ventral ectoderm of Xenopus forms neural tissue, including hindbrain, in response to activin

Development of anterior-posterior polarity in the nervous system of Xenopus laevis requires inductive interactions between dorsal mesoderm and presumptive neurectoderm. Although the molecules that mediate neural induction remain elusive, peptide growth factors have been implicated in the formation of mesoderm. Activin A is of particular interest because it can cause formation of both mesodermal and neural tissues in explants of embryos (Smith et al., 1990; Sokol et al., 1990; Thomsen et al., 1990). However, several lines of evidence demonstrate that activin is not a direct neural inducer. Green and Smith (1990) have shown that activin does not induce dispersed cells to become neural but can induce dispersed cells to become neural inducers. Moreover, dorsal mesoderm can induce neural tissue in animal cap explants at a later stage than can activin, suggesting that activin induces dorsal mesoderm which in turn induces neural tissue (Kintner and Dodd, 1991; Sharpe and Gurdon, 1990; Green et al., 1990). The finding that activin A is able to induce a complete anterior-posterior axis in pieces of animal pole ectoderm from the blastula has suggested that it can account for the full series of dorsoanterior struc tures (Sokol et al., 1990). However, further studies indicated that at least part of activin’s patterning activity must be due to a predisposition of the dorsal blastula ectoderm to become neural. While the dorsal halves of animal caps are able to form a complete axis in response to activin, ventral halves cannot and by histological criteria do not make notochord or neural tissue (Sokol and Melton, 1991). Furthermore, the generation of a complete pattern in response to activin requires a relatively large piece of animal pole tissue; smaller pieces do not differentiate anterior structures (Rebagliati and Dawid, personal communication). These results have raised the possibility that activin is a “permissive” inducer allowing the realization of fates that were prepatterned in the animal cap rather than an “instructive” inducer that prescribes alternate fates for cells (see Gurdon, 1987).

To determine what tissues are inducible by activin alone and what tissues require the dorsal prepatterning signal, we have used molecular markers of neural tissue and anterior-posterior position in the nervous system: N-CAM, a general neural marker (Kintner and Melton, 1987), En-2, (Hemmati-Brivanlou et al., 1991) a marker of the midbrain-hindbrain junction, and Krox-20, a hindbrain-specific marker (Wilkinson et al., 1989; Papalopulu et al., 1991). Whole-mount in situ hybridiz ation (Hemmati-Brivanlou et al., 1990a; Harland, 1991) was used to compare the expression of these markers in activin-treated animal caps from control embryos and ventralized embryos (derived from eggs which were irradiated with UV light, reviewed in Gerhart et al., 1989). Whole-mount in situ hybridization offers advantages over total RNA assays such as northern blotting and RNAase protection. First, the technique allows us to distinguish between low levels of expression in almost all explants and high levels of expression in a few explants. Second, we can determine if the gene is expressed throughout the explant or in a particular pattern. Finally, the method is sufficiently sensitive to detect messages, such as En-2 and Krox-20, that are present at low levels in the embryo.

Our results show that, contrary to previous conclusions (Sokol and Melton, 1991), neural tissue is abundant and that the hindbrain marker Krox-20 is expressed at high frequency in ventralized animal caps treated with activin. Thus, activin must induce dorsal tissues which in turn can induce neural tissue, including hindbrain. We do confirm the earlier findings that ventralized animal caps do not make notochord ef ficiently. Although anterior tissues (that express En-2) can be induced in ventralized animal caps in response to dorsal mesoderm (Hemmati-Brivanlou et al., 1990b), we show here that ventralized animal caps treated with activin develop only a limited amount of the anterior posterior axis, and cannot make structures anterior of the hindbrain. Thus formation of the complete neural axis, and of the notochord, is dependent on signals other than activin in the developing embryo.

Embryos were fertilized and cultured as described in Condie and Harland (1987) and staged according to Nieuwkoop and Faber (1967). For UV treatment, embryos were dejellied in 2% cysteine in 1/3 x modified Ringers at 10 minutes after fertilization. Between 25 and 30 minutes after fertilization, embryos were placed in quartz-bottomed Petri dishes and irradiated in an upside-down Stratalinker (Stratagene) at an energy reading of 1000 mJoules. Immediately, embryos were placed in agarose-coated Petri dishes in 1/3 × modified Ringers plus either 10 μg/ml oxytetracycline or 50 μg/ml gentamycin. To assess the effectiveness of the UV treatment, embryos were scored for their dorsal-anterior index (DAI) (Kao and Elinson, 1988) around stage 25. The average DAI in these experiments was 0.4.

Animal caps were explanted at stage 9 in 1/3 × modified ringers. The explants were immediately rinsed in a large volume of LCMR (Stewart and Gerhart, 1990) and then placed in 50 μI of LCMR + 0.5% BSA with antibiotics plus a defined amount of activin in a microtiter dish well coated with agarose. After overnight incubation at 15°C, the explants were rinsed in fresh LCMR + antibiotics (without activin) and transferred to LCMR in a new agarose-coated Petri dish. Highly purified porcine Activin A was generously provided by Drs W. Vale and J. Vaughan (Salk Institute, LaJolla).

Explants and control intact embryos were homogenized in 0.1 M NaCl, 20 mM Tris, pH 7.8, 10 mM EDTA, 1% SDS and digested in 0.2 mg/ml proteinase K at 37°C for 30 minutes. The samples were then phenol extracted and precipitated with 0.3 M sodium acetate and ethanol. After resuspension in water, the samples were precipitated with 2.5 M LiCl by incubation on ice for at least an hour. An equivalent of 1.5 embryos or 5 animal caps was fractionated on a denaturing agarose gel, and the RNA was subsequently transferred to a gene screen membrane by capillary blotting in 20 × SSC overnight. The membrane was UV-irradiated in a Stratalinker and baked 1–2 hours at 80°C under vacuum.

Gel-isolated fragments of N-CAM (Kintner and Melton, 1987), cardiac actin (Dworkin-Rast! et al., 1986), E13 (Sive et al., 1989; Jonas et al., 1985), and EFlcx (Krieg et al., 1988) were used to prepare probes by random primed synthesis. The labelled DNAs were hybridized to the membrane overnight at 42°C in 50% formamide, 5 × SSC, 1 ×Denhardt’s, 25 mM sodium phosphate, pH 6.5, and 0.5 mg/ml torula RNA. The filters were washed for 15 minutes in 2 × SSC; 1% SDS at room temperature and then two times 30 min at 65°C in 50 mM Tris, pH 8, 2 mM EDTA, 0.5% sodium pyrophosphate, 1 × Denhardt’s, 1% SDS and 0.05% n-lauroyl sarcosine. Finally the filters were exposed to X-ray film for 1 to 168 hours.

In situ hybridization was performed as described in Harland (1991). The Krox-20 probe was generated from a clone of the Xenopus Krox-20 homologue kindly provided by D.G. Wilkinson and L.C. Bradley. Hybridizing with the Krox-20 probe gives a very strong signal in rhombomeres 3 and 5 of the hindbrain as well as some neural crest staining beyond the limits of the hindbrain in rhombomere 5. There is a low level of epidermal staining that occurs with use of this probe; however, the signal from the CNS is much more intense than the epidermal signal and consequently easily distinguished. En-2 probe was made from a plasmid containing a 1.5 kb cDNA clone of the Xenopus En-2 (Hemmati-Brivanlou et al., 1991). In situ hybridization with all markers gives an intense blue signal that is very distinct from background and dust.

Tor 70 (Kushner, 1984) had previously been found to stain zebrafish notochord (P.D. Kushner and B. Mendelson, unpublished information). In Xenopus the antibody is specific Ct to differentiating notochord from the late neurula stage to t I about stage 28. All the vacuolating notochord cells which are detectable by Nomarski microscopy of cleared specimens stain with the antibody. The notochord staining declines, as staining appears in the otic vesicle, cranial ganglia and a fewproctodeal and stomodeal cells. At stage 40, staining is absent from the notochord, but prominent in lateral line, procto deum, stomodeum and nasal pits. We have not investigated the nature of the antigen in Xenopus further.

Antibody staining was carried out using the whole-mount immunocytochemistry protocol described in Hemmati Brivanlou and Harland (1989). Secondary goat anti-mouse F b fragments coupled to HRP were obtained from Jackson

L:bs. The samples stained by this procedure had previously b en assayed by in situ hybridization, fixed, cleared and subsequently stored in methanol at − 20°C.

Although it has been found that activin does not induce eyes or notochords in animal cap explants from UV embryos (Sokol and Melton, 1991), we wanted to determine by rigorous RNA assays if and to what extent neural tissue can be formed in isolated ectoderm in response to activin. Because the size of the animal cap used affects the range of structures generated by activin, we wanted to assay relatively large pieces of ectoderm. Thus, instead of dividing animal caps into dorsal and ventral pieces, we cut caps from control and UV-ventralized embryos. This comparison is valid since animal caps from ventralized embryos are equivalent to ventral halves of animal caps and by histology they respond to activin in the same way (Sokol and Melton, 1991).

To examine the expression of the tissue-specific genes N-CAM, muscle actin and E13, a marker of epidermal cytokeratin, animal caps from UV embryos (UV animal caps) and non-UV-treated embryos (control animal caps) were explanted at stage 9 and incubated in a range of activin concentrations from 0 to 100 pM. At control stage 20, at least 10 cultured animal caps were pooled for the isolation of total RNA, and an equivalent of 5 animal caps per sample was analyzed by northern blotting. Fig. 1 shows that N-CAM, a general neural marker, can be induced in UV animal caps although no N-CAM is present in intact UV embryos also homogen ized at stage 20. 10 pM activin will induce N-CAM expression in both UV and control caps. At these activin concentrations, both types of explants respond to the factor in a dose-dependent manner only over a limited range, from 10 pM to 50 pM. Expression appears to be fully induced at 50 pM activin in the explants and no change in the level of message is seen at higher concentrations (the lower N-CAM expression seen in control caps at 100 pM is not reproducible).

Hybridization with a cardiac actin probe shows that mesodermal derivatives, in this case muscle, are also being induced by activin (Fig. 1). No muscle actin is found in either control or UV caps incubated in buffer alone or in 1 pM activin; this result indicates that no contaminating marginal zone is taken during explan tation (Gurdon et al., 1984). Furthermore, we find that no muscle actin is detected in whole UV embryos, thus confirming that the UV treatment was effective. In both UV and control animal caps, the amount of actin RNA induced is similar at all activin concentrations above 10 pM. In agreement with the results of Sokol and Melton (1991), both the control and the UV animal caps are competent to make muscle.

In the embryo, ectodermal cells follow one of two mutually exclusive paths to become either epidermal or neural (Jamrich et al., 1987). We used an epidermal-specific marker E13 to ask if the explants make exclusively mesodermal and neural tissue or if some of the cells have remained epidermal. Fig. 1 shows that both UV and control explants incubated in buffer alone express E13. Incubation in 50 pM to 100 pM activin diminishes the expression of E13 in the explants although more so in the control than in the UV animal caps. This decrease may be due to epidermal cells being transformed into neural type cells.

Finally, EFlcx is used as a control for the amount of RNA loaded. In Fig. 1, signal from the EFlcx probe and thus RNA is present in every lane although differences in the amount of RNA loaded are apparent. Five explant equivalents of RNA were assayed in each lane, but it is likely that the recovery of RNA will vary from explant to explant.

Thus, by northern analysis we have shown that activin induces both mesodermal and neural tissues and reduces the level of an epidermal marker in UV and control explants. The induction of both muscle and neural markers is similar in both types of explants; thus, activin is able to induce neural tissue in ventralized as well as in control caps.

Since we have found that UV caps do make neural tissue in response to activin, we wanted to examine the expression of regional neural markers in the isolates to determine the anterior extent to which regional neural induction occurs in the ventralized ectoderm. In order to look at the pattern of expression of genes specific to certain regions of the central nervous system, we assayed animal cap explants by in situ hybridization. Northern analysis showed that the expression of mesodennal and neural markers in both the UV and control caps reached a maximum at or below 50 pM activin; therefore, animal caps were treated with either a control solution or 50 pM activin solution and allowed to develop until intact sibling controls reached stage 20. At this time they were fixed and subjected to hybridiz ation with N-CAM, En-2, which is expressed at the midbrain-hindbrain junction, or Krox-20, which is expressed in rhombomeres 3 and 5 of the hindbrain.

N-CAM in situ hybridization on control embryos fixed at stage 20 confirms the results of Kintner and Melton (1987) that the RNA is present in and is specific to the central nervous system at this stage (Fig. 2a). Staining both UV and non-UV animal caps reveals that N-CAM is expressed in the two types of explants in response to activin. Neither set of animal caps incubated in control media exhibited detectable N-CAM expression (data not shown). In both samples treated with 50 pM activin, the RNA appears to be predominantly in elongated explants, and the appearance of the staining does not differ significantly between the two types of explants (Fig. 2d and g). Moreover, more than two-thirds of both UV and control explants expressed the marker in response to activin (Table 1). Since none of the intact UV embryos assayed showed any N-CAM expression and the average dorsal-anterior index (DAI) of the UV embryos used in the assay was 0.4, it was not possible that the N-CAM staining could be due to ineffective UV treatment.

Animal caps were next stained for the hindbrain marker Krox-20. Fig. 2b presents in situ hybridization of Krox-20 to intact control embryos fixed at stage 20. As in the mouse (Wilkinson et al., 1989) the marker is expressed strongly in two bands in the hindbrain of the embryo corresponding to rhombomeres 3 and 5 (Papa lopulu et al., 1991); moreover, the posterior stain extends more weakly beyond the limit of the neural plate comprising most likely a neural crest component of the expression (L.C. Bradley, A. Snape, S. Bhatt, D.G. Wilkinson, unpublished). No specific staining is seen in either UV or control animal caps incubated in buffer alone (data not shown), but Krox-20 RNA is detected in both types of animal caps after incubation in 50 pM activin (Fig. 2e and h). In both types of explants a high percentage of caps stained for the marker, 92% of control and 71% of UV (Table 1), and the intensity of staining appeared to be quite similar. The intensity of stain is comparable to that in the hindbrain rather than the neural crest of whole embryos, so we conclude that hindbrain cells are present in the explants. However, a small number of the control animal caps had two rings of Krox-20 staining while all of the UV caps had only one, and many UV caps had staining at the very tip of the explant while staining of the control caps was more to the middle. This result suggests a possible difference in the anterior extent of induction in UV and control caps: the anterior limit of induction in UV caps may be between rhombomeres 3 and 5, whereas it can extend beyond rhombomere 3 in control caps.

Finally we assayed the animal caps for expression of En-2 RNA. Fig. 2c is a whole control embryo fixed at stage 20 and stained for En-2 RNA which, like the protein, is detected in a tight band of expression at the midbrain-hindbrain junction (Hemmati-Brivanlou and Harland, 1989). At later stages En-2 is also expressed in the mandibular arch (Hemmati-Brivanlou et al., 1991) but at lower intensity than in the brain. Because of the age of the explants and the strong pattern of staining found in the brain, we conclude that the En-2 expression seen in the explants is neural in nature. As expected when treated with buffer alone, explants from both UV and control embryos were negative for En-2 RNA (data not shown). Fewer control caps incubated in 50 pM activin stained for En-2 than for either N CAM or Krox-20, and a nominal number of the UV caps treated with activin were positive for En-2 (11.1% as opposed to 46% of the control caps; Table 1). The staining in those few UV caps that did express En-2 was in very small patches compared with the staining in control caps which appears in bands or rings around the explant (Fig. 2f and i). Because the frequency of expression was so low, the staining we do see may result from caps explanted from embryos with a DAI of greater than 0 or 1.

Fig. 3 provides a graphical representation of the numbers of explants expressing the various neural markers as assayed by in situ hybridization. Explants were considered positive for a particular marker if they exhibited a significant and coherent patch of stained cells. While a high proportion of both UV and control animal caps stain for N-CAM and Krox-20, fewer of the control explants and only one-tenth of the UV explants express the more anterior marker En-2. There are_ t o possible explanations for the low freq ency of act!vm inducible En-2 expression in the UV ammal caps: either the explants are not able to express the midbrain hindbrain marker or higher concentrations of the factor are required for induction. To test this_ rossibility, explants were incubated in a range of achvm concen trations from 1 pM to 200 pM. In Fig. 4, the percentage of UV as well as control explants expressing En-2 is plotted against activin concentration. The proportion of explants positive for En-2 increases only between 1 pM and 10 pM activin and remains essentially the same up to 200 pM; however, the percentage of En-2-positive explants never goes above 45%, whereas N-CAM and Krox-20 are both seen in 70% and 93%, respectively, of explants treated with 50 pM activin (Table 1). Thus, he control explants are less able to make the more antenor En-2 than the more posterior Krox-20 marker, and the UV animal caps never make En-2 above a basal level even at 200 pM activin. Since no UV treatment w s 100% effective, the few UV animal caps that did express En-2 may be from those less severely ventra lized embryos.

It has been shown that the most dorsal mesoderm, namely notochord, is most effective at inducing neural tissue (Jones and Woodland, 1989; Hemmati-Brivanlou et al., 1990b); Sokol and Melton (1991) find that by histological criteria no notochord is made in ventral ectoderm explants treated with activin. Thus, we wanted to test if those explants positive for general and regional neural markers did contain notochord. Us ng monoclonal antiserum, Tor 70 (Kushner, 1984), which specifically stains notochord, we countersta n d explants that had been subjected to m situ hybnd1zation. Fig. 2j shows a control stage 20 embryo previously assayed for En-2 RNA that has be n stamed with, Tor 70 supernatant. The notochord 1s clearly visible, although at this stage the posterior-most section of the notochord has yet to express the antigen. Expression of the Tor 70 antigen in activin-treated explants appears as very distinct notochord-like structures (see Fig. 2k and 1). Among control explants exposed to 50 pM activin, 34% expressed the Tor 70 antigen while only 5.4% of the UV animal caps were positive for the marker (Table 1).

Furthermore, expression of the neural and no ochord markers do not correlate. Among the control isolates previously stained for En-2 and Krox-20, the explants which expressed the regional marker often lacked notochords and, conversely, those that failed to stain for the regional markers sometimes had notochords (Fig. 2k and 1). Moreover, a high percentage of the UV animal caps treated with activin express N-CAM and Krox-20 while none of them have notochords (data not shown). Thus, activin must induce some non noto chordal mesoderm which can direct the format10n of neural structures.

Using whole-mount in situ hybridization assays, we demonstrate that activin can induce the expression of both N-CAM and Krox-20 in explanted ventralized ectoderm (UV animal caps) as effectively as it an in control ectoderm; however it is unable to ehc1t the expression of the more anterior En-2 marker efficiently in ventralized ectoderm. Even at high concentrations of activin, 200 pM, both the frequency and intensity of En-2 expression in UV explants are low and the frequency in control explants is not increased. Thus the lack of induction is not due to a requirement for more factor by ventral ectoderm. Because UV animal caps are able to express En-2 when recombined with anterior notochord (Hemmati-Brivanlou et al., 1990b), the lack of En-2 expression in response to activin in most UV explants does not result from an inability of the explant to make anterior neural structures. From our results and the finding of Sokol and Melton (1991) that activin-treated ventral ectoderm does not make eyes, we conclude that hindbrain is the anterior extent of neural induction in response to activin.

Our finding that neural markers such as N-CAM are efficiently induced in ventralized animal caps contrasts with that of Sokol and Melton (1991) who found that ventral halves of animal caps and UV animal caps differentiate only epidermis, muscle and mesenchyme in response to activin. However, these explants were examined morphologically and not with the use of molecular markers, and consequently the explants may also have induced neural structures not apparent in the assay. The discrepancy in our conclusions highlights the power of specific molecular markers and whole-mount staining techniques. Our data are consistent with the experiments of Thomsen et al. (1990) which show that activin mRNA injected into ventral blastomeres leads to the duplication only of tail and trunk but never head structures.

It is clear from experiments reviewed in the introduction that activin is a mesoderm inducer and not a neural inducer. Activin must therefore be able to induce types of inesoderm which in turn induce regional neural markers. In response to activin, UV animal caps are unable to form mesoderm that can induce formation of anterior neural structures; in contrast they are com petent to make mesoderm that can induce other neural markers. To determine what types of mesoderm were made in the activin-treated explants, we assayed the isolates for muscle actin by northern analysis and for notochord by immunocytochemistry. While actin was induced to about the same extent in UV and control animal caps, notochord was made in 34% of control animal caps and only 5.4% of UV animal caps. Taken together, these results are consistent with the conten tion that control explants make anterior structures and UV explants do not because the two types of isolates make different spectra of mesodermal types in response to activin.

Although it is a formal possibility that activin treatment of ventralized animal caps renders the cells unable to respond to the anterior neural inducer, the results of our northern analysis suggest that this is not the case. Green and Smith (1990) showed that low doses of activin turn off epidermal markers, yet we find that epidermal cells (cells expressing E13) are present in activin-treated UV animal caps. Therefore, we conclude that, as in the normal caps, there remain cells untransformed by activin that should remain competent to express En-2 in the presence of the proper inducer. So, although activin can direct the differentiation of posterior brain structures, other signals, presumably emitted from the dorsal mesoderm, are required for the specification of the most anterior neural structures.

Although we previously found that anterior noto chord is the strongest inducer of En-2, here we show that the expression of En-2 does not necessarily depend on the presence of notochord. Others (Jones and Woodland, 1989; Hemmati-Brivanlou et al., 1990b; Green et al., 1990; Green and Smith, 1990) have shown that types of dorsal mesoderm besides notochord are able to induce neural marker expression in naive ectoderm. Thus a cell’s ability to induce neural tissue may depend on properties that are not directly related to the cell’s final differentiated state.

The activin-treated animal caps show a remarkable coherence of induced structures; mRNAs such as Krox-20 and En-2 are found in discrete patches or stripes and not distributed throughout the explants. Green et al. (1990) have previously suggested that differential exposure of cells to activin in animal caps leads to heterogeneity of response; the heterogeneity is abolished by dispersion of the cells (Green and Smith, 1990). Even when activin-treated cells were reaggregated in a random fashion, expression of muscle-and notochord-specific markers was seen in patches instead of isolated spots and the patches of muscle and notochord were often in association with each other. These results suggest that, after an initial inductive signal, cell-cell interactions play a role in organizing the tissue differentiation. Similar interactions may be responsible for the patterned expression of Krox-20 and En-2 in the animal caps.

It is important to stress that although dorsal and ventral animal caps yield different responses to peptide growth factors, or to weak inducers from the embryo, the dorsal mesoderm does not require a “predisposed” ectoderm to induce a full range of dorsoanterior structures. In Xenopus, this is most apparent from organizer grafts, where dorsal mesoderm can induce complete heads (see for example, Gimlich and Cooke, 1983; Smith and Slack, 1983; Cooke, 1989). Neverthe less, there are clear differences in dorsal and ventral animal pole cells which are apparent from an early stage. Both Kageura (1990) and Gallagher et al. (1991) found differences in the inducing ability or fate of dorsal blastomeres at the cleavage stage. London et al. (1988) showed that an epidermal marker is already specified to turn off in dorsal cells at early cleavage stages. As animal caps are taken at successively later stages the specification of dorsal animal cap to extinguish ex pression becomes stronger, and the boundary between prospective expressing and non-expressing cells is sharpened. This progressive change suggests that the predisposition of the dorsal side is reinforced by signals that are outside the animal cap. In the blastula, the dorsal animal cap differs in response to activin from the ventral animal cap (Sokol and Melton, 1991). By the time gastrulation begins, there are differences in the ability of dorsal and ventral animal cap to differentiate neural tissue in response to a weak inducer of neural tissue, the early involuted mesoderm (Sharpe et al., 1987) or in response to activin (Ruiz i Altaba and Jessell, 1991). In a urodele, the predisposition is so strong at the onset of gastrulation that dorsal animal caps can differentiate neural tissue autonomously (Barth, 1941).

What is the nature of the “predisposition” of dorsal animal cap to form neural tissue and neural tissue of more anterior character? Sharpe et al. (1987) suggested that this predisposition occurs “before induction has commenced,” and the results of Kageura (1990) and Gallagher et al. (1991) support this idea. However, it is likely that much of the predisposition in the late blastula and early gastrula arises through early inductive signals that spread through the plane of tissue from the mesoderm to the prospective neural tissue. Over an extended period, such “planar” induction is sufficient to induce large amounts of neural tissue (Kintner and Melton, 1987; Dixon and Kintner, 1989).

The next step is to identify the molecules required for the induction of anterior neural structures. Members of the Xenopus wnt gene family, in particular Xwnt-8, have been shown to rescue an axis in UV-irradiated embryos (Smith and Harland, 1991; Sokol et al., 1991). However, so far no wnt gene has been identified that is actually expressed in the proper spatial and temporal context to direct axis formation. Recently, using expression cloning, we have isolated a novel factor that can rescue a complete axis and is present at a time and in a place consistent with its dorsalizing activity (W. C. Smith and R.M.H., unpublished results). Like Xwnt-8 (Christian et al., 1992) this factor may synergize with factors such as FGF and activin to promote formation of anterior dorsal structures. The expression cloning technique may prove fruitful in isolating other mol ecules that induce anterior neural tissues.

We are indebted to Wylie Vale and Joan Vaughan for the gift of the purified activin A and David Wilkinson and L.C. Bradley for the Xenopus Krox-20 cDNA clone. We thank Scott Staebell for suggesting the use of Tor antibodies, and Chris Kintner, Doug Melton and David Wilkinson for communication of results prior to publication. This work was supported by a grant from the N.I.H. toR.M.H. and from the ALS foundation to P.D.K.