FRS2α is a major intracellular substrate of the ligand-activated FGF and NGF receptors and is rapidly and highly tyrosine phosphorylated in cells upon FGF or NGF stimulation. Structurally, FRS2α bears an N-terminal myristylation site, a PTB domain, four potential Grb2(SH2) recognition sites, and two Shp2(SH2) recognition sites (
19). Tyrosine-phosphorylated FRS2α forms a complex with Grb2-Sos and Shp2, which is itself tyrosine phosphorylated and bound to the Grb2-Sos complex (
14). Thus FRS2α mediates the recruitment of the Grb2-Sos complexes directly and indirectly via Shp2 in signaling via the FGF and NGF receptors. The overexpression of wild-type FRS2α enhances FGF-induced MAPK activation and neurite outgrowth in PC12 cells. Conversely, overexpression of FRS2α mutated to abolish the recruitment of Shp2 and Grb2-Sos complexes abrogates FGF-induced MAP kinase activation and neurite outgrowth (
14,
19). In addition, microinjection of anti-FRS2α antibodies inhibits FGF-induced DNA synthesis (
19). These observations establish that FRS2α serves as a physiological link between ligand-activated FGF and NGF receptors and the Ras/MAPK signaling pathway (
14,
19).
To achieve regulation and specificity in signaling, many RTKs utilize a class of proteins called docking proteins. These proteins contain multiple phosphotyrosine residues that serve as recruitment sites for a variety of signaling molecules. Docking proteins commonly possess an N-terminal PTB domain involved in direct interaction with RTKs (
30). It was previously demonstrated that certain docking proteins bind to a consensus NPX
pY sequence motif in several RTKs. For example, the activated insulin receptor possesses an NPX
pY sequence in the juxtamembrane region which is the recognition site for the PTB domains of IRS1 and Shc (
13). A similar NPE
pY motif is also present in the EGF receptor, to which the PTB domain of Shc binds (
3,
20). Upon ligand stimulation the NGF receptor becomes autophosphorylated on an NPQY motif in the juxtamembrane region which serves as a high-affinity binding site for the PTB domain of Shc. This motif is conserved in TrkB, TrkC (
36), and glial cell-derived neurotrophic factor receptor, Ret (
1).
Since the FRS2 proteins are major intracellular substrates and downstream effectors of the FGF and NGF receptors, we set out to investigate the mechanism of interaction between the FRS2 proteins and these two receptors. The interaction between FRS2 and FGF receptor represents a novel paradigm. The association of FRS2 with the FGF receptor is constitutive and does not require tyrosine phosphorylation of the receptor. We have shown in this report that the interaction occurred specifically through the PTB domain of FRS2 binding to the juxtamembrane region of FGF receptor despite the absence of an NPXY motif. We further mapped the binding site and showed that it resides at amino acids 419 to 430 bearing the sequence KSIPLRRQVTVS. This sequence is conserved throughout the mammalian FGF receptor family (Fig. B), and amino acids K419, I421, P422, L423, R425, V427, and V429, when mutated to alanine, strongly diminished the interaction between FGFR1 and the PTB domains of the FRS2 proteins. The inefficient recruitment of FRS2α by FGFR1 bearing combined substitutions of alanine for these amino acids resulted in decreased tyrosine phosphorylation of FRS2α. We have earlier shown that tyrosine-phosphorylated FRS2α bound directly to the SH2 domains of Grb2 and Shp2, leading to potentiation of FGF-induced MAPK activation (
14,
19). Elimination of Grb2 and Shp2 binding by mutation of the tyrosine residues responsible for Grb2 and Shp2 binding to phenylalanine led to reduction in FRS2α-mediated MAPK activation by 30 to 40% (
14,
19). Indeed, the reduced tyrosine phosphorylation of FRS2α by FGFR1 due to inefficient binding to FGFR1 with alanine substitutions for amino acids in the binding domain resulted in reduced MAPK activation to a similar extent. The incomplete inhibition of MAPK activation by the FRS2 and FGFR1 mutants could be attributed to recruitment of Grb2-Sos complexes by other docking proteins, such as Shc and Gab1, that are tyrosine phosphorylated in response to FGF stimulation.
Interestingly, the sequence of the binding domain does not seem to acquire the β-turn conformation which is found in PTB domain-interacting ligands (reviewed in references
7 and
15). The PTB recognition site on FGFR1 contains a putative protein kinase C phosphorylation site bearing the sequence RQVT (amino acids 425 to 428) (
11). The phosphorylation of this threonine is not required for interaction with the PTB of FRS2α since the T428A mutant FGFR1 was still able to interact with FRS2α. In addition, a bacterially produced protein of the juxtamembrane domain of FGFR1 (amino acids 399 to 470) which was not phosphorylated and a synthetic peptide corresponding to the binding site on FGFR1 bind directly to the PTB domains of FRS2 proteins and compete effectively with the binding of the full-length FGFR1.
Unlike what is found for FGFR1, the binding of FRS2 to the NGF receptor (TrkA) is mediated through the classical NPQ
pY motif of TrkA. FRS2 binds in vivo a phosphopeptide corresponding to the binding site on TrkA, resulting in reduced interaction between the activated receptor and the PTB domains of the FRS2 proteins. Indeed, mutation of the tyrosine residue in this motif (Y490) to phenylalanine on the TrkA receptor completely abrogated its signaling capacity. PC12 cells stably transfected with the Y490F mutant TrkA do not exhibit NGF-induced MAPK activation and neurite outgrowth (
28,
41). Recently, it was demonstrated that the PTB domain of FRS2α bound directly to the Shc binding site on TrkA and TrkB (
10,
26).
Our results suggest that the PTB domains of the FRS2 proteins are capable of recognizing diverse sequences specifically. As the synthetic peptides corresponding to the sequences of the binding regions of FGFR1 and TrkA prevent the interaction between these receptors and the PTB domains of the FRS2 proteins, it is likely that the recognition of the diverse ligand sequences occurs through similar or overlapping sites on the PTB domains. Thus, contrary to the previously established paradigms of interaction between RTKs and PTB domains of docking proteins, the phosphorylation of the FGF receptor is not obligatory for recognition by the PTB domains of FRS2 proteins. However, the binding of the PTB domains of the FRS2 proteins to TrkA is dependent upon tyrosine phosphorylation of the NPXY motif.
An emerging notion on the recognition of ligands by PTB domains is that PTB domains exist as a family of structurally conserved protein modules with diverse ligand-binding specificities that is not restricted to recognition of the β-turn-forming sequence NPX
pY. The PTB domains of Shc and IRS1 preferentially bind to targets containing the NPX
pY motif. Those of X11 and FE65 recognize as their target the β-amyloid precursor protein (βAPP) on an NPTY motif, but the phosphorylation of the tyrosine residue is not obligatory (
4,
43). Indeed the replacement of the tyrosine residue with alanine results in no significant loss of binding affinity (
42). The binding of the PTB domain of FE65 to its ligand βAPP requires the presence of an extra 28 residues flanking the NPXY site, but unlike that of X11 this PTB domain does not require the asparagine residue of the NPXY motif (
42). An example of the diverse binding specificity to the same PTB domain is that of the Numb protein and its ligands. Numb is a protein involved in asymmetric cell division in
Drosophila melanogaster. It binds to the adapter protein Lnx through an NPXY sequence, where tyrosine phosphorylation decreases the binding affinity (
8). It also binds to the Numb-associating kinase via the sequence GFSNMSFEDFP, which does not contain a tyrosine residue (
6). In a degenerate phosphopeptide library screen, it was found that the PTB domain of Numb preferentially bound GPY motifs and that the affinity of the nonphosphorylated version of this peptide is significantly lower than that of the phosphorylated peptide (
24). Recently, competitive binding studies showed that the PTB domain of Numb binds to its three ligands with similar affinities (
23). Moreover, since the peptides compete with each other, it is likely that they bind overlapping sites in the PTB domain. The solution structure of the PTB domain of Numb in a complex with the GP
pY peptide shows that rather than the typical type I β-turn conformation observed in other PTB domain-peptide structures, the GP
pY peptide assumes a helical-turn conformation that determines the contact between this peptide and the PTB domain (
23). Based on the above examples, it seems that PTB domains can recognize a wide range of ligands and that their specificities will be defined by their spatial contacts. Our results show that the PTB domains of the FRS2 proteins can be classified in the group of PTB domains which can interact with multiple ligands that have no sequence homology.
FRS2 proteins recognize multiple receptors in phosphorylation-dependent and -independent manners. This property endows FRS2 proteins with a capacity to modulate a signal common to the receptors for FGF, NGF, and probably other neurotropic factors. The experiments presented in this report demonstrate that the relative levels of expression of these receptors and their activation states may influence the degree of recruitment of a limiting element (i.e., FRS2α or -β) that is crucial for activation of a common signal transduction pathway (i.e., Grb2/Sos/Ras). This may provide a plausible mechanism for transmodulation of signals between several different RTKs. The increased expression of FGF receptors in the absence of ligand activation could lead to changes in the distribution of FRS2 binding to ligand-activated TrkA and serve to downregulate the signals originating from TrkA via FRS2α.
Other studies have shown that activation of other members of the neurotrophin receptor family of tyrosine kinases, such as Ret (
32) and TrkA and TrkB (
10,
14) induce tyrosine phosphorylation of FRS2 in PC12 cells and cortical neurons, respectively. These observations are of particular interest since they suggest that transmodulation between RTKs could be a general mechanism that controls the strength of signals initiated by various tyrosine kinases and the fate of the cell in which these tyrosine kinases are expressed. It is therefore important to establish more physiologically relevant systems for exploring the possibility that tyrosine phosphorylation of a limiting docking protein could be an important step that defines specificity and provides a control element crucial for regulation of several families of RTKs that control neuronal cell differentiation and survival.