Development 115, 607-616 (1992)Printed in Great Britain © The Company of Biologists Limited 1992

607

Activation of the easter zymogen is regulated by five other genes to define

dorsal-ventral polarity in the Drosophila embryo

REBECCA CHASAN*-t, YISHI JIN* and KATHRYN V. ANDERSON

Division of Genetics, Department of Molecular and Cell Biology, University of California, Berkeley, California 94720, USA

*The first two authors contributed equally to this worktCurrent address: The Plant Cell, American Society of Plant Physiologists, 15501 Manona Drive, Rockville, MD 20855, USAtCurrent address: Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Summary

The product of the Drosophila easter gene, a member ofthe trypsin family of serine proteases, must be moreactive ventrally than dorsally to promote normalembryonic polarity. The majority of the easter protein inthe embryo is present in the unprocessed zymogen formand appears to be evenly distributed in the extracellularspace, indicating that the asymmetric activity of wild-type easter must arise post-translationally. A dominantmutant form of easter that does not require cleavage ofthe zymogen for activity (eaAN) is active both dorsallyand ventrally. The eaAN mutant bypasses the require-

ment for five other maternal effect genes, indicating thatthese five genes exert their effects on dorsal-ventralpatterning solely by controlling the activation of theeaster zymogen. We propose that dorsal-ventral asym-metry is initiated by a ventrally-localized molecule in thevitelline membrane that nucleates an easter zymogenactivation complex, leading to the production of ven-trally active easter enzyme.

Key words: easter, dorsal-ventral patterning, Drosophila,serine protease, zymogen.

Introduction

The products of 11 maternally expressed genes requiredfor the development of lateral and ventral structures ofthe Drosophila larva act by creating a ventral to dorsalgradient of the dorsal protein in nuclei of the blasto-derm embryo (Steward et al., 1988; Roth et al., 1989;Rushlow et al., 1989; Steward, 1989). These dorsalgroup genes appear to encode components of a signaltransduction pathway that relays information from anexternal asymmetric cue to the cytoplasm to promotethe graded translocation of dorsal protein from thecytoplasm to the nucleus.

Genetic experiments have defined a flow of infor-mation among the dorsal group genes (Anderson et al.,1985; P. Hecht, D. Morisato, Y. J. and K. V. A.,unpublished data). The Toll gene, which acts upstreamof dorsal, encodes a transmembrane protein that isuniformly distributed in the plasma membrane of theearly embryo (Hashimoto et al., 1988; 1991). Seven ofthe dorsal group genes act upstream of Toll. Recentdata indicate that all seven of these genes are necessaryfor the asymmetric activation of the Toll protein,apparently by producing a ventrally localized ligand forToll (Stein et al., 1991; Stein and Niisslein-Volhard,1992). Three of these genes, pipe, nudel, and wind-beutel, are somatic-dependent (Stein et al., 1991;

Manseau and Schiipbach, 1989); they are probablyexpressed during oogenesis in follicle cells, whichsecrete the eggshell. The other four genes, easter,snake, gastrulation defective and spdtzle, are germline-dependent (Stein et al., 1991; Seifert et al., 1987;Konrad et al., 1988) and encode proteins that areapparently translated from maternal RNAs in the earlyembryo and secreted to the perivitelline space betweenthe eggshell and the plasma membrane (Fig. 1; DeLottoand Spierer, 1986; Konrad and Marsh, 1990; D.Morisato and K. V. A., unpublished). Both easter andsnake have significant structural similarity to extracellu-lar serine proteases of the trypsin family (DeLotto andSpierer, 1986; Chasan and Anderson, 1989).

It is not known how the products of the seven genesupstream of Toll act together to generate a ventrally-localized Toll ligand, but asymmetric easter activityappears to be essential for localized ligand production.Females carrying EMS-induced dominant alleles ofeaster produce ventralized or lateralized embryos inwhich lateral structures are expanded at the expense ofdorsal structures (Chasan and Anderson, 1989; Jin andAnderson, 1990). These dominant easter alleles, whichare caused by mis-sense mutations in the catalyticdomain (Jin and Anderson, 1990), exert their effects bycausing a more uniform distribution of nuclear dorsalprotein (Steward, 1989). From these results we have

608 R. Chasan, Y. Jin and K. V. Anderson

m H iB C D E F Q H I Fig. 1. The easter protein inyoung embryos is largely in thezymogen form, as assayed onwestern blots. (A) Totalembryonic extract from 0-3hour wild-type (Oregon R)embryos. (B) Total embryonicextract from 0-3 hour embryosfrom ea-^aVeo5022"1) females.(C) Perivitelline extract from0-12 hour wild-type embryos.(D) Perivitelline extract from0-12 hour embryos from 77"(Df(3R)roXB3/Df(3R)r/9QRX)females. (E) Perivitellineextract from 0-12 hourembryos from ea^/ea5022™1

females. (F) Perivitellineextract from 0-12 hour

embryos from females carrying the dominant lateralizing allele, ea831'1253 and one copy of ea+. ea831"1253 is an allele thatcarries the point mutations in both ea831 and ea125 3, introduced into the genome by P-element transformation (Jin, 1991).(G) Total embryonic extract from wild-type embryos. (H) Total embryonic extract from ea~ embryos. (I) Perivitelline fluidcollected by micropipette after pricking the vitelline membrane of 200 TV (ea+) embryos. Lines to the left of lane (G)correspond to the position of markers of the same size as shown at the left of the figure (A/rxl0~3). The two blots in lanesA-E and lanes G-I were probed with two batches of affinity-purified antibodies, hence the difference in background bands.The antibodies recognize a protein of ~50xl03Afr in wild-type and Tl~ embryos (arrowhead) that is absent in embryosfrom eaA/ea5O22rxl females. Since antibodies fail to detect either of the two smaller chains that would be present in thecleaved, processed enzyme, we conclude that the processed form represents only a small fraction of the total easter proteinpresent in the embryo.

2 0 0 -

97 -

69 -

46 -

3 0 -

2 1 -1 4 -

inferred that wild-type easter activity is spatiallyasymmetric, allowing the activation of Toll and thesubsequent nuclear localization of dorsal on the ventralside of the embryo only. The dominant easter alleleslead to a more symmetric distribution of easter activity,leading to activation of Toll and dorsal nuclearlocalization on the dorsal side of the embryo. Tounderstand how dorsal-ventral polarity arises, it istherefore essential to understand how easter activitybecomes spatially asymmetric.

The sequence of the predicted product of the eastergene has all the features of the trypsin family of serineproteases (Chasan and Anderson, 1989), including azymogen activation site at the amino terminus of theprotease catalytic domain. By analogy to other serineproteases, the easter protein would need to be proteoly-tically cleaved at this site to be catalytically active. Theactivities of many proteases, such as those of the bloodcoagulation cascade, are regulated by cleavage of thezymogen (Furie and Furie, 1988). This precedent raisesthe possibility that easter activity might be spatiallyregulated by asymmetric zymogen cleavage.

In this report, we investigate the role of zymogenprocessing in the regulation of easter activity. Most ofthe easter protein in the embryonic extracellular space isin the unprocessed zymogen form. However both wild-type easter and the EMS-induced dominant easter alleleproducts require a normal zymogen activation site tohave any activity, which implies that zymogen cleavageis a necessary step in activation of both the wild typeand mis-sense dominant proteases. We also studied therole of zymogen cleavage by constructing a mutant formof the easter protein that lacks the amino-tenninal

domain. This mutant, which should be active withoutzymogen cleavage, has a dominant phenotype like thatof the EMS-induced dominant alleles: it promotes thedevelopment of lateral structures, but does not promotedorsal-ventral asymmetry. Using this mutant, we showthat five of the six other genes upstream of Toll exerttheir effects on dorsal-ventral patterning solely throughthe activation of the easter zymogen. These resultssuggest that spatial regulation of easterns activity bylocalized zymogen activation is a key initial event indefining the polarity of the dorsal-ventral embryonicpattern.

Materials and methods

Stocks and mutant phenotypesThe EMS-induced dominant alleles of easter have beendescribed (Chasan and Anderson, 1989; Jin and Anderson,1990; Erdeiyi and Szabad, 1989). Most recessive dorsal groupalleles are described in Tearle and Niisslein-Volhard (1987) orSchiipbach and Wieschaus (1989). The windbeutel alleleswbr98 and wbf*46, the kind gift of T. Schupbach, are stronglydorsalizing. Df (3R) ea831™ and Df (3R) ea5022"1 aredeficiencies including the easter locus, isolated as phenotypicrevertants of dominant easter alleles (Chasan and Anderson,1989; Erd61yi and Szabad, 1989; Chasan, 1991). The ea~embryos were produced by eaYeo5022™1 females; theseembryos have no detectable easter mRNA (Chasan, 1991).The phenotypes of mutant embryos were evaluated in thepattern of gastrulation and in the structures of the differen-tiated cuticle (Wieschaus and Nusslein-Volhard, 1986).

Production of antibodiesAnti-easter antibodies were obtained by immunizing rats with

Regulated easter zymogen activation 609

a trpE-easter fusion protein (Dieckman and Tzagaloff, 1985)containing the entire easter protein coding region except forthe first four amino acids, which are part of the predictedsignal peptide. The fusion protein was purified by electro-elution from preparative SDS-PAGE gels. Rats were injectedwith 80-100 mg of fusion protein emulsified in syntheticadjuvant (MPL + TDM; RIBI Immunochem) for both initialinjections and boosts. Antisera were affinity-purified (Harlowand Lane, 1988) on columns of the trpE-easter fusion proteinbound to Affigel 10/15 (Driever and Niisslein-Volhard, 1988).

Preparation and analysis of Drosophila embryoextractsWhole embryo extracts were prepared as described (Hashi-moto et al., 1991). Perivitelline fluid was removed bymicropipetting as described (Stein et al., 1991) from embryoslaid by Toll~ females. Perivitelline extracts were prepared byvortexing embryos in buffer with silicon carbide particles,which preferentially releases the contents of the extracellular,perivitelline space (Jin, 1991). We have found that -25% ofthe total easter protein is released into the perivitellineextracts and that easter is ~25-fold enriched in perivitellinecompared to total embryonic extracts. Extracts were electro-phoreticaOy separated on SDS-PAGE gels under reducingconditions and transferred to nitrocellulose for western blotanalysis (Towbin et al., 1979). Only an easter protein the sizeof the unprocessed zymogen was seen on western blots (Fig. 1and data not shown) or by immunoprecipitation of in vivolabeled easter protein (Chasan, 1991).

Site-directed mutagenesis and assays for the activity ofmutant allelesSite-directed mutagenesis was performed as described (Jinand Anderson, 1990). All mutations were confirmed bysequencing (Sanger et al., 1977).

The primers used for easter zymogen activation sitemutation were: the Arg-127 to Gin: 5'TGTCGAATCA-GATCTATGGC3'; the Arg-127 to Leu: 5'CTTTCGAATCT-CATCTATGG3'. The cDNAs containing both the zymogensite mutation and a dominant point mutation were con-structed by replacing the appropriate fragment of the cDNAcarrying the zymogen site mutation with the correspondingfragment of the dominant allele genomic DNA. The serine-338 to alanine mutation was constructed previously (Jin andAnderson, 1990). For transformation of the efl

83'-ala3:fe allele,the Xhol-Pstl fragment of the wild-type genomic DNA (Jinand Anderson, 1990) in Bluescript (Stratagene) was replacedwith the corresponding fragment of the doubly mutantcDNA. The 2.5 kb EcoKl-Pstl fragment containing thecomplete easter gene (Jin and Anderson, 1990) including theea and the serine-338 to alanine mutations was cloned intothe CaSpeR-2 or -3 vectors (Pirrotta, 1988) and injected intow1118 embryos. Transformants carrying mutant easter DNAswere selected on the basis of w+ eye color.

The mutagenic primer for the eaAN deletion had thesequence 5'TCCGCCATAGATGCCCGCGGATCATTT3'.The 12 nucleotides in the 5' part of the primer hybridize to thesequence encoding the first four amino acids of the catalyticdomain (IYGG) and the 15 nucleotides in the 3' part of theprimer hybridize to the sequence encoding the last five aminoacids of the predicted signal peptide (KSSAG) (Chasan andAnderson, 1989). The template, which was cloned in thevector pGEM-7Zf(+) (Promega Biotec), had been isolated ina separate mutagenesis experiment and contained a deletionwithin the N-terminal region, which reduced the size of theregion to be looped out. By analogy with other serine

proteases, the cysteine at position 260 in the catalytic domainwould form a disulfide bond with a cysteine residue in the N-terminal domain (Chasan and Anderson, 1989). To eliminatethis unpaired cysteine, it was changed to a serine in a secondmutagenesis. This mutagenic primer had the sequence5'AGGCAGGGATATCGGTCGCAC3'.

For transcript injection assays, the templates were linear-ized and capped SP6 transcripts were made essentially asdescribed (Krieg and Melton, 1987). Injections were aspreviously described (Chasan and Anderson, 1989).

Construction of stocks carrying dominant easter andrecessive dorsal group mutationsDouble mutants of dominant easter (eaD) alleles with dorsal,gastrulation defective and windbeutel were constructed byordinary crosses. To make third chromosomes that carriedboth ea513 and a recessive dorsal group mutation, hetero-zygous larvae were irradiated as described (Anderson et al.,1985) to produce X-ray induced mitotic recombinants in themale germ line. Recombinants were initially identified on thebasis of the exchange of flanking markers and confirmed bytesting for maternal effect phenotypes. The double mutants ofea125 or ea831 and recessive dorsalizing alleles were con-structed using one of two dominant suppressors, TF26 or asuppressor on the second chromosome (Y. J. and K. V. A.,unpublished). TF26 is an incompletely penetrant dominantdorsalizing allele of Toll that acts specifically as a dominantsuppressor of dominant easter ventralizing mutations. Fromthe progeny of females heterozygous for TF26 and ea1253 orea , recombinants were recovered that carried TF26 and theeaP allele on the same chromosome. To construct doublemutant chromosomes of other third chromosomal dorsalgroup alleles with these ea° alleles, double recombinantprogeny from eaD rr^/dorsal group allele females thatretained the ea° allele, lost TF26 and gained the other dorsalgroup allele were identified by marker exchange. Recombi-nant lines were tested for the presence of the desired maternaleffect mutations in test crosses. The second chromosomalsuppressor was used in a similar manner to construct some ofthe double mutants with eam.

Results

The zymogen form of the easter protein is present inthe perivitelline spaceAlthough the phenotypes caused by the dominantalleles indicate that easter activity must be dorsoven-trally asymmetric in the embryo, the easter proteinappears to be uniformly synthesized in the earlyembryo, since both the easter transcript (Jin, 1991) andthe newly synthesized easter protein (Chasan, 1991)appear to be uniformly distributed in the blastodermembryo. The N-terminal signal sequence in the easteropen reading frame (Chasan and Anderson, 1989) andthe presence of easter-rescuing activity in the perivitel-line fluid (Stein and Niisslein-Volhard, 1992) suggestedthat the easter protein is secreted from the embryo intothe extracellular perivitelline fluid that lies between theplasma membrane and the vitelline layer of theeggshell. Conventional techniques for antibody stainingof Drosophila embryos require removal of the vitellinemembrane and would cause the loss of proteins that aresoluble in the perivitelline space. We were able,however, to confirm the presence of easter protein in

610 R. Chasan, Y. Jin and K. V. Anderson

this compartment by micropipetting perivitelline fluidout of embryos (Stein et al., 1991), and assaying foreasier protein on western blots (Fig. 1). These exper-iments showed that the easter protein was soluble andtherefore presumably evenly distributed in the perivi-telline fluid.

Only a 50 x 103 Mr protein, which corresponds to thefull-length unprocessed zymogen form of the easterprotein, was detected in whole embryo or perivitellinefluid extracts from wild type or dominant mutantfemales (Fig. 1 and data not shown). The size of theprotein is somewhat larger than the predicted size (41 x103 Mr) of the easter zymogen (Chasan and Anderson,1989), but is the same size as the protein made by invitro translation and translocation using SP6 transcriptsof the easter cDNA (Chasan, 1991). Cleavage of thezymogen at a defined site at the N-terminal end of thecatalytic domain (Chasan and Anderson, 1989) wouldyield two smaller polypeptides, but two smaller bandsthat would correspond to the cleaved, activated easterprotein were not detected on reducing and denaturinggels. We estimate that if a processed form of easter ispresent, it constitutes less than 10% of the steady-stateamount of easter protein.

Wild-type easter requires a normal zymogen activationsite for activityZymogen activation is a key step in allowing serineprotease activity (Stroud et al., 1977). However, someserine proteases, such as human tissue plasminogenactivator, have appreciable protease activity in thezymogen form (Tate et al., 1987). Because we were ableto detect only the zymogen form of easter in embryos, itwas important to test whether zymogen cleavage wasrequired for wild-type easter activity.

Based on homology with other members of thetrypsin family, activation of the easter zymogen shouldoccur by a cleavage after the arginine-127 preceding the

IYGG of the catalytic domain (Chasan and Anderson,1989), suggesting that the enzyme that activates eastercleaves after basic residues. We used site-directedmutagenesis to change arginine-127 to glutamine orleucine, both of which are similar to arginine in size, butlack the basic group. Unlike mutants at other positionsin the cleavage site, these mutants should not disruptprotease activity of the zymogen, because a similarmutant blocked normal processing but did not destroyprotease activity of the zymogen form of tissueplasminogen activator (Tate et al., 1987). Transcripts ofthe wild-type easter cDNA fully rescue the dorsalizedphenotype of embryos produced by easter~ females(Chasan and Anderson, 1989). In contrast, transcriptsof an easter cDNA that encoded either glutamine orleucine at position 127 had no activity when injectedinto either wild-type embryos or easter~ embryos(Table 1). Thus, changing arginine-127 to glutamine orleucine abolished easter activity completely, suggestingthat, even though we have not detected a cleaved formof easter biochemically, the wild-type easter is activeonly after zymogen cleavage.

The products of the EMS-induced dominant alleles actas processed proteasesThe phenotypes of the dominant easter alleles indicatethat the activities of these alleles are more spatiallyuniform than the activity of the wild-type easter. Tounderstand how easter activity is normally spatiallyregulated, it is important to define the aspect of easteractivity that is altered in the dominant alleles. Wetherefore carried out a series of genetic experiments totest whether the dominant alleles differ from the wild-type allele in their requirements for protease function,zymogen processing and the other dorsal group geneproducts for activity.

Although the sequence of the easter protein suggeststhat it acts as a protease, it is possible that the easter

Table 1. Dominant easter allele products are not active as zymogens

Transcripts

wild type (Arg127)Gin127

Gin127-**831

Gin127-**5-13

Leu127

Leu127-ea125 3

Leu127-**5-13

ng;iic

23;2388;1753; 1049;773; 1280;376:11

ea

dorsalizedat gastrulation

0885349738076

Recipient

dorsalizedcuticle

017107

123

11

n^jnc

45;4531;2336;2343;4044;3736;3018; 15

wild-type

normalgastrulation

45313643443618

cuticle

45232340373015

Transcripts of cDNAs containing zymogen activation site mutations were assayed by injection into ea or wild-type embryos. The wild-type transcript (Arg127) always rescued ea~ embryos both at gastrulation and in the pattern of the differentiated cuticle and had nodominant effect when injected into wild-type embryos. The zymogen site mutations destroyed the ability of transcripts to rescue thedorsalized phenotype when injected into ea~ embryos and blocked the dominant activity of the dominant alleles when injected into wild-type embryos. Gin and Leu127transcripts carry the zymogen site mutations in an otherwise wild-type cDNA. Leu>Z7-e(r 13 and Leu127-ea 3 transcripts contain the Arg to Leu zymogen activation site mutation as well as the nucleotide change in the lateralizing ea5 1 3 or theventralizing ea . Gln '^-ea5 1 3 and Gin -ea transcripts contain the Arg to Gin zymogen activation site mutation and the nucleotidechange in the lateralizing ea5 1 3 or the ventralizing ea831. ng: number of embryos scored at gastrulation. n^: number of embryos scored inthe pattern of the differentiated cuticle.

Regulated easter zymogen activation 611

Table 2. Phenotypes of dominant easter alleles in combination with strongly dorsalizing alleles at other locindl pip wbl snk spz Tl pit tub dl

ea1

DODO DO

DODO

DODO DO

DO

DODODO

DODODO

DODODO

DODO

DODO

All double mutant combinations tested were strongly dorsalized. The dominant alleles alone have a lateralizing (ea513), moderatelyventralizing (ea831) or weakly ventralizing (ea125 3) phenotype. Each combination of recessive alleles alone produces a strongly dorsalizing(DO) phenotype. The recessive alleles used were: TliBRE6/T(K'RE (with ea5 1 3); TF*</TiLB1 (with ea831 and ea1253); spz^/spz™1; P^/74

tub3/tubna; tutf/tub73*- pip^/pip4; ndP^/ndl133 and ndllulndlxm (with ea5 13); ndllts/ndlm (with ea831); gcf/gd7; dl'/dl1; ^/and snk^/DtiSKjry36 (with ea5-13); snk^/snk™ (with ea^; wb^/wb!**8*; wblR?jP**6

protein has additional functions and that the dominantmutations alter an aspect of easter function that isindependent of protease activity. To test this hypoth-esis, we constructed an intragenic double mutant,efl83i-aia338 j^jg m u t a n t contains both the point mu-tation in the dominant ventralizing allele ea and analanine residue in place of the active site serine-338 (Jinand Anderson, 1990). The serine-338 to alanine change,which is a conservative amino acid replacement thatshould abolish protease function without affectingprotein structure, destroyed the activity of the wild-typeeaster (Jin and Anderson, 1990). When the ea831""1338

genomic DNA was introduced into the genome by P-element mediated transformation, western blot analysisshowed that it produced a stable full-length protein(data not shown). However, this protein was unable topromote the production of any ventral or lateralstructures in an easter~ background. The double mutantallele also had no dominant effect: females carrying theea83i-aia338 ane i e m (jje presence of one or two copies ofea+ produced 100% hatching embryos. Thus, allactivity of ea831 absolutely requires the active site serine,and the dominant activity of the allele most probablyreflects a change in the activity of the protease.

Since the zymogen form of easter is spatially uniformin its distribution, the more spatially uniform activity ofthe dominant easter alleles could be accounted for if thedominant proteins have some activity in the zymogenform. To test the activity of the zymogen form of thedominant alleles, we constructed cDNAs that containboth an amino acid change causing a dominantphenotype and a mutation at the zymogen activationsite. Transcripts containing the point mutations thatcause the dominant easter mutations produce ventra-lized or lateralized embryos when injected into youngwild-type or ea~ embryos (Jin and Anderson, 1990).Transcripts from cDNAs containing both the arg-127-to-leucine or arg-127-to-glutamine in addition to one ofthe mutations causing the dominant alleles ea1253, ea831

or ea513 did not rescue any lateral or ventral structureswhen injected into easter' embryos, nor did they alterthe dorsal-ventral pattern of wild-type embryos (Table1). Because all activity of the dominant easter alleleswas lost when the zymogen activation site was mutated,we conclude that, like the wild-type allele, the domi-nant alleles require zymogen cleavage to have anyactivity.

To help understand how the EMS-induced dominantalleles escape normal spatial regulation, we examined

the phenotypes of double mutants of dominant easteralleles with mutations in other dorsal group genes.Double mutants of dominant ventralizing or lateralizingeaster alleles and recessive strongly dorsalizing alleles atthe other ten dorsal group loci were all stronglydorsalized, indicating that the dominant alleles do notact by bypassing a regulatory step imposed by one of theother dorsal group genes (Table 2).

From the site-directed mutagenesis experiments andthe analysis of double mutant phenotypes, we concludethat the EMS-induced dominant easter alleles aresimilar to wild-type easter in a number of respects. Boththe wild-type and dominant alleles require the activesite serine and a normal zymogen processing site for anyactivity. In addition, all activity of both wild type anddominant easter alleles depends absolutely on theactivity of all other dorsal group genes.

A form of easter that does not require zymogencleavage is a dominant lateralizing alleleTo investigate directly whether zymogen cleavage is aregulated step in the control of easter activity, we usedsite-directed mutagenesis to delete the amino-terminaldomain in an easter cDNA (Fig. 2). In the N-terminaldeletion mutant, which we term eaAN, the signalsequence is followed directly by the catalytic domain.This mutant form of easter should be secreted to theperivitelline space and cleavage by signal peptidaseshould release the active C-terminal catalytic domain,bypassing the normal requirement for zymogen cleav-age. Similar mutant forms of trypsin and of tissueplasminogen activator retain protease activity (Vasquezet al., 1989; MacDonald et al., 1986).

To assay the activity of the eaAN mutant, we injectedin vitro-synthesized transcripts into embryos fromeaster" females. As Fig. 3 shows, transcripts of theeaAN cDNA had biological activity, promoting theproduction of lateral structures never made in unin-jected easter~ embryos. The rescued pattern, however,lacked dorsal-ventral asymmetry. The pattern of gastru-lation was like that of lateralized embryos: both ventraland dorsal pattern elements are absent and all cellsbehave like the normal lateral cells (Anderson et al.,1985). The cephalic furrow, which begins midlaterallyin the wild-type embryo, initiated equally at all dorsal-ventral positions, and neither dorsal folds nor a ventralfurrow formed. Also like lateralized embryos, theinjected embryos differentiated laterally derived ven-

612 R. Chasan, Y. Jin and K. V. Anderson

WT

eaAN

ss

ss

N-term domain

catalytic domain

catalytic domain

Cy.

Fig. 2. Structure of an easiermutant that lacks the N-terminal domain. The eaANmutant of easier was createdby site-directed mutagenesis,using an oligonucleotide thatlooped out the region betweenwhat we predict to be the end

sw of the signal sequence and theisoleucine residue that is the

first amino acid of the catalytic domain (Chasan and Anderson, 1991). In addition, the cysteine in the catalytic domain thatshould form a disulfide bond with a cysteine in the N-terminal domain was changed to a serine to eliminate the unpairedcysteine.

tral denticle bands around the entire embryo circumfer-ence. When the eaAN transcripts were injected intowild-type embryos, the pattern of gastrulation wasventralized: a ventral furrow formed, but dorsal foldswere absent and the germ band failed to extend. Thedifferentiated cuticle of these embryos also appearedventralized, lacking all dorsal and dorsolateral struc-tures and having ventral denticle bands around theembryonic circumference. In these injection assays, the

Fig. 3. Transcripts of the eaANmutant lateralize easier"embryos. Embryos fromeasier" females {ea^/ea5022™1)were injected with SP6transcripts of the eaAN mutantcontaining the Cys-260 to Serchange. At gastrulation (A),the head fold encircled theinjected embryos, a phenotypecharacteristic of lateralizedmutant embryos (Anderson etal., 1985). (Anterior to theleft; dorsal up). The dorsalfolds, which are situateddorsally in the wild-typeembryo and encircle easter"embryos, are absent, as is theventral furrow, which formsventrally in the wild-typeembryo. The injected embryosdifferentiate cuticle (B) inwhich laterally derived ventraldenticles are found around thedorsal-ventral circumference.SP6 transcripts of the eaANmutant that retained the Cysat position 260 were alsoinjected into easter" embryos(not shown); the recipientswere also lateralized atgastrulation, although not alldifferentiated circumferentialventral denticles. Each embryowas injected, prior to pole cellformation, with approximately200 pi (-2% egg volume) of~1 mg/ml transcript.

eaAN mutant behaves like the dominant lateralizingeaster point mutant alleles, transcripts of which latera-lize easier" embryos and ventralize wild-type embryos(Jin and Anderson, 1990).

Epistatic relationships of the eaAN mutant and otherdorsal group mutantsIf the eaAN mutant bypasses a normal regulatory step,

Regulated easter zymogen activation 613

it should be active in the absence of the gene productsnecessary for regulation. We therefore assayed eaANactivity in embryos from females lacking the function ofeach of the six dorsal group genes besides easter thathad been defined as genetically upstream of Toll. Asshown in Table 3 and Fig. 4, eaAN transcripts elicitedthe production of lateral structures in embryos fromfemales mutant for pipe, nudel, windbeutel, gastrulationdefective, and snake. Recipient embryos were latera-lized and were indistinguishable from easter~ embryosinjected with the same transcripts. Thus, unlike theEMS-induced dominant alleles, the eaAN allele doesnot require these five dorsal group genes for activity.The epistasis of eaAN over these genes shows that thesefive genes act genetically upstream of easter and arerequired for activation of the easter zymogen.

. Among the genes upstream of Toll, the spdtzle genewas unique in this epistasis assay. Embryos fromspdtzle' females remained dorsalized when injectedwith the eaAN transcript, indicating that even in thepresence of the preactivated easter, the spdtzle product'

Fig. 4. Cuticle phenotypes ofwindbeutel (A) and spdtzle (B)embryos injected with theeaAN transcript. Embryos laidby females of the genotypewbl^/wbl™88 are normallydorsalized, but are lateralizedby the injection of eaANmutant transcripts, both atgastrulation (not shown) andin the pattern of the cuticle,which differentiates laterallyderived ventral denticles at alldorsal-ventral positions. Incontrast, embryos fromspz^/Df^R)!?***1 femalesremain dorsalized wheninjected with the mutanttranscript: the embryosdifferentiate the dorsalizedcuticle characteristic ofuninjected embryos.

is still needed for the activity of Toll and of thesubsequent steps of the dorsal-ventral pathway.

Discussion

Although the easter protein appears to be uniformlydistributed in the embryonic perivitelline space, thephenotypes of the previously described EMS-induceddominant alleles and of the eaAN mutant describedhere argue strongly that a wild-type dorsal-ventralpattern can form only if easter activity is spatiallyasymmetric. Our studies on the two classes of dominantalleles allow us to infer how easter activity is confined toventral regions of the wild-type embryo.

The point dominant mutations redistribute the activityof the processed proteaseEach of the nine EMS-induced dominant mutations iscaused by a single mis-sense mutation at a conservedsite in the catalytic domain (Jin and Anderson, 1990).

614 R. Chasan, Y. Jin and K. V. Anderson

Table 3. Phenotypes of dorsal group mutant embryosinjected with the eaAN mutant transcript

Recipient(maternal genotype)

easterpipenudelwindbeutelgastrulation defectivesnakespatzle

Gastrulationphenotype

L

122

516249680

D

2

6213

70

Cuticlephenotype

L

3129

7101224

0

D

840003

39

Embryos from females homozygous for strong dorsalizing allelesof easier, pipe, nudel, windbeutel, gastrulation defective, snake orspStzle were injected prior to pole cell formation. Embryos werescored as being lateralized (L) or completely dorsalized (D) atgastrulation and in the pattern of the embryonic cuticle.Lateralized embryos have a prominent, dorsoventrally symmetrichead fold at gastrulation, but no ventral furrow or dorsal folds,and have circumferential ventral denticle bands in thedifferentiated cuticle. Dorsalized embryos make a series ofdorsoventrally symmetric transverse folds at gastrulation anddifferentiate cuticle covered with fine dorsal hairs. Maternalgenotypes: ea4/^5022"1; pip^/pip66*; ndl™/ndllu; wbl^lwbl™ orwbrfwbP188; gd'/gd1; snk^/snk™; ^zmVDf(3R)775BRkl.

To help determine how these alleles lead to a change inthe distribution of easter activity, we carried outexperiments to test whether the products of thedominant alleles act as processed proteases.

We found that, like wild-type easter, the ventralizingea831 allele absolutely requires the active site serine foractivity. In addition, the dominant alleles ea83', ea513

and ea1253 require a normal zymogen activation site.Like wild-type easter, these dominant alleles depend onthe activity of all other dorsal group gene products toaffect the dorsal-ventral pattern, indicating that thesealleles do not bypass a regulatory step imposed by anyof the known dorsal group genes. Our data suggest thatboth the wild-type and dominant forms of easter act asprocessed proteases and that the ventralizing andlateralizing effects of the dominant alleles are caused bya more uniform spatial distribution of the activity of theprocessed protease.

Cleavage of the easter zymogen is regulated by otherdorsal group genesThe vast majority of the easter protein in the embryo ispresent as the unprocessed zymogen form. However,we have shown that the zymogen site, and we inferzymogen processing, is necessary for activity. Becausewe cannot detect processed easter, we assume that onlya small fraction of the total easter protein in the embryois in the catalytically active form. It seemed possible,therefore, that the production of the active easterprotease by cleavage of the easter zymogen could be arate-limiting, regulated step in dorsal-ventral pattern-ing.

To assess the role of easter zymogen cleavage indorsal-ventral patterning, we used site-directed muta-genesis to create a mutant form of easter, eaAN, thatwould not require activation by zymogen cleavage

because the N-terminal domain was deleted. The eaANmutant promotes the development of lateral structures,indicating that the mutant retains easter proteaseactivity and normal substrate specificity. However, incontrast to wild-type easter, which appears to be activeonly ventrally, the eaAN mutant is equally active at alldorsal-ventral positions. Thus deletion of the amino-terminal domain, thereby bypassing the requirementfor zymogen cleavage, abolishes normal spatial regu-lation of activity.

Unlike the EMS-induced dominant alleles, the eaANmutant is epistatic to several dorsal group genes. Thismade it possible to order partially the function of thedorsal group genes upstream of Toll. Five of thesedorsal group genes, the somatic-dependent genes pipe,nudel, and windbeutel and the germ line-dependentgenes snake and gastrulation defective, are not requiredfor the production of lateral structures in the presenceof the eaAN protein, and therefore act upstream ofwild-type easter to promote its activity. Since all five ofthese genes are required for activation of easter and arebypassed by a mutant form of easter that does notrequire zymogen activation, we infer that these fivegenes are required, directly or indirectly, for cleavageand activation of the easter zymogen.

Generating an asymmetric ligand for TollRecent data indicate that the products of the sevendorsal group genes upstream of Toll all act to produce aspatially localized ligand for the Toll protein, but thesedata do not order the activities of these genes orestablish which of those genes could encode the Tollligand (Stein et al., 1991). Our results show that pipe,nudel, windbeutel, snake and gastrulation defective exerttheir effects on dorsal-ventral patterning solely throughthe regulation of easter activity, and therefore none ofthese genes encodes a ligand that binds to and activatesToll. Because spatzle is the only dorsal group geneupstream of Toll that is required for eaAN to exert itseffect on the dorsal-ventral pattern, easter and spdtzleare required most directly to activate Toll. The easterprotein and an activity that rescues the spatzle pheno-type both appear to be initially uniformly distributed inthe perivitelline space (Fig. 1; Stein and Niisslein-Volhard, 1992). One attractive hypothesis that wouldexplain how a Toll ligand is asymmetrically produced isthat asymmetrically active easter protease cleaves thespdtzle protein, and cleaved spatzle protein then bindsto and activates the Toll protein.

Models for localization of easter activityBecause asymmetric easter activity is crucial for gener-ating a localized Toll ligand, and because the easterzymogen is uniformly distributed, it is crucial tounderstand how other gene products act post-translatio-nally to confine easter activity to the ventral part of theembryo. The dominant easter alleles provide powerfultools to dissect that spatial regulation. Two verydifferent kinds of mutations in easter disrupt the dorsal-ventral asymmetry of easter activity: the eaAN mutantlacks the amino-terminal domain and does not require

Regulated easter zymogen activation 615

• easter zymogen

^ pipe, nudel, wlndbeutel,gastrulatlon defective, snake(zymogen activation complex)

Fig. 5. A model for the spatial regulation of easter activity.The products of pipe, nudel, windbeutel, gastrulationdefective and snake are required for the production of aventrally localized zymogen activation complex. Just one ofthese proteins, most likely the product of one of thesomatic-dependent genes, would need to be asymmetricallydistributed for zymogen activation to be spatiallyasymmetric. The model assumes that the easter zymogen isuniformly distributed in the perivitelline space, thatzymogen is activated by cleavage on the ventral side of theembryo, that the active easter protease acts before itdiffuses far from the ventral side, and that its cleavedsubstrate also does not diffuse freely.

zymogen cleavage, while the EMS-induced dominantalleles require the normal machinery of zymogenactivation as well as an intact zymogen activation sitefor activity.

Two classes of models of how easter activity isspatially regulated can explain how the dominantmutations alter the dorsoventral pattern. In the firstmodel, the EMS-induced dominant alleles and theeaAN mutant both disrupt a single regulatory step. Forinstance, there could be a spatially localized activator orrepressor that requires both the amino-terminal domainand part of the catalytic domain for binding to the easterprotein. In the absence of the amino-terminal domainor in the presence of mis-sense mutations in criticalpositions of the catalytic domain, easter would some-how be active, independent of this regulator. Becausethe activities of the EMS-induced dominant alleles aredependent on all the known dorsal group genes, thisregulator would have to be the product of someuncharacterized gene.

We prefer a second model in which the two kinds ofdominant mutations disrupt two different steps in easterprotease function (Fig. 5). This model does not requireinvoking the existence of unknown genes and explainswhy several genes are required for activation of theeaster zymogen. As in the model proposed by Stein etal. (1991), embryonic dorsal-ventral asymmetry istriggered by a molecule localized ventrally in thevitelline membrane, perhaps the product of the so-matic-dependent pipe, nudel, or windbeutel genes. Inour model, this ventrally localized molecule nucleatesthe assembly of an easter zymogen activation complex,which could incorporate the products of the snakeand/or gastrulation defective genes, on the ventral side

of the vitelline membrane (Fig. 5). This complex couldbe analogous to the membrane-localized prothrombi-nase complex, which accelerates the rate of prothrom-bin activation lO'-fold (Furie and Furie, 1988; Krishnas-wamy, 1990). Because the activation complex islocalized, the uniformly distributed easter zymogen isactivated only ventrally. Once the zymogen is cleaved,the activity of the wild-type easter must still be confinedto ventral regions. Localization of protease activitycould occur if the processed easter has a very short half-life and decays before it diffuses to the dorsal side of theembryo. Alternatively, the processed easter proteasecould remain bound ventrally after proteolytic process-ing.

In the second model, the event which initiates ventraleaster activity is the localized activation of the easterzymogen. The eaAN mutant does not require zymogenactivation and is therefore active everywhere. Theproducts of the EMS-induced dominant alleles, incontrast, would be activated ventrally by the normalzymogen activation machinery, but would be altered inthat property of the processed enzyme that normallyrestricts its activity to the ventral side. Because ofgreater stability of the processed protease or loweraffinity of the processed protease for a ventral bindingsite, the active products of the EMS-induced dominantalleles would diffuse within the perivitelline space afterzymogen cleavage and cut the easter substrate dorsallyas well as ventrally.

After the ventrally active easter protease cleaves itssubstrate, the product of that proteolytic reaction mustalso remain ventral. If easter's substrate is the ligandthat activates Toll, diffusion of the processed ligandaway from the ventral side could be prevented byimmediate binding of the ligand to Toll. The eastersubstrate would therefore not need to be prelocalized.If the soluble perivitelline rescuing activity for spdtzle(Stein and Nusslein-Volhard, 1992) is the spdtzle geneproduct and if spdtzle is easter's substrate and ToWsligand, then the rescuing activity would be the unpro-cessed spdtzle, while processed spdtzle would be tightlybound to Toll on the ventral side of the embryo.

Thus it seems likely that dorsal-ventral polaritydepends on localized activation of a receptor, Toll, inwhich neither receptor nor ligand, nor even theprotease that activates the ligand, is prelocalized.Instead, localized activation of a protease zymogeninitiates a cascade of protein interactions that directsthe spatially coordinated response of a group of targetcells.

We thank Chip Ferguson for help with the figures. Wethank Chip Ferguson, Peter Hecht, Donald Morisato, othermembers of the Anderson laboratory and Jasper Rine forcomments on the manuscript. This work was supported bygrants from the National Institutes of Health (GM 35437) andthe National Science Foundation (DCB 8452030) to K. V. A.

References

Anderson, K. V., JUrgens, G. and Nusslein-Volhard, C. (1985).

616 R. Chasan, Y. Jin and K. V. Anderson

Establishment of dorsal-ventral polarity in the Drosophila embryo:genetic studies on the role of the Toll gene product. Cell 42, 779-789.

Chasan, R. (1991). Analysis of the role of easter in dorsal-ventralpattern formation in Drosophila melanogaster. Ph.D. Thesis.University of California, Berkeley.

Chasan, R. and Anderson, K. V. (1989). The role of easter, anapparent serine protease, in organizing the dorsal-ventral patternof the Drosophila embryo. Cell 56, 391^00.

DeLotto, R. and Spierer, P. (1986). A gene required for thespecification of dorsal-ventral pattern in Drosophila appears toencode a serine protease. Nature 323, 688-692.

Dieckman, C. L. and TzagalofT, A. (1985). Assembly of themitochondrial membrane system. CBP6, a yeast nuclear genenecessary for synthesis of cytochrome b. /. Biol. Chem. 260, 1513-1520.

Drlever, W. and Nusslein-Volhard, C. (1988). A gradient of bicoidprotein in Drosophila embryos. Cell 54, 83-93.

Erd l̂yi, M. and Szabad, J. (1989). Isolation and characterization ofdominant female sterile mutations of Drosophila melanogaster. I.Mutations on the third chromosome. Genetics 122, 111-127.

Furie, B. and Furie, B. C. (1988). The molecular basis of bloodcoagulation. Cell 53, 505-518.

Harlow, E. and Lane, D. (1988). Antibodies: A Laboratory Manual.Cold Spring Harbor Press, Cold Spring Harbor.

Hashimoto, C , Gerttula, S. and Anderson, K. V. (1991). Plasmamembrane localization of the Toll protein in the syncytialDrosophila embryo: importance of transmembrane signaling fordorsal-ventral pattern formation. Development 111, 1021-1028.

Hashimoto, C , Hudson, K. L. and Anderson, K. V. (1988). The Tollgene of Drosophila, required for dorsal-ventral embryonic polarity,appears to encode a transmembrane protein. Cell 34, 269-279.

Jin, Y. (1991). Generating dorsal-ventral asymmetry in theDrosophila embryo: studies on the easter gene and the polarizingactivity. Ph.D. Thesis. University of California, Berkeley.

Jin, Y. and Anderson, K. V. (1990). Dominant and recessive alleles ofthe Drosophila easter gene are point mutations at conserved sites inthe serine protease catalytic domain. Cell 60, 873-881.

Konrad, K. and Marsh, L. (1990). 31st Annual Drosophila Res.Conf. Abstr. 28.93.

Konrad, K. D., Goralski, T. J. and Mahowald, A. P. (1988).Developmental genetics of the gastrulation defective locus inDrosophila melanogaster. Dev. Biol. 127, 133-142.

Krieg, P. A. and Melton, D. A. (1987). In vitro RNA synthesis withSP6 RNA polymerase. Meth. Enzymol. 155, 397-415.

Krishnaswamy, S. (1990). Prothrombinase complex assembly.Contributions of protein-protein and protein-membraneinteractions toward complex formation. /. Biol. Chem. 265, 3708-3718.

MacDonald, M. E., van Zonneveld, A. J. and Pannekoek, H. (1986).Functional analysis of the human tissue-type plasminogen activatorprotein: the light chain. Gene 42, 59-67.

Manseau, L. and SchUpbach, T. (1989). cappuccino and spire: twounique maternal-effect loci required for both the anteroposterior

and dorsoventral patterns of the Drosophila embryo. Genes Dev. 3,1437-1452.

Pirrotta, V. (1988). Vectors for P-mediated transformation inDrosophila. Biotech. 10, 437-456.

Roth, S., Stein, D. and Nusslein-Volhard, C. (1989). A gradient ofnuclear localization of the dorsal protein determines dorsoventralpattern in the Drosophila embryo. Cell 59, 1189-1202.

Rushlow, C. A., Han, K., Manley, J. L. and Levine, M. (1989). Thegraded distribution of the dorsal morphogen is initiated by selectivenuclear transport in Drosophila. Cell 59, 1165-1177.

Sanger, F., NIcklen, S. and Conlson, A. R. (1977). DNA sequencingwith chain-terminating inhibitors. Proc. Natl. Acad. Sci. USA 74,5463-5467.

SchUpbach, T. and Wieschaus, E. (1989). Female sterile mutations onthe second chromosome of Drosophila melanogaster. I. Maternaleffect mutations. Genetics 121, 101-117.

Seifert, E., MMler-Holtkamp, F., Marcey, D. and Jackie, H. (1987).Injection of wild-type cytoplasm and poly (A)+ RNA provokesphenotype rescue in spUtzle mutant Drosophila embryos. Roux'sArch. Dev. Biol. 196, 78-82.

Stein, D. and Nusslein-Volhard, C. (1992). Multiple extracellularactivities in Drosophila egg perivitelline fluid are required forestablishment of embryonic dorsal-ventral polarity. Cell 68, 429-440.

Stein, D., Roth, S., Vogelsgang, E. and Nusslein-Volhard, C. (1991).The polarity of the dorsoventral axis in the Drosophila embryo isdefined by an extracellular signal. Cell 65, 725-735.

Steward, R. (1989). Relocalization of the dorsal protein from thecytoplasm to the nucleus correlates with its function. Cell 59, 1179-1188.

Steward, R., Zosman, S. B., Huang, L. H. and Schedl, P. (1988). Thedorsal protein is distributed in a gradient in early Drosophilaembryos. Cell 55, 487-495.

Stroud, R. M., KossiakofT, A. A. and Chambers, J. L. (1977).Mechanisms of zymogen activation. Ann. Rev. Biophys. Bioeng. 6,177-193.

Tate, K. M., Hlggins, D. L., Holmes, W. E., Winkler, M. E.,Heyneke, H. L. and Vehar, G. A. (1987). Functional role ofproteolytic cleavage at arginine-275 of human tissue plasminogenactivator as assessed by site-directed mutagenesis. Biochem. 26,338-343.

Tearle, R. and NUssldn-Volhard, C. (1987). Tubingen mutants andstock list. Dros. Inf. Serv. 66, 209-269.

Towbin, H., Staehelln, T. and Gordon, J. (1979). Electrophoretictransfer of proteins from polyacrylamide gels to nitrocellulosesheets: procedure and applications. Proc. Natl. Acad. Sci. USA 76,4350-4354.

Vasquez, J. R., Evnin, L. B., Hlgaki, J. N. and Cralk, C. S. (1989).An expression system for trypsin. J. Cell. Biochem. 39, 265-276.

Wieschaus, E. and Nusslein-Volhard, C. (1986). Looking at embryos.In Drosophila: A Practical Approach, pp. 199-227. (ed. D. B.Roberts). Oxford: IRL.

(Accepted 24 February 1992)