Xenopus mothers against decapentaplegic is an embryonic ventralizing agent

Transcription

1 Development 122, (1996) Printed in Great Britain The Company of Biologists Limited 1996 DEV Xenopus mothers against decapentaplegic is an embryonic ventralizing agent that acts downstream of the BMP-2/4 receptor Gerald H. Thomsen Department of Biochemistry and Cell Biology, Institute for Cell and Developmental Biology, State University of New York, Stony Brook, NY , USA SUMMARY Dorsal-ventral patterning in vertebrate embryos is regulated by members of the TGF-β family of growth and differentiation factors. In Xenopus the activins and Vg1 are potent dorsal mesoderm inducers while members of the bone morphogenetic protein (BMP) subclass pattern ventral mesoderm and regulate ectodermal cell fates. Receptors for ligands in the TGF-β superfamily are serinethreonine kinases, but little is known about the components of the signal transduction pathway leading away from these receptors. In Drosophila the decapentaplegic protein (dpp), a homolog of vertebrate BMP-2 and BMP-4, functions in dorsal-ventral axial patterning, and a genetic screen for components involved in signaling by dpp has identified a gene named mothers against decapentaplegic (Mad). Mad encodes a unique, predicted cytoplasmic, protein containing no readily identified functional motifs. This report demonstrates that a gene closely related to Drosophila Mad exists in Xenopus (called XMad) and it exhibits activities consistent with a role in BMP signaling. XMad protein induces ventral mesoderm when overexpressed in isolated animal caps and it ventralizes embryos. Furthermore, XMad rescues phenotypes generated by a signalingdefective, dominant-negative, BMP-2/4 receptor. These results furnish evidence that XMad protein participates in vertebrate embryonic dorsal-ventral patterning by functioning in BMP-2/4 receptor signal transduction. Key words: mesoderm, pattern formation, signal transduction, Mad, BMP, Xenopus, Drosophila INTRODUCTION Bone morphogenetic proteins (BMPs) are members of the transforming growth factor β (TGF-β) superfamily, and in Xenopus embryos they participate in patterning the ventral mesoderm and the ectoderm. BMP-2 and BMP-4 are capable of inducing ventral mesoderm and re-specifying prospective dorsal mesoderm to differentiate into ventral tissues in Xenopus embryos (Koster et al., 1991; Dale et al., 1992; Jones et al., 1992; Fainsod et al., 1994; Clement et al., 1995; Hemmati-Brivanlou and Thomsen, 1995). BMP signals are required in the embryo to form ventral mesoderm because blocking BMP-2/4 receptor activity in the ventral part of the embryo eliminates blood formation and dorsalizes the mesoderm (Graf et al., 1994; Suzuki et al., 1994). BMP-4 protein also promotes the differentiation of epidermis from Xenopus ectoderm (Wilson and Hemmati-Brivanlou, 1995), and if BMP signals in the ectoderm are blocked by a dominantnegative BMP-2/4 receptor (Sasai et al., 1995), or if ectodermal cells are disaggregated (Grunz and Tacke, 1989), they differentiate into neural tissue. Embryonic dorsal-ventral patterning in Drosophila is accomplished partly through the action of decapentaplegic (dpp), which is a TGF-β member most closely related to vertebrate BMP-2 and BMP-4. These factors are probably true homologs since the signaling systems in which they function have been conserved in evolution as a mechanism to generate dorso-ventral polarity in insect and vertebrate embryos. The axial polarity of these organisms is, however, reversed relative to one another (Hogan, 1995; Holley et al., 1995; Winnier et al., 1995). This conservation of molecular players in dorsalventral patterning across a wide evolutionary distance is underscored by the capacity of BMP-4 and dpp to functionally substitute for one another in vertebrate and Drosophila embryos (Padgett et al., 1993; Holley et al., 1995). Thus it is anticipated that vertebrate homologs of molecules involved in transmitting or receiving dpp signals may participate in inductive signaling by BMPs in vertebrate embryos. Receptors for TGF-β-related ligands consist of heterodimers of single transmembrane serine/threonine kinases that are classified as type I or type II according to structural and functional features (Massague et al., 1994). Types I and II BMP receptor subunits have been cloned from animals as diverse as nematodes and mammals (Estevez et al., 1993; Graf et al., 1994; Suzuki et al., 1994; ten Dijke et al., 1994; Liu et al., 1995), including a receptor in Xenopus that binds BMP-2 and BMP-4 (BMP-2/4 receptor; Graf et al., 1994; Suzuki et al., 1994). Type II receptor subunits can bind ligand and are constitutively phosphorylated, but receptor signals are not transmitted until a type II receptor pairs with a type I receptor in a

2 2360 G. H. Thomsen ligand-dependent step. This results in phosphorylation of the type I subunit and activation of the receptor complex (Wrana et al., 1994; Liu et al., 1995). Little is known, however, about the signal transduction components downstream of TGF-βrelated receptors. Studies in Xenopus suggest that crosstalk can occur between the signal transduction pathway of activin receptors and receptor tyrosine kinases (Cornell and Kimelman, 1994; LaBonne and Whitman, 1994; Umbhauer et al., 1995), but signal transduction molecules for tyrosine kinase receptors probably do not directly function in signaling by TGF-β-related receptors. A genetic screen for enhancers of weak dpp alleles uncovered a maternal-effect gene named mothers against dpp, or Mad (Raftery et al., 1995; Sekelsky et al., 1995), which plays a role in dpp signaling. Drosophila Mad (herein referred to as DMad) encodes a unique protein, predicted to be intracellular but containing no recognizable functional motifs, and it is a member of a new family of proteins (Sekelsky et al., 1995). Mad-related genes, dubbed dwarfins, have also been identified in C. elegans and they correspond to the small alleles, sma-2, sma-3 and sma-4 (Savage et al., 1996). A pancreatic tumor suppressor gene, DPC4, has also been identified that is related to Mad and the dwarfins (Hahn et al., 1995). A prediction from genetic studies in the fly and nematode is that Mad proteins may function in TGF-β receptor signal transduction. This paper reports on a Xenopus homolog of Drosophila Mad that functions in the signal transduction pathway from BMP receptors. The Xenopus Mad gene (XMad) is closely related to DMad, and both encode factors that induce and pattern ventral mesoderm in Xenopus embryos. Significantly, XMad rescues mesodermal and ectodermal patterning defects generated by a dominant-negative BMP-2/4 receptor, placing XMad downstream of the BMP-2/4 receptor. The evidence I present strongly supports a function for Mad proteins in signal transduction from TGF-β-related receptors. This study also provides additional support for the hypothesis that components of the vertebrate BMP and Drosophila dpp signaling systems are homologous, and constitute a mechanism to generate dorsal-ventral pattern that has been conserved during evolution. MATERIALS AND METHODS Cloning of Xmad A Xenopus blastula (stage 9) cdna library in Lambda Zap was screened with a random-primed 32 P-labeled DMad probe. About cpm were hybridized at 42 C to plaques on nylon filters, in 50% formamide, 5 Denhardt s, 5 SSC and 0.1 mg/ml denatured salmon sperm DNA. Filters were washed in 0.5 SSC, 0.5% SDS, 55 C. More than 25 cdnas were initially isolated and among the purified clones a full-length cdna was identified and sequenced (Sanger et al., 1977) on both strands by primer walking. DNA and protein sequences were analyzed with Lasergene software (DNAStar Inc.). Xmad expression studies In the developmental northern blot, 20 µg of total RNA were loaded per lane on the gel, which was run and transferred to nylon membrane (Micron Separations, Inc.) in 10 SSC according to standard procedures (Sambrook et al., 1990). The blot was pre-hybridized and probed in 50% formamide, 5% SDS, 1 SSC, 1 Denhardt s and 0.1 mg/ml sonicated salmon sperm DNA. The probe was a 32 P-labeled, random-primed fragment of the noncoding 3 end of XMad. Final blot wash conditions were 60 C, 0.2 SSC, 0.2% SDS. In situ hybridization was performed with an antisense, digoxygenin-labeled RNA probe (Harland, 1991) using BMPurple colorimetric substrate (Boehringer-Mannheim). Embryonic ventralization and mesoderm induction assays BMP-4 (Nishimatsu et al., 1992) and XMad were cloned into the CS2+ vector (Rupp et al., 1994), and DMad was cloned into psp64t (Krieg and Melton, 1987). Capped mrna was synthesized from linearized vectors using mmessage Machine kits (Ambion). In ventralization and animal cap assays, 1.5 ng of mrna were injected per blastomere. Animal caps were injected at the 2-cell stage and cut at blastula stages 8 to early 9. RNA was prepared from whole embryos and embryonic explants and northern blots were performed as described (Thomsen and Melton, 1993). VMZs were isolated as described (Graf et al., 1994; Hemmati-Brivanlou and Thomsen, 1995) RT-PCR assays and primer sequences were as published (Wilson and Melton, 1994; Wilson and Hemmati-Brivanlou, 1995), except for αt1 globin (accession no. J00976), which were: upstream, 5 -TTGCTGTCTCA- CACCATC-3 (bases 30-37); downstream, 5 -TCTGTACTTGGAG- GTGAG-3 (bases ). Some primer sequences were obtained from the Xenopus Molecular Marker Resource on the internet (http//: vize222.zo.utexas.edu/). The NCAM primers span a putative intron (unpublished observations), and the sequences are: upstream primer, 5 -CACAGTTCCACCAAATGCCG-3 (bases ); downstream primer, 5 -GCTGGGGTGCCCTTGACATC-3 (bases ). Amplification with NCAM primers was performed by adding cdna template to the PCR reaction during the first 94 C denaturation step. RESULTS Sequence and embryonic expression of Xenopus Mothers against dpp In preliminary studies DMad was tested for activity in Xenopus (by microinjecting synthetic mrna) and was found to ventralize embryos (Fig. 3E) in a fashion similar to the BMP-4 ligand. Therefore it was anticipated that a Xenopus homolog of DMad might exist and function in inductive signaling by BMP ligands. A full-length Xenopus Mad gene (XMad) was isolated from a blastula stage cdna library using DMad as a probe, and XMad encodes a protein of 464 amino acids that is 75% identical to DMad protein and 85% similar when conservative substitutions are considered (Fig. 1). This remarkable level of conservation between XMad and DMad proteins is especially striking within the N- and C-terminal regions. Within the N-terminal portion of XMad, 133 of 145 residues (92%) are identical to DMad. At the C terminus, 112 of the last 120 amino acids (93%) are identical to DMad, and each difference represents a conservative substitution, with the exception of residue 339. Several mutations in DMad alleles map to the C-terminal 50 amino acids, highlighting the functional importance of this very conserved domain (Sekelsky et al., 1995). The final four amino acids at the C terminus of DMad and XMad comprise a serine tail (SSVS), and XMad contains 47 serine, 21 threonine and 16 tyrosine residues, altogether constituting 18% of the protein. These represent an abundance of potential target sites for phosphorylation (or other modifications), which might be important in regulating

3 Xenopus mothers against dpp 2361 Mad activity given that TGF-β-related receptors are serine/threonine kinases. Surprisingly, XMad and other Madrelated proteins contain no conserved protein domains that would suggest particular functions for these factors, other than predicting that they will be cytoplasmic. Analysis by northern blot revealed that transcripts of XMad are stored maternally and that the gene is expressed throughout development through swimming tadpole stages (Fig. 2A). The most abundant transcript is 3.4 kb in size, but minor transcripts of 3.0 and 4.3 kb are also detected. The full-length XMad cdna cloned in this study is 3.2 kb long, closely correlating with these transcript sizes, but it is not known whether these transcripts derive from several related XMad genes or are alternative splicing products of a single gene. By whole mount in situ hybridization in embryos (Fig. 2B- E) XMad mrna is not localized in the blastula and early gastrula stages, a result confirmed by a northern blot on RNA of dissected embryonic regions (data not shown). XMad transcripts begin to show localized expression at mid-gastrula stages, at which time they become enriched within involuting mesoderm around the yolk plug, in addition to being abundantly expressed in the ectoderm (Fig. 2C). At the end of neurulation XMad is localized to the nervous system (not shown), and by early tadpole stages XMad transcripts are restricted to the central nervous system, eye and head neural crest cell populations (Fig. 2D,E). XMad transcripts are located in areas adjacent to, or overlapping with, regions where BMP-2 and BMP-4 transcripts are found (Fainsod et al., 1994; Hemmati- Brivanlou and Thomsen, 1995; Schmidt et al., 1995), consistent with expectations that a molecule involved in BMP receptor signaling would be expressed in cells near a source of BMP ligands. Ventral mesodermal patterning by Xenopus Mad In Xenopus embryos and animal cap assays XMad and DMad display similar activities when expressed from microinjected mrna. Expression of 3 ng of XMad mrna in the dorsal marginal zone completely ventralizes the embryos (Fig. 3C), yielding a dorsoanterior index (Kao and Elinson, 1989), or DAI of 0 (n=18). This is also reflected by the highly elevated levels of red blood cell differentiation (a ventral mesodermal derivative) and a complete loss of muscle (a dorsal mesodermal tissue) in those embryos (compare lanes a and c in Fig. 3F). DMad also ventralizes Xenopus embryos, albeit less effectively than XMad (DAI=1.9, n=20; Fig. 3E). Over-expression of XMad on the ventral side results in embryos with a slightly enlarged posterior-ventral region, but which otherwise appear normal (DAI=5, n=20; Fig. 3D). Perhaps this reflects an increase in posterior mesoderm. Red blood differentiation is also slightly higher in these embryos compared to controls (compare lanes b and d, Fig. 3F). XMad and DMad induce ventral-posterior mesoderm in animal caps when expressed from microinjected mrna (Fig. 4). Injection of 3 ng of mrna for either factor induces mesoderm-specific genes that mark ventral (globin, Xwnt-8), posterior (Xhox-3 and XlHbox-6) and lateral plate (Xtwist) territories, but markers of dorsal mesoderm such as goosecoid and cardiac (muscle) actin are not induced (data not shown). XMad and DMad do not induce neural tissue (see Fig. 6). The fact that mesoderm induction by XMad and DMad is essentially identical to that of BMP-4 suggests that Mad activates the BMP-4 signaling pathway. XMad functions downstream of the BMP-2/4 receptor If XMad operates in inductive signaling by BMP-2 or BMP-4, does it act before or after the reception of a BMP ligand by its receptor? If XMad functions within cells that receive BMP signals, elevated levels of XMad protein might compensate for a block in the activity of BMP-2/4 receptors. If, however, XMad functions upstream of the receptor, for example in steps XMad DMad 1 MN MDTDDVESNTSSAMST VTSLFSFTSPAVKRLLGWKQGDEEEKWAEEAVDALVKKLKKKKGAIQELEKALTCPGQPSNCVTIPRSLDGRLQVSHRKG :. :. : : : : : : LGSLFSFTSPAVKKLLGWKQGDEEEKWAEKAVDSLVKKLKKRKGAIEELERALSCPGQPSKCVTIPRSLDGRLQVSHRKG LPHVIYCRVWRWPDLQSHHELKPLECCEYPFGSKQKEVCINPYNYKRVESPVLPPVLVPRHSEYNPQHSLLAQFRNLEPS : :: : :. :.:: : LPHVIYCRVWRWPDLQSHHELKPLELCQYPFSAKQKEVCINPYHYKRVESPVLPPVLVPRHSEFAPGHSML-QFNHV--A EPHMPHNATFPDSFQQPNSHPFPHSPNSSYPNSPGSSSTYPHSPASSDPGSPFQIPADTPPPAYMPPEDQMTQDNSQPMD.:::: :.:.:::. :. :.: :. :: : : :::: EPSMPHNVSYSN SGF-----NSHSLSTSNTSVGSPSSVNSNPNSPYDSLAGTPPPAYSPSED---GNSNNPND TNMMVPNISQDINRADVQAVAYEEPKHWCSIVYYELNNRVGEAFHASSTSVLVDGFTDPSNNRNRFCLGLLSNVNRNSTI.. ::..:: :.. :... ::. : :.: -GGQLL--DAQM--GDVAQVSYSEPAFWASIAYYELNCRVGEVFHCNNNSVIVDGFTNPSNNSDRCCLGQLSNVNRNSTI ENTRRHIGKGVHLYYVGGEVYAECLSDSSIFVQSRNCNFHHGFHPTTVCKIPSGCSLKIFNNQEFAQLLAQSVNHGFETV. : : : : : : : ENTRRHIGKGVHLYYVTGEVYAECLSDSAIFVQSRNCNYHHGFHPSTVCKIPPGCSLKIFNNQEFAQLLSQSVNNGFEAV YELTKMCTIRMSFVKGWGAEYHRQDVTSTPCWIEIHLHGPLQWLDKVLTQMGSPHNPISSVS : YELTKMCTIRMSFVKGWGAEYHRQDVTSTPCWIEIHLHGPLQWLDKVLTQMGSPHNAISSVS Fig. 1. Protein sequence comparison of Xenopus Mad with Drosophila Mad. The predicted Xenopus Mad (XMad) protein sequence is aligned with Drosophila Mad protein (DMad). The XMad cdna encodes a predicted protein of 464 amino acids. XMad protein is 75% identical to DMad at the amino acid level (identical residues are indicated by vertical bars) and 85% similar when conservative amino acid substitutions are considered (double dots). Note the high degree of conservation in the N-terminal and C-terminal portions of the proteins. The GenBank accession number for XMad is U58834.

4 2362 G. H. Thomsen Fig. 2. Developmental expression of Xenopus Mad. (A) A developmental northern blot of total embryonic RNA shows that XMad is maternal and expressed at all stages of early development. Three transcripts are detected with the sizes (in kb) indicated on the left. The most abundant mrna is 3.4 kb, and minor transcripts of 4.3 kb and 3.0 kb are also detected. It is not known whether these transcripts derive from closely related genes or splicing variants. Lane numbers correspond to developmental stages (Nieuwkoop and Faber, 1967): (7, 9) blastula, (11) gastrula, (15, 18) neurula, (26) tailbud tadpole and (38) swimming tadpole. (B) In situ hybridization shows that XMad transcripts are uniformly distributed in the early gastrula (stage 10). Dense staining covers the entire prospective ectoderm and marginal zone, but vegetal cells (the lighter region) do not stain efficiently by this procedure (Harland, 1991). A northern blot on isolated dorsal, ventral, animal and vegetal regions confirmed the uniformity of XMad expression at early stages (not shown). (C) A mid-gastrula embryo (stage 12) split sagitally reveals that XMad is expressed in the ectoderm and neurectoderm (arrows mark the ectodermal-mesodermal boundary), and expression in the underlying mesoderm is greater in the posterior, adjacent to the yolk plug (YP). This embryo is lightly pigmented and the brown line anterior to the yolk corresponds to involuted bottle cells. The embryo is positioned with the anterior to the left and dorsal at the top. (D) At tailbud tadpole stage 26 XMad expression is high in the central nervous system and head. (E) A close-up of the head of the embryo in D, highlighting XMad expression in the brain (b), eye (e) and head neural crest derivatives (mc, mandibular crest; hc, hyoid crest; abc, anterior branchial crest; pbc, posterior branchial crest). Expression in the otic vesicle, between the hyoid crest and anterior branchial crest, is also visible. Scale bars, 0.1 mm. associated with ligand synthesis, processing or secretion, XMad would not be expected to complement defective receptors. Truncated BMP-2/4 receptors lacking an intracellular ser/thr kinase domain act as dominant-negative inhibitors of cellular responses to BMP-2 and BMP-4 ligands in Xenopus Fig. 3. XMad and DMad ventralize Xenopus embryos. Synthetic mrnas encoding control (pgem vector) or Mad sequences were injected into the equatorial region of two dorsal or two ventral blastomeres at the 4-cell blastula stage, and phenotypes were scored at tadpole stage 40. Dorsal (A) or ventral (B) injections of control (pgem vector) mrna resulted in normal embryos. (C) Dorsal injection of XMad mrna caused severe ventralization. (D) Ventral injection of XMad mrna caused posterior thickening and a slight reduction in the tail. The average dorso-anterior index, a measure of the degree of dorsal and anterior mesodermal patterning (Kao and Elinson, 1989), was for each group: A, DAI=5 (n=18); B, DAI=5 (n=18); C, DAI=0 (n=18); D, DAI=5 (n=20). (E) Expression of DMad in dorsal blastomeres ventralized the embryos (DAI=1.9, n=20), while ventral expression yielded relatively normal embryos (DAI=5, n=21; not shown). (F) Northern blot analysis of muscle actin (a dorsal mesodermal marker) and αt1 globin (a ventral mesodermal marker) gene expression in the control and XMadinjected embryos shown in A-D. Lanes a-d correspond to embryos in A-D. Note the loss of muscle actin and the significant increase in globin expression in embryos injected dorsally with XMad mrna (lane c). This indicates that overexpression of XMad protein in the presumptive dorsal mesoderm respecified its fate to that of ventral mesoderm. Ventral expression of XMad boosted globin expression slightly (lane d), but dorsal or ventral injections of vector mrna had no effect (lanes a and b). Histone H4 mrna (Perry et al., 1985) was scored as a control for RNA loading. embryos (Graf et al., 1994; Maeno et al., 1994; Suzuki et al., 1994). Ectopic expression of dominant-negative BMP receptors in the ventral side of Xenopus embryos causes prospective ventral mesoderm to differentiate into dorsal mesoderm, resulting in development of ectopic dorsal axial structures. Cultured VMZ explants expressing a dominantnegative BMP receptor are similarly dorsalized, do not form blood and instead develop muscle (Graf et al., 1994; Suzuki et al., 1994; and see Fig. 5). In Xenopus ectoderm expression of a dominant-negative BMP-2/4 receptor triggers neural differentiation by blocking receptor activation by endogenous BMP-2 or BMP-4 (Sasai et al., 1995). The likelihood that BMP-4 signals specify epidermal rather than neural differentiation was demonstrated by exposing dissociated animal cap cells to BMP-4 protein

5 Xenopus mothers against dpp 2363 followed by reaggregation (Wilson and Hemmati-Brivanlou, 1995). This treatment induces epidermis in the animal cap cells, which otherwise differentiate into neurons (Grunz and Tacke, 1989; Wilson and Hemmati-Brivanlou, 1995). Neural differentiation can be considered the default program that ectodermal cells follow in the absence of BMP-4 signals. If XMad functions downstream of the BMP-2/4 receptor in Xenopus, over-expression of XMad might rescue phenotypes generated by interfering with BMP-2/4 receptor activity. This possibility was tested by co-expressing XMad together with a dominant-negative BMP-2/4 receptor (tbr, (Graf et al., 1994). Expression of XMad together with tbr reverses the formation of ectopic axial structures caused by expression of tbr alone in the ventral side of whole embryos (compare Fig. 5A with 5B). This rescue was 100% effective at a dose of XMad mrna (50 pg) equivalent to that of tbr mrna needed to generate a secondary axis. Defects caused by higher doses of tbr (up to 200 pg mrna) are also rescued by an equivalent amount of XMad mrna (data not shown). Cultured VMZ explants injected with tbr mrna elongate and produce melanocytes indicating neural induction (Fig. 5D), and they form muscle but not erythrocytes, as demonstrated by a northern blot on the VMZ RNA (Fig. 5G, lane d). Co-expression of XMad and tbr (each at 50 pg mrna) in VMZ explants completely reverses these effects and the rescued VMZs form blood, lack muscle, and look like control VMZ explants (Fig. 5E-G). Intact animal cap explants form epidermis in culture, but they undergo neural differentiation if signaling by BMP-2/4 receptors is blocked (Sasai et al., 1995; Xu et al., 1995). Fig. 6 demonstrates that XMad can suppress neural differentiation triggered by the dominant-negative BMP-2/4 receptor, tbr. Expression of tbr alone induces N-Cam and NRP-1 (a ribonucleoprotein gene enriched in neural tissue), but induction of these markers is reversed by expressing XMad (XM) together with tbr. XMad alone does not induce neural tissue. The results of this dominant-negative BMP receptor rescue experiment, like those done in the marginal zone (Fig. 5), demonstrate that XMad functions downstream of the BMP-2/4 receptor. The results in animal caps also suggest that control over the activity of XMad could be a mechanism to regulate the decision between epidermal or neural differentiation. It is curious that it takes much less XMad mrna to rescue the dominant-negative-bmp receptor than it does to induce mesoderm or ventralize embryos. An interpretation of these results is that a relatively low level of BMP signaling is sufficient to suppress both the neuralization of the ectoderm and the latent capacity of the presumptive ventral mesoderm to develop into dorsal, trunk-type mesoderm when BMP signals are blocked. Mesoderm induction in the animal cap, and ventral- Fig. 4. XMad and DMad induce mesoderm in animal caps. The panel shows an RT-PCR analysis of mesodermal marker gene expression in animal caps injected with 3.0 ng of mrna for pgem (C), DMad (DM), XMad (XM) or Xenopus BMP-4 (B4). The last two lanes show RT-PCR products from stage-18 embryonic cdna synthesized in the presence (emb RT+) or absence (emb RT ) of reverse transcriptase to control for cdna synthesis and DNA contamination, respectively. Note that expression of DMad, XMad and BMP-4 induced each of the ventro-posterior mesoderm markers assayed. EF1-α was scored as a positive control for cdna synthesis. Animal caps were harvested at stage 30 (tadpole) to score αt1 globin expression, stage 11 (mid-gastrula) to score Xwnt 8 and Xtwist, and stage 18 (neurula) to score Xhox-3 and Xlhbox6. The ef1-α signal shown corresponds to that of stage 11 cdna, but all other cdna samples treated with reverse transcriptase were positive. Fig. 5. XMad rescues dominant-negative BMP2/4 receptor phenotypes. The panels display phenotypes of whole embryos (A-C) and isolated ventral marginal zones (VMZs, D-E) expressing the dominant-negative BMP-2/4 receptor (Graf et al., 1994), alone or in combination with XMad. 50 pg of each mrna were injected into the marginal zone of two ventral blastomeres at the 4-cell stage. VMZs were explanted at stage 10.5, and embryos and VMZ explants were scored at stage 40. (A) Expression of the dominant-negative BMP- 2/4 receptor (tbr) from injected mrna resulted in tadpoles with secondary axial structures in 33% of the cases (n=18). A typical secondary axis is indicated by the arrows. (B) Co-expression of XMad together with tbr resulted in 100% normal embryos (n=18). (C) Injection of control (pgem) mrna resulted in 100% normal embryos (n=21). (D) VMZs expressing tbr elongated and developed pigmented melanocytes, a neural derivative. (E) VMZs from embryos co-expressing tbr and XMad were rescued to normalcy and formed oblong belly pieces like control VMZs (F). (F) Injection of control (pgem) mrna into VMZs resulted in a typical wild-type VMZ morphology. (G) A northern blot on RNA from the VMZs shown in D-F. VMZs expressing tbr (lane d) developed dorsal mesoderm, as revealed by the expression of muscle actin, and they lacked ventral mesoderm (red blood) as reflected by the absence of αt1 globin expression. When XMad was co-expressed with tbr (lane e) dorsalization of the VMZ was reversed; the explants lacked muscle and expressed globin, similar to control VMZs (lane f). Cytoplasmic actin mrnas cross-hybridize with the muscle actin probe and migrate as two bands above the muscle-specific message, providing a positive control for RNA loading in the gel.

6 2364 G. H. Thomsen Fig. 6. Neural induction by tbr expression in animal cap ectoderm is inhibited by co-expression of XMad. Lanes 1-4 correspond to an RT- PCR analysis of cdna from animal caps injected at the 2-cell stage with 2 ng pgem RNA (lane C),1.0 ng XMad mrna (lane XM), 50 pg tbr mrna (lane tbr), and 50 pg tbr mrna plus 50 pg XMad mrna (lane tbr + XM). Caps were cut at stage 8 and harvested at neurula stage 18. Note the induction of N-CAM and NRP1 (a neuralenriched ribonucleoprotein; Richter et al., 1990) mrnas when BMP signals were blocked by tbr. This effect was suppressed by coexpression of XMad. Some background expression of NRP1 is normal. Furthermore, XMad alone did not induce neural tissue (lane XM). Lanes 5 and 6 are positive and negative controls for RT-PCR as in Fig. 2. EF1-α is a positive control for cdna synthesis. ization of the dorsal/anterior mesoderm of the organizer, however, appear to require a much higher level of XMad (or BMP) activity. In fact, there is evidence that Xenopus animal cap ectoderm has low and high response thresholds for BMP- 4 ligand. Wilson and Hemmati-Brivanlou (1995) showed that it takes only 0.3 ng/ml of BMP-4 to switch ectodermal cells from a neural to epidermal fate, yet it takes a much higher dose, 1000 ng/ml, to induce mesoderm. DISCUSSION Drosophila mothers against dpp was isolated as an enhancer of dpp alleles, placing it somewhere in the pathway of Dpp signaling. The Drosophila genetic studies did not, however, establish whether Mad protein acts in cells that deliver or respond to dpp ligand. The present study establishes that a vertebrate homolog of Drosophila Mad exists and functions in the specification of embryonic ventral mesodermal cell fates. XMad may also function in the decision between neural and epidermal cell fates in the ectoderm. XMad mimics all the activities of the BMP-4 ligand: it can induce mesoderm, respecify presumptive dorsal mesoderm to form ventral mesoderm and suppress neural differentiation. Most importantly, this study demonstrates that XMad can rescue defects caused by blocking signals from a BMP-2/4 receptor, providing strong evidence that XMad functions in BMP signal transduction. The possibility that XMad is a component of signal transduction pathways for receptors other than the BMPs in Xenopus seems unlikely. Among the known mesoderminducing factors in Xenopus only BMP-4 is capable of inducing ventral mesoderm, ventralizing dorsal mesoderm and inhibiting neural differentiation, each of which I have shown in the present study to be properties of XMad. Of the known mesoderm inducers, activin, Vg1 and fibroblast growth factor (FGF) each induce some types of ventral mesoderm, but they do not induce blood (Kessler and Melton, 1994). Conversely, activin and Vg1 induce primarily dorsal mesoderm, but maximum doses of XMad, BMP-2 and BMP-4 do not, arguing that XMad does not function in signal transduction pathways for activin, Vg1 or indeed other factors capable of inducing or patterning dorsal mesoderm such as noggin (Smith et al., 1993), chordin (Sasai et al., 1994), or nodal-related factors (Jones et al., 1995). In the ectoderm BMP-4 protein, like XMad, inhibits neural differentiation, but activin does not (Wilson and Hemmati-Brivanlou, 1995). In addition, FGF induces and patterns neural tissue (Doniach, 1995), which is a property opposite to that of XMad. Thus I conclude that XMad functions in signal transduction from BMP-2/4 receptors. XMad may also function in BMP-7 signal transduction, since this BMP can also specify ventral mesoderm like BMP-2 and BMP-4 (S. Nishimatsu and G. H. Thomsen, unpublished observations). It is, of course, formally possible that XMad operates in another, perhaps parallel, signal transduction pathway that triggers biological responses similar to those of BMPs. Generally speaking the sum of the data predict that various Mad proteins will function in signal transduction for the family of TGF-β receptors, but direct biochemical tests will be required to firmly establish the link between TGF-β-related receptors and Mad proteins. In this study overexpression of XMad was shown to induce and pattern ventral but not dorsal mesoderm, which prompts the suggestion that other Mad-related factors may transduce signals for TGF-β-related ligands involved in dorsal mesoderm induction or patterning, such as activin, Vg1 or nodal-related proteins. Similarly, the TGF-βs proper are principally involved in the control of cell growth, so it is tempting to suggest that DPC4, a Mad-related tumor suppressor gene, operates in TGFβ receptor signaling (Hahn et al., 1995), but that remains to be established. The conclusion that Mad proteins function in signal transduction for members of the TGF-β family is also supported by genetic experiments in C. elegans, which demonstrate that mutations in the Mad-related small genes (or dwarfins) phenocopy mutations in the C. elegans BMP-2/4 receptor, daf-4, and are not rescued by daf-4 overexpression (Savage et al., 1996). It is interesting that wild-type Xenopus and Drosophila Mad display activity (mesoderm induction, BMP receptor rescue) when overexpressed as wild-type proteins, in contrast to other signal transduction molecules such as wild-type ras and MAP kinase, which do not induce mesoderm (Whitman and Melton, 1992). Instead only mutant, constitutively activated, forms of ras and MAP-kinase induce mesoderm when overexpressed in animal caps (Whitman and Melton, 1992; Cornell and Kimelman, 1994; LaBonne and Whitman, 1994; Schmidt et al., 1995). My results imply that XMad protein has some level of intrinsic activity when expressed in animal cap and marginal zone cells, even in the absence of BMP-2/4 receptor activity. This behavior of XMad is similar to that seen with components in the wnt signal transduction pathway, such as disheveled (Sokol et al., 1995) and GSK-3 (Dominguez et al., 1995; He et al., 1995), which are also active as wild-type forms in Xenopus mesodermal patterning assays. A molecular explanation for why XMad is active in overexpression and dominant-negative receptor assays may relate to its level of phosphorylation. Recently, Hoodless et al. (1996) found that the human equivalent of XMad (hxmadr1) is

7 Xenopus mothers against dpp 2365 partially phosphorylated in the absence of BMP signals, and it becomes hyper-phosphorylated upon BMP-2 stimulation. This might explain why XMad is active in embryonic cells lacking BMP signals, and why relatively high amounts of injected XMad mrna are required to induce mesoderm. It will be informative to determine whether region-specific differences in XMad phosphorylation exist in Xenopus embryos, and whether phosphorylation differences correlate with activity. The regional expression of XMad mrna correlates with the location of and tissue responsiveness to BMP-2 and BMP-4. At blastula and gastrula stages animal cap ectoderm and presumptive dorsal mesoderm respond to BMP ligands (Jones et al., 1991; Koster et al., 1991; Dale et al., 1992; Hemmati- Brivanlou and Thomsen, 1995; Wilson and Hemmati- Brivanlou, 1995), and XMad is expressed in these areas. At neurula and tadpole stages BMP-2 (Hemmati-Brivanlou and Thomsen, 1995) and BMP-4 (Fainsod et al., 1994; Hemmati- Brivanlou and Thomsen, 1995; Schmidt et al., 1995) are expressed in isolated regions adjacent to or overlapping with the neurally restricted expression of XMad (Fig. 2). Gradual localization of XMad expression in the embryo may be a mechanism that regulates cell competence to respond to BMP ligands during development, thereby affecting patterning decisions. Mad proteins constitute a new and unique family of intracellular proteins represented in phyla ranging from nematodes through chordates. The high degree of similarity (85%) between XMad and DMad proteins, and their identical activities in mesoderm induction and patterning assays, argue that thedmad and XMad genes are true homologs, consistent with the homology displayed between other components in the Dpp and BMP signaling systems, such as the proteases tolloid/bmp-1, and the dpp/bmp ligands and their receptors. The biochemical properties of Mad-related proteins await a full analysis, as does their precise roles in signaling by particular TGF-β receptors. I thank S. Newfeld and W. Gelbart for the Drosophila Mad cdna clone, J. Graf and D. Melton for the tbr cdna clone, and E. DeRobertis for the stage 9 cdna library. I thank members of my laboratory for sharing reagents and helpful suggestions. I thank R. Padgett and Jeff Wrana for communicating results prior to publication, and I thank P. Wilson, W. Theurkauf, S. Nishimatsu, W. Lennarz, and M. Horb for critical comments on the manuscript. This work was supported by grants to G. H. T. from the National Science Foundation and American Heart Association. REFERENCES Cornell, R. and Kimelman, D. (1994). Activin-mediated mesoderm induction requires FGF. Development 120, Clement, J. H., Fettes, P., Knochel, S., Lef, J. and Knochel, W. (1995). Bone morphogenetic protein 2 in the early development of Xenopus laevis. Mech. Dev. 52, Dale, L. Howes, G. Price, B.M.J. and Smith, J.C. (1992). Bone Morphogenetic Protein 4: a ventralizing factor in Xenopus development. Development 115, Dominguez, I. Itoh, K. and Sokol, S.Y. (1995). Role of glycogen synthase kinase 3 beta as a negative regulator of dorsoventral axis formation in Xenopus embryos. Proc. Natl Acad. Sci. USA 92, Doniach, T. (1995). Basic FGF as an inducer of anteroposterior neural pattern. Cell 1067, Estevez, M. Attisano, L. Wrana, J. Albert, P. Massague, J. and Riddle, D.L. (1993). The daf-4 gene encodes a bone morphogenetic protein receptor controlling C. elegans larval development. Nature 365, Fainsod, A. Steinbeisser, H. and De Robertis, E.M. (1994). On the function of BMP-4 in patterning the marginal zone of the Xenopus embryo. EMBO J. 13, Graf, J. Theis, R. Song, J. Celeste, A. and Melton, D. (1994). Studies with a Xenopus BMP receptor suggest that ventral mesoderm-inducing signals override dorsal signals in vivo. Cell 79, Grunz, H. and Tacke, L. (1989). Neural differentiation of Xenopus laevis ectoderm takes place after dissagregation and delayed reaggregation without inducer. Cell Differ. Dev. 28, Hahn, S.A. Schutte, M. Hoque, A.T.M.S. Moskaluk, C.A. da Costa, L.T. Rozenblum, E. Weinstein, C.L. Fischer, A. Yeo, C.J. Hruban, R.H. and Kern, S.E. (1995). DPC4, a candidate tumor suppressor gene at human chromosome 18q21.1. Science 271, Harland, R.M. (1991). In situ hybridization: an improved whole mount method for Xenopus embryos. Meth. Cell Biol. 36, He, X. Saint-Jeannet, J.P. Woodgett, J.R. Varmus, H.E. and Dawid, I.B. (1995). Glycogen synthase kinase-3 and dorsoventral patterning in Xenopus embryos. Nature 374, Hemmati-Brivanlou, A. and Thomsen, G.H. (1995). Ventral mesodermal patterning in Xenopus embryos: Expression patterns and activities of BMP-2 and BMP-4. Dev. Genet. 17, Hogan, B.L.M. (1995). Upside-down ideas vindicated. Nature 376, Holley, S.A. Jackson, P.D. Sasai, Y. Lu, B. De Robertis, E.M. Hoffmann, F.M. and Ferguson, E.L. (1995). A conserved system for dorsal-ventral patterning in insects and vertebrates involving sog and chordin. Nature 367, Hoodless, P. A., Haerry, T., Abdollah, S., Stapleton, M.O Conner, M. B., Attisano, L. and Wrana, J. L. (1996). MADR1, a Mad-related protein that functions in BMP2 signalling pathways. Cell (in press). Jones, C.M. Kuehn, M.R. Hogan, B.L.M. Smith, J.C. and Wright, C.V.M. (1995). Nodal-related signals induce axial mesoderm and dorsalize mesoderm during gastrulation. Development 121, Jones, C.M. Lyons, K.M. and Hogan, B.L.M. (1991). Involvement of Bone Morphogenetic Protein-4 (BMP-4) and Vgr-1 in morphogenesis and neurogenesis in the mouse. Development 111, Jones, C.M. Lyons, K.M. Lapan, P.M. Wright, C.V.E. and Hogan, B.J.M. (1992). DVR-4 (Bone Morphogenetic Protein-4) as a postero-ventralizing factor in Xenopus mesoderm induction. Development 115, Kao, K.R. and Elinson, R.P. (1989). Dorsalization of mesoderm induction by lithium. Dev. Biol. 132, Kessler, D. and Melton, D. (1994). Vertebrate embryonic induction: Mesodermal and neural patterning. Science 266, Koster, M. Plessow, S. Clement, J.H. Lorenz, A. Tiedemann, H. and Knochel, W. (1991). Bone Morphogenetic Protein 4 (BMP4), a member of the TGF-β family, in early embryos of Xenopus laevis: analysis of mesoderm inducing activity. Mech. Devel. 33, Krieg, P.A. and Melton, D.A. (1987). In vitro RNA synthesis with SP6 RNA polymerase. Meth. Enzymol. 155, LaBonne, C. and Whitman, M. (1994). Mesoderm induction by activin requires FGF-mediated signals. Development 120, Liu, F. Ventura, F. Doody, J. and Massague, J. (1995). Human type II receptor for bone morphogenetic proteins (BMPs): Extension of the twokinase receptor model to the BMPs. Mol. Cell. Biol. 15, Maeno, M. Ong, R.C. Suzuki, A. Ueno, N. and Hung, H.F. (1994). A truncated bone morphogenetic protein 4 receptor alters the fate of ventral mesoderm to dorsal mesoderm: roles for animal pole tissue in the development of ventral mesoderm. Proc Natl Acad Sci USA 91, Massague, J. Attisano, L. and Wrana, J.L. (1994). The TGF-β family and its composite receptors. Trends Cell Biol. 4, Nieuwkoop, P.D. and Faber, J. (1967). Normal Table of Xenopus laevis (Daudin). Amsterdam, North Holland Publishing Company. Nishimatsu, S. Suzuki, A. Shoda, A. Murakami, K. and Ueno, N. (1992). Genes for bone morphogenetic proteins are differentially transcribed in early amphibian embryos. Biochem Biophys. Res. Com. 186, Padgett, R.W. Wozney, J.M. and Gelbart, W.M. (1993). Human BMP sequences can confer normal dorsal-ventral patterning in the Drosophila embryo. Proc. Natl Acad. Sci. USA 90, Perry, M. Thomsen, G.H. and Roeder, R.G. (1985). The genomic organization and nucleotide sequence of two distinct histone gene clusters from Xenopus laevis: Identification of novel, conserved upstream sequence elements. J. Mol. Biol. 185,

8 2366 G. H. Thomsen Raftery, L.A. Twombly, V. Wharton, K. and Gelbart, W.M. (1995). Genetic screens to identify elements of the decapentaplegic signaling pathway in Drosophila. Genetics 139, Richter, K. Good, P.J. and Dawid, I.B. (1990). A developmentally regulated, nervous system specific gene in Xenopus encodes a putitive RNA-binding protein. New Biologist 2, Rupp, R.A. Snider, L. and Weintraub, H. (1994). Xenopus embryos regulate the nuclear localization of XMyoD. Genes Dev. 8, Sambrook, J. Fritsch and Maniatis, T. (1990). Molecular Cloning, a Laboratory Guide. Cold Spring Harbor, Cold Spring Harbor Laboratory. Sanger, F. Nicklen, S. and Coulson, A. (1977). DNA sequencing with chainterminating inhibitors. Proc. Natl Acad. Sci. USA 74, Sasai, B. Lu, H. Steinbeisser, H. Geissert, D. Gont, L.K. and DeRobertis, E.M. (1994). Xenopus chordin, a novel dorsalizing factor activated by organizer-specific homeobox genes. Cell 79, Sasai, Y. Lu, B. Steinbeisser, H. and De Robertis, E.M. (1995). Regulation of neural induction by the Chd and Bmp-4 antagonistic patterning signals in Xenopus. Nature 376, Savage, C. Das, P. Finelli, A.L. Townsend, S. Sun, C.-Y. Baird, S.E. and Padgett, R.W. (1996). The C. elegans sma-2, sma-3, and sma-4 genes define a novel conserved family of TGF-β pathway components. Proc. Natl Acad. Sci. USA 93, Schmidt, J.E. Suzuki, A. Ueno, N. and Kimelman, D. (1995). Localized BMP-4 mediates dorsal/ventral patterning in the early Xenopus embryo. Dev. Biol. 169, Sekelsky, J.J. Newfeld, S.J. Raftery, L.A. Chartoff, E.H. and Gelbart, W.M. (1995). Genetic characterization and cloning of mothers against dpp, a gene required for decapentaplegic function in Drosophila melanogaster. Genetics 139, Smith, W.C., Knecht, A. K., Wu, M. and Harland, R.M. (1993). Secreted noggin protein mimics the Spemann organizer in dorsalizing Xenopus mesoderm. Nature 361, Sokol, S.Y. Klingensmith, J. Perrimon, N. and Itoh, K. (1995). Dorsalizing and neuralizing properties of Xdsh, a maternally expressed Xenopus homolog of dishevelled. Development 121, Suzuki, A. Theis, R.S. Yamaji, N. Song, J.J. Wozney, J. Murakami, K. and Ueno, N. (1994). A truncated BMP receptor affects dorsal-ventral patterning in the early Xenopus embryo. Proc. Natl Acad. Sci. USA 91, ten Dijke, P. Yamashita, H. Sampath, T.K. Reddi, A.H. Estevez, M. Riddle, D.L. Ichijo, H. Heldin, C.H. and Miyazono, K. (1994). Identification of type I receptors for osteogenic protein-1 and bone morphogenetic protein-4. J. Biol. Chem. 269, Thomsen, G.H. and Melton, D.A. (1993). Processed Vg1 protein is an axial mesoderm inducer in Xenopus. Cell 74, Umbhauer, M. Marshall, C.J. Mason, C.S. Old, R.W. and Smith, J.C. (1995). Mesoderm induction in Xenopus caused by activation of MAP kinase. Nature 376, Whitman, M. and Melton, D.A. (1992). Involvement of p21 ras in Xenopus mesoderm induction. Nature 357, Wilson, P.A. and Hemmati-Brivanlou, A. (1995). Induction of epidermis and inhibition of neural fate by BMP-4. Nature 376, Wilson, P.A. and Melton, D.A. (1994). Mesodermal patterning by an inducer gradient depends on secondary cell-cell communication. Curr. Biol. 4, Winnier, G. Blessing, M. Labosky, P.A. and Hogan, B.L.M. (1995). Bone orphogenetic protein-4 (BMP-4) is required for mesoderm formation and patterning in the mouse. Genes Dev. 9, Wrana, J.L. Attisano, L. Wieser, R. Ventura, F. and Massague, J. (1994). Mechanism of activation of the TGF-beta receptor. Nature 370, Xu, R.H. Kim, J. Taira, M. Zhan, S. Sredni, D. and Kung, H.F. (1995). A dominant negative bone morphogenetic protein 4 receptor causes neuralization in Xenopus ectoderm. Biochem. Biophys. Res. Commun. 212, (Accepted 25 May 1996)