Dorsoventral Axis Formation in theDrosophila Embryo — Shaping andTransducing a Morphogen Gradient

Bernard Moussian1 and Siegfried Roth2

The graded nuclear location of the transcriptionfactor Dorsal along the dorsoventral axis of the earlyDrosophila embryo provides positional informationfor the determination of different cell fates. Nuclearuptake of Dorsal depends on a complex signallingpathway comprising two parts: an extracellular pro-teolytic cascade transmits the dorsoventral polarityof the egg chamber to the early embryo and gener-ates a gradient of active Spätzle protein, the ligandof the receptor Toll; an intracellular cascade down-stream of Toll relays this graded signal to embryonicnuclei. The slope of the Dorsal gradient is not deter-mined by diffusion of extracellular or intracellularcomponents from a local source, but results fromself-organised patterning, in which positive and neg-ative feedback is essential to create and maintainthe ratio of key factors at different levels, therebyestablishing and stabilising the graded spatial infor-mation for Dorsal nuclear uptake.

In the past 10 years, compelling evidence has high-lighted the importance of morphogens for pattern for-mation during development [1,2]. A morphogen isdefined by two crucial properties. First, it exhibits agraded distribution across a developmental field, sothat the value of morphogen concentration at anygiven point of the gradient corresponds to a particularposition within that field. Second, distinct ranges ofmorphogen concentrations elicit different cell fates.The combination of these two properties provides anelegant and powerful mechanism for specifying cellfate as a function of space. Recent work has focusedon the cellular mechanisms of morphogen spreadingand long-range gradient formation. A surprising diver-sity of cellular processes and mechanisms of consid-erable kinetic complexity have been found to berequired for morphogen transport and for the shapingof morphogen gradients [3–7]. Despite the highlycomplex picture that has emerged from these studies,all cases considered so far have one feature incommon: gradient formation depends either onspreading of the morphogen from a local source or onthe spreading of a localised inhibitor, which forms agradient opposing that of the morphogen. As mor-phogen (or inhibitor) production is localised while itsdegradation occurs within the entire developmental

field, these cases can be classified as ‘local-source-dispersed-sink’ (LSDS) mechanisms [8–10].

Here, we summarise recent findings on the estab-lishment of the morphogen gradient that patterns thedorsoventral axis of the early Drosophila embryo, theformation of which can apparently not be explained bythe LSDS model. This gradient forms in the fluid-filledperivitelline space surrounding the embryo in whichthe key components required for gradient formationare evenly distributed. Positive and negative feedbackloops among these compounds, together with theirdifferential spreading within the perivitelline fluid,appear to constitute a reaction-diffusion system withself-organising properties.

The Ventralising Cascade of Drosophila — anOverviewEstablishment of dorsoventral asymmetry in the earlyDrosophila embryo is largely under maternal control(Figure 1) [11]. The initial cue for axis formation arisesat the ventral side of the embryo and is stably trans-duced into the embryo via the Toll pathway (Figures 2and 3). This pathway is evolutionarily conserved andalso acts during insect and mammalian immuneresponses, where it is activated upon microbial chal-lenge and induces the expression of antimicrobialpeptides or inflammatory cytokines [12–14]. DuringDrosophila embryogenesis, the Toll pathway estab-lishes at least three different regions along thedorsoventral axis, resulting in the separation of themesoderm and the neuroectoderm from the non-neu-rogenic (dorsal) ectoderm (Figure 4) [15,16]. The dorsalectoderm is patterned by the zygotically regulatedDecapentaplegic (Dpp) pathway [17].

Generation of the ventralising information involves17 maternally provided factors, which were identifiedin various genetic screens (Table 1) [18–23]. Thespatial cues for the embryonic dorsoventral axis orig-inate during oogenesis in the follicle cell layer, asomatic, monolayered epithelium, which surroundsthe oocyte–nurse cell complex (Figure 1) [24–27].Factors derived from the follicle cells comprise theintracellular proteins Pipe (Pip), Slalom (Sll) and Wind-beutel (Wbl) and the secreted protease Nudel (Ndl).Together they are required to activate a cascade ofproteases which are secreted as inactive precursorsby the oocyte or the early embryo. This cascade iscomposed of Gastrulation defective (GD), Snake (Snk),and Easter (Ea) and has similarity to the blood clottingand complement activating cascades of vertebrates[28]. Three-dimensional modelling of the protease-substrate complexes supports the sequence of actionof the proteases (GD–Snk–Ea), which was derived bygenetic epistasis analysis [29] and suggests that allcomponents of the cascade have been identified. Theprotease cascade is modulated by several intrinsic

Review

Current Biology, Vol. 15, R887–R899, November 8, 2005, ©2005 Elsevier Ltd. All rights reserved. DOI 10.1016/j.cub.2005.10.026

1Department of Genetics, Max-Planck Institute forDevelopmental Biology, Spemannstr. 35, 72076 Tübingen,Germany. 2Institute for Developmental Biology, University ofKöln, Gyrhofstr. 17, 50923 Köln, Germany.E-mail: siegfried.roth@uni-koeln.de

positive and negative feedback loops, as well as bythe Easter inhibitor Serpin27A. The activation of thecascade culminates in the generation of the mor-phogen Spätzle (Spz), which binds to the transmem-brane receptor Toll (Tl). The signal is transduced intothe embryo by the adaptor proteins Krapfen/dMyd88(Kra/dMyd88) and Tube (Tub) and the kinase Pelle(Pll). This leads to the degradation of the IκκB homo-logue Cactus (Cact), which allows the NFκκB transcrip-tion factor Dorsal (Dl) to enter the nucleus and activateor inhibit zygotic genes required for dorsoventral cellfate specification.

At several positions of the pathway, the means ofsignal transduction is not yet understood. In particu-lar, the substrate of Pipe specifying the ventral domainremains to be identified, and the role of Nudel is con-troversial as neither its activating factors nor its sub-strates are known. Furthermore, it is not clear howPelle transmits the signal to Cactus. Finally, two newmaternal factors, Weckle (Wek) and Seele (Sle) havebeen found which have loss-of-function phenotypessimilar to those of other pathway components. Theirrole within the pathway remains to be established [19].

In order to generate a pattern of distinct cell typesalong the dorsoventral axis, the intensity of the signalregulating the entry of Dorsal into the nucleus is high inthe ventral most region and drops at the lateral posi-tions, whereas in the dorsal region Dorsal continues tobe sequestered by Cactus and remains in the cytoplasm[30–32] (Figure 4 and 5B). High nuclear concentrationsof Dorsal activate the transcription of twist (twi) and snail(sna) in the ventralmost 18–20 cells (20%), which invagi-nate into the embryo to form the mesoderm [16] (Figure5C). Abutting this zone, the lateral cells (16–18 cells oneach side) at intermediate and low Dorsal concentrationexpress single minded (sim), rhomboid (rho) and shortgastrulation (sog) and a number of other genes requiredfor the establishment and patterning of the neuroecto-derm [33–35]. The neuroectoderm gives rise to theventral nerve cord and the ventral epidermis; finally, theremaining dorsal portion of the embryo (40%) lacksDorsal activity and expresses genes required for theestablishment of the amnioserosa and the dorsal epi-dermis, like zerknüllt (zen) and dpp. These genes arerepressed by Dorsal in the ventrolateral and ventralcells. A combination of microarray experiments and

ReviewR888

Figure 1. Major events during dorsoventral axis formation in Drosophila.

(A) Gurken (green), a TGFαα protein emanating from a dorsal source in the oocyte, represses pipe (orange) in the follicular epithelium,restricting its expression to a ventral stripe. (B) Pipe leads to the modification of an unknown ECM component (red) which is secretedby the follicle cells and maintained within the extracellular space surrounding the late oocyte and early embryo. (C) Subsequently, anextracellular signal is generated within the region defined by the modified ECM component. This signal induces the nuclear transportof the transcription factor Dorsal. The signal has peak levels within a narrow stripe straddling the ventral midline (dark blue). In thearea between the dark blue stripe and the border of the red stripe in (B) a gradient is established, which accounts for the differentventrolateral cell fates of the embryo (compare to Figure 6).

A Restriction of Pipe expression

B ECM modification

C Ventralising signal and nuclear Dorsal gradient

Oocyte

Follicle epithelium

Nurse cells Nucleus

Perivitelline space

Embryo

Embryo

Perivitelline space

Nuclei

Dorsal

Embryo

Ventral viewLateral view

Oocyte

Cross section

Perivitelline space

Perivitelline space

Current Biology

Embryo

bioinformatics has recently been used to identify a largenumber of new Dorsal target genes and their cis-regula-tory elements, which mediate concentration dependentregulation by Dorsal [33,34,36,37]. Thus, Dorsal and itstarget genes constitute one of the best-understoodgene regulatory networks [38].

Origin of the Ventralising InformationThe ventralising signal is derived from the dorsoven-tral asymmetry of the egg chamber. Four factors havebeen identified so far, which are required to transmitthis asymmetry to the embryo by specifying a domainat the ventral side of the egg where the Toll ligand isgenerated (Figure 2) [21,39–42]. The dorsoventralasymmetry arises from the repression of pipe tran-scription in the dorsal half of the follicular epitheliumby Gurken–EGF receptor signalling during oogenesis[39,43], which is a direct readout of a gradient of EGFreceptor activity and does not require secondary sig-nalling processes (Figure 1) [44–46].

Pipe expression at the ventral side of the develop-ing oocyte is sufficient to induce axis formation in theembryo, as ectopic expression of pipe in the follicularepithelium activates the Toll pathway [43,45].However, it is crucial to note that expression of pipe isinsufficient to explain the shape of the Dorsal gradient.Pipe is expressed evenly in a domain encompassing~40% of the circumference of the egg chamber withsharp lateral borders (Figure 1 and Figure 5A,B).During egg maturation, this domain maintains its sizerelative to the egg circumference [43,45]. The entirebell-shaped Dorsal gradient spans 40% of the egg cir-cumference and thus lies within the domain of uniform

pipe expression (Figure 1 and Figure 5A,B). Moreimportantly, in mutants of grk or egfr, pipe repressionis compromised and the pipe domain expands. TheDorsal gradient is not just expanding, however, butrather splits and acquires two peaks, one at each sideof the ventral midline [39,43] (Figure 5C,D and Figure6). These observations demonstrate that pipe doesnot just localise a source, from which a gradient formsby diffusion. Pipe rather defines a ventral region of theegg, in which a complex patterning system leads togradient formation by the dynamic interactions of itscomponents. Therefore, a LSDS model cannot beused to explain the formation of the Dorsal gradient.

The pipe locus encodes ten related proteins, whichshow similarity to heparansulfate 2-O-sulfotrans-ferases of vertebrates [39,47]. Only one of the proteinisoforms is expressed in the ventral follicle cells andresponsible for dorsoventral axis formation. Otherisoforms appear to be required for the late morpho-genesis of the salivary glands. As mentioned above,the substrate modified by Pipe is not known. A func-tional analysis of enzymes essential for glycosamino-glycan synthesis showed that heparan sulfate andother glycosaminoglycans are not required fordorsoventral patterning within follicle cells and thusare unlikely to be substrates for Pipe [39,47].However, several observations suggest that Pipeindeed acts as a sulfotransferase. The universalsulfate donor in sulfotransferase reactions is 3´-phos-phoadenosine 5´-phosphosulfate (PAPS). Both thePAPS synthetase [47] and Slalom, which importsPAPS from the cytoplasm into the Golgi apparatus[21], are required for dorsoventral pattern formation.

Current BiologyR889

Table 1. Genes involved in dorsoventral axis formation in Drosophila.

Gene Protein Molecular function Biological function

pipe Heparan-sulfate 2-sulfotransferase ECM modification Specification of the ventral domain

slalom Translocator Import of PAPS1 into Golgi Specification of the ventral domain

windbeutel ER protein Localisation of Pipe to the Golgi Specification of the ventral domain

nudel Serine protease ? Specification of the ventral domain

gastrulation defective Serine protease Activation of Snake Generation of the ventralising signal

snake Serine protease Activation of Easter Generation of the ventralising signal

easter Serine protease Interaction with Gd Generation and refinement of the Processing of Spätzle ventralising signal

serpin27A Serine protease inhibitor Inhibition of Easter Modulation of Easter activity

spätzle NGF-like Ligand of Toll Ventralising signal, morphogen

Toll Receptor Receptor for Spätzle Transmission into the embryo

tube Adaptor Recruitment of Pelle Signal transduction

dMyd88/krapfen Adaptor Binding of Toll and Tube Localisation of Tube

pelle Serine/threonine kinase Phosphorylation leading to Stable copy of gradient within the Cactus degradation embryoInactivation of Toll and Tube

cactus IκκB homologue Inhibition of Dorsal Suppression of the ventralising signal

dorsal NFκκB homologue Regulation of zygotic genes Morphogen

weckle Zinc-finger transcription factor Regulation of transcription ?

seele ? ? ?

In addition, the molecular analysis of different pipealleles and feeding experiments with an inhibitor ofPAPS synthetase support the assumption that Pipetransfers sulfate groups to the carbohydrate moietiesof a yet unknown protein or glycolipid [47] .

Unlike pipe, windbeutel is expressed in all folliclecells around the oocyte. However, genetic data clearlyshow that Windbeutel like Pipe is involved in definingthe ventral domain [41]. Windbeutel encodes a novelprotein implicated in ER trafficking, and is required forproper localisation of Pipe to the Golgi apparatus[40,48]. Loss of Windbeutel function causes retentionof Pipe in the ER.

No connection has yet been established betweenPipe and the protease Nudel. Nudel is expressed in allfollicle cells, and the protein localises to the surface ofthe oocyte and early embryo [42,49]. It is a largeprotein (>300 kDa) with multiple domains, including acentral 33 kDa protease domain. To be active, Nudelrequires autocatalytic processing as well as cleavageby a yet unknown factor. The fragments segregate todifferent compartments within the embryo. The pro-tease domain, for instance, which is crucial for gener-ation of the ventralising signal, localises to vesicles inthe embryo. Entry of this fragment into the embryo isdependent on its protease activity and temporallycoincides with activation of the Toll receptor, but isnot asymmetric along the dorsoventral axis. Moreover,carboxy- and amino-terminal fragments of Nudel havealso been detected in the embryo, in another subcel-lular compartment. These domains of Nudel havemotifs found in many proteins of the extracellularmatrix, including three potential glycosaminoglycanaddition sites in the amino-terminal portion. Modifica-tion of these sites is important for activation of theprotease domain, but, interestingly, does not dependon Pipe function [49,50].

Besides autocatalytic processing, the target of theNudel protease remains to be identified; it has beenshown that the best candidate, GD, is processed inde-pendently of Nudel [51], but computational modellingof Nudel and GD suggests that Nudel may indeed bindand cleave GD [29]. Generation of the 33 kDa Nudelprotease domain is also independent of Pipe and

Windbeutel function, as it is present in embryos frompipe and windbeutel mutant females and it occursevenly throughout the circumference of the embryo.Hence, the follicular factors do not act in a linearpathway to activate the downstream proteasecascade, but rather in parallel.

Starting the Protease Cascade — the Role of gdHow is the information provided by the ventral Pipedomain used to generate the ventralising signal? Theprotease cascade in the perivitelline fluid starts withGD, which encodes an unusual serine protease, struc-turally similar to mammalian complement factors C2and B [52–54]. In the absence of GD, the downstreamserine proteases show a basic level of activation,which however lacks dorsoventral polarity [52–54].Thus, GD provides the critical link between pipeexpression and local activation of the proteasecascade. The gd mRNA is stored in the oocyte andthe GD protein is secreted as an inactive precursorinto the perivitelline space. GD acts prior to egg depo-sition during late stages of oogenesis.

The GD precursor is proteolytically processed,giving rise to an amino-terminal propeptide and thecarboxy-terminal catalytic chain. GD acts locally at theembryo surface [52,54,55]. Indeed, it has been shownin vitro that GD associates with heparin [51], suggest-ing that in vivo it might associate with sulfated glyco-proteins of the extracellular matrix. Injection assayssupport the idea that GD is not freely diffusible as theventralising effect of GD is restricted to the side ofinjection. Interestingly, Pipe appears not to be essen-tial for GD localisation and activation as injections ofgd RNA in the dorsal region of the embryo can initiateaxis formation. However, the product of Pipe seems tofacilitate GD activation as injection of GD into theventral side of the embryo shows stronger and spa-tially more extended effects. Thus, Pipe may justprovide a bias for the activation of the cascade. Thiswould be sufficient to induce pattern formation if thesystem has self-organising properties [10].

What actually activates GD remains uncertain,although a direct interaction of GD with Nudel cannotbe ruled out [29,56]. However, as stated above,

ReviewR890

Figure 2. A model for the activation of theToll pathway.

The ventral domain (light grey) of theembryo is specified by Pipe in the follic-ular epithelium. Windbeutel is required toallow the export of Pipe from ER toGolgi. Slalom is required to transport thesulfate donor PAPS into the Golgi whereit functions as substrate for the sulfo-transferase reaction carried out by Pipe.The autocatalytically processed Nudelcentral protease domain either pro-cesses the ECM component modified byPipe or directly triggers the proteasecascade through the activation of GD.GD activates Snake, which in turn acti-vates Easter to process Spätzle, theligand for the Toll receptor. Easter activ-ity is spatially and temporally regulated

by Serpin27A. Moreover, Easter might control its own activation through inactivation of GD. Finally, the amino-terminal fragmentof Spätzle interferes with Toll activation through an unknown mechanism, possibly by activating Serpin27A.

Gd

Snake

Easter

Spätzle

Toll

Serpin27A

?

Hep

aran

Pipe

Windbeutel

Nudel

Slalom

?

Ventral domain

?

?

?

Embryo

ECM

Plasma membrane

Follicular epithelium

Perivitelline fluid

Current Biology

exogenous GD can induce the Toll pathway in a nudelmutant background suggesting that Nudel is notabsolutely required for GD activation [53]. Two mech-anisms have been proposed for GD activation at theventral side of the embryo [51]: either the GD precur-sor is processed dependent on the Pipe substrate andNudel by a yet unknown factor and subsequently trig-gers the downstream cascade, or GD has a low pro-tease activity as a precursor, which is inhibited by ayet unknown factor. As in other complement-likefactors [57], a conformational change within GDinduced by the Pipe substrate and Nudel may releaseGD from the suppressor.

We would like to propose a third model of GD acti-vation, which is in agreement with an interpretationput forward by Ellen LeMosy and colleagues [56].This model is based on the observation that GD pro-tease activity is enhanced in the presence of Snake,the downstream target of GD [51]. In addition, Snakelike GD has been shown to associate with heparin[51]. In our model, activation of GD requires Snake tobe locally concentrated in the vicinity of GD. This inturn would depend on the Pipe substrate and/orNudel. As GD is bound to the embryo surface, thesignal could not spread beyond the domain definedby Pipe, which would ensure that the dorsoventralasymmetry within the follicular epithelium is faithfullyconveyed to the embryo.

How does GD act on the downstream proteases?Interestingly, interallelic complementation occursbetween mutations affecting the catalytic chain intrans to mutations within the amino-terminal polypep-tide. Indeed, injection assays have shown that bothpolypeptides are important for signalling and havedifferent roles during axis formation. The GD proteasedomain serves to transmit the ventralising signal,while the amino-terminal fragment appears to haveseveral functions. It is required to link GD to theembryo surface, but it might also repress the ventral-ising signal [54]. Interestingly, both polypeptides pro-vided in parallel have no effect on the mutantphenotype indicating that an intact unprocessed GD

protein is required to correctly transmit the signal.This requirement kinetically links the activationprocess to the formation of the potentially inhibitoryamino-terminal fragment.

In vitro data suggest that GD is subject to feedbackregulation after activation of Snake [51]. First,processed GD is subject to autoproteolysis, andsecond, Easter, the target of Snake, cleaves GD. So farit has not been established whether these proteolyticprocesses lead to activation or inactivation of GD, butboth possibilities are interesting from a theoreticalpoint of view (see below). In case of negative feedback,they might, together with the generation of aninhibitory amino-terminal fragment, provide a first stepby which the broad domain of activation defined bypipe is narrowed down towards the ventral midline.

New Insights into the Role of EasterRecent work has put Easter into the centre of interestas a major factor responsible for the shape of theDorsal gradient [58–63]. Easter codes for a serine pro-tease that requires proteolytical processing by Snakein order to be active [51,63]. Regulation of Easteractivity involves the serine protease inhibitorSerpin27A, which irreversibly binds to the catalyticcentre of Easter once it is cleaved from the amino-ter-minal pro-domain [59–61].

Interestingly, the interaction of Easter withSerpin27A is subject to feedback regulation, as theamount of the Easter–Serpin27A complex beingformed depends on downstream components of thepathway. In dorsalised embryos, the level ofEaster–Serpin27A is increased, whereas in Cactusmutants it is decreased [62]. This signal dependenceof Serpin levels is difficult to explain. One couldimagine that the secretion of Serpin27A into the extra-cellular space is controlled by Toll signalling. At theventral side of the embryo, secretion is inhibited,whereas at the lateral and dorsal sides outside thePipe domain Serpin27A is secreted to inhibit Easter.An inverse correlation of Toll pathway activity andSerpin27A secretion would explain the observation

Current BiologyR891

Figure 3. The cytoplasmic eventsdownstream of the Toll receptor.

The activated Toll receptor recruits themembrane-localised dMyd88–Tubeheterodimer to the intracellular domainof the activated Toll receptor. Subse-quently, the kinase Pelle is recruited tothe heterotrimeric Toll–dMyd88–Tubecomplex and undergoes autophos-phorylation, thereby enhancing itskinase activity. By phosphorylation ofToll and Tube Pelle is released fromthe heterotetrameric complex. At thisstep, signalling is disrupted, and theToll–Spätzle complex is presumablyinternalised. Next, Pelle transduces thesignal to downstream factors, resultingin the degradation of Cactus. Thisallows binding of Dorsal to Tamowhich regulates nuclear entry of Dorsalthrough a nuclear pore containingDNTF-2 and Mbo.

Spätzle

Toll

Pelle

EmbryoPlasma membrane

PP

Dorsal

Cactussog

rho

sim

snail

twist

Tamo

DNTF-2

Mbo

Nucleus

dMyd88

Tube

dMyd88

Tube

Pelle13

2

1

2

3

Current Biology

Perivitelline fluid

that the amount of the Easter–Serpin27A complex isincreased by dorsalising mutations, but decreased byventralising ones.

What is the biological significance of theEaster–Serpin27A interaction? Some dominant eastermutations within the conserved Serpin binding pocketcause partial ventralisation of the embryo withoutincreasing the total amount of Spätzle indicating thatEaster spreads around the embryo to ectopically acti-vate Spätzle [61]. Dorsoventral polarity is maintainedin these mutants, confirming that Easter is asymmet-rically activated, but the slope of the Dorsal gradientis dramatically flattened [61]. Likewise, in embryosderived from Serpin27A mutant females ectopicEaster activity causes a complete ventralisation of theembryo. Hence, Serpin27A is required to represssignal transduction through the interaction with Easteroutside the domain defined by Pipe.

Additional insight into the mechanisms of regula-tion of the protease cascade comes from the analy-sis of the dominant mutation Easter5.13, whichcauses an inhibitor-independent but low proteaseactivity sufficient to induce sog but not twist expres-sion leading to a loss of polarity that results in later-alised embryos [61]. Easter5.13 mutant embryos haveincreased amounts of processed Spätzle. Injection ofthe perivitelline fluid from gastrulating ea5.13;spzdouble mutant embryos, in which Toll signalling hasceased, into embryos from pipe mutant females still

induced axis formation. The increased amount ofprocessed Spätzle thus results from prolonged activ-ity of Easter. This indicates that Easter–Serpin27Ainteraction is also required to temporally restrictEaster activity.

The spatial and temporal regulation of Easter arethus both crucial for establishing a peak of the ven-tralising signal and a graded decrease of the signaltowards the borders of the ventral domain.

The Morphogen SpätzleRecent genetic and molecular findings underscorethat Spätzle does not simply respond to the refinedEaster activity, but assumes a more active role in gra-dient formation. Spätzle encodes a NGF-like proteinwith a cystine knot motif, found in many vertebrategrowth factors, and alternative splicing creates multi-ple isoforms, which are present at different timepoints of development [64,65]. The isoforms actingduring dorsoventral axis formation are processed intotwo active molecules, the constant carboxy-terminalC-106 ligand that dimerises to bind two monomers ofToll [65,66] and the variable amino-terminal pro-domain [64].

Controlling the amount of Spätzle is crucial forcorrect pattern formation. Feedback regulationapparently acts to reduce Spätzle activity. This isnotably observed in embryos derived from Tollmutant females [67], in which the amount of Spätzleis clearly increased, suggesting that degradation ofSpätzle probably through internalisation of the recep-tor–ligand complex is important in the regulation ofsignalling [68–70].

How does Spätzle contribute to the generation ofthe dorsoventral gradient? As noted above, anincrease of the amount of Spätzle does not affect theslope of the gradient, as assayed by sog expression,but rather causes the enlargement of the twistdomain, which yet never exceeds the ventral regionspecified by Pipe [67]. This is even true for eggs laidby egfr mutant females, in which the pipe domainincreases from 20% to 70% of the circumference.Reduction of the Spätzle dose by half causes areduction of the twist domain from 20% to 15% of theembryo circumference without affecting the domainof sog expression [67]. In conclusion, Spätzle seemsto respond stably to the slope of Serpin-regulatedEaster activity.

However, it appears that this stable transmission isnot resulting from an exact copying of the Easteractivity gradient. Evidence for a rather complicatedmechanism of Spätzle activity comes from injectionessays showing that increased local amounts ofSpätzle cause axis duplication, a phenomenon thathas been observed also in embryos laid by grk andegfr mutant females [24,25]. In these embryos, twoventral furrows are formed within the enlarged pipedomain (Figure 5D) [67]. However, embryos fromToll10b mutant females that induce twist expressionaround the embryo circumference do not exhibit aduplicated axis [25,30,67]. Hence, expanding theembryonic ventral domain per se is not sufficient toinduce axis duplication. Rather, axis duplication

ReviewR892

Figure 4. Positional information generated by the Toll pathway.

A schematic half cross-section of an early Drosophila embryo;graded activation of the Toll pathway in the ventral half leadsto the subdivision of the embryo into three broad regions alongthe dorsoventral axis — the presumptive mesoderm, neuroec-toderm and non-neurogenic ectoderm. Subsequent signallingprocesses lead to further subdivisions within these regions.Indicated here are the mesectoderm, the dorsal ectoderm andthe amnioserosa.

sogrhosimtwist

Mesoderm

Dorsalectoderm

Amnioserosa

HighnuclearDorsal

LownuclearDorsal

Mesectoderm

Neuroectoderm

Current Biology

IntermediatenuclearDorsal

occurs upstream of the Toll receptor, outside theembryo.

Indeed, recently, using an elegant approachMorisato [67] has shown that axis duplicationdepends on Spätzle itself, as it is suppressed inembryos expressing high levels of a dominant versionof Spätzle that enhances its interaction with thereceptor Toll. This can be explained if the amino-ter-minal fragment of Spätzle that is formed by Easter-mediated proteolysis acts as a negative regulator ofsignalling. Injections of this part of the Spätzle proteincause a dorsalisation of the embryo, thus antagonis-ing the effect of the carboxyl terminus of Spätzle.However, the dorsalising effect of the amino-terminalfragment is weaker compared to the ventralising effectof the carboxy-terminal fragment.

It has been proposed that the amino-terminal frag-ment of Spätzle directly or indirectly inhibits Easter. AsEaster is also the substrate for inhibition by Serpin27A,it is attractive to propose a link between both types ofinhibition. The amino-terminal fragment of Spätzlemight either bind to Easter and facilitate the interac-tion with the Serpin or bind and activate the Serpin.However, irrespective of the biochemical details, theproduction of an inhibitory peptide linked to the acti-vation process is likely to be essential in generatingthe Spätzle morphogen gradient. The more Easteractivates Spätzle, the more the Spätzle amino-termi-nal fragment inactivates Easter. Assuming a higher dif-fusion rate for the amino-terminal Spätzle fragmentthan for the carboxy-terminal fragment, the amino-ter-minal fragment would gradually inhibit Easter at theborders of the position with high concentration of the

carboxy-terminal Spätzle fragment, thereby creating aSpätzle slope where sog expression is activatedaround a Spätzle peak that induces twist expression(Figure 6).

Axis duplication upon injection of high amounts ofSpätzle or after expansion of the pipe domain mightresult from perturbation of the delicate balancebetween the ligand (Spätzle) and its inhibitor (Spätzleamino-terminal fragment). Higher concentrations ofunprocessed Spätzle or an expansion of the ventraldomain where Spätzle is processed are likely to causeincreased inhibitor concentrations at ventralmostpositions. Easter inactivation would thus be maximalat this position and, consequently, the ventralmaximum of Spätzle activity would split into twomaxima at more lateral positions.

In summary, generation of the ventralising signalcomprises at least three potential pairs of activatorsand inhibitors: the carboxy- and amino-terminal frag-ments of GD, Easter and Serpin27A, and the carboxy-and amino-terminal fragments of Spätzle. While Easterand Serpin27A physically interact to shape theborders of the Spätzle gradient, the role of the GD andSpätzle amino-terminal fragments is not entirely clear.

Theoretical ApproachesAs the formation of the Spätzle gradient is due to acti-vating and inhibiting interactions among components,which are diffusible within a fluid-filled space, theperivitelline space, it should be ideally suited for mod-elling based on reaction-diffusion kinetics. Pattern for-mation mechanisms based on reaction-diffusionkinetics can be divided into two groups [71,72]:

Current BiologyR893

Figure 5. The pipe expression domainand the Dorsal gradient.

(A) Cross-section of a stage 10 eggchamber at the level of the oocytenucleus (n) showing the distribution ofpipe mRNA (blue). pipe is evenlyexpressed in ventral follicle cells (fc),encompassing 40% of the egg chambercircumference. (B) Cross-section of asyncytial blastoderm embryo showingthe distribution of Dorsal protein (red).The nuclear Dorsal gradient is confinedto a region corresponding to the pipedomain of the follicular epithelium.(C,D) Cross-sections of embryos at thebeginning of gastrulation showing Twistprotein distribution (brown). In wild-type(C), Twist is expressed in a group ofventral cells which have received highlevels of Toll signalling and, therefore,accumulated high levels of nuclearDorsal protein. These cells start toinvaginate to form the mesoderm. In anembryo derived from gurken mutant(grkHK/grkWG) females (D), the reductionof Grk signalling leads to an expansion ofthe pipe domain. As a result, Toll sig-nalling is active in a broader region. Thisresults in the splitting of the peak of Tollactivation and consequently in tworegions accumulating high levels ofnuclear Dorsal. Therefore, two separategroups of cells express twi and start toinvaginate. (A and B) modified from [26].

(1) Mechanisms in which the final pattern is stableover time. After a transition period, a steady-state isreached in such systems at which dynamic interac-tions of the components stably maintain the pattern.Patterns formed under such steady-state conditionshave been termed Turing structures. (2) Chemicalwave mechanisms in which the pattern is transientand constantly changes its shape. It is important todecide whether the pattern to be modelled representsa steady-state or a non-steady state situation sincethis decision leads to certain constraints regardingmolecular interactions and diffusion rates.

As in Drosophila the Spätzle gradient appears to bestable during the relevant developmental stages, asteady-state approach was used for modelling [10,73].Theoretical considerations show that the productionof a stable pattern requires the link between a processof local self-activation and long-range or lateral inhibi-tion [74,75]. The inhibition can be realized in two dif-ferent ways either through production of an inhibitor,which spreads from the peak of local activation or bythe depletion of a substrate, which is consumed bythe activation process [75].

The biochemical data described above show thatthe formation of the Spätzle gradient involves a localactivation process, which may have multiple positivefeedback loops, as required by the theory. The self-activation appears to be linked to the production ofinhibitors — in this case the amino-terminal pro-domains of the proteases and Spätzle — which mighthave higher diffusion rates than the catalytic domainsof the proteases and the receptor-binding domain ofSpätzle. This suggests lateral inhibition by diffusibleinhibitors. However, simulations using this type oflateral inhibition are unable to reproduce a Spätzlegradient with one peak, but rather result in two peaksof high activation at the lateral borders of the pipedomain [73].

The correct pattern can be simulated only with thehelp of a substrate-depletion model. Because the sub-strate is depleted around the activation peak, the peakremains stably localised in the centre. Maybe the ECMcomponents modified by Pipe lead to the productionof a diffusible substrate for the proteolytic cascade[73]. Substrate depletion might also result if down-stream proteases inactivate up-stream proteases,which are diffusible in the perivitelline space. Morebiochemical details, however, are required to recon-struct the full kinetic complexity for more realisticmodelling approaches.

In addition, some of the assumptions made in thesimulations so far should be reconsidered. It is possi-ble that the system never reaches a steady state. Forinstance, in a more primitive insect, the red-flour beetleTribolium castaneum, the nuclear Dorsal gradientforms only transiently and undergoes progressiveshape changes before it disappears [76]. Thus, model-ling approaches should also consider presteady-statesolutions or non-steady-state mechanisms. Interest-ingly, the blood coagulation cascade, which sharesmany organisational features with the dorsoventralpathway was successfully modelled using a non-steady state travelling-wave mechanism [77–79].

Inside the EmbryoThe shape of the Dorsal gradient seems to be estab-lished mainly outside the embryo, but how does thecytoplasm respond to these cues (Figure 3)? Thecentral factor for robust transmission of the ventralis-ing information into the embryo appears to be theserine/threonine kinase Pelle [80–87]. Graded Pelleactivity is sufficient to induce all zygotic genes speci-fying different regions of the dorsoventral axis, as hasbeen shown by monitoring zygotic gene expression inresponse to ectopic activation of Pelle in absence ofthe ventralising signal [88].

How is this accuracy achieved? Pelle acts down-stream of the two adaptor proteins Tube andKrapfen/dMyd88 [83,84,89]. These two factors associ-ate through their DEATH domains and localise to theplasma membrane independently of the signal [84]. Inthe absence of signal, Pelle is distributed in the cyto-plasm [90,91]. Upon ligand binding and Toll dimerisa-tion, the intracellular domain of Toll recruits theTube–Krapfen/dMyd88 complex through the associa-tion of the TIR (Toll and Interleukin Receptor) domainsof Krapfen/dMyd88 and Toll. This complex thenrecruits Pelle that binds to the DEATH domain of Tubethrough its own DEATH domain [92]. The amount ofTube and Krapfen/dMyd88 at the ventral plasmamembrane of the embryo is now two times higher thanat the dorsal side.

Multimerisation of the Toll receptor together withthe formation of the tetrameric complex of Toll–Tube–Krapfen/dMyd88–Pelle leads to a local increaseof Pelle concentration [83,85,93,94]. At high concen-trations Pelle undergoes autophosphorylation, thusenhancing its kinase activity and phosphorylating Tolland Tube, thereby getting released from the complexto activate downstream targets [95]. Phosphorylationof Tube and Toll stops signal transduction, which isthought to prevent over-amplification of the signal andensures that Pelle activity stably transmits the Spätzlegradient into the embryo. Consistently, lack of Pellefunction increases the amount of Tube andKrapfen/dMyd88 at the ventral side of the embryo,indicating negative feedback regulation.

How does Pelle activation lead to a robust responsein the transcription of zygotic target genes? In vitroassays suggest that Cactus and Dorsal are directtargets of Pelle [95–97]. In the mammalian Tollpathway, however, the Pelle homologue IRAK (Inter-leukin 1 Receptor-Associated Kinase) does not phos-phorylate the Cactus homologue IκκB (nor NfκκB, theDorsal homologue) [98,99]. Rather, IκκB is phosphory-lated by the IκκB-Kinase (IKK) complex to release NfκκB.Thus, it might be possible that in vivo IKK relatedkinases act between Pelle and Cactus in Drosophila.Indeed, two IKKββ exist in Drosophila, one of which,Ird5, has an essential role in immune response[100–102]. In addition, Ird5 seems to be required forcorrect Toll signalling during dorsoventral axis forma-tion, but only a very small portion of the embryos fromird5 mutant females display a dorsalised phenotype(0.5%), suggesting that redundant factors might actjust upstream of Cactus and Dorsal. This redundancymay explain why these factors escaped identification

ReviewR894

in genetic screens and it may contribute to the robust-ness of the Toll pathway during axis formation.Notably, a role for IKKαα and IKKγγ homologues indorsoventral axis formation has been ruled out[101,102], but they act in the activation of the NfκκBhomologue Relish, which activates expression ofimmune response factors. Furthermore, the Cactusprotein, besides its canonical IκκB kinase recognitionmotif harbours another redundant motif that is respon-sible for signal dependent degradation [97,103].

Besides Cactus, also Dorsal has to be phosphory-lated upon Toll signalling in order to translocate to thenucleus [96,97]. This is revealed by a mutant allele ofDorsal, DorsalS234P, which fails to interact with Cactus.Nuclear translocation of the mutant protein was abol-ished in a GD mutant background. Hence, nucleartranslocation of Dorsal seems to be a three-stepprocess. First, Cactus is phosphorylated releasingDorsal into the cytoplasm. Then, Cactus is ubiquiti-nated and degraded by the proteasome [104,105].Third, Dorsal is phosphorylated, dimerises and entersthe nucleus [106]. It will be interesting to knowwhether Cactus and Dorsal are phosphorylated by thesame kinase.

Surprisingly, there is yet another mechanism ofcontrolling Dorsal nuclear import independent ofcactus [107,108]. WntD (Wnt inhibitor of Dorsal) canblock the nuclear import of Dorsal, even in theabsence of Cactus. However, wntD is expressed onlyat the anterior and posterior termini of the earlyembryo and its deletion does not lead to embryonicpatterning defects, suggesting that it does not play amajor role in establishing the Dorsal gradient. WntDacts also as a feedback inhibitor in the Drosophilainnate immune system where its expression is acti-vated by Toll signaling and downregulates theimmune response. This immune function of WntDappears to be the reason why this regulatory loop hasbeen maintained in evolution.

Regulation of Dorsal activity also occurs at the levelof nuclear entry which is under the control of thenuclear transport machinery. Three factors modulat-ing nuclear transport, Tamo, Drosophila NuclearTransport Factor-2 (DNTF-2) and Members-only(Mbo), the Drosophila homologue of the nuclear poreprotein Nup88, selectively interact with Dorsal[109,110]. Tamo attenuates nuclear translocation ofDorsal, whereas Members-only and DNTF-2 are

Current BiologyR895

Figure 6. Model for the generation of the Spätzle gradient.

(A) Spatial distribution of Easter protease activity (green), initiated by GD in the ventral domain defined by Pipe, is tightly regulated byits inhibitor Serpin27A. Serpin27A activity (light blue) in turn may be under the control of the amino-terminal Spätzle fragment (darkblue), which is produced together with the carboxy-terminal Toll binding fragment by Easter (green). As a result of its presumablyhigher rate of diffusion, the Spätzle amino-terminal fragment is present in a broader domain than the carboxy-terminal fragment. Inlateral regions, the amino-terminal fragment leads to an inhibition of Easter. The resulting Spätzle gradient (red) is stably copied intothe embryo through the graded activation of Pelle (red) which regulates the nuclear transport of Dorsal (pink). Ventrally, high nuclearDorsal concentrations initiate twist transcription, and at more lateral positions, lower nuclear Dorsal concentrations lead to sog tran-scription. (B) An expansion of the Pipe domain reveals the self-regulatory capacity of the system. The inhibition exerted by the amino-terminal fragment of Spätzle has only a certain range. If the region in which the protease cascade is activated expands beyond thisrange of inhibition the activation peak splits and two peaks form, one at each side of the ventral midline (see Figure 5D).

twist sogtwistsogtwistsog sog sog

Easter

Spätzle C-106 (low diffusion rate)

Spätzle amino-terminal fragment (high diffusion rate)

Serpin27A activated Pelle

BA

Pipe Pipe

Cytoplasm

Extracellular

Plasmamembrane

Dorsal

Nucleus

Current Biology

essential for this process. However, these interactionsdo not depend on Toll signalling.

While the question of how Cactus and Dorsalreceive the signal allowing Dorsal to enter the nucleusis still not completely resolved, it seems that down-stream of Pelle no further refinement takes place andthat redundant factors regulate Dorsal to ensure arobust transcriptional response.

However, the system shows plasticity at the level ofthe Dorsal target genes and thus is able to correctpossible fluctuations of the ventralising signal. Forinstance, a reduction of the dose of Spätzle leads to anarrower Twist domain and shifted ventrolateralanlagen. However, such embryos develop into normallarvae [67]. It is not clear at what level the size regula-tion of the anlagen occurs. Twist itself might play animportant role as it forms a gradient, which extendsbeyond the mesodermal region. It is not only requiredfor mesoderm development, but also acts synergisti-cally with Dorsal to regulate the expression of genesrequired for more lateral fates [88].

Dorsoventral Axis Formation and Immunity —Evolutionary ConsiderationsIn Drosophila, there are multiple connections betweendorsoventral axis formation and immunity, both withinthe Toll pathway and in its activation [13,111]. Thelatter relates to an important difference in the way theToll pathway acts in innate immunity in mammals andinsects. The Toll receptors of mammals appear to bedirectly activated by microbial molecules [112], withindividual classes of receptors being specific to par-ticular pathogen-derived molecules. However, inDrosophila, which has nine Toll receptor genes, onlyToll-1, the founding member involved in dorsoventralpatterning, appears to be required for innate immunity[13]. Some of the other Toll receptors are required forparticular developmental processes [113–119], sug-gesting that the diversity of Toll receptors is notrelated to the diversity to microbial patterns. IndeedToll-1 is not activated by microbial molecules, butrather requires activation by Spätzle [13,111]. Micro-bial infections are sensed upstream of Spätzle throughsecreted peptidogylcan recognition proteins (PGRPs)and ββGlucan recognition proteins (ββGRPs) [111,120,121], which are believed to activate specific proteasesthat cleave Spätzle [122,123]. Activation of Toll-1leads to the rapid and massive secretion of antimicro-bial peptides, which contribute to the clearing ofinvading microorganisms from the hemolymph. Thismechanism constitutes together with another signal-ing pathway responsible for the recognition of Gram-negative bacteria (the IMD pathway) the humoralbranch of the pathogen defense [111,121].

While so far none of the upstream proteases actingduring dorsoventral pattern formation have beenfound to be required for the humoral response, oneimportant regulator of the protease cascade,Serpin27A, is required for another branch of thepathogen response, the melanisation reaction[124,125]. In arthropods, melanisation is required forwound healing, encapsulation and sequestration ofmicrobes and production of cytotoxic reactive oxygen

species. Melanisation is controlled by a hemolymph-borne protease cascade, the terminal step of which isinhibited by Spn27A.

Why is there such a close relationship betweenpathogen defense and dorsoventral patterning ininsects? Either axis formation or innate immunity isthe ancestral function of the pathway. While there isno clear evidence that Toll signaling is required fordorsoventral patterning in vertebrates, the function ininnate immunity appears to be conserved. Moreover,similar receptors are even involved in pathogen resis-tance in plants [126]. Thus, it is likely that the role ofthe pathway in axis formation in insects stems from itsearlier role in pathogen defense.

A possible evolutionary scenario can be envisagedfrom certain embryological features of phylogeneti-cally more basal insects [127]. The adaptation to theterrestrial life-style is linked to the formation of largeyolk-rich eggs, which may become the target ofmicrobial infection. During early development, theblastoderm covers the yolk, but only a small portiongives rise to the embryo proper, while the remainderforms a protective extra-embryonic tissue, the serosa.Interestingly, there is evidence that the serosa ofsome insects has an immune function [76,128]. Forinstance, in beetles Dorsal protein is highly expressedin the serosa and transported to the nuclei afterpathogen challenge [76]. By extension, the largeserosa of more basal insects may also have animmune function which was acquired early in evolu-tion to provide protection from infections and thus aselective advantage. We suggest that this was thereason why the Toll pathway and some componentsof the upstream activating cascade were expressed inthe serosa during early embryonic development. Fromthere, only a small shift in temporal and spatialexpression would have been necessary to shift thefunction of the Toll pathway towards axis formation.Initially, it might have only been used to induce axispolarity, while most of the dorsoventral patterning wasachieved by interactions of zygotic genes. In beetleembryos, which may reflect such an intermediate sit-uation, the Dorsal gradient is more transient and influ-ences its fewer target genes less directly compared toDrosophila [76] (S. Roth, unpublished). Thus, the highamount of spatial information encoded in the Dorsalgradient in Drosophila is likely to be a late product ofinsect evolution.

ConclusionsIn Drosophila, the regulatory circuitry comprising thethree serine proteases GD, Snake and Easter, as wellas the Toll ligand Spätzle, accounts for the robustnessof a pathway that establishes a delicate balance ofpattern elements that organise the embryo (Figure 6).Stable entry of the signal into the embryo is mediatedby the kinase Pelle. The intracellular segment of theaxis-forming pathway is, hence, not contributing toformation of the Dorsal gradient, but rather to the sta-bilization and transmission of the spatial informationgenerated outside the embryo. In turn, the Dorsal gra-dient converts this information into patterns of geneexpression, which specify the cell fates of the embryo.

ReviewR896

AcknowledgementsWe thank Sam Mathew for providing the picture used inFigure 5C, and Francesca Peri and Jana Krauss for criticalreading of the mansucript. We are grateful to Hans Meinhardtfor many discussions about pattern formation theory.

References1. Gurdon, J.B., and Bourillot, P.Y. (2001). Morphogen gradient inter-

pretation. Nature 413, 797–803.2. Freeman, M., and Gurdon, J.B. (2002). Regulatory principles of

developmental signaling. Annu. Rev. Cell Dev. Biol. 18, 515–539.3. Lander, A.D., Nie, Q., and Wan, F.Y. (2002). Do morphogen gradi-

ents arise by diffusion? Dev. Cell 2, 785–796.4. Bollenbach, T., Kruse, K., Pantazis, P., Gonzalez-Gaitan, M., and

Julicher, F. (2005). Robust formation of morphogen gradients. Phys.Rev. Lett. 94, 018103.

5. Kruse, K., Pantazis, P., Bollenbach, T., Julicher, F., and Gonzalez-Gaitan, M. (2004). Dpp gradient formation by dynamin-dependentendocytosis: receptor trafficking and the diffusion model. Develop-ment 131, 4843–4856.

6. Eldar, A., Rosin, D., Shilo, B.Z., and Barkai, N. (2003). Self-enhancedligand degradation underlies robustness of morphogen gradients.Dev. Cell 5, 635–646.

7. Eldar, A., Dorfman, R., Weiss, D., Ashe, H., Shilo, B.Z., and Barkai,N. (2002). Robustness of the BMP morphogen gradient inDrosophila embryonic patterning. Nature 419, 304–308.

8. Slack, J.M.W. (1991). From egg to embryo, 2nd Edition (Cambridge:Cambridge University Press).

9. Solnica-Krezel, L. (2003). Vertebrate development: taming the nodalwaves. Curr. Biol. 13, R7–R9.

10. Meinhardt, H., and Gierer, A. (2000). Pattern formation by local self-activation and lateral inhibition. Bioessays 22, 753–760.

11. Morisato, D., and Anderson, K.V. (1995). Signaling pathways thatestablish the dorsal-ventral pattern of the Drosophila embryo. Annu.Rev. Genet. 29, 371–399.

12. Lemaitre, B., Nicolas, E., Michaut, L., Reichhart, J.M., and Hoff-mann, J.A. (1996). The dorsoventral regulatory gene cassettespatzle/Toll/cactus controls the potent antifungal response inDrosophila adults. Cell 86, 973–983.

13. Imler, J.L., and Zheng, L. (2004). Biology of Toll receptors: lessonsfrom insects and mammals. J. Leukoc. Biol. 75, 18–26.

14. Khush, R.S., and Lemaitre, B. (2000). Genes that fight infection:what the Drosophila genome says about animal immunity. TrendsGenet. 16, 442–449.

15. Stathopoulos, A., and Levine, M. (2002). Dorsal gradient networksin the Drosophila embryo. Dev. Biol. 246, 57–67.

16. Rusch, J., and Levine, M. (1996). Threshold responses to the dorsalregulatory gradient and the subdivision of primary tissue territoriesin the Drosophila embryo. Curr. Opin. Genet. Dev. 6, 416–423.

17. Raftery, L.A., and Sutherland, D.J. (2003). Gradients and thresholds:BMP response gradients unveiled in Drosophila embryos. TrendsGenet. 19, 701–708.

18. Mohler, J.D. (1977). Developmental genetics of the Drosophila egg.I. Identification of 59 sex-linked cistrons with maternal effects onembryonic development. Genetics 85, 259–272.

19. Luschnig, S., Moussian, B., Krauss, J., Desjeux, I., Perkovic, J., andNusslein-Volhard, C. (2004). An F1 genetic screen for maternal-effect mutations affecting embryonic pattern formation inDrosophila melanogaster. Genetics 167, 325–342.

20. Schupbach, T., and Wieschaus, E. (1989). Female sterile mutationson the second chromosome of Drosophila melanogaster. I. Mater-nal effect mutations. Genetics 121, 101–117.

21. Luders, F., Segawa, H., Stein, D., Selva, E.M., Perrimon, N., Turco,S.J., and Hacker, U. (2003). Slalom encodes an adenosine 3¢-phos-phate 5¢-phosphosulfate transporter essential for development inDrosophila. EMBO J. 22, 3635–3644.

22. Nusslein-Volhard, C., Lohs-Schardin, M., Sander, K., and Cremer,C. (1980). A dorso-ventral shift of embryonic primordia in a newmaternal-effect mutant of Drosophila. Nature 283, 474–476.

23. Anderson, K.V., and Nusslein-Volhard, C. (1984). Information for thedorsal–ventral pattern of the Drosophila embryo is stored as mater-nal mRNA. Nature 311, 223–227.

24. Schupbach, T. (1987). Germ line and soma cooperate during ooge-nesis to establish the dorsoventral pattern of egg shell and embryoin Drosophila melanogaster. Cell 49, 699–707.

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

26. Roth, S. (2003). The origin of dorsoventral polarity in Drosophila.Philos. Trans. R. Soc. Lond. B Biol. Sci. 358, 1317–1329.

27. Stein, D., Roth, S., Vogelsang, 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.

28. Krem, M.M., and Cera, E.D. (2002). Evolution of enzyme cascadesfrom embryonic development to blood coagulation. TrendsBiochem. Sci. 27, 67–74.

29. Rose, T., LeMosy, E.K., Cantwell, A.M., Banerjee-Roy, D., Skeath,J.B., and Di Cera, E. (2003). Three-dimensional models of proteasesinvolved in patterning of the Drosophila Embryo. Crucial role of pre-dicted cation binding sites. J. Biol. Chem. 278, 11320–11330.

30. 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.

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

32. 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.

33. Markstein, M., Zinzen, R., Markstein, P., Yee, K.P., Erives, A.,Stathopoulos, A., and Levine, M. (2004). A regulatory code for neu-rogenic gene expression in the Drosophila embryo. Development131, 2387–2394.

34. Stathopoulos, A., Van Drenth, M., Erives, A., Markstein, M., andLevine, M. (2002). Whole-genome analysis of dorsal-ventral pat-terning in the Drosophila embryo. Cell 111, 687–701.

35. Stathopoulos, A., and Levine, M. (2005). Localized repressors delin-eate the neurogenic ectoderm in the early Drosophila embryo. Dev.Biol. 280, 482–493.

36. Markstein, M., Markstein, P., Markstein, V., and Levine, M.S. (2002).Genome-wide analysis of clustered Dorsal binding sites identifiesputative target genes in the Drosophila embryo. Proc. Natl. Acad.Sci. USA 99, 763–768.

37. Papatsenko, D., and Levine, M. (2005). Quantitative analysis ofbinding motifs mediating diverse spatial readouts of the Dorsal gra-dient in the Drosophila embryo. Proc. Natl. Acad. Sci. USA 102,4966–4971.

38. Levine, M., and Davidson, E.H. (2005). Gene regulatory networks fordevelopment. Proc. Natl. Acad. Sci. USA 102, 4936–4942.

39. Sen, J., Goltz, J.S., Stevens, L., and Stein, D. (1998). Spatiallyrestricted expression of pipe in the Drosophila egg chamber definesembryonic dorsal-ventral polarity. Cell 95, 471–481.

40. Sen, J., Goltz, J.S., Konsolaki, M., Schupbach, T., and Stein, D.(2000). Windbeutel is required for function and correct subcellularlocalization of the Drosophila patterning protein Pipe. Development127, 5541–5550.

41. Nilson, L.A., and Schupbach, T. (1998). Localized requirements forwindbeutel and pipe reveal a dorsoventral prepattern within the fol-licular epithelium of the Drosophila ovary. Cell 93, 253–262.

42. LeMosy, E.K., Kemler, D., and Hashimoto, C. (1998). Role of Nudelprotease activation in triggering dorsoventral polarization of theDrosophila embryo. Development 125, 4045–4053.

43. Nilson, L.A., and Schupbach, T. (1999). EGF receptor signaling inDrosophila oogenesis. Curr. Top. Dev. Biol. 44, 203–243.

44. Pai, L.M., Barcelo, G., and Schupbach, T. (2000). D-cbl, a negativeregulator of the Egfr pathway, is required for dorsoventral pattern-ing in Drosophila oogenesis. Cell 103, 51–61.

45. Peri, F., Technau, M., and Roth, S. (2002). Mechanisms of Gurken-dependent pipe regulation and the robustness of dorsoventral pat-terning in Drosophila. Development 129, 2965–2975.

46. Roth, S., Neuman-Silberberg, F.S., Barcelo, G., and Schupbach, T.(1995). cornichon and the EGF receptor signaling process are nec-essary for both anterior-posterior and dorsal-ventral pattern forma-tion in Drosophila. Cell 81, 967–978.

47. Zhu, X., Sen, J., Stevens, L., Goltz, J.S., and Stein, D. (2005).Drosophila Pipe protein activity in the ovary and the embryonic sali-vary gland does not require heparan sulfate glycosaminoglycans.Development 132, 3813–3822.

48. Konsolaki, M., and Schupbach, T. (1998). windbeutel, a generequired for dorsoventral patterning in Drosophila, encodes aprotein that has homologies to vertebrate proteins of the endoplas-mic reticulum. Genes Dev. 12, 120–131.

49. LeMosy, E.K., Leclerc, C.L., and Hashimoto, C. (2000). Biochemicaldefects of mutant nudel alleles causing early developmental arrestor dorsalization of the Drosophila embryo. Genetics 154, 247–257.

50. Turcotte, C.L., and Hashimoto, C. (2002). Evidence for a gly-cosaminoglycan on the nudel protein important for dorsoventralpatterning of the drosophila embryo. Dev. Dyn. 224, 51–57.

51. Dissing, M., Giordano, H., and DeLotto, R. (2001). Autoproteolysisand feedback in a protease cascade directing Drosophila dorsal-ventral cell fate. EMBO J. 20, 2387–2393.

Current BiologyR897

52. DeLotto, R. (2001). Gastrulation defective, a complement factorC2/B-like protease, interprets a ventral prepattern in Drosophila.EMBO Rep. 2, 721–726.

53. Han, J.H., Lee, S.H., Tan, Y.Q., LeMosy, E.K., and Hashimoto, C.(2000). Gastrulation defective is a serine protease involved in acti-vating the receptor toll to polarize the Drosophila embryo. Proc.Natl. Acad. Sci. USA 97, 9093–9097.

54. Ponomareff, G., Giordano, H., DeLotto, Y., and DeLotto, R. (2001).Interallelic complementation at the Drosophila melanogaster gas-trulation defective locus defines discrete functional domains of theprotein. Genetics 159, 635–645.

55. Konrad, K.D., Goralski, T.J., Mahowald, A.P., and Marsh, J.L. (1998).The gastrulation defective gene of Drosophila melanogaster is amember of the serine protease superfamily. Proc. Natl. Acad. Sci.USA 95, 6819–6824.

56. LeMosy, E.K., Tan, Y.Q., and Hashimoto, C. (2001). Activation of aprotease cascade involved in patterning the Drosophila embryo.Proc. Natl. Acad. Sci. USA 98, 5055–5060.

57. Arlaud, G.J., Volanakis, J.E., Thielens, N.M., Narayana, S.V., Rossi,V., and Xu, Y. (1998). The atypical serine proteases of the comple-ment system. Adv. Immunol. 69, 249–307.

58. Rushlow, C. (2004). Dorsoventral patterning: a serpin pinned downat last. Curr. Biol. 14, R16–R18.

59. Hashimoto, C., Kim, D.R., Weiss, L.A., Miller, J.W., and Morisato, D.(2003). Spatial regulation of developmental signaling by a serpin.Dev. Cell 5, 945–950.

60. Ligoxygakis, P., Roth, S., and Reichhart, J.M. (2003). A serpin reg-ulates dorsal-ventral axis formation in the Drosophila embryo. Curr.Biol. 13, 2097–2102.

61. Chang, A.J., and Morisato, D. (2002). Regulation of Easter activity isrequired for shaping the Dorsal gradient in the Drosophila embryo.Development 129, 5635–5645.

62. Misra, S., Hecht, P., Maeda, R., and Anderson, K.V. (1998). Positiveand negative regulation of Easter, a member of the serine proteasefamily that controls dorsal-ventral patterning in the Drosophilaembryo. Development 125, 1261–1267.

63. Smith, C., Giordano, H., and DeLotto, R. (1994). Mutational analysisof the Drosophila snake protease: an essential role for domainswithin the proenzyme polypeptide chain. Genetics 136, 1355–1365.

64. DeLotto, Y., and DeLotto, R. (1998). Proteolytic processing of theDrosophila Spatzle protein by easter generates a dimeric NGF-likemolecule with ventralising activity. Mech. Dev. 72, 141–148.

65. DeLotto, Y., Smith, C., and DeLotto, R. (2001). Multiple isoforms ofthe Drosophila Spatzle protein are encoded by alternatively splicedmaternal mRNAs in the precellular blastoderm embryo. Mol. Gen.Genet. 264, 643–652.

66. Weber, A.N., Tauszig-Delamasure, S., Hoffmann, J.A., Lelievre, E.,Gascan, H., Ray, K.P., Morse, M.A., Imler, J.L., and Gay, N.J. (2003).Binding of the Drosophila cytokine Spatzle to Toll is direct andestablishes signaling. Nat. Immunol. 4, 794–800.

67. Morisato, D. (2001). Spatzle regulates the shape of the Dorsal gra-dient in the Drosophila embryo. Development 128, 2309–2319.

68. Campenot, R.B., and MacInnis, B.L. (2004). Retrograde transport ofneurotrophins: fact and function. J. Neurobiol. 58, 217–229.

69. Lai, E.C., Deblandre, G.A., Kintner, C., and Rubin, G.M. (2001).Drosophila neuralized is a ubiquitin ligase that promotes the inter-nalization and degradation of delta. Dev. Cell 1, 783–794.

70. Neet, K.E., and Campenot, R.B. (2001). Receptor binding, internal-ization, and retrograde transport of neurotrophic factors. Cell. Mol.Life Sci. 58, 1021–1035.

71. Epstein, I.R., and Showalter, K. (1996). Nonlinear chemical dynam-ics: oscillations, patterns, and chaos. J. Phys. Chem. 100, 13132.

72. Kapral, R., and Showalter, K. (1995). Chemical Waves and Patterns(Dordrecht: Kluwer Academic Publishers).

73. Meinhardt, H. (2004). Different strategies for midline formation inbilaterians. Nat. Rev. Neurosci. 5, 502–510.

74. Meinhardt, H. (2004). Models for the generation of the embryonicbody axes: ontogenetic and evolutionary aspects. Curr. Opin.Genet. Dev. 14, 446–454.

75. Meinhardt, H., and Gierer, A. (1974). Applications of a theory of bio-logical pattern formation based on lateral inhibition. J. Cell Sci. 15,321–346.

76. Chen, G., Handel, K., and Roth, S. (2000). The maternal NF-kappaB/dorsal gradient of Tribolium castaneum: dynamics of earlydorsoventral patterning in a short-germ beetle. Development 127,5145–5156.

77. Zarnitsina, V.I., Ataullakhanov, F.I., Lobanov, A.I., and Morozova,O.L. (2001). Dynamics of spatially nonuniform patterning in themodel of blood coagulation. Chaos 11, 57–70.

78. Zarnitsina, V.I., Pokhilko, A.V., and Ataullakhanov, F.I. (1996). Amathematical model for the spatio-temporal dynamics of intrinsicpathway of blood coagulation. II. Results. Thromb. Res. 84,333–344.

79. Zarnitsina, V.I., Pokhilko, A.V., and Ataullakhanov, F.I. (1996). Amathematical model for the spatio-temporal dynamics of intrinsicpathway of blood coagulation. I. The model description. Thromb.Res. 84, 225–236.

80. Medzhitov, R., Preston-Hurlburt, P., Kopp, E., Stadlen, A., Chen, C.,Ghosh, S., and Janeway, C.A., Jr. (1998). MyD88 is an adaptorprotein in the hToll/IL-1 receptor family signaling pathways. Mol.Cell 2, 253–258.

81. Charatsi, I., Luschnig, S., Bartoszewski, S., Nusslein-Volhard, C.,and Moussian, B. (2003). Krapfen/dMyd88 is required for the estab-lishment of dorsoventral pattern in the Drosophila embryo. Mech.Dev. 120, 219–226.

82. Kambris, Z., Bilak, H., D’Alessandro, R., Belvin, M., Imler, J.L., andCapovilla, M. (2003). DmMyD88 controls dorsoventral patterning ofthe Drosophila embryo. EMBO Rep. 4, 64–69.

83. Sun, H., Towb, P., Chiem, D.N., Foster, B.A., and Wasserman, S.A.(2004). Regulated assembly of the Toll signaling complex drivesDrosophila dorsoventral patterning. EMBO J. 23, 100–110.

84. Towb, P., Galindo, R.L., and Wasserman, S.A. (1998). Recruitmentof Tube and Pelle to signaling sites at the surface of the Drosophilaembryo. Development 125, 2443–2450.

85. Towb, P., Bergmann, A., and Wasserman, S.A. (2001). The proteinkinase Pelle mediates feedback regulation in the Drosophila Tollsignaling pathway. Development 128, 4729–4736.

86. Shen, B., and Manley, J.L. (2002). Pelle kinase is activated byautophosphorylation during Toll signaling in Drosophila. Develop-ment 129, 1925–1933.

87. Horng, T., and Medzhitov, R. (2001). Drosophila MyD88 is anadapter in the Toll signaling pathway. Proc. Natl. Acad. Sci. USA 98,12654–12658.

88. Stathopoulos, A., and Levine, M. (2002). Linear signaling in the Toll-Dorsal pathway of Drosophila: activated Pelle kinase specifies allthreshold outputs of gene expression while the bHLH protein Twistspecifies a subset. Development 129, 3411–3419.

89. Galindo, R.L., Edwards, D.N., Gillespie, S.K., and Wasserman, S.A.(1995). Interaction of the pelle kinase with the membrane-associ-ated protein tube is required for transduction of the dorsoventralsignal in Drosophila embryos. Development 121, 2209–2218.

90. Grosshans, J., Bergmann, A., Haffter, P., and Nusslein-Volhard, C.(1994). Activation of the kinase Pelle by Tube in the dorsoventralsignal transduction pathway of Drosophila embryo. Nature 372,563–566.

91. Grosshans, J., Schnorrer, F., and Nusslein-Volhard, C. (1999).Oligomerisation of Tube and Pelle leads to nuclear localisation ofdorsal. Mech. Dev. 81, 127–138.

92. Hu, X., Yagi, Y., Tanji, T., Zhou, S., and Ip, Y.T. (2004). Multimeriza-tion and interaction of Toll and Spatzle in Drosophila. Proc. Natl.Acad. Sci. USA 101, 9369–9374.

93. Norris, J.L., and Manley, J.L. (1996). Functional interactionsbetween the pelle kinase, Toll receptor, and tube suggest a mech-anism for activation of dorsal. Genes Dev. 10, 862–872.

94. Shen, B., and Manley, J.L. (1998). Phosphorylation modulates directinteractions between the Toll receptor, Pelle kinase and Tube.Development 125, 4719–4728.

95. Edwards, D.N., Towb, P., and Wasserman, S.A. (1997). An activity-dependent network of interactions links the Rel protein Dorsal withits cytoplasmic regulators. Development 124, 3855–3864.

96. Drier, E.A., Huang, L.H., and Steward, R. (1999). Nuclear import ofthe Drosophila Rel protein Dorsal is regulated by phosphorylation.Genes Dev. 13, 556–568.

97. Drier, E.A., Govind, S., and Steward, R. (2000). Cactus-independentregulation of Dorsal nuclear import by the ventral signal. Curr. Biol.10, 23–26.

98. Israel, A. (2000). The IKK complex: an integrator of all signals thatactivate NF-kappaB? Trends Cell Biol. 10, 129–133.

99. Karin, M., and Ben-Neriah, Y. (2000). Phosphorylation meets ubiq-uitination: the control of NF-[kappa]B activity. Annu. Rev. Immunol.18, 621–663.

100. Wu, L.P., Choe, K.M., Lu, Y., and Anderson, K.V. (2001). Drosophilaimmunity: genes on the third chromosome required for theresponse to bacterial infection. Genetics 159, 189–199.

101. Lu, Y., Wu, L.P., and Anderson, K.V. (2001). The antibacterial arm ofthe drosophila innate immune response requires an IkappaB kinase.Genes Dev. 15, 104–110.

102. Silverman, N., Zhou, R., Stoven, S., Pandey, N., Hultmark, D., andManiatis, T. (2000). A Drosophila IkappaB kinase complex requiredfor Relish cleavage and antibacterial immunity. Genes Dev. 14,2461–2471.

103. Fernandez, N.Q., Grosshans, J., Goltz, J.S., and Stein, D. (2001).Separable and redundant regulatory determinants in Cactusmediate its dorsal group dependent degradation. Development 128,2963–2974.

ReviewR898

104. Bergmann, A., Stein, D., Geisler, R., Hagenmaier, S., Schmid, B.,Fernandez, N., Schnell, B., and Nusslein-Volhard, C. (1996). A gra-dient of cytoplasmic Cactus degradation establishes the nuclearlocalization gradient of the dorsal morphogen in Drosophila. Mech.Dev. 60, 109–123.

105. Reach, M., Galindo, R.L., Towb, P., Allen, J.L., Karin, M., andWasserman, S.A. (1996). A gradient of cactus protein degradationestablishes dorsoventral polarity in the Drosophila embryo. Dev.Biol. 180, 353–364.

106. Isoda, K., and Nusslein-Volhard, C. (1994). Disulfide cross-linking incrude embryonic lysates reveals three complexes of the Drosophilamorphogen dorsal and its inhibitor cactus. Proc. Natl. Acad. Sci.USA 91, 5350–5354.

107. Ganguly, A., Jiang, J., and Ip. Y.T. (2005). Drosophila WntD is atarget and an inhibitor of the Dorsal/Twist/Snail network in the gas-trulating embryo. Development 132, 3419–3429.

108. Gordon, M.D., Dionne, M.S., Schneider, D.S., and Nusse, R. (2005).WntD is a feedback inhibitor of Dorsal/NF-kappaB in Drosophiladevelopment and immunity. Nature 437, 746–749.

109. Minakhina, S., Yang, J., and Steward, R. (2003). Tamo selectivelymodulates nuclear import in Drosophila. Genes Cells 8, 299–310.

110. Uv, A.E., Roth, P., Xylourgidis, N., Wickberg, A., Cantera, R., andSamakovlis, C. (2000). members only encodes a Drosophila nucle-oporin required for rel protein import and immune response activa-tion. Genes Dev. 14, 1945–1957.

111. Ferrandon, D., Imler, J.L., and Hoffmann, J.A. (2004). Sensing infec-tion in Drosophila: Toll and beyond. Semin. Immunol. 16, 43–53.

112. Athman, R., and Philpott, D. (2004). Innate immunity via Toll-likereceptors and Nod proteins. Curr. Opin. Microbiol., 7, 25–32.

113. Kambris, Z., Hoffmann, J.A., Imler, J.L., and Capovilla, M. (2002).Tissue and stage-specific expression of the Tolls in Drosophilaembryos. Gene Expr. Patterns 2, 311–331.

114. Chiang, C., and Beachy, P.A. (1994). Expression of a novel Toll-likegene spans the parasegment boundary and contributes to hedge-hog function in the adult eye of Drosophila. Mech. Dev. 47, 225–239.

115. Eldon, E., Kooyer, S., D'Evelyn, D., Duman, M., Lawinger, P., Botas,J., and Bellen, H. (1994). The Drosophila 18 wheeler is required formorphogenesis and has striking similarities to Toll. Development120, 885–899.

116. Williams, M.J., Rodriguez, A., Kimbrell, D.A., and Eldon, E.D. (1997).The 18-wheeler mutation reveals complex antibacterial gene regu-lation in Drosophila host defense. EMBO J. 16, 6120–6130.

117. Seppo, A., Matani, P., Sharrow, M., and Tiemeyer, M. (2003). Induc-tion of neuron-specific glycosylation by Tollo/Toll-8, a DrosophilaToll-like receptor expressed in non-neural cells. Development 130,1439–1448.

118. Halfon, M.S., and Keshishian, H. (1998). The Toll pathway isrequired in the epidermis for muscle development in the Drosophilaembryo. Dev. Biol. 199, 164–174.

119. Halfon, M.S., Hashimoto, C., and Keshishian, H. (1995). TheDrosophila toll gene functions zygotically and is necessary forproper motoneuron and muscle development. Dev. Biol. 169,151–167.

120. Gobert, V., Gottar, M., Matskevich, A.A., Rutschmann, S., Royet, J.,Belvin, M., Hoffmann, J.A., and Ferrandon, D. (2003). Dual activationof the Drosophila toll pathway by two pattern recognition receptors.Science 302, 2126–2130.

121. Royet, J., Reichhart, J.M., and Hoffmann, J.A. (2005). Sensing andsignaling during infection in Drosophila. Curr. Opin. Immunol. 17,11–17.

122. Ligoxygakis, P., Pelte, N., Hoffmann, J.A., and Reichhart, J.M.(2002). Activation of Drosophila Toll during fungal infection by ablood serine protease. Science 297, 114–116.

123. Levashina, E.A., Langley, E., Green, C., Gubb, D., Ashburner, M.,Hoffmann, J.A., and Reichhart, J.M. (1999). Constitutive activationof toll-mediated antifungal defense in serpin-deficient Drosophila.Science 285, 1917–1919.

124. Ligoxygakis, P., Pelte, N., Ji, C., Leclerc, V., Duvic, B., Belvin, M.,Jiang, H., Hoffmann, J.A., and Reichhart, J.M. (2002). A serpinmutant links Toll activation to melanization in the host defence ofDrosophila. EMBO J. 21, 6330–6337.

125. De Gregorio, E., Han, S.J., Lee, W.J., Baek, M.J., Osaki, T., Kawa-bata, S., Lee, B.L., Iwanaga, S., Lemaitre, B, and Brey, P.T. (2002).An immune-responsive Serpin regulates the melanization cascadein Drosophila. Dev. Cell. 3, 581–592.

126. Zipfel, C., and Felix, G. (2005). Plants and animals: a different tastefor microbes? Curr. Opin. Plant Biol. 8, 353–360.

127. Roth, S. (2004). Gastrulation in other insects. In Gastrulation: FromCells to Embryos, C. Stern, ed. (New York: Cold Spring Harbor Lab-oratory Press), pp. 105–121.

128. Gorman, M.J., Kankanala, P., and Kanost, M.R. (2004). Bacterialchallenge stimulates innate immune responses in extra-embryonictissues of tobacco hornworm eggs. Insect Mol. Biol. 13, 19–24.

Current BiologyR899