Cell, Vol. 70, 431-442,

August

7, 1992, Copyright

0 1992 by Cell Press

The SH2 and SH3 Domain-Containing Protein GRB2 Links Receptor Tyrosine Kinases to ras Signaling E. J. Lowenstein,’ R. J. Daly,” A. G. Batzer,* W. Li, l 6. Margolis,’ R. Lammers,t A. Ullrich,t E. Y. Skolnik,’ D. Bar-Sagi,* and J. Schlessinger’ l Department of Pharmacology New York University Medical Center New York, New York 10016 tMax-Planck lnstitut fijr Biochemie Am Klopfrspitz 16A 8033 Martinsried Germany *Cold Spring Harbor Laboratories 1 Bungtown Road Cold Spring Harbor, New York 11724

Summary A cDNA clone encoding a novel, widely expressed protein (called growth factor receptor-bound protein 2 or GRBP) containing one src homology 2 (SH2) domain and two SH3 domains was isolated. lmmunoblotting experiments indicate that GRB2 associates with tyrosine-phosphorylated epidermal growth factor receptors (EGFRs) and platelet-derived growth factor receptors (PDGFRs) via its SH2 domain. Interestingly, GRB2 exhibits striking structural and functional homology to the C. elegans protein sem-5. It has been shown that sem-5 and two other genes called 181-23 (EGFR like) and let-50 (ras like) lie along the same signal transduction pathway controlling C. elegans vulva1 induction. To examine whether GRB2 is also a component of ras signaling in mammalian cells, microinjection studies were performed. While injection of GRB2 or H-ras proteins alone into quiescent rat fibroblasts did not have mitogenic effect, microinjection of ORB2 together with H-ras protein stimulated DNA synthesis. These results suggest that GRB2/sem-5 plays a crucial role in a highly conserved mechanism for growth factor control of ras signaling. Introduction Polypeptide growth factors mediate their physiological responses by binding to and activating cell surface receptors with intrinsic protein tyrosine kinase activities (reviewed in Ullrich and Schlessinger, 1990). Both receptor activation and tyrosine autophosphorylation were shown to be mediated by an allosteric intermolecular process (reviewed in Schlessinger, 1988). Receptor autophosphorylation appears to be essential for association with a group of cytoplasmic target proteins (reviewed in Koch et al., 1991; Heldin, 1991; Margolis, 1992). Following binding, certain target proteins, such as phospholipase Cr (PLC-$ become tyrosine phosphorylated and activated (Meisenhelder et al., 1989; Wahl et al., 1989; Margolis et al., 1989; Kazlauskas and Cooper, 1989; Kim et al., 1991). Other

target proteins, such as phosphoinositide-3 (PI-3) kinaseassociated ~85, function as adaptors or regulatory subunits to couple tyrosine kinase receptors to effector proteins (Cantleyet al., 1991; Hu et al., 1992; McGlade et al., 1992). The association between signaling proteins and growth factor receptors is strictly dependent upon tyrosine phosphorylation of specific short sequence motifs. A short consensus sequence was described in the kinase insert region of the platelet-derived growth factor (PDGF) and colony-stimulating factor 1 receptors and shown to act as a binding site for PI-3 kinase-associated ~85 (Kazlauskas and Cooper, 1990; Cantley et al., 1991; Escobedo et al., 1991 a; Reedijk et al., 1992). Another sequence motif was found in the carboxy-terminal tails of the fibroblast growth factor (FGF) and epidermal growth factor receptors (EGFRs) and shown to act as a binding site for PLC-r (Margolis et al., 1990a, 1990b; Mohammadi et al., 1991, 1992; Rotin et al., 1992a). These autophosphorylation sites in growth factor receptors represent recognition structures for specific target proteins containing src homology 2 (SH2) domains. SH2 domains are conserved sequences of approximately 100 aa found in various signaling molecules and oncogenic proteins (reviewed in Koch et al., 1991; Heldin, 1991; Margolis, 1992). SH2 domains have been found in a diverse group of proteins, some containing enzymatic activities, such as PLC-y, GTPase-activating protein, and pp60c-wc (Stahl et al., 1988; Suh et al., 1988; Sadowski et al., 1986; Vogel et al., 1988; Trahy et al., 1988) while others, such as rick, v-crk, and ~85 (Lehmann et al., 1990; Mayer et al., 1988, 1992; Skolniket al., 199l;Otsuet al., 1991; Escobedoet al., 1991b), lack any apparent enzymatic activity. Binding of SH2 domains to tyrosine-phosphorylated regions of growth factor receptors is thought to provide a common mechanism by which diverse enzymatic and regulatory proteins can interact specifically with growth factor receptors and thereby couple growth factor stimulation to multiple intracellular signaling pathways (reviewed in Koch et al., 1991; Heldin, 1991; Margolis, 1992). SH2 domains are often accompanied by a stretch of conserved sequence of approximately 50 aa, termed SH3 domain, whose function is not currently known. We have recently utilized the tyrosine-autophosphorylated carboxy-terminal tail of EGFR as a specific probe for direct expression/cloning of novel EGFR-binding proteins from hgtl 1 libraries in a method we refer to as cloning of receptor targets (CORT). The first growth factor receptorbound (GRBl) protein cloned by this method was shown to be the human counterpart of PI-3 kinase-associated p85a (Skolnik et al., 1991; Escobedo et al., 199lb; Otsu et al., 1991). We describe the cloning and characterization of the second protein cloned by the CORT method, GRBP. GRB2 is a small, widely expressed protein, whose entire sequence is composed of a single SH2 domain flanked by two SH3 domains. Here we demonstrate GRBP association with ligand-activated EGFRs and PDGF receptors

Cell 432

A.

813

GCCAGTGAATTCGGGGGCTCAGCCCTCCTCCCTCCCTTCCCCCTGCTTCAGGCTGCTGAG

60

GACGAGCTGAGCTTCMMGGGGGGACATCCTCMGGTTTTGMCGh4GAATGTGATCAG

180 34

PEJ.SF~~CQ~L~VJ.~EECDO MCTGGTA~GGCAGAGCTTAATGGAAAAGACGGCTTCATTCCCAAGAACTACATAGAA

240 54

ta”YIAELUClc~Gr~PlcNYIE ATTGAMCCACATCCGTGGT’TTTTTGGCAAAATCCCCAGAGCCAAGGCAGAAGAAATGCTT

300 74

MKPHP~FIGIIPrcAK*EEYL

360 94 TTCTCCCTCTCTGTCMGTTTW~CGATGTGCAGCACTT

818

420 114

F==S”=FGNn’JOuFK’J’.RDG GCCGGGMGTACTTCCTCTGGGTGGTGAAGTTCAATTCTTTGAATGAGCTGGTGGATTAT

480 134

bGKYFLw”“KPNS=NEL”n”

540 154 CCACAGCAGCCGACATACGTCCAGGCCCTCTTTGACTTTGATCCCCAGGAGGATGGAGAG P-TYVQQLFDFDPDEDDE

600 174

~~GCTTGCCACGGGCAGC~CATGTTTCCCCGCMTTATGTCACCCCCGTGMC 8GAC”G”TGMFPRNY”TP”N

720 214

CGGMCGTCTMGAGTCAAGMGCMTTATTTAAAGAAAGTGAAMATGTAAAACACATA R N ”

780 217

Figure 1. Nucleotide of GRB2

and Peptide

Sequence

(A) cDNA and protein sequence of G/W2 clone 10-53, with 5’ and 3’ untranslated flanking sequences; SH2 (thick line) and SH3 (thin line) domains are indicated. (6) A schematic representation of the overall domain structure of GRBP. (C and D) Sequence alignments of GRBZ SH2 and SH3 domains with other proteins. N and C refer to N-terminal and C-terminal domains, respectively. The one-letter code is used to indicate amino acid residues. Bold letters identify those positions where the same or a conservative amino acid substitution is present at that position. Compared are PLC-11, GTPase-activating protein, v-src, v-abl, v-crk, and ~85. The SH2 domain of GRBZ is most similar to the SH2 domain of v-fgr (43% similarity), and the N-terminal SH3 domain is most similar to the SH3 domain of human vav (48% similarity).

840 900 960 1020 1080 1109

B.

GRBZ

SH2

SH3 j--~-t I 0

I 50

I loo

SH3

I 150

1 200

(PDGFRs). The association of GRBP is strictly dependent upon tyrosine autophosphorylation of growth factor receptors and is mediated by the SH2 domain of GRBP. Interestingly, the amino acid sequence of GRBP is 58% identical (63% similar) to the sequence of the Caenorhabditis eleg-

ans protein semd (Clark et al., 1992). sem-5 was identified as a gene that regulates vulva1 development and sex myoblast migration in C. elegans, based on evidence that mutations in semd disrupt this process. Similar defects in vulva1 development are caused by mutations in the

Receptor

Tyrosine

Kinases

and ras Signaling

433

C. elegans proteins let-23 (EGFR like) or let-60 (ras like). This suggests that let-23, let-60, and semd are part of the same signaling pathway. The finding that activated let-601 ras can rescue vulva1 development in these mutants supports this hypothesis (Aroian et al., 1990; Horvitz and Sternberg, 1991). In this report, we demonstrate that while microinjection of GRBP or H-ras protein alone did not have a mitogenic effect, injection of GRB2 together with H-ras protein into rat fibroblasts stimulated DNA synthesis. Hence, GRBSlsem-5 appears to be involved in an evolutionarily conserved pathway that mediates growth factor control of ras signaling. Results 6. Isolation of a cDNA Clone Encoding a Protein with Novel SH2 and SH3 Domains The carboxy-terminal tail of the EGFR was used as a probe to screen a human brain stem Igtl 1 protein expression library as previously described (Skolnik et al., 1991). One of the clones isolated utilizing this technique, clone 2-4, contained an insert of 1100 bp found to contain a reading frame encoding novel SH2 and SH3 domains. The insert from clone 24 contained a 3’ stop codon followed by a polyadenylation signal but did not contain the 5’ start site. To isolate the 5’ end of the gene, the library was rescreened using DNA probes generated by amplifying DNA from clone 2-4. This approach enabled identification of clone 10-53, which was found to encode the full-length protein. Clone 10-53, while overlapping with clone 2-4 at the 3’end, contained a S’ATG codon meeting Kozak translation initiation criteria (Kozak, 1969) giving a 660 bp open readingframefrom theinitiatingmethionine(Fickett, 1982) (Figure 1A). Analysis of the protein sequence of clone lo-53 using GenBank revealed that the full-length protein contained a single SH2 domain flanked by two SH3 domains and that these three domains comprise the bulk of the protein (Figure 1B). The SH2 and SH3 domains of GRBP are compared with those in other proteins in Figures 1 C and 1D. The full-length protein encoded by clone IO-53 was named GRB2 and had a predicted molecular size of about 25 kd. The sequence also contains two potential protein kinase C phosphorylation sites (amino acids 18 and 98) two potential casein kinase 2 phosphorylation consensus sequences (amino acids 12 and 127) (Woodget et al., 1966; Kishimota et al., 1985; Marin et al., 1986; Kuenzel et al., 1987) and two Arg-Gly-Asp motifs (Ruoslahti and Pierschbacher, 1986). Northern Analysis and Protein Expression To determine tissue distribution of GRB2, Northern hybridization analysis of various mouse tissue RNAs was performed, using as a probe the insert from clone 10-53. This analysis demonstrated GRBP expression in every tissue examined, with the highest expression in the brain, spleen, lung, and intestine (Figure 2A). Upon longer exposure, GM32 transcripts were also visible in the thymus. We have thus far been unable to identify a tissue or cell line that does not express GRBS, further demonstrating the ubiquitous nature of GR62 expression. GM32 hybridized to two

kD

k[

160

46.5

48.5

365 36.5

12

Figure 2. Analysis and Cell Lines

of Expression

of GRBP in Various

Murine

Tissues

(A) Northern analysis in murine tissues, with tissue of origin as indicated, with 20 ug total RNA loaded per lane. Hybridization with actinspecific probes indicated that the thymus sample was underloaded (data not shown). The sizes of the GM2 transcripts (relative to Bethesda Research Laboratories size markers indicated) are 3.6 kb and 1.5 kb. (B) lmmunoprecipitation of GRBP from [%]methionine HER14 lysates with preimmune (lane 1) and immune GRB2 antiserum (Ab50) (lane 2). lmmunoblot analysis of GRB2 from lysates of HER14 cells with Ab86 (lane 3). Molecular size markers (in kilodaltons) are indicated. Arrow indicates band corresponding to GRBP protein. Exposure times are 24 hr.

transcripts of 1.5 and 3.8 kb. The 1.5 kb transcript corresponds to the expected size of clone 10-53. At present, the nature of the 3.8 kb transcript is not known. Several polyclonal rabbit antisera against GRB2 were generated (see Experimental Procedures) and used to analyze the GRB2 protein by immunoblotting or immunoprecipitation experiments. Figure 28 shows that a protein of 25 kd is recognized by the immune, but not by the preimmune, antiserum utilizing either immunoprecipitation analysis of [%]methionine cells or an immunoblotting approach. The various antisera recognized a 25 kd protein in every cell line and tissue examined, consistent with the distribution of the GRBP transcript found in Northern analysis.

Cell 434

Elt

Ab

A-EGFR

dP-Tyr

EGFI-

N-SH3

d-GR02

T

SH2

C-SH3

GRB2

I--+I

kD

kD

I

SH2

180

N-SH3 116

84

84

C-SH3 N-SH3SH2

48.5

SH2 C-SH3 40.5

36 5 Figure

29

36 5 1 IP

Figure Cells

Ab

3. Association

2

3

5

4

6

4-EGFR

of Endogenous

GRBP

with

EGFR

in HER14

HER14 cells mock treated (lanes 1,3, and 5) or EGF treated (lanes 2, 4, and 6) were lysed, immunoprecitated with anti-EGFR antibodies (MAb 106) subjected to SDS-PAGE, and (after transfer to nitrocellulose) blotted with polyclonal anti-EGFR antibodies (anti-C) (lanes 1 and 2) anti-PY antibodies (lanes 3 and 4) or anti-GRBP antibodies (Ab66) (lanes 5 and 6). The immunoblots were labeled with [‘251]protein A followed by autoradiography at -70%. Anti-GRBP blots were exposed for 24 hr. Anti-EGFR and anti-PY blots were exposed for 16 hr. The positions of molecular size markers (in kilodaltons) are indicated.

GRB2 Associates with Growth Factor Receptors in Living Cells Receptor substrates that contain SH2 domains are endowed with the ability to associate physically with certain activated growth factor receptors. Since the goal of the CORT technique is to identify target proteins for particular growth factor receptors, we assessed whether GRB2 associates with the EGFR. HER14 cells, which are NIH 3T3 cells (clone 2.2) that express approximately 400,000 wildtype human EGFRs per cell (Honegger et al., 1987) were treated with or without EGF, lysed, and subjected to immunoprecipitation analysis according to published procedures (Margolis et al., 1990a, 1990b). lmmunoblotting of anti-EGFR immunoprecipitated with antibodies to GRB2 demonstrated association of the 25 kd GRBP protein with activated EGFR (Figure 3, lane 8). As shown for PLCr and ~85, the association between EGFR and GRBP was strictly dependent upon ligand activation and tyrosine autophosphorylation (Figure 3, lanes 5 and 8) (Margolis et al., 1989, 1990a, 1990b; Wahl et al., 1989; Meisenhelder et al., 1989; McGlade et al., 1992; Hu et al., 1992). Thus, GRB2 associates only with the activated tyrosine-phosphorylated EGFR. We have also demonstrated GRB2 association with EGFR by immunoprecipitation of GRBP followed by immunoblotting with anti-EGFR antibodies (data not shown). Similar results were obtained with PDGFR; activated PDGFR associated with GRB2 in HER14 cells in a growth factor-dependent manner. However, no association between GRB2 and the FGF receptor

4. Schematic

Representation

of GRBP-GST

Fusion

Proteins

GST fusion proteins of full-size GRBP and various regions of GRB2 were generated and purified by affinity chromatography utilizing glutathione-agarose beads, as described in Experimental Procedures. Shown are the SH2 domain of GRBP (SH2). the amino-terminal SH3 (N-SH3), carboxy-terminal SH3 (C-SHS), the amino-terminal SH3 and SH2 domains (N-SH3SH2), and the SH2 domain with the carboxy terminal SH3 domain (SH2 C-SHI). GST region of fusion proteins is not shown.

(FGFR) was detected when similar experiments, using anti-GRB2 for immunoprecipitation and anti-FGFR antibodies for immunoblotting, were performed with cell lines expressing FGFR (data not shown). Interaction of GRB2 with Growth Factor Receptors Is Mediated via the SH2 Domain It has been shown that SH2 domains mediate the interaction of signaling molecules, such as PLC-y or GTPaseactivating protein, with tyrosine-phosphorylated growth factor receptors(Koch et al., 1991; Heldin, 1991; Margolis, 1992). To determine whether the interaction between GRBP and growth factor receptors is mediated via the SH2 domain of GRBP, we constructed bacterial expression vectors that were designed to express GRB2 as well as the various domains of GRBP as glutathione S-transferase (GST) fusion proteins (Figure 4). These fusion proteins were purified by affinity chromatography on glutathioneagarose beads (Smith and Johnson, 1988) and subsequently incubated with lysates from EGF- or PDGF-treated HER14 cells. The ability of the fusion proteins to bind the activated EGFR or PDGFR was assessed by immunoblotting the washed complexes with either anti-phosphotyrosine (anti-PY) or anti-receptor antibodies. Both the full-length GRB2 fusion protein and a fusion protein containing only the SH2 domain of GRBP were capable of binding tyrosine-phosphorylated proteins, which comigrated with the activated EGFR or PDGFR (Figure 5, lanes 4,8,12, and 14). In contrast, neither receptorbound GST alone (Figure 5, lane 2) nor GST fusion proteins containing the amino- or carboxy-terminal SH3 domains could bind to activated receptors (data not shown). Binding was ligand dependent, since immunoblotting with anti-EGFR antibodies detected association of the EGFR with the fusion proteins only when incubated with lysates from growth factor-stimulated cells (Figure 5, lanes 7-10). Thus, in agreement with data about other SH2 domain-containing proteins, the association be-

Receptor 435

Bit

Tyrosine

Kinases

Ah

and ras Signaling

1P-Tyr

EGF-+-+-+ PDGF _

_

-

4 P-Tyr

d-EGFR

_

-

-

_

+ ---

-

+ +

-

+

180-

I,

116NC) 64-

5 8-

1

2

3

4

5

6

7

--F us I 0 II Protein

8

9

10

--

GST

GUI32

IP Ab

, P

32P -++

SH2

SH2

GRB2

IPI

-S I

+ -PIPI

kD

84-

29

184-

1234

12

13

14

Fusion ProReceptors In

Binding of fusion proteins to the tyrosinephosphorylated proteins (lanes 1-6) and EGFRs (lanes 7-10) in control and EGFstimulated HER14 cell lysates, and tyrosinephosphorylated proteins in control and PDGFstimulated lysates (lanes 11-14). Lysates were incubated with equal amounts of fusion proteins immobilized on glutathione-agarose beads. Bound proteins were washed, subjected to SDS-PAGE, and immunoblotted with anti-PY (lanes l-6 and 11-14) or anti-EGFR (lanes 7-10) antibodies. The immunoblots were labeled with [‘“llprotein A followed by autoradiography at -70%. Exposure time is 16 hr. The positions of the molecular size markers are indicated (in kilodaltons).

I-

tween GRBP and growth factor receptors is mediated by theSH2domain (Kochet al., 1991; Heldin, 1991; Margolis, 1992). It is noteworthy that the full-length GRB2 fusion protein bound several other tyrosine-phosphotylated proteins in EGF- and PDGF-stimulated cell lysates (Figure 5, lanes 3, 4,11, and 12). While these bound proteins failed to interact with the SH2-GST fusion protein (Figure 5, lane 6) or either

EGF

I1

Figure 5. Binding of GST-GRB2 teins to Activated Growth Factor Vitro

56

Figure 6. Lack of Significant Phosphorylationof Following Stimulation with EGF

7

8

GRBP in HER14Cells

[zP]orthophosphate (lanes l-4) or [“Slmethionine (lanes 5-6) metabolically labeled HER14 cells were lysed following mock or EGF treatment. The precleared lysates were immunoprecipitated with either preimmune or antiGRB2 antibodies (Ab50) and subjected to SDSPAGE and autoradiography. Exposure times of 2 hr PP) and 2 days (“S) are shown. The positions of GRB2 and the coimmunoprecipitating 55 kd phosphoprotein are marked with arrows.

GRB2

SH2

SH3 domain of GRB2 expressed independently, they did interact with a fusion protein containing both the N-terminal SH3 and SH2 domains (data not shown). The ability of the N-terminal SH3 domain of GRB2 to enhance the binding activity of the SH2 domain suggests that the N-terminal SH3 domain is important for binding to various cellular proteins and that binding to these proteins may require the concerted action of both SH2 and SH3 domains. GRBP Binds to Activated Growth Factor Receptors without Being Phosphotylated in Living Cells After demonstrating that GRBP was able to bind to the activated EGFRs and PDGFRs, we were next interested in determining if GRB2 was a substrate for receptor tyrosine kinases. We examined the capacity of EGF to stimulate phosphorylation of GRBP in HER1 4 cells metabolically labeled with [32P]orthophosphate. These cells were treated with EGF, lysed, and immunoprecipitated with antibodies toGRB2. Whileanti-GRBPantibodies immunoprecipitated GRB2 from [%]methionine cell lysates (Figure 6, lanes 6 and 6). phosphorylated GRB2 was not detected in the anti-GRB2 immunoprecipitates from [%P]orthophosphate cells. Despite marked overexposure of this gel, no detectable band corresponding to GRBP was evident in the otthophosphate-labeled immunoprecipitates. In similar experiments, stimulation of HER14 cells with PDGF also did not result in detectable phosphorylation of GRBP (data not shown). The failure to detect phosphorylated GRB2 was not due to poor stimulation of the cells by EGF, since anti-PY immunoprecipitation of the 32P-labeled lysates demonstrated a marked increase in tyrosine phosphorylation of numerous cellular substrates following EGF stimulation (data not shown). Similarly, anti-PY immunoblotting of GRB2 immunoprecipitated from EGF- or PDGF-stimulated HER1 4 cell lysates did not reveal tyrosine-phosphorylated GRBP (data not shown). To determine if the failure to detect tyrosine-phosphorylated GRBP was due to the rapid dephosphorylation by a protein tyrosine phosphatase, a potent tyrosine phospha-

Cell 436

Figure 7. AlignmentofAmino of GRBP and sem-5

GRBZ SEM-5

GRB2 SEM-5

KVLNeEcDonWYKAELnGkDGFIPk KVLNkDeDphWYKAELdGnEGFIPs

s '" 50

SH3

GRB2 SEM-5

GRBZ SEM-5

Acid Sequences

Alignment of the amino acid sequences of GRBL and sem-5 (single letter code). Boxes surround the SH2 and SH3 domains, as indicated. Bold capital letters indicate identical amino acids; capital letters indicate conservative substitutions.

GRB2 SEM-5

GRB2 SEM-5

FNSLNELVdYHRstSVSRnqqIfLr FNSLNELVaYHRtaSVSRthtIlLs

GRBZ

149 150 174 172

SEM-5

LaFrRGDfIhVmdnsDPNWWkGach LaFkRGDvItLinkdDPNWWeGqln

199 197 2 17

SEM-5

gqtGmFPrNYVtPvNrNv nrrGlFPsNYVcPyNsNkscsnvap

22;’

SEM-5

gfnfgn

226

GRB2 SEM-5

GRB2

tase inhibitor, vanadate, was tested for its effects upon GRBP phosphorylation. [=P]orthophosphate cells were incubated with or without vanadate at 37OC for 20 min prior to the addition of EGF, and GRB2 phosphorylation was assessed as described above. Vanadate pretreatment of EGF-stimulated cells similarly did not result in detectable GRBP phosphorylation. The inability to demonstrate GRBP phosphorylation was further corroborated in a double immunoprecipitation experiment. 32P-labeled HER1 4 lysates were immunoprecipitated with anti-PY antibodies bound to beads and eluted, with the eluates subjected to a second immunoprecipitation with anti-GRB2 antibodies. While clear stimulation of tyrosine phosphorylation was demonstrated in these lysates, no significant phosphorylation of anti-PY-associated GRB2 was detected. Thus, our data demonstrate that while GRB2 associates with the EGFRs and PDGFRs, it is not a good substrate for either receptor. In addition, GRB2 is not phosphorylated by a tyrosine or serinejthreonine kinase that functions distal to the receptor in the signaling pathway. These data suggest that growth factor regulation of GRB2 is not mediated through GRB2 phosphorylation. GRB2 tyrosine phosphorylation was detected in 293 cells transiently overexpressing PDGFR and GRB2, as determined by anti-PY and anti-GRBP blotting (data not shown). A shift in the mobility of GRBP was detected on anti-GRB2 (Ab88) blots in the presence of activated PDGFR, and the lower mobility form was shown to be tyrosine-phosphorylated by anti-PY blotting. Similar experiments have confirmed that the immunoprecipitating antibody (Ab50) will recognize tyrosine-phosphorylated GRBP. These data demonstrate that it is possible to tyrosine phosphorylate GRB2 under conditions of overexpression of both receptor and GRB2 protein. Interestingly, a phosphoprotein of approximately 55 kd was found to coimmunoprecipitate with GRB2 using immune, but not preimmune, sera in lysates from EGF- or

SH3

PDGF-stimulated HER14 cells (Figure 6, lanes 3-4 and 7-8). The 55 kd protein was shown to be phosphorylated on tyrosine, serine, and threonine residues (data not shown). The association of the 55 kd protein with GRB2 immunoprecipitates was dependent upon growth factor stimulation, since this interaction was not observed in GRBP immunoprecipitates from unstimulated cell lysates. The identity of this protein is unknown. GRB2 Represents the Human Homolog of the Product of the C. elegans Gene sem-5 As mentioned earlier, GRB2 is composed of one SH2 domain flanked by two SH3 domains in the order of SH3 SH2 SH3. A C. elegans gene encoding a protein with similar size and domain order has been cloned (Clark et al., 1992). This gene, called sem-5, plays a crucial role in C. elegans development, as mutations in semd impair both vulva1 development and sex myoblast migration. Figure 7 shows a comparison of the amino acid sequences of GRB2 and semd. The N-SH3 domains of GRBP and sem-5 are 65% identical (69% similar), their SH2 domains are 58% identical (63% similar), and the C-terminal SH3 domains are 58% identical (60% similar), respectively. The overall sequence identity and similarity is 58% and 63%, respectively. Considering the evolutionary distance between human and nematode, these two genes are very similar, suggesting that semd represents the C. elegans homolog of GRB2. Effects of Microinjection of GRB2 on Mitogenic Signaling Since GRBP is closely related to semd, a C. elegans gene involved in ras-mediated cell signaling processes, we sought to determine whether GRB2 is a component of the ras signaling pathway in mammalian cells. To address this, the GRBP-GST fusion protein and the normal human H-ras protein were microinjected into quiescent rat embryo fibroblasts (REF-52) and DNA synthesis was determined

Receptor 437

Tyrosine

Kinases

and ras Signaling

Figure 8. Stimulation of DNA Coinjection of H-ras and GRBP

GRB2

Synthesis

by

GAB2 (A), H-ras (B), or GRBP and H-ras (C) were microinjected into quiescent REF-52 cells located to the right of the grid line. Within 1 hr after injection, BrdU was added to the culture medium. At 24 hr after injection, cells were fixed and processed for DNA synthesis as described in Experimental Procedures. Photographs were taken with a 10 x lens.

GRB2 + H-ros

by incorporation of the thymidine analog 5bromo-2’ deoxyuridine (BrdU). In confluent monolayers of FIEF-52 cells, DNA synthesis was detected only in 5% of the cells (Figure 8A), indicating the quiescent state of the cells. Proteins were microinjected into the cytoplasm of quiescent cells within a defined area of the tissue culture dish. Within 1 hr after injection, cultures were treated with BrdU, and DNA synthesis was assayed 24 hr after injection. As shown in Figure 8A, microinjection of the GRBP protein had no effect on DNA synthesis. Similarly, in agreement with previous studies, microinjection of the same concentration of the H-ras protein had no detectable effect on DNA synthesis (Figure 88) (Feramisco et al., 1984; BarSagi and Feramisco, 1988). In contrast, coinjection of the GRB2 and the H-ras proteins resulted in a significant stimulation of DNA synthesis. As shown in Figure 8C, approximately 50% of the cells in the injected area showed stimulated DNA synthesis. These results demonstrate a cooperative interaction between GRB2 and H-ras in the signaling pathway leading to DNA synthesis. To investigate further the functional similarities between GRB2 and semd, GRB2 fusion proteins containing mutations shown to impair sem-5 activity were coinjected with the H-ras protein (Table 1). These point mutations in GRB2, P49L, and G203R correspond to the sem-5 alleles n1619 and n2195 (Clark et al., 1992) respectively. The results of these experiments, summarized in Table 1, show that coinjection of either P49L or G203R GRBP mutants together with H-ras protein did not result in DNA synthesis. Hence, mutations in GRB2 that affect these highly conserved SH3 residues render the protein ineffective in the mitogenic assay. These observations suggest that the SH3 domains of GRBP constitute an essential functional component of the protein, and since these mutants are still capable of binding, the activated EGFRs (data not shown)

are probably involved in coupling to the downstream effector molecule(s). To investigate the role of individual GRB2 domains, we examined the effect on DNA synthesis of coinjecting various domains of GRB2 with H-ras. Table 1 shows that neither the SH2 domain nor the SH3 domains when injected alone could interact synergistically with H-ras to induce stimulation of DNA synthesis. Discussion We have cloned CORT expression

a novel EGFR-binding protein by the method (Skolnik et al., 1991), which we

Table 1. Effect of Microinjection of GRB2 and GRBP Mutants on DNA Synthesis in Quiescent Cells

Protein H-ras GST GRBP H-ras H-ras H-ras H-ras H-ras H-ras

+ + + + + +

Injected

Number Injected

GRB2 P49L G203R SH2 N-SH3 C-SH3

190 148 160 210 172 181 137 140 121

of Cells

Stimulation of DNA Synthesis” w 5 4 7 62 5 5 8 6 5

Quiescent REF-52 cells were maintained in DMEM containing 10% FCS. Purified proteins were microinjected at 2 mg/ml into the cytoplasm of cells. Within 1 hr after injection, BrdU was added to the cultures. At 24 hr after injection, cells were fixed and stained for BrdU incorporation as described in Experimental Procedures. Results represent the average of three experiments. P49L and G203R are GBRP point mutants corresponding to semd alleles n1619 and n2195. respectively. SH2, N-SH3, and C-SH3 are deletion mutants composed of the SH2. amino SH3, and carboxy SHIdomainsof GBRP, respectively. ’ Percentage of injected cells that showed BrdU incorporation.

Cell 438

have called GRB2. This 25 kd protein contains one SH2 domain and two SH3 domains and has wide tissue and cell distribution. GRBP is expressed in ten different murine tissues and in every human, monkey, and murine cell line tested, as demonstrated by Northern blotting, immunoprecipitation, and immunoblotting experiments. Also shown is that GRB2 associates with EGFRs and PDGFRs in a ligand-dependent manner, both in vitro and in living cells. Like other SH2 domain-containing proteins, the association between GRB2 and growth factor receptors is mediated by the SH2 domain, is strictly dependent upon receptor tyrosine autophosphorylation, and involves a direct interaction between GRB2 and the tyrosine-phosphorylated receptors. Despite the fact that GRBP forms stable complexes with tyrosine-phosphorylated EGFRs and PDGFRs, both in vitro and in living cells, GRB2 does not appear to be phosphorylated on tyrosine, serine, or threonine residues at physiologic levels of expression to any significant extent. The fact that pretreatment of cells with vanadate did not increase GRB2 phosphorylation indicates that GRBP is not rapidly dephosphorylated by tyrosine phosphatases. Since GRB2 can be phosphorylated upon transient overexpression with PDGFRs, it is most likely that GRB2 is a poor substrate for EGFRs and PDGFRs in living cells under physiological conditions. In this regard GRBP behaves like the PI-3 kinase-associated p85 protein (Hu et al., 1992). Recent studies suggest that p85 is not tyrosine phosphorylated at physiological levels of expression and that the binding of p85 to tyrosine-phosphorylated PDGFRs via its SH2 domain is essential for activation of PI-3 kinase (McGlade et al., 1992; Hu et al., 1992). Like GRB2, p85 is a poor substrate for the PDGFR at physiological levels of expression; only upon overexpression of PDGFR and p85 could PDGF-induced tyrosine phosphorylation of p85 be detected (Hu et al., 1992). Different results were obtained for PLC-v. This SH2 domain-containing protein binds to EGFRs, PDGFRs, and FGFRs and is tyrosine phosphorylated both in vitro and in living cells (Margolis et al., 1989; Meisenhelder et al., 1989; Wahl et al., 1989; Burgess et al., 1990). Moreover, it has been shown that binding of PLC-7 to receptors via its SH2 domain is critical for tyrosine phosphorylation of PLC-y (Rotin et al., 1992b) and that tyrosine phosphorylation of PLC-y is essential for its activation (Nishibe et al., 1990; Kim et al., 1991). It appears therefore that SH2 domain-containing proteins can be divided into at least two functional classes. The first class includes signal transduction molecules such as PLC-7, whose SH2 domains promote binding to and tyrosine phosphorylation by growth factor receptors(Margolis et al., 1989; Meisenhelder et al., 1989; Wahl et al., 1989; Burgesset al., 1990). Thesecond class includes proteins like p85 and GRBP, which bind to tyrosine-autophosphorylated growth factor receptors but do not become phosphorylated at physiological levels of expression (Hu et al., 1992). These two proteins probably represent regulatory subunits of downstream signaling molecules whose activity is modulated by receptor binding. It is not clear yet whether the catalytic subunit of PI-3 kinase is phosphorylated upon binding of ~85 to activated

growth factor receptors or whether phosphorylation is crucial for activation. It is of note that in preliminary experiments, in which GRBP was immunoprecipitated from [35S]methionine or [32P]orthophosphate metabolically labeled cell lysates, a 55 kd tyrosine-phosphorylated protein was found to associate with GRB2 upon EGF or PDGF stimulation (see Figure 6). The identity of this protein is presently unknown. The extent of sequence homology between GRB2 and sem-5 is striking considering the evolutionary distance between nematodes and humans. The 58% sequence identity (63% similarity) and the conserved overall architecture of these two proteins suggest that semd is the C. elegans homolog of GRBP or a closely related member of the same gene family. The structural similarity between GRB2 and sem5 is reflected in the analogous function of these two proteins. We have recently shown that GRB2 is able to rescue sem-5 mutations in C. elegans (M. Stern et al., unpublished data) and that the sem-5 protein is able to bind specifically the tyrosine-phosphorylated human EGFR (data not shown). By detailed genetic studies, genes crucial for C. elegans vulva1 development and sex myoblast determination have been identified (Horvitz and Sternberg, 1991; Aroian et al., 1990; Clark et al., 1992). It was shown that mutations in lef-23 (EGFR like), semd (GRB2 like), or let-60 (ras like) lead to defects in vulva1 development, while semd also functions in sex myoblast migration. It was therefore proposed that the products of these genes lie along the same signal transduction pathway crucial for normal vulva1 development. We have used a microinjection assay to examine the cooperation between H-ras protein and GRB2 in mammalian cells. It was previously shown that microinjection of oncogenic H-ras protein into cultured fibroblasts stimulated DNA synthesis, while microinjection of normal H-ras protein did not enhance DNA synthesis (Feramisco et al., 1984). Similarly, microinjection of GRB2 protein alone did not cause DNA synthesis (Figure 8). However, microinjection of GRB2 together with H-ras protein was followed by strong stimulation of DNA synthesis in the injected cells (Figure 8). Interestingly, point mutations in GRB2 corresponding to the mutations in sem-5 that cause defects in vulva1 development were unable to stimulate DNA synthesis when coinjected with the H-ras protein into the rat fibroblasts (see Table 1). Similarly, coinjection of H-ras with GRB2 SH2 or SH3 domains alone also did not result in DNA synthesis. Only coinjection of intact GRB2 with H-ras resulted in mitogenic stimulation. Hence, on the basis of genetic studies of C. elegans and the results presented in this report, it is possible to propose a model for the information flow and interaction among these proteins in C. elegans and mammalian cells (Figure 9A). We have shown that GRB2 (see Figure 5) and semd (data not shown) are able to bind to tyrosine-autophosphorylated EGFRs. It is therefore likely that semd will bind tyrosinephosphorylated let-23 via its SH2 domain according to the scheme presented in Figure 9A. Since mutations in let-60 cause a similar phenotype as mutations in either let-23 and semd and since activated ras can rescue let-23 and sem-5 mutations, it is reasonable to assume that let60lras func-

Receptor 439

Tyrosine

Kinases

and ras Signaling

tyrosine residues (Margolis et al., 199Ob). The “P-labeled carboxy terminal tail was then used as a probe to screen a @II human brain stem expression library, as previously described (Skolnik et al., 1991).

EGF

EGFR-

11

C.elegans

W

GRBP -

P%?

-

ras

-

proliferation

11

Northern Analysis Total cellular RNA was prepared with the Stratagene RNA isolation kit. For Northern analysis, RNA was size fractionated on a 1.2% agarose2.2 M formaldehyde gel, transferred by capillary action to a Nytran membrane (Schleicher & Schuell, Incorporated), and prehybridized and hybridized at 65OC in 0.5 M sodium phosphate (pH 7.2), 7% SDS, 1 mM EDTA, and 100 uglml salmon sperm DNA. The GffB.2 cDNA insert was labeled with 3ZP using the random prime method (US Biochemical Company) and used at 2 x 106 cpmlml. The membrane was then washed once at room temperature and then two times at 6W.Z in 40 mM sodium phosphate (pH 7.2), 1% SDS, and 1 mM EDTA.

GDPlGTP Exchange factor

pjyfi@

GTPase

Figure 9. A Model for the Interaction semd and Their Role in ras Signaling

between

Screening of bgtll cDNA Library and DNA Sequence Analysis An oligo (deoxyribosylthymine) lgtll library, constructed from mRNA isolated from human brain stem, was obtained (RhBne Poulenc-Rorer Pharmaceuticals). Screening of the library was performed as previously described (Skolnik et al., 1991). cDNA inserts isolated from positive recombinant phage that bound the EGFRs were subcloned into Ml3 and sequenced by the dideoxy chain termination method, using the Sequenase 2.0 kit (US Biochemical Company). Since the initial clone isolated by expression/cloning did not contain the 5’ ends of the gene, the library was rescreened, using the clone 2-4 insert as a DNA probe.

EGFRs

and GRB2/

Tyrosine-autophosphorylated EGFR (or let-23) binds to the SH2 domain of GRBP (or semd). ras (or let60) acts downstream, leading to either cell proliferation or vulva1 development. In response to growth factor stimulation, GRBP binding to EGFR is followed by tyrosine phosphorylation of P55 protein (A). GRBP may control growth factor-induced ras signaling by stimulating a GDP-GTP exchange factor or by inhibiting a ras GTPase activity or both (B).

tions downstream of EGFR and GRBP and that GRB2 is somehow involved in regulation of ras activity. Moreover, intact GRBP protein and H-ras cooperate to stimulate DNA synthesis when microinjected into rat fibroblasts. In this regard, the 55 kd phosphoprotein that binds to GRBP in response to growth factor stimulation is a candidate for the downstream signaling molecule regulated upon GRB2 binding to activated growth factor receptors. GRBP and its as yet unidentified associated protein(s) may control growth factor-induced ras signaling by stimulating the activity of a GDP-GTP exchange factor and/or by inhibiting a ras GTPase activity or both (Figure 9B). These functions of GRB2/sem-5 are presently speculative and require further investigation. Nevertheless, it is clear that the functional homology between GRBP and semd and their involvement in ras signaling indicate that GRBP plays a crucial role in a highly conserved mitogenic signaling pathway.

Ieolatlon and Labeling of the Carboxy-Terminal Domain of EGFR The intracellular domain of the EGFR, which includes the tyrosine kinaee and carboxy-terminal domain, was purified from a recombinant baculovirus expression system as described (Margolis et al., 1990b; Skolnik et al., 1991). The recombinant protein was phosphorylated with [y-PP]ATP, washed, and digested with cyanogen bromide to yield a 203 residue carboxy-terminal tail contaihing all five phosphorylated

Cell Lines, Immunoblottlng, and Immunopreclpltatlon HER14 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% calf serum (CS). Prior to stimulation, cells were cultured for 18 hr in DMEM, 1% CS. Cells were then stimulated with either EGF (275 @ml) or PDGF-fl(50 @ml) (Intergen, Purchase, New York) for 2 min in DMEM containing 1 mg/ml bovine serium albumin (BSA) and 20 mM HEPES (pH 7.5), following which the cells were immediately washed and lysed. Lysate protein content was normalized as described (Bradford, 1976). Cell lysis, immunoprecipitation, and immunoblotting were performed as previously described (Margolis et al., 1969). Using a modification of the calcium phosphate precipitation method (Chen and Okayama. 1967), 293 cells were transfected. Antlbodies Several polyclonal antibodies were generated against GRB2. A synthetic peptide derived from the N-terminal SH3 domain (residues 3650) and the full-length GRB2-GST fusion protein were used to produce rabbit polyclonal antisera called Ab66 and Ab55, respectively. Both of these antisera are effective at recognizing denatured GRBP in immunoblots. A third polyclonal rabbit antisera, called Ab50, was generated against the GRBP-GST fusion protein containing the C-terminal SH3 domain of GRB2 (residues 167-221) and is capable of immunoprecipitating native GRBP from solubilized cells. Monoclonal anti-PY antibodies (lG2) covalently coupled to agarose were purchased from Oncogene Science (Manhasset, New York). lmmunoblotting with anti-PY antibodies was performed with a rabbit polyclonal antibody. Anti-EGFR immunoprecipitates were performed with monoclonal antibody MAb 106 (Bellot et al., 1969). Anti-EGFR immunoblots were performed with anti-C terminus peptide (residues 1176-I 166) antisera (Margolis et al., 1969). Generation of GST Fuslon Proteins Using the cDNA of GRBP as a template, DNA fragments corresponding to the various GRBP domains were synthesized using the polymerase chain reaction and oligonucleotides that contained appropriate restriction sites and bordered the domains of interest. The amplified DNA was isolated, digested with BamHl and EcoRI. and cloned into pGEX3X (Pharmacia), which was then used to transform Escherichia coli HB 101 to ampicillin resistance. Large-scale cultures were then grown, induced with isopropyl fl-D-thiogalactopyranoside, and the GST fusion proteins purified on glutathione-agarose beads as previously described (Smith and Johnson, 1966). The following fusion proteins were prepared: GST-GRB2 full length (amino acids 2-217); GST-SHP (amino acids 50-161); GST-N-terminal SH3 (amino acids 2-59); GSTC-terminal SH3 (amino acids 156-217); GST-N-terminal SH3-SH2

Cell 440

(amino acids 2-161); GST-SHP-C-terminal SH3 (amino acids 50217). MutationsinGRS2corresponding to thesem-5alleles n1619and n2195 (Clark et al., 1992) were engineered as follows, A DNA fragment encoding the open reading frame of GRBP was synthesized from the GRBP cDNA by the polymerase chain reaction, using flanking oligonucleotides suitable for in-frame insertion into pGEX-PT (Pharmacia). This DNA fragment was cloned into M13mp18 (Amersham) and subjected to mutagenesis using the Amersham oligonucleotide-directed mutagenesis system. The mutagenic oligonucleotides were 5’-GCTTCATTCTCAAGAACTA-3’. which changes the proline residue at position 49 of GRBP to leucine, and 5’GGGCAGACCCGCATGTTTC3’, which converts the glycine residue at position 203 of GRB2 to arginine. The mutations were confirmed by DNA sequencing, and then the DNA fragments were subcloned into pGEX-PT for fusion protein production. The human H-rasoncogene protein was produced in an E. coli expression system and purified by ion-exchange, gel permeation, and hydrophobic column chromatographies as described (Gross et al., 1965). Binding Assays To assay the binding of native growth factor receptors to GST fusion proteins, 500 ul of HER14 cell lysate was incubated for 90 min at 4OC with approximately 5 frg of fusion protein coupled to glutathioneagarose beads. The beads were then washed three times with HNTG (20 mM HEPES [pH 7.51, 10% glycerol, 0.1% Triton X-100, and 150 mM NaCI) and after boiling in sample buffer, the proteins were separated on an 6% SDS polyacrylamide gel. Bound proteins were transferred to nitrocellulose and blotted with antibodies as described (Margolis et al., 1990a, 199Ob, 1992). Cell Metabolic Labeling Labeling cells with [“Plorthophosphate was carried out as previously described (Li et al., 1991). In brief, confluent HER14 cells, starved for 16 hr in 1% fetal calf serum (FCS)-DMEM, were incubated for 2 hr in P,-free media (GIBCO) containing 1% dialyzed fetal bovine serum (FBS), washed twice with P,-free media, and labeled for 2 hr in P,-free medium, 1% dialyzed FBS, 1 m Ci/ml [UP]orthophosphate (carrier free, 314.5-337.5 TBqlmmol, purchased from NEN, Wilmington, Delaware) at 37°C. For some experiments, cells were labeled with [YP]orthophosphate for 15 hr to reach equilibrium. Where appropriate, cells were incubated with vanadate (200 sM) at 37“C for the last 20 min of cell labeling. Cells were then stimulated for 2 min with EGF (250 nglml) or PDGF (50 nglml), rapidly washed two times with ice-cold phosphatebuffered saline (PBS), and solubilized immediately in lysis buffer (10 mM Tris-HCI [pH 7.61, 50 mM NaCI, 30 mM sodium pyrophosphate, 50 mM sodium fluoride, 100 pM sodium orthovanadate. 5 KM ZnCI,, 1 mM phenylmethylsulfonyl fluoride, and 0.5% Triton X-100). After centrifugation, cell extracts were precleared for 1 hr with 50 pl of Sepharose G25 and then incubated overnight with anti-GRBP antiserum (Ab50) at 4OC. The immune complexes were then precipitated with protein A-Sepharose for 45 min at 4OC, washed 8-l 5 times with RIPA buffer (20 mM Tris-HCI ]pH 7.61.300 mM NaCI, 2 mM EDTA, 1% Triton X-100,1 % sodium deoxycholate, and 0.1% SDS), heated in Laemmli’s sample buffer containing 0.1 M f.t-mercaptoethanol and 1% SDS at 95OC for 5 min, resolved by SDS-PAGE (6%-15% gradient), and visualized byautoradiographyof dried gels. To isolatetyrosine-phosphorylated proteins, the cell lysates were incubated with anti-PY antibody (Oncogene Science) beads for 2 hr at 4OC. The anti-PY beads were washed five times with lysis buffer, followed by elution with phenylphosphate (2 mM) in the presence of ovalbumin. Cell Culture and MicroInjection REF-52 cells were maintained in DMEM supplemented with 10% FCS. Cells were plated onto gridded coverslips (Belloc Incorporated) and grown to confluency. Microinjections were performed with glass capillaries drawn to a tip diameter of