Inhibition of vascular endothelial cell growth by activin-A.

The demonstration of type 2 activin receptor expression in human umbilical vein endothelial cells prompted an investigation of the effects of the activin/inhibin family of hormones on vascular endothelial cell growth. Recombinant activin-A inhibited [3H]methylthymidine uptake and growth of a panel of endothelial cell types; recombinant inhibin-A was without effect. Affinity cross-linking studies demonstrated the presence of type 1 and type 2 activin receptors on the surface of bovine aortic endothelial cells, while detailed analysis of type 2 activin receptor expression revealed both type 2 and type 2B activin receptor mRNA in all endothelial cell types analyzed. In addition, capillary endothelial cells were found to express activin-beta A subunit mRNA and protein, the levels of which were increased in response to transforming growth factor (TGF)-beta. Furthermore, activin-A and TGF-beta caused additive inhibition of capillary endothelial cell [3H]methylthymidine uptake. These findings implicate the activins in the regulation of endothelial cell function, and suggest that TGF-beta and activin may interact to inhibit capillary endothelial cell growth.


The demonstration of type 2 activin receptor expression in human umbilical vein endothelial cells prompted an investigation of the effects of the activin/ inhibin family of hormones on vascular endothelial cell growth. Recombinant activin-A inhibited ['Hlmethylthymidine uptake and growth of a panel of endothelial cell types; recombinant inhibin-A was without effect.
Affinity cross-linking studies demonstrated the presence of type 1 and type 2 activin receptors on the surface of bovine aortic endothelial cells, while detailed analysis of type 2 activin receptor expression revealed both type 2 and type 2B activin receptor mRNA in all endothelial cell types analyzed. In addition, capillary endothelial cells were found to express activin-bA subunit mRNA and protein, the levels of which were increased in response to transforming growth factor (TGF)-/3. Furthermore, activin-A and TGF-8 caused additive inhibition of capillary endothelial cell ['Hlmethylthymidine uptake. These findings implicate the activins in the regulation of endothelial cell function, and suggest that TGF-/3 and activin may interact to inhibit capillary endothelial cell growth.
Abnormal endothelial cell proliferation is associated with a number of pathological states, for example, tumor vascularization (Folkman and Shing, 1992). Mechanisms by which endothelial cell growth is controlled are therefore of widespread interest, and although much attention has been paid to the role of endothelial cell mitogens in this respect, it is equally likely that endothelial proliferation could result from a release from inhibitory growth control. Several inhibitors of endothelial cell growth have been identified, including the transforming growth factor-@, interleukin-1, tumor necrosis factor, and platelet factor-4 (Folkman and Shing, 1992;Bick-ne11 and Harris, 1991).
The activins and inhibins were originally characterized as gonadal regulators of follicle stimulating hormone release from the anterior pituitary, and are now known to exert widespread influence over mammalian reproductive function (De Jong, 1988Vale et al., 1990. Members of the TGF-8' superfamily, the mature proteins are disulfide-bridged homo-* This work was supported by the Imperial Cancer Research Fund.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
High affinity cell surface receptors for activin have been identified on a number of gonadal and non-gonadal cell types (Campen and Vale, 1988;Sugino et al., 1988;Hino et al., 1989;Kondo et al., 1989;Centrella et al., 1991;Shao et al., 1992). By analogy to the TGF-fi receptor system, these can be classified as type 1 and type 2 receptors (Roberts and Sporn, 1990;Massague, 1990). Two subtypes of the type 2 activin receptor have recently been characterized by cDNA cloning and shown to define a novel class of transmembrane receptor proteins, the putative receptor serine-threonine kinases (Mathews and Vale, 1991;Attisano et al., 1992;Kondo et al., 1991;Mathews et al., 1992;Legerski et al., 1992). A type 2 TGF-fi receptor has also recently been characterized by cDNA cloning, and, as predicted by cross-linking studies, shown to belong to the same receptor family as the type 2 activin receptors . Thus, it is possible that all members of the TGF-P superfamily employ receptors of this class to ellicit cellular responses (Massague, 1992).
In addition to regulation of reproductive function, the activins/inhibins regulate many other cellular processes. Activin-A is identical to erythrocyte differentiation factor, a potent stimulator of red blood cell formation (Et0 et al., 1987;Yu et al., 1987), influences cells of the central nervous system (Sawchenko et al., 1988;Schubert et al., 1990), induces mesoderm and axial structures in early vertebrate development van der Eijnden-Van Raaji et al., 1990;Mitrani et al., 1990;Thomsen et al., 1990;Hemmati-Brivanlou and Melton, 1992) and cooperates with bone morphogenetic proteins to promote bone formation (Ogawa et al., 1992). We report here a study of the role of activin-A in regulation of endothelial cell growth, prompted by the finding that human umbilical vein endothelial cells express the type 2 activin receptor, ACTRP.

EXPERIMENTAL PROCEDURES
Materials-Recombinant human activin-A and inhibin-A were a gift of Genentech Inc., San Francisco, CA. All radionucleotides were from Amersham Corp. TGF-@1 was from British Biotechnology Ltd., Oxford, UK. T3 RNA polymerase was from Boehringer Mannheim, T7 RNA polymerase from Life Technologies, Inc. Disuccinimidyl suberimidate was from Pierce Chemical Co. Anti-activin-PA subunit monoclonal antibody (Groome and Lawrence, 1991) was the gift of Professor Nigel Groome, Oxford-Brookes University, Oxford, UK. All tissue culture media were from ICRF Central Services.
Cell Isolation and Culture-Human umbilical vein endothelial cells were isolated from fresh umbilical cords by dissociation with a 20% solution of "Dispase" (Collaborative Research Ltd.) in phosphatebuffered saline and used between passage 2 and 4. Bovine adrenal capillary endothelial cells were isolated and characterized as described elsewhere (Fawcett et al., 1991;McCarthy and Bicknell, 1992). HU-VECs and BACE were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, heparin (50 pglml), and basic FGF (1 ng/ml). Bovine aortic endothelial cells were isolated from fresh bovine aorta by collagenase digestion followed by fluorescence-activated cell sorting of the resultant cell population using dilacetyl low density lipoprotein (Biogenesis) as an endothelial label the American Type Culture Collection, Bethesda, MD (ATCC CCL (Voyta et al., 1984). Calf pulmonary artery endothelial cells were from 209). BAEC and CPAE were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. All endothelial cells were cultured in Petri dishes coated with a solution of 0.1% gelatin in phosphate-buffered saline.
Ribonuclease Protection Assays-RNase protection assays were performed as described in Ausubel et al. (1987). All template cDNAs used in transcription reactions were generated by polymerase chain reaction from appropriate libraries, ligated into the EcoRV site of pBluescript and sequenced to determine orientation. Probes were as follows; ACTR2-bovine ACTR2 cDNA complementary to nucleotides 232-661 of mACTR2 (Mathews and Vale, 1991), ACTR2B-bovine ACTR2B cDNA template complementary to nucleotides 641-1093 of mACTR2B (Attisano et al., 19921, activin-PA subunit-nucleotides 176-476 bovine P A cDNA (Forage et al., 1986). Riboprobes were labeled using T3/T7 RNA polymerases and [LY-~'P]CTP and hybridized at 45 "C to 10-pg samples of total cellular RNA. Free probe was digested with RNase T1 at room temperature, and protected fragments were analyzed on polyacrylamide sequencing gels followed by autoradiography.
Activin Binding and Cross-linking-Activin-A was iodinated to a specific activity of between 10 and 100 pCi/pg by chloramine T oxidation. Briefly, 1-5 pg of protein in 50 pl of phosphate-buffered saline was mixed with 0.5 mCi carrier-free NalZ5I (Amersham Corp.) and 25 pl 4 mg/ml chloramine T. Iodination was carried out for 2 min at room temperature and terminated by addition of 100 p1 of 100 mM L-tyrosine. '251-Activin-A was separated from unincorprated label by gel filtration on Sephadex PD-10 columns (Pharmacia LKB Biotechnology Inc.). For competition binding assays, cells were grown to confluence in 24-well plates, washed once with phosphate-buffered saline supplemented with 0.9 mM CaClz and 0.5 mM MgC12, and incubated with the indicated concentrations of lZ5I-activin-A and unlabeled ligands for 45 min at room temperature. Cells were then washed once with ice-cold supplemented phosphate-buffered saline and solubilized in 0.5 M NaOH, and counts were determined on a ycounter. For cross-linking studies, cells were grown to confluence on 50-mm dishes, incubated with the indicated concentrations of IZ5I. activin-A and unlabeled ligands as described above, washed once with PBS, and receptor-bound ligand cross-linked at room temperature for 10 min with 1 mM disuccinimidyl suberimidate. Membrane proteins were extracted in 50 mM Tris, pH 7.4, 1% Triton X-100 at 0 "C for 60 min, and subjected to reducing SDS-polyacrylamide gel electrophoresis and autoradiography to visualize cross-linked receptors. Western Analysis-2-ml samples of conditioned medium were concentrated by precipitation with 5 volumes of acetone, and the precip-itates were subjected to reducing SDS-polyacrylamide gel electrophoresis (15% gel). Proteins were transferred to Immobilon-P membrane (Millipore), and filters were blocked overnight in Tris-buffered saline containing 5% "Marvel" and 0.1% Tween. ACtiVin-PA protein was detected using the E4 anti-PA monoclonal antibody. Bound antibody was visualized using an anti-mouse-peroxidase conjugate (Dako) and ECL chemiluminescent detection (Amersham Corp.).

RESULTS
Effect of Actiuin-A on Endothelial Cell Growth-The finding that prompted this series of experiments was the cloning of a full-length cDNA encoding a type 2 activin receptor (Mathews and Vale, 1991) from a human umbilical vein endothelial cell (HUVEC) cDNA library by polymerase chain reaction amplification (data not shown). Sequence analysis of the HUVEC activin receptor revealed complete identity with human ACTR2 cDNAs isolated from testis (Donaldson et al., 1992;Matzuk and Bradley, 1992). Since activin-A exerts diverse effects on cellular growth and differentiation, ranging from induction of mesoderm during embryogenesis to stimulation of erythroid development, the effects of activin-A/inhibin-A on endothelial cell growth were subsequently investigated.
Recombinant human activin-A caused dose-dependent inhibition of [3H]methylthymidine incorporation into DNA in four endothelial cell types, namely BACE, CPAE, BAEC, and HUVEC (Fig. 1). The dose dependence of the inhibition for each cell type was similar, with half-maximal inhibition observed between 1 and 10 ng/ml(36 and 360 pM, respectively). Recombinant human inhibin-A had no effect on endothelial cell [3H]methylthymidine incorporation, neither did inhibin-A antagonize the growth inhibitory effects of activin-A (data not shown). To confirm that inhibition of [3H]methylthymidine uptake reflected growth inhibition, the effect of activin-A on endothelial cell growth was assessed over a 6-day period. Cells were treated every 2 days with activin-A at the indicated concentration. Consistent with inhibition of endothelial cell DNA synthesis, activin-A caused growth inhibition of HU-VEC, BACE, CPAE, and BAEC (Fig. 2). Activin-A did not affect cell attachment in any experiment. As with [3H]methylthymidine uptake, inhibin-A did not antagonize inhibition of endothelial cell growth by activin-A (data not shown).
Actiuin Receptor Expression in Endothelial Cells-The nature of the receptor system mediating growth inhibtory actions of activin-A on vascular endothelial cells was subsequently investigated. Affinity cross-linking of lZ5I-activin-A to BAECs using the bifunctional cross-linking reagent, disuccinimidyl suberimidate, revealed the presence of three activin binding proteins of 160, 80, and 65 kDa, binding to each of which was fully competed by unlabeled activin-A (Fig. 3a). Competition was complete in the presence of 25 ng/ml (0.9 nM) unlabeled ligand. Subtraction of the molecular mass of the cross-linked activin-PA subunit yields receptor molecular masses of approximately 50, 65, and 145 kDa. By analogy to the TGF-P receptor system, the 45-and 60-kDa species represent types 1 and 2 activin receptors, respectively (Massague, 1992). The identity of the 145-kDa component is unclear. Displacement binding analysis was also performed to investigate the specificity of activin binding sites on BAEC (Fig.  3b). Half-maximal displacement of bound lZ5I-activin-A was observed a t 2 ng/ml (70 PM) unlabeled activin-A, in close agreement with the affinity of the cross-linked activin receptor complexes for activin-A demonstrated in Fig. 3a. Bound lZ5I-activin-A was partially displaced by high concentrations of inhibin-A, but not by TGF-P1. '251-A~tivin-A binding was also detected on CPAE, but was undetectable in BACE or HUVEC under identical conditions. Expression of type 2 activin receptors was therefore investigated by RNase protec-   (Fig. 4). Both ACTR2 (Mathews and Vale, 1991) and ACTR2B  mRNAs were detected in BAEC, CPAE, BACE, and HUVEC confirming that all four endothelial cell types found to be activin responsive express activin receptors. staining of endothelium in vessels of many organs,' suggesting that in addition to responding to activin-A, vascular endothelial cells also express this ligand. Expression of activin-DA mRNA in capillary endothelial cells (BACE) was therefore investigated. As shown in Fig. 5, activin-pA mRNA was expressed at a low level in BACE cells treated with control S. Fox, manuscript in preparation. and IL-la (lane 6 ) , had no effect on PA mRNA levels. Thus, capillary endothelial cells express activin-pA subunit, expression of which is stimulated specifically by TGF-p. In similar experiments, inhibin-a subunit mRNA was not detectable in BACE cells, either by RNase protection or polymerase chain reaction of reverse-transcribed mRNA (data not shown).

Actiuin-pA Expression in Capillary Endothelial Cells-Im-
Expression of activin-pA protein by BACE cells was also investigated. Confluent cultures were treated either with control medium or with TGF-Bl (10 ng/ml) for 72 h, following In experiments using human-specific probes, strong signals for both ACTR2 and ACTR2B were obtained from HUVEC RNA (data not shown).

2 3 4 5 6 7
Activin-pa which conditioned medium was collected and analyzed for activin-pA subunit protein by Western blotting. As shown in Fig. 6, untreated BACE cells secreted mature 14-kDa activin-P A subunit at a low level (lane 1 ). TGF-pl-treated cells displayed an approximately 2-fold increase in secreted activin-PA subunit (lane 2), confirming that TGF-8 regulated expression of activin-BA mRNA results in increased secretion of mature activin-BA protein from BACE cells.

Actiuin-A and Endothelial Cells
Cell DNA Synthesis-Taken together with the lack of detectable inhibin-a mRNA expression in BACE cells, the ability of TGF-(3 to stimulate activin-pA expression pointed to TGF-(3 as a stimulator of activin release from capillary endothelial cells. The effect of simultaneous addition of activin-A and TGF-Pl on BACE cell [3H]methylthymidine uptake was therefore investigated to determine whether these two related molecules can interact to regulate capillary endothelial cell growth (Fig. 7). TGF-Pl caused dose-dependent inhibition of BACE cell [3H]methylthymidine uptake, as previously described (McCarthy and Bicknell, 1992). At each concentration of TGF-Pl, activin-A at both 5 and 50 ng/ml caused increased inhibition of BACE cell DNA synthesis compared to TGF-@1 alone. These data suggest that TGF-@-stimulated release of activin-A may augment the growth inhibitory response to TGF-B.

DISCUSSION
The TGF-(3 superfamily is comprised of several homologous growth regulatory molecules that exert diverse effects on cellular growth and differentiation, many of which are shared by more than one family member. Identification by cDNA cloning of type 2 receptors for activin and TGF-@ as transmembrane serine-threonine kinases has shed light for the first time on signaling mechanisms employed by these pleiotropic molecules (Massague, 1992) and permitted analysis of their expression in diverse cell types. The identification of type 2 activin receptor expression in HUVECs suggested for the first time that the activin/inhibin family of peptide hormones may regulate endothelial cell function. We have therefore investigated the relevance of the activins/inhibins to endothelial cell biology using a representative panel of endothelial cell types.
Although first isolated and characterized in the context of paracrine and endocrine actions on cells of the reproductive system (De Jong, 1988;Vale et al., 1990), the activins and inhibins are now known to be expressed in a variety of nongonadal tissues (Meunier et al., 1988), and are becoming increasingly recognized as regulators of diverse cellular processes. In common with TGF-@, activin-A can either stimulate or inhibit DNA synthesis depending upon the cell type (Kojima and Ogata, 1989;Hedger et al., 1989;Gonzalez-Manchon and Vale, 1989;Centrella et al., 1991;Shao et al., 1992). DNA synthesis in four endothelial cell types. These findings were extended to endothelial cell growth, which in each cell type was decreased by activin-A. Thus, in common with the TGF-89, activin-A is an inhibitor of vascular endothelial cell growth. Inhibin-A had no effect on endothelial cell growth, nor did it antagonize the inhibitory effect of activin-A.
Cross-linking of lZ5I-activin-A revealed three activin-binding proteins on the surface of BAECs, the sizes of which were similar to those observed on other activin-responsive cell lines (Hino et al., 1989;Centrella et al., 1991;Shao et al., 1992). By analogy to the TGF-P receptor system, two of these receptors were interpreted as type 1 (50 kDa), and type 2 (65 kDa) components (Roberts and Sporn, 1990;Massague, 1990Massague, ,1992. The identity of the 145-kDa receptor component remains to be established, although, again by analogy to the TGF-(3 receptor system, this may represent a type 3 activin receptor (Lopez-Casillas et al., 1991;Wang, et al., 1991). Specific binding of lZ5I-activin-A to BAECs was competed fully by activin-A, partially competed by inhibin-A, and unaffected by TGF-(31. In experiments to address whether inhibin-A could antagonize the action of activin-A on endothelial cells, as observed in other cell types (De Jong et al., 1988;Vale et al., 1990), a %fold excess of inhibin-A over activin-A had no effect on activin-A-mediated growth inhibition of CPAE, BACE or HUVEC.3 However, since competition of activin-A binding required a 50-fold excess of inhibin-A (Fig. 3b), it remains possible that higher inhibin concentrations may be effective.
Although a low level of lZ5-I activin-A binding to CPAE could also be detected, no binding could be obtained to either BACE or HUVEC. Thus, as with the TGF-Bs, not all activinresponsive cell types readily display ligand binding . However, expression of both ACTR2 (Mathews and Vale, 1991) and ACTR2B  mRNAs was subsequently demonstrated in all four endothelial cell types used in this study, consistent with responsiveness of each cell type to activin-A, and with type 2 activin receptors being essential signal generating components of the activin receptor system (Massague, 1992;.
Endothelial cells are known to synthesize a number of polypeptide growth factors to which they respond, for example basic FGF (Schweigerer et al., 1987) andTGF-(31 (Antonelli-Orlidge et al., 1989;Sato and Rifkin, 1989;McCarthy and Bicknell, 1992). Investigation of expression of the activins/ inhibins in capillary endothelial cells (BACE) revealed low level expression of activin-pA subunit mRNA in the absence of added factors. Moreover, activin-PA mRNA expression in BACE cells was stimulated specifically in response to TGF-@I or TGF-(32, and suppressed in the presence of FGF. Activin-@, mRNA expression is known to be regulated by various stimuli in other cell types (Takahashi et al., 1990(Takahashi et al., , 1992. Mechanisms responsible for the regulation of activin-@A mRNA expression described here or elsewhere remain to be established. Analysis of activin-pA protein expression in BACE cells confirmed that TGF-Pl stimulates activin-pA protein secretion into BACE conditioned medium. Thus, both mRNA and protein data indicate that TGF-P stimulates activin-pA expression in BACE cells. Since expression of inhibin-a subunit mRNA was not detected in BACE cells, it is assumed that this reflects TGF-P stimulated release of activin. In addition to stimulation of activin release, TGF-@1 also stimulates its own expression in capillary endothelial cells (Mc-Carthy and Bicknell, 1992). Interestingly, simultaneous addition of TGF-B and activin-A to BACE cells caused additive S. McCarthy, unpublished results. inhibition of DNA synthesis, suggesting that TGF-P may promote its own inhibitory effects by induction of activin-A in addition to TGF-/3. Further work is required to formally prove whether activin-A mediates any component of TGF-/3 growth inhibition, and to determine whether this mechanism operates in other cell types. It will also be interesting to determine whether capillary endothelial cells express other TGF-/3 superfamily members, and if so, whether these are similarly regulated by TGF-/3. Recent work has established that BACE cells express mRNAs encoding the bone morphogenetic proteins BMP-2 and BMP-4.3 Future studies will be aimed at understanding how these and other factors interact to regulate vascular endothelial cell growth.