Visualization of SHP-1-target interaction

Receptor tyrosine kinases (RTK) present an important class of signal transducers playing roles in development, regulation of cell proliferation, differentiation and apoptosis, as well as in diseases such as cancer and diabetes (Hunter, 1998; Schlessinger, 2000). Termination and fine-tuning of their signaling activity occurs through different mechanisms including phosphorylation by other protein kinases, internalization, degradation via the endosome or the ubiquitin-proteasome pathway and dephosphorylation by protein-tyrosine phosphatases (PTPs) (Chiarugi et al., 2002; Haglund et al., 2003; Östman and Böhmer, 2001; Schlessinger, 2000). Some of these processes, such as internalization and dephosphorylation, may also occur in concert (Haj et al., 2002). For many RTKs, these processes are not well understood. While it is clear that PTPs are biologically important negative regulators for at least some RTKs (Östman and Böhmer, 2001), little is known about the mechanisms and the specificity of these interactions. We have recently discovered an efficient interaction of the SH2 domain PTP SHP-1 with the RTK Ros (Keilhack et al., 2001). Ros is an epithelial RTK with a proven role in differentiation and regionalization of the epididymis; in addition, it has a transforming capacity (Riethmacher et al., 1994; Sonnenberg et al., 1991; Sonnenberg-Riethmacher et al., 1996). Hyperphosphorylation of Ros in mice with impaired SHP-1 activity [motheaten viable (mev) mice] suggests that Ros signaling is attenuated by SHP-1 in vivo. SHP-1 binds directly to Ros and inhibits Ros signaling and Ros-dependent transformation in cell lines very efficiently (Keilhack et al., 2001). In order to elucidate the mechanism of Ros-SHP-1 interaction, we have characterized the binding of SHP-1 to Ros and the dephosphorylation of Ros by SHP-1 in more detail. Phosphopeptides representing the potential SHP-1 binding sites in the Ros C-terminal part pY2267 and pY2327 bind the SHP-1 N-terminal SH2 domain with high affinity. Both sites represent SHP-1 binding motifs, which are unique among RTKs. In intact cells, we observed FRET between a functional TrkA-Ros-ECFP fusion protein and EYFP-SHP-1 fusion proteins, provided that the N-terminal SH2 domain of SHP-1 was intact. Nerve growth factor (NGF), a functional ligand for the TrkA-Ros fusion proteins, stimulated the complex formation in intact cells as detectable by fluorescence lifetime imaging of TrkA-Ros-ECFP and membrane translocation of EYFP-SHP-1 fusion proteins. Ligand-dependent stimulation of complex formation required Ros pY2267, whereas Ros pY2327 mediated constitutive SHP-1 association. These results present the first example for visualizing interaction of SHP-1 with a target in intact cells.

Blue fluorescent protein (BFPsg50) was fused to the C-terminus of the chimeric TrkA-Ros by amplifying the TrkA-Ros cDNA using HF-PCR (Roche) with suitable primers and inserting the product into pQB150-fN1 (Q-BIOgene) vector. In this construct, sgBFP and the Ros C-terminus are spaced by six glycine residues. To generate TrkA-Ros-ECFP, a DNA fragment coding ECFP with MluI/EcoRI restriction sites in the ends was amplified by HF-PCR with pECFP-ER (Clontech) as template and exchanged for the BFPsg50 coding sequence in the TrkA-Ros-sgBFP construct. EYFP-SHP-1 fusion proteins were generated by cloning SHP-1 variants into pEYFP-C1 (Clontech) using HF-PCR and appropriate restriction to yield the constructs illustrated in Fig. 2A. In EGFR-Ros2267 the wild-type epidermal growth factor receptor (EGFR) sequence encompassing tyrosine 1173 (AEYLRV) was replaced by the Ros sequence encompassing tyrosine 2267 (LNYMVL) by generating a corresponding expression construct using PCR and standard cloning techniques (details available on request). All constructs were verified by DNA sequencing (MWG Biotech).

All peptide syntheses were performed by Biosyntan (Berlin). Designation and sequence of the peptides is as follows: Ros2267 EGLNY₂₂₆₇MVLATKSS-CONH₂, Ros2327 EGLNY₂₃₂₇ACLAHSE-CONH₂, Epo429 DPPHLKY₄₂₉LYLVVSDSK-CONH₂. Cys2274 in the wild-type Ros sequence had to be replaced by a serine in the corresponding peptide Ros2267, because of the instability of the cysteine-containing peptide. In pRos2267, pRos2327 and pEpo429 the indicated tyrosine residues are phosphorylated. KKKK-Ros2267 and KKKK-Ros2327 contain four lysine residues at the N-terminus. To synthesize the biotinylated peptides Bio-pRos2267, Bio-Ros2267, Bio-pRos2327 and Bio-pEpo429 the peptides were coupled to Biotin using Fmoc-8-amino-3,6-dioxaoctanoic acid (Neosystem, France). pppRos (AREIpY₂₁₀₃KNDpY₂₁₀₇pY₂₁₀₈RKRGEG-CONH₂) is a triple phosphorylated peptide corresponding to the Ros activation loop sequence amino acid 2099-2114.

Activation assays for free SHP-1 were performed with pNPP as a substrate as described earlier (Keilhack et al., 1998). Dephosphorylation assays with synthetic phosphopeptides were done with a malachit green assay (Upstate Biotechnology) according to the manufacturer's instructions. Phosphopeptides were incubated at the concentrations indicated with 3 μg/ml recombinant SHP-1 catalytic domain glutathione-S-transferase (GST)-fusion protein at room temperature for 20 minutes.

GST-fusion proteins were purified as described earlier (Keilhack et al., 1998). For GST pulldown assays, HEK293 cells were transiently transfected with expression constructs for wild-type EGFR or EGFR Ros2267, and 200 μl lysate were used to perform GST pulldowns with 5 μg free GST (control) or 7 μg GST-SHP-1 N-SH2, as described earlier (Keilhack et al., 1998).

Surface plasmon resonance measurements were performed by Biaffin (Kassel, Germany) using a Biocore200 SPR-biosensor (BIAcore AB, Uppsala, Sweden). Biotinylated peptides were immobilized on a streptavidin-coated sensor chip (BIAcore AB) with a surface density of 40-170 resonance units (RU). All experiments were conducted at 22°C in running buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 1 mM DTT, 0.005% Tween). GST-SHP-1 N-SH2 at concentrations ranging from 11 nM to 1.4 μM was injected over the sensor chip with a flow rate of 30 μl/minute for 5 minutes. Dissociation was measured during subsequent washing with running buffer for 5 minutes. The peptide surface was regenerated after every experiment by injection of 0.1% SDS in running buffer (20 seconds, 30 μl/minute). The signal obtained with the Ros2267-coated control surface shows no difference compared with an uncoated sensor chip and was subtracted from the signals obtained for the phosphopeptides. The data were analyzed using the BIAevaluation software 3.1 (BIAcore AB).

Human embryonal kidney (HEK) 293 cells were grown in DMEM/F12 (Phenol-Red-free), COS7 cells in DMEM (Gibco), supplemented with 10% FCS (Gibco) in a humidified atmosphere at 37°C and 5% CO2. HEK293 cells were transiently transfected by using the calcium phosphate method as described earlier (Keilhack et al., 2001), COS7 cells with Lipofectamine Plus (Invitrogene), or Metafectene (Biontex) according to the manufacturer's instructions. Stable transfection of NIH3T3 cells with TrkA-Ros-sgBFP or TrkA-Ros-ECFP expression constructs was performed with Superfect or Effectene (Qiagene) according to the manufacturer's instructions and cells were selected in medium supplemented with 1 mg/ml G418 (Invitrogen).

If required, cells were starved in 0.5% FCS overnight, or 0% FCS for 6 hours and subsequently stimulated with 100 ng/ml NGFβ (Biomol) for the time indicated in the figure legends. Cells were extracted with lysis buffer containing 1% Triton X-100 or 1% NP40 plus protease and phosphatase inhibitors as described earlier (Keilhack et al., 2001).

Cell fixation for microscopy was performed with 4% paraformaldehyde in PBS and mounting with IMMUMOUNT solution (Shandon).

Proliferation and focus formation assays were performed in a similar way to previous studies (Keilhack et al., 2001). To measure proliferation, 50,000 stably transfected or parental NIH3T3 cells/well were seeded into 6 well plates, cultivated in DMEM with 0.5% FCS with or without 50 ng/ml NGF. After 5 days, the cells were trypsinized and counted with a Casy 1 cell counter. For focus formation, cells were grown in the presence of 10% FCS for 14 days with a medium change every 2 days and were then stained with 0.5% crystal violet. To measure tyrosine phosphorylation and Erk-1/2 activation, cells were serum-starved for 6 hours and stimulated with 100 ng/ml NGF for 10 minutes. Aliquots of total cell lysates were subjected to immunoblotting with polyclonal anti-phosphotyrosine antibodies (BD Transduction Laboratories), or anti-pErk antibodies (Cell Signaling Technology). Expression levels were verified on stripped blots using anti-Ros (Riethmacher et al., 1994) and anti-panErk (Upstate Biotechnology) antibodies. Autophosphorylation of TrkA-Ros-sgBFP was measured after immunprecipitation as described by Riethmacher et al. (Riethmacher et al., 1994).

For testing association of TrkA-Ros fusion proteins with SHP-1 variants, coexpression was allowed for 24-48 hours and immunoprecipitation of TrkA-Ros was performed as described earlier (Keilhack et al., 2001) with anti-Ros antibodies or with monoclonal anti-GFP antibodies. Fluorescent fusion proteins were detected by immunoblotting using monoclonal anti-GFP antibodies (Santa Cruz Biotechnology) and untagged SHP-1 with polyclonal anti-SHP-1 antibodies (Santa Cruz). Tyrosine phosphorylation was detected with monoclonal anti-phosphotyrosine antibodies (4G10, Upstate Biotechnology).

To obtain active Ros kinase for phosphorylation of synthetic peptides, HEK293 cells, transiently transfected with TrkA-Ros in pcDNA3 were stimulated with 50 ng/ml NGF for 5 minutes at room temperature and lysed. Ros was immunoprecipitated and the immunoprecipitate was washed twice with 1% and once with 0.1% Triton X-100 in kinase buffer (50 mM Hepes-HCl pH 7.4, 6 mM magnesium chloride, 100 μM sodium orthovanadate). For the kinase reaction 5 μl beads were added to 20 μl of reaction mixture containing the peptide KKKK-Ros2267 or KKKK-Ros2327 in concentrations ranging from 0.1-1 mM, 10 μM ATP and 0.25 μM [γ-³²P]ATP in kinase buffer. The reaction mixture was incubated at 30°C under shaking for 30 minutes and the reaction was stopped by addition of 5 μl 100 mM EDTA and 2.5 mg/ml BSA. The peptides were isolated from the reaction mixture and the incorporated radioactivity was measured as described earlier (Waltenberger et al., 1999). K_m was calculated using the software Enzyme kinetics 2.0.

To evaluate cellular distribution of ECFP and EYFP fusion proteins in living cells, confocal laser scanning microscopy (LSM) was performed with an inverted LSM 510 Rel. 1 (Carl Zeiss GmbH, Göttingen, Germany) using a Zeiss C-Apochromat 63× (NA 1.20) water immersion objective. ECFP and EYFP were excited with the 458 nm or 488 nm line of an argon ion laser, respectively. Fluorescence was recorded with a 475 nm long pass or a 505-550 nm band pass filter, respectively. EYFP-SHP1 translocations were recorded in living, transiently transfected COS7 cells with the same equipment. Expression of TrkRos-ECFP and various EYFP-SHP-1 variants was allowed for 24 hours, cells were serum-starved for 6 hours and then mock-treated or stimulated with 100 ng/ml NGF.

Fluorescence spectroscopy was carried out with transiently transfected, living COS7 cells as described previously (Majoul et al., 2001). About 1-3 ×10⁶ cells were washed with PBS, gently removed from the surface into 1 ml of PBS and immediately used for recording spectra at 37°C using a Fluoromax-2 spectrofluorimeter. The donor (TrkA-Ros-ECFP) was excited at λ_ex=424 nm, and the acceptor (EYFP-SHPs) at λ_ex=494nm. The emission spectra of the donor (λ_em=450 nm-600 nm) and the acceptor (λ_em=510 nm-600 nm) were plotted and analysed using the Origin and Sigma Plot software.

Fluorescence lifetime measurements were performed as described previously (Biskup et al., 2004a; Biskup et al., 2004b). In brief, appropriate cells were selected with a confocal laser scanning microscope (Zeiss LSM 510 Rel. 2, Carl Zeiss GmbH, Göttingen, Germany) using a Zeiss C-Apochromat 63× (NA 1.20) water immersion objective. A mode-locked Titanium:Sapphire laser system (Mira 900, Coherent GmbH, Dieburg, Germany), which was pumped by a 5 W frequency doubled Nd:YVO₄ laser (5W Verdi, Coherent) and tuned to an emission wavelength of 860 nm, was used as excitation source. For streak-camera measurements, the pulse repetition rate was decreased to 2 MHz by a pulse picker (Model 9200, Coherent GmbH, Dieburg, Germany), whereas lifetime imaging was performed with the full repetition rate of 78 MHz. For both types of measurement, the second harmonic of the laser beam (430 nm) was generated in a β-barium borate (BBO) crystal and directed to the scan head of the LSM.

For the acquisition of fluorescence lifetime images the specimen was scanned continuously. Fluorescence light was directed to an external photomultiplier (MCP-PMT, 3809U-51, Hamamatsu), which was connected to a time-correlated single photon counting (TCSPC) module (SPC-730, Becker & Hickl, Berlin, Germany), where a 3D histogram of the photon density over spatial (x,y) and temporal (t) coordinates was built up.

The LSM scanning software was used to direct the laser beam to a spot of interest. Fluorescence was guided via an optical fiber to a spectrograph (Model 250is, Chromex, Albuquerque, NM) and a streak camera (Model C5680 with S20 photocathode and M5677 sweep unit, Hamamatsu Photonics Deutschland, Herrsching, Germany).

for a series of measurements (j) are presented as mean±s.e.m.

According to previously obtained biochemical data, the interaction between Ros and SHP-1 in vitro required the SHP-1 N-terminal SH2 (SHP1-N-SH2) domain and Ros pY2267. The features of this interaction suggested that it occurs with high affinity (Keilhack et al., 2001). We therefore determined binding constants for Ros-derived phoshopeptides and the SHP-1 N-terminal SH2 domain using surface plasmon resonance measurements (Fig. 1A). An earlier characterized efficient binding sequence on the human erythropoietin receptor (EpoR), pY429 (Klingmüller et al., 1995), was used for comparison. SHP1-N-SH2 binds to the EpoR-derived phoshopeptide with a K_d of 1960 nM (Table 1), which is close to previously reported values (Beebe et al., 2000; Pei et al., 1996). The phosphopeptide representing the Ros pY2267 site bound SHP1-N-SH2 with a K_d of 217 nM, i.e. with an affinity almost one order of magnitude higher. Thus, Ros pY2267 represents one of the best binding sites for SHP-1 characterized to date. Two further candidate binding sites for SHP-1 exist in the C-terminus of Ros, pY2166 and pY2327. Previous mutational analysis showed that both are apparently not significantly contributing to the SHP-1 Ros association in vitro (Keilhack et al., 2001). Peptides representing these sites were also subjected to binding assays. While Ros pY2166 exhibited little binding (not shown), Ros pY2327 bound with a K_d of 171 nM, i.e. with an affinity similar to Ros pY2267. The efficient interaction of Ros pY2267 and 2327 with SHP1-N-SH2 is further supported by SHP-1 activation assays (Fig. 1B). Phosphopeptide ligands of SHP1-N-SH2 are known to activate SHP-1 since the binding releases SHP-1 from an inhibited conformation (Pei et al., 1994; Pei et al., 1996; Yang et al., 2002). Again, peptides representing Ros pY2267 and pY2327 activated SHP-1 much more effectively than the EpoR-derived phosphopeptide (Fig. 1B). We finally proved high affinity-binding of SHP-1 to Ros pY2267 by transferring the site to another RTK. We have previously shown that SHP-1 can bind to the epidermal growth factor receptor (EGFR) and that both SH2 domains are required for binding (Keilhack et al., 1998). SHP1-N-SH2 alone cannot bind to EGFR (Keilhack et al., 1998). When the sequence around pY1173 in the EGFR C-terminus was engineered to the sequence around Ros pY2267, the modified EGFR now acquired the ability to bind SHP1-N-SH2 (Fig. 1C). Thus, Ros pY2267 is sufficient to confer high affinity interaction with SHP1-N-SH2 also in the context of another RTK.

The question arose why mutation of Ros Y2327 has little effect on SHP-1 binding in vitro, despite the high affinity of the corresponding phosphopeptide to the SHP-1 SH2 domain. One obvious possibility would be the poor phosphorylation of this site. We, therefore, tested the capacity of Ros kinase to phosphorylate peptides corresponding to Ros Y2267 and 2327. Interestingly, the Ros Y2267 peptide was efficiently phosphorylated (K_m 0.14 mM) while the Ros Y2327 peptide was not (K_m ≫ 1 mM, not shown). Thus, in vitro Ros Y2327 may not function as a high-affinity SHP-1 binding site because of an inefficient phosphorylation by Ros kinase.

It has previously been proposed that SHP-1 is highly active against its own SH2 domain binding site(s) (Bone et al., 1997) leading to decomposition of SHP-1-substrate complexes. We tested this hypothesis for Ros and SHP-1 by measuring activity of the isolated catalytic domain of SHP-1 against phosphopeptides corresponding to different Ros phosphorylation sites. These assays revealed that SHP-1 can dephosphorylate the binding site pY2267 phosphopeptide but with only low efficiency (Fig. 1D). By contrast, efficient dephosphorylation was detectable for the activation loop phosphopeptide containing pY2103, 2107 and 2108 (Fig. 1D). These data suggest that SHP-1 may destabilize the complex with Ros not primarily by dephosphorylation of binding site(s), but rather by inactivation of the Ros kinase.

The high affinity of interaction between Ros and SHP-1 encouraged us to attempt to visualize this interaction in intact cells. To this end, fusion proteins of different SHP-1 variants with enhanced yellow fluorescent protein (EYFP) were constructed (Fig. 2A). Since the ligand of Ros is still unknown, a chimeric protein was employed, consisting of the TrkA NGF-receptor extracellular domain and Ros transmembrane and intracellular domain (Riethmacher et al., 1994). This protein was fused at the C-terminus either to a modified blue fluorescent protein (sgBFP) or to an enhanced cyan fluorescent protein (ECFP) (Fig. 2A). Expression of these fusion proteins in HEK293 cells showed localization in both intracellular membranes and in the plasma membrane (Fig. 2B). Localization of a fraction of TrkA-Ros fusion protein in the plasma membrane was further supported by immuno-electron microscopy (Fig. 2C). As described earlier (Tenev et al., 2000), fluorescent SHP-1 fusion proteins localize largely in the cytoplasm (Fig. 2B). Functionality of TrkA-Ros and SHP-1 fusion proteins was first tested in transient expression experiments. All fusion proteins were expressed with the expected size. Tagged TrkA-Ros expression in HEK293 cells led to abundant tyrosine phosphorylation of multiple proteins, comparable with expression of wild-type TrkA-Ros. Expression of full-length intact EGFP-SHP1 together with TrkA-Ros suppressed tyrosine phosphorylation, demonstrating that the PTP is active (data not shown). To further prove the functionality of the tagged TrkA-Ros, we established stable NIH3T3 cell lines expressing TrkA-Ros-sgBFP or TrkA-Ros-ECFP. As shown in Fig. 3, ligand-dependent activation of (1) TrkA-Ros autophosphorylation, (2) Erk1/2, (3) cell proliferation, and (4) focus formation could be demonstrated in cells stably expressing tagged TrkA-Ros, similar to that described earlier for NIH3T3 cells expressing TrkA-Ros (Riethmacher et al., 1994). In several assays, some constitutive TrkA-Ros activity was detectable in the absence of NGF, which may be related to overexpression of TrkA-Ros.

To assess complex formation between Ros and SHP-1 in intact cells, we checked whether FRET could be observed between TrkA-Ros-ECFP and different coexpressed EYFP-SHP-1 variants.

FRET occurs when a donor molecule is brought into close vicinity (<10 nm) of a suitable acceptor molecule. In this case, the excited donor molecule can transfer its excitation energy to the acceptor molecule. In turn, the intensity and the lifetime of the donor fluorescence decreases, and the acceptor fluorescence increases. We measured changes in the fluorescence lifetime of the donor molecule because this approach has, in contrast to fluorescence intensity measurements, the advantage of being independent of donor and acceptor concentrations.

One of the techniques used in this study to measure fluorescence lifetimes consists of a combination of a spectrograph and a streak camera. This setup allows so-called streak images to be recorded, from which both fluorescence spectra and fluorescence decay curves can be extracted (Biskup et al., 2004b). Fig. 4A shows typical fluorescence emission spectra (upper panel) and fluorescence decay curves (lower panel) obtained from HEK293 cells expressing TrkA-Ros-ECFP only (blue curve) and from HEK293 cells coexpressing TrkA-Ros-ECFP and SHP-1 variants. Compared with the spectrum of TrkA-Ros-ECFP alone, the spectrum obtained from cells coexpressing TrkA-Ros-ECFP and EYFP-SHP1-SH2 (green curve) shows that the donor (CFP) fluorescence in the wavelength range between 465 and 495 nm was decreased, whereas the acceptor (YFP) fluorescence between 520 and 540 nm was increased. Furthermore, the fluorescence decay of the CFP moiety was accelerated upon coexpression with EYFP-SHP1-SH2. Whereas the fluorescence decay of ECFP or TrkA-Ros-ECFP could be fitted by a monoexponential function yielding a lifetime of 2.03±0.11 ns (nanoseconds), the ECFP fluorescence decay recorded from cells coexpressing TrkA-Ros-ECFP and EYFP-SHP1-SH2 had to be approximated by a biexponential function, yielding a mean fluorescence lifetime of 1.15±0.06 ns. A similar lifetime was observed for the EYFP-SHP1-C455S mutant (Table 2). Importantly, ECFP lifetimes for TrkA-Ros-ECFP were significantly longer when it was coexpressed with EYFP-SHP-1 variants with a defective N-terminal SH2 domain (EYFP-SHP1-R32K) (Fig. 4A, lower panel, Table 2). These results indicate a pronounced constitutive association of TrkA-Ros-ECFP with SHP-1 fusion proteins in overexpressing HEK293 cells which requires an intact N-terminal SH2 domain.

Consistent with these results obtained in individual HEK293 cells, a constitutive association was also indicated by spectra obtained from transfected living COS7 cells. For these experiments, cells were removed from the cell culture dish, suspended in PBS and subjected to fluorescence spectroscopy (Majoul et al., 2001) (Fig. 4B). Again, cells coexpressing TrkA-Ros-ECFP and EYFP-SHP1-SH2 showed a decrease of donor (ECFP) fluorescence and a corresponding increase in acceptor fluorescence (EYFP) compared with cells expressing identical amounts of EYFP-SHP1-R32K (estimated by excitation of EYFP).

To visualize ligand-stimulated complex formation, a different method based on time-correlated single photon counting (TCSPC) was used to measure fluorescence lifetimes in COS7 cells transfected with different variants of TrkA-Ros-ECFP and EYFP-SHP1-SH2 (Biskup et al., 2004a). The technique is based on a 3D histogramming process that records photon density over both time and spatial coordinates of the scanning area. Fluorescence decays recorded in each pixel were approximated by a mono- or a biexponential function and mean lifetimes (τ_m) were encoded by color as indicated in the figure legends.

The fluorescence decays recorded from COS7 cells expressing ECFP or TrkA-Ros-ECFP alone could be approximated in all pixels by a monoexponential decay function, yielding a mean of fluorescence lifetimes for all cells (⁠

⁠) of 2.02±0.03 ns and 1.91±0.02 ns, respectively. Moreover, fluorescence lifetimes do not differ throughout the cell as shown by the fluorescence lifetime image of a cell expressing TrkA-Ros-ECFP alone (Fig. 5A). Thus, the lifetime of ECFP does not change when it is fused to proteins, nor does it depend on the intracellular localization.

In COS7 cell coexpressing TrkA-Ros-ECFP and EYFP-SHP1-SH2, the fluorescence decay of ECFP could not be approximated in all regions of the cell with a monoexponential decay function and only a biexponential fit (with τ_s fixed to 2.0 ns) yielded satisfying χ² values (Fig. 5B). As discussed in more detail below, this finding indicates the existence of two populations of TrkA-Ros-ECFP. The mean lifetime in these cells was significantly (P<0.05) reduced compared with cells expressing TrkA-Ros-ECFP only (⁠

ns, Fig. 5A,E) suggesting a certain level of constitutive association of both molecules, in agreement with the streak camera and spectroscopy measurements described above.

When cells coexpressing TrkA-Ros-ECFP and EYFP-SHP1-SH2 were subjected to stimulation with NGF, zones and clusters of short lifetimes could be detected in the cell periphery, indicating complex formation stimulated by NGF (Fig. 5B). Mean lifetimes of cells coexpressing TrkA-Ros-ECFP and EYFP-SHP1-SH2 were significantly (P<0.05) but only moderately decreased after NGF stimulation (⁠

ns, Fig. 5E). Thus, owing to the already existing constitutive association, and despite local effects, NGF could not enhance the overall degree of association by much. These data were supported by measurements of living cells, which have the advantage that NGF induced changes are not averaged over the entire population, but can be followed up in individual cells. When cells were selected that exhibited intermediate or long mean lifetimes, which would be indicative of a low degree of constitutive association, a considerable decrease of

\({\tau}_{m}^{c}\)

could be observed (Fig. 6A,B). However, for all measurements made in living cells (n=7) the mean values (⁠

\({\tau}_{m}^{c}=1.71{\pm}0.09\)

ns before and

\({\tau}_{m}^{c}=1.58{\pm}0.09\)

ns after application of NGF) were close to the respective values of fixed cells. ECFP lifetimes recovered to control values when the acceptor (EYFP) was photobleached (Fig. 6C), which indicates that the decrease in fluorescence lifetime observed after NGF stimulation was only due to FRET between TrkA-Ros-ECFP and EYFP-SHP1-SH2 and not to any other quenching process or photobleaching. In accordance with the lifetime imaging data, we also observed membrane accumulation of EYFP-SHP-1 derivatives with intact SH2 domain upon NGF-stimulation of living COS7 cells expressing TrkA-Ros-ECFP by laser scanning microscopy (Fig. 6D,E).

In Fig. 7 a more detailed analysis of the fluorescence decay is depicted, observed in each pixel after NGF-induced complex formation of TrkA-Ros-ECFP and EYFP-SHP1-SH2. As in the examples shown before, the fluorescence decay of some regions had to be approximated by a biexponential function to achieve a satisfying fit. Again, the fluorescence lifetime (τ_m) is predominantly decreased in the cell periphery (Fig. 7B, left) and the distribution of τ_m over the entire cell exhibits two distinct peaks at 1.5 and 1.9 ns (Fig. 7C, left). As before, the slow lifetime component (τ_s) was set to the lifetime of unquenched ECFP. It can be attributed to the fraction of noninteracting TrkA-Ros-ECFP molecules, which are not subject to FRET. Associated TrkA-Ros-ECFP molecules, which were efficiently quenched by FRET, gave rise to the fast lifetime component (τ_f), which can be regarded as a characteristic property of associated molecules. Thus, the relative amplitudes of the fast (A_f) and slow (A_s) lifetime component provide a rough estimate for the fraction of interacting versus noninteracting molecules.

This interpretation is confirmed by visualization of the parameters recovered in the fit (Fig. 7B). In the pixels where τ_m is low, the amplitude (A_f) of the fast lifetime component (τ_f) is high, indicating that a high fraction of TrkA-Ros-ECFP molecules is subjected to FRET (and thus associated with EYFP-SHP1-SH2). By contrast, A_f is low in pixels where τ_m is close to the lifetime of unbound TrkA-Ros-ECFP, indicating that only a small fraction of TrkA-Ros-ECFP molecules is associated (Fig. 7B, right). In some pixels the values of τ_f were close to τ_s (Fig. 7B, middle, blue) and the biexponential fit converged to a quasi monoexponential decay. Accordingly, the distribution of τ_f exhibits two well separated peaks (Fig. 7C, middle). The first peak can be attributed to pixels in which complexes of TrkA-Ros-ECFP and EYFP-SHP1-SH2 are present and the second peak can be assigned to regions where the fraction of quenched (and associated) TrkA-Ros-ECFP molecules is minimal. Comparison of the lifetime images (Fig. 7B) with the confocal image and the intensity profile along a representative cross-section (Fig. 7A) shows that τ_m is not always decreased and that A_f is not always increased in pixels with high ECFP and EYFP fluorescence intensities. Thus, FRET between TrkA-Ros-ECFP and EYFP-SHP1-SH2 cannot merely be induced by high concentrations of both proteins. Only specific interaction of EYFP-SHP1-SH2 with TrkA-Ros-ECFP brings both fluorescence tags into close vicinity so that FRET occurs.

In COS7 cells coexpressing the mutant TrkA-Ros-ECFPY2267F and EYFP-SHP1-SH2, the extent of constitutive complex formation is not significantly different from cells containing wild-type TrkA-Ros-ECFP and EYFP-SHP1 SH2 (⁠

ns vs. 1.69±0.03 ns, Fig. 5E). Upon stimulation with NGF no significant change of the mean lifetime could be observed (⁠

\({\tau}_{m}^{c}=1.73{\pm}0.04\)

ns, Fig. 5D,E). Thus, loss of Y2267 abrogated the NGF-stimulated complex formation typically detectable with wild-type TrkA-Ros-ECFP. Similar data were observed in living cells (⁠

\({\tau}_{m}^{c}=1.68{\pm}0.10\)

ns before NGF-stimulation vs. 1.59±0.05 ns after NGF stimulation). By contrast, mutating Y2327 clearly reduced the level of constitutive complex formation (⁠

\({\tau}_{m}^{c}=1.87{\pm}0.07\)

ns, Fig. 5D,E). NGF-stimulation led to a significant (P<0.05) decrease of the mean lifetime to levels similar to wild-type TrkA-Ros-ECFP (⁠

\({\tau}_{m}^{c}=1.53{\pm}0.07\)

ns). Again these data could be confirmed by measurements in living cells (⁠

\({\tau}_{m}^{c}=1.85{\pm}0.05\)

ns before NGF-stimulation vs. 1.55±0.08 ns after NGF stimulation). Biochemical experiments supported a role of both pY2267 and pY2327 for interaction of SHP-1 with TrkA-Ros. Coimmunoprecipitation revealed a considerable degree of constitutive association of SHP-1 with TrkA-Ros in intact cells (Fig. 8A). NGF-stimulation enhanced the association to some extent, an effect which was lost in the Y2267F mutant. Mutation of Y2327 strongly reduced association, but NGF stimulation was still detectable. We have previously shown that the association of SHP-1 via pY2267 is important for the efficacy of TrkA-Ros dephosphorylation (Keilhack et al., 2001). Interestingly, as shown in Fig. 8B, Y2327F mutation also reduced the efficacy of TrkA-Ros dephosphorylation significantly. In summary, in intact cells both pY2267 and pY2327 contribute to TrkA-Ros-ECFP/SHP1-SH2 association and SHP-1 mediated dephosphorylation. pY2327 seems to mediate mainly a constitutive association. When pY2327 is lost, pY2267 suffices to mediate ligand-dependent association. Both sites appear to play a complementary or cooperative role for TrkA-Ros interaction.

Characterization of the SHP-1 binding sites pY2267 and pY2327 on the Ros RTK revealed that both are of high affinity. pY2267 was sufficient to confer SHP-1 binding to the heterologous EGFR RTK. The sequences LNpY(2267)MVL and LNpY(2327)ACL correspond to the previously proposed consensus sequence for the SHP-1 N-terminal SH2 domain, hXpYhXh (h, hydrophobic) (Burshtyn et al., 1997; Pei et al., 1996). Using a phosphopeptide library approach, Beebe et al. have recently described two classes of phosphopeptide ligands for this domain: LXpY(M/F)X(F/M/L) (class I) and LXpYAXL (class II) (Beebe et al., 2000). The pY2267 and pY2327 sites match the consensus of class I and II, respectively. Beebe et al. also proposed two classes of binding sites for the SHP-1 C-terminal SH2 domain (Beebe et al., 2000) and pY2327 matches also the corresponding class I consensus (V/I/L)XpYAX(L/V). Therefore, it seems possible that SHP-1 binds to both sites simultaneously, via the tandem SH2 domains. The previous mutational analysis of SHP-1 binding to Ros in vitro revealed only participation of the N-terminal SH2 domain and of Ros pY2267 (Keilhack et al., 2001). By contrast, FRET measurements and biochemical experiments in intact cells described here support the view that also pY2327 is involved in complex formation between TrkA-Ros and SHP-1, mediating a constitutive interaction and contributing to efficient dephosphorylation. Since we found that a peptide containing the Y2327 site is only a poor substrate for Ros kinase it seems likely that Y2327 is phosphorylated by another tyrosine kinase in intact cells. Interestingly, the extent of constitutive association increased with prolonged starvation of the cells (data not shown) suggesting that this kinase may be more active under conditions of serum depletion. Phosphorylation of both Y2267 and Y2327 may potentially allow binding of SHP-1 to Ros via both SH2 domains. Our data are, however, more compatible with an alternative association of SHP-1 with either the pY2327 or the pY2267 site. Mutation of Y2327 abrogated constitutive association. Nevertheless, ligand-dependent association occurred and, based on the FRET measurements, reached levels similar to those of wild-type TrkA-Ros-ECFP. Thus, pY2267 appears sufficient to mediate efficient ligand-dependent association. It is tempting to speculate that, depending on the conditions of cell stimulation, phosphorylation of Y2327 may either enhance the efficiency of SHP-1 interaction, and thus cooperate with Y2267 in mediating downregulation of the signal. In addition, Y2327 may complement the function of Y2267 and play a role in downregulating spontaneous TrkA-Ros activity in the absence of ligand.

Database searches revealed that the pY2267 and pY2327 binding sites are conserved across human, mouse and chicken sequences and are unique for the Ros RTK (not shown). Equally high affinity binding sites for SHP-1 have not yet been described to occur in any other RTK. To our knowledge, a higher affinity binding than the ones described here, has only been observed for the SHP-1 N-terminal SH2 domain to PIRLα, a transmembrane adaptor protein in leukocytes (Mousseau et al., 2000). A direct interaction of SHP-1 has also been shown to occur with other RTKs, including the EGFR (Keilhack et al., 1998), and c-Kit (Kozlowski et al., 1998). However, binding sequences in these receptors are, by comparison, of only low or moderate affinity. Binding of SHP-1 via adaptor molecules, in addition to direct binding, might be important in SHP-1-mediated regulation of these RTKs (Östman and Böhmer, 2001). Thus, interaction of Ros with SHP-1 stands out as an example of direct interaction, concomitant with negative regulation of signaling. Still, regulation of Ros signaling by SHP-1 might additionally occur through indirect mechanisms. Also, other PTPs might participate in Ros dephosphorylation. Interestingly, the Ros catalytic domain has a high homology to the insulin receptor kinase domain (Riethmacher et al., 1994; Sonnenberg-Riethmacher et al., 1996), including a conserved DYY motif in the activation loop. In its phosphorylated form, the latter sequence has been described as an efficient PTP1B substrate and it is possible that Ros is also a target for this ubiquitously expressed PTP (Salmeen et al., 2000). An interaction of one RTK with multiple PTPs has been recurrently discussed (Haj et al., 2003; Östman and Böhmer, 2001) and it may explain why single PTP knockouts have often no obvious RTK signaling phenotypes.

The high affinity of SHP-1 Ros interaction enabled us to visualize the complex formation between both partners in intact cells. Our data, mainly obtained by ECFP lifetime measurements, show a constitutive association of SHP-1 with TrkA-Ros upon coexpression, provided that the SHP-1 variant has an intact N-terminal SH2 domain. In addition, we show that a fraction of SHP-1 translocates to the plasma membrane in response to TrkA-Ros activation by NGF. To our knowledge, these results represent the first demonstration of a SHP-1-target interaction in intact cells. Complex formation between an RTK and a PTP has recently also been demonstrated by FRET for the EGFR and platelet-derived growth factor receptor (PDGFR) and for PTP1B (Haj et al., 2002). In the study of these authors, FRET was measured between GFP-tagged RTKs and a Cy3-labeled anti-PTP1B mAb. Interaction required ligand-triggered internalization and occurred only with the catalytically inactive PTP1B D181A mutant but was undetectable with wild-type PTP1B (Haj et al., 2002) suggesting a dominant role of the PTP1B catalytic domain for complex formation. In the case described here, the SHP-1 SH2 domain interaction with the Ros target is obviously critical for complex formation. From the translocation experiments and lifetime imaging data it can be suggested that SHP-1 binds to activated Ros RTK in the plasma membrane. In contrast to PTP1B, no stringent binding of SHP-1 to any cellular substructure has yet been described. It thus may not be restricted to a certain cellular location for target interaction. Our analysis of constitutive interaction between SHP-1 and Ros shows, however, that intracellular complex formation also occurs. While a part of this interaction may be triggered by inappropriate localization of Ros in the used overexpression setting, it is possible that SHP-1 indeed functions to suppress intracellular signaling of newly synthesized and maturing RTK. SHP-1 is partly localized in a perinuclear compartment (Tenev et al., 2000) and thus in close proximity to newly synthesized RTKs.

Previous biochemical studies revealed that wild-type SHP-1-TrkA-Ros complexes are less stable than such of TrkARos with catalytically inactive SHP-1 versions. The dephosphorylation studies described here suggest that this may not be due to SHP-1 mediated dephosphorylation of the SH2-domain binding site, but may rather be caused by dephosphorylation of activation loop phosphotyrosines and inactivation of the kinase. The pronounced SHP-1 activity against Ros activation loop phosphotyrosine(s) is an interesting observation that requires deeper investigation. For the insulin receptor, efficient dephoshorylation of the activation loop by PTP1B has been shown (Salmeen et al., 2000). Kinase inactivation, in contrast to dephosphorylation of SH2 domain binding sites, appears as an important general mechanism of PTP-mediated RTK regulation.

Taken together, our biochemical and FRET data support a model for SHP-1-Ros interaction as depicted in Fig. 9. (1) Ligand stimulation leads to Ros kinase activation and to phosphorylation of Y2267. (2) SHP-1 is recruited to phosphorylated Ros in the plasma membrane by high-affinity binding to the pY2267 sequence and is in turn activated. (3) SHP-1 dephosphorylates activation loop phosphotyrosines pY2103, pY2107 and pY2108 of Ros, which leads to kinase inactivation, provided the ligand is no longer present. (4) The binding sites are then eventually destroyed by dephosphorylation, mediated either (inefficiently) by SHP-1, or by other PTPs. (5) In turn, the SHP-1-Ros complex is decomposed. (1a) Even in the absence of a ligand, phosphorylation of Ros Y2327 by a heterologous protein tyrosine kinase can occur. (2a) This leads to constitutive complex formation, enabling SHP-1 to suppress spontaneous TrkA-Ros activation. (2b) The formed SHP-1/TrkARos complexes may also undergo ligand activation, and pY2327-associated SHP-1 then contributes to TrkA-Ros dephosphorylation. In this setting, pY2627 and pY2327 may cooperate with respect to TrkA-Ros regulation by SHP-1.

Our data explain why Ros is an efficient target of SHP-1 in vivo. The differential and cooperative role of two high-affinity binding sites is an interesting aspect of this regulation mechanism and may apply also to other cases of SHP-1-target interaction.

We thank Solveig Hehl for help with LSM measurements, Martin Westermann for the immuno-electron microscopy, Carmen Burkhardt and Dirk Schmidt-Arras for help with cloning and various assays. The SPC730 imaging module used for the lifetime imaging was kindly provided by Becker & Hickl GmbH, Berlin. Work was supported by a grant from Deutsche Forschungsgemeinschaft (SFB604, A1) to F.D.B.