Development 109, 967-973 (1990)Printed in Great Britain © T h e Company of Biologists Limited 1990

967

Identification in Xenopus of a structural homologue of the Drosophila gene

Snail

MICHAEL G. SARGENT* and MICHAEL F. BENNETT

Laboratory of Embryogenesis, National Institute for Medical Research, Mill Hill, London, NW71AA, UK

• Author for correspondence

Summary

We have cloned a Xenopus cDNA that is related to snail,a gene that is required for mesoderm formation inDrosophila. The cDNA encodes a protein that containsfive zinc-fingers that closely resemble those of snail. Inthe non-canonical parts of the DNA-binding loop, thereis almost 90 % homology between snail and xsna. Thecorresponding mRNA (xsna) is expressed strongly at thestart of zygotic transcription simultaneously with thetranscription factor EFla. In early gastrulae, xsna isequally distributed between the dorsal and ventralhalves of the equatorial zone. The possibility that thecapacity to synthesise xsna is more localised before thestart of zygotic transcription has been investigated by

culturing fragments of stage 8 embryos until xsna issynthesised. The capacity to synthesise xsna at stage 8 islocated principally in the dorsal half of the equatorialzone. A small amount of maternal xsna is localised in thevegetal hemisphere before zygotic transcription starts.xsna is not present in isolated animal caps but can beinduced by the mesoderm-inducing factors XTC-MIFand bFGF. Synthesis of xsna does not occur auton-omously in dispersed cells but is restored when cellsreaggregate in the presence of calcium and magnesium.

Key words: snail, Xenopus, zygotic transcription,mesoderm-inducing factors, zinc-finger proteins.

Introduction

Two distinct systems generate the anterior-posteriorand dorsal-ventral axes of Drosophila embryos(Anderson, 1990; Levine and Harding, 1990). Some ofthe zygotic gene products in the anterior-posteriorsystem of Drosophila such as homeobox and gap geneshave structural homologues in vertebrates that are alsoconcerned functionally with anterior-posterior organ-isation (Gaunt, 1988; Wilkinson et al. 1989). Just assegmentation in insects and vertebrates was probablyestablished independently, we presume these proteinsacquired their role independently. However, structuraland functional conservation, such as this, suggests theseproteins may have originated in the extremely ancientcommon ancestors of insects and vertebrates.

Only three known vertebrate genes have structuralsimilarities to members of the Drosophila dorsoventralsystem. The Xenopus protein xtwi has high homology totwist, a zygotic Drosophila gene that is positivelyregulated by the maternal gene dorsal (Hopwood et al.1989). c-rel, a chicken oncogene is similar to dorsal(Steward, 1987) and the vertebrate growth factor familyTGF-/Jhas similarities to decapentaplegic, a Drosophilazygotic gene that is negatively regulated by dorsal(Padgett et al. 1987). Here we report the identificationin Xenopus of a structural homologue of snail another

zygotic Drosophila gene which like twist is positivelyregulated by dorsal and specifies ventral pattern, snailencodes a protein with five zinc-finger domains, whichis expressed in the ventral furrow and mesoderm ofDrosophila and is required for mesoderm formation(Simpson, 1985; Boulay et al. 1987). The Xenopushomologue of snail (xsna), described here, has a highlyconserved zinc-finger region, and is also associated withsites of mesoderm formation.

Mesoderm formation in amphibia is governed bysignals that emanate from the vegetal hemisphere whichinduce formation of characteristically mesodermal celltypes such as muscle. Using animal cap explants as anassay two classes of inducing factors have been ident-ified. These are the transforming growth factor fi(TGF#) family, the most potent of which is a homol-ogue of activin A (Smith et al. 1990) and the fibroblastgrowth factor (FGF) (Slack et al. 1987; Kimelman andKirschner, 1987). At least three mesoderm-inducingsignals have been operationally identified in earlyembryos of Xenopus (Slack et al. 1989; Smith et al.1989). Two of these signals are thought to emanate fromthe vegetal hemisphere. The signal from the dorsalvegetal region (signal 1) induces characteristically dor-sal mesodermal tissues (notochord and muscle) whilethe signal from the ventral vegetal region (signal 2)induces ventral tissues like mesenchyme and blood

968 M. G. Sargent and M. F. Bennett

(Dale and Slack, 1987). The third signal is produced bynewly induced dorsal mesoderm cells and instructsventral mesoderm to develop dorsal characteristics(Dale and Slack, 1987). The effects of XTC-MIF (nowknown to be a homologue of activin A; Smith et al.1990) and bFGF on animal cap explants suggest thatthese two substances mimic the dorsal and ventralsignals, respectively (Smith et al. 1989; Slack et al.1989). No candidate molecule for the third signal hasyet been identified. The precise relationship betweenthe known factors and the operationally identifiedsignals is, however, quite confused as there may be anumber of rather similar candidate molecules. Further-more, bFGF, which is present in blastulae has no signalsequence (Abraham et al. 1986) and therefore noapparent mechanism of secretion. We show below thatxsna is induced in animal caps by mesoderm-inducingfactors.

Materials and methods

Identification of a Xenopus homologue of snailA stage 17 library in phage lambda was screened at lowstringency (Kintner and Melton, 1987) using the Drosophilasnail cDNA as a probe (Boulay et al. 1987). Inserts of theseclones were subcloned into pSP72 (Kreig and Melton, 1987)and M13 and were sequenced by the dideoxy procedure. The

nucleotide sequence data reported will appear in the EMBL,GenBank and DDBJ Nucleotide Sequence databases underthe accession number X53450.

Dissection of embryos, handling of explantsXenopus eggs were artificially fertilized and dejellied bystandard methods. Stages were according to Nieuwkoop andFaber (1969). Embryos were dissected using tungsten needlesand the fragments were cultured in full strength NAM on 1 %agarose (Slack, 1984; Smith, 1987). Prospective DMZ andVMZ at stage 8 could be dissected accurately in embryos withclear polarity in pigmentation (chosen at the 2-cell stage). Theventral side was distinguished from the dorsal side by greaterpigmentation and larger cell size. For the dissections de-scribed in Fig. 4A, embryos were arranged with their ventralside to the left and their dorsal to the right. A square of animalcap was then excised such that one diagonal separated ventraland dorsal sides precisely into two equal parts. Vertical cutswere then made at each end of this diagonal to separate DMZand VMZ. Marginal zone material was then removed fromthe vegetal pole region by cutting round the yolk mass asshown in Fig. 4A. Explants cultured to the equivalent of stage20 gave more than 90 % of the characteristic phenotype of thetwo kinds of explants (i.e. DMZ explants elongated axially,while VMZ explants were spherical as described by Dale andSlack (1987). We note that Dale and Slack (1987) believe thatthe characteristic 'dorsalizing activity' in the DMZ is confinedto less than 90° of the circumference of the equator. Ourdissections, therefore, are relatively precise as the limits of theDMZ are substantially wider than this. At stage 10.5, VMZ

30 60 90 120

MetProArgSerPheLeuValLyeLyBHisPheSerAlaSerLyaLyaProAsnTyrSerGluLeuGluSerClnThrValTyrlleserProPhelle150 180 210 240

TyrAspLysPheProVflllleProGlnProGluIleLeuSerThrGlyAlaTyrTyrThrProLeuValTrpAspThrGlyLeuLeuThrThrPhaPheThrSerGluserAspTyrLys270 300 330 360

LysSerProIleSerProSerSerSerAspAspSerSerLysProLeuAspLeuThrSerPheSerSerGluAspGluGlyGlyLyaThrSerAspProProSarProAlaSerSerAla

390 iii £££ 480

ThrGluAlaGluLycPheGlnCysAsnLeuCysSerLysSerTyrSerThrPheAlaGlyLeuSerLysHiBLyBGlnLeuHisCysAspSarGlnThrArgLyBSerPheSerCysLys540 570 600

TyrCyaGluLysGluTyrValSerLauGlyAlaLeuLy3MetHi9lleArgSerHi8ThrLeuProCysValCysLysIleCy8GlyLy8AlaPheSerArgProTrpLeuLeuGlnGly630 660 690 720

HisIleArgThrHiBThrGlyGluLyaProPheSerCysThrHigCyaABnArgAlaPheAlaAflpArgSerAsnLeuArgAlaHiaLeuGlnThrHiaserABpValLyaLyBTyrGln750 780AGAATGTCCCTCCTTCACAAGCACGAGGAAACCG

CyaLygSorCysSerArgThrPheSerArgMetSerLeuLeuHisLyBHIBGluGluThrGlyCyBThrValAlaHis870 900 930 960

ATATTTTTTGGCAGGATATATATAAATATATATATTATTTTTTTTTTCTTTTTTAAATGATGCCATTGCTAAATTTGTT^

990 1020 1050 1080

1110 1140 1200

1230 1260 1290 1320TCTCCAATCGTCTATCCAAAAAGTGTATTAAAAGTTTGTGTGCCAACTTTTATTACACTTTTTCTGTCCGTATATTTT

1380 1440

1470 1500 1530 1560

1590 1620 1650 1680TAGACTATTTATOTCTTCTAAGAGGTCKCAAGTGCATGCGTCTTTTTTTTTTCATAATTTTTTTTAGCTAACCCAATCCAAGAATC

1710 1740 1770 1800TTTTTATAAAAGTGCCACATrTTGTACTGAGGATTCTATGTTTrTATAGTTATGCACATTTATTGATATTCAATAAAATGTTTTTT^TTCATCCGTAGCCGGGTACCGAGCTCAGAATTC

Fig. 1. DNA Sequence and deduced translation of xsna. Boxed area, zinc finger domain.

and DMZ were identified unequivocally using the position ofthe blastopore lip.

Animal caps were dissected and treated with mesoderm-inducing factors as described by Smith (1987).

Single cells were obtained by culturing embryos in calcium/magnesium-free medium (CMFM: Sargent et al. 1986), fromstage 2. These were demembranated at about stage 6 (64-cellstage), dispersed on agarose and cultured as single cells untilthe equivalent of stage 13 (3h after stage 10.5). To initiatereaggregation, calcium and magnesium chloride (IIDM) wereadded to single cell cultures at the equivalent of stage 10.5.The cells were swirled into a heap to facilitate cell-to-celladhesion and were cultured for a further 3h.

Nuclease protection studiesProbes for xsna were antisense RNA made using SP6 RNApolymerase from nucleotide 692 (Pst) to 529 (Msp) of xsna(Fig. 1) in pSP72 (Kreig and Melton, 1987) and gave aprotected fragment of 163 bp. EFla" (which is expressed in allembryonic cells when zygotic transcription starts and there-fore serves as a control for RNA recovery) (Kreig et al. 1989)was probed using antisense RNA from nucleotides 194 to 100(Hinf) prepared in pSP72 using SP6 RNA polymerase. Twobands of about this size were always seen and probablyindicate the existence of two variants of the EFla-sequence inthe Xenopus genome. The protected fragment was 94 bp.UTP was used at a specific activity of 800 or 160 /*Ci per nmolefor xsna and EFlix, respectively. Solution hybridisation of thetwo probes was carried out at 45 °C for 16 h with the samesample to minimise differences in recovery of the two pro-tected fragments essentially as described by Kreig and Melton(1987). Unhybridised RNA was digested with RNAse A andRNAse Tl. Autoradiograms were usually exposed for 4 days.Most data shown were derived from one third of a sample of10 caps.

Results

Isolation of a Xenopus cDNA related to the snail geneo/DrosophilaA stage 17 library of Kintner and Melton (1987) wasscreened at low stringency using Drosophila snail as aprobe. Sequence data from one purified positive cloneindicated that we had identified a zinc-finger protein.This clone was used to rescreen the library for a full-

xsna-Xenopus snail 969

length clone, which was then subcloned and sequenced.A region of high homology with snail was found in anopen reading frame that began at the first Met codon.The deduced amino acid sequence of this protein,called xsna, (259 amino acids) is shorter than snail, (390amino acids). However, the full-length cDNA clonewas about the same size as the single transcript ob-served in Northern blots (1.8kb) (see below). Thefinger regions of snail and xsna have an almost identicallayout (Fig. 2) and are 66 % homologous in this regionoverall. Moreover the non-canonical parts of the fingermotif are 46% homologous overall, increasing to 90%in the third finger. The % identity of each sectionintervening between the canonical elements of theprotein are shown in Fig. 2. Sections D and E, whichare likely to be in the centre of the cr-helical DNA-binding region (Gibson et al. 1988), are 84 and 90%identical. The sequence TGEKP, present in the thirdfinger of both proteins, however, is found in manyfinger proteins (Gibson et al. 1988; Nietfeld et al. 1989).Although there are many published zinc-finger se-quences, we are aware of no sequence that approachesthis level of similarity to snail or xsna. We thereforebelieve that xsna is likely to be a true homologue ofsnail. Outside of the finger region, xsna has no signifi-cant extended homology to snail although there are afew small islands of similarity.

Expression of xsna mRNA in developing embryosThe entire cDNA was used as a probe to examine xsnamRNA accumulation by Northern blot analysis. Onlyone transcript of about 1.8 kb was found. This was firstobserved at stage 10, increased to a plateau from stage12 to 18, and decreased to some extent after stage 25although the mRNA persisted until at least stage 40(data not shown). A more sensitive study of the earlytime course of xsna accumulation was made usingRNAse protection. The rate of synthesis of xsna wascompared with the rate of synthesis of translationelongation factor EFlar. This mRNA is synthesised assoon as zygotic transcription starts in all embryonic cells(Kreig et al. 1989). It provides an internal standard for

Xsna

Snail

Nan canonicalsactlon

o/feldonllty

F

F

F

H

C

C

F

F

Y

Y

Q

K

S

S

V

K

S

Q

Q

A

c

c

c

C

C

C

c

c

c

c

NL

DE

KY

EE

KI

PI

TH

PD

KS

QV

c

c

c

c

c

c

c

c

c

c

SKS

QKM

EKE

GKL

GKA

GKA

NRA

PRS

SRT

HKS

Y

Y

Y

Y

F

F

F

F

F

F

STFAG

STSMG

VSLGA

TTIGA

SRPWL

SRPWL

ADRSN

ADRSN

SRMSL

SRMSL

L

L

L

L

L

L

L

L

L

L

SK

SK

KM

KM

QG

QG

RA

RA

HK

NK

H

H

H

H

H

X X

X

X

X

KQL

RQF

IRS

IRT

IRT

IRT

LQT

QQT

EETG

SSSN

XX

X

H

H

H

H

H

C

C

CD

CPAAE

TLP

TLP

TGEKP

TGEKP

SDVKK

VDVKK

TVAH

TITIA

SQTRKS

A

20

B

10

C

40

D

80

E

SO

F

SO

G

64

Fig. 2. Relationship with Drosophila snailamino acid sequence shown in single lettercode. Canonical features of finger proteins inboxes. Identical amino acids (:), conservativesubstitutions (.). Non-canonical regions A-Gwith % identity.

970 M. G. Sargent and M. F. Bennett

comparison of zygotic mRNAs. As EFlir is extremelyabundant we have prepared the probe for this mRNAusing UTP of one fifth of the specific activity used forxsna, furthermore the protected fragment for EFlcHsabout one half of the size of xsna so that the relativeabundance of xsna appears about 10 times its trueabundance relative to EFla. The RNAse protectionassay shows the two mRNAs increase simultaneouslyjust after stage 9 (Fig. 3B). The great sensitivity of thismethod of assaying mRNA reveals that there is also alow level of mRNA of both genes present before thestart of zygotic transcription that we presume to be ofmaternal origin.

Oo Egg 6 8 10 12 14 16 18

B

7 8 9 10-25 Stage

3-5 4-5 5-5 6 65 7 7-5 9 hrs

••"••* Xsna

EFIot

Fig. 3. (A) Northern Blot using RNA preparations fromoocytes (Oo), eggs and stages 6-18. (B) Nucleaseprotection study of xsna development in Xenopus embryos.Position of protected fragments shown at side.

Xsna

Vegetal PoleEF1«

B

Localisation of xsna within the embryoThe location of xsna transcripts within early embryoscan be determined quantitatively and with considerablesensitivity using RNAse protection to assay mRNA indissected parts of the embryo. Sets of 10 early gastrulae(stage 10.5) were dissected into animal caps, dorsal andventral marginal zones (DMZ and VMZ respectively),and vegetal pole regions (Fig. 4A) from which RNAwas extracted for RNA protection assays. DMZ andVMZ could be unequivocally identified by orientationrelative to the blastopore lip. The bulk of new synthesisof xsna mRNA after the start of zygotic transcription isin the marginal zone. The amount of xsna RNA in theDMZ and VMZ at stage 10.5 is not significantlydifferent, xsna is almost completely absent from theanimal cap although EFIOMS present (Fig. 4B, i-1). Inthe vegetal zone, there is a significant enrichment ofxsna relative to EFlir (determined by densitometry,data not shown) compared with the marginal zonealthough only a small proportion of the total xsnamRNA is in this region (Fig. 4B, 1).

A small amount of xsna is present in embryos beforezygotic transcription starts. Its distribution in dissectedfragments at stage 8 is shown in Fig. 4B (a-d). Morethan 80% of xsna mRNA at this stage is located in thevegetal hemisphere. Dissections into vegetal and ani-mal pole at earlier stages gave similar results (data notshown).

By the time zygotic transcription starts, it is possiblefor an initial focus of xsna synthesis to be obscured byinduction processes. This possibility was investigated bydissecting embryos at stage 8 before xsna synthesis hadstarted and then culturing these explants until xsna wasfully expressed. Prospective DMZ and VMZ weredissected using pigmentation as a marker as describedabove. The amount of xsna in these explants was takenas an indication of the capacity to synthesise xsna atstage 8 in the dissected regions. The explants werecultured to the equivalent of stage 12 to compensate forthe slight reduction in rate of synthesis of EFla inexplants compared with intact embryos. At stage 12 the

f i ) I

Fig. 4. (A) Dissection of embryos at stage 8 and 10.5. Animal caps, vegetal pole, dorsal marginal zone (DMZ) and ventralmarginal zone (VMZ) are marked. (B) Location of xsna in early embryos and development of capacity to express xsna.Embryos dissected as in Fig. 4A. (a-d) Stage 8 embryos; (a) AC (b) DMZ (c) VMZ (d) VG. (e-h) Cultured to theequivalent of stage 12; (e) AC, (f) prospective DMZ, (g) prospective VMZ, (h) VG. (i-1) Embryos dissected and processedat stage 10.5; (i) AC (j) DMZ (k) VMZ (1) VG.

xsna-Xenopus snail 971

DMZ typically contained 4 times more xsna than VMZexplants (Fig. 4B, e-h) while the amount of EFla inDMZ and VMZ was essentially identical. In contrast, inembryos dissected at the start of gastrulation (stage10.5) there is no indication of preferential localisation inthe DMZ. (Fig. 4B, j -k) . The capacity to synthesisexsna is, therefore, localised in the DMZ but not theVMZ before zygotic transcription starts but this soonspreads to the entire marginal zone once zygotic tran-scription starts.

Effect of mesoderm-inducing factors on xsna synthesisin isolated animal capsThe presence of xsna in the marginal zone and absencefrom the animal cap of early gastrulae and also theeffect of DMZ on VMZ noted above suggests that xsnamRNA synthesis may be regulated by mesoderm-inducing factors. This was investigated by culturing setsof stage 8 animal caps in the presence of the mesoderm-inducing factors XTC-MIF and bFGF at 9u.ml"1 and9.9u.ml"1 respectively (equivalent to 1.8ng and Ingrespectively of standard pure preparations of the twofactors; Smith, 1987; Slack et al. 1987). At theseconcentrations animal caps are completely induced bymorphological criteria, but would not normally give themost extreme tissue-type expression. RNA wasextracted from a set of 10 caps at intervals. Both XTC-MIF and bFGF induce xsna significantly at the equival-ent of stage 10.5 and strongly at stage 12 whereascontrol caps have none (Fig. 5). The presence of EFlcvin all three sets of caps after stage 10.5 shows that totalRNA synthesis is similar in all three samples.

Dependence of xsna synthesis on cell-to-cell contactIn Xenopus embryogenesis, some gene products areexpressed in a cell autonomous fashion while othersrequire cell contact (Sargent et al. 1986). To distinguish

b c f 8Xsna

EFK

Fig. 5. Time course of xsna induction in isolated stage 8animal caps, (a-d) Animal caps cultured without additionsto the stage equivalents shown; (a) stage 8, (b) stage 10.5(c) stage 12, (d) stage 15. (e-g) Animal caps cultured in thepresence of XTC-MIF (9u.ml~'); (e) stage 10.5, (f) stage12, (g) stage 15. (h-j) Animal caps cultured in the presenceof bFGF (9.9 u.ml"'); (h) stage 10.5, (i) stage 12, (j) stage15.

Fig. 6. Effect on synthesis ofxsna of addition of calcium and

Xsna magnesium to embryonic cellscultured as single cells,(a) Embryos in CMFM fromstage 2 to stage 6 weredemembranated, and cultured assingle cells on agarose untilequivalent of stage 13 (3 h afterstage 10.5). (b) As for a exceptthat calcium and magnesiumchloride (1 ITIM) were added at

EF1<* the equivalent of stage 10.5. Thecells were swirled into a heap tofacilitate cell-to-cell adhesion andwere cultured for a further 3h.

between these alternatives, the rate of synthesis of xsnahas been examined in dispersed embryonic cells cul-tured in media lacking calcium and magnesium(CMFM, see methods) or after aggregation has beeninitiated by restoration of calcium and magnesium(Fig. 6). When dispersed cells were cultured in CMFM,there was no significant synthesis of xsna while EFlacontinued to be synthesised until the equivalent of stage13. Addition of calcium and magnesium at any timeafter stage 10 restored synthesis of xsna and thiscorrelated with aggregation of the cells. On restorationof calcium and magnesium, xsna synthesis took placeover a two hour period (data not shown).

Discussion

We have described a Xenopus mRNA, xsna, thatencodes a protein that is very similar to Drosophila snailin the putative DNA-binding domain that contains fivezinc-fingers. The finger regions of xsna and snail re-semble each other more closely than either resembleany of the many zinc-finger containing proteins thatexist (Gibson etal. 1988; Nietfeld et al. 1989). Althoughthe possibility of the two genes evolving independentlycannot be excluded, it seems more likely that the DNA-binding regions of xsna and snail are truly homologous,an observation that adds to the increasing body ofinformation that suggests that the genetic systems thatdetermine the two major axes of Drosophila are evol-utionarily conserved in vertebrates and may be thelegacy of an extremely ancient common ancestor. Onemay speculate that homology in the DNA-bindingregion and lack of it in the remainder of the proteinindicates that the conserved function is the capacity toregulate specifically the target genes of these putativetranscriptional activators. It may be significant too, thatalthough the anatomy of regions of mesoderm forma-tion in Drosophila and Xenopus is dissimilar, theputative Xenopus homologues of twist and snail (xtwiand xsna) are associated with these sites in Xenopus(Hopwood et al. 1989). Conversely, xsna is not foundsolely on the ventral side of the embryo as snail is, inDrosophila.

At stage 10.5, the bulk of both xsna and EFla- was

972 M. G. Sargent and M. F. Bennett

localised in the marginal zone. We were a little sur-prised to discover that EFlcrwas not more abundant inthe vegetal zone as this region probably represents atleast 30 % of the volume of the embryo. There has beenvery few investigations of the location of housekeepingmRNAs in early embryos of Xenopus. However, anearly autoradiographic study by Bachvarova et al.(1966) indicated that newly synthesised mRNA waslocated predominantly in the equatorial region of blas-tulae at the start of zygotic transcription and in thedeveloping mesoderm in stage 10.5 gastrulae. An RNAwith a maternal component would probably have beenmore abundant in the vegetal region and would there-fore have given a false impression of the specificity oflocation of xsna. We therefore believe that a zygotichousekeeping mRNA such as EFlo-is the most appro-priate internal standard, notwithstanding that its distri-bution is similar to an inducible mRNA such as xsna.

Although only a small part of the xsna mRNA atstage 10.5 is located in the vegetal region, it is moreenriched relative to EFlc than in the marginal zone.This may, in part, be due to the maternal contributionfound in vegetal cells at stage 8. However, if, as wesuggest below, xsna is synthesised in response to in-ducing factor emanating from the vegetal pole, the siteof xsna synthesis may be contiguous with the vegetalpole and would therefore tend to enrich this fraction.This is currently being investigated by in situ hybridis-ation.

The behaviour of xsna at the start of zygotic tran-scription can largely be explained in terms of thecurrent view of the three-signal model of mesoderminduction as developed by Slack et al. (1989) and Smithet al. (1989) (see Introduction). The relative abundanceof xsna in the DMZ rather than in VMZ explantscultured from stage 8 to 12 indicates that the initialcapacity to synthesise xsna is located in the same regionas the source of the dorsovegetal signal (signal 1). Inview of the inducibility of xsna by XTC-MIF, whichmimics the dorsal signal when applied to animal caps,we infer that the dorsovegetal signal controls xsna and isactive before zygotic transcription starts in establishingthe capacity to synthesise xsna in contiguous tissues.

In intact embryos the amount of xsna in the DMZand VMZ is equal by stage 10.5. We believe that thisshows that xsna expression in the VMZ is increased as aresult of its proximity to the DMZ. This is reminiscentof the 'dorsaJisation' of tissues to the ventral sideeffected by the blastopore lip region (Smith and Slack,1983), and may indicate that xsna expression in theVMZ is controlled by the third 'signal'. The possibilitythat dorsalisation is achieved by cell migration from theDMZ to the VMZ is thought to be unlikely (Smith andSlack, 1983).

The rate of synthesis of xsna in VMZ explants is verylow compared with DMZ explants. This suggests thatsignal 2 which emanates from the ventral-vegetal region(Slack et al. 1989) does not induce xsna. Paradoxically,FGF, which is the best candidate at present for signal 2,is able to induce xsna in animal caps. We have noexplanation for this, but we note that Slack et al. (1989)

have recognised that bFGF has no signal sequence(Abraham et al. 1986), and no apparent mechanism ofsecretion, and therefore may not be signal 2 in vivo. Asmore inducing factors are discovered the paradoxshould be resolved. We would predict that the real invivo ventral mesoderm inducer would not induce xsnain animal caps.

Induction of xsna in animal caps by the two meso-derm-inducing factors and the inferred response to thethird dorsalising signal suggests that synthesis of xsnacan be a response to more than one signal. Furthermorethe timing of the xsna response in animal caps suggeststhat induction is not an 'immediate early' response tomesoderm-inducing factors as with Mix.l (Rosa, 1989)but takes place over several hours. There is no contra-diction between this observation and the observed rapidonset of synthesis at the start of zygotic transcription, asthe tissues that respond probably have developed thecapacity to produce xsna at an earlier time.

The dependence of xsna synthesis on cell-to-cellcontact and the lack of autonomous expression of xsnais consistent with its inducibility by mesoderm-inducingfactors. Thus the capacity to synthesise xsna could belost from dispersed cells when a diffusible factor be-comes too dilute. In addition, there may be a require-ment for cell-to-cell contact or a community effect as foractin in muscle differentiation (Gurdon et al. 1985;Gurdon, 1988) or even a specific requirement fordivalent metal ions. These alternatives are being inves-tigated.

Whilst there is no real anatomical homology betweenDrosophila and Xenopus early embryos, the presenceof the structurally homologous snail and xsna at sites ofinvolution and of presumptive mesoderm formation inearly embryos supports the idea that the system deter-mining dorsoventral pattern identified in Drosophilaexists in vertebrates in a greatly modified form.

We are indebted to A. Alberga who provided us with thesnail cDNA and Jim Smith who provided the mesoderm-inducing factors and to our colleagues Jonathan Cooke, JimSmith, Jeremy Green, Linda Essex and Peter Vize for theircritical comments, and to Marilyn Brennan who prepared themanuscript for publication.

References

ABRAHAM, J. A., MERGIN, A., WHANG, J. L., TUMOTO, A.,FRIEDMAN, J., HJERRILD, K. A., GOSPODAROWICZ, D. ANDFIDDES, J. C. (1986). Nucleotide sequence of a bovine cloneencoding the angiotensic protein, basic fibroblst growth factor.Science 233, 545-548.

ANDERSON, K. V. (1990). Drosophila: the maternal contribution. InGenes and Embryos, (ed. D. M. Glover and B. D. Hames.) pp.1-38. IRL Press.

BACHVAROVA, R., DAVIDSON, E. H., ALLFREY, V. G. AND MIRSKY,A. E. (1966). Activation of RNA synthesis associated withgastrulation. Proc. Naln. Acad. Sci. U.S.A. 55, 358-365.

BOULAY, J. C , DENNEFIELD, C. AND ALBERGA, A. (1987). TheDrosophila developmental gene snail encodes a protein withnucleic acidbinding fingers. Nature 330, 395-398.

DALE, L. AND SLACK, J. M. W. (1987). Regional specificationwithin the mesoderm of early embryos of Xenopus laevis.Development 100, 279-295.

xsna-Xenopus snail 973

GAUNT, S. J. (1988). Mouse homeobox gene transcripts occupydifferent but overlapping domains in the embryonic germ layersand organs; a comparison of Hox-3.1 and Hox-1.5. Development103, 131-144.

GIBSON, T. J., POSTMA, J. P. M., BROWN, R. S. AND ARCOS, P.

(1988). A model for the tertiary structure of the 28 residuesDNA-binding motif (zinc-finger) common to many transcriptionalregulatory proteins. Protein Engineering 20, 209-218.

GURDON, J. B. (1988). Cell movements and a community effect intissue morphogenesis. Nature 336, 772-774.

GURDON, J. B., MOHUN, T. J., FAIRMAN, S. AND BRENNAN, S.

(1985). All components necessary for the eventual activation ofmuscle specific actin genes are localised in the subequatorialregion of an uncleaved amphibian egg. Proc. Natn. Acad. Sci.USA 82, 139-143.

HOPWOOD, N. D., PLUCK, A. AND GURDON, J. B. (1989). A

Xenopus mRNA related to Drosophila twist is expressed inresponse to induction in the mesoderm and the neural crest. Cell59, 893-903.

KIMELMAN, D. AND KIRSCHNER, M. (1987). Synergistic induction ofmesoderm by FGF and TGF-/3 and the identification of anmRNA coding for FGF in the early embryo. Cell 51, 869-877.

KINTNER, C. R. AND MELTON, D. A. (1987). Expression of XenopusN-CAM RNA is an early response of ectoderm to induction.Development 99, 311-325.

KREIG, P. A. AND MELTON, D. A. (1987). Synthesis of RNA invitro using SP6 RNA polymerase. Methods in Enzymology 155,397^115.

KREIG, P. A., VARNUM, S. M., WORMINTON, W. M. AND MELTON,

D. A. (1989). The mRNA encoding elongation factor l-c^EF-la-)is a major transcript at the midblastula transition in Xenopus.Devi Biol. 133, 93-100.

LEVINE, M. S. AND HARDING, K. W. (1990). Drosophila: the zygoticcontribution. In Genes and Embryos, (ed. D. M. Glover and B.D. Hames), pp. 39-94. IRL Press.

NIETFELD, W., EL-BARADI, T., MENTZL, H., PIELER, T., KOSTER,

M., POTING, A. AND KNOCHEL, W. (1989). Second order repeatsin Xenopus laevis finger proteins. J. molec. Biol. 208, 639-659.

NIEUWKOOP, P. D. AND FABER, J. (1969). Normal Table of Xenopuslaevis. 2nd Ed. Amsterdam, North Holland.

PADOETT, R. W., JOHNSTON, R. D. S. AND GELBART, W. M. (1987).

The decapentaplegic gene complex of Drosophila encodes aprotein homologous to transforming growth factor-/? gene family.Nature 325, 81-84.

ROSA, F. M. (1989). Mix.l, a homeobox mRNA inducible bymesoderm inducers, is expressed mostly in the presumptiveendodermal cells of Xenopus embryos. Cell 57, 965-974.

SARGENT, T. D., JAMJUCH, M. AND DAWID, I. B. (1986). Cell

interactions and the control of gene activity during earlydevelopment of Xenopus laevis. Devi Biol. 114, 238-246.

SIMPSON, P. (1985). Maternal zygotic interactions during formationof the dorso-ventral pattern in Drosophila embryos. Genetics105, 615-632.

SLACK, J. M. W. (1984). Regional biosynthetic markers in the earlyamphibian embryo. J. Embryol. exp Morph. 80, 289-319.

SLACK, J. M. W., DARLINGTON, B. G., HEATH, J. K. AND GODSAVE,

S. F. (1987). Mesoderm induction in early Xenopus embryos byheparin binding growth factors. Nature 326, 197-200.

SLACK, J. M. W., DARLINGTON, B. G., GILLESPIE, L. L., GODSAVE,

S. F., ISAACS, H. V. & PATERNO, G. D. (1989). The role offibroblast growth factor in early Xenopus development.Development, Supplement. The Molecular Basis of PositionalSignalling, pp. 141-148.

SMITH, J. C. (1987). A mesoderm-inducing factor is produced by aXenopus cell line. Development 99, 3-14.

SMITH, J. C , COOKE, J., GREEN, J. B. A., HOWES, G. AND SYMES,

K. (1989). Inducing factors and the control of mesodermalpattern in Xenopus laevis. Development, Supplement. TheMolecular Basis of Positional Signalling, pp. 149-155.

SMITH, J. C , PRICE, B. M. J., VAN NIMMEN, K. AND HUYLEBROECK,

D. (1990). Identification of a potent Xenopus mesoderm-inducingfactor as a homologue of activin A. Nature (in press).

SMITH, J. C. AND SLACK, J. M. W. (1983). Dorsalisation and neuralinduction: properties of the organiser in Xenopus laevis. J.Embryol. exp Morph. 78, 299-317.

STEWARD, R. (1987). dorsal an embryonic polarity gene ofDrosophila is homologous to the vertebrate oncogene c-rel.Science 238, 692-694.

WILKINSON, D. G., BHATT, S., COOK, M., BONCINELU, E. AND

KRUMLAUF, R. (1989). Segmental expression of Hox-2homeobox-containing genes in the developing mouse hindbrain.Nature 341, 405^109.