endothelium (20) also suggest that TGFP-related peptides might regulate cardiac organogenesis itself. However, mechanistic tests of this hypothesis require a suitable inhibitor of the TGFp signaling cascade and TGFP-dependent gene expression.
Neonatal cardiac myocytes possess all three of the characteristic cell surface receptors for TGFp (TOR) visualized by receptor cross-linking (21). Expression cloning proved the type I1 TPR, a 75-kDa glycoprotein, to possess an intracellular domain distinct from the four classes of tyrosine kinase found in the receptors for platelet-derived, epidermal, insulin-like, and fibroblast growth factors (22). Instead, TpRII resembles the type I1 receptor for activin, a distant member of the TGFp superfamily, and Daf-1, a protein controlling larva formation in Caenorhabditis elegans; all three constitute a novel class of transmembrane protein with a consensus serinehhreonine kinase as the predicted cytoplasmic signaling domain (22)(23)(24)(25).
TpRII is fully functional in the absence of the type I11 receptor @-glycan), illustrated by the absence of this proteoglycan from L6 myoblasts (14,26,27). TpRII is competent to bind TGFp in the absence of the type I TpR, a 53-kDa protein whose structure has not yet been defined, but signal generation apparently requires a heteromeric protein complex involving both receptors I and 11 (28).
Kinase-defective mutations of receptor tyrosine kinases are known to inhibit the function of wild-type receptors, possibly by a block to the intermolecular autophosphorylation that follows ligand-induced dimerization (29-31). Although the corresponding initial aspects of TGFp signal transduction are less completely understood, we reasoned that a truncated TpRII, lacking the serinekhreonine kinase domain, would function as a dominant inhibitor of TGFP-dependent transcription. We have used the cardiac myocyte model to demonstrate that the truncated TpRII confers resistance to TGFp control of developmentally regulated cardiac genes.

EXPERIMENTAL PROCEDURES
Plasmids-To generate the truncated human TPRII by PCR amplification, each 100-pl reaction mixture contained 10 ng of TPRII clone H2-3FF (22), 600 ng of the primers shown, 200 VM of each dNTP, 50 m M KCl, 1.5 m M MgCl,, 10 m M Tris-HC1, pH 8.0, and 5 units of Taq polymerase (Promega). Amplification comprised 5 min of initial denaturation at 94 "C, then 30 cycles (1 min at 94 "C, 1.5 min at 72 "C, and 1 min at 60 "C) using a Perkin-Elmer Cetus DNA thermal cycler. The final extension reaction was for 7 min at 72 "C. The resulting PCR product was analyzed on an 8% polyacrylamide gel and had the expected size of 883 nucleotides. For directional subcloning, the products of three PCR reactions were combined, purified with a Centricon 100 spin column, digested with EcoRI and HindIII, and loaded on a 1.2% agarose gel. The DNA band was excised, and DNA was isolated with the Quiaex gel extraction kit (Qiagen). For expression in eukaryotic cells, AkTpRII was subcloned between the EcoRI and HindIII sites of pSV-Sport1 (GIBCO/ BRL), under the control of the SV40 early promoter and enhancer. The truncated activin receptor cDNA comprised nucleotides 36-787 of pmActR2 (23). Clones were sequenced by the dideoxy method using Sequenase 2.0 (U. S. Biochemical Corp.).
The skeletal a-actin reporter, -39#+24SkALuc, was constructed by subcloning nucleotides -394 to +24 of the chicken SkA gene as an RsaI-Hind111 fragment between the SmaI and HindIII sites of the firefly luciferase reporter expression vector pXPl (32). The aMHC-luciferase reporter, 5500aMHCLuc, was prepared by subcloning the intergenic region between the murine pMHC and aMHC gene loci as a 5.5-kilobase pair Kpn I-Hind111 fragment (33) into pXPl (Fig. 2 B ) . This fragment was selected because previous studies had proven its fidelity to tissue-restricted, stage-specific, and thyroid hormone-dependent expression of endogenous aMHC (33), and because TGFP-inhibited elements cannot not yet be ascribed to smaller portions of the gene. The constitutive P-galactosidase expression vector, pCMVp, places the Escherichia coli lac2 gene under the transcriptional control of the cytomegalovirus (CMV) immediate-early promoter (34). lated as previously described from 1-2-day-old rats (12, 13). Myocytes Cell Culture and Zhnsfection-Neonatal cardiac myocytes were isowere purified by density centrifugation through a Percoll step gradient (1.050 g.ml", 1.060 g.ml-' and 1.082 gml" Percoll in 116 mM NaCl, 406 p~ MgCl,, 11 mM NaH2P04, 5.5 mM glucose, 39 mM HEPES, pH 7.3, 0.002% phenol red) and were plated a t a density of 1 X lo6 cells/35-mm dish (Primaria, Falcon). Cells were cultured overnight in Dulbecco's modified Eagle's m e d i u d a m ' s n u t r i e n t mixture F-12 (l:l), 17 mM HEPES, 3 mM NaHCO,, 2 mM L-glutamine, 50 pg.ml-l gentamicin, 10% horse serum, and 5% fetal bovine serum. Cells were transfected 24 h after plating by a diethylaminoethyl-dextran sulfate method (10 pg of of CMV-ZacZ). Cells were incubated with the DNA-DEAE-dextran com-AkTpRII or pSV-Sportl, 7.5 pg of a luciferase reporter construct, 2.5 pg plex for 3 h, then for 60 s with 10% dimethyl sulfoxide in Dulbecco's modified Eagle's medium. Cells were cultured overnight in the medium described above, which was replaced on the following day by serum-free medium containing 1 pgml-' insulin, 5 pg.ml-' transferrin, 1 nM Na,Se04, 1 nM LiCl, 25 pg,ml-' ascorbic acid, and 0-1.0 nM thyroxine (12,13). Porcine TGFpl and -2 and chicken TGFp3 (R&D Systems) were added at final concentration of 1 ngm-', and the medium and growth factor were replaced at 16 h.
Luciferase and p-Galactosidase Assays-Cells were harvested after 36 h in the presence or absence of TGFp, in 150 pl of 25 mM Tris phosphate, pH 7.8, 2 mM dithiothreitol, 2 mM CDTA, 10% glycerol, 0.1% Triton X-100. Luciferase activity was monitored as the oxidation of luciferin in the presence of coenzyme A (35), using an Analytical Luminescence model 2010 luminometer. For lac2 determinations, extracts were incubated with 4.85 m g m -' chlorophenol red-P-D-galactosidase (Boehringer Mannheim), 62.3 mM Na,HPO,, 1 mM MgCl,, 45 mM p-mercaptoethanol for 1 4 h at 37 "C (36) and activity was measured as absorbance at 575 nm. Results were compared by Scheffe's multiple comparison test for analysis of variance and the unpaired two-tailed t test, using a significance level o f p < 0.05.

RESULTS
To construct the truncated TPRII (AkTpRII), we subjected the human TORI1 cDNA H2-3FF (22) to PCR amplification, using primers that correspond to nucleotides 306-326 and 1153-1172 and incorporated asymmetric linkers for directional cloning (Fig. 1) To determine whether AkTPRII could prevent up-regulation of a "fetal" cardiac gene by TGFp1, we co-transfected neonatal rat cardiac myocytes with (i) AkTpRII or the equivalent vector lacking insert, to control for promoter competition by SV40 sequences; (ii) the SkA-luciferase reporter, -394/+24SkALuc; and (iii) the constitutive lac2 gene, pCMVp, a to correct for transfection efficiency, cell recovery, and potential global effects of the growth factors (Fig. 2). Numerical values for expression of the actin and myosin reporter genes shown below (luciferase, corrected for lacZ) are normalized to that in simultaneous cultures in the absence of TGFP and AkTpRII. In overall agreement with previous results using a related SkA reporter (13), the -394/+24SkA-luc construct was expressed at levels at least 100-fold greater than in parallel cultures of cardiac fibroblasts' and was up-regulated by 1 ngm-' TGFpl (2.748 5 0.382, compared to the vehicle control; p = 0.0005; Fig. 2.4). AkTpRII reduced SkA expression in the presence of TGFpl to the basal level found in vehicle-treated cells ( p = 0.0005, uersus the vector control). By contrast to expression triggered by exogenous TGFpl, whose suppression by the truncated TGFp receptor was virtually complete, AkTPRII had little effect on basal, tissue-specific transcription of the SkA promoter. The modest inhibition which was observed was reproducible and significant (0.670 2 0 . 0 7 2 ;~ = 0.0156), consistent with findings by Roberts et al. (11) that TGFp is secreted in culture by neonatal rat cardiac myocytes and acts in an autocrine fashion on the cells. The block to SkA induction was specific to AkTPRII; no inhibition of basal or TGFp-induced transcription resulted from AkAcRII, a corresponding truncation of the murine type I1 activin receptor. By contrast, as shown in Fig. 3, exogenous fulllength TPRII amplifies the induction of SkA transcription by TGFp1. Thus, the block to TGFP-dependent transcription by AkTpRII is contingent on truncation of the cytoplasmic domain. One stringent test for the specificity of dominant-negative mutations is whether exogenous wild-type protein can rescue the mutant phenotype. Increasing the amount of fulllength TpRII cDNA progressively restored the responsiveness of cardiac muscle cells to TGFP1, despite a constant amount of the expression vector encoding AkTpRII. As was true for the truncated activin receptor (37), complete rescue required less than stoichiometric amounts of the wild-type receptor (cf, .  Biological actions of the TGFp isoforms, while often similar, differ drastically in some systems. As one illustration, only TGFP3 is implicated in the epithelial-mesenchymal transformation required for creation of cardiac valves (20). Therefore, to ascertain whether AkTPRII might disrupt signaling by more than one form of TGFP, we first tested if neonatal rat cardiac myocytes in fact possess transcriptional responses to TGFp2 and 4 3 (Fig. 2). TGFP2 and $3 each induced the SkApromoter to at least the same extent as TGFpl(3.882 2 0.506 and 3.910 To facilitate analysis of a TGFP-inhibited pathway in cardiac muscle cells, we generated an cYMHC-luciferase construct, since aMHC is the cardiac gene whose expression, at the mRNA level, is repressed most completely by TGFPl (12,13). As shown in Fig. 4, cYMHC-luciferase activity was highly dependent on thyroid hormone (1.000 t 0.117 uersus 0.093 2 0.028 at 1 and 0 nM; p = 0.0017) and was inhibited nearly 70% by TGFPl (0.344 ? 0.088; p = 0.0110). AkTPRII specifically abolished down-regulation by TGFP1, with no effect on up-regulation by T3. Thus, AkTpRII impairs TGFP-dependent signals for both negative and positive control of gene expression, without spurious effects on a TGFP-independent pathway. In agreement with this evidence that AkTPRII specifically disrupts TGFP-dependent transcription, activity of the CMV-driven lac2 gene was indistinguishable in AkTPRII-and vector-transfected cells.3 DISCUSSION Identification of the cDNA sequence for TpRII has provided a critical opportunity to construct a truncated receptor variant, as a reagent to interdict TGFp signal transduction at the level of the receptor itself. These experiments indicate that deletion of the serindthreonine kinase domain generates a trans-dominant inhibitor of TGFP signal transduction. In overall agreement with this conclusion, Melton and colleagues (37) have recently shown that a kinase-defective form of the homologous type I1 activin receptor can block activin-dependent events in early Xenopus embryos. Thus, alteration of the cytoplasmic signaling domain may be a generic strategy for producing lossof-function mutations, not only in receptor tyrosine kinases but also in those whose action depends on a serindthreonine kinase domain. An additional inference to be drawn from both studies is that mutation of the respective type I1 receptors is sufficient to repress signal transduction with no need for concomitant mutation of other proteins in the ligand-binding complex (25, 28). By analogy t o receptor tyrosine kinases, mutations of TpRII and the type I1 activin receptor might be expected to act through a block to autophosphorylation in trans. Recently, Massague and colleagues (28) have confirmed the prediction that kinase activity is essential for signaling but is superfluous for ligand binding by TPRII. Thus, other mechanisms are plausible to account for the dominant-negative action, including sequestration of TGFp and activin by the truncated receptors or, conceivably, impaired expression of normal type I1 receptor. An additional caveat is the potential, for which credible support exists (381, that unexpected tyrosine kinase activity could also be inherent to this class of transmembrane protein. The ability of all three isoforms of TGFp t o activate the SkA promoter in neonatal rat ventricular myocytes concurs with their shared ability to antagonize depressive effects of interleukin-1B on beating rate and equivalence for binding to cardiac cells (11). Analogously, the fact that AkTpRII blocks gene activation by all three peptides agrees with their equal potency for inhibition of DNA synthesis in receptor-defective DR-27 mink lung cells transfected with the full-length human TpRII (28). Thus, our results with the dominant-negative TpRII corroborate the conclusion that TpRII acts as a receptor for all three mammalian TGFp isoforms (28).
In contrast to other TGFP-regulated genes, activation of SkA transcription by TGFpl appears to be mediated largely via a proximal serum response factor (SRF)-binding element (SRE) and a potential TEF-1 site, which are indispensable as well for basal, tissue-restricted expression.2 However, the 3' arm of this SRE possesses a n overlapping recognition site for a second SRE-binding protein, the bifunctional transcription factor W1 (39), a competitive antagonist for SRF at this location (40, 41). It is unknown whether TGFp acts through up-regulation of SRF, modification of SRF, or, conceivably, decreased W1 activity. cis-Acting sequences for TGFp repression of aMHC have not yet been delineated, but candidate elements within the 5"flanking region that was required for cardiac-specific expression in vivo include consensus sites both for SRF and the SRFrelated MADS box protein, 42).
Genetic methods to obtain a mechanistic understanding of growth factor signal transduction can be confounded by ambiguous of counterintuitive results from conventional techniques used to create gainor loss-of-function mutations. For example, induction of endogenous Fos and Jun by mechanical stress is T. Brand, W. R. MacLellan, and M. D. Schneider, unpublished observations. associated with up-regulation of atrial natriuretic factor in ventricular muscle cells (171, not repression as seen with forced expression (43). Similarly, despite the importance of homologous recombination, an increasingly encountered shortcoming of this strategy is the risk of a misleading or false-negative outcome after disrupting only one member of a redundant multi-gene family. This may be the case for knock-out mutations in TGFpl (44); other recent examples include Fos (4.9, MyoD (461, and E2A proteins (47). Thus, dominant-inhibitory genes such as AkTPRII offer a crucial alternative to other procedures for generating loss-of-function mutations. Indeed, there exist corresponding dominant-negative forms of TGFP itself (48). The finite growth capacity of cardiac muscle cells in culture precludes stable transfection as a means to uniformly modify ventricular myocytes, which would be required to assess the impact of AkTPRII on other aspects of the cardiac phenotype such as endogenous genes and gene products, signaling intermediaries, or DNA synthesis. This limitation can be overcome using replication-defective recombinant adenovirus to achieve efliciencies for gene transfer that approach 100% in neonatal and even adult ventricular muscle cells.4 However, cell culture model systems substitute only partially for investigations of cardiac organogenesis itself. By analogy to their use in Xenopus oocytes (31, 371, dominant-negative genes like AkTPRII might serve as a generic approach, complementary to gene ablation, to create loss-of-function mutations in transgenic mammals.