cactus, a maternal gene required for proper formation of the dorsoventral morphogen gradient in Drosophila embryos

Development 11, 371388 (1991) Printed in Great Britain The Company of Biologists Limited 1991 371 cactus, a maternal gene required for proper formation of the dorsoventral morphogen gradient in Drosophila embryos SIEGFRIED ROTH 1 *, YASUSHI HIROMI *, DOROTHEA GODT 3 and CHRISTIANE NUSSLEIN VOLHARD 1 1 MaxPlanckInstitut filr Entwicklungsbiologie, Spemannstrasse 35/111, 7400 Tubingen, FRG Department of Molecular and Cell Biology, University of California, Berkeley, CA 9470, USA 3 Institut filr Entwicklungsphysiologic, Universitat Kdln, Gyrhofstrasse 17, 5000 Kb'ln 41, FRG Present address: Department of Molecular Biology, University of Princeton, Princeton, NJ 08544, USA Summary The dorsoventral pattern of the Drosophila embryo is mediated by a gradient of nuclear localization of the dorsal protein which acts as a morphogen. Establishment of the nuclear concentration gradient of dorsal protein requires the activities of the 10 maternal 'dorsal group' genes whose function results in the positive regulation of the nuclear uptake of the dorsal protein. Here we show that in contrast to the dorsal group genes, the maternal gene cactus acts as a negative regulator of the nuclear localization of the dorsal protein. While loss of function mutations of any of the dorsal group genes lead to dorsalized embryos, loss of cactus function results in a ventralization of the body pattern. Progressive loss of maternal cactus activity causes progressive loss of dorsal pattern elements accompanied by the expansion of ventrolateral and ventral anlagen. However, embryos still retain dorsoventral polarity, even if derived from germline clones using the strongest available, zygotic lethal cactus alleles. In contrast to the lossoffunction alleles, gainoffunction alleles of cactus cause a dorsalization of the embryonic pattern. Genetic studies indicate that they are not overproduces of normal activity, but rather synthesize products with altered function. Epistatic relationships of cactus with dorsal group genes were investigated by double mutant analysis. The dorsalized phenotype of the dorsal mutation is unchanged upon loss of cactus activity. This result implies that cactus acts via dorsal and has no independent morphogen function. In all other dorsal group mutant backgrounds, reduction of cactus function leads to embryos that express ventrolateral pattern elements and have increased nuclear uptake of the dorsal protein at all positions along the dorsoventral axis. Thus, the cactus gene product can prevent nuclear transport of dorsal protein in the absence of function of the dorsal group genes. Genetic and cytoplasmic transplantation studies suggest that the cactus product is evenly distributed along the dorsoventral axis. Thus the inhibitory function that cactus product exerts on the nuclear transport of the dorsal protein appears to be antagonized on the ventral side. We discuss models of how the action of the dorsal group genes might counteract the cactus function ventrally. Key words: developmental genetics, dorsoventral pattern, dorsal nuclear gradient, inhibition of nuclear transport, maternaleffect gene, Drosophila, cactus. Introduction In Drosophila, polarity and primary subdivisions of the embryonic body axes are dependent on the activities of maternaleffect genes (NussleinVolhard et al. 1987; Anderson, 1987; NussleinVolhard and Roth, 1989). One integrated system of twelve components, encoded by the eleven dorsal group genes and cactus, is required to establish the dorsoventral pattern of the embryo (Anderson and NussleinVolhard, 1986; Schiipbach and Wieschaus, 1989). Females lacking the activity of any of the dorsal group loci produce completely dorsalized embryos, whose cells behave and differentiate like the dorsal cells of wildtype embryos. For many of the dorsal group genes, hypomorphic alleles have been isolated which cause only a dorsalization (Anderson and NussleinVolhard, 1986). The phenotypes produced by such alleles can be arranged in a hypomorphic series, which covers a continuous spectrum of pattern alterations ranging from wild type to complete dorsalization. The coordinated and continuous fate map shifts underlying these phenotypes suggested that the positional information of the dorsoventral pattern is graded and that the action of

37 S. Roth, Y. Hiromi, D. Godt and C. NilssleinVolhard the dorsal gtoup genes culminates in the establishment of a morphogen gradient (NiissleinVolhard, 1979a,b; Anderson et al. 1985). Classical as well as molecular experiments led to a functional dissection of the group of eleven genes with identical phenotypes. The double mutant combinations between dominant ventralizing Toll alleles and dorsalizing mutants demonstrated that Toll acts downstream of gastrulation defective, nudel, pipe, snake and easter, but upstream of dorsal (Anderson et al. 1985). Thus, the product encoded by the dorsal gene functions at the end of the developmental pathway, dorsal is distinct from all other dorsal group genes in two further respects. First, it is dosagesensitive. At high temperature, embryos produced by females lacking one dorsal copy are weakly dorsalized (NiissleinVolhard, 1979a; Niisslein Volhard et al. 1980). The dosagesensitivity indicating a concentrationdependent action might be characteristic of morphogens as it was also observed for bicoid, which encodes the anterior morphogen (Frohnhofer and NiissleinVolhard, 1986). Second, cytoplasmic transplantation experiments revealed a ventral enrichment of rescuing activity, indicating an asymmetric distribution of active dorsal product (Santamaria and NussleinVolhard, 1983; NiissleinVolhard and Roth, 1989). Together, these observations suggested that dorsal encodes the morphogen of the dorsoventral axis. In agreement with this assumption, dorsal protein was found to form a nuclear concentration gradient along the dorsoventral axis with a maximum at the ventral side (Steward et al. 1988; Steward, 1989; Rushlow et al. 1989; Roth et al. 1989). The analysis of dorsal protein distributions in mutant embryos shows that different concentration ranges of nuclear dorsal protein correspond to the main subdivisions of the dorsoventral pattern: the mesoderm; the ventral neuroectoderm, which gives rise to the central nervous system and the ventral epidermis; the dorsolateral ectoderm; and a dorsal region, from which dorsal epidermis and amnioserosa are derived. The cell fates of these regions are determined by the corresponding nuclear dorsal protein concentrations in a largely autonomous manner (Roth et al. 1989). dorsal protein probably acts as a transcriptional regulator. It has sequence similarities with the nuclear protooncogene crel and with the transcription factor NFJCB/KBFI (Steward, 1987; Kieran etal. 1990; Ghosh et al. 1990). A comparison of the dorsal protein distribution with the expression pattern of zygotic dorsoventral genes suggests that it exerts both activating and repressing functions. The highest nuclear dorsal protein levels are required to initiate directly or indirectly the transcription of the zygotic genes twist and snail, both needed to specify the mesoderm (Thisse et al. 1987; Leptin and Grunewald, 1990). Low nuclear levels cause a direct or indirect repression of genes, like zen or dpp, required for the development of dorsal epidermis and amnioserosa (Rushlow et al. 1987; Irish and Gelbart, 1987; St. Johnston and Gelbart, 1987; Roth et al. 1989). No overall asymmetry in the concentration of dorsal protein is observed along the dorsoventral axis of syncytial blastoderm embryos. Therefore, the formation of the nuclear gradient must occur via a spatial regulation of the nuclear localization of dorsal protein (Steward, 1989; Rushlow et al. 1989; Roth etal. 1989). Mutations in any of the 10 dorsal group genes lead to a complete cytoplasmic dorsal protein localization, indicating that they are involved in the nuclear uptake of dorsal protein at ventral positions. The products of the dorsal group genes form a signal transduction pathway. Toll encodes a transmembrane protein that is evenly distributed in the plasma membrane of syncytial blastoderm embryos (Hashimoto et al. 1988; C. Hashimoto, personal communication). Its ventral activation requires the activities of two serine proteaselike molecules encoded by snake and easter (DeLotto and Spierer, 1986; Chasan and Anderson, 1989), which are probably secreted into the perivitelline space. The activated Toll protein stimulates the spatially restricted nuclear localization of the dorsal protein. While 10 genes are involved in the positive regulation of the nuclear localization of dorsal protein, only one gene, cactus, has been identified that has opposing effects. Lossoffunction mutations at the cactus locus result in a ventralization of the embryonic pattern (Schiipbach and Wieschaus, 1989), accompanied by an extension of the nuclear dorsal protein gradient towards the dorsal side (Steward, 1989; Roth etal. 1989). Here, we describe the effects of cactus mutations which include both ventralizing lossoffunction and dorsalizing gainoffunction alleles. The dosage sensitivity of cactus and the continuity of phenotypic alterations produced by hypomorphic alleles demonstrate that the concentration of active cactus product is crucial for the determination of the dorsoventral anlagen. However, using double mutant combinations, we show that the concentrationdependent action of cactus occurs entirely via dorsal. Based on genetic studies and cytoplasmic transplantation experiments, we propose that the cactus product is evenly distributed along the dorsoventral axis. The dorsalizing function of cactus is normally suppressed in the ventral half of the egg circumference (EC) by the action of dorsal group genes. This process is probably blocked in dorsalizing gainoffunction cactus alleles. The dorsal protein distributions exhibited by mutant and doubly mutant embryos suggest that cactus encodes an inhibitor of the nuclear concentration of dorsal protein. We propose a direct interaction of cactus and dorsal proteins. Materials and methods Fly strains The wildtype stock was Oregon R. The references for the cactus alleles used in this study are given in Table. The cactus deficiencies Df(L)E10RN and Df(L)TE116GW1 are described in Ashburner etal. 1990. dl' (NussleinVolhard, 1979a). In(L)dl T (Steward and NiissleinVolhard, 1986). The alleles of dorsal group genes spz m7, plf 8, TF 9 (designated as mel(3)7, mel(3)8, mel(3)9 by T. Rice. 1973); Tl SBRE, T,9QRE 7 r O Df(3L)ro*63 t nd^6 t nd fa, ^ (former

Developmental genetics of cactus 373 designation ea" R6s \ ea, tub" 8, tub 38, snk 073, snk 9, pip 386, pip, pll 078, spz, spz 67 are described in Anderson and NussleinVolhard (1984, 1986), Anderson et al. (1985). ea 5 " (Chasan and Anderson, 1989). wind RP (Schiipbach and Wieschaus, 1989). All mutant chromosomes carried visible markers allowing genotypic identification. Flies were grown and eggs collected under standard conditions (NussleinVolhard et al. 1984). Staging of embryos was according to CamposOrtega and Hartenstein (1985). Double mutant analysis The choice of the cactus allele is crucial for the investigation of the epistasis relations of cactus and the dorsal group genes. If weak (V4) or medium (V3) alleles are used, all double mutants with completely dorsalizing alleles of dorsal group genes result in completely dorsalized embryos. Although these cactus alleles cause a deletion of dorsal and dorsolateral pattern elements on their own they do not, if simultaneously a dorsal group gene is removed. Therefore, it is necessary to perform the double mutant analysis with strong cactus alleles (V). Even in this case the phenotype exhibited by the double mutant combination is very sensitive to the alleles or allelic combinations used, as demonstrated in Fig. 8. To get consistent results, we constructed all double mutants with the strongest viable allele, cact 4 : cact^/cact* ; ndf 93 / ndf 46 cac^/cact* ; pip 664 '/pip 386 C4ct A wind RP /cact A wind RP cart* 1 cad* ; spz m7 /spz 197 gd /gd ; cac^/cac^ cacfi/aict* ; snk<" 3 / snk 9 cac^/cact* ; ea J /ea cad* 1 cad* ; T? QRE / Df(3R)w XB3 cac^/cact* ; plf 78 [pir 8 cac^/cact* ; tub^/tub" 8 cacf^dl'/cact^dl' Xray mutagenesis for reversion of cact E1 cacfi 10 was isolated by excision of the Pelement from a Pinduced lossoffunction allele cact" 5 (Table 1). cact 810 / CyO males were Xray irradiated and mated to In(L)dl T b pr en sea/cyo females. Approximately 1000 cacf 10 '/'In(L)dl T females were screened at 18 C for the production of viable progeny. Putative revertant lines were established from male progeny exhibiting cn Cy phenotype. Temperatureshift experiments To determine the temperaturesensitive periods of cact HE and cacf 8 eggs were collected from cact HE or cact* 18 homozygous females on yeasted agar plates for 3h at 9 C or for 6h at 18 C and shifted to the other temperature at the end of the egg collection period. The eggs were covered with Voltalef 3S oil and visibly staged by selecting gastrulating embryos (stage 6; CamposOrtega and Hartenstein, 1985) at regular intervals after the shift (15min at 9 C; 30min at 18 C). The selected embryos developed at 18 C (downshift) or 9 C (upshift). Cuticle preparations of the differentiated embryos were examined. Cytoplasmic transplantation Cytoplasmic transplantations were performed according to Santamaria and NussleinVolhard, 1983. Pole cell transplantation The protocol for the transplantation of pole cells has been described (Lehmann and NussleinVolhard, 1986). Pole cells were transplanted into the progeny of a cross between Oregon R females and ovo DI v males (Busson et al. 1983). In this experimental design, all female progenies are expected to be sterile unless they have received functional pole cells by transplantation. Antibodies The production of antibodies against twist and dorsal protein is described in Roth et al. 1989. Antibodies against zen protein were obtained from C. Rushlow (Rushlow et al. 1987). Immunological staining of wholemount embryos with biotinylated HRPavidin complexes bound to biotinylated second antibody (Vector Laboratories, Avidin/Biotin ABC system) was carried out as described by Macdonald and Struhl (1986), with the modification that during the washes we added 100 DIM NaCl to the solutions. For sectioning, stained embryos were dehydrated (lomin 70% ethanol, xl0min 100% ethanol, x100% acetone) and mounted in DurcupanACM (Fluka). A complete series of transverse sections (10 /an) was prepared for each embryo to study changes of the staining pattern along the anteroposterior body axes. Cuticle preparations of embryos For the observation of cuticular structures, differentiated embryos with vitelline membrane or dissected out of the vitelline membrane were mounted in a mixture of Hoyer's medium (Van der Meer, 1977) and lactic acid (1:1). Results Cactus has the meiotic position 51,7 (Roth, 1990). and maps to the chromosomal region 35F1.;36A1. (Ashburner et al. 1990). In the course of several mutagenesis experiments, 45 cactus alleles have been isolated (Table ). They include homozygous viable as well as zygotic lethal alleles. In the following, we describe the maternal effects cactus mutations exert on the embryonic dorsoventral pattern. The mutant phenotypes are characterized by the cuticle patterns, the morphogenetic movements during gastrulation, the expression patterns of the zygotic genes zen (Rushlow et al. 1987) and twist (Thisse et al. 1988) and dorsal protein distribution. We apply the phenotypic classification of Anderson et al. 1985 with minor modifications as summarized in Table 1. The haploinsufftcient dominant phenotype Deficiencies of the cactus locus as well as lethal alleles and strong viable lossoffunction alleles cause a dominant maternal effect. Depending on the genetic background, 5090% of the larvae produced by heterozygous females do not hatch. In contrast to the dominant effect of the dorsal gene (NussleinVolhard, 1979a), the dominant phenotype of cactus is not dependent on temperature. 040% of the nonhatching larvae have cuticle phenotypes resembling those of weakly ventralizing zygotic mutations such as zen (Wakimoto et al. 1984) or twisted gastrulation (Zusmann and Wieschaus, 1985). They form all cuticular pattern elements found in wildtype embryos,

374 5. Roth, Y. Hiromi, D. Godt and C. NiissleinVolhard DO Dl D D3 V4 V3 V VI vo LI (apolar) LI (polar) L (apolar) L (polar) L3 Amnioserosa Table 1. Classification of mutant phenotypes Dorsal epidermis Filzkdrper, antennal sense organs / Lateral epidermis, anal plates Ventral epidermis Mesoderm The classification of phenotypes is according to Anderson et al. 1985 with modifications. Deleted structures are designated as ''. '' means that the structure is reduced in size or deleted at some positions along the anteroposterior axis. In most cases, the structures designated as ' ' are extended to compensate for the loss of those structures designated as ' '. Dorsauzed embryos (D) are characterized by the loss of ventral pattern elements and an accompanying expansion of more dorsally derived structures. In the most extreme case (DO), only dorsal epidermis and amnioserosa are expressed along the entire embryonic circumference (EC). Ventralized embryos (V) reveal an expansion of lateral and ventral pattern elements at the expense of more dorsally derived structures. In contrast to V embryos, VI embryos exhibit an expansion of the mesoderm at all positions along the anteroposterior axis, a feature that cannot be shown in the representation used in the table. A complete ventralization (VO) is characterized by the uniform expression of mesoderm. Lateralized embryos (L) are those that have lost both dorsauy and ventrally derived structures. They may be completely apolar at different positional levels (e.g. ventrolateral, dorsolateral) or they exhibit residual polarity. but head involution does not occur and the telson is frequently pulled inside the posterior abdominal region (Fig. 1A). The dorsoventral extent of the zen expression domain is reduced compared to wild type and frequently no expression can be detected anteriorly at 75% EL (egg length; anterior tip=100% EL) and posteriorly at 15 % EL (Table 3a, Fig. 3B). This reveals that the reduction of cactus function causes a loss of dorsal or dorsolateral anlagen. Thus, cactus is a haploinsufficient gene and the normal amount of its product appears to be crucial for the determination of dorsal structures. The lossoffunction alleles A phenotypic continuum ranging from wild type to strong ventralization is produced in embryos derived from females that are homozygous or transheterozygous for different lossoffunction cactus alleles. For simplicity, we divide the ventralized phenotypes into four classes of increasing strength (V4, V3, V and VI; Table 1). The class of weakest phenotypes, designated as V4, also includes the haploinsufficient phenotype of cactus. The class of strongest phenotypes, designated as VI, can be observed only in gerrnline mosaics with lethal alleles (see below). The cuticle phenotype caused by weak alleles shows slightly wider than normal ventral denticle belts. All dorsally and dorsolaterally derived pattern elements are still present (V4, Fig. 1A,B). Stronger mutations result in the replacement of the dorsal epidermis by naked cuticle characteristic of more lateral positions in wildtype embryos. In addition, the dorsolaterally derived filzkorper and the sense organs of the head are lost (V3, Fig. 1C). The strongest viable alleles cause a complete lack of all dorsally and laterally derived structures accompanied by the expansion of the ventral epidermis around the entire circumference (V; Fig. ID). Often the cuticle is poorly differentiated and may be reduced in amount. The alterations of morphogenetic movements in cactus mutant embryos reflect early changes of the dorsoventral anlagen. During gastrulation even weak cactus mutations cause a shift of the cephalic fold, normally a lateral structure, to more dorsal positions indicating an expansion of lateral at the expense of dorsal anlagen (Fig. C,D). Germ band extension can be used as a measure of ventralization. In wild type, the germ band extends up to 65% EL, in weakly ventralized embryos up to 30% EL, and in strongly ventralized embryos no extension occurs. However, ventral furrow formation is normal even in the strongest mutations investigated (Fig. E). Fig. 1. Darkfield photographs of the cuticle produced by cactus mutant embryos. (A) cacfi 6 j cacfi 6, C (=V4). This phenotype results also from Df cact~/. (B) cacr E / cacr*, 18 C (=V4). Note that the embryos shown in A and B have their Fk (arrow) and 8th abdominal segment pulled inside the abdomen. (C) cact PD /cact PD (=V3). (D) cact HE /cact HE, 9 C (=V). (E) cact 013 / Df(L)E10RN (=V1). The embryo is derived from a germ line chimeric female (see Table 4). (F) Wild type. (G) cacf 10 1cact* 010 (=D3). This phenotype results from a loss of mesoderm as shown in Fig. 3L. (H) cact 8 ' 0 / cact "(=D). The embryo has smaller ventral denticle belts as compared to wild type indicating that it not only lacks mesoderm, but also exhibits reduced ventrolateral anlagen. (I) cacf 10 1 caa 013 (=D1). (K) cact 30 /cact 80 (=D0). For phenotypic classification, see Table 1. ASO, antennal sense organs; Fk, filzkorper; VE, ventral epidermis.

Developmental genetics of cactus 375 * w / 'I'

376 S. Roth, Y. Hiromi, D. Godt and C. NussleinVolhard Table. cactus alleles Allele Zygotic Phenotypic designation Mutagen lethality* class Reference d AA58 AB10 H8 Q6 Yll PI P3 P6 P7 PD SG HE 09 Oil H4 P5 A me FU E19 99 G8 SI U7 D1 D13 AA51 AB13 X6 55 E E4 E6 E8 E13 E15 El 7 E0 E5 E9 E10RJ1 E10R01 EP E10 BQ X ray X ray Wildtype Wildtype (ts) c V4 (ts) c V4 (ts) c V4 (ts) c V4 V4 V4 V4 V3 V3 V3 (ts) c V3 V3 V3 V3 V V V V V» V» LD1* LD0 4 4 3 3 3 3 1 1 1 3 0 5 0 4 4 4 5a 7 7 6 5 6 'Zygotic lethality of individuals when homozygous or transheterozygous with a cactus deficiency. b Classification of the phenotypes of embryos denved from homozygous females or from females carrying the allele in trans to cacr* (*) according to the criteria given in Table 1. c Temperaturesensitive allele. d The known cactus alleles derive from eight different mutant screens: (0) NussleinVolhard et al. 1984 (1) Schupbach and Wieschaus, 1989 () U. Mayer, R. Lehmann and C. NUssleinVolhard, unpublished (3) S. Richstein and C. NussleinVolhard, unpublished (4) Ashburner et al. 1990, Roth 1990 (5) Y. Hiromi, unpublished; 5a, original insertional mutant;, dysgenic revertants of cact 55 (6) D. St Johnston, unpublished (7) Xray revertants of the gainoffunction allele cacf 0, see text for further details. c cacfi 10 is zygotic lethal when homozygous, but viable when transheterozygous with lethal cactus alleles, cactus deficiencies or revertants of cac^10. Therefore, cacr 610 exhibits a gainoffunction zygotic lethality. Table 3. zen and twist expression (a) zen expression in embryos with reduced cactus activity % egg length (number of cells around EC) 90(59) 75 (71) 60(85) 45 (91) 30(88) 15 (81) 5(65) zendomain (% EC) Wild type 34% 40% 4% 4% 40% 35% 100% cacr/* (V4) 0% 1%** 5% 8% 8% 1%** 100% (b) twist expression in cactus mutant embryos nvisrdomain (% EC) % egg length (number of cells around EC) 85(55) 70(80) 55(94) 40(100) 5(96) 10 (67) Wild type 40% 9% 6% 5% 6% 3% cacf D l cacf 6 (V3) 48% 8% 4% 4% 9% 61% cact* / cact 013 / cact" DfcacC*** (V) (VL) 71% 8% 9% 30% 3% 86% 100% 3% 30% 30% 49% 100% Wildtype and mutant embryos were stained using antizerc or antinvir( antibodies as described in Materials and Methods. After embedding, the embryos were sectioned to yield complete series of transverse sections each 10/on thick. At distances of 50/mi the total number of cells around the embryonic circumference (EC) and number of zen or twist expressing cells were counted. The position along the anteroposterior axis is indicated as % egg length (0%, posterior end). The results of 710 embryos were averaged. The width of the expression domains (%EC) are calculated by relating the averaged numbers of twist or zen expressing cells to the averaged number of cells around the EC at a given egg length. The values varied by 3% for different embryos. The phenotypic classification (V4V1) is indicated in brackets (see Table 1). * cacfl : cact A /b pr. ** At these positions the zen expression is very weak and sometimes hard to detect in embryos derived from cacf/ females. ***The embryos are derived from germ line chimeric females (see Table 4). The genotype of the germ line was: cact 13 / Df(L)E10RN. Although the weakly ventralized phenotypes classified as V4 include cuticular defects, similar or more severe than those caused by zen mutations (Fig. IB), they always show residual zen expression between 30 % and 60 % EL (Fig. 3B) and differentiate some amnioserosa (data not shown). This suggests that structures derived from the dorsalmost positions do not disappear totally before more lateral regions are affected. A complete deletion of the dorsal zen expression domain is first seen in ventralized embryos which also delete dorsolaterally derived structures (V3; Rushlow et al. 1987; Fig. 3C). As in the case of zen expression, twist expression is not uniformly affected along the anteroposterior axis. Weak alleles show no deviation from wild type (data

Developmental genetics of cactus 377 PMG Fig.. The morphogenetic movements of cactus mutant embryos. (A,B) Wild type. (C,D) cacfi 6 / cacfi 6. C (=V4). (E,F) cac^/cact* (=V). (G,H) cart* 10 [Df(L)TE116GW1 (=L with polarity). The germ band extension visible in H demonstrates residual polarity. (I,K) cacr Q /cact BQ (=D0). For each genotype, the same embryo is shown at gastrulation (stage 6: A,C,E,G,I) and lomin later at midgut invagination, stage 7: B,D,F,H,K). Staging is according to CamposOrtega and Hartenstein (1985). For phenotypic classification, see Table 1. CF, cephalic fold; DF, dorsal folds; AMG, anterior midgut; PMG, posterior midgut. not shown); intermediate alleles cause an expansion of the twist stripe only in terminal regions (Table 3b). This effect is more pronounced in strong alleles where the twist stripe is slightly wider also in the abdominal region (Table 3b; Fig. 3H). The observation that the twist expression and therefore the anlagen of the mesoderm are only weakly altered by cactus mutations suggests that the deletion of dorsal and dorsolateral anlagen is compensated by an expansion of predominantly ventrolateral anlagen. Transheterozygous combinations of the lossoffunction alleles including cactus deficiencies reveal no

378 5. Roth, Y. Hiromi, D. Godt and C. NussleinVolhard

Fig. 3. zen and twist expression in cactus mutant embryos. Embryos at blastoderm stage were stained using antizen or antitwist antibodies as described in Materials and methods. Wholemount preparations were photographed using Nomarski optics. Panel G, K and L show ventral surface views, all others show optical sagittal sections. (AE) Embryos stained with antizen antibodies. (FL) Embryos stained with antitwist antibodies. (A,F,G) Wild type. (B) cact" 8 / cact" 8, C (=V4). Similar expression patterns are seen in embryos derived from Df cacc/ females. (C) cact" 8 /cact" 8, 8 C (=V3). Note that the embryos shown in B and C have no change of zen expression in terminal regions (marked by stars). The terminal zen expression is not affected by alterations of the dorsoventral anlagen; it rather depends on terminal group genes (Rushlow et al. 1987). (D) cacp 0 / Df(L)TE116GW1 (=L with polarity). (E) cact 80 /cact 80 (=D0). The arrows in B and D mark a domain of residual zen expression. (H) cact* '/'cact*. (I,K) cact 013 / Df(L)E10RN (=V1). The embryos are derived from germ line chimeric females (see Table 4). (L) cact* 10 / cact* 810 (=D3). Similar expression patterns as in L are seen in embryos derived from dl~/ females at 8 C. The arrows in F, H, I and L mark the extent of twist expression. For phenotypic classification, see Table 1. unusual complementation behavior (Roth, 1990). Weak alleles (giving rise to V4 phenotypes) when tranheterozygous with strong alleles (giving rise to V phenotypes) result in an intermediate ventralization (V3). In transheterozygous combinations, some strong alleles behave like cactus deficiencies and thus may represent amorphic mutations. However, the group of mutations denned by this criterium is not uniform, but includes viable as well as zygotic lethal alleles. Therefore, the determination of the amorphic phenotype requires the analysis of the maternal effect of zygotic lethal cactus mutations. Germline clones with lethal cactus alleles The zygotic lethal cactus alleles cause late larval and pupal death associated with the formation of melanotic tumors (Sparrow, 1978; Roth, 1990). In order to analyse their maternal effects, we produced females with wildtype soma and mutant germline by pole cell transplantation. About one quarter of the fertile, chimeric Donor Cross cact 0 ' y cact 51 CyO CyO Developmental genetics of cactus 379 females produced strongly ventralized embryos while the others produced either wildtype or weakly ventralized (haploinsufficient phenotype) embryos (Table 4) reflecting the distribution of genotypes expected from the donor cross. This shows that the genotype of the germline determines the embryonic phenotype. Thus, cactus, like most of the dorsal group genes (Schiipbach and Wieschaus, 1986; Seifert et al. 1987; Konrad et al. 1988; Stein et al. in preparation) is a germlinedependent maternaleffect gene. We have produced germline clones with three different lethal alleles (Table 4). Embryos derived from a germline of the genotype cact /Df cact~ produce only small fragments of poorly differentiated ventral cuticle (Fig. IE). Their ventral twist domain is wider at all anteroposterior positions and they express twist uniformly in broad terminal regions (Fig. 31,K, Table 3b), indicating that they are more strongly ventralized (VI) than the embryos derived from the strongest viable alleles described above. The localized expression of the twist protein (Fig. 3I,K) demonstrates, however, that they are not completely ventralized and retain polarity. Germline chimeras with other lethal cactus alleles resulted in weaker phenotypes (V, Table 4). Thus, some of the lethal alleles have residual cactus activity. In summary, the lossoffunction alleles reveal that cactus activity is required in wildtype embryos to specify the development of dorsal, dorsolateral and, in part, ventrolateral structures. The hypomorphic alleles produce a continuous phenotypic spectrum characterized by coordinated alterations of the entire dorsoventral anlagen. The ventralmost part of the pattern, the mesoderm, is the least affected structure. The region most sensitive to a reduction of cactus activity is not the dorsalmost anlage of amnioserosa and dorsal epidermis, but the dorsolateral and ventrolateral ectoderm. Despite the various cuticular defects of head, thorax and telson and the nonuniform changes of zen and twist expression along the anteroposterior axis, cactus mutant embryos have normal segmental and terminal anteroposterior anlagen (Caroll et al. 1987; Roth, 1990). The gainoffunction alleles We have identified two mutations (cact*' 0 and Table 4. Germ line chimeras with lethal cactus alleles Adults" 104 Fertile females Genotype of germ line cact1 CyO CyO/CyO ' cact1 cact: 5 Phenotype c WT, V4 V cact 013 CyO Dfcacr* CyO 18 17 Df/CyO. CyO/CyO ' cacti Df: 3 14 WT, V4 VI 'Pole cells of embryos derived from donor crosses were transplanted into ovo D / recipients. The number of adult flies derived from injected embryos was recorded. "Dfcocr: Df(L)E10RN. c The phenotype of embryos derived from chimeric females was characterized using cuticle preparations (Fig. IE) and staining with twist antibodies (Fig. 31,K). For phenotypic classification see Table 1.

380 S. Roth, Y. Hiromi, D. Godt and C. NiissleinVolhard which in trans to lossoffunction cactus alleles produce dorsalized, rather than ventralized embryos (Fig. 1GK). These two mutations also fail to complement amorphic mutations of dorsal, giving rise to ly dorsalized embryos (data not shown). In order to distinguish whether E10 is a mutation in cactus, dorsal, or another interacting gene, we attempted to revert the maternal lethality of E10 in transheterozygotes with dorsal. Six Xray revertants of E10 were obtained, which produce viable progeny in trans to dorsal. Of these, two behave as strong cactus alleles while four are deficiencies that uncover cactus, but not dorsal (Table ; Ashburner et al. 1990). This experiment proves that E10 is indeed a cactus allele. For BQ, a spontaneous revertant was isolated, which exhibits the lossoffunction phenotype, suggesting that BQ is a cactus allele, too. When transheterozygous with weak, intermediate and strong cactus alleles, cacf 10 and cart 8 show a progressive deletion of ventral anlagen compensated by an expansion of dorsal anlagen (Fig. 1GK). The weak dorsalization exhibited by embryos derived from cact E10 /cact AB females is characterized by the loss of mesoderm (D3; Fig. 31). In transheterozygotes with cacf 11, the mesoderm is completely and the ventral epidermis is ly deleted (Fig. 1H). Finally, in trans to cacf 13 or a cactus deficiency the embryos differentiate only dorsal and dorsolateral cuticular structures (Fig. II). The results for cact E1 and cact BQ are similar. However, while cact E1 /Df cact~ leads to a lateralization at dorsolateral level (L, visible in gastrulation phenotype, Fig. G,H and zen expression, Fig. 3D), 40% of the embryos produced by cacf Q /Df cacc or cacf Q I cacf Q females are completely dorsalized. Cuticle phenotype (Fig. IK), gastrulation (Fig. 1,K) and zen expression extending around the entire EC (egg circumference) (Fig. 3E) resemble those caused by completely dorsalizing mutations of dorsal group genes. The observation that the dorsalizing effect of cact E1 and cad 8 is dependent on a reduction of cactus activity and can be suppressed almost completely by a wildtype copy of cactus indicates that these alleles do not represent overproducers. Rather, they lead to products which at ventral positions exert the dorsalizing function normally restricted to dorsal and lateral positions. They are therefore classified as neomorphic alleles that interfere with the spatial regulation required for normal dorsoventral pattern formation. As shown earlier, following loss of cactus activity more nuclei take up dorsal protein and consequently the nuclear dorsal protein gradient extends further towards the dorsal side than in wildtype (Roth et al. 1989, Fig. 4A,B). The opposite effect results from neomorphic cactus alleles. In agreement with their dorsalized phenotype they cause more nuclei to exclude the dorsal protein. In the most severe cases nuclear localization is abolished around the entire EC and the dorsal protein is predominantly cytoplasmic (Fig. 4C). The protein distribution cannot be distinguished from that caused by strongly dorsalizing alleles of dorsal Fig. 4. dorsal protein distribution in cactus mutant embryos. Embryos at blastoderm stage were stained using antidorsal antibodies as described in Materials and methods. Wholemount preparations were photographed using Nomarski optics. The maternal genotypes are: (A) Wild type. (B) cac^/cact*. (C) cacf'^/cacf' 3. group genes (Roth et al. 1989). In summary, lossoffunction and gainoffunction alleles of cactus act via opposing effects on the nuclear localization of dorsal protein. The time of action of the cactus gene product Most of the induced weak cactus alleles are heat sensitive: females homozygous or transheterozygous for cacr 48, cact Q6, cact Y ", cacf 8 and cact HE produce embryos that are more strongly ventralized at 9 C than at 18 C. The alleles cact Q6, cact Y ", cacf 8 lead to hatching larvae at 18 C when homozygous. We used cacf 8 and cacf E, which show the strongest temperaturedependent phenotypic changes, and performed temperatureshift experiments to determine the phenocritical period (Fig. 5). For both alleles the temperaturesensitivity is largely restricted to the syncytial blastoderm stage indicating that the cactus product is acting at the same time when the nuclear dorsal protein gradient is established and when nuclear dorsal protein exerts its function on zygotic gene

Developmental genetics of cactus 381 I8C.9C 9"C > I8 C 100 80 60 40 0 o Developmenta] stage at shift Fig. 5. The temperaturesensitive period of cact HE and cacf 18. The diagrams show the percentage of embryos with strong phenotypes (V for cacf le and V3 for cact 1^ O O) for the different developmental stages at which the temperatureshift occurred (see Materials and methods). For each time point, the phenotypes of 50100 differentiated embryos were scored. The developmental stages are indicated by numbers according to Campos Ortega and Hartenstein (1985): 1,, preblastoderm; 3, pole cell formation; 4, syncytial blastoderm; 5, cellularization; 6, gastrulation. expression (Roth et al. 1989; Rushlow et al. 1989; Steward, 1989). Phenotypic rescue by injection of wildtype cytoplasm The cactus mutant phenotype is only weakly suppressed by injections of wildtype cytoplasm. Embryos derived from cact PD /cact 011 females exhibit a ventralization of medium strength and do not produce filzkorper (V3). After transplanting wildtype cytoplasm into posterior dorsal positions, 560 % of the recipient embryos form patches of filzkorper material (Table 5; Fig. 6). The rescue response occurs only locally, as the injected embryos exhibit no weakening of the ventralization at more anterior positions. To investigate the spatial distribution of the rescuing activity the transplanted cytoplasm was taken from either the dorsal or the ventral side of wildtype embryos. No significant spatial asymmetry of the rescuing activity was observed (Table 5). Despite the variability in our results the experiments demonstrate Table 5. Partial rescue of cactus by the injection of wildtype cytoplasm Donor: Wlldtype Origin of cytoplasm" Dorsal Ventral Age Cleavage Syncytial Cleavage Syncytial Recipient: cacf"/cact PDb Number of recipient embryos Developed Rescued c 87 79 «101 JK 116 m % Rescued 5 60 40 30 "The cytoplasm was taken from either dorsal or ventral positions of wildtype embryos. b Recipient embryos (stage 4b) derived from cac/"'/cact PI> females were injected into the posterior dorsal side in all experiments. 5080% of the injected embryos developed cuticular structures. c The criterion for rescue was the production of filzkorper. Fig. 6. Cuticle phenotype of injected cactus mutant embryos. Cleavage stage cact p /cact PD embryos were injected with cytoplasm from wildtype cleavagestage embryos. The amount of transplanted cytoplasm corresponded to approximately 3% of the total egg volume. The cytoplasm was transplanted to the dorsal side of the posterior region. The injected embryos were allowed to differentiate. (A) The darkfield photograph of an injected embryo shows that the rescue response is weak and locally restricted. (B) Phasecontrast magnification of the posterior, dorsal region of the same embryo shows nonextended filzkorper (Fk) and spiracles (Sp) never seen in uninjected control embryos. that the cactus product, although predominantly required in the dorsal half of the EC (as inferred from the lossoffunction phenotype), is also present ventrally. Double mutants of cactus and completely dorsalizing mutants To analyse the relationship between cactus and the 11 dorsal group loci, we constructed females simultaneously homozygous for a strong cactus allele and an amorphic allele of a dorsal group gene (for complete description of genotypes, see Materials and methods). In ten of the eleven possible combinations, the completely dorsalized phenotype caused by the dorsal group mutation is changed upon the loss of cactus activity. Embryos are produced that differentiate ventral epidermis around the entire EC (Fig. 7A). In contrast to cactus embryos, they lack polarity: they do not form a ventral furrow during gastrulation and the

38 5. Roth, Y. Hiromi, D. Godt and C. NiissleinVolhard Antidorsal Fig. 7. Double mutants of cactus with amorphic alleles of dorsal group loci. (AD) cacr 4 / cacr 4 ; Tf QRE /Df(3R)ro XB3 (=L1). (E) ndf 93 /ndf 46 (=D0). (F) cac^/cact 4 ; ndf 93 '/ndf 46 (=L1). (G,H) cacr 4 dl'/cacr 4 dl' (=D0). (A,G) Darkfield photographs of cuticle. While the embryo in A shows ventral denticle belts surrounding the entire embryonic circumference, the embryo in G shows only dorsal epidermis and cannot be distinguished from embryos derived from dorsal homozygous females. (B) Living embryo during gastrulation (stage 6). The laterally derived CF is visible at both dorsal and ventral sides. No ventral furrow formation occurs. Compare to ventralized gastrulation shown in Fig. E,F. (C) Blastoderm embryo (stage 5) stained with antitwist antibodies. (D,H) Blastoderm embryo (stage 5) stained with antizen antibodies. (E,F) Blastoderm embryos (stage 5) stained with antidorsal antibodies. While the embryo in E reveals an exclusively cytoplasmic dorsal protein localization, all nuclei of the embryo in F contain medium levels of dorsal protein. Compare to Fig. 4A,B. CF, cephalic fold; VE, ventral epidermis. laterally derived cephalic fold is present dorsally and ventrally (Fig. 7B). They express neither zen nor twist (outside the terminal regions of the anteroposterior axis) (Fig. 7C,D), which demonstrates the loss of both dorsalmost and ventralmost pattern elements accompanied by the expansion of ventrolateral structures as characteristic for a lateralized phenotype (LI). All double mutant combinations share the complete lack of polarity with the amorphic single mutants of the dorsal group loci. Therefore, the dorsal group genes are all needed to polarize the embryonic dorsoventral pattern, but they are not absolutely necessary to express fates different from that of the dorsalmost anlagen. There is only one exception from this rule: the embryos produced by cact~dl~ females are completely dorsalized and indistinguishable from embryos produced by dl~ females. This can be seen by the cuticle phenotype, exhibiting only dorsal epidermis (Fig. 7G), and the uniform expression of zen (Fig. 7H). Hence, only the activity of the dorsal gene is an absolute prerequisite for the production of more ventral structures. As shown earlier, dorsal protein is present in embryos mutant for dorsalizing alleles of all dorsal

Developmental genetics of cactus 383 Fig. 8. Double mutants of hypomorphic cactus alleles with amorphic alleles of dorsal group loci. Maternal genotypes: (A,B) cact HE /cac HE ; spz m7 /spz 197 (9 C). (CF) cac^/cact 4 ; spz m7 /spz 197. (A,C) Darkfield photographs of cuticle preparations. (B,D) Phasecontrast photographs of cuticle preparations. (E) Gastrulating embryo (stage 6). (F) Blastoderm embryo (stage 5) stained with antizen antibodies. The arrows mark a domain of residual zen expression. The phenotype depicted in A,B is an apolar lateralization at a dorsolateral level (L). Dorsolaterally derived Fk material is visible at all positions of the embryonic circumference. The phenotype shown in CF resembles that of completely dorsalized embryos with respect to the cuticle pattern, however, in 30% of the embryos antennal sense organs are visible (D). The deep cephalic fold during gastrulation (E) and the lack of zen expression (F), features visible in all mutant embryos, demonstrate that this phenotype is lateralized, however, at a more dorsal level (L3) compared to L. See Table 1 for classification of phenotypes. ASO, antennal sense organs; CF, cephalic fold; Fk, filzkorper; Sp, spiracles. group genes (Steward, 1989; Roth et al. 1989). However, it is excluded from the nuclei and remains localized in the cytoplasm. The exclusively cytoplasmic localization is due to the presence of cactus product, because in doubly mutant embryos lacking both cactus and a dorsal group activity nuclei at all dorsoventral positions take up dorsal protein equally (Fig. 7E,F). The intermediate nuclear dorsal protein levels correspond to the uniform development of ventrolateral structures (Roth et al. 1989). If, instead of a strong cactus mutation (cacr 4 ), weaker alleles (e.g. cact^e or cact PD ) are combined with amorphic alleles of dorsal group genes, the level of apolar lateralization exhibited by the doubly mutant embryos is shifted to more dorsal positions. Thus, the developmental fate can be altered at all positions of the EC as a consequence of a change of cactus activity and hence, of the level of nuclear dorsal protein. We are able to distinguish three types of apolar lateralized embryos resulting from such double mutant combinations: (1) the ventrolateral level described above (LI, Fig. 7AD,F), () a dorsolateral level with rings of filzkorper material (L, Fig. 8A,B), (3) a dorsolateral level with dorsal epidermis, but reduced zen expression and a lateralized gastrulation phenotype (L3, Fig. 8CF).

384 S. Roth, Y. Hiromi, D. Godt and C. NiissleinVolhard Fig. 9. Double mutants of cactus with ly dorsalizing and ventralizing mutants. Embryos at blastoderm stage were stained using antitwist antibodies as described in Materials and methods. Wholemount preparations were photographed using Nomarski optics. The maternal genotypes are: (A) spz 67 /spz m? ( C, =D). (B) cact* /cact* ; spz^/spz"' 7 ( C, = L1 with polarity). The reduction of twist expression exhibited by the embryo shown in B is not visible in optical sagittal sections. (C) ea sl3 / (=L1 with polarity). (D) cact* / cacf 4 ; ea*' 3 / (=V1). (E) Tl rm9 /Tl"" 9 (=L1). (F) cact* / cacf* ; ji rm9 1Ti rm9 (=V0). The arrows demarcate the domain of twist expression. See Table 1 for classification of phenotypes. In summary, the double mutant analysis reveals that cactus and dorsal act downstream of all other dorsal group genes and that cactus is not an independent factor required for the determination of dorsal and lateral fates. Instead, it exerts its function via the dorsal product, dorsal can act as a morphogen in the absence of all other dorsal group genes. The only requirement for its morphogen function is the nuclear localization of the dorsal protein, whose extent is closely linked to the residual amount of cactus activity. Double mutants of cactus and ly dorsalizing and ventralizing mutants A variety of alterations of the dorsoventral pattern can be produced using ly dorsalizing, ventralizing and lateralizing alleles of different dorsal group genes (Anderson et al. 1985; Anderson and NiissleinVolhard, 1986). The different phenotypes correspond to different dorsal protein distributions (Steward, 1989; Roth et al. 1989). We wondered how these aberrant patterns change if cactus activity is reduced. A weakly dorsalizing mutation (e.g. spz 67 /spz 67 ) leads only to the deletion of mesoderm (D). In a double mutant with a strong cactus allele (cacf* / cacr 4 ; spz 67 /spz 67 ), the ability to form mesoderm is ly restored (Fig. 9A,B). A dominant easier allele (ea 5, Chasan et Anderson, 1989) can be used to produce lateralized embryos that still have residual polarity although they have lost dorsalmost and ventralmost structures and reveal only a weak ventral twist expression (LI with polarity). The reduction of the cactus activity in cact A /cact A ; ea 513 / females leads to a dramatic increase of the twist domain. However, the doubly mutant embryos still retain polarity as revealed by differences in the level of twist expression along their dorsoventral axis (VI, Fig. 9C,D). Finally, the removal of the cactus activity in a mutant background that already causes an apolar lateralized phenotype {Tl rm9 / Tl rm9, LI) leads to a complete ventralization (cacr 4 / cact 4 ; Tl rm9 /Tl rm9 ; Roth et al. 1989). The embryos express the twist protein evenly along their dorsoventral axis (Fig. 9E,F). They show no polarity during development and do not produce any cuticular structures (VO). In these three double mutant combinations, the reduction of cactus activity preserves the polarity of the pattern present in the single dorsal group mutant. The reduction of cactus activity affects all positions of the

Developmental genetics of cactus 385 EC and it leads to the formation of mesoderm not or only weakly present in the single mutant. Discussion Among the 1 maternaleffect genes that encode components of the dorsoventral pattern formation process cactus is unique in that only lossoffunction mutations of cactus lead to ventralized phenotypes. The observation that the ventralization exhibited by cactus embryos is accompanied by increased levels of nuclear dorsal protein suggested that cactus is a negative regulator of the nuclear localization of dorsal protein (Roth etal. 1989; Steward, 1989). This hypothesis raises several questions, which we address in this paper using mainly a genetic approach. (1) Does cactus indeed exert its function entirely via dorsal or is it also required independently for dorsoventral pattern formation? () Does cactus inhibit the nuclear localization of dorsal protein indirectly (mediated by other components) or via direct interactions with dorsal protein? (3) How is the cactus product distributed in the embryo? (4) What is the relationship between cactus and the dorsal group genes? (1) The lossoffunction alleles of cactus demonstrate that in wildtype embryos cactus activity is required (at least) for the determination of the dorsal, the dorsolateral and part of the ventrolateral anlagen. The strong dosage sensitivity of the phenotype might lead to the assumption that cactus, like bicoid or dorsal, acts as a morphogen and determines the anlagen of the dorsal half of the EC in a concentrationdependent manner. However, cactus is not required for the formation of dorsally derived structures if dorsal function is missing. Therefore, cactus is not an independent morphogen. It acts exclusively via dorsal. The dosage sensitivity and the continuity of phenotypic alterations produced by hypomorphic cactus alleles can be explained if the amount of cactus activity is closely linked to the morphogen function of dorsal. () In the double mutant combinations of cactus and completely dorsalizing alleles of dorsal group genes, the removal of cactus confers some potential to form ventrolateral pattern elements upon all dorsal group mutants with the only exception of dorsal. Hence, none of the sofaridentified components acts between cactus and dorsal and the products of both cactus and dorsal function at the end of the developmental pathway. Although we cannot be sure that all genes of the dorsoventral pattern formation process are known, the presented data are consistent with a close relationship, potentially a formation of a complex including cactus and dorsal products. It has been shown that the amorphic alleles of the dorsal group loci lead to embryos that contain normal amounts of dorsal protein. However, the protein fails to be taken up by the nuclei (Steward, 1989; Rushlow etal. 1989; Roth et al. 1989). The double mutants with cactus demonstrate that dorsal protein is not only present but also functional in the absence of dorsal group activities. Obviously the only requirement for its function is its nuclear localization, which can occur when cactus activity is reduced. One simple model would be that the cytoplasmic localization of dorsal protein is controlled by the formation of a complex with cactus protein. Only dorsal protein released from the cytoplasmic complex with cactus enters the nucleus. The investigation of several hypomorphic dorsal alleles that are defective in the cacfmsdependent inhibition of nuclear localization supports this view (Roth, 1990). The proposed biochemical interactions are reminiscent of those of the transcription factor NFJCB and its inhibitor IKB (Baeuerle and Baltimore, 1988; Zabel and Baeuerle, 1990). NFKB is required for the inducible expression of a variety of genes in different cell types (Lenardo and Baltimore, 1989). The active, nuclear form of NF*B is a complex of two polypeptides, p65 and p50 (Baeuerle and Baltimore, 1989). p50 represents the DNAbinding component of the complex and, as recently shown, has sequence similarities to the dorsal protein (Kieran et al. 1990; Ghosh et al. 1990). The other polypeptide of the complex, p65, is required for the interaction with the inhibitor protein, IKB (Baeuerle and Baltimore, 1989). If complex formation with IRB occurs, p50/p65 remains in the cytoplasm. Therefore, I^B acts as an inhibitor of p50/p65 nuclear transport. Given the sequence similarities between dorsal and p50, the structure of cactus might be related to IKB. (3) The formation of the nuclear gradient of dorsal protein requires that its nuclear localization is regulated in a graded manner along the dorsoventral axis. The inhibition that cactus exerts on the nuclear localization should be highest dorsally and it should continuously decrease towards more ventral positions. Unequal activity of cactus along the dorsoventral axis could arise from an asymmetric distribution of its product, the spatially regulated inhibition of its function or other activities competing with its function in a spatially controlled way. In cytoplasmic transplantation experiments, wefindcactus activity at both dorsal and ventral sides of wildtype embryos. However, the rescue response is so weak that we could only have excluded a very strong asymmetry of distribution. In double mutant combinations with various dorsalizing and ventralizing mutations (Figs 7, 8, 9), the reduction of cactus activity causes fate map changes at all positions along the dorsoventral axis. This indicates that cactus acts both dorsally and ventrally when dorsal group activities are absent or ly reduced. Because some of the mutations used in this study (e.g. spz 67 ) cause only weak phenotypes, we think that this conclusion is valid for wildtype embryos, too. Therefore, we propose that the cactus product is equally distributed along the dorsoventral axis. According to this interpretation the phenotypic effects caused by gainoffunction alleles of cactus are not due to a mislocalization of cactus products. We believe that the neomorphic alleles produce mutant cactus protein that like the wildtype protein is equally distributed, but leads to an unregulated inhibition of the nuclear localization of dorsal protein. The phenotypes caused by these alleles

386 S. Roth, Y. Hiromi, D. Godt and C. NiissleinVolhard resemble those of dorsal group mutations, because their products block the action of the dorsal group genes, which normally promote the graded nuclear localization of dorsal protein. Aside from information relating to the cactus function, the double mutant combinations of amorphic dorsal group alleles with cactus alleles of different strength (Figs 7, 8) are relevant to the interpretation of the morphogen function of dorsal protein. The embryos that result from these double mutants accumulate the same amount of dorsal protein in all nuclei as illustrated for LI embryos (Fig. 7F). They show no polarity during development and express the same pattern elements around the entire EC. Therefore, the pattern elements differentiated by such embryos cannot result from interactions of regions with different developmental fate, but are determined by the respective nuclear dorsal protein concentration in a largely autonomous way (Roth et al. 1989). Using cactus alleles of different strength, the nuclear dorsal protein concentrations can be altered in an almost continuous manner. In this way, we can determine how many distinct regions of the dorsoventral axis are directly dependent on the morphogen gradient. Here, we have described three different forms of lateralization (L3, L, LI). Thus, together with the complete dorsalization (DO) and ventralization (VO), so far, five different positional levels of the dorsoventral axis have been shown to depend on different nuclear dorsal protein concentrations. (4) The action of the dorsal group genes in the ventral half of the egg, which causes the formation of the nuclear concentration gradient of the dorsal protein, can be explained in two different ways. The dorsal group genes could act either via dorsal or via cactus (Roth et al. 1989). The first model explains the regulation of nuclear transport of dorsal protein assuming that the dorsal group genes and cactus act in parallel (Fig. 10). In this model, cactus could be a Mode) 1 Model dorsal group genes dorsal group genes cactus cactus nuclear transport of dorsal protein nuclear transport of dorsal protein Fig. 10. Two models explaining the relation of the dorsal group genes and cactus. According to model 1 cactus and the dorsal group genes act in parallel to regulate the nuclear transport of dorsal protein. Model assumes a serial action. cytoplasmic anchor that retains dorsal protein in the cytoplasm. Upon modification of dorsal protein due to the activity of dorsal group products (in the ventral half of the egg) dorsal protein is released.from its binding to cactus and enters the nucleus. The second model proposes a serial action of the dorsal group genes and cactus; this model formally represents the inhibition of an inhibition (Fig. 10). Also in this model cactus could be a cytoplasmic anchor that binds dorsal protein. However, the action of the dorsal group genes would cause a modification of cactus that leads to the release of dorsal protein. The second model proposing that the dorsal group genes act exclusively via cactus postulates that a complete loss of cactus function results in totally ventralized embryos without residual polarity. The double mutants between dorsal group genes (dorsal excepted) and cactus should also lead to a complete ventralization. At variance with these predictions, the strongest cactus mutations that we found so far have residual polarity and the double mutants investigated are lateralized at a ventrolateral level. However, despite the large number of available cactus alleles, it is possible that we have not observed the amorphic cactus phenotype. Some of the zygotic lethal cactus alleles have residual maternal activity. This could also apply to the strongest zygotic lethal cactus allele that we tested in germline chimeras. Furthermore, a currently unidentified component might exist that acts similar to cactus, so that only the lack of both activities would result in totally ventralized embryos. Thus, the existing genetic data do not exclude the possibility that the dorsal group genes act via cactus. If, as assumed in the first model, the dorsal group genes act directly on dorsal, even in the complete absence of cactus some residual polarity could be retained. Dorsal group activities might modify the dorsal protein not only to prevent its binding to cactus, but also to enhance the rate of its nuclear uptake. While the unmodified dorsal protein, when released from its complex with cactus, would reach only intermediate nuclear concentrations, the highest nuclear protein levels would require a modification of dorsal protein. A nuclear dorsal protein gradient with a high point ventrally could exist then even in the absence of cactus activity. The cytoplasmic localization of dorsal protein at ventral positions caused by the neomorphic cactus alleles can be explained in both models. The altered cactus proteins derived from these alleles may lack the ability to interact with dorsal group products, so that they cannot be modified and, hence, do not release bound dorsal protein. Alternatively, the proteins derived from neomorphic cactus alleles may bind the dorsal protein in a way that the modification of dorsal protein by dorsal group activities is blocked or that in spite of the modification no release from the complex with cactus occurs. Both alternatives explain why the neomorphic cactus alleles prevent the dorsal group products from stimulating the nuclear localization of dorsal protein.

Developmental genetics of cactus 387 Some results of our double mutant analysis impose certain constraints on models explaining the action of the dorsal group genes. The double mutants of amorphic dorsal group mutants with cactus alleles of different strength lead to apolar lateralized embryos which express ventrolateral or dorsolateral structures (LI, L, L3; Fig. 7AD, Fig. 8). These double mutants demonstrate that all major pattern elements present in the dorsal half of the embryonic circumference can be expressed in an apolar fashion upon reduction of cactus activity without any change in the activity of dorsal group products. Further, if intermediate levels of dorsal group activity corresponding to ventrolateral structures are present like in TF m9, the additional reduction of cactus activity in cact;tt" 9 double mutants, causes the formation of the ventralmost structure (mesoderm, Fig. 9). The examples show that the different nuclear concentrations of dorsal protein need not be in a onetoone correspondence with different levels of dorsal group activity. Rather, they can be generated in two distinct ways, either by a change in activity of cactus or of the dorsal group. These results can be easily explained using the second model of the double negative serial action of dorsal group genes and cactus, however they do not exclude the first model. Furthermore, both mechanisms do not exclude each other; thus, the possibility exists that they are realized simultaneously. The dorsal group genes could act on both cactus and dorsal protein to prevent complex formation and to stimulate nuclear import respectively. In the case of NFKB/IKB, it was shown that the dissociation of the cytoplasmic complex is mediated by a protein modification of the inhibitor. In response to a variety of extracellular signals, IJCB is phosphorylated, presumably by protein kinase C (Ghosh and Baltimore, 1990"). This phosphorylation causes the release of p50/p65 which then enters the nucleus and activates transcription. The molecular analysis of cactus will elucidate whether IKB has structural similarities to cactus, and, if so, whether functional similarities also exist. We are very grateful to T. Schupbach for the gift of the original cactus alleles and to D. St Johnston for the discovery of cact 8. We thank J. Habeck for help in preparing sections, E. Vogelsang for help with pole cell transplantations, C. Rushlow for the gift of antibodies. We thank R. Geisler, M. Mullins, M. Leptin, D. Stein, and L. Stevens for stimulating discussion and suggestions on the manuscript. R. Groemke Lutz prepared photographs. The contribution by Y. H. was done while he was a postdoctoral fellow with Corey S. Goodman. Y. H. would like to thank C. S. G. for support, him and K. V. Anderson for stimulating discussions. This work was supported by the DFG (Leibniz program). Y. H. was a senior postdoctoral fellow of the American Cancer Society. References ANDERSON, K. V. (1987). Dorsalventral embryonic pattern genes of Drosophila. TIG 3, 9197. ANDERSON, K. V., JORGENS, G. AND NOSSLEINVOLHARD, C. (1985). Establishment of dorsalventral polarity in the Drosophila embryo: genetic studies on the role of the Toll gene product. Cell 4, 779789. ANDERSON, K. V. AND NOSSLEINVOLHARD, C. (1984). Information for the dorsalventral pattern of the Drosophila embryo is stored as maternal mrna. Nature 311, 37. ANDERSON, K. V. AND NOSSLEINVOLHARD, C. (1986). Dorsalgroup genes of Drosophila. In Gametogenesis and the Early Embryo (ed. J. Gall) pp. 177194. New York: Alan R. Liss, Inc. ASHBURNER, M., THOMSON, P., ROOTE, J., LASKO, P. F., GRAU, Y., EL MESSAL, M., ROTH, S. AND SIMPSON, P. (1990) The genetics of a small autosomal region of Drosophila melanogaster, including the structural gene for Alcohol dehydrogenase. VII. Characterization of "the region around the snail and cactus loci. Genetics 16, 679694. BAEUERLE, P. A. AND BALTIMORE, D. (1988). IRB: a specific inhibitor of the NFxrB transcription factor. Science 4, 540546. BUSSON, B., GANS, M., KOMITOPOULOU, K. AND MASSON, M (1983). Genetic analysis of three dominant female sterile mutations on the Xchromosome of Drosophila melanogaster. Genetics 105, 30935. CAMPOSORTEGA, J. A. AND HARTENSTEIN, V. (1985). The Embryonic Development of Drosophila melanogaster. Berlin: Springer. CARROLL, S. B., WINSLOW, G. M., TWOMBLY, V. J. AND SCOTT, M. P. (1987). Genes that control dorsoventral polarity affect gene expression along the anteroposterior axis of the Drosophila embryo. Development 99, 3733. CHASAN, R. AND ANDERSON, K. V. (1989). The role of easter, an apparent serine protease, in organizing the dorsalventral pattern of the Drosophila embryo. Cell 56, 391400. DELOTTO, R. AND SPIERER, P. (1986). A gene required for the specification of dorsalventral pattern in Drosophila appears to encode a serine protease. Nature 33, 68869. FROHNHOFER, H. G. AND NOSSLETNVOLHARD, C. (1986). Organisation of anterior pattern in the Drosophila embryo by the maternal gene bicoid. Nature 34, 1015. GHOSH, S. AND BALTIMORE, D. (1990). Activation in vitro of NFfcB by phosphorylation of its inhibitor IkB. Nature 344, 67868. GHOSH, S., GIFFORD, A. M., RIVIERE, L. R., TEMPST, P., NOLAN, G. P. AND BALTIMORE, D. (1990). Cloning of the p50 DNA binding subunit of NF*rB: homology to rel and dorsal. Cell 6, 1019109. HASHIMOTO, C, HUDSON, K. L. AND ANDERSON, K. V. (1988). The Toll gene of Drosophila, required for dorsalventral embryonic polarity, appears to encode a transmembrane protein. Cell 5, 6979. IRISH, V. F. AND GELBAJTT, W. M. (1987). The decapentaplegic gene is required for dorsalventral patterning of the Drosophila embryo. Genes Dev. 1, 868879. KIERAN, M., BLANK, V., LOCEAT, F., VANDEKERCKHOVE, J., LOTTSPEICH, F., LE BAIL, O., URBAN, M. B., KOURILSKY, P., BAEUERLE, P. A. AND ISRAEL, A. (1990). The DNA binding subunit of NFJCB is identical to factor KBF1 and homologous to the rel oncogene product. Cell 6, 10071018. KONRAD, K. D., GORALSKI, T. J. AND MAHOWALD, A. P. (1988). Developmental genetics of the gastrulation defective locus in Drosophila melanogaster. Devi Biol. 17, 13314. LEHMANN, R. AND NOSSLEINVOLHARD, C. (1986). hunchback, a gene required for segmentation of an anterior and posterior region of the Drosophila embryo. Devi Biol. 119, 40417. LENARDO, M. J. AND BALTIMORE, D. (1989). NF*rB: a pleiotropic mediator of inducible and tissuespecific gene control. Cell 58, 79. LEPTIN, G. AND GRUNEWALD, B. (1990). Cell shape changes during gastrulation in Drosophila. Development 110, 7384. MACDONALD, P. M. AND STRUHL, G. (1986). A molecular gradient in early Drosophila embryos and its role in specifying the body pattern. Nature 34, 537545. NOSSLEINVOLHARD, C. (1979a). Maternal effect mutations that alter the spatial coordinates of the embryo of Drosophila melanogaster. In Determinants of Spatial Organization (ed. I.

388 S. Roth, Y. Hiromi, D. Godt and C. NilssleinVolhard Koenigsberg and S. Subtelney), pp. 18511. New York: Academic Press. NOSSLEINVOLHABD, C. (1979b). Pattern mutants in Drosophila embryogenesis. In Cell Lineage, Stem Cells and Cell Determination. INSERM Symposium No. 10, (ed. N. Le Douarin) pp. 698. New York: NorthHolland. N0SSLE1NVOLHARD, C, FROHNHOFER, H. G. AND LEHMANN, R. (1987). Determination of anteroposterior polarity in Drosophila. Science 38, 16751681. NOSSLEINVOLHARD, C, LOHSSCHARDIN, M., SANDER, K. AND CREMER, C. (1980). A dorsoventral shift of embryonic primordia in a new matemaleffect mutant of Drosophila. Nature 83, 474476. NOSSLEINVOLHARD, C. AND ROTH, S. (1989). Axis determination in insect embryos. In Cellular Basis of Morphogenesis. Ciba Foundation Symposium 144. pp. 3755. New York: John Wiley and Sons. NOSSLEINVOLHARD, C, WIESCHAUS, E. AND KLUDING, H. (1984). Mutations affecting the pattern of the larval cuticle in Drosophila melanogaster. I. Zygotic loci on the second chromosome. Wilhelm Roux's Arch, devl Biol. 183, 678. ROTH, S. (1990). Die Rolle der maternalen Gene cactus und dorsal bei der dorsoventralen Musterbildung im Drosophila Embryo. PhD Thesis. University of Tubingen. ROTH, S., STEIN, D. AND NOSSLEINVOLHARD, C. (1989). A gradient of nuclear localization of the dorsal protein determines dorsoventral pattern in the Drosophila embryo. Cell 59, 118910. RUSHLOW, C, FRASCH, M., DOYLE, H. AND LEVINE, M. (1987). Maternal regulation of zerknullt: a homoeobox gene controlling differentiation of dorsal tissues in Drosophila. Nature 330, 583586. RUSHLOW, C. A., HAN, K., MANLEY, J. L. AND LEVINE, M. (1989). The graded distribution of the dorsal morphogen is initiated by selective nuclear transport in Drosophila. Cell 59, 11651177. SANTAMARIA, P. AND NOSSLEINVOLHARD, C. (1983). Partial rescue of dorsal, a maternal effect mutation affecting the dorsoventral pattern of the Drosophila embryo, by the injection of wildtype cytoplasm. EMBO J., 16951699. SCHOPBACH, T. AND WIESCHAUS, E. (1986). Germline autonomy of maternaleffect mutations altering the embryonic body pattern of Drosophila. Devi Biol. 113, 443448. SCHOPBACH, T. AND WIESCHAUS, E. (1989). Female sterile mutations on the second chromosome of Drosophila melanogaster. I. Maternal effect mutations. Genetics 11, 101117. SEIFERT, E., MOLLERHOLTKAMP, F., MARCEY, D. AND JACKLE, H. (1987). Injection of wildtype cytoplasm and poly(a) RNA provokes phenotype rescue in sptitzle mutant Drosophila embryos. Roux's Arch, devl Biol. 196, 788. SPARROW, J. C. (1978). Melanotic 'Tumours'. In The Genetics and Biology of Drosophila. vol. b (ed. M. Ashburner and T. R. F. Wright), pp. 77313. New York und London: Academic Press. STEWARD, R. (1987). Dorsal, an embryonic polarity gene in Drosophila, is homologous to the vertebrate protooncogene, c rel. Science 38, 69694. STEWARD, R. (1989). Relocalization of the dorsal protein from the cytoplasm to the nucleus correlates with its function. Cell 59, 11791188. STEWARD, R. AND NOSSLEINVOLHARD, C. (1986). Tha genetics of the dorsalbicaudald region of Drosophila melanogaster. Genetics 113, 665^78. STEWARD, R., ZUSMAN, S. B., HUANG, L. H. AND SCHEDL, P. (1988). The dorsal protein is distributed in a gradient in early Drosophila embryos. Cell 55, 487495. ST JOHNSTON, R. D. AND GELBART, W. M. (1987). Decapentaplegic transcripts are localized along the dorsalventral axis of the Drosophila embryo. EMBO J. 6, 785791. THISSE, B., STOETZEL, C, EL MESSAL, M. AND PERRINSCHMITT, F. (1987). Genes of the Drosophila maternal dorsal group control the specific expression of the zygotic gene twist in presumptive mesodermal cells. Genes Dev. 1, 709715. THISSE, B., STOETZEL, C, GOROSTIZATHISSE, C. AND PERRIN SCHMITT, F. (1988). Sequence of the twist gene and nuclear localization of its protein in endomesodermal cells of early Drosophila embryos. EMBO J. 7, 175183. VAN DER MEER, J. M. (1977). Optical clean and permanent whole mount preparations for phasecontrast microscopy of cuticular structures of insect larvae. Dros. Info. Serv. 5, 160. WAKIMOTO, B. T., TURNER, F. R. AND KAUFMAN, T. C. (1984). Effects in embryogenesis in mutants associated with the antennapedia gene complex of Drosophila melanogaster. Devl Biol. 10, 14717. ZABEL, U. AND BAEUERLE, P. A. (1990). Purified human IATB can rapidly dissociate the complex of the NFirB transcription factor with its cognate DNA. Cell 61, 5565. ZUSMANN, S. AND WIESCHAUS, E. (1985). Requirements for zygotic gene activity during gastrulation in Drosophila melanogaster. Devl Biol. Ill, 359371. {Accepted 0 March 1991)