Sprouty proteins are in vivo targets of Corkscrew/SHP-2 tyrosine phosphatases

RTK signaling pathways regulate many cellular and developmental events,including respiratory system branching and photoreceptor differentiation in Drosophila. The proper magnitude and duration of signaling is crucial for such events, because altering these parameters by constitutively activating an RTK pathway or altering its kinetics can lead to different outcomes (Marshall, 1995). Indeed, many RTK pathways are equipped with positively acting regulators and negative feedback loops, the balance of which dictates the cellular response(Perrimon and McMahon, 1999). Two of the best known but least well understood classes of regulators are the Corkscrew/SHP-2 family of tyrosine phosphatases that promote RTK signaling and the Sprouty family of RTK feedback inhibitors, whose genetic and molecular interactions we explore here.

corkscrew was identified by its requirement for development of the terminal regions of the embryo. It encodes an SH2 domain-containing protein tyrosine phosphatase that functions downstream of, and promotes signaling through, the RTK Torso (Perkins et al.,1992). A vertebrate homolog, SHP-2 (now known as Ptpn11), was found to promote PDGFR signaling (Freeman et al., 1992; Bennett et al.,1994). Further work demonstrated that Csw and SHP-2 promotes signaling downstream of many RTKs in a variety of systems(Perkins et al., 1996)(reviewed by Feng, 1999), and that this activity requires their tyrosine phosphate-binding SH2 domains and tyrosine phosphatase activity (Allard et al., 1998; Deb et al.,1998).

How can tyrosine phosphatases promote tyrosine kinase signaling? To account for their positive effect on signaling, Csw and SHP-2 were postulated to dephosphorylate either a positive RTK effector that is inactivated by phosphorylation or a negative regulator that is activated by phosphorylation(Stein-Gerlach et al., 1998; Huyer and Alexander, 1999). This initiated searches for Csw/SHP-2 substrates by genetic screens(Herbst et al., 1996; Firth et al., 2000),biochemical screens using substrate-trapping forms of the phosphatases(Herbst et al., 1996; Agazie and Hayman, 2003a), and tests of candidate proteins. These identified several signal transduction scaffolding proteins, including DOS, GAB1, GAB2, SHPS1 and IRS1(Qu, 2002). However,dephosphorylation of these substrates by Csw/SHP-2 inhibits RTK signaling, so they cannot account for the positive effects of Csw/SHP-2.

Two other types of substrates may account for some positive effects of Csw/SHP-2 on RTK signaling. One is the phosphotyrosines on Torso, PDGFR and EGFR that mediate binding and signal inhibition by RasGAP(Cleghon et al., 1998; Ekman et al., 2002; Agazie and Hayman, 2003b). The other is the CSK-binding proteins Paxillin and PAG/Cbp, dephosphorylation of which prevent CSK-mediated inactivation of Src kinase activity and its positive effect on RTK signaling (Ren et al., 2004; Zhang et al.,2004). However, neither of these is likely to be a general target and other crucial substrates remain to be identified.

Several observations make Sprouty proteins appealing candidates for Csw/SHP-2 substrates. First, although Sprouty (Spry) was identified by its role as an Fgf feedback inhibitor in Drosophila tracheal(respiratory) development (Hacohen et al.,1998), Spry and its four mammalian homologs, Spry1-Spry4, are now known to function more broadly as RTK feedback inhibitors that can regulate a variety of RTK pathways including FGF, EGF, VEGF and PDGF pathways(Casci et al., 1999; Kramer et al., 1999; Reich et al., 1999; Impagnatiello et al., 2001; Lee et al., 2001; Nutt et al., 2001), all of which are also regulated by Csw/SHP-2(Perkins et al., 1996; Feng, 1999). Both Spry and SHP-2 influence RTK signaling kinetics(Saxton et al., 1997; Dikic and Giordano, 2003). Second, although the mechanism by which Spry proteins modulate signaling is not understood, like Csw/SHP-2 they localize to the membrane following signaling and associate with RTK signaling complexes(Allard et al., 1996; Herbst et al., 1999; Qu, 2000; Hanafusa et al., 2002). Third,cell culture studies of Spry1 and Spry2 demonstrate that they are phosphorylated on a conserved tyrosine in response to RTK signaling, and the modification is required for Spry function(Hanafusa et al., 2002; Hall et al., 2003; Rubin et al., 2003; Li et al., 2004; Mason et al., 2004).

Here, we investigate the relationship between Csw/SHP-2 and Sprouty in RTK pathways in vivo and in vitro. We establish a close functional relationship by manipulating gene activity during RTK signaling in Drosophiladevelopment and in cultured mammalian cells. We use the cell culture system to demonstrate that SHP-2 controls phosphorylation on the essential tyrosine of Spry1. Biochemical experiments show that SHP-2 associates in a complex with Spry1 and can dephosphorylate the essential tyrosine. Substrate-trapping forms of Csw bind Spry in cultured Drosophila cells and during development. The results demonstrate that Spry proteins are targets of Csw/SHP-2 tyrosine phosphatases in Drosophila and vertebrate RTK pathways, and explain how Csw/SHP-2 can enhance RTK signaling by inactivating a feedback inhibitor. While this manuscript was in preparation, complementary work showing that SHP-2 can reduce tyrosine phosphorylation on a Spry protein in PC12 cells was reported (Hanafusa et al.,2004).

spry^Δ5 null allele was used(Hacohen et al., 1998). csw transgenes under sevenless enhancer control were used for Csw misexpression during eye development: SE-csw^G547E, SE-csw^C583S, SE-csw^src90, SE-csw^G547Esrc90 and SE-csw^C583Ssrc90(Allard et al., 1996; Allard et al., 1998). We refer to csw^src90 as myr-csw. Protein misexpression with GAL4/UAS system (Brand and Perrimon,1993) used sev-GAL4(Brand and Perrimon, 1993), btl-GAL4 (Shiga et al.,1996), UAS-GFP (Jarecki et al., 1999), UAS-csw^src90(Johnson Hamlet and Perkins,2001) and UAS-spry(Hacohen et al., 1998). UAS-spry^Y201F was made by introducing a TAT>TTT mutation in pUAST-spry (Hacohen et al., 1998) and establishing transgenic lines by P element transformation; homozygous viable insertions on chromosome II were used(insertions 13, 15, 37). Crosses were carried out at 25°C.

Tracheae were visualized by fluorescence microscopy of heat-killed(65°C for 5 seconds) third instar larvae carrying FRT2A and viable insertions of btl-GAL4 and UAS-GFP on chromosome II. Sections (2 μm) through adult eyes were prepared as described(Tomlinson and Ready,1987).

HEK293 cells (ATCC #CRL-1573) were grown in a humidified chamber containing 5% CO₂ at 37°C in Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum (Invitrogen). For transfections,2×10⁵ log phase cells were cultured in 35 mm plates and transfected with 0.5 μg of each expression plasmid using Fugene 6 lipid transfection reagent (Invitrogen). Total DNA per transfection was 1 or 2 μg depending on number of plasmids used; empty vector was used to keep total DNA constant.

Drosophila S2 cells were maintained at 22°C in M3 insect medium (Sigma) or Schneider's medium (Invitrogen) supplemented with 10%heat-inactivated fetal calf serum. Stably transfected S2 cell lines expressing Csw, Csw^C583S and Breathless with C-terminal FLAG epitope(Herbst et al., 1996; Toering, 2003) were maintained in medium containing 200 μg/ml hygromycin.

Expression plasmids for HEK293 cells were: pRc/CMV/SHP-2 and pRc/CMV/SHP-2^C459S (Paul Khavari, Stanford) with CMV promoter driving expression; pEFBOS/mFGFR3 and pEFBOS/mFGFR3^K644E(thanatophoric dysplasia type II mutation)(Su et al., 1997) with human EF-1α promoter; pMO/IRES/FGFR1c (David Ornitz, Washington University)with MoLTR; and pBJ5/ERK2-HA (Gerald Crabtree, Stanford) with SRαpromoter. pTA/HA-Spry1, with CMV promoter driving expression of mouse Spry1 with N-terminal HA epitope (YPYDVPDYA), was constructed by PCR amplification of Spry1 cDNA (Minowada et al.,1999) using a forward primer encoding initiator methionine, HA epitope and residues 2-6 of Spry1, and insertion into pTA vector (Invitrogen). Spry1^Y53F (TAC>TTC) and Spry1^Y89F (TAC>TTC)mutations were introduced into pTA/HA-Spry1 by site-directed mutagenesis. pCDNA/mFL-Spry1 was constructed as above except FLAG epitope (DYKDDDDK) and pCDNA vector (Invitrogen) were used. pCDNA/SHP-2-V5 was constructed in similar manner by amplifying SHP-2 sequence in pRc/CMV/SHP-2 with reverse primer encoding V5 epitope (GKPIPNPLLGLDST). Coding sequences of constructs were verified by DNA sequencing.

Rabbit antisera were raised against an N-terminal peptide of Drosophila Sprouty (residues 19-37, LPRVHRPRAPEPTLSGVDH) and against a C-terminal peptide (574-591, RKGDLTPEKRLLDSSPDY) (Biosynthesis, Lewisville,TX). Sera were affinity purified on columns containing the immobilized peptide and used at 1:1000 dilution (immunoprecipitation) or 1:2000-1:5000(immunoblots). Rabbit anti-pY53 mouse Spry1 antiserum was raised against phosphopeptide GCGSNEpYTEGPSVARRPAPR that includes Spry1 residues 49-66. Eight week post-immunization bleed was used at 1:2000.

Twenty-four hours after transfection of HEK293 cells with pBJ5/ERK2-HA,human recombinant bFGF (Invitrogen) was added to the medium to 25 ng/ml. At times indicated after bFGF addition, cells were rinsed with PBS and lysed with 0.4 ml RIPA buffer (50 mM Tris pH 7.4, 1% NP-40, 0.25% Na-deoxycholate, 150 mM NaCl, 1 mM EDTA) containing protease inhibitors (Complete-Mini, Roche) and 0.2 mM sodium vanadate to inhibit tyrosine phosphatases. HA-ERK2 was immunoprecipitated by incubating lysate at 4°C with 10 μg anti-HA mAb(Roche). After 3 hours, 35 μl protein A agarose (Sigma) was added and incubation continued for 1 hour. Immunoprecipitate was washed three times at 4°C with 1 ml NP40 wash buffer (1% NP40, 1 mM EDTA) and once with 1 ml TNE(10 mM Tris pH 8, 100 mM NaCl, 1 mM EDTA), boiled in Laemmli loading buffer and resolved on 10% SDS-PAGE gel. Gel was probed with anti-dpERK mAb (Sigma),HRP-conjugated secondary antibody and HRP chemistry (Enhanced Chemiluminescence, Amersham). dpERK levels were quantitated by scanning densitometry of fluorograms. Blots were reprobed with anti-HA to detect total ERK2-HA. Expression of other proteins was monitored by SDS-PAGE and immunoblotting of aliquots of cell lysates.

HEK293 cells were transfected with pTA/HA-Spry1, pMO/IRES/FGFR1c and the plasmids noted. Twenty-four hours later, bFGF was added to medium at 200 ng/ml. After 30 minutes, cells were lysed and HA-Spry1 was immunoprecipitated by incubation with 35 μl anti-HA agarose (Santa Cruz Biotechnology) at 4°C for 4 hours. Immunoprecipitates were washed and separated by SDS-PAGE. Immunoblots were probed with anti-phosphotyrosine mAb (Upstate Biotechnology)and HRP immunochemistry. Same conditions were used for analysis of Spry1 Y53 phosphorylation, except cells were treated with bFGF for times noted, HA-Spry1 was immunoprecipitated with anti-HA and immunoblots were probed with anti-pY53.

To identify Spry1 phosphotyrosines, 2×10⁶ HEK293 cells were transfected with pTA/HA-Spry1 and pMO/IRES/FGFR1c. Forty-eight hours later, bFGF was added to 200 ng/ml. After 30 minutes, cells were lysed,HA-Spry1 was immunoprecipitated with anti-HA, resolved by SDS-PAGE, and stained with Coomassie Blue. HA-Spry1 band was excised, treated with trypsin and proteolytic fragments were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) and identified using Mascot MS/MS Ions Search (Mass Spectrometry Lab, Stanford, CA).

To analyze tyrosine phosphorylation of Drosophila Spry,5×10⁶ exponentially growing S2 cells expressing Breathless-FLAG were treated for 15 minutes with 0.1 mM pervanadate, then washed and lysed. Endogenous Spry was immunoprecipitated with anti-Spry C-terminal antiserum, resolved by SDS-PAGE and probed on immunoblots with anti-phosphotyrosine.

HEK293 cells were transfected with plasmids expressing SHP-2-V5, HA-Spry1 and FGFR3 or FGFR3^K644E to activate the FGF pathway. After 48 hours, cells were lysed with 0.4 ml NP40 lysis buffer (150 mM NaCl, 1% NP40,50 mM Tris pH 8) containing protease inhibitors. SHP-2-V5 was immunoprecipitated with anti-V5 mAb (Invitrogen), washed and resolved by SDS-PAGE. HA-Spry1 in immunoprecipitate was detected on immunoblots probed with rabbit anti-HA (Santa Cruz Biotechnology).

To assay Csw-Spry association in stably transfected S2 cells expressing Csw or Csw^C583S, ∼2×10⁷ exponentially growing cells in M3 medium were harvested by centrifugation and lysed in 1 ml NP-40 lysis buffer containing protease inhibitors. Csw was immunoprecipitated at 4°C for 2-4 hours with 1 μl anti-Csw-CT(Allard et al., 1996) and 50%(v/v) protein G Sepharose beads (Sigma). Beads were washed three times with lysis buffer, boiled in Laemmli loading buffer and proteins resolved on 8%SDS-PAGE gels. Endogenous Spry that co-immunoprecipitated was detected on immunoblots probed with anti-Spry C-terminal. Spry immunoprecipitation was carried out as above using anti-Spry and protein A beads;co-immunoprecipitated Csw was detected with anti-Csw-CT.

For Drosophila imaginal discs, 100 pairs of third instar eye-antennal discs of each genotype were dissected into ice-cold PBS(Fig. 5E) or snap frozen and stored at –80°C, homogenized as above and cleared by centrifugation. Csw immunoprecipitation and subsequent analysis were as above.

GST-SHP-2 fusion proteins were expressed in E. coli and purified as described (O'Reilly et al.,2000). To prepare lysates containing Spry1, HEK293 cells were transfected with pTA/HA-Spry1 and pEFBOS/FGFR3^K644E to induce HA-Spry1 phosphorylation. Cells were lysed in NP40 lysis buffer with protease inhibitors. Lysate was mixed with 100 μl of 50% (v/v) glutathione-Sepharose beads coated with a GST-fusion protein and incubated at 4°C. After three hours, beads were washed three times with NP40 wash and once with TNE. Proteins were separated by SDS-PAGE, and immunoblots were probed with anti-HA.

Phosphatase assay (O'Reilly et al.,2000) used purified GST-SHP-2 proteins. To prepare phosphorylated HA-Spry1 substrate, transfected HEK293 cells expressing FGFR1c and HA-Spry1 were treated with bFGF at 200 ng/ml and 0.1 mM pervanadate for 30 minutes. Cells were lysed in NP40 lysis buffer and HA-Spry1 was immunoprecipitated with anti-HA. Immunoprecipitates were washed with NP40 wash buffer and TNE, and resuspended in phosphatase buffer (25 mM HEPES 7.4, 150 mM NaCl, 5 mM EDTA, 10 mM DTT). Phosphatase reactions (25 μl) containing phosphatase buffer, 50 ng GST-SHP-2 fusion protein, and substrate immunoprecipitated from 2×10⁵ cells, were incubated at 37°C. After 30 minutes,Laemmli sample buffer was added, and products were resolved by SDS-PAGE. Immunoblots were probed with anti-phosphotyrosine, anti-pY53 or anti-HA. ERK2-HA substrate was prepared from pBJ5/ERK2-HA-transfected cells and analyzed on immunoblots with anti-dpERK.

We compared the effects of Spry and Csw/SHP-2 on RTK signaling processes in two different Drosophila tissues and a human cell line. In the Drosophila tracheal system, Spry negatively regulates the Branchless(Bnl) FGF pathway that controls branch sprouting. In spry^– mutants, this RTK pathway is overactive and extra cells become terminal cells and form ectopic branches(Hacohen et al., 1998). csw^– mutations also cause tracheal defects among other embryonic defects (Perkins et al.,1996; Firth et al.,2000), and we found that csw^G547E, a dominant-negative allele (Allard et al.,1996) (Fig. 1A),enhanced the penetrance of ganglionic branch outgrowth defects in bnlheterozygous embryos (data not shown), as does partial inactivation of other Bnl pathway genes (Sutherland et al.,1996). Furthermore, expression of myr-Csw, an activated membrane-localized form of Csw containing the c-Src myristylation site(Allard et al., 1996), in the developing tracheal system using a btl-GAL4 driver induced ectopic terminal cells and branches, just like spry null mutations(Fig. 1B-E). These results imply that Csw and Spry have opposite effects on the Bnl RTK pathway.

In the developing eye, both csw and spry regulate EGFR and/or Sevenless RTK pathways that specify the pattern and fates of photoreceptors and neighboring cells(Allard et al., 1996; Allard et al., 1998; Casci et al., 1999; Kramer et al., 1999). Indeed,chromosomal deficiencies encompassing spry were identified in a screen for suppressors of the eye and wing vein phenotypes caused by a csw weak loss-of-function allele(Firth et al., 2000),indicating that these genes have opposing roles during RTK signaling in the developing eye and wing. Consistent with this, removing one copy of spry enhanced the phenotype of extra R7 photoreceptors caused by eye-specific expression of myr-Csw (Allard et al., 1996), increasing the number of R7 cells per ommatidium from 2.1±0.06 (mean ± s.e.m.) to 3.0±0.07(Fig. 1F). Likewise, removing one copy of spry suppressed the loss of photoreceptor phenotype caused by eye-specific expression of dominant negative Csw^G547E(Allard et al., 1998),increasing the number of outer photoreceptors per ommatidium from 3.3±0.05 to 4.1±0.03 (Fig. 1G). Reducing spry dose did not suppress the effect of another dominant-negative Csw, Csw^C583S(Fig. 1H; 5.1±0.03 for spry⁺ versus 4.8±0.03 for spry⁺/spry^Δ5). Csw^C583S is a substrate-trapping form of the enzyme that, unlike Csw^G547E, can bind its substrates, although it does not hydrolyze or release them (Herbst et al.,1996). We consider the implications of this result in the Discussion.

To compare effects of Spry and Csw/SHP-2 on RTK signal transduction, we assessed their effects on MAPK activation induced by FGF signaling in human embryonic kidney 293 (HEK293) cells, using an antiserum specific for diphosphorylated (activated) MAPK (dpERK). Transient transfection of a Spry1 expression plasmid reduced activation of MAPK in response to basic FGF (bFGF),as did transfection of a plasmid expressing dominant negative SHP-2^C459S (Fig. 1I). When both Spry1 and SHP-2^C459S were expressed,MAPK activation was almost completely abolished. Spry1 and SHP-2 both predominantly affected the duration of MAPK activation in response to bFGF,rather than the magnitude of the response, consistent with the idea that they regulate the same step in signal transduction(Fig. 1J). Thus, in all three systems investigated, Spry and Csw/SHP-2 proteins regulated the same RTK signaling processes but in opposite directions.

To determine the regulatory relationship between Csw/SHP-2 and Spry during RTK signaling, the effect of SHP-2 on tyrosine phosphorylation of Spry1 was analyzed in HEK293 cells. FGF signaling induces tyrosine phosphorylation of Spry proteins in several cell lines including HEK293 cells(Hanafusa et al., 2002; Tefft et al., 2002)(Fig. 2B, lanes 1,4). If Spry1 is a substrate of SHP-2 (Fig. 2A, arrow 2), then inhibiting SHP-2 should increase tyrosine phosphorylation on Spry1, even though it reduces signaling overall. However,if SHP-2 functions upstream of Spry1 phosphorylation (arrow 1), then inhibiting SHP-2 should decrease Spry1 phosphorylation. If SHP-2 acts downstream or parallel to Spry1 (arrow 3), no change in Spry1 phosphorylation is expected.

To assess the effect of SHP-2 inhibition on Spry1, HEK293 cells were co-transfected with plasmids expressing dominant-negative SHP-2 and Spry1 with a hemagglutinin epitope (HA-Spry1). Tyrosine phosphorylation of HA-Spry1 was assayed following FGF induction. Expression of SHP-2^C459S increased phosphorylation on Spry1 (Fig. 2B, lanes 1,2,4,5), as did expression of another dominant-negative SHP-2, SHP-2^ΔP (data not shown). Thus, SHP-2 negatively regulates tyrosine phosphorylation on Spry1.

Overexpression of wild-type SHP-2 reduced Spry1 tyrosine phosphorylation,consistent with this model (Fig. 2B, lanes 3,6). However, the effect was small and variable,presumably because SHP-2 levels are not limiting and its phosphatase activity is controlled by binding via its SH2 domain to scaffolding proteins as in other contexts (Barford and Neel,1998).

There are five conserved tyrosines in vertebrate Sprouty proteins, one of which (Y53 in Spry1 and Spry4, Y55 in Spry2) is crucial for Spry1 and 2 activity in PC-12 cells (Hanafusa et al.,2002) and for Spry2 and Spry4 in HEK293 cells(Sasaki et al., 2001). Y53 is also crucial for Spry1 activity in HEK293 cells, because changing it to phenylalanine (Spry1^Y53F) abolished its ability to block bFGF activation of MAPK (data not shown). To determine whether phosphorylation on this critical tyrosine residue is regulated by SHP-2, a phospho-specific antiserum was generated that recognizes the pY53 form of Spry1(Fig. 2C). Expression of dominant-negative SHP-2^C459S increased phosphorylation on Y53, both in the absence and presence of added FGF(Fig. 2D). Thus, SHP-2 regulates phosphorylation of this crucial Spry1 tyrosine.

To determine whether SHP-2 regulates phosphorylation of other Spry1 tyrosines, the effects of dominant negative SHP-2 on Spry1^Y53F were examined using a general anti-phosphotyrosine antibody. The Y53F mutation did not completely eliminate tyrosine phosphorylation of Spry1, and the residual phosphorylation increased in the presence of dominant negative SHP-2(Fig. 2E, lanes 1-4). Thus,there is at least one other Spry1 phosphotyrosine regulated by SHP-2.

To identify additional Spry1 phosphotyrosines, HA-Spry1 was purified from FGF-stimulated HEK293 cells and analyzed by tandem mass spectrometry. This confirmed phosphorylation of Y53 and identified Y89 as a second phosphorylation site. Mutation of Y89 to phenylalanine (Spry1^Y89F)reduced Spry1 tyrosine phosphorylation in response to FGF(Fig. 2E, lanes 1,5), implying that Y89 is also a major phosphorylation site. However, Spry1^Y89Finhibited MAPK activation by FGF signaling in the HEK293 cell assay,demonstrating that phosphorylation of Y89 is not essential for this activity(data not shown). Although dominant-negative SHP-2 increased tyrosine phosphorylation on both Spry1^Y53F and Spry1^Y89F(Fig. 2E, lanes 3-6), there was little tyrosine phosphorylation and only a small effect of dominant-negative SHP-2 on the Spry1^Y53F/Y89F double mutant(Fig. 2E, lanes 7,8). Thus,tyrosines 53 and 89 are the major tyrosine phosphorylation sites on Spry1 and both are regulated by SHP-2.

If Spry1 is a SHP-2 substrate, the proteins must associate at least transiently in vivo. To test this, HA-Spry1 and SHP-2 with a V5 epitope(SHP-2-V5) were co-expressed in HEK293 cells. SHP-2-V5 was immunoprecipitated from cell extracts, and HA-Spry1 that co-immunoprecipitated was detected on immunoblots. Little HA-Spry1 was detected in immunoprecipitates from unstimulated cells (Fig. 3A,lane 1), comparable with that observed in controls lacking SHP-2-V5 (lanes 7-9). However, when FGF signaling was activated by overexpression of FGFR3, an association between the two proteins was detected, and more was observed when constitutively active FGFR3^K644E(Su et al., 1997) was expressed (lanes 2,3). Similar results were obtained when the antibody treatments were reversed (data not shown). Thus, FGF signaling induces formation of a complex containing SHP-2 and Spry1. Complex formation does not require the major Spry1 tyrosine phosphorylation sites, because it was not affected by Spry1^Y53F or Spry1^Y53F/Y89F mutations(Fig. 3A, lanes 4-6 and data not shown).

Interaction between SHP-2 and Spry1 was confirmed by GST pull-down experiments using a full-length, constitutively-active SHP-2(SHP-2^E76A) GST fusion protein(O'Reilly et al., 2000). When purified SHP-2^E76A-GST attached to beads was incubated with extracts of HEK293 cells expressing HA-Spry1, HA-Spry1 bound to the beads,whereas little bound to control beads coated with GST(Fig. 3B). HA-Spry1 also bound to beads coated with a GST-fusion containing only the SH2 domains of SHP-2(SHP-2^SH2), but with reduced efficiency. Thus, SHP-2 and Spry1 associate in a complex, and the interaction is mediated at least in part by SHP-2 SH2 domains.

To test whether SHP-2 can dephosphorylate Spry1, purified SHP-2^E76A-GST was incubated with HA-Spry1 protein isolated from FGF-stimulated HEK293 cells. The SHP-2 E76A mutation prevents auto-inhibition of the phosphatase domain, alleviating the need for factors that might be necessary in vivo to relieve auto-inhibition(O'Reilly et al., 2000). SHP-2^E76A-GST eliminated tyrosine phosphorylation on HA-Spry1,including the crucial phosphotyrosine (Y53), whereas control proteins lacking the phosphatase domain (SHP-2^SH2, SHP-2^ΔP) and GST alone had no effect (Fig. 4A,B). Under identical conditions, SHP-2^E76A-GST did not dephosphorylate ERK2 phosphotyrosine(Fig. 4A, lower panel). Thus,Spry1 is a substrate of SHP-2 in vitro.

The mammalian cell culture and biochemical experiments demonstrate that SHP-2 negatively regulates tyrosine phosphorylation on Spry1, that the proteins associate in a signal-induced complex, and that SHP-2 selectively dephosphorylates Spry1 in vitro. Genetic interactions between csw and sprouty in tracheal and eye development suggest that a similar regulatory relationship might exist for the Drosophila proteins.

To determine if Drosophila Spry is tyrosine phosphorylated,endogenous Spry in Breathless-expressing S2 cells was immunoprecipitated and probed with anti-phosphotyrosine antiserum(Fig. 5A). Spry isoforms of 70 kDa and 42 kDa are present in S2 cells(Toering, 2003). Tyrosine phosphorylation of the 70 kDa species was detected at low levels in untreated cells, and phosphorylation of both forms was apparent when cells were treated with the phosphatase inhibitor pervanadate, implying that Spry phosphorylation is regulated by an endogenous phosphatase.

The tyrosine crucial for vertebrate Spry function is conserved in Drosophila Spry (Hanafusa et al.,2002). To determine if the conserved tyrosine (Y201) is important for function, transgenes encoding wild-type Spry (UAS-spry) or mutant Spry with the tyrosine substituted with phenylalanine(UAS-spry^Y201F) were expressed using sev-GAL4driver, and their effects on eye development examined. Spry expression caused misrotation and disorganization of ommatidia, missing photoreceptors in 5% of ommatidia, and external roughening of the eye(Fig. 5B and data not shown). Spry^Y201F expression had little or no effect. In two out of three UAS-spry^Y201F insertions analyzed, ommatidia had normal organization and no missing photoreceptors or eye roughening(Fig. 5C). The same was true of the third insertion, except it caused a low frequency of photoreceptor loss(0.4% of ommatidia). Thus, the conserved tyrosine is crucial for biological activity of Drosophila Spry.

The above results and the genetic interactions between csw and spry suggest that Csw might regulate Spry directly by dephosphorylation. In vivo substrates of Csw/SHP-2 can be identified with substrate-trapping forms of the enzymes, such as Csw^C583S, which bind but do not dephosphorylate or release their substrates(Herbst et al., 1996; Agazie and Hayman, 2003a). To determine if Csw^C583S can trap endogenous Spry in S2 cells,wild-type Csw and Csw^C583S were expressed in S2 cells and the amount of Spry that bound to each was determined by co-immunoprecipitation(Fig. 5D). A small amount of both Spry isoforms bound wild-type Csw (lane 3). Substantially more of each isoform bound Csw^C583S (lane 4). This implies that Spry is a direct target of Csw in S2 cells.

To test for a substrate-trapping interaction between Csw^C583Sand Spry during development, binding was analyzed in a similar manner in extracts of eye-antennal imaginal discs dissected from transgenic larvae expressing modified Csw proteins (myr-Csw, myr-Csw^C583S or myr-Csw^G547E). Myristylated forms of Csw were used to facilitate membrane localization of the enzyme and detection of interaction with substrates (Allard et al.,1996). There was low but detectable interaction between myr-Csw and endogenous Spry (Fig. 5E,lane 4). Stronger interaction was observed with myr-Csw^C583S (lane 5). The enhanced interaction was not simply due to stronger interaction between the SH2 domains of the mutant Csw and a scaffolding protein such as Dos that cannot be dephosphorylated by it. If so, other catalytically inactive forms of Csw, such as myr-Csw^G547E, should show the same enhanced interaction, which was not observed (lane 6). We conclude that the enhanced interaction between Spry and myr-Csw^C583S is a substrate-trapping effect, providing strong evidence that Spry is a substrate of Csw during eye development.

Since discovery over a decade ago of the key roles of Csw/SHP-2 tyrosine phosphatases in promoting RTK signaling, identification of their crucial substrates has remained an important goal. Our results identify Spry proteins as in vivo substrates of Csw/SHP-2, and show how Csw/SHP-2 can promote RTK signaling by dephosphorylating and inactivating these RTK feedback inhibitors.

Four lines of evidence support the conclusion that Csw/SHP-2 inactivate Spry proteins by direct binding and dephosphorylation. First, genetic experiments in developing Drosophila eye and trachea and HEK293 cells demonstrated that Csw/SHP-2 and Spry act in the same RTK signaling events but in opposite directions. Indeed, manipulating their activity in opposite directions caused similar Drosophila phenotypes and similar effects on MAPK activation in HEK293 cells, and reducing spry dose suppressed the csw loss-of-function phenotype in the eye and enhanced the gain-of-function phenotype, supporting the idea that they regulate the same step in signaling. Second, molecular epistasis experiments in HEK293 cells demonstrated that SHP-2 functions upstream of, and negatively regulates,phosphorylation of the critical tyrosine residue (Y53) of Spry1. Third,biochemical studies of extracts of HEK293 cells, Drosophila S2 cells,and eye discs demonstrated that Csw/SHP-2 proteins associate in complexes with Spry proteins. Interaction was enhanced in S2 cells and eye discs when a substrate-trapping Csw was used. Interaction involves more than just binding of Csw/SHP-2 to the crucial tyrosine, because complex formation was observed with SHP-2 mutants lacking the phosphatase domain and with a Spry mutant lacking the tyrosine. Finally, purified SHP-2 selectively dephosphorylated Spry1 in vitro. These data support the conclusion that Spry proteins are direct targets of Csw/SHP-2 in all three systems examined.

One genetic result did not readily fit with the model that Csw functions by inactivating Spry by dephosphorylation. Whereas reduction of sprydose suppressed the eye phenotype of a hypomorphic csw allele(Firth et al., 2000) and dominant-negative Csw^G547E, consistent with the model, it did not suppress the milder phenotype of dominant-negative Csw^C583S (Fig. 1G,H). This catalytically inactive, substrate trapping form of Csw has unusual properties: it behaves in a dominant-negative fashion, interfering with wild-type Csw function, but also retains some wild-type Csw function because it partially rescues other dominant-negative and hypomorphic csw alleles (Allard et al.,1998). This residual activity of Csw^C583S is proposed to result from its ability to partially mimic the effect of dephosphorylating a substrate by binding to it tightly(Allard et al., 1998). Spry binds Csw^C583S and could be such a substrate(Fig. 5D,E). If so, this could explain the lack of suppression of Csw^C583S phenotype by reduction in spry dose: decreasing spry levels would not reduce spry function under conditions in which it is already trapped in an inactive or partially inactive form by Csw^C583S.

Csw/SHP-2 binding and dephosphorylation of Spry creates an interesting regulatory circuit downstream of RTKs (Fig. 6A). Both components of the circuit are induced and activated following receptor activation, Csw/SHP-2 by SH2 domain interactions with phosphotyrosines, and Spry proteins by transcriptional induction of their genes and tyrosine phosphorylation of the proteins. One induced component(Spry) is a signaling inhibitor, the other (Csw/SHP-2) is a signaling promoter that acts by inactivating the inhibitor.

Why does a signaling pathway induce both a feedback inhibitor and a protein that inactivates it? One possibility is that this double-negative circuit provides a mechanism for rapidly resetting the signaling system: the inhibitor terminates signaling and the deactivator restores the inhibitor to its original (inactive) state, readying the cell for another round of signaling. This may be important when cells experience successive waves of signaling,such as the waves of EGFR and Sevenless signaling in eye development(Freeman, 1996).

Another possibility is that the double-negative circuit allows precise control of the signal output profile (Fig. 6B). In the absence of feedback, the response to a signal is simple and sustained, increasing monotonically until reaching saturation(black curve). If a basic negative-feedback system is operative, the magnitude and duration of the response are limited, generating a parabolic response profile (red curve). However, if the pathway contains both a feedback inhibitor (Spry) and an inducible component (Csw/SHP-2) that deactivates it,this creates more complex output profiles, such as the irregularly shaped curve observed for MAPK activation following FGFR activation in HEK293 cells(Fig. 1J; Fig. 6B, green curve). By altering activity of individual feedback components, other complex profiles can be generated (Fig. 1J). If cells can distinguish different profiles, as some cells distinguish different calcium oscillations (Lewis,2003), this could lead to different outcomes. The shape of the RTK response profile could be as important to outcome as the magnitude and duration of the response. In a similar way, differential induction of individual components of a double-negative feedback circuit can transform simple signaling gradients into complex spatial patterns of signal output.

Although our results imply that the Csw/SHP-2-Spry circuit operates in a variety of RTK signaling processes, it is unlikely to operate in all such events. Csw and SHP-2 are widely expressed and required in many and perhaps all RTK signaling processes, whereas Spry genes are expressed in more limited domains during development and appear to function in only a subset of such processes. For example, expression of spry genes in Drosophila and mouse embryos is largely confined to FGF signaling centers (Hacohen et al., 1998; Minowada et al., 1999).

In RTK signaling processes where Spry proteins are not expressed, Csw/SHP-2 must stimulate signaling by dephosphorylating other substrates, such as the autophosphorylation site on EGFR and Torso/PDGFRs that recruits RasGAP(Cleghon et al., 1998; Ekman et al., 2002; Agazie and Hayman, 2003b) or sites on Src kinase regulators (Ren et al., 2004; Zhang et al.,2004). In some pathways, such as the Torso pathway, more than one Csw/SHP-2 substrate is likely to be present and regulated by the enzyme(Cleghon et al., 1998; Casci et al., 1999). The specific substrates present in each cell should alter signaling kinetics in different ways, creating a rich diversity of output profiles. It will be interesting to determine the extent and importance of this diversity, and whether alterations in signal output profile and Spry feedback inhibition contribute to pathogenesis of human diseases associated with misregulation of SHP-2 activity (Tartaglia et al.,2001; Musante et al.,2003).

Work was supported by National Institutes of Health grants (to M.A.S. and M.A.K.), National Science Foundation Predoctoral Fellowships (to R.K.S.-B. and S.J.T.) and National Cancer Institute Training Grant (to L.A.J.). M.A.K. is an Investigator of Howard Hughes Medical Institute. R.K.S.-B. conceived the project. All authors conceived and designed experiments. L.A.J., R.K.S.-B. and S.J.T. performed experiments. All authors analyzed data and wrote the paper.