Tyrphostin A23 Inhibits Internalization of the Transferrin Receptor by Perturbing the Interaction between Tyrosine Motifs and the Medium Chain Subunit of the AP-2 Adaptor Complex*

Several intracellular membrane trafficking events are mediated by tyrosine-containing motifs within the cytosolic domains of integral membrane proteins. Many such motifs conform to the consensus YXXΦ, where Φ represents a bulky hydrophobic residue. This motif interacts with the medium chain (μ) subunits of adaptor complexes that link the cytosolic domains of integral membrane proteins to the clathrin coat involved in vesicle formation. The YXXΦ motif is similar to motifs in which the tyrosine residue is phosphorylated by tyrosine kinases. Tyrphostins (structural analogs of tyrosine) are inhibitors of tyrosine kinases and function by binding to the active sites of the enzymes. We previously showed that, in vitro and in yeast two-hybrid interaction assays, some tyrphostins can inhibit the interaction between YXXΦ motifs and the μ2 subunit of the AP-2 adaptor complex (Crump, C., Williams, J. L., Stephens, D. J., and Banting, G. (1998) J. Biol. Chem. 273, 28073–28077). A23 is such a tyrphostin. We now show that molecular modeling of tyrphostin A23 into the tyrosine-binding pocket in μ2 provides a structural explanation for A23 being able to inhibit the interaction between YXXΦ motifs and μ2. Furthermore, we show that A23 inhibited the internalization of125I-transferrin in Heb7a cells without having any discernible effect on the morphology of compartments of the endocytic pathway. Control tyrphostins, active as inhibitors of tyrosine kinase activity, but incapable of inhibiting the YXXΦ motif/μ2 interaction, did not inhibit endocytosis. These data are consistent with A23 inhibition of the YXXΦ motif/μ2 interaction in intact cells and with the possibility that different tyrphostins may be used to inhibit specific membrane trafficking events in eukaryotic cells.

Several intracellular membrane trafficking events are mediated by tyrosine-containing motifs within the cytosolic domains of integral membrane proteins. Many such motifs conform to the consensus YXX⌽, where ⌽ represents a bulky hydrophobic residue. This motif interacts with the medium chain () subunits of adaptor complexes that link the cytosolic domains of integral membrane proteins to the clathrin coat involved in vesicle formation. The YXX⌽ motif is similar to motifs in which the tyrosine residue is phosphorylated by tyrosine kinases. Tyrphostins (structural analogs of tyrosine) are inhibitors of tyrosine kinases and function by binding to the active sites of the enzymes. We previously showed that, in vitro and in yeast two-hybrid interaction assays, some tyrphostins can inhibit the interaction between YXX⌽ motifs and the 2 subunit of the AP-2 adaptor complex (Crump, C., Williams, J. L., Stephens, D. J., and Banting, G. (1998) J. Biol. Chem. 273, 28073-28077). A23 is such a tyrphostin. We now show that molecular modeling of tyrphostin A23 into the tyrosinebinding pocket in 2 provides a structural explanation for A23 being able to inhibit the interaction between YXX⌽ motifs and 2. Furthermore, we show that A23 inhibited the internalization of 125 I-transferrin in Heb7a cells without having any discernible effect on the morphology of compartments of the endocytic pathway. Control tyrphostins, active as inhibitors of tyrosine kinase activity, but incapable of inhibiting the YXX⌽ motif/2 interaction, did not inhibit endocytosis. These data are consistent with A23 inhibition of the YXX⌽ motif/2 interaction in intact cells and with the possibility that different tyrphostins may be used to inhibit specific membrane trafficking events in eukaryotic cells.
The trafficking and intracellular targeting of many transmembrane proteins depend on motifs found within their cytosolic domains. One such motif contains a critical tyrosine residue within the sequence YXX⌽, where ⌽ represents a bulky hydrophobic residue (1,2). Tyrosine-based motifs conforming to this consensus sequence can interact directly with the medium chain () subunits of heterotetrameric adaptor complexes involved in several different intracellular trafficking pathways. The AP-2 adaptor complex facilitates incorporation of trans-membrane proteins containing YXX⌽ motifs into clathrincoated vesicles at the cell surface (1,(3)(4)(5)(6)(7)(8)(9), whereas AP-1, AP-3, and AP-4 adaptor complexes are involved in other intracellular vesicular transport steps (reviewed in Refs. 10 -13). The subunits from all four adaptor complexes have been shown to interact with the YXX⌽ motif, with the precise sequence and context of this motif determining the specificity of the interaction (14,15). Mutation of residues within and around the YXX⌽ motif affect interaction with subunits; but, in all cases, mutation of the tyrosine residue in the motif (e.g. to alanine) blocks interaction with the subunit (4,6,8). Structural studies on peptides containing these motifs have generated data that have been interpreted as showing that the YXX⌽ motif exists as part of a tight turn, an ␣-helix, or a ␤-strand (16 -20). However, whatever the predicted overall structure, each of these studies has shown the side chain of the critical tyrosine residue to project away from the peptide backbone.
The YXX⌽ internalization motif is remarkably similar to sequences in which the tyrosine residue can be phosphorylated and, once phosphorylated, bind to Src homology 2 domains (21). It is now clear that although both tyrosine kinases and medium subunits of adaptor complexes recognize essentially the same motif, there is discrimination, such that very few tyrosinebased motifs that have been shown to interact with chains also act as substrates for tyrosine kinases (22)(23)(24). Furthermore, although both chains and tyrosine kinases accommodate the tyrosine side chain as part of their interaction with the YXX⌽ motif, there is no great similarity between the YXX⌽binding sites on chains and those on tyrosine kinases (20).
The similarity between motifs that interact with chains and those that are recognized by some tyrosine kinases suggested that structural analogs of the tyrosine side chain might disrupt the interaction between chains and their target tyrosine motifs. Tyrphostins are structural analogs of tyrosine. They were originally developed as substrate-competitive inhibitors of the epidermal growth factor tyrosine kinase (25)(26)(27). Tyrphostins have subsequently been used to investigate the physiological roles of many different tyrosine kinases. Some tyrphostins have also been reported to inhibit endocytosis and autophagy (28) and vesicle formation from the trans-Golgi network (29), thus implying a possible role for tyrosine kinases in these processes.
Our previous studies using tyrphostins involved the type I integral membrane protein TGN38, which cycles between the trans-Golgi network and the plasma membrane, but has a steady-state localization at the trans-Golgi network (reviewed in Ref. 30). The cytosolic domain of TGN38 contains a consensus tyrosine-based motif, SDYQRL, which is essential for internalization of the protein (31)(32)(33)(34). Both this sequence and the complete cytosolic domain of TGN38 can specifically interact with the 2 subunit of the plasma membrane-associated adap-tor complex AP-2 (4,5,8). Furthermore, this interaction is critically dependent on the tyrosine residue (4,8). We previously used an in vitro assay and yeast two-hybrid growth assays to show that tyrphostins A23 and A46 inhibit the interaction between TGN38 and 2 (35). However, a range of other tyrphostins did not disrupt the interaction, demonstrating that the effect was independent of tyrosine kinase inhibition.
We have now modeled the structure of A23 and other tyrphostins into the YXX⌽-binding cleft in 2. This has provided us with a structural explanation for the inhibition of the YXX⌽ motif/2 interaction by A23.
If tyrphostin A23 were to block the interaction between YXX⌽ motifs and 2 in intact cells, we would predict that clathrin-mediated endocytosis would be inhibited. We have now tested this by following the internalization of transferrin (Tf) 1 in cells incubated in the presence or absence of various tyrphostins. Tf is internalized upon binding to its receptor at the cell surface. This receptor is internalized via a clathrinmediated pathway by virtue of an interaction between a tyrosine-based motif, YTRF, and 2 (16,36). Thus, inhibition of clathrin-mediated endocytosis leads to a reduction in internalized Tf (for example, see Ref. 37). We now show that tyrphostin A23 inhibits the internalization of the Tf receptor (TfR) from the plasma membrane. This effect is specific for A23; other tyrphostins that act as inhibitors of tyrosine kinase activity do not inhibit Tf internalization. Thus, this inhibition of Tf internalization in intact cells correlates with the capacity of A23 to inhibit the interaction between YXX⌽ motifs and 2 in vitro.

EXPERIMENTAL PROCEDURES
Molecular Modeling-The crystal structure (Protein Data Bank Code 1i31) of the 2 adaptin subunit (AP50) of the AP-2 clathrin adaptor complexed with the peptide FYRALM (20) was used as the starting point for modeling. Hydrogen atoms were added to the structure consistent with pH 7. All residues with an atom within 25 Å of lysine 203 were selected from the structure, and the rest were discarded. The remaining volume in this sphere was filled with water, and all residues in the concentric sphere between 20 and 25 Å were fixed in space during subsequent calculations. Those residues in the 20-Å sphere were energyminimized to an average derivative of 0.01 kcal/Å. Partial charges for the tyrphostin atoms were derived from Austin Model 1 semi-empirical molecular orbital calculations performed on the respective anions in vacuum. The tyrphostin anions were docked simply by superimposing their phenoxy groups onto the tyrosine ring of the FYRALM peptide, removing this peptide, and resoaking the binding site in water. Two conformations (P and Q) of tyrphostins A8 and A23 were explored, distinguished by the exocyclic torsion between the aromatic ring and the double bond.
Tissue Culture-Heb7a cells were grown in Dulbecco's modified Eagle's medium (Sigma) plus 10% fetal calf serum (Labtech International) and 1% penicillin/streptomycin (Invitrogen) for at least 2 days prior to use in experiments.
Tyrphostins-Tyrphostins (Sigma and Calbiochem) were stored as 350 mM (A23 and A63) or 8 mM (A51) (1000-fold) stock solutions in Me 2 SO at Ϫ20°C and added to solutions just before use.
Radiolabeled Transferrin Internalization Assays-The kinetics of Tf internalization in Heb7a cells were measured following the method described by McGraw and Subtil (38) using 0.5 g/ml 125 I-labeled human diferric Tf (PerkinElmer Life Sciences) for labeling. Cells exposed to tyrphostins were incubated with the drug for a total of 30 min (i.e. inclusive of the incubation time for Tf uptake). The data were processed using Microsoft Excel and plotted with the linear regression function in SigmaPlot 2001 (Version 7.1).
Internalization of Fluorescent Transferrin-Cells were washed with serum-free medium and labeled for 30 min with 25 g/ml human serum Tf conjugated to Alexa 594 (Molecular Probes, Inc.). The cells were then fixed with methanol at Ϫ20°C for 4 min.
Immunofluorescence-Cells were fixed with methanol (as described above) or 3% paraformaldehyde, 1 M CaCl 2 , and 1 M MgCl 2 in phosphate-buffered saline (pH 7.4) at 20°C for 20 min. Paraformaldehydefixed cells were quenched with 30 mM glycine in phosphate-buffered saline and permeabilized with 0.1% Triton X-100 in phosphate-buffered saline for 4 min at 20°C. All cells were blocked with 3% bovine serum albumin in phosphate-buffered saline for 15 min prior to incubation with primary antibodies (diluted in the blocking buffer) for 1 h. Murine monoclonal antibodies raised against LAMP-1 (CD107a) (clone H4A3, BD Biosciences), LAMP-2 (CD107b) (clone H4B4, BD Biosciences), CD63 (Biogenesis), and EEA1 (clone 14, Transduction Laboratories) were used at 1:200 dilutions. Alexa 488-conjugated goat anti-mouse IgG (H ϩ L) (Molecular Probes, Inc.) diluted 1:800 in the blocking buffer was used as secondary antibody. The coverslips were mounted with Mowiol (as described by Harlow and Lane (39)) and examined on an inverted Leica TCS NT (UV) confocal laser-scanning microscope with a ϫ100 plan apo objective lens. Care was taken to ensure equal laser intensity, photomultiplier tube setting, and pinhole size for Tf internalization assays. Optical slices were taken at 0.5-or 1-m intervals through the cells and displayed as maximum projections in the microscope software. The images were then imported to Adobe Photoshop Version 6.0 to produce the final figures.
Mean pixel intensity measurements were determined over four quadrants of cells at zoom ϫ1 and magnification ϫ63 using the Leica confocal microscope software. The data were processed with Microsoft Excel and plotted with SigmaPlot 2001(Version 7.1) using the linear regression function.
Transferrin Receptor Degradation Assay-Heb7a cells were incubated in the presence of 350 M tyrphostin A23 for 0, 15, or 30 min. Post-nuclear cell lysates were prepared (40), and equal amounts of protein, as determined by the Bradford assay (41), were electrophoresed under reducing conditions on a 10% SDS-polyacrylamide gel. The gel was immunoblotted with a murine monoclonal antibody (H68.4) raised against the cytosolic domain of the human TfR (a gift from C. R. Hopkins) (42).

Molecular Modeling of Tyrphostins into the 2 Tyrosine-
binding Pocket-We have previously shown that, in vitro, both tyrphostins A8 and A23 can inhibit the interaction between YXX⌽ motifs and the 2 subunit of the AP-2 adaptor complex, with tyrphostin A23 doing so more efficiently (35). In contrast, other tyrphostins failed to inhibit this interaction (35). In an initial attempt to provide a molecular explanation for these observations, we chose to model the structures of various tyrphostins into the tyrosine-binding cleft in 2. Tyrphostins A8, A23, and A46 have previously been shown to be efficient inhibitors of the interaction between YXX⌽ motifs and 2 in recombinant protein interaction assays, whereas tyrphostins A47, A25, and A51 are not (35). We observed that the phenyl rings of tyrphostins A8, A23, A46, and A47 were accommodated within the tyrosine-binding cleft in 2, whereas those of tyrphostins A25 and A51 were not. Tyrphostins A8 and A63 are monohydroxylated at position 4 of the phenyl ring in their structure (i.e. the same as a tyrosine side chain); tyrphostins A23, A46, and A47 are 3,4-dihydroxylated on their phenyl ring; and tyrphostins A25 and A51 are 3,4,5-trihydroxyphenyl compounds (Fig. 1). The energy-minimized complexes of A8 and A23 in orientation/conformation P (see "Experimental Procedures") resulted in the tyrphostins binding snugly in the pocket occupied by tyrosine in the FYRALM peptide complex. The O Ϫ atom forms a hydrogen-bonding network with Lys 203 , Asp 176 , and Arg 423 at the base of the binding cleft, whereas the phenyl ring lies in the hydrophobic groove formed by the side chain of Arg 423 , Phe 174 , and Trp 421 , as shown in Fig. 2. In contrast, the complexes in orientation/conformation Q (see "Experimental Procedures") break this hydrogen-bonding network because steric clashes between the dinitrile group and Val 422 force the tyrphostins ϳ2 Å along the groove, away from the O-binding site. In the P complex of tyrphostin A23, the second ring oxygen abuts the guanidinium group of Arg 423 . This probably accounts for the fact that this compound is more efficient than tyrphos-tin A8 as an inhibitor of the interaction between YXX⌽ motifs and 2 (35). The reason for the failure of tyrphostins with three-ring hydroxy groups (e.g. A25 and A51) to inhibit the interaction between YXX⌽ motifs and 2 also becomes evident from inspection of Fig. 2. The third hydroxy group would necessarily be forced into the hydrophobic part of the cleft (lined by Leu 175 ) if the compound were to bind in a similar fashion to that proposed for A8 and A23. Thus, it appears that the tyrosine-binding cleft in 2 can accommodate a 3,4-dihydroxy derivative, but not a 3,4,5-trihydroxy derivative, of a phenyl ring. This is consistent with our previous in vitro observations on the inhibition of YXX⌽ motif/2 interactions by different tyrphostins (35). In fact, addition of an extra hydroxyl group in the 3-position of the phenyl ring is beneficial for the interaction with 2 (see above). Thus, a 3,4-dihydroxyphenyl compound (such as tyrphostin A23 or A46) is predicted, by molecular modeling studies, to both fit well in the tyrosine-binding cleft of 2 and be stabilized in that binding by hydrogen bonding and other interactions (Fig. 2). This provides an explanation for our earlier observation (35) that tyrphostin A23 inhibits the interaction between 2 and a YXX⌽ motif more efficiently than tyrphostin A8 (which is structurally identical to tyrphostin A23 apart from the fact that A8 is a 4-monohydroxyphenyl compound and A23 is a 3,4-dihydroxyphenyl compound) (Table I). However, these data do not explain why tyrphostin A63 should not be equally as efficient as tyrphostin A8 in inhibiting the interaction between 2 and a YXX⌽ motif because both are 4-monohydroxyphenyl compounds (Fig. 1). In fact, the only difference between the structures of tyrphostins A8 and A63 is that the former has a double bond linking the malononitrile group to the aromatic ring, whereas the latter does not (Fig. 1). This led us to suspect that the pK a values for the hydroxyl groups on the phenyl rings of the two molecules might be different. Measurements of the tyrphostin pK a values showed this to be the case, with the pK a for tyrphostin A8 being 6.73 and that for A63 being 9.61 (Table I). This implies that the tyrphostin anion is the species that binds into the 2 active site and complements the overall positive charge at the base of the cleft, which is composed of the side chains of Lys 203 , Asp 176 , and Arg 423 (Fig. 2).
Tyrphostins A46 and A47 are structurally very similar, both being 3,4-dihydroxyphenyl compounds, and both can be modeled into the tyrosine-binding cleft in 2. Furthermore, the pK a values for A46 and A47 (7.13 and 6.87, respectively) ( Table I) are such that these compounds will be mostly ionized at physiological pH. However, our previous in vitro binding studies showed that only tyrphostin A46 efficiently inhibits the interaction between 2 and a YXX⌽ motif (35). Because the structure and acidity of these tyrphostins are so similar, it was unclear why A47 was inactive, whereas A46 showed good activity. A possible explanation lies in the relative stability of these tyrphostins toward hydrolysis. Tyrphostins A46 and A47 were incubated in the dark in pH 7 buffer at 37°C, and the absorbance due to the anion was measured at regular intervals.

TABLE I pK a values for Trphostins
A buffer comprising sodium acetate, BisTris, Tris, Ches, and Chaps (20 mM each) was used for the pK a measurements at 25°C. For each compound, UV spectra were measured between 250 and 500 nm at 0.5-pH unit intervals between pH 5 and 11 immediately after addition of the Me 2 SO stock solution (5 l of 2-10 mg/ml to 1 ml of buffer). An extra point for the titration of tyrphostin A63 at pH 12 was obtained using 10 mM NaOH. Spectra corresponding to the first ionization of each compound showed clear isosbestic points, indicating clean conversion between the neutral and mono-anionic species. Absorbances at the wavelengths shown below, corresponding to the mono-anionic species, were plotted against pH. These data were fitted to the equation A A ϩ (A B Ϫ A A ) ⅐ (alog(pH Ϫ pK a ))/(1 ϩ alog(pH Ϫ pK a )) (where A A is the absorbance at low pH, and A B is that at high pH) in Graft Version 3.01 (Erithacus Software Ltd.) to determine pK a values. Although the signal due to A46 persisted, decaying Ͻ5% over 24 h, that due to the anion of A47 decayed significantly over this period (Fig. 3). The hydrolysis followed first-order kinetics with a half-life of 3.4 h. Thus, the most likely explanation for the lack of activity shown by A47 is that the compound is rapidly hydrolyzed in in vitro interaction assays (performed in the light, a condition that further reduces the stability of the compound).
The preceding data demonstrate that specific tyrphostins inhibit the interaction between tyrosine-based internalization motifs and 2 by binding to the tyrosine-binding cleft of 2. This interaction is stabilized by hydrogen (and other) bonds at the base of the cleft and is optimal when a 3,4-dihydroxyphenyl tyrphostin, in mostly the anionic form at physiological pH, is used. Tyrphostin A23 is such a tyrphostin, previously shown to inhibit the interaction between tyrosine-based internalization motifs and 2 both in vitro and in yeast two-hybrid interaction assays (35). This tyrphostin was selected for use in further studies.
125 I-Transferrin Internalization Assays-Having established the structural basis for the inhibition of the YXX⌽ motif/2 interaction by tyrphostin A23, it was pertinent to investigate whether this inhibition occurs in vivo to inhibit trafficking pathways in a mammalian system. To address this question, we focused on the internalization of Tf from the plasma membrane. Tf is internalized following binding to the TfR at the cell surface, with the internalization of the TfR being dependent upon interactions between 2 and the tyrosine-based internalization motif (YTRF) in the TfR cytosolic domain (16,36,44). A biochemical assay was used to probe the effects of selected tyrphostins on Tf internalization. The assay followed the internalization of 125 I-Tf into Heb7a cells incubated at 37°C in the presence or absence of different tyrphostins (Fig. 4). Incubation of cells in the presence of 350 M tyrphostin A23 (the concentration selected as a result of observed effects in in vitro assays (35)) for 30 min led to a marked decrease in the rate (ϳ4-fold) of 125 I-Tf internalization compared with that observed in control cells (Fig. 4A). Control tyrphostins (A51 and A63, previously shown to have no effect on the interaction between 2 and the TGN38 YXX⌽ internalization motif (35)) had no significant effect on 125 I-Tf internalization (Fig. 4, B and C). Tyrphostins A51 and A23 (IC 50 for inhibition of epidermal growth factor receptor tyrosine kinase activity of 0.8 and 35 M, respectively) were both efficient inhibitors of tyrosine kinase activity, but only A23 inhibited Tf internalization (Fig. 4, compare A and B). Tyrphostin A63 (IC 50 for inhibition of epidermal growth factor receptor tyrosine kinase activity of 6500 M) has been used as a negative control in assays of inhibition of tyrosine kinase activity by tyrphostins (27); it was also ineffective as an inhibitor of endocytosis (Fig. 4C).
Tyrphostins A23 and A51 were both used at 10 times their IC 50 values for inhibition of epidermal growth factor receptor tyrosine kinase activity in these assays. The fact that only A23 inhibited Tf internalization indicates that it exerts its effect by a mechanism other than inhibition of tyrosine kinase activity. Given our previously published in vitro data (35), this mechanism is most FIG. 3. Stability of tyrphostins A46 and A47. Shown is the evolution of the UV absorption signal of tyrphostin anions at pH 7.2 and 37°C with time. The buffer used is described in the legend to Table I. The data for A46 (E) were fitted to a straight line, and there was Ͻ5% loss in the signal over 24 h. The data for A47 (q) were fitted to a single exponential decay with a half-life of 230 min, resulting in Ͼ95% signal loss over 24 h. Absorbance is given in arbitrary units. likely the inhibition of the interaction between 2 and the YTRF internalization motif in the cytosolic domain of the TfR.
Internalization of Fluorescently Labeled Transferrin-Human serum Tf conjugated to Alexa 594 (Alexa 594-Tf) was also used as a tool to investigate any possible effects of tyrphostin A23 on trafficking pathways (Fig. 5). Heb7a cells were treated with 350 M tyrphostin A23 for varying times and allowed to internalize Alexa 594-Tf. Cells exposed to tyrphostin A23 for 30 min in total (Fig. 5B) showed reduced internalization of the fluorescent probe compared with untreated cells (Fig. 5A), a result consistent with the 125 I-Tf uptake assays. Heb7a cells exposed to tyrphostin A23 for longer periods (Fig. 5, C-F) showed no significant morphological changes in the distribution of the fluorescent probe compared with control cells (Fig.   5A). This demonstrates that, although the rate of Tf internalization was inhibited by tyrphostin A23, this tyrphostin had no significant effect on the distribution of Alexa 594-Tf throughout the recycling pathway over the time course of the experiment. Alexa 594-Tf uptake experiments were also performed on Heb7A cells incubated in the presence of different concentrations of tyrphostin A23 for 30 min. The inhibitory effect of tyrphostin A23 on Alexa 594-Tf uptake was found to be dosedependent (Fig. 6).
Internalization of TGN38 -Because our original in vitro studies focused on the effects of tyrphostins on the interaction between 2 and the tyrosine-based internalization motif in the cytosolic domain of TGN38 (35), the effect of A23 on the internalization of TGN38 was also assayed. This was done by an The cells were fixed with methanol, mounted, and examined using an inverted Leica TCS NT (UV) confocal laser-scanning microscope with a ϫ100 plan apo objective lens. Heb7a cells exposed to tyrphostin A23 for 30 min in total (B) showed reduced internalization of the fluorescent probe compared with untreated cells (A). Heb7a cells exposed to tyrphostin A23 for longer periods (C-F) showed no significant morphological changes in the distribution of the fluorescent probe compared with control cells (A). Scale bar ϭ 10 m.
antibody uptake assay using a monoclonal antibody that recognizes the extreme N terminus of TGN38 (2F7.1), followed by detection of the internalized antibody using a fluorescently labeled secondary antibody as described previously (34). Normal rat kidney cells were incubated in the presence or absence of 350 M A23 for 5 min at 37°C and then with antibody 2F7.1 in the continued presence or absence of 350 M A23 for 30 min at 37°C prior to methanol fixation, incubation with secondary antibody, and processing for fluorescence microscopy. Mean pixel intensities were recorded from both sets of cells under identical imaging conditions (as described under "Experimental Procedures"). The mean pixel intensity (in arbitrary units) in control cells was 121.25 Ϯ 22.10 (n ϭ 50 cells), and that in A23-treated cells was 38.09 Ϯ 10.05 (n ϭ 50 cells). Thus, in addition to inhibiting Tf uptake, A23 also inhibited the internalization of TGN38. No other tyrphostin tested had any effect on the internalization of TGN38.
Internalization of Fluorescent Dextran-Our earlier in vitro assays (35) and the modeling studies presented here suggest that the inhibition of Tf uptake and of TGN38 internalization by A23 is a result of inhibition of the interaction between 2 and the tyrosine-based internalization motifs in the cytosolic domains of the TfR and TGN38. If this is the case, then internalization of a fluid-phase marker should not be affected by incubation of cells in the presence of A23. To address this, Heb7a cells were incubated in the presence of 1 mg/ml fluorescently labeled 10-kDa dextran (as described under "Experimental Procedures") for 30 min at 37°C in the presence or absence of 350 M A23. Mean pixel intensities were then recorded from both sets of cells under identical imaging conditions. The mean pixel intensity (in arbitrary units) in control cells was 34.0 Ϯ 9.9 (n ϭ 30 cells), and that in A23-treated cells was 38.0 Ϯ 11.0 (n ϭ 30 cells). Thus, incubation of cells in the presence of A23 for 30 min at 37°C had no effect on fluid-phase uptake of dextran. No other tyrphostin tested had any effect on fluidphase uptake of dextran.
Analysis of the Morphology of Endocytic Compartments-The fact that the distribution of internalized Alexa 594-Tf in cells that had been incubated in the presence of tyrphostin A23 was indistinguishable from that in control cells suggested that the compartments of the endocytic recycling pathway were not perturbed by the presence of tyrphostin A23. To confirm this and to address possible effects of tyrphostin A23 on the morphology of other endocytic compartments, we used immunofluorescence microscopy to localize a range of endomembrane markers in both A23-treated (350 M, 30 min) and untreated Heb7a cells. No significant changes in the localization of the lysosomal/late endosomal markers LAMP-1 (Fig. 7, A and B), LAMP-2 (C and D), and CD63 (E and F) and the early endosomal marker EEA1 (G and H) were observed. Similar analysis using markers for the Golgi apparatus and trans-Golgi network also failed to show any difference between control cells and those that had been incubated with tyrphostin A23 (data not shown).
Transferrin Receptor Degradation Assay-We considered it possible that, although there were no gross changes in the morphology of endocytic compartments, the presence of tyrphostin A23 might affect the internalization/recycling of TfRs  Post-nuclear cell lysates were prepared, and equal amounts of protein were electrophoresed under reducing conditions on a 10% SDS-polyacrylamide gel prior to transfer to nitrocellulose and incubation with a murine monoclonal antibody (H68.4) raised against the cytoplasmic domain of the human TfR. Antibody binding was detected by ECL as described under "Experimental Procedures." Antibody H68.4 detected a 95-kDa band of equal intensity in each of the lanes. This shows that the number of TfRs remained constant throughout the experiment and that they were not degraded in response to treatment with tyrphostin A23.
such that an elevated proportion of internalized TfRs might be routed to late endosomes and lysosomes for degradation. This possibility was investigated by immunoblot analysis of postnuclear supernatants from Heb7a cells that had been incubated in the presence of 350 M tyrphostin A23 for different times (Fig. 8). The blot was probed with a murine monoclonal antibody (H68.4) (42) raised against the cytosolic domain of the human TfR. No reduction in the level of TfR expression was detected during a 30-min incubation in the presence of 350 M tyrphostin A23 (Fig. 8). DISCUSSION We had previously shown that, in vitro and in yeast twohybrid interaction assays, certain tyrphostins are capable of inhibiting the interaction between 2 and tyrosine-based internalization motifs (35). Molecular modeling of different tyrphostins into the tyrosine-binding cleft in 2 has now provided a structural explanation for the finding that certain tyrphostins inhibit this interaction, whereas others do not. It is clear that, although mono-and dihydroxylated phenyl rings can be accommodated in the tyrosine-binding cleft in 2, trihydroxylated phenyl rings cannot. Thus, certain tyrphostins are excluded. Furthermore, tyrphostins with dihydroxylated phenyl rings are likely to bind to 2 with higher affinity than those with monohydroxylated phenyl rings because the dihydroxylated derivatives are capable of a more extensive bonding network at the base of the cleft than their monohydroxylated counterparts. We also observed that the pK a values for the hydroxyl group(s) on the phenyl ring of the tyrphostins can influence the interaction with the tyrosine-binding cleft in 2; those with a pK a of ϳ7 effectively inhibit the interaction, whereas those with a higher pK a (ϳ9) do not. This is consistent with the anionic form of the tyrphostin being the active species, interacting with the overall positive charge at the base of the binding cleft in 2.
Having provided a structural explanation for the inhibition of the interaction between 2 and tyrosine-based internalization motifs by specific tyrphostins, we went on to address whether tyrphostin A23, a known inhibitor of the tyrosine motif/2 interaction in vitro, has any effect on endocytosis in mammalian cells. Consistent with both the in vitro data and the modeling studies, tyrphostin A23 significantly reduced (ϳ4-fold) the endocytosis of Tf in Heb7a cells. Control tyrphostins, known to have no effect on the tyrosine motif/2 interaction in vitro, had no effect on Tf uptake. These results are consistent with tyrphostin A23 inhibition of the interaction between 2 and tyrosine-based internalization motifs in intact cells. It is interesting to note that Nesterov et al. (43) made a dominant-negative epitope-tagged mutant of 2 and assessed its effect on AP-2 and endocytosis. Elevated expression of this D176A/W421A 2 mutant resulted in metabolic replacement of endogenous 2 by the mutant in AP-2 complexes. This led to complete abrogation of AP-2 interaction with the tyrosine-based internalization motifs because the D176A/W421A mutations lie within the tyrosine-binding pocket of 2 (20,43). Furthermore, expression of D176A/W421A 2 led to a 4-fold decrease in the internalization of Tf from the cell surface (43). This decrease in Tf internalization matches that which we observed following incubation of cells in the presence of tyrphostin A23.
The data presented here show that tyrphostin A23 reduced the rate of Tf internalization by ϳ4-fold. However, there was no effect on the gross morphology of the endomembrane system over the time course of the experiment. These results are consistent with tyrphostin A23 inhibition of the interaction between the tyrosine-based internalization motif on the TfR and the 2 medium chain subunit of the AP-2 adaptor complex. These data show that tyrphostin A23 can be used as an inhib-itor of endocytosis in mammalian cells and that, when using tyrphostins as inhibitors of tyrosine kinases, any observed effects on membrane trafficking events should be interpreted with caution due to possible interaction with chains. It is possible that different structural analogs of tyrosine could inhibit the interaction between tyrosine-based motifs and other adaptor complex medium chains. Future molecular modeling could potentially identify new drugs designed to inhibit other membrane trafficking pathways. These drugs could target different adaptor complex medium chains or their as yet unidentified homologs. Thus, specifically designed tyrphostins may become useful tools for dissecting membrane trafficking pathways in eukaryotic cells.