Argos Mutants Define an Affinity Threshold for Spitz Inhibition in Vivo*

Argos, a secreted antagonist of Drosophila epidermal growth factor receptor (dEGFR) signaling, acts by sequestering the activating ligand Spitz. To understand how different domains in Argos contribute to efficient Spitz sequestration, we performed a genetic screen aimed at uncovering modifiers of an Argos misexpression phenotype in the developing eye. We identified a series of suppressors mapping to the Argos transgene that affect its activity in multiple developmental contexts. These point mutations map to both the N- and C-terminal cysteine-rich regions, implicating both domains in Argos function. We show by surface plasmon resonance that these Argos mutants are deficient in their ability to bind Spitz in vitro. Our data indicate that a mere ∼2-fold decrease in KD is sufficient to compromise Argos activity in vivo. This effect could be recapitulated in a cell-based assay, where a higher molar concentration of mutant Argos was needed to inhibit Spitz-dependent dEGFR phosphorylation. In contrast, a ∼37-fold decrease in the binding constant nearly abolishes Argos activity in vivo and in cellular assays. In agreement with previously reported computational studies, our results define an affinity threshold for optimal Argos inhibition of dEGFR signaling during development.

The epidermal growth factor receptor belongs to a family of receptor tyrosine kinases that are well conserved from lower metazoans to humans (1,2). In humans, mutations or genetic alterations that alter receptor activity have been correlated with cancer progression and poor clinical outcome, validating the ErbB family as a target for therapeutic agents (3)(4)(5)(6)(7).
In Drosophila melanogaster, EGF 5 receptor signaling is utilized reiteratively throughout development to mediate a wide array of cellular decisions (8). Remarkably, this versatility is accomplished with a single receptor (dEGFR) and four activating ligands: Spitz (Spi), Gurken, Keren, and Vein. The functional diversity of dEGFR signaling has been partially attributed to the differential use of ligands throughout development (8). For instance, the ligand Gurken is produced exclusively in the germline to specify eggshell structures and the embryonic axes (9,10). In contrast, Spi and Vein participate in numerous processes, sometimes sharing additive roles (such as in ventral ectoderm patterning) (11)(12)(13)(14) and in some instances acting as the main dEGFR ligand (such as in photoreceptor recruitment or wing vein differentiation, respectively) (15)(16)(17)(18)(19)(20). Interestingly, no knock-out phenotype or expression pattern has been reported for the fourth ligand, Keren (21). In addition to the four agonists, the Drosophila EGF receptor signaling system includes two extracellular inhibitors, which function in negative feedback loops to antagonize dEGFR signaling. Kekkon 1 is a transmembrane molecule of the leucine-rich repeat-immunoglobulin (LIG) superfamily that attenuates dEGFR activity via a direct interaction (22,23). Argos (Aos) is a secreted molecule that was initially proposed to bind and inhibit dEGFR by virtue of its atypical EGF domain (24). Recent work, however, has demonstrated that Aos instead exerts its antagonistic effect on dEGFR signaling by sequestering the activating ligand Spi, although its effect on the remaining three ligands has not yet been reported (25). Aos was also shown to associate with the surface of cultured S2 cells in an interaction that is dissociable with excess soluble heparin (25), suggesting a potential regulatory mechanism for Aos activity in vivo. However, the functional and physiological significance of this finding remains unclear.
Structurally, Aos is composed of N-and C-terminal cysteinerich regions (NCR and CCR, respectively), separated by a largely unconserved linker. The NCR, which contains 4 cysteines, has no known direct function, and misexpression of this domain alone displays no biological activity. The CCR includes 12 cysteines (see Fig. 1B) and contains a putative EGF-like domain (residues 363-424) (24). Misexpression of the entire CCR from Aos (residues 225-444) displays partial activity in vivo and is sufficient for binding Spi in vitro (albeit with a ϳ20fold decreased affinity) (25,26). In contrast, misexpression of the putative EGF-like domain alone does not rescue Aos mutant phenotypes (26), arguing that (if it does adopt an EGFlike fold) it is not sufficient for Aos function.
Aos participates in multiple developmental processes where it is expressed as a "high threshold" gene, in response to high levels of Spi-induced dEGFR signaling (27,28). Consistent with its role as an inhibitor of dEGFR signaling, Aos knock-outs exhibit phenotypes typical of dEGFR gain-of-function mutants, and Aos misexpression inhibits dEGFR signaling (26,29). Aos contributes importantly to the spatio-temporal regulation of dEGFR signaling through its participation in a negative feedback loop as a result of Spi-dependent dEGFR signaling. For example, Aos is required for the proper timing of pulsations during oenocyte delamination (30). Aos has also been described as a long range inhibitor during eye development (and other tissues), acting to ensure the formation of steep Spi gradients close to the source of Spi production (31). Recent computational studies have proposed two key roles for Aos in dEGFR signaling based on a model of the dEGFR/Spi/Aos module in embryonic ventral ectoderm patterning (32). First, sequestration by Aos limits the spatial range of Spi action. Second, the Aos negative feedback loop counteracts fluctuations in gene dosage (Spi secretion rate and dEGFR levels), imparting robustness to the system (32). One important prediction of the model was that Spi sequestration by Aos must be nearly irreversible (or of very high affinity) to provide a robust feedback loop. As such, an increase in the off rate would result in a loss of robustness, although this remains to be tested experimentally.
To test the prediction that even small reductions in Spi/Aos affinity cannot be tolerated and to investigate the domain requirements for Aos activity, we screened for mutations in an Aos transgene that suppress the strong misexpression phenotype caused in the developing eye (29). We report the identification of a series of point mutations in Aos that reduce or impair its activity in multiple tissues. These lesions map to both the NCR and CCR regions, demonstrating that both modules are necessary for Aos function in vivo. We have also correlated reductions in the phenotypic strength of Aos mutants with decreases in their in vitro binding affinity for Spi. Whereas a mere ϳ2-fold reduction in affinity appears to be sufficient to reduce the effectiveness of Aos as a dEGFR inhibitor, a ϳ37fold decrease in affinity greatly compromises its activity. We also show that these mutants are correspondingly less efficient in abolishing Spi-dependent dEGFR phosphorylation in cellular studies. Our data thus show that both the NCR and CCR in Aos are necessary for establishing a high affinity complex with Spi, which is critical for imparting dEGFR signaling robustness.

EXPERIMENTAL PROCEDURES
Identification of Aos Orthologs-Publicly available genome sequences were searched using the tblastn algorithm for se- quences related to Aos. Sequence alignments were performed with ClustalW1.8 (searchlauncher.bcm.tmc.edu/multi-align/ multi-align.html) and prepared for publication using the BOX-SHADE server (www.ch.embnet.org/software/BOX_form.html).
For rescue analysis, a sev-GAL4 driver was combined with each of the represented UAS-aos loss of function alleles (see Fig.  5) and then introduced into an aos null background (aos ⌬7 / Df(3L)Exel6129). For scanning electron microscopy of the adult eye, the females were dehydrated in an increasing ethanol: dH 2 O series, as described by Tio et al. (18).
Sequence and Western Blot Analysis of aos Alleles-For each putative loss of function allele genomic DNA was isolated from 10 -20 adult flies with Qiagen DNeasy columns (Qiagen). The aos transgene was then amplified by PCR, purified by gel extraction (Qiagen), and sequenced using cycle sequencing according to the manufacturer's instructions (Applied Biosystems). At least two independent rounds of genomic DNA purification, PCR, and sequencing were carried out for each allele. For Western analysis, ovaries from four females misexpressing the suppressors during stages 9 -11 in the follicle cells (CY2-GAL4, UAS-aos) were dissected in phosphate-buffered saline, transferred into 50 l of phosphate-buffered saline ϩ25 l of 4ϫ sample buffer on ice, and homogenized. 15 l of each sample was loaded on an 8% SDS-PAGE gel and transferred to nitrocellulose. The blots were probed with anti-Aos (Developmental Studies Hybridoma Bank) at 1:100, stripped, and reprobed with anti-␣-tubulin (12G10; Developmental Studies Hybridoma Bank) at 1:5000 as a loading control.
Molecular Cloning and Protein Production-Full-length Aos was amplified by PCR, incorporating a SpeI restriction site in the 5Ј primer and a His 6 tag followed by a NotI restriction site in the 3Ј primer, and subcloned into pFastbac (Invitrogen) giving pFbAosHis. Aos mutants were generated by QuikChange (Stratagene) using pFb-AosHis as a template. Baculoviruses encoding different Aos alleles were generated and amplified according to the manufacturer's instructions. For protein purification, 1 liter of Sf9 cells were infected with each corresponding baculovirus (except Aos V146D , which required 2-2.5 liters) for 3 days. Conditioned medium was dialyzed against 12 volumes of 10 mM HEPES, pH 8, 150 mM NaCl, and flowed through a nickel-nitrilotriacetic acid column with a 2-ml bed volume (Qiagen). The column was washed with 25 ml of 20 mM imidazole in buffer A (25 mM Tris, pH 8, 100 mM NaCl), and protein was eluted in 5 ml of 300 mM imidazole/buffer A. The eluted protein was concentrated and further purified by gel filtration using a Superose 6 column (Amersham Biosciences) equilibrated with 25 mM HEPES, pH 8, 150 mM NaCl. Secreted Histagged Spi (amino acids 1-128) was purified from transfected S2 cells as described previously (25). The protein concentrations were determined using absorbance at 280 nm.
Surface Plasmon Resonance (SPR)-SPR experiments were performed on a BIAcore 3000. Spi was immobilized onto a CM5 sensorchip by standard amine coupling using 10 mM acetate, pH 5.5, for preloading the CM-dextran surface. Purified Aos FIGURE 2. A screen to identify mutations affecting Aos activity. A, a strain exhibiting a rough eye phenotype caused by misexpression of Aos in the developing eye (GMRϾAos) was constructed and maintained over a balancer chromosome (CyO). This strain was ethyl methane sulfonate-mutagenized and outcrossed to a wild type strain. Progeny misexpressing Aos were then selected by the absence of the dominant curly wing phenotype associated with CyO balancer and then screened for suppression of the eye phenotype. Refer to the text and "Experimental Procedures " for additional details. B, graphic view of Aos from D. melanogaster depicting the signal sequence (SS), the two conserved regions, NCR and CCR, in black, the variable regions in gray, and the putative EGF domain marked with a black line. Class I mutants (non-cysteine missense) are shown as asterisks; class II mutants (cysteine missense) are shown as dots; and class III alleles (nonsense) are depicted with hexagons. mutants were flowed over the Spi-containing sensorchip from lower to higher concentrations at 10 l/min. The sensorchip was regenerated after running each sample with 10 mM glycine, pH 3, and 1 M NaCl. Binding curves were generated by normalizing the total response units at equilibrium against the maximal saturation response (B max ) and plotting them as a function of protein concentration. Curves were fit to a single-site binding model using the program Prism, from which the K D values were derived. The experiments were done at least three times, generating error bars and standard error values. The experiments were also carried out by flowing protein from higher to lower concentration, yielding slightly higher K D values (likely because of sample aggregation and incomplete regeneration), but similar wild type to mutant ratios.
Activation Studies-dEGFR activation assays in S2 cells were performed as described previously (25). Briefly, dEGFR-expressing D2f cells (kindly provided by Benny Shilo) were serumstarved overnight, and dEGFR production was induced with 60 M CuSO 4 for 3 h. The cells were incubated on ice with purified Spi alone or in the presence of the indicated amount of purified Aos proteins for 10 min. The cells were lysed in radioimmune precipitation assay buffer (25 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, and 0.1% SDS with phosphatase and protease inhibitors), and lysates were sepa-rated by SDS-PAGE and transferred to nitrocellulose. The blots were probed with anti-pY20 (Santa Cruz), stripped, and reprobed with anti-dEGFR as a loading control (23).

Isolation of Mutations
Affecting Aos Activity-Aos was first described in D. melanogaster as an antagonistic ligand of dEGFR with an atypical EGF motif that was hypothesized to direct its association with dEGFR and preclude Spi binding (24,34). However, recent studies have shown that Aos inhibits dEGFR signaling by binding to the activating ligand Spi rather than associating with the receptor itself (25), bringing the significance of the atypical EGF-like domain in Aos into question. As pointed out by Howes et al. (26), sequences outside the putative EGF motif of Aos are strongly conserved in the house fly Musca domestica (representing an evolutionary distance of ϳ100 million years) and are required for Aos function in vivo, further suggesting that more than just the putative EGF-like domain is required for Aos function. We have expanded upon these observations by identifying Aos orthologs outside the dipteran lineage (Fig. 1). We found Aos orthologs in two additional arthropods, the honeybee Apis mellifera and the beetle Tribolium castaneum, indicating conservation of Aos over a span of ϳ500 million years. We were unable to identify Aos orthologs in vertebrates, suggesting that, like the dEGFR inhibitor Kekkon1, the existence of Aos is phylogenetically restricted (35,36). Sequence conservation in Aos extends well beyond the boundaries of the putative EGF motif and reveals two strongly conserved regions (Fig. 1) that we term the NCR and CCR, which are separated by a region of variable length. The NCR and CCR are defined primarily by sets of 4 and 12 conserved cysteines, respectively. Previous studies have indicated that the CCR represents the primary ligand-binding domain and can provide a measure of activity in vivo (25,26). In contrast, no direct functionality has been ascribed to the NCR.
Given these regions of strong sequence conservation and the reported function of Aos as a "Spi sink," we sought to investigate the contributions of each domain to Aos function. We therefore carried out a screen for mutations that disrupt the ability of an Aos transgene to affect eye development upon misexpression, as diagrammed in Fig. 2A. During eye development Aos limits Spi availability, thus attenuating dEGFR signaling and preventing the specification of excess photoreceptors (24,29,31). Alterations in the dosage of aos (and consequently dEGFR activity) are readily observed as disruptions in the highly organized array of facets in the adult compound eye. Misexpression of an Aos transgene in the developing eye (using GMR-GAL4) leads to inhibition of dEGFR signaling (presumably through sequestration of Spi) and causes a severe loss of photoreceptors and morphological defects in the adult eye ( Figs. 2A and 3B) (37)(38)(39). Reducing transgene activity would restore normal morphology to the eye and thus provides a means to identify mutations in Aos that impair its activity in vivo.
To carry out a screen for functionally defective aos transgenes, we utilized the GAL4/UAS system (37,40). In this bipartite expression system, regulation of the transgene of interest, here aos, is controlled by UASs bound by the yeast transcriptional activator GAL4. Drosophila strains expressing GAL4 in specific spatio-temporal patterns are then used to direct UAS transgene expression in the tissue of interest. For this screen we generated a recombinant fly strain with the UAS-aos transgene under the control of a GAL4 line, P{GAL4-ninaE.GMR}, that drives GAL4 expression (and thus Aos misexpression) in the developing eye. This strain, referred to as GMRϾaos ϩ , was mutagenized and outcrossed, and the progeny were screened for suppression of the eye defects associated with Aos misexpression (Fig. 2A). In addition to recovering mutations in the aos transgene, the primary focus of this work, we anticipated the recovery of two additional classes of mutations that would also suppress the GMRϾaos ϩ eye phenotype shown in Fig. 3B. Mutations in the GAL4 driver would prevent misexpression of the aos transgene and thus revert the associated eye phenotype. These mutations were identified as progeny that failed to show eye defects when combined with a distinct UAS responder transgene, here a dominant negative version of dEGFR (UAS-dEGFR DN ), and were discarded ( Fig. 2A). Progeny that retained an intact GMR-GAL4 driver (and thus Aos misexpression) were further characterized. Mutations in genes essential for Aos function might also be recovered and as such would be unlinked to the UAS-aos transgene (38,39). Here we report the isolation and characterization of 16 mutations linked to the aos transgene that resulted in suppression of the misexpression phenotype in the eye (Table 1 and Figs. 2B and 3). Of these 16 lines, 15 showed strong suppression in the eye, consistent with minimal Aos function. The remaining mutant (aos 334 , identified as Aos P372S below) showed only partial suppression in the eye (Table 1 and Fig. 3C), suggesting that it retained moderate activity. In addition to its function in the eye, Aos has been implicated in regulating dEGFR signaling in other contexts, including wing development and embryogenesis (13,34). Suppressor activity was assessed in these two contexts as well, using wing-specific (MS1096-GAL4) and ubiquitous (Tubulin-GAL4) drivers. Misexpression of wild type aos with MS1096-GAL4 leads to a loss of wing veins, whereas embryonic misexpression with Tubulin-GAL4 results in lethality. All of the a Activity based on the severity of the rough eye phenotype when the UAS-aos allele is combined with the GMR-GAL4 driver. ϩϩ indicates a severe rough eye phenotype, whereas ϩ and Ϫ indicate moderate or weak phenotypes, respectively. b Percentage of wings with gap in the L4 vein when the UAS-aos allele is combined with the MS1096-GAL4 driver. The numbers refer to the percentages of times a gap in the L4 wing vein was observed. c Expression of aos mutants was driven ubiquitously during embryogenesis using the Tubulin-GAL4 driver. The percentage of viability was determined as described under "Experimental Procedures." The numbers in parentheses represent the total number of flies scored. ND, not determined.
mutated aos transgenes showed similar activity profiles in other tissues (Table 1), indicating that the suppressor lines from which they were derived did not have tissue-specific alterations.
The one exception to this is aos P372S , which showed only partial suppression in the eye (and in viability) but apparently strong suppression in the wing (Table 1). Molecular Characterization of aos Alleles-To ascertain the molecular nature of each mutation, we sequenced the aos transgene in each of the 16 suppressor lines (Table 1). We grouped the sequenced alleles into three general classes, based on the nature of the mutation uncovered (Table 1 and Fig. 2B). We recovered three missense mutations in residues other than cysteine, which we named class I alleles. We designated as class II alleles missense mutations that alter conserved cysteines and as class III alleles nonsense mutations that generated truncated gene products. To avoid confusion, all of the alleles will be henceforth referred to by their amino acid change. All three allelic classes highlight the relative importance of both the NCR and the CCR. Class III alleles delete all or a portion of the CCR and lack activity in our misexpression assays (Figs. 2B and 3G and Table 1), in agreement with the findings with C-terminal truncations engineered by Howes et al. (26). Class II alleles (cysteine mutants) behaved as nearly complete suppressors, consistent with the notion that proper Aos folding depends upon the formation of a correct pattern of disulfide bonds, including those outside the putative EGF domain. A cysteine in the NCR (C141) was altered in two alleles, and three cysteines in the CCR (C341, C407, and C413) were mutated (Figs. 2B and 3F and Table 1). Class I mutants functionally implicate both the NCR and CCR in Aos activity. aos S371F and aos P372S affect well conserved neighboring residues in the CCR but display different strengths in vivo ( Fig. 2B and 3, C and D, and Table 1). aos P372S retains moderate activity in the eye and embryo (the latter inferred from viability studies), whereas aos S371F exhibits minimal activity in the misexpression assays. The only class I allele uncovered in the NCR, aos V146D , alters a conserved valine (isoleucine in TcAos) to aspartate and displays minimal activity in the misexpression assays (Figs. 2B and 3E and Table 1). Western blot analysis confirmed that the differences in allelic activity are due neither to nonsense-mediated decay nor decreased protein stability. Misexpression of aos mutants in the follicle cells of the ovary (using a CY2-  A is a scanning electron micrograph of an eye lacking all endogenous aos activity (aos null ; aos ⌬7 / Df(3L)Exel6129), whereas B demonstrates an almost complete restoration of the disrupted eye morphology to wild type by an aos transgene driven by the sev-GAL4 driver (B, aos null sevϾaos ϩ ). In C (aos null sevϾaos P372S ), the morphology of the aos null eye was also greatly improved, suggesting that this class I allele retains substantial rescuing capability. The morphology of aos null eyes expressing the remaining class I alleles (D, aos null sevϾaos S371F ; E, aos null sevϾaos V146D ) was only moderately improved, suggesting they both retain a measure of activity, with aos S371F consistently displaying slightly better morphology. In contrast, the morphology of aos null eyes was not improved by expression of the class II allele aos C141Y , consistent with a lack of activity (F, aos null sevϾaos C141Y ).
GAL4 driver), where Aos has also been reported to inhibit dEGFR signaling (27), led to high levels of protein expression (Fig. 4). The sole exception to this was aos Q107X , which encodes an early nonsense mutation that eliminates the Aos antibody epitope (Fig. 4). As a more accurate and sensitive indicator of the physiological activity of the class I mutations relative to wild type Aos, we also assessed their ability to compensate for the loss of endogenous Aos in the eye. Loss of aos activity in the eye results in a roughened appearance along with a prominent blistering along the posterior margin of the adult compound eye (Fig. 5A) (29). Misexpression of wild type aos in developing photoreceptors (using sevenless-GAL4) in an aos mutant background rescues the knock-out phenotype and restores the eye to nearly wild type morphology (Fig. 5B) (26,29). Consistent with the observation that aos P372S retains considerable activity in the suppression assays, we observed significant (but not complete) rescue of the aos null phenotype with this allele (Fig. 5C). We also noted differences in the relative abilities of the other mutations to rescue the aos null eye phenotype. Both aos S371F and aos V146D provide weak rescue, with aos S371F appearing to have slightly more activity, indicating that these alleles retain some degree of activity. In contrast, the class II mutant aos C141Y , which behaved as a complete suppressor in the screen, was unable to rescue the aos null phenotype in the eye, consistent with its lack of activity (and likely misfolding).
Aos Mutants Are Defective in Their Ability to Bind and Inhibit Spi-Given the recently identified action of Aos as a ligand sink, we next asked whether the loss of activity displayed by the Aos alleles was due to a reduced affinity for Spi. To address this question, we purified recombinant Aos proteins encoding the class I mutations uncovered in the screen (Fig. 6A) and assessed their ability to bind immobilized Spi using SPR, as described under "Experimental Procedures." We excluded from this analysis mutants that modified cysteines or incorporated premature stop codons, because these alleles were inactive in vivo and are likely to produce proteins with folding defects that would be difficult to purify. We found that each class I Aos allele exhibited a weaker Spi binding affinity than the wild type protein, showing an overall correspondence with its phenotypic strength ( Fig. 6B and Table 2). The strong suppressor Aos V146D , which maps to the NCR, bound Spi with an apparent K D of 646 nM (ϳ43-fold weaker than wild type), suggesting that the NCR contributes substantially to Aos function both in vitro and in vivo. In the CCR, the strong suppressor Aos S371F bound with an apparent K D of 554 nM (ϳ37-fold weaker than wild type). In contrast, Aos P372S , which behaved as a moderate suppressor in vivo, bound Spi with an apparent K D of 32 nM, indicating that just a 2-fold decrease in affinity is sufficient to interfere with Aos function in vivo. This small decrease in affinity correlated with an increased off rate in SPR sensorgrams (Fig. 6C). We also generated a double mutant encoding both class I changes in the CCR (Aos SP-FS ) and found that the effects of the single mutants on ⌬G for Spi binding are close to additive, because Aos SP-FS bound Spi with a K D of 834 nM (about 55-fold weaker than wild type).
Finally, we investigated the relative abilities of the Aos mutants to inhibit Spi-dependent dEGFR phosphorylation in cultured cells. We treated S2 cells expressing low levels of dEGFR (to prevent self-activation) with 50 nM Spi, titrated in increasing amounts of purified Aos proteins (wild type or mutated), and examined dEGFR tyrosine phosphorylation levels by Western blotting (Fig. 7). In agreement with our studies in vivo and by SPR, the stronger alleles (Aos S371F and Aos V146D ) were unable to inhibit Spi-dependent dEGFR phosphorylation, even when used at a 25-fold molar excess over Spi. In contrast, FIGURE 6. Aos mutants display compromised binding to Spi. A, Coomassie-stained SDS-PAGE gel of purified proteins utilized in these studies. Proteins were loaded at a concentration of ϳ0.25 mg/ml. A degradation product corresponding to the C terminus of Aos was observed in all Aos variants, as determined by immunoblotting against the C-terminal His tag (not shown). B, SPR experiments reveal decreased Spi binding affinities by Aos mutants. Increasing concentrations of Aos mutants (analytes) were flowed over sensorchips with immobilized Spi (ligand), and the normalized responses at equilibrium (B max ) were plotted as a function of analyte concentration in nM. Aos P372S binds strongly to Spi with a K D of ϳ32 nM, only 2-fold weaker than Aos ϩ . Aos S371F and Aos V146D bind Spi much more weakly, with K D values of ϳ555 and ϳ646 nM, respectively. The double mutant Aos SP-FS bound Spi least strongly, with a K D of ϳ834 nM. C, SPR sensorgrams show a qualitative difference in the on and off binding rates of Spi to Aos ϩ versus Aos mutants. Aos proteins were injected at 800 nM onto a Spi sensorchip, and responses were normalized against their respective maximal binding and plotted as a function of time. A clear increase in the off rate can be observed with Aos P372S , which binds ϳ2-fold more weakly to Spi than Aos ϩ . Aos P372S nearly abolished dEGFR phosphorylation when added at a 25-fold excess, compared with the 5-fold excess of wild type Aos required to completely abolish dEGFR phosphorylation ( Fig. 7) (25). These results demonstrate that the loss of activity of Aos mutants is due to decreased binding affinities for Spi.

DISCUSSION
We have identified a series of point mutations in Aos that exhibit a range of phenotypes and have effects in multiple developmental contexts. We have also correlated the phenotypic strength of these mutants with decreased in vitro Spi binding affinity of the encoded Aos protein. The point mutations uncovered in this study implicate both the N-and C-terminal cysteine-rich regions of Aos in Spi binding and appear to define an affinity threshold for Aos function during development.
dEGFR Signaling Is Highly Susceptible to Small Changes in the Affinity of Aos for Spi-Recent computational studies by Reeves et al. (32) suggested that Aos confers robustness to the dEGFR signaling module in Drosophila development by restricting the range of Spi action and predicted that highly efficient sequestration of Spi by Aos is necessary to generate a robust feedback loop. It follows that the system should be highly susceptible to small changes in the Aos-Spi affinity or dissociation rate, and our data are in agreement with this prediction. Remarkably, a decrease of just ϳ2-fold in the affinity of Aos for Spi (caused by Aos P372S ) as assessed in our in vitro studies is sufficient to reduce the in vivo activity of this mutant in the developing eye and embryo and to greatly diminish it in the adult wing. Decreases of 37-or 43-fold in Spi binding affinity, seen with Aos S371F and Aos V146D , respectively, abolished most Aos activity from all of the tested tissues.
We also addressed whether the Aos mutants could rescue an aos null phenotype, to control for potential genetic biases inherent in the suppression studies. The rescue data support the notion that a modest decrease in binding affinity for Spi is sufficient to compromise Aos activity in vivo, because Aos mutants exhibited a similar (although overall higher) pattern of activity than in the suppression assay. The higher level of Aos mutant activity in the rescue assay may reflect a higher sensitivity of the aos null genetic background to Aos protein levels. Importantly, the progressive loss of activity of Aos alleles seen in flies could also be recapitulated in a cultured cell system. In agreement with genetic data, inhibition of Spi-induced dEGFR phosphorylation required a larger excess of purified Aos P372S than wild type Aos. In the cases of Aos S371F and Aos V146D , we were unable to observe any inhibition even when added at a 25-fold excess over Spi. Our results argue that the high binding affinity between Aos and Spi (and perhaps other ligands) is a critical parameter for full Aos activity in vivo and for conferring dEGFR signaling robustness. It will be interesting to determine whether dEGFR signaling is similarly susceptible to small changes in the affinity of dEGFR for its ligands.
One Aos mutant (Aos P372S ) exhibited a more complete suppression phenotype in the wing than in the eye or in viability studies. This could reflect the strength of the GAL4 driver utilized in the wing (MS1096). Another more intriguing possibility is that Vein, which appears to be the primary dEGFR ligand in the wing (where Spi plays no role (41)) is inhibited less efficiently by Aos, and Aos mutations would therefore impair its effect more completely. However, the effects of Aos on other dEGFR ligands have not yet been reported.
Structural Considerations-Aos was originally thought to inhibit dEGFR directly, via a proposed "atypical" EGF domain. It is now known instead to act as a ligand antagonist, sequestering Spi in a manner similar to bone morphogenetic protein inhibition by Chordin (orthologous to Drosophila Short gastrulation) (42)(43)(44)(45), and insulin growth factor 1 binding by the insulin growth factor-binding protein family (46). A common feature of these ligand antagonists is the presence of two cysteine-rich domains. Our results suggest that the two cysteine-rich domains in Aos, NCR and CCR, function interdependently in binding and sequestering Spi. Indeed, mutations in either the NCR or CCR impair Aos function and in vitro Spi binding. Interestingly, Aos V146D binds Spi with a K D similar to that previously reported for an Aos fragment containing only the C-terminal 225 amino acids (and lacking the NCR altogether) (25). This observation suggests that the C-terminal region of Aos provides the majority of Spi binding interactions, with additional weak contributions from the NCR being required for high affinity binding of full-length Aos to Spi. A similar bipartite binding site involving two cysteine-rich regions has been reported for insulin growth factor-binding protein family members (47). It is interesting to speculate that the required contributions of both the NCR and CCR for efficient Spi sequestration, and their linkage by a proteolytically sensitive region could support a proteolysis-based mechanism for relieving Spi inhibition (or attenuating Aos function) during development. A precedent for this exists in bone morphogenetic protein signaling where the protease Tolloid relieves Chordin-mediated sequestration of bone morphogenetic protein (48).
The findings in this report show a clear correlation between the failure of Aos to function in vivo and a reduction in its ability to efficiently sequester Spi in vitro and in cell culture systems. Thus, our studies provide direct in vivo evi- FIGURE 7. Aos mutants are deficient in their ability to inhibit Spi-dependent dEGFR phosphorylation. D2f cells expressing low levels of dEGFR were left unstimulated or were stimulated on ice for 10 min with 50 nM Spi together with increasing concentrations of purified wild type or mutated Aos proteins. dEGFR tyrosine phosphorylation levels were assayed by immunoblotting with anti-phosphotyrosine (top panels). Immunoblots were stripped and reprobed with anti-dEGFR to ensure equal loading (bottom panels). Whereas a 5-fold molar excess of Aos ϩ over Spi is sufficient to abolish dEGFR phosphorylation, a 25-fold excess of Aos P372S is necessary to decrease dEGFR phosphorylation to nearly base-line levels. In contrast, Aos S371F and Aos V146D do not affect Spi activity even at high molar excesses.
dence for the Spi sequestration model presented by Klein et al. (25) based on in vitro analyses of interactions between Aos, Spi, and the dEGFR extracellular region. In addition, the fact that missense mutations in both the N-and C-terminal cysteine-rich regions of Aos impair Spi binding suggests a bipartite Spi-binding site in Aos and provides the first insight into how this ligand antagonist achieves its function. A full understanding of how Aos recognizes its target and how this can be exploited in the design of antagonists for human EGF ligands will require crystallographic and further functional studies of the Aos-Spi complex.