A processed form of the Spätzle protein defines dorsal-ventral polarity in the Drosophila embryo

Extracellular morphogens are molecules distributed in a gradient across a field of cells that elicit different cell fates as a function of their concentration. In biochemical terms, an extracellular morphogen would be a ligand for a membrane receptor and graded receptor activation would lead to different cellular responses. The initial establishment of dorsal-ventral polarity in the Drosophila embryo appears to rely on such an extracellular morphogen gradient defined by a set of maternal effect genes, although the identity of the morphogen and the mechanism of gradient formation have not yet been determined.

Embryonic dorsal-ventral asymmetry is triggered by ventral activation of the transmembrane protein encoded by the maternally transcribed Toll gene. The consequence of ventral activation of Toll is the production of a ventral-to-dorsal concentration gradient of Dorsal protein in the nuclei of the blastoderm embryo (Roth et al., 1989; Rushlow et al., 1989; Steward, 1989). Seven other maternal effect genes, pipe, nudel, windbeutel, gastrulation defective, snake, easter and spätzle, act upstream of Toll and are required for the production of an extracellular signal that activates Toll (Anderson et al., 1985; Hashimoto et al., 1988, 1991; Stein et al., 1991; Stein and Nüsslein-Volhard, 1992). The Toll protein is distributed uniformly around the embryonic circumference (Hashimoto et al., 1991), suggesting that its activity is confined to the ventral side of the embryo by ventral localization of the ligand that activates Toll.

In a series of elegant experiments in which the extracellular fluid that lies between the plasma membrane of the syncytial blastoderm embryo and the egg shell (the perivitelline fluid) was transferred from one embryo to another, Stein et al. (1991) identified a soluble factor that has the properties predicted for a Toll ligand. When this factor, called polarizing activity, was injected into the perivitelline space of a recipient embryo, it defined the polarity of the dorsal-ventral pattern, with the most ventral cell types developing at the injection site. Thus this polarizing activity activated the Toll-dependent nuclear translocation of Dorsal near the site of injection. The polarizing activity was detectable only in the perivitelline fluid of embryos that lacked Toll, and could not be retrieved from embryos with wild-type Toll. This suggested that the polarizing activity might be bound to Toll, when Toll is present. Based on these data, Stein et al. (1991) suggested that the polarizing activity was a ligand that binds to and activates Toll at the site of injection.

In this paper, we use a biochemical approach to characterize the polarizing activity. By measuring the phenotypic response to material injected into the perivitelline space as a bioassay, we found that it was possible to recover polarizing activity from the supernatant of wild-type embryo extracts boiled at pH 4.5. From this starting material, we partially purified a 24×10³M_r protein that has polarizing activity. This protein is recognized by antibodies to the C-terminal half of the Spätzle, indicating the spätzle gene encodes the polarizing activity. Polarizing activity cannot be extracted from embryos laid by females that lack spätzle, but is present in the lowpH/boiled supernatant of mutant embryo extracts that lack other genes that act upstream of Toll. We present evidence that the low-pH/boiling treatment partially proteolyzes Spätzle, thereby bypassing the genes that are normally required for processing of Spätzle from an inactive precursor to an active product. We conclude that a processed form of Spätzle defines embryonic dorsal-ventral polarity by activating Toll on the ventral side of the embryo.

Toll− perivitelline extracts were prepared by vortexing 0-12 hour dechorionated embryos produced by Toll⁻ females [Df(3R)ro^XB3/Df(3R)Tl^9QRX] in buffer (30 mM NaCl, 50 mM Hepes, pH 7.5, 1 mM PMSF) with silicon carbide powder (grain size 1200, VWR) for one minute in 20 mM Hepes, pH 7.5, 30 mM NaCl. RNAase A bound to Sepharose (Boehringer) was used to test RNAase sensitivity. Protease sensitivity was assayed by treating extracts with 2.5 mg/ml proteinase K for 30 minutes at 37°C; proteinase K was inactivated by boiling. Supernatant fractions were concentrated for injection using a Centricon-10.

Polarizing activity from wild-type embryos was prepared by douncing 0-4 hour embryos with a type A pestle until it moved freely, in five volumes of 20 mM sodium acetate pH 4.5, 30 mM NaCl. Extracts were clarified by centrifugation for 5 minutes at 3000 g at 4°C and the supernatants were placed in boiling water for 5 minutes and then immediately placed in ice. Boiled extracts were cleared at 10,000 g for 10 minutes. Extracts were prepared for injection either by replacing the acidic buffer with 20 mM Hepes pH 7.5, 30 mM NaCl using a Centricon 10 microconcentrator or by precipitation with 20% TCA at 4°C followed by an acetone wash and resuspension in 20 mM Hepes pH 7.5, 30 mM NaCl.

We used embryos laid by pip⁶⁶⁴/pip³⁸⁶ females as recipients for perivitelline injections. Polarizing activity was injected into the perivitelline space of dechorionated syncytial blastoderm (stage 4a or 4b) embryos at 50% egg length on the flat (dorsal) side of the embryo. Samples were assayed at a series of two-fold dilutions, until the greatest dilution that could still induce an asymmetric head fold was determined. We defined 1 unit (u)/ml of polarizing activity as the concentration that led 50% of injected pipe⁻ embryos to develop an asymmetric head fold. At least 15 embryos at each concentration of polarizing activity for each sample and column fraction were scored to determine the activity of the sample.

Cuticle preparations were made as described (Wieschaus and Nüsslein-Volhard, 1986). Embryos were assayed for expression of Twist protein with a rabbit anti-Twist antibody (Roth et al., 1989; the kind gift of S. Roth); injected embryos were fixed as described by Stein et al. (1991).

30 g of 0–4 hour thawed frozen Oregon R embryos were homogenized in 5 volumes of 20 mM acetate pH 4.5, 30 mM NaCl, 1 mM PMSF on ice. Extracts were clarified by centrifugation for 10 minutes at 3000 g at 4°C. 10 ml aliquots of the supernatant in 15 ml Falcon tubes (Fisher) were placed in boiling water for 5 minutes and were then immediately cooled to 4°C on ice and then cleared by centrifugation at 10,000 g for 10 minutes. Protein was precipitated by adjusting extracts to an ammonium sulfate concentration of 95%. Precipitated material was collected by centrifugation at 10,000 g for 10 minutes. The particulate matter was resuspended in 5 ml of 20 mM acetate pH 4.5, 30 mM NaCl (Buffer A) and dialyzed overnight at 4°C against 4 liters of the same buffer. The dialysate was clarified by centrifugation at 10,000 g for 10 minutes. The supernatant was adjusted to 25% ammonium sulfate at 4°C and cleared by centrifugation at 10,000 g for 10 minutes. The supernatant was brought to 40% ammonium sulfate concentration at 4°C and the precipitate was collected by centrifugation at 10,000 g for 10 minutes and resuspended in 1 ml of 20 mM Hepes pH 7.5, 30 mM NaCl (Buffer B). This resuspended material was cleared by centrifugation for 5 minutes at 10,000 g. The supernatant was loaded onto a 1.6×68 cm Ultragel AcA 54 sizing column equilibrated with Buffer B and eluted at a flow rate of 12 ml/hour with the same buffer at 4°C. Undiluted fractions were assayed for polarizing activity by injection into pipe⁻ embryos as described above. Pooled active fractions were loaded onto a HiTrap Blue column (Cibacron Blue 3GA Sepharose) at 1.2 ml/hour, equilibrated with Buffer B, washed with 10 ml of buffer B at 6 ml/hour and eluted with 20 mM Hepes pH 7.5, 1 M NaCl. To prepare fractions for injections, each aliquot was adjusted to 20% TCA in the presence of 1 mg of BSA for carrier. Samples were incubated on ice for at least one hour. Precipitated material was washed once with −20°C acetone, resuspended in Buffer B plus 0.3 mg/ml BSA and stored on ice. Protein from active fractions was precipitated by the addition of TCA to 20%. Acetone-washed pellets were resuspended in Laemmli sample buffer without a reducing agent. Samples were run on 15% acrylamide mini gels. The entire lane was divided into 2 mm gel slices which were crushed and then eluted in 0.2 ml of Buffer B plus 0.1% SDS overnight at 4°C. All gel slices were assayed for activity. Protein was prepared for injection by acetone precipitation by adding 5 volumes of −20°C acetone and incubating −20°C for 2 hours. Acetone washed pellets were resuspended in Buffer B.

Western blots were performed as described (Morisato and Anderson, 1994).

The original source of the soluble extracellular component that could activate the Toll pathway (the polarizing activity) was perivitelline fluid from embryos laid by Toll⁻ females (Stein et al., 1991). In those experiments, nanoliter drops of the fluid were recovered with a micropipette after pricking the vitelline membrane of the embryos (Stein et al., 1991). Any biochemical characterization of the active component required a larger source of material. To isolate more perivitelline fluid, we used a procedure that enriched for perivitelline contents by vortexing dechorionated embryos in buffer with silicon carbide particles (Chasan et al., 1992; Materials and Methods).

Perivitelline extracts were made from embryos laid by Toll⁻ mothers and assayed for biological activity. We used the dorsalized embryos laid by pipe⁻ mothers as recipients in this assay, because it has been shown that injection of polarizing activity into these dorsalized embryos leads to the production of ventral structures at the injection site (Stein et al., 1991). Furthermore, the dorsalized pipe⁻ phenotype cannot be rescued by wild-type RNA, cytoplasm or perivitelline fluid injected into the embryo (Anderson and Nüsslein-Volhard, 1984; Stein and Nüsslein-Volhard, 1992), ruling out the possibility of characterizing a pipe rescuing activity rather than polarizing activity. When the Toll⁻ perivitelline extract was injected into the perivitelline space of pipe⁻ embryos, the embryos showed a partial restoration of dorsal-ventral pattern elements, with the most ventral pattern elements differentiating near the injection site. Thus the Toll⁻ perivitelline extracts, like Toll⁻ perivitelline fluid, defined the polarity of the embryonic pattern. The polarizing activity present in the Toll⁻ perivitelline extracts was stable to RNAase A, but was destroyed by treatment with proteinase K, indicating that the active component was a protein. The polarizing activity was very stable: its biological activity was not destroyed by heating to 100°C at either pH 7.0 or pH 4.5 (data not shown). The stability of the polarizing activity was useful in the strategy that we developed to purify the protein.

Perivitelline extracts from Toll⁻ embryos did not provide enough material for more extensive biochemical characterization of the polarizing activity. Due to female sterility, Toll⁻ flies cannot be grown as a homozygous population, and therefore must be generated by crosses. In addition, Toll has an important zygotic function and 95% of the Toll⁻ zygotes expected in a cross die during development (Gerttula et al., 1988). We therefore tried to recover polarizing activity from wild-type embryos.

Stein et al. (1991) proposed that polarizing activity was detectable in perivitelline fluid only in Toll⁻ embryos because the polarizing activity would be bound to the Toll protein in wild-type embryos. By this logic, polarizing activity should be present in wild-type embryos, but in a tightly bound complex. Because the polarizing activity in Toll⁻ perivitelline extracts was stable to boiling at pH 4.5, we hoped that these harsh conditions would release any polarizing activity present in wildtype embryos from a complex. Wild-type embryo homogenates were heated to 100°C at pH 4.5. Although 99.5% of embryonic proteins were precipitated by boiling at pH 4.5, the supernatant contained polarizing activity: concentrated supernatant fractions induced ventral structures at the site of injection when injected into the perivitelline space of pipe⁻ embryos (Fig. 1). To test whether the polarizing activity isolated from wildtype embryos by this procedure activated Toll, we assayed which dorsal group genes were required to respond to the preparations from wild-type embryos (Table 1). As Stein found for the polarizing activity in perivitelline fluid, the polarizing activity in low pH/boiled wild-type embryo extracts induced the formation of ventral structures in embryos lacking the dorsal group genes that act upstream of Toll (nudel, pipe, gastrulation defective, easter and spätzle), but had no activity in Toll⁻ embryos. Therefore, like Stein’s activity, the wild-type polarizing activity worked by activating Toll and bypassed the requirement for all of the genes upstream of Toll. Thus, the low-pH/boiled supernatant fraction of wild-type embryo extracts provided an abundant source of material for further characterization of polarizing activity.

With an extract containing polarizing activity in hand, it became possible to determine a dose-response relationship for the activity. We injected extracts at various concentrations into the perivitelline space of dorsalized pipe⁻ embryos and assayed the effects on the pattern of gastrulation, the cuticle and the expression of the ventral marker protein Twist.

Different concentrations of the polarizing activity preparation induced the formation of a series of progressively more ventral cell types over a 150-fold range in concentration (Fig. 2). In all cases, the rescued pattern was dorsoventrally asymmetric, with the ventral-most structures appearing during gastrulation nearest the site where the polarizing activity was deposited. Uninjected pipe⁻ embryos were completely dorsalized both at gastrulation and in the differentiated cuticle. The first detectable response to injection of low concentrations of polarizing activity into pipe⁻ embryos was the induction of an asymmetric head fold during gastrulation, with a head fold (a lateral pattern element) appearing on the side near the injection site (Figs 1, 2). At progressively higher concentrations, first dorsolateral, then ventrolateral and finally ventral structures were induced. High concentrations of polarizing activity led to the formation of the most ventral pattern element, the mesoderm, as shown by the invagination of a ventral furrow at the site of injection and the expression of the mesodermspecific protein Twist (Fig. 3). The production of mesoderm in 50% of the injected embryos required polarizing activity that was 150-fold more concentrated than the concentration that produced the weakest detectable response, the production of a head fold, in 50% of the injected embryos (Fig. 2). Thus, large differences in the concentration of polarizing activity are required to produce the full range of dorsal-ventral pattern elements.

Based on this dose-response relationship, the most sensitive assay for polarizing activity was the induction of an asymmetric head fold at the time of gastrulation. This assay was also rapid, because gastrulation occurs 2 hours after injection. In addition, this assay was sensitive to small differences in polarizing activity concentrations. We therefore used the morphology of the head fold to follow the polarizing activity during purification.

Using the pattern of gastrulation as a bioassay, we used standard biochemical procedures to purify the polarizing activity. The starting material for the purification was the supernatant from 0-4 hour wild-type embryo extracts heated to 100°C at pH 4.5. Ammonium sulfate precipitation, gel filtration chromatography, dye ligand chromatography and elution from an SDS-polyacrylamide gel were then used as successive steps to purify the polarizing activity (Table 2). At each step, fractions were injected into pipe⁻ embryos and the pattern of gastrulation was monitored in order to identify the peak of activity. The peak fractions were pooled and used as the starting material for the next purification step. Based on the migration of the active fraction on non-reducing gels, the purified polarizing activity was about 24×10³M_r in size. The final fraction was enriched more than 1000-fold for polarizing activity relative to the initial low-pH/boiled supernatant. However, despite this enrichment, there was not enough of the purified protein to detect on a silver-stained gel, suggesting that the protein was very rare in embryos. Because we did not have enough material to determine the N-terminal sequence of the protein, we asked whether the purified protein was encoded by a known dorsal group gene.

To determine which gene products were required for the production of polarizing activity, we tested whether polarizing activity was detectable in embryos laid by mothers homozygous for dorsal group mutations. Extracts were made from embryos laid by mutant mothers and the supernatant after heating to 100°C at pH 4.5 was assayed for the ability to induce ventral structures (Table 3). Embryos from spätzle mutant females did not contain detectable polarizing activity, indicating that spätzle was required for the presence of polarizing activity. However, low-pH/heat-treated extracts from embryos from easter or pipe mutant females did have polarizing activity. This was surprising not only because extracts from pipe mutants rescued the pipe phenotype, but also because Stein et al. (1991) showed that all of the genes acting upstream of Toll were required for the presence of polarizing activity in perivitelline fluid. We conclude that the low-pH/heat treatment bypassed the requirements for easter and pipe for the production of polarizing activity and that the treatment produced polarizing activity in a nonphysiological way.

Genetic experiments had indicated that, of the seven genes that act upstream of Toll, easter and spätzle are most directly required to activate Toll (Chasan et al., 1992; Morisato and Anderson, 1994). Because in our assay spätzle, but not easter, was required for the presence of polarizing activity, we conclude that Spätzle acts downstream of Easter in the pathway that leads to Toll activation. Thus the spätzle gene is the only known dorsal group gene that could encode the polarizing activity.

We tested whether polarizing activity was encoded by spätzle using antibodies raised against fusion proteins containing the spätzle open reading frame (Morisato and Anderson, 1994). In extracts prepared at pH 7.5 from wild-type embryos, there is a family of proteins encoded by Spätzle that migrates at 45 to 60×10³M_r under reducing conditions (Morisato and Anderson, 1994). In embryo extracts boiled at pH 4.5, the Spätzle protein migrated at 14 to 18×10³M_r under reducing conditions and at approximately 24×10³M_r under non-reducing conditions. Thus low-pH/boiling apparently caused partial proteolysis of Spätzle. The protein recognized by Spätzle antibodies was present in the low-pH/heated extracts from wild type, but not spätzle⁻, embryos, indicating that the low molecular weight protein is a product of spätzle. Because low pH-boiled material had polarizing activity but whole embryo extracts did not, we propose that low-pH/boiling converts an inactive form of Spätzle into an active, cleaved product.

The protein recognized by the Spätzle antibodies copurified with the polarizing activity measured by the bioassay through the gel filtration, dye ligand chromatography and polyacrylamide gel steps (Fig. 4A,B). This Spätzle protein was highly enriched in the final steps of the purification, as seen when equal amounts of total protein were loaded on a gel and immuno-blotted with the Spätzle antibodies (Fig. 4B). Because the protein recognized by the Spätzle antibodies copurified exactly with the polarizing activity, we conclude that the polarizing activity is a product of the Spätzle gene.

The purified Spätzle protein that had the polarizing activity was smaller than the 45-60×10³M_r primary translation products of Spätzle (Morisato and Anderson, 1994). Antibodies directed against the C-terminal, but not the N-terminal half, of the spätzle open reading frame recognized the polarizing activity (data not shown), suggesting the polarizing activity included C-terminal, but not N-terminal, portions of the primary Spätzle translation product.

We have used a biochemical approach to characterize an extracellular factor that defines dorsal-ventral polarity in the Drosophila embryo. The factor that we identified biochemically is encoded by the spätzle gene, which by genetic criteria acts immediately upstream of the receptor encoded by Toll (Morisato and Anderson, 1994). Thus, biochemistry and genetics have converged to define a single extracellular molecule that activates Toll on the ventral side of the embryo and thereby defines the polarity of the embryo.

Using a bioassay developed by Stein et al. (1991), we partially purified a protein from wild-type embryos extracts that induces an organized dorsal-ventral pattern in the Drosophila embryo.Like the activity defined by Stein et al. (1991), the purified polarizing activity induces ventral structures at the site of injection and produces a set of lateral and dorsolateral pattern elements in their normal spatial order. Of the genes that act upstream of the receptor Toll, spätzle, but not pipe or easter, is required for the presence of polarizing activity in lowpH/boiled extracts, suggesting that spätzle encodes the polarizing activity. A 24×10³M_r form of Spätzle protein copurifies with the biological activity that induces ventral structures throughout the entire purification procedure. We conclude that a form of the Spätzle protein is the extracellular molecule that induces dorsal-ventral polarity when injected into the extracellular space.

The form of Spätzle that we purified is smaller than the primary translation products of the spätzle gene (Morisato and Anderson, 1994) and appears to represent a proteolytic product of full-length Spätzle protein. The Spätzle protein in wild-type embryos appears to undergo proteolytic processing to generate the active form of the protein (Morisato and Anderson, 1994). The proteolytic processing of Spätzle in wild-type embryos appears to produce a form of the protein in which the Nterminal and C-terminal halves of the cleaved product remained disulfide-bonded together (Morisato and Anderson, 1994). In contrast, the low-pH/boiled form of Spätzle that we purified appears to be a C-terminal domain that is not disulfidebonded to the N terminus of the protein. The polarizing activity of the isolated C-terminal domain indicates that the protein sequences required for activation of Toll are localized in the C-terminal portion of the precursor molecule.

We were fortunate because heating embryo extracts at pH 4.5 processes Spätzle in vitro in a way that converts the fulllength protein to a protein with polarizing activity. While we had hoped to liberate the polarizing activity from a complex with Toll in wild-type embryos by the low pH and heat treatment, instead we generated polarizing activity, apparently by inducing a proteolytic cleavage event. Although most peptide bonds are stable under these conditions, some peptide bonds, aspartate-proline, asparagine-glutamate and asparagineglycine, hydrolyze at high temperature and low pH (Landon, 1977; Voorter et al., 1988; Martin et al., 1990); however, Spätzle does not contain any of these dipeptides. Therefore we infer that the proteolysis of Spätzle is either an unusual acid hydrolysis event or that unfolding of the protein at high temperature and low pH renders it accessible to a protease present in the extracts. In either case, the biological activity of the material suggests that the low-pH/boiling treatment creates a product similar to that produced by the normal proteolytic processing of Spätzle, perhaps because a region of the protein is exposed or strained and therefore particularly labile.

In order to induce an asymmetric dorsal-ventral pattern, the form of Spätzle that we purified must remain localized following injection. We assume that Spätzle binds to the plasma membrane near the site of injection and, as a consequence of Spätzle binding, Toll is activated at that position in the embryo. The simplest way to account for these findings is to infer that processed Spätzle is the ligand that binds directly to Toll and that triggers Toll to relay a signal to the cytoplasm. The available genetic and biochemical data indicate that spätzle acts immediately upstream of Toll and no other genes are known that act at that step in the pathway, which is also consistent with the hypothesis that processed Spätzle is the Toll ligand. Now that both the receptor and its putative ligand have been cloned, it will be possible to test directly whether processed Spätzle binds to and activates Toll.

Whether or not processed Spätzle acts as a classic ligand for Toll, our experiments show that processed Spätzle is the limiting component that determines where Toll is active. The doseresponse relationships that we observed show that different concentrations of processed Spätzle induce different cell types. Therefore it seems likely that processed Spätzle is distributed in a gradient in the perivitelline space and that different concentrations of processed Spätzle lead to a gradient of Toll activity which leads to the nuclear gradient of Dorsal protein.

The dose-response curve for the processed form of Spätzle that we purified is rather shallow: a 150-fold higher concentration of processed Spätzle is required to induce the most ventral pattern element, the mesoderm, than is required to produce a detectable dorsolateral cell fate. Thus if Spätzle acts as a graded morphogen to organize the dorsal-ventral embryonic pattern, its concentration gradient would be very steep, with processed Spätzle two orders of magnitude more concentrated on the ventral midline than at the lateral positions. This suggests that the production of a gradient of processed Spätzle requires stringent mechanisms to insure a high ventral concentration of the processed protein without that protein spreading to more lateral positions.

We thank Sylvia Sanders, David King and Tohru Yoshihisa for advice with protein purification procedures. We thank Siegfried Roth for the kind gift of Twist antibodies. We also thank Lisa Molz, Sylvia Sanders and the members of the Anderson lab for reading this work. D. M. was a Fellow of the Miller Institute for Basic Research in Science and the Dupont Fellow of the Life Sciences Research Foundation. The work was supported by National Institutes of Health grant GM 35437 and National Science Foundation Faculty Awards for Women DCB-9023672 to K. V. A.