The Dorsal-related immunity factor (Dif) can define the dorsal-ventral axis of polarity in the Drosophila embryo

The v-rel oncogene was originally identified as the oncogene associated with the highly virulent turkey retrovirus REV-T (Stephens et al., 1983; Wilhelmsen et al., 1984; Wilhelmsen and Temin, 1984), which induces fatal lymphoid tumors in young birds. Over the last decade a number of proteins have been isolated that share amino acid sequence similarity with an approximately 300-amino-acid region in v-Rel, the Rel domain. These proteins include c-Rel, the cellular counterpart of v-Rel, the p50 and p65 subunits of the immune system transcription factor NFκB, and the Drosophila dorsal-ventral (DV) morphogen, Dorsal (Wilhelmsen et al., 1984; Wilhelmsen and Temin, 1984; Bours et al., 1990; Ghosh et al., 1990; Kieren et al., 1990; Nolan et al., 1991; Ruben et al., 1991; Steward, 1987). Specific segments of the Rel-homologous regions of these proteins have been shown to mediate DNA binding, protein-protein interactions and nuclear localization (see Verma et al., 1995).

The Rel-homologous (RH) proteins share the striking feature of being regulated at the level of nuclear localization (reviewed in Verma et al., 1995). A variety of stimuli result in translocation of RH proteins to the nucleus, where they act to regulate transcription. The nuclear localization of RH proteins is determined by their interactions with other proteins, termed IκBs (for a review, see Beg and Baldwin, 1993). A number of IκB proteins that have been characterized show structural similarity to one another, carrying 6-7 copies of a protein motif known as the ankyrin repeat. IκBs act as inhibitors of nuclear localization by binding directly to their cognate Rel-homologues and retaining them in the cytoplasm. Several types of evidence have suggested a model in which this binding interferes with the ability of the nuclear localization signal (nls) receptor to recognize the nls carried within the Rel domain of the RH protein (Beg et al., 1992; Ganchi et al., 1992; Henkel et al., 1992).

The controlled nuclear localization of the RH protein Dorsal plays an important part in pattern formation along the DV axis during development of the Drosophila embryo (reviewed in Morisato and Anderson, 1995). In addition to Dorsal, this process requires its cognate IκB-homologous partner Cactus (Kidd, 1992; Geisler et al., 1992), as well as the function of ten additional maternal-effect loci. In wild-type embryos, Dorsal protein is distributed in a nuclear gradient with highest concentrations of the protein in the nuclei adjacent to the ventral side of the egg. Females homozygous for loss-of-function mutations in dorsal (dl) or the ten other dorsal group genes produce embryos that are dorsalized. In syncytial blastoderm embryos lacking the function of any of these ten genes, all Dorsal protein is cytoplasmic and a nuclear gradient never forms. In contrast, females carrying loss-of-function alleles of cactus (cact) produce embryos that are ventralized and have ectopically high concentrations of Dorsal protein in nuclei all around the DV axis. Thus, the dorsal group genes and cact act in a pathway in which the dorsal group gene products promote, and Cactus inhibits, nuclear localization of Dorsal. Once localized to the nucleus and depending on its concentration, Dorsal protein activates transcription of some target genes and represses others, depending on the promoter context of the particular target gene. (Thisse et al., 1987, 1988, 1991; Jiang et al., 1991; Pan et al., 1991; Bouley et al., 1987). Complementary studies of transcriptional regulation in the vertebrate immune system and pattern formation in the Drosophila embryo point to a remarkable conservation of the components in the regulatory pathways controlling NFκB and Dorsal nuclear localization (Wasserman, 1992). The polarity of the nuclear gradient of Dorsal protein and in turn of the embryonic DV axis is defined by a ventrally localized extracellular signal that binds to and activates a receptor encoded by the Toll locus (Hashimoto et al., 1988; Stein et al., 1991; Morisato and Anderson, 1994; Schneider et al., 1994). The signal transmitted by the Toll receptor into the cytoplasm of the embryo leads to the degradation of Cactus (Belvin and Anderson, 1995) and ultimately to the graded nuclear uptake of the Dorsal protein. One of the proteins required downstream of Toll for the propagation of the ventralizing signal is the product of the pelle gene, a putative serine-threonine kinase (Shelton and Wasserman, 1993) whose target remains unknown. Interestingly, the cytoplasmic portion of the Toll protein exhibits amino acid sequence similarity to the cytoplasmic portion of the IL-1 receptor (Schneider et al., 1991; Gay and Keith, 1991), which can mediate NFκB activation. In addition, IRAK, a mammalian homologue of the Pelle protein, was identified and shown to complex with IL-1 receptor and become phosphorylated in cells responding to stimulation by IL-1 (Cao et al., 1996). The similarities between the components of the NFκB and Dorsal regulatory pathways indicates a striking evolutionary conservation.

Ip et al. (1993) identified a gene encoding a novel RH protein that they named Dorsal-related immunity factor (Dif). Within the Rel-homologous regions, Dorsal and Dif exhibit 48% amino acid identity. Outside of the Rel-homologous regions, however, they show no marked similarity. Its expression in the fat body and its rapid nuclear uptake in response to bacterial challenge strongly suggest a role for Dif in the activation of antimicrobial genes in the Drosophila immune response. Additionally, Dif is capable of activating the cecropin A1 promoter as well as a promoter carrying the diptericin κB-like target sites in tissue culture cells following cotransfection (Petersen et al., 1995; Gross et al., 1996).

The observations described above suggest that the mechanisms regulating Drosophila DV axis formation as well as immune system function in both Drosophila and man share a common evolutionary origin. In order to investigate the specificity of regulation and activation of the Dif and Dorsal proteins, we have expressed the dif gene product in Drosophila embryos and shown that it is capable of substituting, to a significant degree, for Dorsal. Under the control of cact and the dorsal group of genes, Dif can define a polarized DV axis of polarity with many of the appropriate pattern elements in their normal locations. Our results provide support for the idea that at least one of the functions of the dorsal group of genes in the immune system is to regulate the activity of Dif. In the embryo, Dif can activate and repress known target genes of Dorsal. Our data suggest, however, that Dif may not be able to interact with certain basic helix-loop-helix (bHLH) proteins with which Dorsal normally cooperates to define ventral and ventrolateral regions of the embryo.

All stocks were maintained and eggs were collected employing standard conditions and procedures (Roberts, 1986; Wieschaus and Nüsslein-Volhard, 1986). Staging of embryos was according to Campos-Ortega and Hartenstein (1985).

The wild-type stock used was Oregon R. dl¹ is described by Nüsslein-Volhard et al. (1980). Df(2L)TW119, which uncovers the dl locus, and the pip alleles pip³⁸⁶ and pip⁶⁶⁴, are described in Lindsley and Zimm (1992). The chromosome carrying both the dl¹ and cact^A2 mutations was constructed by Roth et al. (1991). For the sake of simplicity, throughout this report we refer to the progeny of dl mutant females as dl⁻ embryos.

Plasmid pCaSpeRbcdBglII was constructed in order to express genes in the female germ-line. This expression vector, a pCasPeR (Pirrotta,1988) derivative, carries 2 kb of the 5′ region from the bicoid (bcd) gene. This includes the maternally expressed promoter, extending from a BamHI site to the PstI site at nt 1244 (for numerical coordinates, see Berleth et al., 1988). Adjacent to the promoter fragment this vector carries the genomic region extending from the SmaI site at nt 4292 to the EcoRI site at 5.9 kb, which extends 1kb downstream of the polyadenylation signal. This construct does not contain the bcd anterior localization determinants. At the junction between the two genomic fragments (nts 1244/4292), a BglII site has been created for introduction of genes to be expressed. A variety of genes, including dl, cact, easter (D. S., unpublished) and bicoid (Driever et al., 1990) have been expressed using this vector.

For the expression of Dif in the female germ-line, a full-length dif cDNA was excised from plasmid pNB40 (Brown and Kafatos, 1988) as a HindIII/NotI fragment, made blunt-ended and ligated to BclI linkers. This fragment includes dif under the translational control of the Xenopus Globin leader. After cleavage with BclI, the dif fragment was ligated to BglII-cleaved pCasPeRbcdBglII. A plasmid carrying the dif fragment in the appropriate orientation to allow bcd promoter-directed transcription was identified and designated pCasperbcddif.

Fly transformants carrying the dif gene under the control of the bcd promoter were obtained by introduction of pCasperbcddif into the genome of flies of the genotype w/w; dl¹cn sca/CyO by conventional methods of microinjection and P-element-mediated transformation (Spradling, 1986). Two transgene inserts, one located on the second chromosome (4B) and one located on the third chromosome (6A), were used in all the experiments reported here. In studies using flies carrying a single insert, the genotype of was w/w; dl¹cn sca/dl¹cn sca; pCasper[w⁺, bcd-dif6A]/+. In studies using four copies of Dif, the genotype was w/w; dl¹cn sca pCasper[w⁺, bcd-dif4B]/ dl¹cn sca pCasper[w⁺, bcd-dif4B]; [w⁺, bcd-dif6A]/ [w⁺, bcd-dif6A].

Living embryos at the beginning of gastrulation were photographed under bright field illumination after submersion in Voltalef 3S halocarbon oil. For assessment of cuticle patterns, embryos of various phenotypes were allowed to develop for 2-4 days at 18°C. Eggs were dechorionated using 50% bleach in water. Dechorionated embryos were mounted in a drop of Hoyers mounting medium:lactic acid (1:1) (van der Meer, 1977) and placed at 65°C for 24 hours. After clearing, embryos were photographed using dark-field optics. In cases where the level of rescue was quantified, cleared embryos were scored for the presence of Filzkörper, Filzkörper and denticle material, or Filzkörper and well-patterned denticle bands.

In situ hybridization to RNA was carried out according to Tautz and Pfeifle (1989) with the following modifications: 10% dimethylsulfoxide was added to the paraformaldehyde fixative, and the incubation with anti-digoxigenin antibody (Boehringer-Mannheim) was carried out overnight at 4°C at a concentration of 1:5,000 followed by a 30-60 minute incubation at room temperature. Probes consisted of gel-purified DNA fragments labelled with digoxigenin-dUTP by random priming. Full-length cDNAs were used for zen, dpp and rhomboid (rho) (Bier et al., 1990). For single-minded (sim) (Crews et al., 1988), the probe was a 1.2 kb SacI fragment encoding the amino acid residues 511-655 and the 3′ UTR. For twi, a 537 nt BamHI-EcoRI fragment extending from coordinates 1154-1691 (GenBank accession no. x14569) was used. For sna, a DNA fragment corresponding to the entire open reading frame was used.

To determine the width of the twi and sna expression domains, stained cellular blastoderm-stage embryos were selected and oriented with ventral surfaces up. At approximately 50% egg length, the number of cell diameters extending from one lateral boundary of the strongly stained region to the other was counted. As the arrangement of neighboring cells is not identical from embryo to embryo, and as there is some variability in the width of expression domains in Dif-expressing embryos, values presented are the average of at least six determinations.

For two-hybrid studies of the interactions between Twist and the Rel-homologous regions of Dorsal and Dif, dorsal and dif were introduced into plasmid pADNS as HindIII and HindIII/NotI fragments, respectively. Both cDNAs contained translation signals from the Xenopus globin gene. pADNS carries the yeast 2 μ plasmid origin of replication for propogation in Saccharomyces cerevisiae as well as ADH promoter sequences for the expression of introduced genes (Collicelli et al., 1989). Introduction of either pADNS-dorsal or pADNS-dif into a derivative of yeast strain NLY2 carrying the E. coli lacZ gene under the transcriptional control of two consensus binding sites for Rel (Lehming et al., 1994, 1995) resulted in high levels of β-galactosidase expression (data not shown). This result confirmed those obtained by Lehming et al. (1995) using similar dorsal- and dif-expressing plasmids.

Derivatives of pADNS-dorsal and pADNS-dif that included the RH regions but lacked the transcriptional activation domains were constructed using conventional site-directed mutagenesis to introduce stop codons at the C-terminal ends of the RH domains. pADNS-dlRel encodes amino acids 1-348 (leu349 to stop) of the Dorsal protein while pADNS-difRel encodes amino acids 1-380 (leu381 to stop) of Dif.

twi DNA sequences encoding residues 3-490 of the twist open reading frame were introduced as an EcoRI/SalI fragment into similarly digested pEG202, resulting in a plasmid encoding an in-frame fusion between the lexA DNA binding region and Twist. Introduction of this plasmid into a yeast strain (CTY) carrying the E. coli lacZ gene under the transcriptional control of the lexA DNA binding motif (Kalpana and Goff, 1993) resulted in high levels of β-galactosidase expression (data not shown), confirming that the activation region of Twist is expressed and functions in S. cerevisiae.

As described above for twi, T4(scute) DNA sequences encoding residues 100-345 of the T4 open reading frame were introduced as an EcoRI/XhoI fragment into similarly digested pEG202. Previously, Gonzalez-Crespo and Levine (1993) established that this region of T4 conferred the ability to interact with the Dorsal RH domain in vitro. This fusion protein was also capable of directing high-level expression of β-galactosidase under lexA-responsive promoter control (data not shown).

Plasmids were introduced into yeast strains by conventional lithium acetate-mediated transformation (Ito et al., 1983). For each interaction study, three independently isolated colonies containing the appropriate plasmids were first patched onto synthetic drop-out medium plates (Rose et al., 1980) prepared with the appropriate amino acids omitted for selection and grown at 30°C for 72 hours. Using a sterile loop, approximately 100 μl of cells were then transferred to an Eppendorf tube containing 500 μl of water. After centrifugation of the tubes, cell pellets were resuspended in 200 μl breaking buffer (0.1 M Tris, pH 8.0, 20% glycerol, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride). Extracts were prepared after addition of an equal volume of glass beads by vortexing (10× for 30 seconds) at 4°C. Extracts were assayed for β-galactosidase activity by the method of Miller (1972). Values reported are the averages from assays of three independent transformants.

In wild-type embryos, the pattern elements along the DV axis defined by the dorsal group and cact are, from ventral to dorsal: the mesoderm; the ventral ectoderm, which gives rise to the central nervous system and ventral hypoderm; the dorsolateral ectoderm, from which the tracheae and dorsal hypoderm are derived; and the amnioserosa. The ventral and dorsolateral ectoderm give rise to the tissues that secrete the cuticle of the first instar larva. We have used the formation of cuticular structures as sensitive indicators of the ability of Dif to substitute for Dorsal protein in DV pattern formation. The ventralmost cuticle carries the conspicuous ventral denticles (Fig. 1A), while at ventrolateral positions the cuticle is naked. At the posterior, dorsolateral tissue is marked by the presence of Filzkörper, which are tracheal specializations. The dorsalmost cuticle carries segmentally arranged rows of dorsal hairs.

The progeny of females carrying complete loss-of-function alleles of any of the dorsal group genes develop into symmetric tubes of cuticle with rings of dorsal hairs. Neither ventral denticles nor Filzkörper are formed in such embryos (Fig. 1B).

Embryos from dl mutant mothers expressing Dif exhibited partial restoration of cuticular pattern elements along the DV axis. Most embryos produced by females carrying a single copy of p[w⁺, bcd-dif] produced Filzkörper as well as naked cuticle of ventrolateral origin. In addition to these structures, other embryos developed a patch of cuticle containing denticle material (Fig. 1C). More completely rescued embryos produced segmentally distinct bands of ventral denticle material (Fig. 1D). In these cases, the ventral denticles were poorly arranged in comparison to those of wild-type larvae, and bands of ventral denticles were often seen to cross one another. This degree of rescue was representative of 9 out of the 12 independent lines obtained. Two lines showed a slightly lower level of rescue, producing embryos with Filzkörper but no denticle material. Embryos from the remaining transformant line were not rescued. We attribute differences in level of rescue to position effects on expression levels related to P-element insertion sites. None of the transformant lines produced hatching embryos.

We used the morphogenetic movements that occur during gastrulation, and their orientation with respect to the egg shell, as sensitive indicators of the ability of Dif to confer polarity to embryos lacking Dorsal activity. In gastrulating wild-type embryos, ventral furrow formation, which always occurs adjacent to the curved ventral side of the egg, signals the invagination of the mesoderm. In addition, on the dorsal side, folds form ahead of the invaginating posterior midgut as it moves toward the anterior (Fig. 2A).

In embryos derived from females homozygous for loss-of function alleles of any of the dorsal group genes, gastrulation is aberrant. Neither the ventral furrow nor the posterior midgut invagination form. Instead symmetrical ‘dorsal’ folds develop all around the circumference of the embryo (Fig. 2B).

In contrast, dl⁻ embryos expressing Dif exhibited features consistent with an appropriately polarized pattern of gastrulation. Most conspicuous was the anterior movement of the posterior midgut invagination along the dorsal side of the embryo (Fig. 2C). However, these embryos did not develop a ventral furrow, indicating that the mesoderm did not form properly.

Because Dif expression in embryos lacking Dorsal activity was capable of mediating rescue of both DV pattern elements and morphogenetic movements, we wanted to define this rescue at the molecular level by analyzing the expression patterns of known target genes of Dorsal. To accomplish this, the expression patterns of twi, zen, dpp, sna, rho and sim were analyzed by in situ hybridization. As preliminary data indicated that a single copy of the Dif transgene achieved levels of Dif protein that were near the threshold required for ventral activation of sna, we constructed a strain carrying four copies of Dif under the transcriptional control of the bcd promoter. The expression studies described below were carried out with this strain.

Twist is a bHLH protein whose transcription is activated by Dorsal protein (Thisse et al., 1987, 1988). In wild-type embryos, twi RNA is detected at blastoderm stage as a ventral stripe of expression in the developing mesodermal anlagen (Thisse et al., 1987) (Fig. 3), which peaks during gastrulation. In embryos lacking Dorsal activity, twi is not expressed (Kosman et al., 1991; Ray et al., 1991) (Fig. 3). The pattern of twi expression in embryos lacking Dorsal protein but expressing Dif was virtually indistinguishable from that seen in wild-type (Fig. 3). In fact, in embryos derived from females carrying four copies of the Dif construct, twi was expressed in a slightly wider domain than wild-type embryos. In these embryos, strong twi expression extended over approximately 18 cell diameters and faded in the lateral boundaries over an additional few cell diameters. In the case of wild-type embryos that were stained under identical conditions, strongest twi expression was observed over a region of approximately 16 cell diameters.

To confirm that Dif, like Dorsal, was interacting directly with the twi regulatory region, a P-element-based transgene carrying the twi promoter upstream of the E. coli lacZ gene (Pan et al., 1991) was crossed into embryos from females expressing a single copy of Dif in a dl⁻ background. As expected, staining of these embryos with an antibody directed against β-galactosidase showed a ventral stripe of expression similar to that of the same reporter construct in a wild-type background (data not shown).

Both zen and dpp are repressed by Dorsal in all but the most ventral region of the embryo. In the early embryo, the expression patterns of zen (Fig. 3) and dpp are virtually coincident; both are first expressed during the syncytial blastoderm stage in a longitudinal band along the dorsal side of the embryo (Doyle et al., 1986; St Johnston et al., 1987). Expression peaks in the cellular blastoderm stage and decreases during gastrulation. In embryos lacking Dorsal activity, both zen (Rushlow et al., 1987) (Fig. 3) and dpp (Ray et al., 1991) are expressed evenly around the DV circumference, beginning at syncytial blastoderm and continuing through cellular blastoderm and gastrulation stages. In embryos lacking Dorsal activity but expressing Dif, however, both zen (Fig. 3) and dpp (data not shown) expression was repressed on the ventral side of the majority of embryos. In some embryos, however, strong zen expression was detected in ventral regions (data not shown).

The wild-type pattern of sna RNA expression is transient and highly dynamic (Alberga et al., 1991; Kosman et al., 1991; Leptin, 1991). The sna transcript is first detected at syncytial blastoderm in a ventral pattern corresponding to the cells of the developing mesoderm (Fig. 3). This expression persists during cellular blastoderm and strong expression continues in the invaginating mesoderm as gastrulation proceeds. sna is expressed in a precisely bordered domain of approximately 18 cell diameters (see also Ip et al., 1992b). In dl⁻ embryos, no ventral expression of sna is observed (Kosman et al., 1991; Ray et al., 1991) (Fig. 3), consistent with the lack of mesodermal precursor cells. In embryos from dl mutant females carrying Dif, sna was expressed in a broad stripe along the embryonic ventral midline. Although some variability was detected in the width of the sna expression domain when measured in the central regions of the embryo, the expression domain extended over approximately 12 cell diameters in most embryos. In striking contrast to the precise border of expression observed in wild-type embryos, sna expression in Dif-expressing embryos faded off gradually in lateral regions, similar to what is seen for twi expression.

In wild-type embryos, rho is expressed in three stripes, two ventrolateral and one dorsal (Bier et al., 1990) (Fig. 3). In embryos lacking Dorsal activity, rho was expressed evenly around the DV axis (Fig. 3), presumably as a result of the expansion of dorsal fate. In dl⁻ embryos expressing Dif (Fig. 3), the dorsal stripe of rho expression was similar to that seen in wild-type embryos. However the two ventrolateral stripes of rho expression were shifted ventrally and fused into a wide single stripe along the ventral surface. This ventral stripe exhibited decreased levels of expression along the ventral midline.

sim expression is first seen in wild-type embryos in the cellular blastoderm in two narrow lateral stripes of cells corresponding to the mesectoderm (Thomas et al., 1988) (Fig. 3). In embryos lacking Dorsal activity, sim is not expressed in ventral regions (Rushlow and Arora, 1990) (Fig. 3), consistent with the absence of cells corresponding to mesoderm, mesectoderm and neurectoderm. In contrast, in most dl⁻ embryos expressing Dif, sim expression was observed along the ventral midline. In many of these embryos sim expression was patchy and could not be detected as two independent stripes. However, in limited regions of some embryos expression was separated into lateral domains (Fig. 3, arrow).

In wild-type embryos, the sna and rho expression domains depend upon synergistic interactions between Dorsal and transcription factors of the bHLH type. Strong uniform expression within its domain and the establishment of sharp lateral borders of sna expression depend on interactions between Dorsal and Twist (Kosman et al., 1991; Ip et al., 1992b), while the the lateral extent of expression of rho is thought to be mediated through cooperation between Dorsal and bHLH proteins such as Scute/T4 (Rusch and Levine, 1996). Because expression of Dif in dl⁻ embryos led to aberrant patterns of sna and rho expression, we sought to determine whether interactions known to occur between the RH domain of Dorsal and certain bHLH proteins can also be detected between these bHLH proteins and Dif. It had been previously demonstrated (Gonzalez-Crespo and Levine, 1993; Shirokawa and Courey, 1997), using GST-fusion protein pull-down assays, that the amino-terminal 378 amino acids of Dorsal (the RH domain) co-precipitates in vitro with full-length Twist protein. For this reason we focussed our attention on studies of the interaction between Twist and the RH domains of Dorsal and Dif.

We employed components generated by Lehming et al. (1994, 1995), to develop a modified two-hybrid system (Fields and Song, 1989) capable of detecting interactions between Twist and the RH domains of Dorsal and Dif. Starting with an S. cerevisiae strain in which the E. coli lacZ gene is under the control of a Dorsal responsive promoter (Lehming et al., 1995), we first constructed Dorsal and Dif derivatives that lack their transcriptional activation domains and therefore fail to direct high levels of transcription of β-galactosidase. We then introduced Twist protein into these cells to determine whether this would lead to increased levels of lacZ expression via an interaction between Twist protein and either Dif or Dorsal. For the sake of convenience, the Twist protein was expressed in S. cerevisiae as an in-frame fusion to the bacterial lexA DNA binding domain. As seen in Table 1, the LexA protein sequences have no effect on expression of lacZ driven by the Dorsal-responsive promoter.

When a Dorsal derivative expressing truncated protein that lacks the C-terminal activation domain was expressed in yeast cells together with a control plasmid encoding the lexA DNA binding domain alone, β-galactosidase levels were barely detectable (Table 1). However, when the truncated Dorsal derivative was expressed together with the lexA-twi fusion, β-galactosidase expression increased dramatically. As neither the lexA-twi fusion nor the truncated Dorsal protein were independently capable of directing transcription from the Dorsal responsive promoter (Table 1), the increase in lacZ expression seen when they were co-expressed must have arisen through cooperation between the two proteins. A similar synergistic activation of a Dorsal-responsive promoter by Dorsal and Twist has been reported by other workers (Reach et al., 1996; Shirokawa and Courey, 1997) in Drosophila Schneider (SL2) tissue culture cells.

Surprisingly, in contrast to Dorsal, a derivative of Dif truncated immediately C-terminal to the RH domain was capable of directing moderate levels of β-galactosidase when expressed either on its own (data not shown) or together with the lexA DNA-binding domain control plasmid (Table 1). However, co-expression of this Dif derivative with Twist did not lead to any increase in β-galactosidase levels, suggesting that unlike Dorsal, Dif does not cooperate with Twist. The basis for gene activation by this derivative remains unclear. While it is possible that the Dif RH domain is itself capable of supporting low levels of gene activation, an alternative possibility is that some suppression of the stop codon inserted at amino acid 381 is occurring, resulting in low levels of full-length Dif protein that contains a C-terminal activation domain. Importantly, we have noted that truncated Dorsal derivatives that retain portions of the Dorsal activation domain direct levels of gene activation similar to those produced by the Dif derivative, but do demonstrate increased levels of transcription when expressed together with Twist. This indicates that even though the truncated Dif protein alone exhibits some activity, an increase in this activity through an interaction with Twist, if present, would have been detectable in this system.

We also attempted to determine whether Dif is capable of interacting with T4. Although T4 and Dif were not capable of cooperating to activate lacZ in the S. cerevisiae assay, we also were unable to detect an interaction between T4 and Dorsal. As it has previously been shown by Gonzalez-Crespo and Levine (1993) that Dorsal and T4 physically associate in vitro, it is currently unclear whether the situation is different in vivo or whether our assay is incapable of detecting the interaction between T4 and Dorsal. Because of this discrepancy, we cannot determine whether Dif and T4 interact using this system. However, our observations of the phenotype of embryos expressing Dif would suggest that it cannot.

The ability of Dif to restore a range of pattern elements along the DV axis, and to define an appropriately polarized gastrulation pattern, suggests that the Dif protein responds correctly to the signals that normally act to regulate Dorsal activity. Therefore, it is likely that in these embryos Dif, like Dorsal, forms a nuclear concentration gradient along the DV axis. We obtained from three laboratories antibodies directed against the Dif protein. In our hands, however, these antibodies did not detect Dif protein in whole-mount preparations of embryos that exhibited Dif-mediated rescue. Thus, to better document the regulation of Dif activity by the dorsal group, and to obtain additional evidence supporting the formation of a Dif nuclear gradient, we assessed the effects of either removing dorsal group signalling, or of reducing the gene dosage of cact.

When the p[w⁺, bcd-dif]6A transgene insert was crossed into females mutant for the dorsal group gene pipe, the embryos produced were completely dorsalized (Fig. 4). This indicates that Dif is not active in the absence of dorsal group signalling, supporting the suggestion that Dif is regulated by the same effectors that normally act upon Dorsal.

Dorsal activity is exquisitely sensitive to the dosage of Cactus (Roth et al., 1991; Govind et al., 1993), which suggested that it might be possible to observe differences in the phenotypic rescue of dl⁻ embryos expressing one copy of dif by reducing the cact gene copy number. Removing one copy of cact resulted in an overall relative ventralization, as can be seen in Table 2 (see also Fig. 5). The majority of embryos from females heterozygous for cact exhibited denticle belts, while females carrying two copies of wild-type cact produced a minority of embryos with this level of cuticular rescue. These observations suggest that Cactus regulates the nuclear localization of Dif.

The expression pattern of sim also provides molecular evidence for the increased ventralization of these embryos. Embryos from cact⁺ females with only one copy of the dif transgene expressed sim only rarely and in patches. However, in embryos from females heterozygous for cact, sim expression was often observed in a continuous or discontinuous ventral stripe (Fig. 5). Reduction of the cact gene dosage did not lead to detectable differences in the patterns of expression of the other genes examined (data not shown).

The presence of κB-type binding sites in the promoter regions of the cecropin and diptericin genes has led to the suggestion that controlled nuclear localization of an RH protein is involved in the activation of these genes (Hultmark, 1994). Its expression in the Drosophila fat body makes Dif a likely candidate to carry out this function. For this reason, we investigated whether maternally expressed Dif was capable of activating immune response genes in the embryo. We crossed the cec-lacZ (Engstrom et al., 1993) and dipt-lacZ (Reichhart et al., 1992) reporter genes into the background of embryos expressing Dif. These transgenes carry the cecropin and diptericin promoters driving lacZ. They are expressed in the fat body in response to bacterial injection of larvae (Engstrom et al., 1993; Reichhart et al., 1992), presumably through Dif activation. If Dif alone were sufficient to activate these promoters in the embryo, one would expect to see a lacZ expression pattern similar to that seen for twi. In fact, no expression of lacZ from these reporter genes was observed in dl⁻ embryos expressing Dif (data not shown). Therefore, if Dif is involved in activating these promoters in vivo, additional factors that are not expressed in embryos may also be required.

The Dif protein can, to a striking degree, substitute functionally for the Dorsal protein in the process of DV pattern formation. This restoration of DV polarity is evident from the gastrulation pattern, cuticular pattern elements and zygotic gene expression of embryos produced by females expressing Dif in the absence of Dorsal. These results suggest that both DNA binding and transcriptional activation as well as the regulation of nuclear localization have been conserved between Dorsal and Dif throughout the course of Drosophila evolution.

Several lines of evidence indicate that maternally expressed Dif is regulated by the same mechanism that controls the activity of Dorsal. First, the DV axis of embryos expressing Dif instead of Dorsal was appropriate with respect to the intrinsic orientation of the egg shell. Second, although we were not able to detect a protein gradient directly with antibodies directed against Dif, phenotypic evidence suggests that when Dif protein was expressed in the embryo its effects were graded along the DV axis. In addition, the rescue of dl− embryos by Dif required that the dorsal group signalling pathway be intact. Finally, Dif-mediated patterning was sensitive to the dosage of the cact gene.

Maternally expressed Dif, like Dorsal, is capable of effectively activating twi transcription, and of repressing expression of zen and dpp. The intrinsic activity of Dorsal is transcriptional activation (Pan and Courey, 1992; Jiang et al., 1992, 1993; Kirov et al., 1993), whereas Dorsal-mediated repression requires ancillary factors termed corepressors (for a review see Ip, 1995). Mutational analyses of promoters repressed by Dorsal have led to the identification of additional binding sites that are presumed to be occupied by these factors (Doyle et al., 1989; Huang et al., 1993). Recently, corepressors required for repression of the zen and dpp promoters by Dorsal have been identified (Lehming et al., 1994; Huang et al., 1995; Dubnicoff et al., 1997). The fact that Dif is capable of substituting for Dorsal in the repression of zen and dpp indicates that the Dif protein is capable of interacting effectively with these corepressors.

Although the domains of expression of Dorsal target genes in Dif-expressing embryos appear broadly similar, there are differences in the precise positioning of gene expression domains with respect to one another. While small differences in promoter affinity and gene activation by Dorsal and Dif are likely to contribute to these differences, our data also provide insight into more general properties of Dif-versus Dorsal-mediated gene regulation. The differences seen in the expression of sna, rho and sim suggest that different RH proteins may not be cabable of interchangeably cooperating with the same set of transcription factors. While both Dorsal and Dif mediate ventral activation of sna expression, the nature of the expression domain is different. In wild-type embryos at cellular blastoderm, sna is expressed in the most ventral region of the embryo in a band of 18 cells that exhibits a strong on/off pattern at the boundaries. Studies of the sna promoter have indicated that the strength and precision of the sna expression domain require appropriate levels of Dorsal as well as synergy between Dorsal and Twist (Ip et al., 1992b), which is also expressed in a ventral stripe. Although embryos expressing Dif transcribed high levels of sna ventrally, cellular blastoderm embryos never exhibited the strong on/off boundary seen in lateral regions of wild-type embryos. Rather, the sna expression pattern appeared graded with highest levels at the ventralmost point. Consistent with this observation, sna-mediated repression of rho only occurred in the ventralmost cells (where highest levels of sna were seen), and did not extend to the lateral boundaries of the sna expression domain (see also below). The aberrant sna expression pattern observed in Dif-expressing embryos could be explained by a mechanism in which Dif is simply a less efficient activator of sna than Dorsal. However, our observation that Dif and Twist do not interact in an S. cerevisiae system that does allow detection of the Dorsal/Twist interaction leads us instead to favor the notion that Dif is incapable of synergizing with Twist and that this lack of cooperation leads to reduced and imprecise expression of sna expression relative to wild-type.

Dif’s inability to interact with bHLH proteins and the resulting alteration in sna expression may also explain the deviations from the wild-type pattern that we observed in the transcription of rho and sim in Dif-expressing embryos. Two distinct aspects of Dif’s inability to synergize with bHLH proteins may explain the misregulation of the ventrolateral domain of rho expression. First, since the lateral extent of the rho expression domain depends on Dorsal’s interaction with bHLH proteins such as T4, Dif’s apparent failure to cooperate with bHLH proteins such as T4 may be responsible for the narrowing of the rho expression domain. Second, in wild-type embryos, the broad ventrolateral domain of rho activation is resolved into two lateral stripes (Bier et al., 1990; Ip et al., 1992a) by Snail-mediated repression ventrally. In Dif-expressing embryos, the inability of Dif to interact with Twist to accomplish high levels of Snail expression may have resulted in inefficient ventral repression of rho.

Like rho, the sim gene is expressed in the late cellular blastoderm embryo in two lateral stripes of cells (Thomas et al., 1988). Although the transcriptional regulation of sim has not been extensively characterized, the presence of Snail binding sites in the 2.8 kb transcriptional regulatory region (Kasai et al., 1992) and the changes in sim expression seen in sna mutant embryos (Hemavathy et al., 1997) suggest that it, like rho, may be activated in a ventral band by a mechanism involving synergy between Dorsal and bHLH proteins, and repressed by sna in an overlapping but slightly more narrow ventral region. Consistent with this proposal, we observed that in Dif-expressing embryos, sim transcription, when detected, was seen mainly along the ventrolateral midline and that its lateral boundaries were narrow in comparison to wild type.

We propose that the Dorsal and Dif proteins differ in their abilities to physically interact with bHLH proteins in the control of transcription. Dorsal has been shown to be capable of interacting directly with the bHLH proteins Twist and Scute (T4) (Gonzalez-Crespo and Levine, 1992; Jiang and Levine, 1993). Studies of the transcription of various Dorsal target genes indicate that synergy between Dorsal protein and bHLH proteins occurs in two distinct modes (Szymanski and Levine, 1995). In one mode, sharp on/off patterns of transcription in the mesoderm do not require that the Dorsal and bHLH binding sites on the promoter be adjacent. In such promoters (e.g. sna), synergy is thought to occur when Dorsal and Twist interact with different components of the transcriptional machinery (Ip et al., 1992a; Szymanski and Levine, 1995). Although this mode of synergy would not require a physical interaction between Dorsal and Twist, direct interactions between Dorsal and Twist have been observed in vitro (Gonzalez-Crespo and Levine, 1993; Shirokawa and Courey, 1997), and we demonstrate here that Dorsal and Twist are capable of associating in yeast cells. Moreover, Shirokawa and Courey (1997) have demonstrated that the same domains of Dorsal and Twist that mediate protein-protein interactions also cooperate in synergistic activation of transcription in transfected tissue culture cells and in a cell-free in vitro transcription system, and that this synergistic activation of transcription did not seem to involve cooperative binding to target DNA. Thus, while the mechanism by which the Dorsal/Twist interaction yields transcriptional activation is unclear, the data presented here, as well as that of others, indicate that this synergistic interaction requires direct protein-protein contact.

The second mode of Dorsal/bHLH interaction requires that the Dorsal and bHLH binding sites be linked. In such cases, represented by the rho (and perhaps sim) promoter, cooperative binding of the promoter by Dorsal and bHLH proteins such as Scute/T4 results in promoter occupancy by Dorsal even in lateral regions of the embryo where nuclear concentrations of Dorsal are low (Szymanski and Levine, 1995). Our proposal that Dif is incapable of appropriately interacting with bHLH proteins provides an explanation for the ventral shift in rho and sim expression that is observed in Dif-expressing embryos. In the absence of cooperative interactions facilitating promoter occupancy by Dif in lateral regions of the embryo, expression would be limited to more ventral regions where there are sufficient concentrations of Dif to allow binding to low affinity sites. Taken together, these data suggest that one level of specificity in RH protein action is defined by the protein-protein contacts that they are capable of forming with other transcription factors.

The normal function of Dif is thought to be the activation of promoters involved in the fly immune system. If Dif acting alone is capable of directing expression of these promoters, we would have expected that Dif-expressing embryos carrying diptericin-lacZ or cecropin-lacZ transgenes would exhibit ventral expression of β-galactosidase. However, no lacZ expression was detected in these embryos, indicating that Dif alone is insufficient to activate these promoters. Consistent with this finding, in flies carrying Toll gain-of-function alleles, even though both Dif and Dorsal are constitutively nuclear in the fat body cells, neither the diptericin nor the cecropin genes are constitutively active (Lemaitre et al., 1995). These findings imply that additional factors may be required to cooperate with Dif to accomplish activation of these genes. Consistent with this notion, the diptericin promoter also contains functional sequence motifs that exhibit homology to NF-IL6 sites and the GAAANN sequence present in the interferon response elements of some mammalian interferon-induced genes (Meister et al., 1994; Georgel et al., 1995).

Given the similarities between the mechanisms mediating Dorsal and NFκB activation in insects and mammals respectively (Wasserman, 1992), the identification of Dif as an RH protein suggested that a homologous pathway involving components of the IL-1/Toll signal transduction system may regulate gene activation in the fly immune system. Additional evidence for this hypothesis was provided by Lemaitre et al. (1996), who showed that four components of the embryonic dorsal-ventral pathway, spätzle, Toll, tube and pelle, but not dorsal, are required for expression of the antifungal peptide Drosomycin in response to immune challenge. In addition, these genes influence the expression of the antibacterial agents Cecropin A, Attacin and Defensin. Our demonstration that maternally expressed Dif is regulated by Cactus and the dorsal group is consistent with the idea that Dif is regulated by this pathway in the immune system. Our finding also lends in vivo support to the observation that Dif and Cactus interact physically in vitro (Tatei and Levine, 1995).

Much emphasis has been placed on the finding that Dorsal activity is present in a gradient in the embryo. A priori, there is currently no reason to suspect that activation of immune function genes in the fly would also be graded. However, when expressed in embryos, a gradient of Dif activity was detected. We (Bergmann et al., 1996) and others (Reach et al., 1996), have recently detected a cytoplasmic Cactus gradient that is complementary in orientation to the nuclear gradient of Dorsal and whose formation appears be a key event in the process regulating nuclear localization of Dorsal. Our demonstration that Cactus is capable of regulating Dif function suggests that Cactus activity is responsible for the polarized activity of Dif expressed in embryos. Alternative models for the regulation of Dorsal suggest that dorsal group signalling acts directly through modification of Dorsal protein (Norris and Manley, 1992; Whalen and Steward, 1993; Gillespie and Wasserman, 1994). However, the substantial amino acid sequence divergence between Dif and Dorsal (Ip et al., 1993) suggests that potential target sites for modification, which are present in the Dorsal protein and required for regulation of its nuclear localization, would not necessarily be conserved in Dif. Our results with Dif support the model that dorsal group signalling is mediated mainly through Cactus.

Because both insects and vertebrates employ RH proteins in the regulation of their immune systems, but only insects appear to utilize this signalling pathway in DV pattern formation, it has been proposed that the immunological role of this class of proteins preceded its pattern-forming function during evolution (Hultmark, 1994). It is interesting to note that over the course of evolution, the Dorsal protein seems to have evolved new capabilities that its immune system ancestor did not have, such as the ability to interact with developmentally important bHLH proteins. Further analyses will undoubtably illuminate additional capabilities, some shared and some unique, of this fascinating group of proteins.

We wish to thank the following people for their contributions of DNA clones, antibodies, yeast strains and/or Drosophila strains: Al Courey, Ylva Engstrom, Y. Tony Ip, Norbert Lehming, Michael Levine, Robert Ray, Jean-Marc Reichhart, Siegfried Roth, Chris Rushlow and Trudi Schüpbach. We thank anonymous reviewers for comments on an earlier version of this manuscript. Supported by grants from the ACS (D. S.) and the NIH (L. S.).