The maternal NF-κB/Dorsal gradient of Tribolium castaneum: dynamics of early dorsoventral patterning in a short-germ beetle

Transcription

1 Development 127, (2000) Printed in Great Britain The Company of Biologists Limited 2000 DEV The maternal NF-κB/Dorsal gradient of Tribolium castaneum: dynamics of early dorsoventral patterning in a short-germ beetle Gang Chen 1,, Klaus Handel 1, and Siegfried Roth 1, *, 1 Max-Planck-Institut für Entwicklungsbiologie, Spemannstrasse 35/II, D Tübingen, Germany Present address: Universität Tübingen, Zoologisches Institut, LS Entwicklungsphysiologie, Auf der Morgenstelle 28, D Tübingen, Germany Present address: Aventis Research & Technologies GmbH & Co KG, Operative Forschung. Industriepark Höchst. Gebäude G 830, D Frankfurt am Main, Germany Present address: Institut für Entwicklungsbiologie, Universität zu Köln, Gyrhofstrasse 17, D Köln, USA *Author for correspondence ( siegfried.roth@uni-koeln.de) Accepted 25 September; published on WWW 2 November 2000 SUMMARY In the long-germ insect Drosophila melanogaster dorsoventral polarity is induced by localized Toll-receptor activation which leads to the formation of a nuclear gradient of the rel/ NF-κB protein Dorsal. Peak levels of nuclear Dorsal are found in a ventral stripe spanning the entire length of the blastoderm embryo allowing all segments and their dorsoventral subdivisions to be synchronously specified before gastrulation. We show that a nuclear Dorsal protein gradient of similar anteroposterior extension exists in the short-germ beetle, Tribolium castaneum, which forms most segments from a posterior growth zone after gastrulation. In contrast to Drosophila, (i) nuclear accumulation is first uniform and then becomes progressively restricted to a narrow ventral stripe, (ii) gradient refinement is accompanied by changes in the zygotic expression of the Tribolium Toll-receptor suggesting feedback regulation and, (iii) the gradient only transiently overlaps with the expression of a potential target, the Tribolium twist homolog, and does not repress Tribolium decapentaplegic. No nuclear Dorsal is seen in the cells of the growth zone of Tribolium embryos, indicating that here dorsoventral patterning occurs by a different mechanism. However, Dorsal is up-regulated and transiently forms a nuclear gradient in the serosa, a protective extraembryonic cell layer ultimately covering the whole embryo. Key words: Insect embryogenesis, Evolution of development, Toll, twist, decapentaplegic, zerknüllt, Nuclear concentration gradient, Innate immunity, Tribolium castaneum INTRODUCTION Drosophila embryos represent an extreme case of long-germ development, a mode of embryogenesis which is likely to be derived since it is found only in other holometabolous insect orders, but is absent in the more primitive hemimetabolous groups (Sander, 1976). The long-germ type of development is characterised by two features. First, all segments along the anteroposterior (AP) axis form more or less synchronously and their anlagen are established before gastrulation commences. Secondly, the anlagen of the embryo extend from the anterior to the posterior tip of the blastoderm. Most of the blastoderm gives rise to the embryo proper while only a small portion develops into extraembryonic coverings. In short-germ insects on the other hand, only head segments are specified prior to gastrulation while more posterior segments emerge later from a posterior growth zone. Additionally, the anlagen of the embryo may occupy only a small region of the blastoderm while the majority of the blastoderm cells form the serosa, the outer extraembryonic membrane. The long-germ mode of development places high demands on the amount of positional information required before gastrulation for metameric and dorsoventral (DV) patterning. Positional information for DV patterning must exist along the entire egg length since mesoderm invagination, like segmentation, occurs synchronously along the anteroposterior axis. In Drosophila this is accomplished through an extraembryonic signaling cascade, which is activated along the ventral side of the egg. Ventral activation is triggered by modifications of the extracellular matrix, which can be traced back to the establishment of DV polarity during oogenesis (Nilson and Schüpbach, 1998; Sen et al., 1998). The embryo senses the extracellular signal through the transmembrane receptor Toll which is maternally provided and distributed uniformly in the plasma membrane (Morisato and Anderson, 1995). Ventrally activated Toll stimulates the nuclear import of the rel/nf-κb protein Dorsal which, prior to signaling, is evenly distributed in the cytoplasm where is forms a complex with the IκB-like Cactus protein (Belvin et al., 1995; Bergmann et al., 1996). As a result of ventral signaling, a nuclear Dorsal protein gradient forms with peak levels along the ventral midline. Thus, the activation pattern which leads to Dorsal gradient formation is likely to be a stripe along the

2 5146 G. Chen, K. Handel and S. Roth entire ventral side of the egg. How this activation pattern arises with such precision is still the subject of ongoing research (for review see Roth, 1998). We wondered how much DV positional information exists in a short-germ embryo before gastrulation and how much of this information is maternally provided. This question is interesting not only because the embryo proper forms from a more restricted region of the blastoderm and therefore global DV information along the entire egg length might not be necessary. In addition, many classical ligation and fragmentation experiments have provided evidence for extensive regulation of DV patterning in short- and intermediate- germ embryos of hemimetabolous as well as holometabolous insects (Sander, 1976). Impressive examples have been described in which apparently complete embryos form from a variety of DV egg fragments. Although rudimentary forms of DV pattern regulation exist even in Drosophila (Roth and Schüpbach, 1994; Roth et al., 1999), it is unlikely that Drosophila has the regulatory capacity observed in classical experiments with lower insects since important aspects of DV axis specification occur early during Drosophila oogenesis (Nilson and Schüpbach, 1998; Sen et al., 1998). This suggests that during insect evolution major changes of the mechanisms of DV axis formation took place. Besides its function in axis formation the Toll-rel/NF-κB pathway is also required to mediate the innate immune response in Drosophila. Here, pathway activation induces the expression of potent antimicrobial peptides upon septic injury (Anderson, 2000). Not all pathway components are identical in both contexts. Thus, for injury-induced signaling Dorsal is replaced by two other rel/nf-κb proteins, dorsal-related immunity factor (Dif) and Relish (Ip et al., 1993; Dushay et al., 1996). Nevertheless, the employment of Toll signaling both in axis formation and immunity poses the interesting question of its ancestral role and the evolutionary path that leads to its functional diversification. To approach both the evolution of the Toll pathway and the mechanistic alterations of DV pattern formation among insects we have chosen the short-germ beetle Tribolium castaneum for comparison with Drosophila melanogaster. Molecular and genetic techniques have been established for Tribolium (Berghammer et al., 1999) and some zygotic DV genes have already been characterised (Sommer and Tautz, 1994; Falciani et al., 1996; Sanchez-Salazar et al., 1996). On the basis of the expression pattern of the snail, twist and dpp homologs of Tribolium the existence of a Dorsal-like activity had been suggested (Sommer and Tautz, 1994; Sanchez-Salazar et al., 1996). Here, we show that in Tribolium embryos a maternally expressed rel/nf-κb protein indeed exists which is highly similar to Drosophila melanogaster Dorsal (Dm-dl). Like Dmdl, it forms a nuclear concentration gradient at the blastoderm stage. We describe the spatial and temporal dynamics of gradient formation and compare it to the expression of Tribolium Toll (Maxton-Küchenmeister et al., 1999) as well as to the expression of potential Dorsal target genes of Tribolium. This analysis firstly suggests positive feedback control of gradient formation involving Dorsal-dependent activation of Toll. Secondly, is shows that the relationship between different nuclear Dorsal concentrations and the cell fates along the dorsoventral axis is more indirect in Tribolium as compared to Drosophila. Finally, early up-regulation of Dorsal in the serosa of Tribolium embryos leads us to some speculations about the origin of the patterning function of the Toll/Dorsal pathway. MATERIALS AND METHODS Tribolium and Drosophila stocks The beetle stock (ecotype San Bernadino) was kept and eggs collected as described by Beermann (1998). The following Drosophila melanogaster stocks were used. Wild type: Oregon R, dl T : w; In(2L) cn pr bw/cyo, b, dl I5 : w; dl I5 cn bw/cyo DTS, Df(2L)TW119: w; Df(2L)TW119 cn bw/cyo (Roth et al., 1989), and α-tubgal4:vp16: maternal GAL4 driver line on the X chromosome provided by D. St Johnston (Maxton-Küchenmeister et al., 1999). Cloning of the Tc-dl homolog Two fragments covering the entire rel homology domain (RHD) of Dm-dl were used to screen an embryonic cdna library of Tribolium castaneum (Wolff et al., 1995) employing low stringency conditions. One cdna of 2.2 kb was recovered and used subsequently for screening under high stringency conditions, which lead to the isolation of 7 additional cdnas of identical length and the same restriction maps. One was sequenced completely and two others were shown to have identical 5 -ends suggesting that the obtained sequence (2192 bp) represents a complete cdna. It consists of a 233 bp 5 -untranslated region (UTR), an 1671 bp open reading frame and a 288 bp 3 UTR with a typical polyadenlylation signal followed by a poly(a) tail. Rescue constructs Rescue was not obtained using pcasperbcd (Stein et al., 1998) even after the 5 - and 3 -UTRs of Tc-dl were replaced by the corresponding UTRs from Dm-dl. The RNA produced from pcasperbcd [Tc-dl] transgenes appears to be retained in the nurse cells. puast[tc-dl] with α-tubgal4:vp16 as driver also showed no rescue. The final rescue construct was puasp[tc-dl]. The full-length cdna of Tc-dl was cloned into the BamHI/ XbaI site of puasp (Rørth, 1998). puasp [Dmdl]: Dm-dl full-length cdna was cloned into the KpnI/XbaI of puasp. RNA injections Dm-dl or Tc-dl cdnas were cloned into the EcoRI/XhoI site of pcs2 (Rupp et al., 1994) and RNA injections were done as described by Jazwinska et al. (1999). Production and purification of Tc-dl antibodies Two fusion protein constructs were made by subcloning the RHD of Tcdl into the His-tag vector prset-a (Invitrogen) and GST-tag vector pgex-2t (Pharmacia), respectively. The prset-a-derived fusion protein was used to inject rabbits using a standard immunisation protocol (Harlow and Lane, 1988). The serum from boosted rabbits was affinitypurified using the GST fusion protein. The antibodies were tested by western blot analysis using both fusion proteins and embryonic extracts essentially as described previously (Roth et al., 1989). Immunohistochemistry Immunostaining and in situ hybridisation was done essentially as described previously (Roth et al., 1989; Tautz and Pfeifle, 1989). YOYO-1 (Molecular Probes) and DAPI stainings were done subsequently. RESULTS A rel/nf-κb protein from Tribolium castaneum with high sequence similarity to Drosophila Dorsal Three rel/nf-κb genes have been identified from Drosophila melanogaster: dorsal (Dm-dl), dorsal-related immunity factor (Dif) and relish (Steward, 1987; Ip et al., 1993; Dushay et al.,

3 The Tribolium Dorsal gradient 5147 Fig. 1. The Dorsal protein of Tribolium castaneum. (A) Comparison of the amino acid sequences of Dorsal from Tribolium castaneum (Tc-dl), Drosophila melanogaster (Dm-dl) and Gambiae immune factor 1 (Gambif 1) from Anopheles gambiae. Shading of identical amino acids reveals significant conservation only in the rel homology domain (boxed in red). The GenBank Accession Number for Tc-dl is AY (B) Domain structure of Tc-dl, Dm-dl, Gambif1 and Dif. The numbers refer to the percent identity in the rel homology domain between Tc-dl, Dm-dl, Gambif1 and Dif. Although there is no significant sequence conservation outside the Rel homology domain, Dm-dl, Gambif1 and Dif contain polyglutamine, polyasparagine or polyalanine stretches in the C-terminal, potential transactivation domain which are absent in Tc-dl. (C) Phylogenetic comparison with other known rel/nf-κb family members. EMBL Data Library accession nos: Rel A (p65 human), M62399; Rel B (human), M83221; C-Rel (human), M99576; NF-κB1 (p100 human), X61498; NF-κB2 (p105 human), M55643; Dif, L29015; Dm Dorsal, M23702; Gambif 1, X The sequences were aligned using the CLUSTAL method (MEGALIN from Lasergene with Gap penalty = 10 and Gap length penalty = 10, Higgins et al., 1989). Tree reconstruction was done with PAUP (Swofford, 1998) using the neighbour-joining algorithm (standard settings without transition matrix). The bootstrap values support the close phylogenetic relationship between Tc-dl, Dm-dl and Gambif ) which show significant similarity only in an N-terminal domain of approximately 300 amino acids responsible for DNA binding, dimerisation and IκB/Cactus interaction, called the rel homology domain (RHD; Verma et al., 1995). The RHDs of Dif and relish show only 42% and 26% identity, respectively to the RHD of dorsal, being on average, as far diverged from dorsal as the vertebrate members of the rel/nfκb family (Fig. 1C). Thus, potential dorsal orthologs from other species are likely to be recognised on the bases of their RHD sequences. We therefore used probes generated from the RHD of Dm-dl for low stringency screening of an embryonic Tribolium castaneum cdna library and recovered a 2.2 kb cdna with a single open reading frame coding for a 556 amino acid protein (Fig. 1A). The conceptual translation revealed an N-terminal RHD which is highly similar to that of Dm-dl (77% identity) and gambif1 (70% identity), a recently identified Dorsal-like protein form Anopheles gambiae with a possible role in the immune system (Barillas-Mury et al., 1996). The sequences flanking the RHD show no similarity to known proteins and contain no recognisable protein motifs. In particular, the C-terminal region which by analogy to other rel/nf-κb proteins is likely to be a transactivation domain, does not contain the polyglutamine, -alanine or -asparagine stretches characteristic of the C-termini of Dm-dl, Dif and

4 5148 G. Chen, K. Handel and S. Roth Fig. 2. Tribolium dorsal mrna and protein are maternally expressed. (A-D) Tc-dl mrna distribution. (A,C) Anterior to the left and dorsal up. (A,B) Preblastoderm. Whole-mount (A) and cross section (B) show uniform mrna distribution around the DV circumference. The RNA is concentrated in the cortex of the egg, but lower amounts are also present in the central yolk-rich cytoplasm. (C,D) Primitive pit formation and serosal segregation. DAPI staining to visualise the nuclei was performed after RNA in situ hybridisation. The same embryo is shown with bright-field plus fluorescence (C) and fluorescence only (D). The border between presumptive serosa and germ rudiment is indicated by arrows. High levels of Tc-dl mrna are expressed in serosa cells. (E) Western blot with Tribolium embryonic extracts. The age of the embryos for extract preparation is given in hours. Commassie Blue staining was used to normalise the amounts of protein in each lane (data not shown). gambif1 (Fig. 1B). We call this new rel/nf-κb protein Tribolium castaneum-dorsal (Tc-dl) because of the protein distribution in early Tribolium embryos (see below). Both mrna and protein of Tribolium castaneumdorsal are maternally expressed Tc-dl mrna is present in freshly laid eggs and preblastoderm embryos where it is evenly distributed around the embryonic circumference (Fig. 2A,B). Transverse sections show that the mrna is concentrated in the cortex of the embryo while only small amounts are present between big yolk granules in the centre. This distribution remains unchanged during blastoderm stages (data not shown). However, shortly before gastrulation Fig. 3. The formation and refinement of the nuclear Tc-dl gradient. (A,C,E-H) Bright-field images of whole embryos stained with anti- Tc-dl antibodies. (B,D) Fluorescence images of the embryos shown in A and C, respectively that were also stained with a DNA dye (YOYO-1) to visualise the cleavage nuclei. The insets in A,B and C,D show magnified energids under bright-field optics (left) and bright-field optics plus fluorescence (right). (A,B) Nuclear cycle 7. The nuclei have not reached the periphery. Tc-dl protein is present in the cytoplasm surrounding the nuclei. (C,D) Nuclear cycle 8. The nuclei reach the egg cortex and accumulate Tc-dl protein. (E-H). Consecutive stages of gradient refinement, which occur between nuclear cycle 9 (E) and the onset of serosal segregation (H). The arrow points to the anterior cap expression of Tc-dl which coincides with the presumptive serosa. starts the mrna concentrations increase in an anterior cap and, subsequently, high mrna levels are found in all presumptive serosa cells while mrna levels decrease in the germ rudiment (Fig. 2C,D). The RNA continues to be expressed in the serosa during later stages of embryonic development. Tc-dl mrna accumulates in the oocytes of mid and late stage egg chambers (data not shown) suggesting that Tc-dl, like Dm-dl mrna (Steward et al., 1985) is maternally provided. This assumption is supported by western blot analysis using

5 The Tribolium Dorsal gradient 5149 antibodies raised against a bacterially produced fusion protein which contained the entire RHD of Tc-dl (Fig. 2E). In embryonic extracts the antibodies recognise two protein bands with an apparent molecular mass of 63 kda and 35 kda. The upper band is likely to correspond to the full-length protein, which has a predicted molecular mass of 62 kda while the lower band might be a degradation product since its developmental profile mirrors that of the upper band. Tc-dl protein is produced maternally as it can be detected during oogenesis (Fig. 2E). During early embryogenesis protein amounts are low, but they increase significantly when serosal segregation starts (9 hours after egg deposition; Handel et al., 2000) and high levels of protein can be detected throughout embryonic development. Tc-dl protein forms a nuclear concentration gradient in blastoderm embryos As in Drosophila preblastoderm embryos, early nuclear divisions take place in the yolk-rich centre of the Tribolium egg. The cloud of nuclei expands along the anteroposterior egg axis and then the nuclei migrate towards the cortex (Grünfelder, 1997). Before they reach the cortex the nuclei are surrounded by small islands of cytoplasm (energids) which show weak anti-tc-dl staining (Fig. 3A,B). The nuclei appear to exclude the protein (insets of Fig. 3A,B). The cytoplasmic staining of the energids is even along the DV axis, albeit often stronger in the posterior half of the embryo. When they reach the cortex, all nuclei appear to take up Tc-dl protein (Fig. 3C,D). In contrast to Drosophila, where the very first nuclei to reach the cortex take up Dorsal protein in a graded fashion (Roth et al., 1989; Rushlow et al., 1989; Steward, 1989), in Tribolium no asymmetry in nuclear concentrations can be observed at this early stage. However, two cell cycles later, Tc-dl protein preferentially accumulates in nuclei of one half of the embryo (Fig. 3E). A shallow nuclear concentration gradient forms that spans about 40% of the embryonic circumference (Fig. 3E). Unlike Drosophila, the shape and coverings of Tribolium eggs do not allow an unequivocal determination of DV polarity, so it is not possible at this stage to identify the DV position at which nuclear accumulation begins. However, since the gradient persists and later overlaps with the mesodermal marker gene twist (see below), we infer that nuclear accumulation, as in Drosophila, is initiated at the ventral side. During the two subsequent syncytial blastoderm cell cycles this gradient refines by becoming both steeper and more restricted with regard to the embryonic circumference (Figs 3F,G, 4A-D). Refinement is neither produced by nuclear migration nor linked to the division cycles. Fig. 4A-D show transverse sections through embryos of the same nuclear cycle since they both have 69 evenly spaced nuclei around the circumference. The gradient of the upper section spans 23 nuclei (33% of the circumference) with 13 nuclei showing high concentrations flanked by 5 nuclei with intermediate and low levels. In the lower section, presumably from a later embryo of the same cell cycle, the gradient spreads over less than half the number of nuclei (11 corresponding to 16% of the circumference) and is steeper, decreasing from highest to lowest levels over 3 nuclei. Thus, the nuclear Dorsal gradient is more dynamic in Tribolium as compared to Drosophila where its lateral expansion does not change during blastoderm stages (Roth et al., 1989). Tc-dl disappears form the germ rudiment, but becomes highly expressed in the serosa Later during development, the area that is occupied by the gradient shrinks to a narrow five-cell-wide stripe, which harbours only a few scattered nuclei with high levels of Tc-dl (Figs 3H and 4E,F). This Tc-dl distribution corresponds to the stage when the prospective serosa cells first become different from the cells of the germ rudiment giving rise to the embryo proper (Handel et al., 2000). The cell density decreases in the prospective serosal region since the serosa cells stop dividing, start to flatten and expand. In contrast, the cells of the germ rudiment continue to divide and might also be pushed together through the flattening of the serosa cells so that here cell density increases. Since the refinement of the Tc-dl gradient is complete before these changes are obvious, we believe that they do not significantly contribute to the narrowing of the Tcdl domain. While Tc-dl protein expression vanishes from the germ rudiment, it is upregulated in the prospective serosa. The protein distribution thus parallels that of the mrna and forms a clearly visible anterior cap of protein expression (arrow in Fig. 3H). In this anterior cap the protein is mainly cytoplasmic, with the exception of some ventral nuclei located at the border Fig. 4. The refinement of the nuclear Tc-dl gradient does not depend on cell movements. Cross sections through blastoderm embryos stained with anti-tc-dl antibodies showing successive stages of gradient refinement. All sections are derived from about 65% egg length (0% corresponds to the posterior pole). (A,C,E) Complete sections. (B,D,F) Magnification of ventral portion of sections shown in A, C and E, respectively. The embryos in A and C have a uniform distribution of nuclei around the circumference, but the embryo in E shows cell flattening dorsally.

6 5150 G. Chen, K. Handel and S. Roth differentiating serosa cells. Interestingly, before serosal closure is complete, a stripe of serosa cells facing the ventral side of the egg shows nuclear uptake, while the remaining lateral and dorsal serosa cells have strong cytoplasmic staining (Fig. 5C- F). Thus, the asymmetric nuclear uptake, which previously had taken place in the embryo, is now restricted to extraembryonic cells facing the perivitelline space. During later stages of development a few embryos show strong nuclear accumulation of Tc-dl in circular patches of the serosa (Fig. 5G). Since these patches do not always occur at the same location and since they do not have regularly shaped cells in their centre they may represent scars resulting from injuries of the serosa. Tc-dl nuclear uptake might be a consequence of pathogen stimulation suggesting that high levels of Tc-dl in the serosa fulfil a protective function. Fig. 5. Tc-dl protein during later development. (A-D) Ventral surface views of whole embryos doubly stained with anti-tc-dl antibodies and a DNA dye (YOYO-1). (A,C) Bright-field images. (B,D) Fluorescence images of the embryos shown in A and C, respectively. (A,B) Primitive pit formation (arrows). Note the absence of nuclear Tc-dl in the region of the presumptive ventral furrow. (C,D) Early germband extension shortly before the serosal window (arrows) closes. Note the presence of nuclear Tc-dl in a ventral stripe of serosa cells. (E) Dorsal portion and (F) ventral portion of the same cross section through an embryo slightly older than that shown in C, stained with anti-tc-dl antibodies. s, serosa; a, amnion; vf, ventral furrow. At the position at which the section is taken the serosa has closed ventrally. Note that the serosa in this particular embryo is ruptured laterally so that only a small ventral fragment remains. Small amounts of nuclear Tc-dl are still seen in the amnion while Tc-dl is completely absent from the embryonic tissues. High levels of Tc-dl are present in the serosa. The serosal staining is cytoplasmic at the dorsal side (E) and nuclear at the ventral side (F). (G) Accumulation of nuclear Tc-dl in serosal cells surrounding a region that might consist of necrotic tissue (arrowhead). Out of 50 embryos, 4 exhibited such a patch, which in each case was located at a different position in the serosa. between germ rudiment and serosa, which maintain high levels of nuclear Tc-dl (Fig. 3H). By the time the primitive pit forms, the nuclear gradient has disappeared completely from the germ rudiment, and only weak cytoplasmic staining remains in the cells of the future ventral furrow and the primitive pit (Figs 5A,B, 8D,E). During furrow formation, cytoplasmic Tc-dl also vanishes from the germ rudiment (Fig. 5F) except in the region of the primitive pit (data not shown). We have not observed Tcdl nuclear staining or high levels of cytoplasmic staining in any part of the embryo during later stages of embryogenesis. In particular, no nuclear accumulation was found in the cells of the growth zone where together with segment formation DV patterning also has to continue (data not shown). However, high Tc-dl protein levels persist in the Tc-Toll as a potential target gene of Tc-dorsal The restriction of Tc-dl nuclear import to ventral regions suggests the existence of a ventrally localized signal that, as in Drosophila, activates a Toll-like plasma membrane receptor. The Tribolium Toll homolog however, differs from Drosophila Toll by being transcribed only zygotically (Maxton- Küchenmeister et al., 1999). We therefore were interested to compare in detail Tc-Toll expression with gradient formation. While during preblastoderm stages no Tc-Toll expression can be detected (Maxton-Küchenmeister et al., 1999 and data not shown), weak expression becomes visible in early syncytial blastoderm stages when the nuclei reach the cortex. This expression is uniform along the DV axis and parallels the weak nuclear accumulation of Tc-dl (Fig. 6A, data not shown). When the nuclear gradient of Tc-dl forms, Tc-Toll is up-regulated ventrally and becomes strongly expressed in regions of high nuclear Tc-dl concentrations (Fig. 6B). Low Tc-Toll levels still remain at the dorsal side (Fig. 6B). During later blastoderm stages Tc-Toll expression follows a more complex pattern partially independent of Tc-dl (data not shown). However, in the posterior half of the germ rudiment, Tc-Toll expression vanishes in parallel with the refinement and disappearance of the Tc-dl gradient (Fig. 6C). These data are consistent with the assumption that at least during early blastoderm stages a positive feedback loop is established. Ventral Toll receptor activation, through an extraembryonic signal, might enhance the nuclear import of Tc-dl, which in turn would lead to the activation of Toll transcription and thus to the local production of higher amounts of Toll receptor. The relationship of the Tc-dl gradient to the expression of zygotic DV patterning genes In Drosophila three types of Dorsal target genes can be distinguished (Rusch and Levine, 1996): type I targets are turned on by high levels of nuclear Dorsal and are therefore expressed in a ventral stripe (e.g. twist (twi) and snail (sna)); type II targets are activated by low levels of Dorsal being expressed in lateral regions (e.g. rhomboid (rho), short gastrulation (sog), brinker (brk); Jazwinska et al., 1999); type III targets that are uniformly activated by unknown mechanisms are subsequently repressed by Dorsal so that their expression is restricted to the dorsal side of the embryo (e.g. zerknüllt (zen) and decapentaplegic (dpp); Rushlow and Roth, 1996). Most Dorsal targets also receive inputs from the terminal (torso) system, which modulate their expression

7 The Tribolium Dorsal gradient 5151 Fig. 6. The nuclear Tc-dl gradient and its relationship to Tc-Toll expression. All embryos were stained for Tc-dl protein (brown) and Tc-Toll mrna (blue). (A-C) Complete cross sections with magnified portions of the cortex below. (A) Cycle 8 embryo. Arrows indicate the position of the nuclei. Tc-dl protein levels are too low to be detectable in cross sections. Low levels of Tc-Toll mrna are found uniformly around the circumference. (B) Early blastoderm embryo. (C) Blastoderm embryo shortly before serosal segregation. pattern at the anterior and posterior extremities of the embryo (Ray et al., 1991). Homologs of type I and III targets have been cloned from Tribolium and their expression patterns have been shown to differ from their Drosophila counterpart in accordance with the short-germ type of Tribolium embryogenesis (Sommer and Tautz, 1994; Falciani et al., 1996; Sanchez-Salazar et al., 1996). Since the Dorsal gradient of Tribolium resembles that of Drosophila with regard to its position relative to the egg axis, we wondered when and how the differences in the expression patterns of potential target genes arise. As representative of the type I target genes we studied the expression of Tc-twi in relation to gradient formation (Sommer and Tautz, 1994; Handel et al., unpublished data). Tc-twi begins to be weakly expressed along the entire region in which high levels of nuclear Tc-dl can be detected (Fig. 7A). Before a difference between serosa cells and germ rudiment becomes apparent, Tc-twi is repressed in the anterior 20% of the AP axis, and the remaining domain becomes asymmetric along the AP axis by increasing its width at the posterior pole (Fig. 7B). These changes are not preceded by corresponding anteroposterior asymmetries in nuclear Tc-dl distribution. Rather, Tc-dl disappears from the germ rudiment as Tc-twi levels rise (Fig. 7C). This is in marked contrast to Drosophila where twi expression and high nuclear Dm-dl concentrations remain tightly coupled during the cellular blastoderm and early gastrulation. In Tribolium, twi expression might only be initiated by Tc-dl, while its subsequent regulation becomes independent from Tc-dl. In Drosophila, the type III targets, zen and dpp, have early expression patterns composed of a broad dorsal domain and symmetric terminal caps, which depend on activation by the terminal system (Ray et al., 1991). Terminal cap activation and ventral repression occur concomitantly. In Tribolium however, the expression of both genes is first initiated in a broad anterior cap (Falciani et al., 1996; Sanchez-Salazar et al., 1996) which shows no obvious DV asymmetry, even though the nuclear Tcdl gradient has already been established at this stage (Figs 7D, 8A). For Tc-dpp a small posterior cap expression is also visible (Fig. 8A). This suggests that Tc-zen and Tc-dpp are first exclusively under the control of an AP patterning system, presumably the terminal system (Schröder et al., 2000). Subsequently, the Tc-zen expression domain expands to the dorsal side of the embryo (Fig. 7E). At this stage, the lateral expansion of Tc-dl has already decreased so that a big gap occurs between lowest detectable nuclear Tc-dl concentrations and the sharp border of Tc-zen expression at the dorsal side. Thus, the shift of the Tc-zen domain to the dorsal side might not be a direct consequence of repression by Tc-dl. The most striking deviation from Drosophila was found with regard to dpp. After its initial anterior expression, Tc-dpp becomes predominantly expressed in an anterior ventral domain where high nuclear Tc-dl concentrations are present (Fig. 8B,C). This domain resolves into a stripe separating the anterior cap expression of Tc-dl (presumptive serosa) from the region where Tc-dl disappears (germ rudiment) (Fig. 8D,E). When ventral furrow formation starts, Tc-dpp is weakly expressed as a one cell-wide stripe demarcating the serosa (Fig. 8F). The transition from the earliest anterior-cap like to Fig. 7. The nuclear Tc-dl gradient and its relationship to Tc-twi and Tc-zen expression. (A-C) Tc-dl protein (brown) and Tc-twi RNA (blue) distribution in progressively older embryos. (D-F) Tc-dl protein and Tc-zen RNA distribution in progressively older embryos. For all embryos anterior is to the left. (A,B) Ventral surface views. (C,D,F) Optical midsections. (E) Lateral view of slightly tilted embryo. (A,D) Early blastoderm. (B,E) Late blastoderm. (C,F) Primitive pit (pp) formation. The arrows point to the posterior margin of the anterior cap expression of Tc-dl.

8 5152 G. Chen, K. Handel and S. Roth Fig. 8. The nuclear Tc-dl gradient and its relationship to Tc-dpp expression. All embryos show both Tc-dl protein (brown) and Tc-dpp mrna (blue) distributions. The embryos are arranged according to increasing developmental age. Anterior is to the left. (A,C,E) Optical midsections. (B,F) Lateral surface views. The embryo in B is slightly tilted to show the ventral side. (D) Ventral surface view. The arrow marks the border between germ rudiment and presumptive serosa. (A-C) Blastoderm embryos. (D) Serosal segregation. (E) Primitive pit formation. (F) Early gastrulation. s, serosa; pp, primitive pit. the subsequent anteroventral expression domain of Tc-dpp, indicates that Tc-dl does not repress, but rather activates Tcdpp in a direct or indirect manner. Furthermore, since Tc-dpp remains confined to the presumptive serosa region, Tc-dl only modulates the regulatory input from the AP patterning system. In summary, despite similarities in the geometry of the morphogen gradients, Dm-dl and Tc-dl show clear differences in the relationship to their target genes. First, Tc-twi is likely to be a type I target whose initiation, but not its high level expression in the presumptive mesodermal region depends on Tc-dl. Second, neither Tc-zen nor Tc-dpp represent good candidates for type III targets that are directly repressed by Tcdl with Tc-dpp even being a likely candidate for direct or indirect activation. Rescue of the Drosophila dorsal mutant phenotype by Tc dl Given the more dynamic gradient formation and the altered relation to its target genes, it is of interest to determine whether Tc-dl can rescue the Drosophila dl mutant phenotype. In a first attempt, mrna encoding Tc-dl was injected into dl mutant embryos. The injected embryos gastrulated with normal polarity and revealed a rescued DV pattern at the site of injection (Fig. 9E,F). Muscle movement indicates that such embryos formed some mesoderm (data not shown). To more carefully test the functional similarity between Tc-dl and Dmdl, constructs for P-element mediated transformation were generated. Partial rescue of dl mutant embryos was obtained when a Tc-dl transgene in UASp (Rørth, 1998) was driven by maternal α-tub Gal4:VP16. One copy of the transgene leads to differentiation of dorsolateral structures (filzkörper, fk; Fig. 9C); two copies promote, in addition, the formation of ventrolateral structures (ventral epidermis, ve; Fig. 9D). A control construct containing Dm-dl leads to complete rescue or even ventralisation (Fig. 9A,B). We examined the formation of the Tc-dl gradient and the expression of DV patterning genes in dl mutant embryos rescued by two copies of the Tc-dl transgene. Tc-dl protein shows the same early nuclear accumulation at the ventral side of Drosophila embryos as the endogenous Dorsal (Fig. 9G; Roth et al., 1989). The gradient is maintained until cellularisation without the signs of dynamic change seen in Tribolium embryos (Fig. 9H). A combination of antibody staining and in situ hybridisation allows the visualisation of nuclear concentrations of Tc-dl required for repression or activation of target genes. Activation of twi, although expected from the RNA injection experiment, was not observed (Fig. 9G,H). However, in regions of high nuclear Tc-dl accumulation, zen is repressed and sog is activated (Fig. 9I,K). Only low levels of Dm-dl would be required to achieve equivalent changes in gene expression (Roth et al., 1989; Jazwinska et al., 1999). Together, these findings indicate that Tc-dl binds to Drosophila target gene promoters, but that, in the heterologous situation, it acts both as a weaker repressor and a weaker activator than endogenous Dorsal. DISCUSSION Several lines of evidence suggest that Tc-dl is the ortholog of Dm-dl: (1) among all rel/ NF-κB proteins of Drosophila, Dmdl is by far the closest relative of Tc-dl (the RHDs of both proteins share 77% identity); (2) Tc-dl like Dm-dl is maternally expressed and uniformly distributed in the cytoplasm of early embryos; (3) the Tc-dl protein forms a nuclear concentration gradient in Tribolium blastoderm embryos; (4) Tc-dl partially rescues the dl mutant phenotype of Drosophila embryos where it forms a nuclear gradient very similar to the endogenous gradient. The rescue experiment indicates that Tc-dl correctly interacts with the IκB-like Cactus protein of Drosophila (Belvin et al., 1995; Bergmann et al., 1996) and that the Tcdl/Cactus protein complex is subject to Toll induced breakdown at the ventral side. Surprisingly, similar experiments using maternally expressed Dif resulted in an even better restoration of the DV pattern. However, in these experiments the nuclear gradient of Dif could not be visualized and so no data exist about the concentrations of Dif that induced a certain transcriptional response (Stein et al., 1998). The rescue capacity of Dif, despite its divergent RHD, points to the importance of the C-terminal transactivation domain.

9 The Tribolium Dorsal gradient 5153 Fig. 9. Tc-dl partially rescues the Dm-dl mutant phenotype. (A-F) Dark-field photographs of cuticle preparations. (A,B) Control rescue with Dm-dl. The larvae are derived from dl I5 /Df(2L)TW119 females carrying both tubgal4:vp16 and UASp-Dm-dl transgenes. (A) Wild-type-like cuticle. de, dorsal epidermis; ve, ventral epidermis; fk,: filzkörper. The filzkörper are dorsolateral structures. (B) Ventralized cuticle. The ventral epidermis (ve) is expanded at the expense of the dorsal epidermis. This phenotype is identical to that of cactus (Roth et al., 1991) and most likely results from too much Dm-dl produced by the transgene so that endogenous Cactus is unable to retain Dm-dl in the cytoplasm. (C,D) Partial rescue with Tc-dl. The larvae are derived from dl T /Df(2L)TW119 females carrying tubgal4:vp16 and either one (C) or two (D) copies of the UASp-Tc-dl transgene. (C) Rescue of only dorsolateral structures (fk). (D) Rescue of dorsolateral and ventrolateral structures (fk and ve). (E,F) Rescue by mrna injection. (E) Uninjected control embryo forms only dorsal epidermis (de). (F) Injected embryo with extensive rescue of ventral epidermis (ve). (G-K) Cross sections through blastoderm embryos having the same genotype as the embryo shown in D. All embryos were stained with anti-tc-dl antibodies (brown). (G,H) twi in situ hybridisation; no twi expression was detected. (I) Tcdl protein and zen mrna (blue) distribution. (K) Tc-dl protein and sog mrna (blue) distribution. The arrows demarcate regions in which nuclear Tc-dl protein can still be detected although zen is not repressed and sog is not activated. Corresponding nuclear levels of Dm-dl would lead to zen repression and sog activation. Maybe some features of the transactivation domains of Dm-dl and Dif, like the polyglutamine and polyasparagine stretches absent in Tc-dl (Fig. 1B), are species-specific adaptations for efficient interaction with the endogenous basal transcription machinery. This might explain why Tc-dl is a weaker activator and repressor compared to Dm-dl and probably even to Dif. These aspects of protein evolution, however, are less relevant to the main focus of this study concerned with similarities and differences in gradient formation and target gene expression between Drosophila and Tribolium, which raise several interesting questions regarding the evolution of DV patterning mechanisms in insects. Dynamics of gradient formation and the regulatory behaviour of short-germ embryos In contrast to Drosophila, the Tc-dl gradient undergoes a process of dynamic refinement, which leads from weak uniform nuclear up-take via a broad flat gradient to a narrow zone harbouring scattered nuclei with high Tc-dl concentrations (Figs 3 and 4). This process occurs with remarkable precision along the entire AP axis. Therefore, its induction is likely to depend, as in Drosophila, on extraembryonic cues. Since a Toll homolog of Tribolium is expressed at the right stage to mediate gradient formation (Maxton-Küchenmeister et al., 1999) it is highly suggestive that a Spätzle-like extracellular Toll ligand is produced, as in Drosophila, at the ventral side of the egg. However, since Toll expression is itself influenced by gradient formation (Fig. 5) the distribution of the extraembryonic ligand might determine the final shape of the gradient in a less rigid way than in Drosophila (Roth, 1993). A less rigid control by extraembryonic cues might help to understand the high degree of regulation along the DV axis which has been observed with many insect embryos. For example longitudinal fragmentations of preblastoderm embryos of the leaf hopper Euscelis lead to the formation of segmented germ bands in each egg fragment. This holds true even after separation of ventral and the dorsal egg halves (Sander, 1971). Cold treatment of preblastoderm embryos of the Chrysomelide beetle Atrachya can result in up to four complete germ bands forming in one egg (Miya and Kobayashi, 1974). In these cases, DV patterning seems to be initiated from different positions along the embryonic circumference including even the dorsal side of the egg. To

10 5154 G. Chen, K. Handel and S. Roth explain such results a self-organising pattern-forming system is required, which links local self-enhancement with a lateral inhibition process (Meinhardt, 1982, 1989). In Tribolium, Toll and dorsal could be part of a self-enhancing feedback loop. Small levels of Toll-receptor activation would result in Dorsal nuclear import and in up-regulation of Toll transcription. This in turn would locally increase the receptor density and cause sequestration of the activating ligand. If, as in Drosophila, the ligand is diffusible in the perivitelline space (Stein et al., 1991), this loop can in principle be initiated and maintained at any region of the egg. The molecular nature of the lateral inhibition process that needs to accompany the positive feedback loop to prevent uncontrolled spreading, is not known. Two principal ways of inhibition can be envisaged: either the production of a diffusible inhibitor is linked to local self-enhancement or the self-enhancement process depletes its own substrate (Meinhardt, 1982). In Drosophila remnants of an inhibitory process have been found. In certain mutant backgrounds partial duplication of the DV pattern can be observed (Roth and Schüpbach, 1994). Experiments suggest that Easter, the protease which activates Spätzle, is subject to pathwaydependent inhibition (Misra et al., 1998). In Tribolium upregulation of the Toll receptor might lead to the depletion of the ligand in the surroundings and thereby suppress Toll activation in the vicinity of an activation centre. Thus, given the right dynamic parameters, receptor up-regulation might contribute to both the positive feedback and the spatial restriction required for pattern formation. Tc-dl plays a less direct role in establishing dorsoventral cell fates compared to Dm-dl Although the regulatory behaviour of DV patterning found in perturbation experiments suggests reduced dependence on a maternal prepattern, it is also possible that interactions between zygotic DV patterning genes play an important role. In this context, it is interesting that the Tc-dl not only differs from the Dm-dl gradient, but also the relationship of the gradients to their respective target genes and hence to the cell fates along the DV axis appears to be different. This is apparent for both the mesoderm and the ectoderm. In Drosophila, a ventral stripe of high nuclear Dorsal in the trunk region of the blastoderm embryo is congruent with the mesodermal anlagen since it defines the lateral expansion of twi expression which, together with sna, promotes ventral furrow formation. Dm-dl remains present in twi- and snaexpressing cells until the mesoderm has invaginated. In Tribolium, in contrast, the early weak Tc-twi expression domain, which is even along the AP axis and coincides with the highest levels of nuclear Tc-dl, is rapidly replaced by a domain with strong AP asymmetry by becoming repressed anteriorly and broadened towards the posterior pole (Fig. 7A,B). When this expression pattern is fully developed, nuclear Tc-dl has disappeared from the germ rudiment (Fig. 7C). However, this final Tc-twi domain corresponds to the presumptive mesoderm since it presages the position and shape of the ventral furrow (Handel et al., unpublished data). This implies that the shape of the gradient does not fully determine the mesodermal anlagen and that Tc-twi transcription becomes independent of activation by Tc-dl at late blastoderm. The connection between Tc-dl and the patterning of the ectoderm is even more indirect. In Drosophila low levels of nuclear Dorsal activate the expression of type II target genes which are required for the formation of neuroectoderm (Rusch and Levine, 1996; Jazwinska et al., 1999). Simultaneously, Dorsal represses the type III target genes of the Dpp group confining their expression to the dorsal side which gives rise to the amnioserosa and to the non-neurogenic ectoderm (Rushlow and Roth, 1996). No fate maps have been constructed for the Tribolium blastoderm embryo so far. However, the serosa cells seem to derive from a broad anterior domain which is slightly tilted to the dorsal side (Handel et al., 2000). The serosa is likely to be homologous to the amnioserosa of Drosophila. zen and dpp are expressed in both tissues. While in Drosophila these are dorsal expression domains, in Tribolium the zen and dpp domains initially have the shapes of anterior caps, which are symmetric with regard to the DV axis (Falciani et al., 1996; Sanchez-Salazar et al., 1996), even though the Tc-dl gradient has already formed (Figs 7D, 8A). Only later Tc-zen shifts to the dorsal side, and Tc-dpp shifts to the ventral side (Figs 7E, 8B). The latter is remarkable since it indicates that a target gene, which is repressed by Dorsal in Drosophila, might be activated by Dorsal in Tribolium. In Drosophila, zen and dpp are also expressed in terminal caps, which are not influenced by Dorsal (Rushlow et al., 1987; Roth et al., 1989). However, no function has so far been attributed to these terminal caps since all patterning functions seem to reside in the dorsal expression domains. It is tempting to suggest that the terminal caps are evolutionary remnants of short-germ development, where the extraembryonic tissues are derived from anterior rather than dorsally located regions of the blastoderm (Schröder et al., 2000). While the region giving rise to the serosa can be located by following the course of embryonic development, this is harder with regard to the subdivisions of the ectoderm. Therefore, it is not yet clear how early Tc-dpp expression (Fig. 8E,F) relates to the later Tc-dpp domain corresponding to the dorsal ectoderm (Sanchez-Salazar et al., 1996). There is some indication that the dorsal ectoderm expression is independently initiated in more posterior regions of the germband and is thus not continuous with the early Tc-dl associated domain (S. Roth unpublished observations). In summary, a comparison of spatial and temporal aspects of Tc-dl gradient formation with respect to both the expression pattern of potential target genes and the likely position of the blastoderm anlagen suggests that Dorsal has a less direct role in cell fate determination in Tribolium than it has in Drosophila. It is quite possible that its major role in the germ anlage is the initiation of Tc-twi expression while the patterning of the embryonic ectoderm is a secondary consequence of mesoderm formation initiated by Tc-twi. The Dorsal gradient as preadaptation for long-germ development It has been argued that both very early and late development are the major targets of evolutionary change while a certain stage characteristic for each phylum, the phylotypic stage, is relatively stable (Slack et al., 1993; for review see Gerhart and Kirschner, 1997). Besides vertebrates, insects provide the best example for the existence of a stable phylotypic stage. Insect embryos look very similar once segmentation is complete

11 The Tribolium Dorsal gradient 5155 despite diverse forms of oogenesis and dramatic differences in the extent to which the body plan is determined before gastrulation (Sander, 1976). Thus, processes close to the phylotypic stage, such as the final aspects of segmentation, are expected to be conserved, while early events such as axis specification under the control of the maternal genome might be divergent (Patel, 1994). However, this view of evolutionary change is based mainly on phenotypic observations since not enough knowledge has accumulated about the molecular mechanisms of early development in diverse insect groups. With regard to maternal anteroposterior gradients several lines of evidence suggest that bicoid, the anterior determinant of Drosophila might have arisen late in evolution being unique to Dipterans (Stauber et al., 1999). Nevertheless, a Bicoid-like activity needs to be invoked in Tribolium to explain the formation of the maternal Caudal protein gradient (Wolff et al., 1998). The existence of a Dorsal gradient in the short-germ beetle Tribolium provides a good example of the conservation of an early patterning process despite the fact that later aspects of embryogenesis leading to the phylotypic stage are dramatically different. The Dorsal gradient in Tribolium has the same AP extension with regard to the egg shell as that of Drosophila. The spatial information it provides however, is not exploited to the same extent as in the long-germ Drosophila embryo. Thus, it appears that in early Tribolium embryos a developmental mechanism exists which is fully used only in embryos with long-germ development. To find a long-germ type Dorsal gradient in a beetle embryo might not be so surprising because among different beetle species transitions between long, intermediate and short-germ types of development have been observed (Sander, 1976; Patel et al., 1994). It will however be interesting to see whether Dorsal gradients exist in more primitive short-germ embryos of hemimetabolous insects and if so, how they are positioned within the egg. Innate immunity and the origins of the Dorsal gradient The immune function of the Toll-rel/NF-κB pathway is evolutionarily very old since it is conserved between insects and vertebrates (Hoffmann et al., 1999; Anderson, 2000). Although recent data from Xenopus suggest an involvement of the Toll pathway in DV axis formation in amphibians, the problem of the ancestral role has still not been resolved (Armstrong et al., 1998). If the ancestral function was in immunity, our finding that high amounts of Dorsal are expressed in the serosa of Tribolium might provide a hint as to how the system originally devoted to pathogen defence was coopted for patterning. In primitive hemimetabolous insects most of the blastoderm cells give rise to the serosa (Sander, 1976). Since the serosa has a protective function for the whole egg, often evident in the formation of serosal cutical (Machida and Audo, 1998), there might have been an evolutionary advantage to supply serosa cells with pathogen defense mechanisms using the Toll pathway. Furthermore, the distinction between serosa and germ rudiment is the earliest cell differentiation event in an insect embryo, taking place at blastoderm stage before gastrulation starts so that serosa specific components have to be provided very early. Thus, only a small shift in temporal and spatial expression would have been necessary for a shift in function of the Toll pathway from innate immunity to axis formation. When, in evolution, might this functional shift have occurred? Other arthropods have extraembyonic ectoderm, but, to our knowledge, the distinction between an amnion covering the embryo ventrally and a serosa covering the entire egg (embryo plus yolk), is unique to ectognathan insects (Anderson, 1973; Machida and Audo, 1998). Thus, the pathway might have been co-opted for axis formation during early insect evolution. To test this hypothesis, it will be crucial to isolate pathway components from more primitive insects and other arthropods. We thank Anke Beermann, Reinhard Schröder and Martin Klingler for introducing us into the secrets of beetle rearing; John Doctor and Diethard Tautz for molecular probes; Reinhard Schröder for the cdna library; Diethard Tautz for help with the phylogram; Xiang Fan and Oliver Karst for excellent technical assistance; Christian Bökel for teaching us protein techniques; Manfred Schorpp for help with the rabbits; Klaus Sander, Reinhard Schröder, Nicola Berns, Xiang Fan, Maithreyi Narasimha, Thomas Seher and Henry Roehl for valuable comments on the manuscript. We are grateful to Uli Schwarz (MPI für Entwicklungsbiologie, Tübingen) and Wolf Engels (University of Tübingen) for their help throughout the course of this study. Teresa Nicolson kindly provided space for G. C. and K. H. in her laboratory during the last year of the experimental work. The DAAD provided a stipend to G. C. The work was supported by grants from the DFG (SPP 1027). REFERENCES Anderson, D. T. (1973). Embryology and Phylogeny in Annelids and Arthropods. Oxford: Pergamon Press. Anderson, K. V. (2000). Toll signaling pathways in the innate immune response. Curr. Opin. Immunol. 12, Armstrong, N. J., Steinbeisser, H., Prothmann, C., DeLotto, R. and Rupp, R. A. W. (1998). Conserved Spätzle/Toll signaling in dorsoventral patterning of Xenopus embryos. Mech. Dev. 71, Barillas-Mury, C., Charlesworth, A., Gross, I., Richman, A., Hoffmann, J. A. and Kafatos, F. C. (1996). Immune factor Gambif1, a new rel family member from the human malaria vector, Anopheles gambiae. EMBO J. 15, Beermann, A. (1998). Genetische und molekulare Analyse des Distal-less Gens bei Tribolium castaneum. (Dissertation, Tübingen) Belvin, M. P., Jin, Y. and Anderson, K. V. (1995). Cactus protein degradation mediates Drosophila dorsal-ventral signaling. Genes Dev. 9, Berghammer, A., Bucher, G., Maderspacher, F. and Klingler, M. (1999). A system to efficiently maintain embryonic lethal mutations in the flour beetle Tribolium castaneum. Dev. Genes Evol. 209, Bergmann, A., Stein, D., Geisler, R., Hagenmaier, S., Schmid, B., Fernandez, N., Schnell, B. and Nüsslein-Volhard, C. ( 1996). A gradient of cytoplasmic Cactus degradation establishes the nuclear localization gradient of the dorsal morphogen in Drosophila. Mech. Dev. 60, Dushay, M. S., Asling, B. and Hultmark, D. (1996). Origins of immunity: Relish, a compound Rel-like gene in the antibacterial defense of Drosophila. Proc. Natl. Acad. Sci. USA 93, Falciani, F., Hausdorf, B., Schröder, R., Akam, M., Tautz, D., Denell, R. and Brown, S. (1996). Class 3 Hox genes in insects and the origin of zen. Proc. Natl. Acad. Sci. USA 93, Gerhart, J. and Kirschner, M. (1997). Cells, Embryos and Evolution. Malden: Blackwell Science. Grünfelder, C. G. (1997). Vom frisch abgelegten Ei zum Blastoderm: Untersuchungen zur Feinstruktur der frühen Embryogenese des Reismehlkäfers Tribolium confusum, Duval (Coleoptera, Tenebrionidae). Doctoral thesis, University of Freiburg. Handel, K., Grünfelder, C., Roth, S. and Sander, K. (2000). Tribolium embryogenesis: a SEM study of cell shapes and movements from blastoderm to serosal closure. Dev. Genes Evol. 210,