Modulation of BMP Activity in Dorsal-Ventral Pattern ... · Modulation of BMP Activity in Dorsal-Ventral Pattern Formation by ... We analyzed the interactions between mutations in

Developmental Biology 245, 109–123 (2002)doi:10.1006/dbio.2002.0614, available online at http://www.idealibrary.com on

Modulation of BMP Activity in Dorsal-VentralPattern Formation by the Chordinand Ogon Antagonists

Daniel S. Wagner and Mary C. Mullins1

Department of Cell and Developmental Biology, University of Pennsylvania Medical School,Philadelphia, Pennsylvania 19104

We analyzed the interactions between mutations in antagonistic BMP pathway signaling components to examine the rolesthat the antagonists play in regulating BMP signaling activity. The dorsalized mutants swirl/bmp2b, snailhouse/bmp7,lost-a-fin/alk8, and mini fin/tolloid were each analyzed in double mutant combinations with the ventralized mutantschordino/chordin and ogon, whose molecular nature is not known. Similar to the BMP antagonist chordino, we found thatthe BMP ligand mutants swirl/bmp2b and snailhouse/bmp7 are also epistatic to the putative BMP pathway antagonist,ogon, excluding a class of intracellular antagonists as candidates for ogon. In ogon;mini fin double mutants, we observed amutual suppression of the ogon and mini fin mutant phenotypes, frequently to a wild type phenotype. Thus, theTolloid/Mini fin metalloprotease that normally cleaves and inhibits Chordin activity is dispensable, when Ogon antagonismis reduced. These results suggest that Ogon encodes a Tolloid and Chordin-independent antagonistic function. By analyzinggenes whose expression is very sensitive to BMP signaling levels, we found that the absence of Ogon or Chordin antagonismdid not increase the BMP activity remaining in swirl/bmp2b or hypomorphic snailhouse/bmp7 mutants. These results,together with other studies, suggest that additional molecules or mechanisms are essential in generating the presumptivegastrula BMP activity gradient that patterns the dorsal–ventral axis. Lastly we observed a striking increased penetrance ofthe swirl/bmp2b dominant dorsalized phenotype, when Chordin function is also absent. Loss of the BMP antagonistChordin is expected to increase BMP signaling levels in a swirl heterozygote, but instead we observed an apparent decreasein BMP signaling levels and a loss of ventral tail tissue. As has been proposed for the fly orthologue of chordin, shortgastrulation, our paradoxical results can be explained by a model whereby Chordin both antagonizes and promotes BMPactivity. © 2002 Elsevier Science (USA)

Key Words: Zebrafish; Danio rerio; dorsal–ventral; BMP; BMP antagonist; Chordin; pattern formation.

INTRODUCTION

According to current models, the range of cell fates alongthe dorsal–ventral (D-V) axis is specified by the gradedaction of a bone morphogenetic protein (BMP) signalingpathway (reviewed in Holley and Ferguson, 1997; Thomsen,1997). Both ectoderm and mesoderm are sensitive to differ-ent levels of BMP signaling (Dosch et al., 1997; Jones andSmith, 1998; Knecht and Harland, 1997; Nguyen et al.,1998; Wilson et al., 1997). For example, the ectoderm isspecified to form neural tissue dorsally in the absence of, orat low, BMP signaling levels (Hawley et al., 1995; Holley et

1 To whom correspondence should be addressed. Fax: 215-898-
9871. E-mail: [email protected].
0012-1606/02 $35.00© 2002 Elsevier Science (USA)All rights reserved.

al., 1995; Wilson and Hemmati-Brivanlou, 1995). Higherlevels specify the laterally derived neural crest and ectoder-mal placodes (Marchant et al., 1998; Nguyen et al., 1998),whereas the highest levels of BMP signaling establishepidermal ectoderm. The level of BMP signaling at anypoint along the D-V axis requires the concerted action ofBMP signaling pathway components.

Three BMP ligands are expressed in the zebrafish gastrulain a pattern consistent with a role in specifying D-V cellfates: bmp2b, bmp7, and bmp4 (Chin et al., 1997; Dick etal., 2000; Martinez-Barbera et al., 1997; Nikaido et al.,1997; Schmid et al., 2000). The action of these moleculescan be modulated by extracellular BMP antagonists, includ-ing Chordin, Noggin, and Follistatin (Hemmati-Brivanlou
et al., 1994; Sasai et al., 1994; Smith and Harland, 1992).
109

The extracellular antagonists decrease signaling potentialby binding the BMP molecules and preventing activation oftheir receptors. The antagonist Chordin is in turn nega-tively regulated by the astacin metalloprotease Tolloid(Blader et al., 1997; Marques et al., 1997; Piccolo et al.,1997), which cleaves Chordin and decreases its affinity forBMP ligands (Larrain et al., 2000). The BMP moleculessignal through members of the Type I and Type II TGF�receptor family. Ligand binding to a receptor complexresults in the phosphorylation of Smad1, 5, and 8 molecules(Kretzschmar et al., 1997), which then translocate to thenucleus and act in transcriptional regulation.

Several genes in the zebrafish have been identified bymutational analysis to be essential for D-V pattern forma-tion (Hammerschmidt et al., 1996; Mullins et al., 1996;Solnica-Krezel et al., 1996). The molecular nature of mostof these genes has been identified and they are componentsof a BMP signaling pathway. Here we examine throughgenetic analysis the interactions between the BMP antago-nist genes chordino and ogon, which display ventralizedmutant phenotypes, and the BMP signaling componentgenes, swirl, snailhouse, lost-a-fin, and mini fin, whichexhibit dorsalized mutant phenotypes.

The ventralized mutants, chordino (din) and ogon (previ-ously also known as mercedes and short tail), display asimilar expansion of ventrally derived gastrula cell fates,such as the ventral tail fin, the posterior somitic mesoderm,blood and pronephros (Fisher et al., 1997; Hammerschmidtet al., 1996; Miller-Bertoglio et al., 1999; Solnica-Krezel etal., 1996). Dorsally derived gastrula cell types are reduced,as reflected in smaller anterior somites and a variablyreduced head and eyes. din encodes the zebrafish chordingene (Schulte-Merker et al., 1997), whereas the molecularidentity of ogon has not been determined. The ogon alleleused in this analysis, ogonm60, is a �2.5 cM deletion and ispresumed to be a null allele (Miller-Bertoglio et al., 1999). Inaddition to the ventralized phenotype, ogonm60 mutantsdisplay a swelling of the midbrain and hindbrain ventricles,which is likely due to the loss of an additional gene in thisdeletion.

swirl and snailhouse (snh) encode the bmp2b and bmp7ligand genes, respectively (Dick et al., 2000; Kishimoto etal., 1997; Nguyen et al., 1998; Schmid et al., 2000). Nullmutant swirl and snh embryos display a strongly dorsalizedphenotype, characterized by the loss of ventral cell fates anda concomitant expansion of dorsal cell types to the mostventral region of the embryo (Mullins et al., 1996; Nguyenet al., 1998; Schmid et al., 2000). These mutant embryos dieby the 14 somite stage. The mutant phenotype of thehypomorphic snhty68a allele is a weaker dorsalization thanthat observed in snh null alleles (Schmid et al., 2000).Anterior neuroectoderm and somitic mesoderm is ex-panded, but does not encircle the embryo, whereas moreposterior aspects of these tissues circumscribe the embryo.Ventral mesodermal derivatives such as blood and prone-phros are reduced or absent.

The mini fin (mfn) and zygotic lost-a-fin (laf) mutants

have dorsalized phenotypes that are restricted to the tail.mfn and zygotic laf mutants display a partial to completeloss of the ventral tail fin and ventral tail vein (Mullins etal., 1996). mfn encodes Tolloid (Connors et al., 1999), ametalloprotease that can cleave Chordin, reducing its abil-ity to bind BMP molecules and resulting in an increase inBMP activity (reviewed in Mullins, 1998). laf encodes aBMP type I receptor, Alk8, which functions both mater-nally and zygotically in dorsal–ventral axis formation(Bauer et al., 2001; Mintzer et al., 2001). The maternal-zygotic laf loss-of-function phenotype resembles swirl andsnh null alleles, indicating that it is a receptor for one orboth of these BMP molecules.

In this study, we examined the ability of the BMPantagonists Chordin and Ogon to modulate BMP signalinglevels through double mutant analysis with BMP pathwaycomponent mutants. In these investigations we observed amutual suppression phenotype in many ogon;mini findouble mutants, which is not observed in chordino;mini findouble mutants, suggestive of independent antagonisticfunctions of Ogon and Chordin on BMP signaling. Weposition the unidentified gene, ogon, relative to the BMPpathway component genes. In addition we found a combi-natorial expansion of the somitic mesoderm in lost-a-fin;ogon and lost-a-fin;chordino double mutants. By gene ex-pression analysis, we found that Chordin and the putativeBMP pathway antagonist, Ogon, do not significantly regu-late the low BMP signaling levels of the presumptive BMPactivity gradient required for neural crest development. Toour surprise, we observed an increased penetrance of theweakly dorsalized swirl/bmp2b dominant phenotype, whenChordin is absent, suggesting an unexpected role for Chor-din in promoting high BMP signaling levels.

MATERIALS AND METHODS

Cross Construction and AnalysisMutant stocks were constructed with the following alleles; mini

finty130a, mini fintm124, swirltc300a, snailhousety68a, lost-a-fintm110b (Mul-lins et al., 1996), chordinott250, ogontm305 (Hammerschmidt et al.,1996), and ogonm60 (Solnica-Krezel et al., 1996). Families wereconstructed by crossing fish heterozygous for one mutation to fishheterozygous for the other, except in the case of mini fin families,which were established with homozygous mini fin founders. Theprogeny were raised and identified as single heterozygotes ordouble heterozygotes of the mutations of interest. The singleheterozygotes were intercrossed to provide controls for the doubleheterozygote crosses. We determined the distribution of mutantphenotypes in the progeny of the double heterozygote crosses.Observed distributions were compared to ideal frequencies thatwould result from specific relationships between the 2 genes. Theideal distributions of wild type: dorsalized: ventralized phenotypesare as follows: dorsal epistasis, 9:4:3; ventral epistasis, 9:3:4;mutual suppression of the mutant phenotypes, 10:3:3. In the casewhere the double mutant embryos have a phenotype distinct fromeither single mutant the ideal distribution is 9:3:3:1 (wild type:dorsalized:ventralized:double). Goodness of fit was determined byChi square analysis. The resulting P value is the probability that

110 Wagner and Mullins

© 2002 Elsevier Science (USA). All rights reserved.

the observed distribution represents an expected hypothetical case.A P value of 0.05 or smaller was considered the limit for exclusionof a hypothetical case.

Determination of Embryo Genotype

dintt250 was originally reported to be a deletion in the zebrafishchordin cDNA (Schulte-Merker et al., 1997). We determined thelesion to be a point mutation in the splice donor site of exon 2, asalso reported by others (Dick et al., 2000). The mutation liesimmediately 3� of the splice donor site and creates an MspI RFLP.The genotype of dintt250 was determined by amplification of theexon2/intron2 boundary with the primers chd3 CGT TTT AGTTGG TGC TCT GAC G and chd4 CAG GCT TAC ACT TTA TGCTTC CG. The genotype of ogonm60 was determined by amplifica-tion of SSR markers that lie within the deleted region, z22653 andz8380 (Miller-Bertoglio et al., 1999). SSR markers that lie adjacentto the deletion, z6924 and z1477, were amplified in parallel as acontrol. swirltc300a was genotyped by amplification of genomic DNAthat contains the lesion with the primers MCM 16, AAA GCTTCG ACG TGG GTT CA, and MCM 14, CCT CCA AAA TAGCTC GCT C. The mutant allele creates a HaeIII RFLP in thisamplified fragment. The genotype of snailhousety68a was deter-mined by a 3� UTR HhaI RFLP tightly linked to the lesion. Thisregion is amplified with the primers BMP 7I, TAG TGA CTG ACATTT CCA ACA C, and BMP 7E, TCC ATG GTG ACA TGA TAGC (Schmid et al., 2000).

In Situ Hybridization and Photography

In situ hybridizations were performed essentially as described(Schulte-Merker et al., 1992). The following probes were used:msxB (Ekker et al., 1997) at the 15-somite stage, dlx3 (Akimenko etal., 1994) at 24 hpf, foxd3 (formerly fkd6, Odenthal and Nusslein-Volhard, 1998), and myoD (Weinberg et al., 1996).

Images of live embryos and those processed by in situ hybridiza-tion were captured from an Axioplan2 compound microscope(Zeiss) by a Progres 3012 digital camera (Kontron Elektroniks) andthe images were processed with Adobe software.

Embryo Sectioning

Embryos were cleared in 1:2 Benzyl alcohol: benzyl benzoate andyolks were removed. Embryos were sectioned by hand (Allende andWeinberg, 1994) between somites. MyoD staining was used as aguide for somite position.

RESULTS

mini fin;ogon double Mutants Display ReciprocalSuppression, Whereas mini fin;chordinoMutants Do Not

We compared the genetic interactions of din and ogonwith each of the BMP pathway components to distinguishpossible differences in their BMP antagonist functions.Tolloid/Mini fin is a metalloprotease that cleaves Chordinreducing its effectiveness at blocking BMP activity (Larrainet al., 2000). mfn;din double mutants exhibited din epista-sis with no alteration of the din phenotype (Figs. 1A–1C,

Table 1). In these crosses, no din mutants displayed gaps inthe VTF (see later results and discussion). This epistasisresult is consistent with the predicted roles of these pro-teins, whereby Tolloid cleaves Chordin to inhibit its func-tion. The molecular nature of Ogon is unknown. If Ogon isa secreted BMP antagonist, susceptible to Tolloid cleavage,then we would expect ogon to also be epistatic to minifin/tolloid. We found that in contrast to mfn;din doublemutants, embryos from crosses between mfn;ogonm60

double heterozygotes did not display clear epistasis. Thethree expected phenotypic classes were observed, as well asan additional mutant phenotype. The frequency of embryosin the expected phenotypic classes was 58.2% wild type,15.6% mfn dorsalized, and 21.8% ogonm60 ventralized (Figs.1A, 1C, and 1D, Table 2). An additional 4.4% of embryosdisplayed the head defect characteristic of ogonm60 mutants,but did not exhibit the strongly ventralized tail phenotypeof ogonm60. The tail phenotype of these embryos rangedfrom small duplications of the ventral tail fin and minorswelling of the ventral tail vein to an apparently wild typetail (Figs. 1E and 1F). This ogon head phenotype with asuppressed tail phenotype was not observed in embryosfrom single ogon heterozygote intercrosses (n � 572). Thusogon behaves differently to chordino in double mutantcombination with tolloid/mini fin.

The lower than expected frequency of mfn and ogonmutants observed (Table 2), suggested that the embryosexhibiting the ogon head phenotype, but lacking the ogontail phenotype, correspond to mfn;ogon double mutants.Embryos with the ogonm60 head phenotype representedapproximately 25% of the embryos from the double hetero-zygote intercrosses (21.8% ogonm60 phenotype and 4.4%ogonm60 head phenotype only), indicating that all ogonhomozygotes can be identified by this phenotype. In thedouble mutants, we hypothesized that homozygosity ofmfn could suppress the tail phenotype of ogon mutants. Todetermine if the embryos with only the ogonm60 headphenotype were ogon;mfn double homozygotes, we geno-typed these embryos for the mfntm124a mutation and for theabsence of an SSR marker that maps within the ogonm60

deletion (Miller-Bertoglio et al., 1999). We found that theseembryos were indeed homozygous for both ogon and mfn(n � 9, data not shown). This suppression of the ogon

TABLE 1mini fin; chordino

Phenotypic class

Wild type mini fin chordino

Observed (5 crosses,n � 1032) 55.6% 18.6% 25.8%

Predicted bychordino epistasisP � 0.847 56.25% 18.75% 25.0%

111Interactions between Zebrafish D-V Mutants


mutant tail phenotype was not fully penetrant; more thantwo-thirds, but not all, of the double mutants expressed thisphenotype. The suppressed mfn; ogonm60 embryos were not

viable, most likely due to lethality of the ogon headnecrotic phenotype, which appears to be unrelated to theventralized phenotype, but associated with the same dele-tion as ogonm60 (Miller-Bertoglio et al., 1999). We obtainedsimilar mutual suppression results with another presump-tive mfn null allele, mfnty130a (Connors et al., 1999), indicat-ing that this was not an allele-specific effect between themfn and ogon mutations (data not shown).

We found that the reciprocal suppression of the mfn andogon phenotypes was fully penetrant in double mutantswith the hypomorphic allele, ogontm305 (Hammerschmidt etal., 1996; Miller-Bertoglio et al., 1999). Homozygousogontm305 embryos display a duplicated ventral tail fin, haveno head defect, and little edema posterior to the yolkextension. In crosses between fish double heterozygous formfn and ogontm305, we observed that both the dorsalized(19.4%) and ventralized (17.5%) classes were less than the25% expected for an epistatic gene, with a correspondingincrease in the wild type class (63.0% versus 56.25%expected for an epistatic gene, Table 2). This distributionfits the expected probability of all the double mutantsdisplaying wild type morphology (P � 0.69). Interestingly,these results demonstrate that the function of Tolloid ininhibiting Chordin function is dispensable, when the nega-tive regulator of BMP signaling, Ogon, is also reduced.

Considering the complete suppression of the ogon tailphenotype in the ogon;mfn double mutants when ogon washypomorphic, rather than null, we investigated whether theincomplete epistasis of the null ogon allele to mfn could beaccounted for by incomplete loss of function of ogon via thereported maternal contribution of ogon (Miller-Bertoglio etal., 1999). Hence, the mutual suppression of ogon and minifin mutant phenotypes could result from residual wildtype Ogon activity present in ogonm60 homozygous em-bryos, due to the presence of maternal ogon gene productfrom ogonm60/� mothers. To test this hypothesis, we exam-ined mfn;ogon double mutants from females deficient inmaternal ogon contribution by generating females of thegenotype ogonm60/ogontm305;mfntm124/� and crossing them toogonm60/�;mfntm124/� males. We found that loss of thematernal ogon contribution did not lessen the mutualsuppression of the ogon and mfn phenotypes. About 17% ofthe ogonm60 mutants were suppressed by mfn (21/120,confirmed by genotypic analysis for mfn and ogon), identi-cal to the 17% we observed in the crosses above withheterozygous ogonm60 females. These results indicate thatresidual maternal Ogon activity cannot account for themutual suppression of the mini fin and ogon phenotypes.Rather the suppression phenotype can be accounted for bythe increased antagonistic function of Chordin occurring inthe absence of Tolloid/Mfn inhibition, which can partiallyor completely compensate for loss of Ogon antagonism.Altogether these results indicate that Ogon antagonismoccurs independently of Tolloid, and acts additively withChordin.

FIG. 1. mini fin (mfn) double mutant analysis. Phenotypes ob-served in embryos from crosses between mfn;din double heterozy-gous fish. (A) In wild type embryos a single ventral tail fin extendsthe full length of the tail (arrowheads). (B) The din mutant displaysmultiple ventral tail fins (arrowhead), an edema posterior to theyolk extension at the site of blood formation (arrow) and a variablyreduced head and eyes. (C) mfn mutant embryos display a gap in theventral tail fin (arrowheads). Additional phenotypes to (A and C)observed in embryos from crosses of mfn;ogon double heterozygousfish (D and E). (D) The ogon mutant is characterized by multipleventral tail fins (black arrowhead), an edema posterior to the yolkextension (arrow) and an altered head morphology (white arrow-head). (E) Embryos exhibiting the ogon head phenotype (whitearrowhead) and a much less severe or wild type tail phenotype(black arrow head). This embryo had only a small edema posteriorto the yolk extension without the pooling of blood or the multi-plicated ventral tail fins seen in D.





Lack of Modulation of Ventral Gastrula BMPActivity by Chordin and Ogon Antagonism

If Swirl/Bmp2b, Snh/Bmp7, and Bmp4 act together duringgastrulation to specify ventral cell identity and Chordin andOgon can antagonize all 3 BMP molecules, then thestrength of the swirl and snh mutant phenotypes may belessened due to loss of Chordin or Ogon, which couldincrease the potency of the remaining BMP molecules.Redundancy in BMP signaling may occur, because all 3 ofthese BMPs can rescue swirl and snh mutants (Dick et al.,2000; Nguyen et al., 1998). Moreover we can detect residual

BMP signaling activity in swirl/bmp2b null mutants(Nguyen et al., 2000; unpublished observations), indicatingthat some BMP activity remains in this mutant that may beantagonized by Chordin. Hence in the absence of Chordinin a swirl mutant, we may see a weakening of the swirlmutant phenotype due to increase in this remaining Bmpactivity. As previously shown, we found that swirl isepistatic to din (Hammerschmidt et al., 1996) and ogon(Miller-Bertoglio et al., 1999) (Table 3 and see resultsbelow). To determine if loss of din or ogon can reduce theseverity of the swirl mutant phenotype, we examined foxd3

FIG. 2. Loss of chordino has no apparent effect on gene expression in swirl (swr) and snh embryos. msxB expression at the 5-somite stage(A–D). (A) Wild type. (B) din embryos exhibit a variable reduction of anterior msxB staining in the presumptive neural crest (nc). (C and D)A similar small range of msxB expression in an anterior ventral position (arrowhead) is observed in embryos from both swirl singleheterozygous and swirl;din double heterozygous crosses: shown here are weak (C) and strong (D) phenotypes. Expression of foxd3 (formerlyfkd6) in the neural crest progenitors in 5-somite stage snh mutants is not affected by loss of din (E–H, lateral views, dorsal to the right; insetsare dorsal-anterior views with anterior to the left). (E) Expression of foxd3 in the prospective cranial neural crest (bracket) and pax2.1 in themid/hindbrain boundary (white box in inset) of a wild type embryo. (F) din mutant embryos exhibit a variable decrease in the width ofanterior neural tissue as seen by decreased spacing between the domains of foxd3 expression and the width of the mid/hindbrain boundary.(G and H) The range of expanded foxd3 expressing neural crest phenotypes observed in snhty68a embryos. The degree of expansion wasunaffected by din genotype; both embryos shown are homozygous mutant for din as well as snhty68a. Classification of the embryos as eitherstrong, moderate or weak snh mutant phenotypes (G is weak, H is strong), followed by examination of the din genotype revealed thefollowing: strong, 4 din/din, 4 din/�; moderate, 3 din/din, 3 din/�, 4 �/�; weak 2 din/din, 3 �/�. In crosses with ogon, 1 of 5 embryoswith a strong snh phenotype was homozygous mutant for ogon. Of 4 embryos with a weak snh phenotype, none was homozygous mutantfor ogon (data not shown).FIG. 3. swirl and snailhouse double mutant analysis. Phenotypes observed in embryos from crosses between swirl; din doubleheterozygous fish (A–F). (A) Wild type embryo at the 5-somite stage (dorsal to the right, anterior up). (B) A swirl mutant embryo at the5-somite stage. Morphogenetic movements have driven the somitic tissue (arrowhead) and the tail bud off the yolk. These mutants die bythe 14-somite stage (� 16 hpf). (C) Wild type embryo at 1 dpf displaying a single ventral tail fin (arrowhead). (D) The weakly dorsalized,dominant swirl phenotype is characterized by a partial loss of the ventral tail fin (absent between the arrowheads). (E) din mutant embryoshave multiple ventral tail fins (arrowhead) and a large edema posterior to the yolk extension (arrow). (F) The dominant dorsalized swirlmutant phenotype is epistatic to din, as evident by the absence of the ventral tail fin (arrowheads) in this din�/�;swr�/� embryo. Howeversome aspects of ventralization are still present, such as a minor edema (arrow) and a multiplicated ventral tail fin tip (arrowhead in the inset,which is a slightly different view). For snh double mutant analysis, 3 classes of 1 dpf phenotypes were observed in embryos from crossesbetween snh; din double heterozygous fish; wild type embryos as in A, din mutant embryos as in E and snh mutant embryos (G). snhembryos are truncated posteriorly and the trunk tissue is twisted above the yolk (arrowhead). Phenotypes observed in embryos from crossesof swirl;ogon double heterozygous fish (H and I). Wild type and swirl mutant embryos were observed identical to those in A, B, and C. (H)ogon mutants at 1 dpf exhibit multiple ventral tail fins (arrowhead) and an edema posterior to the yolk extension (arrow) as well as aswelling of the hindbrain ventricle (white arrowhead). (I) The dominant dorsalized swirl phenotype is epistatic to ogon as evidenced byembryos with a gap in the ventral tail fin (arrow), which also display an ogon head phenotype (white arrowhead) and multiplicated ventraltail fin tip (black arrowhead).

TABLE 2mini fin; ogon

Phenotypic class

Wild type mini fin ogonogon head

Phenotype only

mini fin: ogonm60 observed (14 crosses, n � 835) 58.2% 15.6% 21.8% 4.4%Predicted by combinatorial phenotype 56.25% 18.75% 18.75% 6.25%mini fin: ogontm305 observed (3 crosses, n � 679) 63.0% 19.4% 17.5%Predicted by mutual suppression (mini fin: ogontm305 P � 0.69) 62.5% 18.75% 18.75%



(formerly fkd6) expression in the neural crest precursors(Odenthal and Nusslein-Volhard, 1998) and msxB expres-sion in dorsal neuroectodermal tissue (Akimenko et al.,1995) at the 5-somite stage. The amount of both foxd3 andmsxB expression is exquisitely sensitive to variations inBMP signaling levels (Nguyen et al., 1998; and unpub-lished). From intercrosses of swirl;din and swirl;ogondouble heterozygous fish, we examined 228 and 134 em-bryos, respectively, for msxB and 168 and 152 embryos,respectively, for foxd3 expression. We detected no signifi-cant difference among swirl homozygous mutant embryosfrom double heterozygous crosses that was not observed inembryos from single swirl heterozygote crosses (Figs. 2A–2D, and data not shown). Thus no increase in BMP activitywas detected in swirl mutants due to the loss of Chordin orOgon antagonism.

We examined whether loss of Chordin or Ogon antago-nism could ameliorate the phenotype of the snhty68a hypo-morph, by increasing the activity of Bmp2b, Bmp4, and/orthe remaining Bmp7 activity that is present in snhty68a

mutants. The snhty68a hypomorphic mutant (Fig. 3G) ischaracterized by a reduction in ventral tissues and anexpansion of dorsolateral cell fates to a lesser degree thanthat observed in swirl or snh null mutant embryos, whichexhibit identical dorsalized phenotypes (Schmid et al.,2000). We found that snhty68a is epistatic to din (see alsoDick et al., 2000) and ogon (Table 4). We used the hypo-morphic snh allele as a sensitized background to examinethe ability of the Chordin and Ogon antagonists to modu-late the presumptive BMP gradient that acts during gastru-lation. Because some Bmp7 activity remains in snhty68a

mutants, we expected that loss of the BMP antagonists dinor ogon would partially suppress the snhty68a mutant pheno-type by relieving repression of either residual Bmp7 and/orother BMP activity (e.g. Bmp2b and Bmp4).

We investigated the expression of foxd3 in the prospec-tive cranial neural crest (Odenthal and Nusslein-Volhard,1998), which is very sensitive to changes in BMP activity, insnhty68a;din or snhty68a;ogon double mutants (Figs. 2E–2H,data not shown). The dorsalized embryos of snhty68a single

mutant or snhty68a;din or snhty68a;ogon double mutants dis-played a range of dorsalizations from a weak expansion of thecranial neural crest foxd3 domain (Fig. 2G, data not shown) toa great expansion of the field of foxd3 positive cells thatencompassed the ventral most regions of the embryo (Fig. 2H).We categorized snh mutant embryos from double mutantcrosses according to the degree of foxd3 expansion. The din(n � 27) or ogon (n � 9) genotype was determined for theseembryos to test if loss of Chordin or Ogon antagonism affectedthe strength of the neural crest phenotype. To our surprise, wefound no correlation between the din or ogon genotype andthe strength of the snhty68a mutant phenotype (Figs. 2G and2H, data not shown). Therefore, loss of Chordin or Ogonantagonism does not weaken the phenotype caused by thehypomorphic snhty68a mutation, suggesting that there is noincrease in BMP signaling activity in snhty68a;din or snhty68a;ogon double mutants.

Enhancement of the swirl DominantHaploinsufficient Phenotype When chordino IsAbsent

In examining the ability of Chordin and Ogon antago-nism to regulate BMP signaling activity in swirl;din and

TABLE 3swirl; chordino and swirl; ogon

Phenotypic class

Wild typeWeakly

dorsalized swirl VentralizedVentralized and

weakly dorsalized

swr;din (7 crosses, n � 995) 57.9% 2.1% 22.9% 8.3% 8.7%swr (4 crosses, n � 368) 76.9% 0.0% 23.1%swr;ogo (7 crosses, n � 1206) 54.3% 3.3% 25.7% 14.3% 2.4%swr (6 crosses, n � 393)a 66.3% 4.3% 29.4%

Note. Single mutant crosses are from the same family as the double mutant crosses to control for background due to genetic variation.a The presence of the dominant swirl phenotype in these crosses is likely due to the AB background in which the ogon mutant was

propagated. In contrast the swirl dominant phenotype is rare in the TU background in which the swirl and din stocks were maintained.

TABLE 4snailhouse; chordino and snailhouse;ogon

Phenotypic class

Wild type snailhouse Ventralized

snh;din observed (7crosses, n � 795)

56.3% 24.8% 18.9%

snh;ogo observed (8crosses, n � 1131)

59.2% 23.4% 17.4%

Predicted by dorsalizedepistasis (snh;dinP � 0.99; snh;ogoP � 0.14)

56.25% 25.0% 18.75%



swirl;ogon double mutants, we made an interesting obser-vation. Embryos from crosses of swirl;din double heterozy-gous fish displayed the 3 expected phenotypic classes plus 2

additional classes (Table 3): the expected classes were57.9% wild type, 22.9% swirl strongly dorsalized, and 8.3%din ventralized (Figs. 3A–3C, and 3E); in addition to em-bryos with a weakly dorsalized phenotype (2.1%, Fig. 3D),which may correspond to the dominant swirl phenotype(Mullins et al., 1996), and a combined weakly dorsalizedand ventralized phenotype (8.74%, Fig. 3F). Intercrosses ofswirl;ogon double heterozygotes also displayed the ex-pected phenotypic classes, as well as the dominant swirland the combined dorsalized and ventralized phenotypes(Table 3, Figs. 3H and 3I). The combined mutant classdisplayed a weakly dorsalized phenotype, a partial loss ofthe ventral tail fin, as well as a ventralized phenotype,expansion of blood or multiplication of the remainingventral tail fin.

In the swirl;din crosses, we genotyped weakly dorsalizedembryos, including those with the combined ventralizedphenotype, for the swirl mutation and found that they wereall swirl heterozygotes (n � 23). We then genotyped theseswirl heterozygotes for din and found that five of six weaklydorsalized mutants and all of the combined dorsalized/ventralized mutants (n � 17) were homozygous for din,demonstrating that the dominant swirl phenotype is epi-static to the din phenotype. We similarly found from theswirl;ogon crosses that the combined dorsalized/ventralized mutants were both heterozygous for swirl andhomozygous for ogon (n � 5). We conclude that, as withdin, the dominant and recessive swirl phenotypes are epi-static to ogon. In the swirl;din intercrosses, the high frac-tion of din homozygotes among the swirl dominant mu-tants is surprising; 5 or 6 din homozygotes, correspondingto 25% of the swirl heterozygotes, are expected, rather thanthe 22 observed. Similar ratios of mutant phenotypes havebeen observed using different alleles of swirl and din(Hammerschmidt and Mullins, 2002), indicating that thiseffect is not specific to the alleles used here or just astatistical aberration (P � 9.8 � 10�13). Furthermore, nodominant swirl phenotypes were observed in sibling crosseswhere din was wild type (Table 3). Intriguingly, theseresults reveal an increased penetrance of the swirl domi-nant phenotype, when Chordin is deficient.

This enhancement of the swirl dominant phenotypesuggests that Chordin may play a role in potentiating BMPsignals. Further evidence for Chordin enhancing BMP sig-nals is the loss in ventral tail fin tissue that is observed insome din mutants (Table 3, Fig. 4). This phenotype re-sembles the din/din; swr/� embryos that display a gap inpart of the ventral tail fin, while the remainder of the fin ismultiplicated. We investigated the rate and the nature ofthis apparent gap in 2 alleles of din, dintt250, and dinm52.Embryos were scored at 1dpf and the presence of true gapsin the ventral fin was confirmed at 2 dpf. In our experienceapproximately 50% of embryos with ventral tail fin gaps at1 dpf had recognizable fin along the length of the tail at 2dpf, which may represent reductions rather than absense offin tissue. True gaps were observed in an average of 7% ofdintt250 homozygous mutant embryos (median is 2%, n �

FIG. 4. Loss of ventral tail fin in din mutants. (A) Wild type embryos at1dpf have a single ventral tail fin (arrows in ventral view inset). (B) Astrong dintt250 mutant embryo at 1 dpf displays multiple ventral tail fins(white arrowheads). (C) Occasional dintt250 embryos display gaps in themultiplicated ventral fin (gap, black arrowheads; ventral fins, whitearrowheads). dinm52 embryos also display alterations in the fin struc-ture. (D) Most dinm52 embryos display a full ventral tail fin with minorduplications (white arrowheads) (E) Some embryos display reducedfins (black arrowhead) or gaps, in addition to duplications (whitearrowheads). All panels lateral views, ventral view in inset.



FIG. 5. lost-a-fin double mutant analysis. Embryonic phenotypes observed from crosses between laf; din double heterozygous fish (A–C).(A) A wild type embryo at 1 dpf. (B) A typical laf mutant embryo, which lacks the ventral tail fin (arrowheads). din mutant embryos weresimilar to the embryo in Fig. 3E. (C) A double mutant embryo displaying a sharp dorsal flexure of the tail, no ventral tail fin (arrowheads)and a mass at the posterior end of the yolk extension (arrow). (D) A laf;ogon double mutant exhibiting a sharp dorsal bend in the tail, noventral tail fin (arrowheads) and a cell mass posterior to the yolk extension (arrowhead). The degree of dorsal flexure and the presence ofthe cell mass are variable compared to the laf; din embryos. Expression of myoD at the 20-somite stage in embryos from laf;din (E–H) andlaf;ogon (I–L) intercrosses. Transverse sections of similar embryos approximately at the level of somite 16 (arrows) are shown as insets(E–H). (E and I) myoD expression in the somitic mesoderm of a wild type embryo. The somite staining does not extend toward the ventralside of the embryo. In contrast to the wild type (E and I), in laf mutant embryos (F) myoD expression in the tail somites extends ventrallybeginning at approximately somite 17 (arrowhead). (G) din mutants also exhibited a ventral expansion of myoD expression, but in moreposterior somites (arrowhead). (H) The double mutant embryos displayed a characteristic dorsal bend in the tail and ventral expansion ofmyoD expression at a more anterior somite than in either single mutant (somite 12, arrowhead). (J) laf embryos from laf; ogon crossesdisplayed a ventral expansion of myoD expression beginning at about somite 19. (K) ogon mutant embryos exhibit a less dramatic expansionof myoD expression in posterior somites at this stage. (L) As with din crosses the laf;ogon double mutants embryos exhibited a ventralexpansion of somitic mesoderm at a more anterior position, beginning at about somite 15 (arrowhead).FIG. 6. Model for ventral tail fin specification in chordino and swirl mutants. The increased frequency of loss of the ventral tail fin indin/din; swirl/� embryos can be explained by a combinatorial decrease in the maximum level of BMP signaling within the tail bud. Indin/din (blue box) or swr/� (red box) embryos, the range of maximum BMP signaling activity is lower than wild type (black box); however,most embryos are above the threshold required to specify the ventral tail fin (dashed line). A combined reduction in maximal BMP signalinglevels is then found in din/din; swr/� embryos (purple box). This lower range of maximal BMP signaling results in a much larger fractionof embryos that are below the level necessary for ventral tail fin specification. The incomplete penetrance of the ventral tail fin phenotypemay reflect differences in maximal BMP signaling levels between individuals, possibly due to genetic background differences or otherunknown regulatory processes.



456, Fig. 4C). Mutants of the weak allele, dinm52 (Solnica-Krezel et al., 1996), display duplications of portions of theventral tail fin, often combined with an opposite reductionin more proximal regions of the ventral tail fin (Figs. 4D and4E), with 2.6% of dinm52 embryos displaying true gaps (n �272, not shown). The presence of embryos displaying thisgap phenotype in both of these mutant alleles indicates thatthe phenotype is not allele specific, and likely reflects afundamental role for Chordin in promoting ventral tail finfate. The low penetrance of this phenotype is likely due toa decrease in BMP signaling levels in din mutants that liesat or just above the threshold required for ventral finspecification (see Discussion). As might be expected for asystem functioning close to threshold levels, the pen-etrance of this phenotype was variable among independentcrosses ranging from 0% (e.g., in the mfn;din crosses) to30% and may reflect a sensitivity to genetic background.

Analysis of lost-a-fin; chordino and lost-a-fin;ogon double Mutants

We investigated the interactions between mutations in thelost-a-fin/alk8 (laf) gene and the din and ogon loci in doublemutant analysis. The laf zygotic mutant phenotype is charac-terized by a loss of the ventral tail fin and ventral tail vein. Incontrast, din and ogon mutants exhibit multiple ventral tailfins, as well as an apparent expansion of ventral tail veintissue in the edema posterior to the yolk extension. Analysisof embryos from intercrosses of laf; din and laf; ogon doubleheterozygous adults, respectively, revealed four distinct phe-notypic classes: wild type, laf dorsalized, ventralized and acombined phenotype with both dorsalized and ventralizedcharacteristics in a ratio of 9:3:3:1 (Figs. 3E, 3H, 5A–5D, Table5). The frequency of the combined phenotypic class is consis-tent with it representing the double mutant embryos. Thesedouble mutant embryos displayed a dilated yolk extension, acell mass posterior to the yolk extension as seen in din andogon mutants, as well as a sharp dorsal bend in the tail, aphenotype not observed in the single mutants. Although thedouble mutant embryos exhibited unique features, they dis-played an absence of the ventral tail fin, the laf mutantphenotype. Thus, laf is epistatic to din and ogon with respectto ventral tail development, consistent with Laf, the type IBMP receptor Alk8, acting downstream of Chordin and Ogon.

Morphological and molecular marker analysis of the

laf;din and laf;ogon double mutants indicates a uniqueeffect on the somitic mesoderm compared to the singlemutants. Somites in the tail of din, ogon and laf singlemutants are expanded ventrally. Analysis of myoD expres-sion at the 20-somite stage revealed that laf;din and laf;ogon double mutants exhibit a ventral expansion of somitictissue at a more anterior position than in the single mutants(Figs. 5E–5L). In laf mutants, the somites encircle the tailbeginning at about somite 17–19 (Fig. 5F). At this level indin mutants, the somites are rarely expanded to the sameextent as in laf mutants (Fig. 5G). In contrast to eithersingle mutant, laf;din double mutants display a dramaticventral expansion of somitic mesoderm beginning at aboutsomite 12 (Fig. 5H). Transverse serial sections confirm theexpansion of myoD expression in more anterior somites inthe double mutants compared to either single mutant(insets in Figs. 5E–5H). Similar results, although to a lesserdegree, were found in the laf;ogon double mutants (Figs.5J–5L). This greater expansion of somitic mesoderm likelyreflects an additive effect of each single mutant on thistissue in the double mutant condition.

DISCUSSION

Mini Fin/Tolloid Is Dispensable When OgonAntagonism Is Reduced

Our analysis revealed that ogon acts similarly to din inmost genetic relationships we assayed, with one exception.As with din, we found that swirl/bmp2b and snh/bmp7 areepistatic to ogon, consistent with Ogon acting upstream ofSwirl/Bmp2b and Snh/Bmp7 as a BMP antagonist, as sug-gested previously (Miller-Bertoglio et al., 1999). However,unlike din, we found that ogon was not clearly epistatic tomfn/tolloid. The majority of mfn;ogon double mutantsdisplayed a mutual suppression of their respective tailmutant phenotypes. Since a fraction of mfn;ogon doublehomozygous embryos display the ogon phenotype, mfn isclearly not epistatic to ogon and therefore, mfn does notfunction downstream of ogon.

The action of Ogon in parallel to Chordin as a BMPantagonist, which is not negatively regulated by Tolloid/Mfn activity, could account for the mutual suppression ofthe ogon and mfn mutant phenotypes. If Chordin and Ogonact independently to antagonize BMP signaling, then ogon;

TABLE 5lost-a-fin; chordino and lost-a-fin; ogon

Phenotypic class

Wild type lost-a-fin Ventralized Combined

laf;din observed (4 crosses, n � 1053) 56.4% 19.6% 18.0% 6.0%laf;ogo observed (7 crosses, n � 720) 56.9% 19.2% 18.3% 5.6%Predicted by additive phenotype

(laf;din P � 0.88; laf;ogo P � 0.86) 56.25% 18.75% 18.75% 6.25%



mfn double mutants will have increased Chordin dorsaliz-ing activity through the loss of Mfn/Tolloid inhibition,resulting in a reduction of BMP signaling activity. Thisreduction may be compensated for by an increase in BMPsignaling caused by loss of Ogon antagonism. Compensa-tion for loss of Ogon by increased Chordin activity has beenpreviously demonstrated by the rescue of ogon mutantembryos by chordin overexpression (Miller-Bertoglio et al.,1999). Thus, Ogon could encode a secreted BMP antagonistor a Bmp2b/Bmp7 inducible antagonist, such as the nega-tive regulatory Smads, Smad6, and Smad7, or the BAMBI/Eyebrow pseudoreceptor (Nakayama et al., 1998; Onicht-chouk et al., 1999; Tsuneizumi et al., 1997). Candidategenes acting similarly to the Drosophila gene brinker,which negatively regulates Dpp transcriptional targets inthe absence of Dpp signaling (Jazwinska et al., 1999; Mi-nami et al., 1999), can be excluded. A brinker-like genewould be epistatic to the swirl/bmp2b and snh/bmp7 genes,whereas we find these genes to be epistatic to ogon. Anumber of BMP antagonists have been excluded as candi-dates for ogon, including follistatin, three noggin homologs(Bauer et al., 1998; Furthauer et al., 1999; Miller-Bertoglioet al., 1999), smad7, and bambi/eyebrow (Miller-Bertoglio,2000); however, other antagonists remain to be investi-gated. Interestingly, our results show that Tolloid/Mfninhibition of Chordin is dispensable for dorsal–ventral axisformation, if Ogon antagonism is reduced. Thus, theseresults illustrate the role of Tolloid/Mfn in finely regulatingBMP activity levels through modulation of BMP antago-nism.

Combinatorial Effects between lost-a-fin andogon and chordino in the Somitic Mesoderm

We found that laf is epistatic to ogon and chordino inspecification of the ventral tail fin, although double mutantembryos also display a combinatorial phenotype in morelaterally-derived cell fates. The epistasis results suggestthat laf acts downstream of ogon and chordino in a BMPsignaling pathway, consistent with the recent identificationof the laf gene as a type I BMP receptor (Bauer et al., 2001;Mintzer et al., 2001). However, unlike the other dorsalizedmutants, the somitic mesoderm was expanded to a greaterdegree in laf; din and laf; ogon double mutants than in thesingle mutants. The combinatorial phenotype observed inthe double mutant embryos is distinct from that observedin moderately dorsalized mutants such as piggytail (Mul-lins et al., 1996) and likely reflects an additive effect of eachsingle mutant on this lateral gastrula cell fate, the posteriorsomitic mesoderm. In single, zygotic laf mutant embryosand also in chordino and ogon mutants the posteriorsomites are enlarged. Thus, the double mutant conditionmay result in a combinatorial increase in posterior pre-somitic mesoderm.

Why does both a dorsalized and ventralized mutant showan expansion of the same tissue, the posterior somiticmesoderm? This may reflect a local effect on the fate map

in the hypomorphic laf/alk8 mutant caused by loss of onlyzygotic, but not maternal Laf/Alk8 activity. In contrast inthe din and ogon mutants, a more global effect on the fatemap is observed. Thus, the enlarged posterior somites mayarise by different mechanisms. In the zygotic, dorsalized lafmutant, tail cell fates are specifically affected, whereasmore anterior trunk tissues appear normal. In these mu-tants, ventral tail fates are absent and appear to be trans-formed into more dorsal tail fates, namely posterior somiticmesoderm (Mullins et al., 1996). In contrast in the ventral-ized mutants, din and ogon, both trunk and tail tissues areaffected. In these mutants, dorsal gastrula fates, includingprospective head and trunk cell fates, are reduced andtransformed into more ventral gastrula fates, which in-cludes prospective posterior somitic mesoderm (Fisher etal., 1997; Hammerschmidt et al., 1996; Miller-Bertoglio etal., 1999). Thus, during gastrulation in the double mutants,when presumptive Laf/Alk8 activity is present, the loss ofChordin or Ogon activity results in a ventralization of thegastrula, generating more ventral cell fates, which includesthe posterior somitic mesoderm. During tail patterningstages, Laf/Alk8 activity is presumed absent and causes aloss of the enlarged ventral tail fin tissue of the ventralizedmutants due to laf/alk8 epistasis. This ventral tail tissue isthen converted into tail somitic mesoderm, which wasadditionally expanded due to loss of ogon or din duringgastrulation. Hence the double mutants display a moreseverely expanded somitic mesoderm phenotype than thesingle mutants. Alternatively, or in addition, it is possiblethat Chordin and Ogon play a different role in patterningposterior as opposed to anterior tissues. We have presentedevidence for Chordin promoting BMP activity within thetail bud during tail patterning. The similar effect of thezygotic laf mutant and the ventralized mutants on theposterior somites may reflect positive functions in BMPsignaling for both Laf/Alk8 and Chordin and Ogon (see alsodiscussion below). Further studies are necessary to establishthe roles of Chordin and Ogon in patterning posteriortissues.

Loss of Chordin or Ogon Antagonism Does NotIncrease BMP Signaling in swirl or Hypomorphicsnailhouse mutants

Our analysis indicates that Chordin and Ogon do notsignificantly modulate the gastrula BMP activity that re-mains in swirl and hypomorphic snh mutants. Three BMPmolecules, bmp2b, bmp4, and bmp7, are expressed in theearly zebrafish gastrula. The expression of all 3 BMP ligandsis initiated, but not maintained in the absence of swirl/bmp2b or snh/bmp7 (Dick et al., 2000; Kishimoto et al.,1997; Nguyen et al., 1998; Schmid et al., 2000). Althoughchordin expression is greatly expanded in swirl/bmp2bmutants (Miller-Bertoglio et al., 1997), our epistasis resultsshow that this expansion is not required for the dorsalizedphenotype. However, we hypothesized that Chordin maystill antagonize the residual Bmp2b, Bmp4, and Bmp7



activity in hypomorphic snh/bmp7 mutants, and Bmp7 andBmp4 activity in swirl/bmp2b mutants. If so, then loss ofChordin in snh;din or swirl;din double mutants may reducethe severity of the BMP mutant phenotype by increasingthe activity of the remaining BMP molecules. We examinedthe prospective neural crest and dorsal neuroectoderm,which is particularly sensitive to changes in BMP signalinglevels, and found that loss of the BMP antagonist Chordinor Ogon had no effect on the amount of these tissues in thedouble mutants. These results indicate that loss of Chordinor Ogon does not substantially increase the BMP signalingactivity that remains in these mutants.

The lack of a detectable change in BMP signaling in theswirl/bmp2b mutant due to loss of Chordin is consistentwith our previous analysis of swirl;snh double mutants. Wefound that presumptive null mutants of swirl or snh orswirl;snh double mutants all display identical mutant phe-notypes, indicating that the Bmp2b/Swirl and Bmp7/Snailhouse ligands have equivalent, nonredundant func-tions in early embryogenesis (Schmid et al., 2000). Becauseeach ligand is independently required, in a swirl;din doublemutant, an increase in Snh/Bmp7 activity due to loss ofChordino would not weaken the mutant phenotype due tothe independent requirement for Swirl/Bmp2b activity forventral cell fate specification.

We expected a modification of the hypomorphic snhty68a

mutant phenotype by loss of Chordin or Ogon antagonism,since some Bmp7 activity, together with Bmp2b and Bmp4activity still remain in this mutant. We can exclude thepossibility that the snhty68a mutation itself inhibits interac-tion with Chordin or Ogon, rendering it insensitive toChordin or Ogon activity, since the snhty68a mutation lies inthe pro-domain and leaves the ligand domain intact, whichhas been shown to interact with Chordin (Piccolo et al.,1996). Possible explanations for the lack of a modificationof the snh hypomorphic phenotype by loss of Chordin orOgon, include that Chordin and Ogon may antagonizeanother BMP signal transduction pathway (e.g. Bmp2b/Bmp4) and not Snailhouse/Bmp7. Thus, any increase inBMP signaling in the double mutants would be limited bythe signaling potential of the snhty68a hypomorph. Alterna-tively, negative regulation of BMP signaling by both Ogonand Chordino may have a quantitative threshold, such thatthey do not negatively regulate BMP signaling at the BMPlevels present in homozygous snhty68a embryos. Hence lossof the negative regulator has no impact on the dorsalizedphenotype.

Lastly, it is important to note that a BMP signalinggradient still forms in snhty68a mutants and snh; din and snh;ogon mutants, as revealed by the presence of the neuralcrest and more ventral cell fates that form in these mutants.These results raise the question of the extent to whichChordin or Ogon normally act in generating the BMPgradient. In din or ogon single mutants, or din; ogon doublemutants a BMP gradient also stills forms, as indicated bythe presence of all cell fates along the DV axis in thesemutants (Fisher et al., 1997; Hammerschmidt et al., 1996;

Miller-Bertoglio et al., 1999). The relative sizes of domainsalong the DV axis are altered in these ventralized mutants,with a reduction in dorsal and an expansion in ventraldomains, indicating that both Chordin and Ogon are impor-tant in shaping the BMP gradient. However, since a gradientstill forms in the absence of these antagonists, other mol-ecules or mechanisms must remain in the single and doublemutants to generate a BMP activity gradient, which mayalso be acting in the snh; din and snh; ogon double mutantsexamined here.

Paradoxical Enhancement of the swirl DominantPhenotype by Loss of Chordin

In swirl;din double mutant analysis, we observed anunexpectedly high frequency of din homozygotes amongthe weakly dorsalized swirl heterozygous mutants.Twenty-two of 23 embryos with a dominant swirl pheno-typewere din homozygotes, rather than the expected 25%or 5–6) of the swirl heterozygotes. Moreover, in siblingcontrol crosses where chordin is wild type, no dominantswirl mutants were observed (see Table 3). Thus, loss ofChordin antagonism greatly increases the penetrance of theswirl haploinsufficient phenotype. This effect is opposite towhat may be expected. Loss of Chordin antagonism isexpected to increase BMP signaling activity. In the swirl/bmp2b heterozygote, BMP signaling is reduced and anadditional loss of Chordin antagonism would be expected toincrease BMP signaling and therefore suppress the swirldominant tail phenotype. Instead we observed an enhance-ment of the weakly dorsalized swirl/bmp2b haploinsuffi-cient phenotype.

One possible interpretation of these results is that loss ofchordin causes a reduction in BMP signaling, rather thanthe expected increase in BMP signaling during tail pattern-ing. The ventral tail fin may be the most sensitive to loss ofBmp2b signaling levels, since this is the primary tissueaffected in swirl heterozygous mutants, as well as sbn/smad5 heterozygous, zygotic mutants (Mullins et al., 1996).Hence, one could argue that specification of the ventral tailfin requires the highest BMP activity levels. Thus Chordinmay be required to achieve the highest BMP signalinglevels. This is reminiscent of the role of the fly orthologueof Chordin, Short gastrulation (Sog), in dorsal–ventral pat-terning of the fly embryo. The amnioserosa, the most dorsalcell fate in the fly embryo, requires the highest BMPactivity levels and also requires Sog activity (reviewed inMullins, 1998). In sog mutants the amnioserosa is stronglyreduced or absent. The mechanism by which a BMP antag-onist acts also as an agonist to promote BMP signaling isstill unclear, although some evidence suggests that theTolloid metalloprotease acts together with Sog and anothersecreted factor, Twisted Gastrulation, to promote high BMPlevels (Chang et al., 2001; Oelgeschlager et al., 2000; Ross etal., 2001; Scott et al., 2001). Recent work has shown thatvertebrate Chordin does not have an agonist function in aDrosophila assay, where Sog does (Decotto and Ferguson,



2001), indicating that not all components regulating thissystem are closely conserved.

We have previously made an analogy between patterningof the zebrafish tail bud and patterning of the dorsal–ventralaxis of the fly embryo, based on the expression patterns ofbmp4 and chordin and analysis of Tolloid/Mfn function inthe zebrafish tail bud (Connors et al., 1999). This model oftail patterning during post–gastrula stages is distinct frommodels of D-V patterning in gastrula stages, since Tolloidappears to modify Chordin activity only during postgastrulastages in the zebrafish. It is possible that this analogy to flyD-V patterning may also extend to the function of Sog andChordin in promoting high BMP signaling levels; that isChordin may act as a BMP agonist and antagonist duringpostgastrula stages in zebrafish tail patterning as does Sogin fly D-V patterning. However, unlike the sog mutant, thezebrafish din mutant does not typically exhibit a loss of theventral tail fin (VTF), which would be expected if Chordin isrequired to generate the highest levels of BMP activity. Wehave observed an incompletely penetrant loss of portions ofthe VTF in din mutants (Fig. 4). Hence, it is possible thatChordin can play a similar role in increasing BMP activityin the zebrafish tail bud; however, this activity may not beas critical in zebrafish development, as in fly embryonicdevelopment. In din mutants the highest BMP activitylevels may be reduced within the tail bud, although theselevels may still be high enough in most din mutants forVTF specification (see Fig. 6). In swirl heterozygotes, thehighest BMP activity levels may be reduced below thatfound in din mutants and be close to the threshold for VTFspecification. In the combined swirl heterozygous, dinhomozygous state, now the highest BMP activity levelsmay be reduced below the threshold and consequently theVTF is not specified in many embryos (see Fig. 6). Hence,the penetrance of the swirl dominant phenotype is in-creased when Chordin is also absent. Further investigationsmust be performed to test this hypothesis and determine if,and the significance of, Chordin acting as a BMP agonist innormal zebrafish development.

ACKNOWLEDGMENTS

We are grateful to S. Fisher, V. Miller-Bertoglio, and M. Halpernfor the BAC clone and other reagents that allowed us to identify themutation in dintt250 mutants. We thank M. Hammerschmidt forcommunicating unpublished results. The authors would like toacknowledge D. Cobb, W. Kilgore, C. Miller, H. Robinson-Bey forcare of the zebrafish. J. Song for technical support. S. Connors, M.Granato, K. Mintzer, G. Runke, and A. Weimelt for critical reviewand comments on the manuscript during preparation. This workwas supported by a University of Pennsylvania Research Grant andNational Institutes of Health grant R01-GM56326 to MCM andtraining program grant HD07516 and NRSA 1 F32 GM19803-01 toDSW.

REFERENCES

Akimenko, M. A., Ekker, M., Wegner, J., Lin, W., and Westerfield,M. (1994). Combinatorial expression of three zebrafish genesrelated to distal-less: part of a homeobox gene code for the head.J. Neurosci. 14, 3475–3486.

Akimenko, M. A., Johnson, S. L., Westerfield, M., and Ekker, M.(1995). Differential induction of four msx homeobox genes dur-ing fin development and regeneration in zebrafish. Development121, 347–357.

Allende, M. L., and Weinberg, E. S. (1994). The expression patternof two zebrafish achaete-scute homolog (ash) genes is altered inthe embryonic brain of the cyclops mutant. Dev. Biol. 166,509–530.

Bauer, H., Lele, Z., Rauch, G. J., Geisler, R., and Hammerschmidt,M. (2001). The type I serine/threonine kinase receptor Alk8/Lost-a-fin is required for Bmp2b/7 signal transduction during dorso-ventral patterning of the zebrafish embryo. Development 128,849–858.

Bauer, H., Meier, A., Hild, M., Stachel, S., Economides, A., Ha-zelett, D., Harland, R. M., and Hammerschmidt, M. (1998).Follistatin and noggin are excluded from the zebrafish organizer.Dev. Biol. 204, 488–507.

Blader, P., Rastegar, S., Fischer, N., and Strahle, U. (1997). Cleavageof the BMP-4 antagonist chordin by zebrafish tolloid. Science278, 1937–1940.

Chang, C., Holtzman, D. A., Chau, S., Chickering, T., Woolf, E. A.,Holmgren, L. M., Bodorova, J., Gearing, D. P., Holmes, W. E., andBrivanlou, A. H. (2001). Twisted gastrulation can function as aBMP antagonist. Nature 410, 483–487.

Chin, A. J., Chen, J.-N., and Weinberg, E. S. (1997). Bone morpho-genetic portein-4 expression characterizes inductive boundariesin organs of developing zebrafish. Dev. Genes Evol. 207, 107–114.

Connors, S. A., Trout, J., Ekker, M., and Mullins, M. C. (1999). Therole of tolloid/mini fin in dorsoventral pattern formation of thezebrafish embryo. Development 126, 3119–3130.

Decotto, E., and Ferguson, E. L. (2001). A positive role for Shortgastrulation in modulating BMP signaling during dorsoventralpatterning in the Drosophila embryo. Development 128, 3831–3841.

Dick, A., Hild, M., Bauer, H., Imai, Y., Maifeld, H., Schier, A. F.,Talbot, W. S., Bouwmeester, T., and Hammerschmidt, M. (2000).Essential role of Bmp7 (snailhouse) and its prodomain in dorso-ventral patterning of the zebrafish embryo. Development 127,343–354.

Dosch, R., Gawantka, V., Delius, H., Blumenstock, C., and Niehrs,C. (1997). Bmp-4 acts as a morphogen in dorsoventral mesodermpatterning in Xenopus. Development 124, 2325–2334.

Ekker, M., Akimenko, M. A., Allende, M. L., Smith, R., Drouin, G.,Langille, R. M., Weinberg, E. S., and Westerfield, M. (1997).Relationships among msx gene structure and function in ze-brafish and other vertebrates. Mol. Biol. Evol. 14, 1008–1022.

Fisher, S., Amacher, S. L., and Halpern, M. E. (1997). Loss ofcerebum function ventralizes the zebrafish embryo. Develop-ment 124, 1301–1311.

Furthauer, M., Thisse, B., and Thisse, C. (1999). Three differentnoggin genes antagonize the activity of bone morphogeneticproteins in the zebrafish embryo. Dev Biol 214, 181–196.

Hammerschmidt, M., and Mullins, M. C. (2002). Dorsoventralpatterning in the zebrafish: Bone morphogenetic proteins andbeyond. In “Pattern Formation in Zebrafish” (L. Solnica-Krezel,Ed.), Vol. in press. Springer-Verlag, Heidelberg, Germany.



Hammerschmidt, M., Pelegri, F., Mullins, M. C., Kane, D. A., vanEeden, F. J., Granato, M., Brand, M., Furutani-Seiki, M., Haffter,P., Heisenberg, C. P., Jiang, Y. J., Kelsh, R. N., Odenthal, J.,Warga, R. M., and Nusslein-Volhard, C. (1996). dino and mer-cedes, two genes regulating dorsal development in the zebrafishembryo. Development 123, 95–102.

Hawley, S. H., Wunnenberg-Stapleton, K., Hashimoto, C., Laurent,M. N., Watabe, T., Blumberg, B. W., and Cho, K. W. (1995).Disruption of BMP signals in embryonic Xenopus ectoderm leadsto direct neural induction. Genes Dev. 9, 2923–2935.

Hemmati-Brivanlou, A., Kelly, O. G., and Melton, D. A. (1994).Follistatin, an antagonist of activin, is expressed in the Spemannorganizer and displays direct neuralizing activity. Cell 77, 283–295.

Holley, S. A., and Ferguson, E. L. (1997). Fish are like flies are likefrogs: conservation of dorsal-ventral patterning mechanisms.Bioessays 19, 281–284.

Holley, S. A., Jackson, P. D., Sasai, Y., Lu, B., De Robertis, E. M.,Hoffmann, F. M., and Ferguson, E. L. (1995). A conserved systemfor dorsal-ventral patterning in insects and vertebrates involvingsog and chordin. Nature 376, 249–253.

Jazwinska, A., Kirov, N., Wieschaus, E., Roth, S., and Rushlow, C.(1999). The Drosophila gene brinker reveals a novel mechanismof Dpp target gene regulation. Cell 96, 563–573.

Jones, C. M., and Smith, J. C. (1998). Establishment of a BMP-4morphogen gradient by long-range inhibition. Dev. Biol. 194,12–17.

Kishimoto, Y., Lee, K. H., Zon, L., Hammerschmidt, M., andSchulte-Merker, S. (1997). The molecular nature of zebrafishswirl: BMP2 function is essential during early dorsoventralpatterning. Development 124, 4457–4466.

Knecht, A. K., and Harland, R. M. (1997). Mechanisms of dorsal-ventral patterning in noggin-induced neural tissue. Development124, 2477–2488.

Kretzschmar, M., Liu, F., Hata, A., Doody, J., and Massague, J.(1997). The TGF-beta family mediator Smad1 is phosphorylateddirectly and activated functionally by the BMP receptor kinase.Genes Dev 11, 984–995.

Larrain, J., Bachiller, D., Lu, B., Agius, E., Piccolo, S., and DeRobertis, E. M. (2000). BMP-binding modules in chordin: Amodel for signalling regulation in the extracellular space. Devel-opment 127, 821–830.

Marchant, L., Linker, C., Ruiz, P., Guerrero, N., and Mayor, R.(1998). The inductive properties of mesoderm suggest that theneural crest cells are specified by a BMP gradient. Dev. Biol. 198,319–329.

Marques, G., Musacchio, M., Shimell, M. J., Wunnenberg-Stapleton, K., Cho, K. W., and O’Connor, M. B. (1997). Produc-tion of a DPP activity gradient in the early Drosophila embryothrough the opposing actions of the SOG and TLD proteins. Cell91, 417–426.

Martinez-Barbera, J. P., Toresson, H., Da Rocha, S., and Krauss, S.(1997). Cloning and expression of three members of the zebrafishBmp family: Bmp2a, Bmp2b and Bmp4. Gene 198, 53–59.

Miller-Bertoglio, V., Carmany-Rampey, A., Furthauer, M., Gonza-lez, E. M., Thisse, C., Thisse, B., Halpern, M. E., and Solnica-Krezel, L. (1999). Maternal and zygotic activity of the zebrafishogon locus antagonizes BMP signaling. Dev. Biol. 214, 72–86.

Miller-Bertoglio, V. E. (2000). Characterization of zebrafish Chor-din and patterning of the dorsal-ventral axis. Ph.D. Dissertation.The Johns Hopkins University, Baltimore, MD.

Miller-Bertoglio, V. E., Fisher, S., Sanchez, A., Mullins, M. C., andHalpern, M. E. (1997). Differential regulation of chordin expres-sion domains in mutant zebrafish. Dev. Biol. 192, 537–550.

Minami, M., Kinoshita, N., Kamoshida, Y., Tanimoto, H., andTabata, T. (1999). brinker is a target of Dpp in Drosophila thatnegatively regulates Dpp-dependent genes. Nature 398, 242–246.

Mintzer, K. A., Lee, M. A., Runke, G., Trout, J., Whitman, M., andMullins, M. C. (2001). Lost-a-fin encodes a type I BMP receptor,Alk8, acting maternally and zygotically in dorsoventral patternformation. Development 128, 859–869.

Mullins, M. C. (1998). Holy Tolloido: Tolloid cleaves SOG/Chordinto free DPP/BMPs. Trends Genet. 14, 127–129.

Mullins, M. C., Hammerschmidt, M., Kane, D. A., Odenthal, J.,Brand, M., van Eeden, F. J., Furutani-Seiki, M., Granato, M.,Haffter, P., Heisenberg, C. P., Jiang, Y. J., Kelsh, R. N., andNusslein-Volhard, C. (1996). Genes establishing dorsoventralpattern formation in the zebrafish embryo: The ventral specify-ing genes. Development 123, 81–93.

Nakayama, T., Snyder, M. A., Grewal, S. S., Tsuneizumi, K.,Tabata, T., and Christian, J. L. (1998). Xenopus Smad8 actsdownstream of BMP-4 to modulate its activity during vertebrateembryonic patterning. Development 125, 857–867.

Nguyen, V. H., Schmid, B., Trout, J., Connors, S. A., Ekker, M., andMullins, M. C. (1998). Ventral and lateral regions of the zebrafishgastrula, including the neural crest progenitors, are establishedby a bmp2b/swirl pathway of genes. Dev. Biol. 199, 93–110.

Nguyen, V. H., Trout, J., Connors, S. A., Andermann, P., Weinberg,E., and Mullins, M. C. (2000). Dorsal and intermediate neuronalcell types of the spinal cord are established by a BMP signalingpathway. Development 127, 1209–1220.

Nikaido, M., Tada, M., Saji, T., and Ueno, N. (1997). Conservationof BMP signaling in zebrafish mesoderm patterning. Mech. Dev.61, 75–88.

Odenthal, J., and Nusslein-Volhard, C. (1998). fork head domaingenes in zebrafish. Dev. Genes Evol. 208, 245–258.

Oelgeschlager, M., Larrain, J., Geissert, D., and De Robertis, E. M.(2000). The evolutionarily conserved BMP-binding proteinTwisted gastrulation promotes BMP signalling. Nature 405,757–763.

Onichtchouk, D., Chen, Y. G., Dosch, R., Gawantka, V., Delius, H.,Massague, J., and Niehrs, C. (1999). Silencing of TGF-betasignalling by the pseudoreceptor BAMBI. Nature 401, 480–485.

Piccolo, S., Agius, E., Lu, B., Goodman, S., Dale, L., and DeRobertis, E. M. (1997). Cleavage of Chordin by Xolloid metallo-protease suggests a role for proteolytic processing in the regula-tion of Spemann organizer activity. Cell 91, 407–416.

Piccolo, S., Sasai, Y., Lu, B., and De Robertis, E. M. (1996).Dorsoventral patterning in Xenopus: inhibition of ventral signalsby direct binding of chordin to BMP-4. Cell 86, 589–598.

Ross, J. J., Shimmi, O., Vilmos, P., Petryk, A., Kim, H., Gaudenz, K.,Hermanson, S., Ekker, S. C., O’Connor, M. B., and Marsh, J. L.(2001). Twisted gastrulation is a conserved extracellular BMPantagonist. Nature 410, 479–483.

Sasai, Y., Lu, B., Steinbeisser, H., Geissert, D., Gont, L. K., and DeRobertis, E. M. (1994). Xenopus chordin: a novel dorsalizingfactor activated by organizer-specific homeobox genes. Cell 79,779–790.

Schmid, B., Furthauer, M., Connors, S. A., Trout, J., Thisse, B.,Thisse, C., and Mullins, M. C. (2000). Equivalent genetic roles forbmp7/snailhouse and bmp2b/swirl in dorsoventral pattern for-mation. Development 127, 957–967.



Schulte-Merker, S., Ho, R. K., Herrmann, B. G., and Nusslein-Volhard,C. (1992). The protein product of the zebrafish homologue of themouse T gene is expressed in nuclei of the germ ring and thenotochord of the early embryo. Development 116, 1021–1032.

Schulte-Merker, S., Lee, K. J., McMahon, A. P., and Hammer-schmidt, M. (1997). The zebrafish organizer requires chordino.Nature 387, 862–863.

Scott, I. C., Blitz, I. L., Pappano, W. N., Maas, S. A., Cho, K. W., andGreenspan, D. S. (2001). Homologues of Twisted gastrulation areextracellular cofactors in antagonism of BMP signalling. Nature410, 475–458.

Smith, W. C., and Harland, R. M. (1992). Expression cloning ofnoggin, a new dorsalizing factor localized to the Spemann orga-nizer in Xenopus embryos. Cell 70, 829–840.

Solnica-Krezel, L., Stemple, D. L., Mountcastle-Shah, E., Rangini,Z., Neuhauss, S. C., Malicki, J., Schier, A. F., Stainier, D. Y.,Zwartkruis, F., Abdelilah, S., and Driever, W. (1996). Mutationsaffecting cell fates and cellular rearrangements during gastrula-tion in zebrafish. Development 123, 67–80.

Thomsen, G. H. (1997). Antagonism within and around the orga-nizer: BMP inhibitors in vertebrate body patterning. TrendsGenet. 13, 209–211.

Tsuneizumi, K., Nakayama, T., Kamoshida, Y., Kornberg, T. B.,Christian, J. L., and Tabata, T. (1997). Daughters against dppmodulates dpp organizing activity in Drosophila wing develop-ment. Nature 389, 627–631.

Weinberg, E. S., Allende, M. L., Kelly, C. S., Abdelhamid, A.,Murakami, T., Andermann, P., Doerre, O. G., Grunwald, D. J.,and Riggleman, B. (1996). Developmental regulation of zebrafishMyoD in wild-type, no tail and spadetail embryos. Development122, 271–280.

Wilson, P. A., and Hemmati-Brivanlou, A. (1995). Induction ofepidermis and inhibition of neural fate by Bmp-4. Nature 376,331–333.

Wilson, P. A., Lagna, G., Suzuki, A., and Hemmati-Brivanlou, A.(1997). Concentration-dependent patterning of the Xenopus ec-toderm by BMP4 and its signal transducer Smad1. Development124, 3177–3184.

Received for publication September 28, 2001Revised January 8, 2002

Accepted February 5, 2002Published online March 19, 2002



Documents

Modulation of BMP Activity in Dorsal-Ventral Pattern ... · Modulation of BMP Activity in Dorsal-Ventral Pattern Formation by ... We analyzed the interactions between mutations in