Current Biology, Vol. 13, 2097–2102, December 2, 2003, 2003 Elsevier Science Ltd. All rights reserved. DOI 10.1016/j .cub.2003.10.062

A Serpin Regulates Dorsal-Ventral Axis Formationin the Drosophila Embryo

the cytokine-like polypeptide Spaetzle (Spz), to its activeform. In addition, a fourth serine protease Nudel (Ndl),provided by the follicle cells during oogenesis, is re-

Petros Ligoxygakis,1,* Siegfried Roth,2

and Jean-Marc Reichhart3

1Genetics UnitDepartment of Biochemistry quired to trigger the cascade [2, 27]. Through activatedUniversity of Oxford Spz, the signal is transduced via the Toll receptor to theSouth Parks Rd Cactus/Dorsal complex, which corresponds to the I�B/OX1 3QU, Oxford NF-�B complex of vertebrates [8]. Cactus is targeted forUK degradation and the NF-�B transcription factor Dorsal is2 Institut fur Entwicklungsbiologie free to enter the nucleus. This regulated nuclear trans-Universitat zu Koln port leads to the formation of a broad Dorsal gradient,Gyrhofstrasse 17 with highest nuclear concentrations in a narrow ventral50931 Koln region and rapidly declining concentrations in lateralGermany positions (Figures 1A and 1K). Dorsal target genes re-3 Institut de Biologie Moleculaire et Cellulaire spond to different levels of the gradient and implementUPR 9022 cell specification and morphogenesis across the d-vCNRS axis [9]. For example, snail (sna) activated by high levels15 rue Rene Descartes of nuclear Dorsal is required for mesoderm formation,67084 Strasbourg while rhomboid (rho) activated by lower levels promotesFrance ectodermal cell fate specification (Figures 1D and 1G).

The spatially restricted activation of the protease cas-cade apparently depends on the modification of an asyet unknown somatically expressed heparan sulfateSummaryproteoglycan (hspg). It is presumed that the gene pipe,which encodes a heparan sulfate 2-O-sulphotransfer-Extracellular serine protease cascades have evolvedase, mediates this modification [10]. The pipe gene isin vertebrates and invertebrates to mediate rapid, localthe only one in the Toll pathway that has a ventrallyreactions to physiological or pathological cues [1]. Therestricted pattern of expression, suggesting that it pro-serine protease cascade that triggers the Toll signal-

ing pathway in Drosophila embryogenesis shares sev- vides the initial spatial cues for the embryonic d-v axiseral organizational characteristics with those involved [10]. Females that lack the activity of any of the proteasein mammalian complement and blood clotting [2]. One genes, spz, Toll, pipe, or dorsal, produce dorsalizedof the hallmarks of such cascades is their regulation embryos in which cells at all positions give rise to dorsalby serine protease inhibitors (serpins) [3]. Serpins act ectoderm [8] (Figure 2E).as suicide substrates and are cleaved by their target It has been hypothesized that the extracellular prote-protease, forming an essentially irreversible 1:1 com- ase cascade leading to the formation of the Dorsal gradi-plex [4]. The biological importance of serpins is high- ent is kept in control of excessive or ectopic activationlighted by serpin dysfunction diseases, such as throm- by a serpin [11]. We have recently generated flies lackingbosis caused by a deficiency in antithrombin [5]. Here, the secreted Drosophila serpin, Serpin-27A (Spn27A),we describe how a serpin controls the serine protease by P element excision mutagenesis [6]. Embryos derivedcascade, leading to Toll pathway activation. Female from such females are severely ventralized, as expectedflies deficient in Serpin-27A produce embryos that lack from a spatially unrestricted activation of the Toll path-dorsal-ventral polarity and show uniform high levels way. Dorsal is uniformly nuclear in all embryonic cells,of Toll signaling. Since this serpin has been recently and as a consequence, mutant embryos lack d-v polarityshown to restrain an immune reaction in the blood of (Figures 1C and 1M). As seen from the uniform expres-Drosophila [6, 7], it demonstrates that proteolysis can sion of snail and the absence of rhomboid expression,be regulated by the same serpin in different biological all cells adopt the ventralmost cell fate, the presumptivecontexts. mesoderm (Figures 1F and 1I). Since the embryos lack

ectodermal cell fates, they do not secrete cuticle andResults and Discussion produce empty egg cases (Figure 2C). A hypomorphic

P allele of Spn27A reveals how the Dorsal gradient isDorsal-ventral (d-v) axis formation in Drosophila is initi- affected by reduced Spn27A levels. The region of peakated by a signal restricted to the ventral side of the levels of nuclear Dorsal expands, while the lateral re-embryo, which is conveyed through the action of three gions of declining concentration are shifted toward theextracellular serine proteases to the transmembrane re- dorsal side. Low levels of Dorsal are maintained in dor-ceptor Toll [2, 8]. The earliest acting protease is Gastru- salmost positions (Figures 1B and 1L). Accordingly, thelation defective (GD), followed by Snake (Snk), and finally snail domain expands to cover 60% of the embryonicEaster (Ea), which appears to process the Toll ligand, circumference, while the rhomboid domains are shifted

toward the dorsal side (Figures 1E and 1H). Thus, re-duced Spn27A levels are not linked to pathway activa-*Correspondence: petros.ligoxygakis@bioch.ox.ac.uk

Current Biology2098

Figure 1. Loss of Spn27A Leads to a Complete Ventralization of the Embryo

Cross-sections through syncytial ([A–C] and [G–I]) and cellular (D–F) blastoderm embryos. The dorsal side is facing up. (A)–(C), Dorsal protein;(D)–(F), snail mRNA; and (G)–(I), rhomboid mRNA distributions. (A) In wild-type, Dorsal protein forms a nuclear gradient restricted to the ventralhalf of the embryonic circumference with high levels straddling the ventral midline. (B) snail mRNA expression is restricted to the regions ofhigh nuclear Dorsal protein. (C) The regions of diminishing Dorsal levels express rhomboid mRNA. (B, E, and H) A hypomorphic P allele ofSpn27A (Spn27AP) in trans to Df(2L)6374 results in a partial ventralisation of the embryo. (B) The region of peak levels of the Dorsal gradientexpands to occupy 60% of the embryonic circumference. (E) This leads to a corresponding expansion of the snail domain, and (H) a shift ofthe lateral rhomboid domains to the dorsal side. (C, F, and I) The protein null allele Spn27Aex32 leads to a complete ventralization and henceto a complete loss of dorsoventral polarity. (C) Uniform high levels of nuclear Dorsal are found around the entire circumference. (F) All cellsexpress snail, and (I) rhomboid expression is absent. (K–M) Schematic representations of the nuclear Dorsal gradient. Abbreviations: v, ventralmidline; d, dorsal midline.

tion in dorsal positions but, rather, of an expansion of is able to counteract the process of axis induction. Onthe other hand, a mutant form of the serpin with thethe ventral region with peak activity.

Injection of Spn27A RNA into embryos lacking Spn27A putative protease cleavage site mutated (see Experi-mental Procedures) lacked rescue activity (Table 1). Thisrestored polarity and differentiation as approximately

60% of the embryos developed a normal cuticle pattern provides further evidence that Spn27A indeed interactswith a protease and, thus, fulfils the predicted biochemi-(Figure 2D). This result confirmed that the phenotypes

described above were due to the lack of Spn27A. One cal function as an inhibitor of the cascade.Since loss of Spn27A function results in apolar em-fifth of the injected embryos were weakly dorsalized

(Table 1; Figure 2H), indicating that an excess of Spn27A bryos, we wondered whether the site of injection of the

Regulation of Early Development by a Serpin2099

Figure 2. Epistasis Analysis and Rescue Ex-periments

Cuticle pattern of wild-type and mutant lar-vae. Anterior faces to the left and dorsal up.

(A) The cuticle of the wild-type larva carriesbelts of strongly pigmented denticles at theventral side (VE, ventral epidermis) andweakly pigmented hairs at the dorsal side(DE, dorsal epidermis). Ventral and dorsal epi-dermis is derived from lateral and dorsal re-gions of the blastoderm, respectively. Theventral most region of the blastoderm is inter-nalized to give rise to the mesoderm.(B) A hypomorphic P allele of Spn27A(Spn27AP) in trans to Df(2L)6374 leads to acomplete loss of dorsal epidermis. The cuti-cle consists of fragments of ventral epi-dermis.(C) The protein null allele Spn27Aex32 resultsin a lack of cuticle production since all cellshave adopted the mesodermal fate (see Fig-ure 1). The picture shows the empty egg case(vitelline membrane).(D) Injection of RNA encoding Spn27A into pre-blastoderm embryos derived from Spn27Aex32/Df(2L)6374 females can restore the wild-typecuticle pattern or (H) cause a mild dorsalisa-tion. The dorsalization is evident from theelongated, u-shaped structure of the cuticle,a phenotype similar to that of twist and snailmutants, which lack the mesoderm [26].

(E) pip mutations (see Table 2 for genotypes) cause a complete dorsalization. The embryo consists of an irregularly shaped hollow tube ofdorsal epidermis (DE).(F) The Spn27A;pip double mutant leads to embryos which are indistinguishable from the pip single mutant. Thus, the ventralisation of theSpn27A mutation is completely suppressed.(G) The gd; Spn27A double mutant results in lateralized embryos that secrete ventral or dorsal-lateral cuticle markers around their entirecircumference. The particular embryo shown is encircled by ventral denticle belts.

Spn27A RNA determined the polarity of the future d-v To further establish the role of Spn27A in the Tollpathway, we explored the epistatic relation to differentaxis. Depositing the RNA to either the dorsal or the

ventral side of the embryo led to the same degree of genes involved in d-v axis formation. Double mutantswith three proteases (GD, Snk, and Ea) as well as withrescue (Table 1). The polarity of gastrulation was always

identical to that of wild-type embryos. This indicates spz were produced. If Spn27A is an inhibitor of the cas-cade, all loss-of-function mutants mentioned abovethat asymmetries in the initial distribution of Spn27A

cannot reorient the d-v axis and that triggering of the should repress the ventralized phenotype because of alack of ventral signal. As seen in Table 2, this is indeedcascade remains strictly dependent on the ventral cues

provided by pipe expression. the case with one notable exception. The gd; Spn27A

Table 1. Rescue of the Spn27A Phenotype by RNA Injection

Side of Injection of Spn27A mRNA Phenotype of Injected Embryos DV Polarity at Gastrulation

Dorsal side 19% weakly ventralized 100% normal58% wild-type23% weakly dorsalizedn � 59

Ventral side 27% weakly ventralized 100% normal55% wild-type18% weakly dorsalizedn � 57

Ventral side, mutated version of Spn27A Strongly ventralized No DV polarity(no rescue) n � 53

The eggs for injection were derived from Spn27Aex32/Df(2L)6374 females. They were oriented such that the RNA could be deposited either todorsal (flat side of the egg case) or ventral (curved side of the egg case) regions of the embryo. After injection, the gastrulation of the embryoswas observed under the dissecting microscope. The embryos injected with the wild-type version of Spn27A RNA, which reached the stageof gastrulation, extended the germband to the dorsal side of the egg. If they formed a ventral furrow, this occurred always at the curved sideof the egg case. Thus, they exhibited wild-type polarity at gastrulation. After completion of development, the cuticle pattern was analyzed.The percentage of different phenotypes is indicated. The embryos injected with the mutant version of Spn27A had a nonpolar gastrulationphenotype and failed to produce cuticle.

Current Biology2100

Table 2. The Ventralized Phenotype of Spn27A Depends on the Activity of the Dorsal Group Genes

Genetic Hierarchy Genotypes of Double Mutants Phenotypes of Double Mutants

pipe Spn27ex32/Df; pip664/pip386 DO↓gastrulation defective gd9/gd9; Spn27Aex32/Df L1/L2↓snake Spn27Aex32/Df; snk073/snk223 DO↓easter Spn27Aex32/DF; ea1/ea2 DO↓spaetzle Spn27Aex32/DF; spzm7/spzm7 DO↓The genetic hierarchy among the Dorsal group genes, which act in the extra-embryonic space, is reviewed in [2, 9]. The arrows do notnecessarily imply a corresponding sequence of biochemical interactions. With the exception of gd, all double mutants result in completelydorsalized phenotypes (D0). gd; Spn27A causes a variable lateralized phenotype that comprises embryos expressing dorsolateral (L2) orventrolateral (L1) cell fates around their circumference. The classification of the phenotypes is according to [28, 29].

double mutant leads to a lateralized phenotype. The nates in the induction of the antifungal peptide genedrosomycin by way of NF-�B-dependent transcriptionembryos express ventral-lateral or dorsal-lateral cell

types around their entire circumference (Figure 2G), indi- in the fat body, the Drosophila analog of the mammalianliver. Although Ea is not involved in the cascade relayingcating a uniform intermediate or low level of pathway

activation. This suggests that the cascade has a basal the infection signal, overexpression of an activated formof Ea (�Ea) results in infection-independent constitutiveactivity in the absence of Gd, which is uncovered by

removing the serpin. Alternatively, the proteins pro- expression of drosomycin [18]. When Spn27A is co-expressed with �Ea, induction of drosomycin in adultduced by the two mutant gd alleles tested could have

residual function (see Experimental Procedures). How- fat body cells is blocked (Figure 3). The most plausibleexplanation for the above result is that of a direct interac-ever, this is unlikely to be the case for gd 9, which carries

a stop codon producing a truncated nonfunctional pro- tion of the two proteins.In this paper, we present evidence that a serine prote-tein variant [12, 13]

It is known that complement and coagulation path- ase inhibitor (Spn27A) provides an important controlelement in the proteolytic cascade leading to d-v axisways can be negatively regulated by serpins bound to

heparan sulfate proteoglycans (hspgs) [14]. Spn27A establishment in early Drosophila embryogenesis. Thefollowing model accommodates our findings. The trig-might require binding to an hspg for its activity and the

ventral Pipe-dependent modification of this hspg might gering of the d-v patterning cascade occurs in a limitedarea at the ventral side of the embryo in the same mannerprevent Spn27A binding. This, in turn, might allow the

cascade to be active at the ventral side. However, if as the mammalian blood-clotting cascade normally gen-erates a blood clot only at the site of the injury. Spn27Apipe would act only via negative control of the serpin,

the Spn27A; pipe double mutant should be ventralized is required for this process of spatial restriction. As itlike the Spn27A mutant alone. As seen in Figures 2E, 2F,and Table 2, this is not the case. Spn27A; pipe femalesproduce dorsalized embryos identical to those pro-duced by pipe mutants. This indicates that Pipe-depen-dent hspg modifications do not act via Spn27A but,rather, have a positive input on the protease cascade.This is in line with the injection data showing that theasymmetry provided by pipe cannot be overridden byan artificially produced asymmetry in the Spn27A distri-bution. Together, these results indicate that the serpinacts as an inhibitory “sink” that prevents the spreadingof cascade activation to lateral and dorsal regions ofthe embryo.

Earlier studies have presented indirect biochemicalevidence for a serpin regulating the Ea protease [11].Since Spn27A is maternally expressed and secreted, asis Ea ([15] and our unpublished data), we tested theability of Spn27A to inhibit Ea in vivo using the UAS/GAL4 system [16].

Figure 3. Spn27A Expression Represses the Immune Challenge-In addition to d-v axis formation, the Toll signaling Independent Induction of drosomycin Resulting from an Activated

pathway controls Drosophila host defense against Form of Ea (�Ea)gram-positive bacteria and fungi [17]. Again, Spz is pro- In wild-type (wt) flies, drosomycin (drs) has a basal level of expres-teolytically cleaved to an active form by the serine prote- sion. In this Northern blot the rp49 probe was used as a loading

control.ase Persephone [18], and signal transduction culmi-

Regulation of Early Development by a Serpin2101

Immunohistochemistryis uniformly distributed throughout the surface of theAntibody stainings to detect Dorsal protein have been describedoocyte (P.L., unpublished data), it could function in pre-[22]. snail and rhomboid mRNAs were detected by in situ hybridiza-venting any misfiring of the cascade in inappropriatetion with Digoxygenin-labled antisense RNA [23]. Embedding and

places. This is particularly relevant with regard to the sectioning techniques have been described elsewhere [22].observation that the downstream proteases have abasal activity in the absence of GD, which (as evident RNA Injections

Capped wild-type and mutated Spn27A RNA were made from corre-from the gd; Spn27A double mutant embryos) has to besponding pCS2 constructs [24] using the Message Machine kit (Am-kept in check by the serpin. The fact that the Spn27A;bion). The RNA was injected into preblastoderm embryos as pre-pipe double mutant females produce dorsalized em-viously described [25]. For both RNAs, the injection mix had the

bryos indicates that the Pipe-modified proteoglycan same concentration of approximately 500 ng/�l.provides a positive input on the protease cascade,which locally enhances its activity. The target of this Molecular Biology

Serpins interact with their target enzyme via an exposed reactiveinput is still unknown. However, it might be neither Ndlcenter loop [4]. This loop places the reactive center in an accessiblenor Gd, but one of the downstream proteases [12, 27].position for interaction with their target, allowing the formation ofOur data indicate that Spn27A exercises its control ata stable complex between enzyme and inhibitor. The amino acidthe level of Ea. The Pipe-dependent enhancement ofsequence and conformation of the loop dictate the selectivity of

the protease cascade may generate sufficient amounts inhibition. More specifically, inhibition relies primarily on the resi-of active Ea to target the serpin and thus overcome dues P1-P1� of the reactive center, which provide a cleavage site

for the target protease. To test whether Spn27A acts as an activeinhibition. This in turn may initiate a positive feedbackserine protease inhibitor, the putative protease cleavage site (P1-loop between Ea and Gd and thereby trigger the fullP1�, [6]) was changed from K-F to P-P by PCR-directed mutagene-activity of the cascade. The existence of feedback loopssis. The resulting construct was sequenced and cloned into thewithin the cascade from Gd to Ea has recently beenpUAST [16] and pCS2 [24] vectors. Northern blot analysis was per-

described [12, 19]. formed as previously described [6].It is interesting to note that a reduction in Spn27A

levels does not lead to uniform low-level triggering of Acknowledgmentsthe cascade but, rather, to an expansion of the ventral

We are grateful to Maria Langen, Ellen Veit, and Wendy Mulvanypeak levels of activation. The cascade remains inhibitedfor technical help with fly genetics and in situ hybridizations; Anniein the dorsalmost regions and a steep decrease of nu-Meunier for technical help in molecular biology and Yiannis Pandis

clear Dorsal can still be observed (Figure 1L). This indi- of Pandis, Inc., for IT support; P.L. is supported by a start up grantcates that normal Spn27A levels are critical in restricting from the University of Oxford and the Royal Society.cascade activation to the pipe domain, but not crucial

Received: September 26, 2003for determining the slope of the Dorsal gradient. DorsalRevised: October 14, 2003gradient formation apparently requires another inhibi-Accepted: October 14, 2003tory process directly linked to the activity of the cascade.Published online: November 3, 2003

Such a process has recently been described [20, 30].The cleavage of Spz provides not only the ligand for Referencesthe receptor Toll, but also a proteolytic fragment thatapparently inhibits the cascade [20]. It will be interesting 1. Krem, M.M., and DiCera, E.D. (2002). Evolution of enzyme cas-

cades from embryonic development to blood coagulation.to analyze at the biochemical level how the interplayTrends Biochem. Sci. 27, 67–74.between the serpin- and Spz-mediated inhibitory pro-

2. LeMosy, E.K., Hong, C.C., and Hashimoto, C. (1999). Signalcesses restrict and shape Dorsal gradient formation.transduction by serine protease cascade. Trends Cell Biol. 9,

We have recently reported that Spn27A controls local- 102–107.ized melanization of pathogen surfaces [6]. Here, we 3. Wright, H.T. (1996). The structural puzzle of how serine protease

inhibitors work. Bioessays 18, 453–464.present evidence for its involvement in the control of a4. Irving, J.A., Pike, R.N., Lesk, A.M., and Whisstock, J.C. (2000).proteolytic cascade in a developmental context. To our

Phylogeny of the serpin superfamily: implications of patterns ofknowledge, this is the first case that such a dual role isamino acid conservation for structure and function. Genomedocumented for a serpin. The paradigm of Spn27A couldRes. 10, 1845–1864.

serve as a model to study the control of proteolysis in 5. Bruce, D., Perry, D.J., Borg, J.Y., Carrell, L.W., and Wardell,different biological contexts. M.R. (1994). Thromboembolic disease due to thermostable con-

formational changes of antithrombin Rouen-VI (187Asn to Asp).J. Clin. Invest. 94, 2265–2274.Experimental Procedures

6. Ligoxygakis, P., Pelte, N., Ji, Ch., Leclerc, V., Duvic, B., Belvin,M., Jiang, H., Hoffmann, J.A., and Reichhart, J.-M. (2002). AGenetic Stocks and Crossesserpin mutant links Toll activation to melanization in the hostThe Spn27A mutant strain was generated as previously describeddefense of Drosophila. EMBO J. 21, 6330–6337.[6]. Double mutant stocks generated in this study were gd7;spn27Aex32,

7. De Gregorio, E., Han, S.-J., Lee, W.-J., Baek, M.-J., Osaki, T.,gd9;spn27Aex32, spn27Aex32;spzrm7, spn27Aex32;ea1/ea2, spn27Aex32;Kawabata, S.-I., Lee, B.L., Iwanaga, S., Lemaitre, B., and Brey,snk233/snk073, and spn27Aex32;pipe386/pipe664. Information for these al-P.T. (2002). An immune-responsive Serpin regulates the melani-leles can be obtained from Flybase (http://flybase.bio.indiana.edu).sation cascade in Drosophila. Dev. Cell 3, 581–592.Other stocks used were previously described:yolkGAL4, UAS-

8. Belvin, M.P., and Anderson, K.V. (1996). A conserved signallingSpn27A, and Df(2L)6374 [6, 16].pathway: the Toll-dorsal pathway. Annu. Rev. Cell Dev. Biol. 12,393–416.

9. Stathopoulos, A., and Levine, M. (2002). Dorsal gradient net-Cuticle PreparationsThe cuticle of nonhatched larvae was prepared after dechorioniza- works in the Drosophila embryos. Dev. Biol. 246, 57–67.

10. Sen, J., Goltz, J.S., Stevens, L., and Stein, D. (1998). Spatiallytion and chemical devitellinization as described [21].

Current Biology2102

restricted expression of pipe in the Drosophila egg chamberdefines embryonic dorsal-ventral polarity. Cell 95, 471–481.

11. Misra, S., Hecht, P., and Maeda, R., and Anderson, K.V. (1998).Positive and negative regulation of Easter, a member of theserine protease family that controls dorsal-ventral patterning inthe Drosophila embryo. Development 125, 1261–1267.

12. Dissing, M., Giordano, H., and DeLotto, R. (2001). Autoproteo-lysis and feedback in a protease cascade directing Drosophiladorsal-ventral cell fate. EMBO J. 20, 2387–2393.

13. Ponomareff, G., Giordano, H., DeLotto, Y., and DeLotto, R.(2001). Interallelic complementation at the Drosophila melano-gaster gastrulation defective locus defines discrete functionaldomains of the protein. Genetics 159, 635–645.

14. Neurath, H. (1989). Proteolytic processing and physiologicalregulation. Trends Biochem. Sci. 14, 268–271.

15. Han, J., Zhang, H., Min, G., Hemler, D., and Hashimoto, C. (2000).A novel Drosophila serpin that inhibits serine proteases. FEBSLett. 468, 194–198.

16. Brand, A.H., and Perrimon, N. (1993). Targeted gene expressionas a means of altering cell fates and generating dominant phe-notypes. Development 118, 401–415.

17. Hoffmann, J.A., and Reichhart, J.-M. (2002). Drosophila innateimmunity: an evolutionary perspective. Nat. Immunol. 3, 121–126.

18. Ligoxygakis, P., Pelte, N., Hoffmann, J.A., and Reichhart, J.-M.(2002). Activation of Drosophila Toll during fungal infection bya novel blood serine protease. Science 297, 114–117.

19. LeMosy, E.K., Tan, Y.-Q., and Hashimoto, C. (2001). Activationof a protease cascade involved in patterning the Drosophilaembryo. Proc. Natl. Acad. Sci. USA 98, 15119–15124.

20. Morisato, D. (2001). Spaetzle regulates the shape of the Dorsalgradient in the Drosophila embryo. Development 128, 2309–2319.

21. Wieschaus, E., and Nusslein-Volhard, C. (1998). Looking at em-bryos. In Drosophila. A Practical Approach, D.B. Roberts, ed.(Oxford: Oxford University Press) pp. 179–214.

22. Roth, S., Stein, D., and Nusslein-Volhard, C. (1989). A gradientof nuclear localization of the dorsal protein determines patternin the Drosophila embryo. Cell 59, 1189–1202.

23. Tautz, D., and Pfeifle, C. (1989). A non-radioactive in situ hybridi-sation method for the localisation of specific RNAs in Drosophilaembryos reveals translation control of the segmentation genehunchback. Chromosoma 98, 81–85.

24. Rupp, R.A., Snider, L., and Weintraub, H. (1994). Xenopus em-bryos regulate the nuclear localisation of XMyoD. Genes Dev.8, 1311–1323.

25. Ferguson, N., and Anderson, K.V. (1992). Decapentaplegic actsas a morphogen to organize dorsal-ventral pattern in the Dro-sophila embryo. Cell 71, 451–461.

26. Simpson, P. (1983). Maternal-zygotic gene interactions duringformation of the dorsoventral pattern in Drosophila embryos.Genetics 105, 615–632.

27. LeMosy, E.K., Kemler, D., and Hashimoto, C. (1998). Role ofNudel protease activation in triggering dorsoventral polarisationof the Drosophila embryo. Development 125, 4045–4053.

28. Roth, S., Hiromi, Y., Godt, D., and Nusslein-Volhard, C. (1991). Amaternal gene required for proper formation of the dorsoventralmorphogen gradient in Drosophila embryos. Development 112,371–388.

29. Anderson, K.V., Jurgens, G., and Nusslein-Volhard, C. (1985).Establishment of the dorsoventral axis in the Drosophila em-bryo: genetic studies on the role of the Toll gene product. Cell42, 779–789.

30. Roth, S., and Schupbach, G. (1994). The relationship betweenovarian and embryonic dorsoventral patterning in Drosophila.Development 120, 2245–2257.