Mutational Analysis of the SRC Homology 2 Domain Protein-tyrosine Phosphatase Corkscrew*

The SRC homology 2 (SH2) domain protein-tyrosine phosphatase, Corkscrew (CSW) is required for signaling by receptor tyrosine kinases, including the Sevenless receptor tyrosine kinase (SEV), which directs Drosophila R7 photoreceptor cell development. To investigate the role of the different domains of CSW, we constructed domain-specific csw mutations and assayed their effects on CSW function. Our results indicate that CSW SH2 domain function is essential, but either CSW SH2 domain can fulfill this requirement. We also found that CSW and activated SEV are associatedin vivo in a manner that does not require either CSW SH2 domain function or tyrosine phosphorylation of SEV. In contrast, the interaction between CSW and Daughter of Sevenless, a CSW substrate, is dependent on SH2 domain function. These results suggest that the role of the CSW SH2 domains during SEV signaling is to bind Daughter of Sevenless rather than activated SEV. We also found that although CSW protein-tyrosine phosphatase activity is required for full CSW function, a catalytically inactive CSW is capable of providing partial function. In addition, we found that deletion of either the CSW protein- tyrosine phosphatase insert or the entire CSW carboxyl terminus, which includes a conserved DRK/GRB2 SH2 domain binding sequence, does not abolish CSW function.

Receptor tyrosine kinases (RTKs) 1 regulate cellular growth and differentiation in response to extracellular signals. The binding of a specific ligand to the extracellular domain of an RTK leads to a well characterized series of biochemical events, including RTK dimerization, activation of the cytoplasmic tyrosine kinase domain, and phosphorylation of the RTK on specific tyrosine residues (reviewed in Ref. 1). The activated and phosphorylated RTK then regulates key cellular target proteins. Frequently, these targets contain either SRC homol-ogy 2 (SH2) domains or phosphotyrosine binding domains that bind to specific phosphorylated tyrosine residues of the RTK. The recruitment of such SH2-and phosphotyrosine binding domain-containing proteins regulates their function and leads to the activation of important intracellular signaling pathways. For example, the interaction of the SH2 domain-containing protein GRB2 with activated RTKs leads to the stimulation of the RAS pathways (2,3).
Among the key signaling proteins, the function of which is required during RTK signaling, is a class of protein-tyrosine phosphatases that includes SHP2 in mammals and Corkscrew (CSW) in Drosophila (reviewed in Ref. 4). In addition to their protein-tyrosine phosphatase (PTP) domains, these proteins are characterized by the presence of two SH2 domains at their amino terminus and a poorly conserved carboxyl-terminal "tail" region. In addition, CSW contains a domain of approximately 150 amino acids that interrupts its PTP domain and is not present in SHP2 (5). The first evidence that SHP2/CSW proteins were important for RTK signaling came from the identification of CSW as a maternally contributed, positively acting component during signaling by the Drosophila TORSO RTK (5). Subsequent studies have implicated CSW in signaling by several other Drosophila RTKs including SEV and the Drosophila epidermal growth factor receptor (6,7). In vertebrates, SHP2 has been shown to participate in signaling by the epidermal growth factor, PDGF, fibroblast growth factor, and insulin receptors (reviewed in Refs. 4 and 8).
Two potential mechanisms for the action of SHP2/CSW during RTK signaling have been proposed. In the case of certain RTKs, such as the PDGF receptor, RTK autophosphorylation creates binding sites for the SH2 domains of SHP2 (9,10). Once bound to receptor, SHP2 becomes phosphorylated at a particular tyrosine residue in its tail region. Phosphorylated SHP2 can then be bound by the SH2 domain of GRB2. The recruitment of GRB2 to the SHP2⅐RTK complex is proposed to lead to the activation of the RAS signaling pathway. Studies of signaling by the SEV RTK in Drosophila have led to an additional model for CSW action during RTK signaling (6,11). The activation of SEV induces neuronal differentiation of the R7 photoreceptor precursor cell (for reviews, see Refs. [12][13][14]. In the R7 cell, CSW acts as a positive component of the signaling pathway initiated by SEV (6). Genetic epistasis experiments have demonstrated that CSW function is required in the developing R7 cell even when constitutively activated forms of RAS and RAF are used to initiate the SEV signaling cascade. These experiments have suggested that although CSW may be involved in RAS activation, CSW must also play a role either downstream of RAS/RAF activation or in a parallel signaling pathway that acts in conjunction with the RAS/MAPK pathway.
One approach to understanding the role of CSW/SHP2 during RTK signaling has been to identify substrates for their phosphatase activity. A CSW substrate, the Daughter of Sev-enless protein (DOS), was identified as a tyrosine-phosphorylated protein trapped in a complex with a catalytically inactive version of CSW (11). DOS contains an amino-terminal pleckstrin homology domain and is most similar to the family of multiadaptor proteins that includes GAB-1 and IRS-1 (15). The analysis of mutations that inactivate the dos gene has indicated that DOS is a positive component of the SEV pathway (11,15). Together, these biochemical and genetic studies have led to the proposal that DOS must be dephosphorylated by CSW during SEV signaling. Searches for SHP2 substrates in mammalian cells have also led to the identification of a number of potential substrates. At present, these include the transmembrane protein SHPS1/SIRP1␣, IRS-1, the PDGF receptor, and several tyrosine-phosphorylated proteins with sizes ranging from 80 to 120 kDa (reviewed in Refs. 4 and 8).
In order to define further the role of CSW/SHP2 during RTK signaling, we have mutated each of the domains of CSW and investigated their effects on SEV signaling and Drosophila development. In this report, we provide evidence that the the PTP domain and at least one functional SH2 domain are essential for CSW function. Analysis of the ability of either mutant or wild type CSW to interact with DOS and SEV indicates that the association in vivo of CSW with SEV is not dependent on either the tyrosine phosphorylation of SEV or the presence of functional SH2 domains in CSW. In contrast, the association of DOS and CSW requires functional CSW SH2 domains. These results suggest that the role of the CSW SH2 domains is not in binding to SEV but is instead in the recognition of substrates, such as DOS. In addition, we found that a deletion of either the unique CSW PTP insert region or the conserved GRB2 binding site-containing carboxyl terminus of CSW does not appreciably reduce CSW function during signaling by SEV, TORSO, and perhaps all RTKs.

MATERIALS AND METHODS
Genetics and Histology-Drosophila culture and crosses were carried out using standard procedures. The csw  and csw C114 alleles used in this study have been characterized as genetic nulls, whereas the csw 6 allele retains partial activity. All three alleles are recessive lethal (5). Tissue sectioning and scanning electron microscopy was performed as described (6).
Generation of csw Mutants and P-element Transformation-Mutations in csw were generated using oligonucleotide mutagenesis as described previously (6). The Arg to Lys substitutions in the SH2 domains of CSW were generated using oligonucleotide 5Ј-AGGAGGAGAGCTT-GGCGAGGAAG-3Ј for csw R32K and oligonucleotide 5Ј-TCTGAGATTC-CTTGACGAGAAAC-3Ј for csw R137K . The csw R32K/R137K construct was made by subcloning fragments from the constructs containing the single point mutations. Deletions of the CSW PTP domain insert (amino acids 290 -444) and the carboxyl terminus (amino acids 551-841) were generated using the oligonucleotides 5Ј-GATGTAGGTCTTAAA-CATCTCCAGGTCGCCGTCGGTGGGCAG-3Ј and 5Ј-GACTTAGAAAT-TCATTTGTCAGGCGATCAGGGTCTGTATATA-3Ј, respectively. The csw construct encoding only the SH2 domains (csw SH2SH2 ) was made by placing a stop codon after residue 231 in the csw cDNA.
The csw mutant cDNAs were subcloned into the transformation vector pKB267, which contains a hybrid promoter (SE) consisting of sevenless enhancers and the hsp70 promoter (16). This vector directs gene expression in a subset of cells in the eye, including R3, R4, and R7, as well as general expression in all cells following heat shock induction. The csw constructs were introduced into the germline by P elementmediated transformation using standard procedures. All flies used in this study were w 1118 . The constructs were marked by the presence of the [w ϩ ] gene in the P-element.
Assays of CSW Function-To assay for rescue of the zygotic lethality associated with homozygous csw loss-of-function mutations, larvae from a cross between males carrying the mutant P[SE-csw] construct and csw  /FM7 females were heat shocked daily for 1 h at 37°C until eclosion. The progeny were scored for the presence of csw 13-87 /Y males, which indicated that the mutant P[SE-csw] construct could provide zygotic CSW function. To assay maternal CSW function, the csw 13-87 /Y; P[SE-csw] males obtained from the zygotic rescue experiment were crossed with csw 13-87 /FM7 females. The resulting progeny were heat shocked daily during their development to induce expression of the mutant P[SE-csw] construct. This resulted in the production of homozygous csw  females that carried the mutant P[SE-csw] construct. These zygotically rescued females were then scored for their ability to produce normal progeny when crossed to wild type flies. The production of normal progeny indicated that the mutant CSW was able to provide maternal CSW function. To measure the effect of mutant P[SE-csw] constructs on the ability of P[SE-csw G547E ] to inhibit R7 development, ommatidia from flies heterozygous for the mutant P[SE-csw] construct and P[SE-csw G547E ] were sectioned and the number of R7 cells was quantitated.
Cell Culture and Protein Analysis-Immunoblotting, immunoprecipitation, and glutathione S-transferase fusion protein precipitation experiments were carried out as described previously (11). The csw and dos cDNAs were cloned into pAT-Hygro, which drives constitutive expression from the Actin5C promoter (6). The SEV S11 and SEV S11KM constructs were described previously (17). The SL2 cells were selected in either Hygromycin B (0.2 mg/ml) or G418 (1 mg/ml). For heat shock induction of SEV S11 and SEV S11KM , cells were incubated at 37°C for 30 min and then returned to 23°C. The antibodies used in this work were anti-CSW-CT (6), anti-DOS-1, which was raised against the carboxyl terminus of DOS (amino acids 856 -878), and anti-PY20 (Signal Transduction Laboratories).
The pGEX-TK-CSW SH2SH2 domain construct contains amino acids 1-217 of CSW (11). The csw R32K and csw R137K SH2 domain mutations were transferred as restriction fragments into pGEX-TK-CSW SH2SH2 . The fusion proteins were affinity purified on glutathione-agarose beads (Sigma). For precipitation of DOS, 4 g of glutathione S-transferase fusion protein was incubated with cell lysates (10 7 cell equivalents) that were prepared from pervanadate-treated SL2-DOS cells.

RESULTS
Our approach to investigating the functional role of each of the domains of CSW was to introduce domain-specific mutations into a construct that contained a csw cDNA expressed under the control of a hybrid sevenless enhancer/heat shock promoter transcriptional control element (16). This transcription unit (SE) directs constitutive expression in a subset of cells in the Drosphila eye, including the photoreceptor R7, R3, and R4 precursor cells and all four of the cone cell precursors. In addition, the SE cassette also provides pulses of ubiquitous expression when the flies are grown at 37°C for short periods.
The mutated SE-csw constructs were introduced into the genome by P element-mediated germline transformation and assayed for function. In order to control for possible variation in transcription due to P element insertions at different sites in the genome, several different insertions were analyzed for each construct. In each case, equivalent results were obtained. The function of the mutated CSW proteins was tested in three ways. The first assay took advantage of a particular allele of csw that we had previously isolated in a genetic screen for mutations that attenuate signaling by SEV (18). Although in vitro analysis of the CSW G547E protein indicates that it retains the ability to dephosphorylate model substrates such as pnitrophenyl phosphate (data not shown), the mutant protein in vivo lacks normal CSW function and inhibits the functioning of wild type CSW (6). The expression of CSW G547E in the developing eye under SE transcriptional control blocks development of the R3, R4, and R7 photoreceptors and leads to flies with small eyes that are rough and disordered (6) (see Fig. 3). Because expression of wild type CSW under SE control suppresses the effects of SE-csw G547E , we assayed the signaling ability of mutant CSW proteins by determining whether their expression in the sevenless transcriptional pattern could also suppress the SE-csw G547E phenotype.
The second assay took advantage of previous observations that animals that lack zygotic CSW function die during pupation due to severe developmental abnormalities in many of their tissues (5, 7). Heat shock promoter-driven expression was therefore used to determine whether the mutant proteins could provide sufficient CSW function to rescue these defects in animals that otherwise lacked CSW function. The test was conducted by crossing flies carrying the SE-csw transgene to females heterozygous for a loss-of-function csw allele in order to generate males that were hemizygous for the defective csw allele but carried the SE-csw transgene. For these assays, we generally used both the csw  and csw C114 alleles. Each of these alleles has been genetically characterized as a loss-offunction allele (5).
The third assay was applicable only to the cases in which expression of the mutated protein was able to rescue the lethality of csw animals. In these cases, the resulting csw 13-87 /Y males carrying the SE-csw transgene were crossed to csw  / FM7c females in order to generate females that were homozygous for the csw 13-87 allele but were alive due to rescue by the expression of the mutant CSW. These females were then tested for their fertility. Since maternal contribution of CSW is essential for TORSO signaling and proper embryonic development, the fertility of these females is a sensitive assay for the ability of the mutant CSW to function during TORSO signaling (5).
SH2 Domain Function Is Essential for CSW Action-SH2 domains possess an invariant arginine residue that is contained in a highly conserved motif (FLVRES) (19). Mutagenesis and crystallography studies have shown that this arginine residue is crucial for the interaction between an SH2 domain and the phosphotyrosine of its target binding sequence (20). We therefore tested the importance of the CSW SH2 domains by mutating the corresponding arginines of each SH2 domain (Arg 32 and Arg 137 ) to lysines and assaying the ability of the resulting proteins to function in the assays described earlier. A construct containing mutations in both SH2 domains (CSW R32K/R137K ) was unable to either rescue the zygotic lethality of male flies lacking csw function (csw 13-87 /Y) or to suppress the eye defects in flies carrying the SE-csw G547E cassette ( Fig.  1 and Table I).
These results demonstrated the importance of CSW SH2 domain function but did not indicate whether both SH2 domains are required. We addressed this question by testing the function of CSW proteins in which only a single SH2 domain was inactivated. Similar results were obtained with both CSW R32K and CSW R137K . In each case, the mutant protein was capable of providing sufficient CSW function to rescue the lethality of male flies (csw 13-87 /Y) otherwise lacking CSW function and of providing maternal CSW function in eggs derived from homozygous csw 13-87 females (Fig. 1). Furthermore, the rescued male flies had smooth eyes with normally constructed ommatidia (data not shown). These results indicated that each of the individual SH2 mutant proteins was capable of providing sufficient CSW function to fulfill the normal roles of CSW and suggest that the two SH2 domains may perform redundant functions.
In contrast to their ability to function as the sole maternal or zygotic source of CSW function during normal development, neither of the single SH2 domain mutated proteins was effective at suppressing the rough-eyed phenotype of SE-csw G547E flies (Table I). This may indicate that expression of CSW G547E interferes with normal CSW function by using both of its SH2 domains to bind to an essential regulator or target of CSW. The CSW R32K and the CSW R137K proteins may be less effective competitors than wild type CSW because of the reduced binding affinity caused by the loss of one of the SH2 domain interactions.
The SH2 Domains Function in Substrate Recognition during SEV Signaling-One possible role for the SH2 domains of CSW during SEV signaling is to allow CSW to bind to activated SEV. Previous studies of the role of SHP2 in signaling by the PDGF and epidermal growth factor receptors has indicated that the SHP2 SH2 domains bind directly to sites of autophosphorylation on the activated RTKs (21)(22)(23) and that this interaction can lead to the tyrosine phosphorylation of SHP2. We therefore asked whether the SH2 domains of CSW might be required to allow CSW to interact with activated SEV. We first sought to FIG. 1. csw constructs and their ability to provide CSW function. Shown is a schematic representation of the mutant constructs generated in this study. Open boxes, CSW SH2 domains; shaded boxes, CSW PTP domain; black box, first 90 amino acids from SRC64, which contain the myristylation site. Point mutations generated by in vitro mutagenesis are indicated by asterisks. The constructs were introduced into the germline and expressed from the hybrid SE cassette. Zygotic CSW function was scored by the ability of hemizygous csw  /Y males to survive in the presence of the csw transgene. A ϩ indicates Ͼ50% of the expected class survived. A Ϫ indicates Ͻ1% of the expected class survived. Maternal CSW function was scored by the ability of zygotically rescued csw  females to produce wild type progeny. A ϩ indicates Ͼ50% of the progeny were wild type. More than 200 animals were scored in each experiment. nd, not determined. examine whether SEV and CSW can be found in a complex in Drosophila tissue culture cells (Schneider line 2, SL2 cells). Because SL2 cells do not express detectable levels of SEV and express only barely detectable levels of CSW, we used SL2 cells that had been engineered to constitutively express elevated levels of CSW under the control of the Actin5C promoter and an activated form of SEV (SEV S11 ) under the control of a heat shock promoter (SL2-CSWSEV S11 cells). Heat shocking of these cells leads to the production of large amounts of catalytically active and tyrosine-phosphorylated SEV (11). Extracts of SL2-CSWSEV S11 cells were immunoprecipitated with monoclonal antibodies to SEV. Examination of these anti-SEV immunoprecipitates by immunoblotting with anti-CSW antisera revealed the presence of CSW ( Fig. 2A). While only a portion of the CSW within the SL2-CSWSEV S11 cells were found in the anti-SEV immunoprecipitates, the results nevertheless indicate that CSW and SEV can associate in Drosophila cells. Among the possible explanations for the low stoichiometry of the CSW⅐SEV complex are limited stability during the immunoprecipitation and the possible requirement for another cellular protein that was not overexpressed in the SL2-CSWSEV S11 cells. Despite the association of CSW and SEV S11 in SL2-CSWSEV S11 cells, CSW was not detectably phosphorylated and thus does not appear to be a target for activated SEV (data not shown). Furthermore, a catalytically inactive CSW (CSW C583S ) also associated with SEV S11 and failed to be tyrosine-phosphorylated, indicating that the lack of CSW phosphorylation was not due to autodephosphorylation (data not shown).
To further investigate whether CSW and SEV interact in a manner similar to that reported for SHP2 and the PDGF re-ceptor, we asked whether the association of CSW and SEV S11 depends on both the tyrosine phosphorylation of SEV and the SH2 domains of CSW. Two experiments were conducted. In the first, a line of SL2 cells (SL2-CSWSEV S11KM cells) was produced that constitutively expressed wild type CSW and inducibly expressed a catalytically inactive version of the SEV S11 protein. Anti-CSW immunoblots were then used to compare the level of CSW found in anti-SEV immunoprecipitates from SL2-CSWSEV S11 and SL2-CSWSEV S11KM cells. Contrary to our expectations, the amount of CSW complexed with SEV was not significantly different between the extracts of the two cell lines despite the lack of detectable tyrosine phosphorylation of SEV S11KM ( Fig. 2A and data not shown). These results indicated that the complex between CSW and SEV is unlikely to involve direct interactions between CSW SH2 domains and SEV phosphorylation sites. In order to more directly assay the participation of the SH2 domains of CSW in the interaction with SEV, we conducted similar co-immunoprecipitation assays with extracts from a SL2 cell line (SL2-CSW R32K/ R137K SEV S11 ) expressing CSW in which the SH2 domains were inactivated by mutation as described earlier. We found that the presence of the SH2 domain mutations did not markedly affect the CSW-SEV interaction. Together, these results indicate that the presence of functional CSW SH2 domains is not required for the association of CSW and SEV. Furthermore, these results raise the possibility that CSW and SEV might interact through another protein rather than directly with each other.
The existence of an SH2-and phosphotyrosine-independent complex containing both CSW and SEV suggests that a crucial role of the SH2 domains of CSW might be in some aspect of CSW function besides recognition of phosphorylated SEV. One possible role might be in the recognition of phosphorylated substrates. Because the only known substrate for the PTP activity of CSW is the DOS protein, we investigated whether the SH2 domains of CSW might interact directly with phosphorylated DOS. Because our mutational analysis of the SH2 domain requirements for CSW indicated that either of the SH2 domains can at least partially fulfill the SH2 domain requirements for CSW, our expectation was that each SH2 domain would be capable of binding to phosphorylated DOS. We tested this prediction by assaying the ability of glutathione S-transferase fusion proteins containing either wild type or mutated CSW SH2 domains to precipitate phosphorylated DOS. For these experiments, the source of phosphorylated DOS was an extract from SL2 cells that constitutively expressed DOS and had been treated with the PTP inhibitor pervanadate. Our previous results had indicated that DOS from such extracts was both tyrosine-phosphorylated and capable of interacting with CSW (11). The results shown in Fig. 2B indicate that the SH2 domains of CSW are capable of directly interacting with DOS. Furthermore, mutations of either SH2 domain result in a only a partial loss of DOS binding, whereas simultaneous mutation of both SH2 domains fully inactivates DOS binding (Fig.  2B). The ability of either SH2 domain to function in DOS binding is similar to our in vivo findings that the SH2 domains are at least partially redundant for CSW function. These results, therefore, are consistent with the idea that recognition of DOS or another substrate may be the primary function of the CSW SH2 domains.
The PTP Activity Is Important for CSW Function-Our previous studies had suggested that the PTP activity of CSW is essential for CSW function during photoreceptor development (6,11). In those studies, we had shown that expression of either of two CSW proteins carrying mutations in the PTP domain (CSW G547E and CSW C583S ) inhibit development of each of the photoreceptors in which they are expressed. Because each mu- tation affects the PTP domain of CSW, these results implied that CSW PTP activity is important for CSW function during photoreceptor development. However, these experiments did not directly test whether the catalytic activity of CSW was absolutely required for all CSW function. In order to verify that the catalytic activity of CSW is crucial for function, these mutant proteins were tested for their ability to rescue the lethality of male animals hemizygous for the null csw  allele. We also tested an additional version of CSW (CSW SH2SH2 ) in which the entire PTP domain and tail region were deleted (Fig. 1). SE expression of the CSW SH2SH2 protein yielded roughened eyes and photoreceptor defects that were similar to those seen in flies expressing either CSW G547E or CSW C583S (data not shown). We found that none of the three mutant proteins was able to rescue the viability of any csw 13-87 individuals (Fig. 1). These results indicate that CSW PTP domain function is crucial for CSW action.
CSW C583S Retains Partial Function in the Absence of Catalytic Activity-Despite this strong evidence that CSW PTP activity is needed for CSW action, one feature of our previous results was inconsistent with the idea that CSW function absolutely requires catalytic activity. As described above, the csw G547E , csw SH2SH2 , and csw C583S alleles each encode a protein that interferes with the function of wild type CSW. A probable explanation for these inhibitory effects is that the mutant proteins use their SH2 domains to compete with wild type CSW for binding to the CSW substrate DOS but then fail to dephosphorylate DOS. Surprisingly, expression of the CSW C583S protein has a much less profound inhibitory effect on photoreceptor development than does the expression of either the CSW G547E or CSW SH2SH2 proteins (Table I and Fig. 3). These results were puzzling because the csw C583S mutation changes an evolutionarily conserved cysteine residue that is absolutely required for the PTP activity (4,24). Thus, CSW C583S entirely lacks catalytic activity (data not shown), and hence, its expression might have been expected to yield the most severe phenotype.
These results suggested that CSW C583S might retain some of its signaling abilities despite its lack of catalytic activity. How-ever, another possible explanation for these results was that the CSW C583S protein was merely expressed at lower levels than either the CSW G547E or the CSW SH2SH2 protein. We tested this possibility by generating flies that carried both the P[SEcsw G547E ] and the P[SE-csw C583S ] transgenes. If the difference between the phenotypes caused by the two transgenes was merely a result of differences in the expression or stability of the two mutant CSW proteins, then the combined phenotype of both transgenes would be expected to be at least as severe as the P[SE-csw G547E ] phenotype alone. The eyes of the P[SEcsw G547E ], P[SE-csw C583S ] flies were less severely roughened than those of P[SE-csw G547E ] flies and more similar to those of P[SE-csw C583S ] flies (Fig. 3). These results suggested that the CSW C583S protein retains some of its function and can thus partially rescue the SE-csw G547E inhibitory phenotype.
Since the effect of expression of CSW C583S on the P[SEcsw G547E ] phenotype was relatively weak, we sought further confirmation of the positive signaling ability of the CSW C583S protein. Our approach was to create a version of CSW C583S that might possess greater function. Since our previous experiments had indicated that wild type CSW function could be enhanced by constitutive localization to the plasma membrane, we targeted CSW C583S to the plasma membrane with the myristylation sequences at the amino terminus of the SRC64 protein.
The resulting protein, CSW C583Ssrc90 , was then placed under the control of the SE expression cassette, and the resulting P element (P[SE-csw C583Ssrc90 ]) was inserted into the genome by P element-mediated transformation. The first apparent difference between the effects of the P[SE-csw C583Ssrc90 ] element and either the P[SE-csw C583S ] or P[SE-csw G547E ] elements was that presence of the P[SE-csw C583Ssrc90 ] element did not yield a rough eye phenotype (Fig. 3). Sectioning of the eyes of the P[SE-csw C583Ssrc90 ] flies did reveal minor abnormalities, usually the absence of the R7 cell, in approximately 10% of the ommatidia.
We then generated flies carrying both P[SE-csw C583Ssrc90 ] and the P[SE-csw G547E ] elements. The resulting flies were entirely smooth-eyed and showed few defects in their photoreceptor development (Fig. 3). These results indicated that the abil-

FIG. 3. A catalytically inactive CSW retains function.
A-E, scanning electron micrographs of adult eyes. All genotypes are hemizygous for the w 1118 allele and heterozygous for the various SE-csw constructs. A, SE-csw/ϩ; expression of wild type CSW does not affect the wild type ommatidial array. However, the normal ommatidial array is disrupted due to a loss of photoreceptor cells, including R3, R4, and R7, in flies that are heterozygous for SE-csw C583S (B) or SEcsw G547E (D). C shows the phenotype associated with heterozygous expression of SEcsw C583Ssrc90 , which encodes a catalytically inactive version of CSW that is targeted to the plasma membrane by myristylation sequences. The eye is nearly wild type in appearance. E and F show the suppression of the SE-csw G547E phenotype by co-expression of SE-csw C583S and SE-csw C583Ssrc90 , respectively.
ity of the CSW C583Ssrc90 protein to rescue the P[SE-csw G547E ] phenotype was nearly comparable to that of wild type CSW and thus supported the idea that some of the functions of CSW can be provided by a catalytically inactive CSW. We further investigated the ability of CSW C583Ssrc90 to function by examining whether the P[SE-csw C583Ssrc90 ] element could rescue the lethality of animals hemizygous for either the null csw 13-87 allele or for the hypomorphic csw 6 allele (Table II). csw 6 is a weak allele of csw that supports development to a later stage than loss-of-function alleles, such as csw   (7). CSW C583Ssrc90 expression was unable to restore the viability of the csw  individuals. In contrast, CSW C583Ssrc90 expression was able to efficiently rescue the lethality of csw 6 individuals. These results demonstrated that a catalytically inactive CSW protein retains a significant level of CSW function.
The PTP Insert and the Carboxyl-terminal Tail of CSW Are Not Required for CSW Function-One difference between the structures of CSW and its vertebrate homolog SHP2 is that CSW contains an insert of approximately 150 amino acids in the amino-terminal region of the PTP domain (5). In order to determine whether the insert performs any crucial function, an allele of csw was constructed in which this region, amino acids 290 -444, was deleted. The mutated CSW protein (CSW ⌬in ) was expressed under SE control and assayed for function in each of the three assays. In each case, the action of CSW ⌬in was indistinguishable from wild type CSW ( Fig. 1 and Table I).
These results indicate that the PTP insert region of CSW is dispensable for CSW function. This conclusion is consistent with the previously reported observation that expression of mammalian SHP2, which lacks this region, can substitute for CSW function during Drosophila development (7). Nevertheless, our experiments do not eliminate a possible role for the insert region in modulating CSW function in a manner that is too subtle to be detected in these assays.
As described earlier, one potential function for the mammalian homolog of CSW, SHP2, is as an adaptor protein used to recruit GRB2⅐SOS complexes to activated RTKs such as the PDGF receptor (9, 10). The first steps in this proposed mechanism are SHP2 binding to the activated RTK via its SH2 domains. SHP2 is then phosphorylated on a particular tyrosine residue (Tyr 542 in hSHP2) in its carboxyl-terminal tail. This phosphorylation site then serves as a docking site for the SH2 domain of GRB2. Although the lack of CSW tyrosine phosphorylation in response to SEV activation suggests that this mode of action does not apply to SEV signaling, two lines of evidence have suggested that CSW might function in this manner during signaling by other RTKs. First, the carboxyl-terminal tail of CSW includes a tyrosine residue (Tyr 666 ) contained within a sequence closely related to the sequences around Tyr 542 in hSHP2 (5). Second, the activated TORSO RTK has been shown to be phosphorylated at a site (Tyr 630 ) that is required for the association of CSW with TORSO and the subsequent phosphorylation of CSW (25).
In order to test the role of the carboxyl-terminal tail of CSW, we generated a mutant version of CSW (CSW ⌬tail ) in which the entire tail region (amino acids 650 -841) was deleted. The mutant protein was placed under the control of the SE expression cassette, inserted into the germline, and tested in each of the three assays of CSW function. The activity of CSW ⌬tail was equivalent to that of wild type CSW in each assay ( Fig. 1 and Table I). These results indicate that the tail region of CSW is dispensable for CSW function and suggest that the proposed mechanism for SHP2 action as an adaptor protein after phosphorylation of Tyr 542 is either not applicable to CSW or is not important for CSW function. These results suggest that the ability of SHP2 to act as an adaptor molecule may be similarly dispensable for SHP2 function, as previously suggested (26). In each case, the presence of other mechanisms for recruiting and activating the DRK⅐SOS complex might allow RTK signaling systems to function in the absence of the putative adaptor role of CSW. However, the possibility that CSW does act as an adaptor after becoming phosphorylated at presently unidentified DRK binding sites cannot be eliminated. DISCUSSION We have investigated the functional contribution of each of the domains of CSW by generating CSW proteins with domainspecific mutations and evaluating the function of these proteins. Our results indicate that SH2 domain function is required for in vivo CSW function but that this requirement can be met by either CSW SH2 domain. Furthermore, the ability of either SH2 domain to support CSW function implies that the crucial binding partner(s) for the CSW SH2 domain is likely to possess binding sites for both SH2 domains. Candidates for such binding partners during SEV signaling included SEV itself and DOS, a putative multi-adaptor protein and CSW substrate. In this report, we have shown that although SEV and CSW can be co-immunoprecipitated from Drosophila cells, their association is independent of CSW SH2 domain function and SEV tyrosine phosphorylation. Thus, SEV is unlikely to be a binding partner for the CSW SH2 domains. We cannot presently evaluate whether the observed phosphotyrosine-independent association of CSW and SEV actually occurs in the developing R7 cell or is important for SEV signaling. In contrast to SEV, DOS does possess binding sites for each of the CSW SH2 domains and thus is a good candidate to be an important target for the CSW SH2 domains.
The ability of either SH2 domain to support CSW function suggests that the two domains have similar roles during CSW action. This finding contrasts with biochemical studies of SHP1, a mammalian SH2 domain PTP closely related to CSW and SHP2 that negatively regulates signaling by many cytokine receptors and RTKs (27,28). These studies showed that deletions of the amino-terminal SH2 domain of SHP1 led to the activation of SHP1 PTP activity, whereas deletion of the carboxyl-terminal SH2 domain did not markedly effect PTP activity. These results have led to the proposal that the activity of SHP1 can be regulated in two ways in response to the activation of different receptors. Receptors, such as the erythropoietin receptor, which contain a binding site for the amino-terminal SH2 domain of SHP1, could both recruit and activate SHP1, thus leading to the rapid termination of receptor signaling. In contrast, receptors possessing only binding sites for the carboxyl-terminal SH2 domain would be expected to recruit but not activate SHP1. This might result in a slower termination of signaling. Although it is not known whether such differential regulation of CSW PTP activity occurs, several studies have shown that the binding of SHP2 to phosphotyrosyl peptide ligands can increase the PTP activity of SHP2 (29 -33). However, our results suggest that whether or not such regulation of CSW occurs, neither the binding of CSW to ligands possessing   or csw 6 males to survive in the presence of the transgene as described under "Materials and Methods." Construct Rescue csw  csw 6 Ϫ ϩ only one CSW SH2 domain binding site nor the activation of CSW PTP activity by occupancy of its amino-terminal SH2 domain is necessary for sufficient CSW function to support normal development. Two lines of evidence suggest that the activity of the PTP domain is also important for CSW function. First, the expression of CSW proteins that either contain mutations in the catalytic domain (CSW G547E or CSW C583S ) or lack the entire PTP domain (CSW SH2SH2 ) inhibits the action of wild type CSW. Second, none of these three proteins is able to support the development of animals that otherwise lack CSW function. The ability of CSW SH2SH2 to act as an inhibitor suggests that the mechanism of inhibition is competition with wild type CSW for SH2 domain binding partners, such as DOS. Presumably, these PTP domain mutant proteins then fail to dephosphorylate the target molecules and thus fail to support effective RTK signaling.
Despite this clear evidence supporting a crucial role for PTP activity in CSW function, our results have indicated that CSW proteins that lack catalytic activity can display significant CSW function. In particular, we have shown that expression of the CSW C583S protein can partially rescue the inhibitory effects of the CSW G547E protein. The ability of CSW carrying the C583S substitution to function became even more apparent when the protein was targeted to the plasma membrane with SRC64 myristylation sequences. CSW C583Ssrc90 almost completely eliminated the dominant inhibitory effects of CSW G547E expression and also supported the development of animals that would normally fail to survive due to the presence of the hypomorphic csw 6 allele.
Since no evidence for CSW function was detected with expression of either of the other PTP domain mutant CSW proteins, the ability of CSW C583S proteins to signal is likely due to a unique property of the C583S mutation. Previous crystallographic and enzymatic studies have indicated that one unique property of PTPs in which the catalytic cysteine is mutated to serine is that such molecules can still bind to substrates with high affinity. For instance, our previous work has shown that CSW C583S can form a stable complex with tyrosine-phosphorylated DOS (11). One possible explanation for the ability of CSW C583S and CSW C583Ssrc90 to function may be that these proteins can partially mimic the effects of substrate dephosphorylation by merely binding to the site of phosphorylation. For example, if DOS is normally maintained in an inactive state by an intramolecular interaction involving a phosphorylated tyrosine residue, then either dephosphorylation by wild type CSW or binding by an inactive CSW might relieve the inhibitory effect. However, for such a model to be consistent with our finding that expression of CSW C583S causes a mildly inhibitory phenotype in an otherwise wild type animal, one must postulate that binding by the catalytically inactive CSW is less effective than dephosphorylation by wild type CSW.
Since this work was submitted, a study examining the functional importance of the various domains of SHP2 during fibroblast growth factor-induced mesoderm induction in Xenopus has been published (34). Mutated SHP2 proteins were assayed for their ability to overcome the inhibitory effects on fibroblast growth factor signaling caused by an SHP2 carrying a deletion within the PTP domain. This test was similar to our assay of mutated CSW function in the presence of the inhibitory CSW G547E protein and yielded similar results. In each case, the mutation of either SH2 domain dramatically reduced SHP2/ CSW function. However, the Xenopus studies were able to further show that whereas amino-terminal SH2 domain mutations abolished SHP2 function, SHP2 lacking carboxyl-terminal SH2 domain function retained partial function that was apparent when the protein was expressed at elevated levels. Our inability to readily vary the level of CSW expression in the developing eye has prevented us from similarly determining whether CSW proteins lacking carboxyl-terminal SH2 domain function also retain a low level of activity in our assay. However, our studies were able to show that the function of each of the SH2 domains was dispensable when assayed by the ability to provide the sole source of CSW rather than by the ability to overcome the effects of an inhibiting protein.