late blastula and gastrula (7-9). These secreted signals, from the organizer, are necessary to form intermediate types of mesoderm, like the somites and kidneys.
Proc. Nati. Acad. Sci. USA
Vol 91, pp. 10243-10246, October 1994
Commentary The transforming growth factor , family and induction of the vertebrate mesoderm: Bone morphogenetic proteins are ventral inducers Richard M. Harland University of California, Department of Molecular and Cell Biology, Division of Biochemistiy and Molecular Biology, 401 Barker Hall, Berkeley, CA 94720
Experimental embryology has shown that the ground state for the mesoderm is ventral, and special signals from the dorsal vegetal region (Nieuwkoop center) and then the dorsal mesoderm (Spemann organizer) cause dorsal development to occur (1). However, now that the molecular mechanisms for mesoderm formation are being elucidated, it is becoming clearer that various different growth factors contribute to both dorsal and ventral pattern (2). The most recent additions to our understanding come from dominantnegative mutations in bone morphogenetic protein (BMP) receptors (3-5). These experiments show that the ventral mesoderm requires an active signal to form the most ventral tissues, such as red blood cells. When mutant BMP receptors are overexpressed and inhibit the normal BMP signal, the tissue develops into muscle, a more dorsal fate (and occasionally the most dorsal fate of notochord). The simplest conclusion is that BMPs act on the ventral mesoderm in normal development to induce the most ventral tissues, blood and mesenchyme. Ideas for Mesoderm Induction and Patterning: Early and Late Signals
In amphibia, the mesoderm is formed at the equator, or marginal zone, of the embryo during cleavage and blastula stages. Mesoderm-inducing cells are in the yolky, vegetal half, while cells that can respond to the inducers are in the cytoplasm-rich animal half (2). Mesoderm patterning to form the dorsalventral series of tissues (notochord, muscle, mesenchyme, and blood) occurs in two distinct steps. Early signals in the blastula induce just two types of mesoderm: the most dorsal sector is fated to become head mesoderm and notochord; the rest is usually considered to be a ground state of mesoderm, which is fated to form mesenchyme and blood (6, 7). A second phase of inductions occur in the late blastula and gastrula (7-9). These secreted signals, from the organizer, are necessary to form intermediate types of mesoderm, like the somites and kidneys (6). The ventral mesoderm is considered to be ground state because tissue-grafting
experiments have shown that the organizer is always dominant, inducing muscle in ventral mesoderm (8, 12). In contrast, ventral grafts have little effect on the fate of dorsal cells (8). Additional evidence that the ventral state is the ground state comes from eggs that have been ventralized by UV-irradiation of vegetal cortex (10). These embryos apparently lack the early Nieuwkoop center signals and, although they make the right amount of mesoderm, it is all ventral (11). Grafting of a Nieuwkoop center induces a proper dorsal-ventral axis, suggesting again that ventral is ground state and dorsal is dominant (1, 10). However, these embryological experiments do not address the details of how ventral tissues are formed. Recent experiments show that even after the initial mesoderm induction, the ground state of ventral is not a cell-autonomous phenomenon; instead, ventral fates require an active signal, mediated by serine/ threonine kinase receptors. Dominant-Negative Mutations in Receptors
Experiments using dominant-negative mutations in receptors have been the most telling as to which signals may be important in ventral cell fates. The first such experiments used a cytoplasmic truncation of the fibroblast growth factor (FGF) receptor, a receptor tyrosine kinase whose activation is mediated by ligand binding. Deletion of the enzymatic domain, and expression ofthis truncation in large excess, prevent activation of the receptor (13). It is presumed that wildtype receptors that bind FGF only interact with the excess mutant receptors, forming nonproductive complexes. Another possibility is that the excess mutant receptor acts as a sponge, binding all available ligand with no possibility of transducing a signal. Whatever the details of the molecular mechanism, it is clear that the inhibition of FGF signaling is quite effective (13, 14) and quite specific for the ras/raf pathway (15). The conclusion of these experiments is that FGF signaling is required in the embryo for formation of the notochord, 10243
somites, and all ventral mesoderm (13, 14). The only survivor is the most anterior-dorsal mesoderm, which in turn is sufficient to organize a reasonable head in the animal. Similar experiments worked very effectively for the activin receptor (16, 17). When disruption of activin signaling is most effective, no mesoderm forms (16, 17). Such experiments led to a view that FGF and activin signaling is responsible for formation of the mesoderm. While formation of the head may be more complicated, requiring activin and something else, like a wnt, noggin, or Vg-1 (2), it seemed that formation of the ventral mesoderm may be fairly simple, requiring just FGF and activin. Ventralizing Inducers
Usually in Xenopus, the experimental paradigm is to find a molecule that does something to embryonic development and worry about whether it is required, or even important, later. This approach has been very successful in identifying candidates for inducers in the embryo. However, it becomes important to satisfy a number of criteria before the molecule is proven to be important. The molecule should be made in the right place at the right time and be made in sufficient amounts to account for the biological effects. Most important, removal of the molecule, or the signaling pathway, should disrupt the biological effect. In addition to FGF and activin, both Xwnt-8 and BMP4 have been suggested as ventralizing signals in the embryo. Xwnt-8 was isolated in a screen for members of the wnt family of signaling molecules (18). Overexpression of Xwnt-8 on the dorsal side during gastrulation partially ventralizes the embryo (19). Xwnt-8 is normally made in the ventral and lateral mesoderm during late blastula and gastrula stages, so it is in the right place to be carrying out a ventralizing function (1820). BMP4 was first isolated from bone as a protein that could induce ectopic bone formation but was also isolated from early frog embryos in a screen for transforming growth factor (3 (TGF-,S) family members (21, 22). When injected
10244
Proc. Natl. Acad. Sci. USA 91 (1994)
Commentary: Harland
as mRNA it has even more extreme effects than Xwnt-8, resulting in complete loss of dorsal structures and excess ventral development (22, 23). The mRNA is present at low levels maternally, and higher amounts are made after transcription starts in the mid-blastula (22). Until recently, the evidence suggested that the BMP-4 mRNA was not localized, but now it is clear that at least in the gastrula BMP-4 has a localization entirely consistent with a ventralizing function. Like Xwnt-8, BMP-4 mRNA is expressed in ventral and lateral mesoderm of the midgastrula embryo (24). A good test of its function by dominant-negative mutation had to await the cloning of the BMP receptor.
default on the ventral side. The embryological experiments still show that the physiological amount of BMP-4 signal cannot override the fate of neighboring organizer tissue. However, it is intriguing that the most dorsal region is devoid of BMP-4 expression in the gastrula. Perhaps it is not only the production of dorsal determinants and dorsal signals in the organizer that allow it to develop as notochord and head mesoderm but also the absence of an antagonistic ventral signal (24). How Do the New Results on BMP-4 Affect Interpretation of Animal Cap Experiments?
The other intriguing aspect of the BMP-4 Dominant-Negative Experiments with the mRNA localization is the expression in the animal cap (24). It is quite reasonable BMP Receptor to think that this expression could acThe identification of BMP receptors and count for some puzzling properties of the some of the potential complications of animal cap. Animal caps are used widely interpreting these experiments are dis- in assays for induction, but while animal cussed later; the main point now is that cap cells are often thought of as an unputative serine/threonine kinase recep- biased assay system, it has become clear tors with high affinity for BMP-2 and -4 in recent years that they have their own were cloned from various species. Graff predispositions to respond to inducers. A and Melton, in collaboration with scien- case in point is the different response of tists from Genetics Institute, cloned a dorsal and ventral animal caps to activin Xenopus BMP receptor (3), while Ueno's (25-27). While it has been suggested that group (4), also working with Genetics this may be due to the presence of a Institute, cloned a mouse BMP receptor. wnt-like signal in the dorsal half of the Using the approach of deleting the kinase cap (28, 29), it could also be due to the domain and expressing the truncated re- presence of BMP-4 in the ventral side. When animal cap cells are dispersed, ceptor in excess, both groups reached similar conclusions. First, the mutant they are more sensitive to activin than was effective in preventing the action of aggregated cells (30), and whereas veninjected BMP-4 mRNA (3, 4), though it tral animal caps do not make notochord did not prevent induction by activin (3). in response to activin (25, 27), caps Injection of the truncated receptor into treated as dissociated cells make notothe ventral side frequently resulted in chord efficiently (30, 31). If BMP-4 is a twinned embryos, though the secondary powerful antagonist of notochord formaaxes induced on the ventral side were not tion, dissociation of the animal cap cells complete. Expression of the truncated could dilute out the endogenous BMP-4, receptor in explants of the prospective allowing more dorsal fates to be induced ventral mesoderm changed their fate by activin. This supports the idea that from blood and mesenchyme to dorsal absence of BMP-4 mRNA from the nomesoderm, differentiating a high amount tochord field during gastrulation could be of muscle and often notochord (3, 5). important in allowing development of the Both early and late molecular markers notochord. mirrored the phenotypic effects (3-5). Two other signals mediated by the anEven organizer-specific markers like imal cap are worthy of note. First, animal goosecoid (3) were induced by the trun- cap, and even injected animal cap RNA, cated receptor. Crucially, the effects of potentiate the formation of red blood the truncated receptor could be over- cells in ventral explants (32). Furthercome by injecting excess wild-type re- more, the animal cap has been shown to antagonize the formation of muscle (33). ceptor (3). The observations make clear the re- Now, it has been shown that a truncated quirement for an active BMP-like signal BMP receptor expressed only in animal for production of the ventral mesoderm. cap cells is sufficient to dorsalize ventral At first sight this may appear to contra- marginal zones (5). Thus, it appears that dict the embryological experiments that BMP-4 expressed in the animal cap may show the ventral mesoderm as the ground affect the tissues of the neighboring mestate for mesoderm, but in fact there is no soderm, both suppressing dorsal tissues real need to alter our views of the em- and enhancing formation of ventral tisbryology. Since BMP-4 is expressed in sues. The animal cap normally covers the the ground state tissue (24), blood and ventral mesoderm after gastrulation is mesenchyme will still be produced as a completed, so once again the signal is in
the right place, and inhibition of the signaling pathway results in changed fates. Antagonistic Dorsal and Ventral Sigals It has always been an attractive idea that antagonistic signals result in a final pattern. The experiments with the truncated BMP receptor make clear that a ventral signal is important in production of the full range of tissues. On the dorsal side, there are no experiments with the same degree of rigor. However, wnts (20, 25), noggin (34), and the TGF-(3 Vg-1 (35) have experimental support as candidates for the early dorsal signals. Among these, only noggin has also been shown to have dorsalizing properties in the gastrula (36). The reciprocal expression of noggin and BMP-4 in the gastrula makes these attractive candidates for the antagonistic secreted signals that compete to determine dorsal, lateral, or ventral fates of the mesoderm. Although the location of noggin and BMP-4 transcripts suggests a patterning role in the gastrula, unfortunately it is not clear from the experiments with the truncated, dominant-negative receptor when BMP-4 signaling is important. By the nature of the experiment, the mutant mRNA must be injected early on, and so the mutant protein that is translated from it could be acting during the early phase of signaling, when mesoderm is first set up into organizer and nonorganizer territories. The mutant could instead be acting predominantly during gastrulation. Even with the useful tools of dominant-negative receptors available, it is very difficult to determine when and where signals are normally active. Growth factors are notoriously difficult to localize with antibodies, and even if an antibody gives a reaction, one must then determine whether the growth factor is processed and presented. Ideally there would be a probe for ligand bound to receptor, but it is hard to see how one could detect such a transient complex with current technology.
Cloning of BMP Receptors There are some aspects of the properties of the BMP receptor, and the properties of the dominant-negative truncated receptor, that remain to be clarified. In particular, how specific are these dominant-negative mutants? Could they be interfering in pathways that are normally activated by other TGF-,B family members? These questions that have not been answered can be posed in the context of what is currently known about TGF-,B receptors. A large number of putative receptors for the TGF-.8 family have now been cloned. The first were the activin and TGF-P receptors, which were identified
Commentary: Harland by expression cloning (37, 38). These cDNA clones were selected because they confer extremely high ligand binding efficiency on transfected cells. The sequences showed surprising similarity to daf-1 and subsequently daf-4, genes that were identified because they disrupt a program of dormancy in the nematode Caenorhabditis elegans (39, 40). The daf-4 sequence was strikingly similar to the activin and TGF-3 receptors and was tested for affinity to known ligands. Interestingly, this nematode protein binds mammalian BMP-4 with fairly high affinity (40). Understanding how these receptors work has relied on biochemistry and somatic cell genetic experiments with TGF-,B and activin (see discussion and references in ref. 41). TGF-,3 can be cross-linked to three cell surface proteins. The largest (type Ill) has a low affinity and is important in ligand presentation. The high-affinity form that was isolated by expression cloning is the type II and can bind ligand by itself. The type I has no demonstrable affinity for TGF-,3 in the absence of the type II but is required for signaling. Ligand induces dimerization of the type I and type II receptors; the type II phosphorylates a conserved domain on the type I and signal is transduced to downstream proteins. From these experiments it was expected that a general paradigm would emerge of a high-affinity type II receptor and a low-affinity type I receptor that is important for signaling. Numerous receptors have now been cloned by lowstringency hybridization to other receptors or by polymerase chain reaction (PCR) with primers based on the sequences of TGF-l3 and activin receptors as well as daf-1 and -4. All of the receptors that have been isolated by PCR, including those used in the dominantnegative experiments, have sequence motifs that put them in the type I class (see, for example, refs. 42 and 43). In contrast, the degenerate primers would probably not pick up type II receptors, which appear to be more divergent in sequence (44). In view of the observation that only type II activin and TGF-,B receptors show high affinity for their cognate ligand, it was surprising that some of the type I receptors have shown high affinity for BMP-2 and -4. Though the picture has yet to be completely clarified, it is currently clear that there are some BMP type I receptors with low affinity and others with high affinity (45-47). The work on the role of BMP signaling in formation of ventral mesoderm has used high-affinity BMP type I receptors, which are likely to be homologues of ALK-3 (43, 47). It still seems likely that there are BMP type II receptors. This contention is largely
Proc. Natl. Acad. Sci. USA 91 (1994)
based on the daf-4 example, which resembles type II receptors in sequence (40). daf-4 can also be cross-linked into a complex with both low- and high-affinity type I receptors (45, 46, 47). In turn, BMP-4 can be cross-linked to mammalian cell surface proteins with the sizes expected of both type I and type II receptors (48). However, in the absence of cloned vertebrate BMP type II receptors, the details of BMP binding and signaling are still not clear. It is particularly alarming that type II receptors can interact with many type I receptors (see ref. 41 for references). Although this is not likely at physiological concentrations of the receptor, the dominant-negatives are expressed at huge levels, so cross-talk of the mutant with numerous receptors is a concern. A further complication is highlighted by recent experiments with the FGF receptor (15, 49), where signal transduction by the FGF pathway is required to permit activin signaling to be effective. So, although blockage of FGF signal results in blockage of some activin signals, it does not mean that the dominant-negative is promiscuous. The good news for the TGF-,S family receptors is that the dominant-negatives for activin and BMP4 have very different effects; each dominant-negative has been shown to block only the activity of the cognate ligand, so they are not acting promiscuously on the same downstream components. However, it appears that the dominant-negative activin receptor is blocking some related signals. Vg-1 signaling is blocked by the dominantnegative activin receptor. While this could be a parallel pathway, like the case of FGF described above, experiments with follistatin, an activin-neutralizing protein, suggest an alternate interpretation. While follistatin can block the experimental effects of activin (50, 51), it does not block mesoderm formation in the embryo, suggesting that something other than activin is inducing mesoderm (51). If this interpretation holds up, it is a necessary conclusion that the dominantnegative activin receptor must be acting somewhat promiscuously. In contrast, experiments using a point mutant of activin that acts as a receptor antagonist suggest that activin is required directly
(52).
Until it is clear from more direct biochemical experiments how the truncated receptors are acting, it is worth being cautious. One peculiar aspect of the truncated BMP receptor is the amount of wild-type receptor that is needed to rescue the effect of the dominant-negative (3). This kind of experiment is absolutely necessary to eliminate the possibility that the effects of the dominant-negative are due to some unanticipated and nonspecific effect. In the case of the activin
10245
receptor, a small minority of wild-type receptor overcomes the effects of the mutant (16); in contrast, an excess of wild-type BMP receptor is necessary to rescue the effects of the mutant (3). This highlights the problem that we do not really know how these dominant-negatives work and they may not all work in exactly the same way. It is clearly premature to expect a full understanding of how dominant-negatives work, since the details of the normal pathway are still undergoing biochemical scrutiny. The important aspect to focus on is the striking new insights that have resulted from this very powerful approach in an animal that is not suited to deletion of genetic information. The dominant-negative approach has been informative and has repeatedly given unexpected results; FGF is not just a ventral inducer but is required for most ofthe mesoderm. Activin is not just required for dorsal signaling but is required for production of all mesoderm, and inhibition of its signaling causes neural tissue to form. Now we find that BMP-4 is required for production of the most ventral types of mesoderm and is probably important in defining the territories that can form the lateral mesoderm and somites. I thank Joan Massague, Jonathan Graff, Ali Hemmati Brivanlou, and Eddy DeRobertis for invaluable discussions on TGF-,Bs and their receptors. 1. Gerhart, J., Doniach, T. & Stewart, R.
(1991) Gastrulation ed. Keller, R. E. (Plenum, NY), pp. 57-77. 2. Slack, J. M. W. (1994) Curr. Biol. 4, 116-126. 3. Graff, J. M., Thies, R. S., Song, J. J., Celeste, A. J. & Melton, D. A. (1994) Cell, in press. 4. Suzuki, A., Thies, R. S., Yamaji, N., Song, J. J., Wozney, J. M., Murakami, K. & Ueno, N. (1994) Proc. NatI. Acad. Sci. USA, in press. 5. Madno, M., Ong, R. C., Suzuki, A., Ueno, N. & Kung, H.-f. (1994) Proc. Natl. Acad. Sci. USA, in press. 6. Stewart, R. M. & Gerhart, J. C. (1990) Development (Cambridge, U.K.) 109, 363-372. 7. Dale, L. & Slack, J. M. (1987) Development (Cambridge, U.K.) 100, 279-95. 8. Smith, J. C. & Slack, J. M. (1983) J. Embryol. Exp. Morphol. 78, 299-317. 9. Lettice, L. A. & Slack, J. M. W. (1993) Development (Cambridge, U.K.) 117, 263-271. 10. Gerhart, J., Danilchik, M., Doniach, T., Roberts, S., Rowning, B. & Stewart, R. (1989) Development (Cambridge, U.K.) 107, 37-51. 11. Cooke, J. & Smith, J. C. (1987) Development (Cambridge, U.K.) 99, 197-210. 12. Gimlich, R. L. & Cooke, J. (1983) Nature (London) 306, 471-473. 13. Amaya, E., Musci, T. J. & Kirschner, M. W. (1991) Cell 66, 257-270. 14. Amaya, E., Stein, P. A., Musci, T. J. &
10246
15. 16. 17.
18.
19. 20. 21.
22. 23.
24. 25. 26. 27.
28.
Commentary: Harland
Kirschner, M. W. (1993) Development (Cambridge, U.K.) 118, 477-487. LaBonne, C. & Whitman, M. (1994) Development (Cambridge, U.K.) 120, 463472. Hemmati-Brivanlou, A. & Melton, D. A. (1992) Nature (London) 359, 609-614. Hemmati-Brivanlou, A. & Melton, D. A. (1994) Cell 77, 273-281. Christian, J. L., McMahon, J. A., McMahon, A. P. & Moon, R. T. (1991) Development (Cambridge, U.K.) 111, 1045-1055. Christian, J. L. & Moon, R. T. (1993) Genes Dev. 7, 13-28. Smith, W. C. & Harland, R. M. (1991) Cell 67, 753-765. Koster, M., Plessow, S., Clement, J. H., Lorenz, A., Tiedemann, H. & Knochel, W. (1991) Mech. Dev. 33, 191-199. Dale, L., Howes, G., Price, B. M. & Smith, J. C. (1992) Development (Cambridge, U.K.) 115, 573-585. Jones, C. M., Lyons, K. M., Lapan, P. M., Wright, C. V. & Hogan, B. L. (1992) Development (Cambridge, U.K.) 115, 639-647. Fainsod, A., Steinbeisser, H. & DeRobertis, E. M. (1994) EMBO J., in press. Sokol, S. & Melton, D. A. (1991) Nature (London) 351, 409-11. Ruiz i Altaba, A. & Jessell, T. (1991) Genes Dev. 5, 175-187. Bolce, M. E., Hemmati-Brivanlou, A., Kushner, P. D. & Harland, R. M. (1992) Development (Cambridge, U.K.) 115, 681-688. Christian, J. L., Olson, D. J. & Moon, R. T. (1992) EMBO J. 11, 33-41.
Proc. Natl. Acad. Sci. USA 91 (1994) 29. Sokol, S. Y. & Melton, D. A. (1992) Dev. Biol. 154, 348-355. 30. Green, J. B. A., Smith, J. C. & Gerhart, J. C. (1994) Development (Cambridge, U.K.) 120, 2271-2278. 31. Green, J. B., New, H. V. & Smith, J. C. (1992) Cell 71, 731-739. 32. Madno, M., Ong, R. C., Xue, Y., Nishimatsu, S., Ueno, N. & Kung, H. F. (1994) Dev. Biol. 161, 522-9. 33. Kato, K. & Gurdon, J. B. (1994) Dev. Biol. 163, 222-229. 34. Smith, W. C. & Harland, R. M. (1992) Cell 70, 829-840. 35. Thomsen, G. H. & Melton, D. A. (1993) Cell 74, 433-441. 36. Smith, W. C., Knecht, A. K., Wu, M. & Harland, R. M. (1993) Nature (London) 361, 547-549. 37. Mathews, L. S. & Vale, W. W. (1991) Cell 65, 973-982. 38. Lin, H. Y., Wang, X. F., Ng-Eaton, E., Weinberg, R. A. & Lodish, H. F. (1992) Cell 68, 775-785. 39. Georgi, L. L., Albert, P. S. & Riddle, D. L. (1990) Cell 61, 635-645. 40. Estevez, M., Attisano, L., Wrana, J. L., Albert, P. S., Massague, J. & Riddle, D. L. (1993) Nature (London) 365, 644649. 41. Wrana, J. L., Attisano, L., Wieser, R., Ventura, F. & Massagu6, J. (1994) Nature (London) 370, 341-347. 42. He, W. W., Gustafson, M. L., Hirobe, S. & Donahoe, P. K. (1993) Dev. Dyn. 196, 133-142. 43. ten Dike, P., Ichijo, H., Franzen, P., Schulz, P., Saras, J., Toyoshima, H.,
Heldin, C. H. & Miyazono, K. (1993) Oncogene 8, 2879-2887. 44. Baarends, W. M., van Helmond, M. J., Post, M., van der Schoot, P. J. C. M., Hoogerbrugge, J. W., de Winter, J. P., Uilenbroek, J. T., Karels, B., Wilming, L. G., Me~ers, J. H. C., Themmen, J. H. C. & Grootegoed, J. A. (1994) Development (Cambridge, U.K.) 120, 189197. 45. Brummel, T. J., Twombly, V., Marquds, G., Wrana, J. L., Newfeld, S. J., Attisano, L., Massagud, J., O'Connor, M. B. & Gelbart, W. M. (1994) Cell 78, 251-261. 46. Penton, A., Chen, Y., Staehling-Hampton, K., Wrana, J. L., Attisano, L., Szidonya, J., Cassill, J. A., Massagud, J. & Hoffmann, F. M. (1994) Cell 78, 239250. 47. ten Dijke, P., Yamashita, H., Sampath, T. K., Reddi, A. H., Estevez, M., Riddle, D. L., Ichijo, H., Heldin, C. H. & Miyazono, K. (1994) J. Biol. Chem. 269, 16985-16988. 48. Koenig, B. B., Cook, J. S., Wolsing, D. H., Ting, J., Tiesman, J. P., Correa, P. E., Olson, C. A., Pecquet, A. L., Ventura, F., Grant, R. A., Chen, G.-X., Wrana, J. L., Massagu6, J. & Rosenbaum, J. S. (1994) Mol. Cell. Biol. 14, 5%1-5974. 49. Cornell, R. A. & Kimelman, D. (1994) Development 120, 453-462. 50. Hemmati-Brivanlou, A., Kelly, 0. G. & Melton, D. A..(1994) Cell 77, 283-295. 51. Schulte-Merker, S., Smith, J. C. & Dale, L. (1994) EMBO J. 13, 3533-3541. 52. Wittbrodt, J. & Rosa, F. (1994) Genes Dev. 8, 1448-1462.
