Determinants of substrate recognition in the protein-tyrosine phosphatase, PTP1.

Photoaffinity labeling has been used to identify amino acids involved in recognition of protein substrates by the protein-tyrosine phosphatase PTP1. The photoactive amino acid p-benzoylphenylalanine (Bpa) was incorporated into a phosphotyrosine-containing peptide derived from epidermal growth factor autophosphorylation site Tyr (EGFR). This peptide photoinactivated PTP1 in a time- and concentration-dependent manner. Three lines of evidence indicate that the interaction between PTP1 and the photoaffinity label was specific: 1) photoinactivation was inhibited in the presence of a non-Bpa-containing peptide from EGFR Tyr in molar excess. 2) The photoaffinity label-containing phosphopeptide was rapidly dephosphorylated by PTP1 with kinetic constants similar to those of the non-Bpa-containing peptide under identical conditions. 3) After complete photoinactivation, the level of incorporation of radioactive photoaffinity label into PTP1 was approximately 0.9 mol of label/mol of enzyme, consistent with a 1:1 stoichiometry of photolabeling. Radiolabeled peptide was used to identify sites of cross-linking to PTP1. Bpa peptide-PTP1 was digested with trypsin, and radioactive fragments were purified by high performance liquid chromatography (HPLC) and analyzed by Edman sequencing. In two parallel experiments which were analyzed using different HPLC columns, a site in the α2′ region of PTP1, most likely Ile, was labeled by the Tyr-derived peptide. The results are discussed in light of the crystal structure of human PTP1B and suggest that an additional mode of substrate recognition must exist for PTP1 catalysis.

Photoaffinity labeling has been used to identify amino acids involved in recognition of protein substrates by the protein-tyrosine phosphatase PTP1. The photoactive amino acid p-benzoylphenylalanine (Bpa) was incorporated into a phosphotyrosine-containing peptide derived from epidermal growth factor autophosphorylation site Tyr 992 (EGFR 988 -998 ). This peptide photoinactivated PTP1 in a time-and concentration-dependent manner. Three lines of evidence indicate that the interaction between PTP1 and the photoaffinity label was specific: 1) photoinactivation was inhibited in the presence of a non-Bpa-containing peptide from EGFR Tyr 992 in molar excess. 2) The photoaffinity label-containing phosphopeptide was rapidly dephosphorylated by PTP1 with kinetic constants similar to those of the non-Bpacontaining peptide under identical conditions. 3) After complete photoinactivation, the level of incorporation of radioactive photoaffinity label into PTP1 was approximately 0.9 mol of label/mol of enzyme, consistent with a 1:1 stoichiometry of photolabeling. Radiolabeled peptide was used to identify sites of cross-linking to PTP1. Bpa peptide-PTP1 was digested with trypsin, and radioactive fragments were purified by high performance liquid chromatography (HPLC) and analyzed by Edman sequencing. In two parallel experiments which were analyzed using different HPLC columns, a site in the ␣2 region of PTP1, most likely Ile 23 , was labeled by the Tyr 992 -derived peptide. The results are discussed in light of the crystal structure of human PTP1B and suggest that an additional mode of substrate recognition must exist for PTP1 catalysis.
Tyrosine phosphorylation of proteins is a fundamental mechanism for the control of cell growth and differentiation. In vivo, this process is reversible and dynamic; the phosphorylation states of proteins are governed by the opposing actions of protein-tyrosine kinases, which catalyze protein-tyrosine phosphorylation, and protein-tyrosine phosphatases (PTPases), 1 which are responsible for dephosphorylation (1,2). The functional role of PTPases in cellular signaling processes is just beginning to be appreciated (3). PTPases constitute a growing family of transmembrane (receptor-like) and intracellular enzymes that rival the protein tyrosine kinases in terms of structural diversity and complexity. Although many PTPases are proteins of greater than 400 amino acids, their catalytic domains are usually contained within a span of 250 residues referred to as the PTPase domain. This domain is the only structural element that has amino acid sequence identity among all PTPases from bacteria to mammals (4).
A central question in the field is how PTPases distinguish the diversity of substrates that they encounter in the cell. Despite the rapid progress in the identification and characterization of new PTPases, there have been relatively few biochemical analyses of the mechanisms that govern PTPase substrate specificity. Evidence suggests that both catalytic and noncatalytic regions of PTPases are important for phosphotyrosyl substrate recognition. In some cases, noncatalytic domains localize the enzymes to specific intracellular compartments in which the effective local concentration of a substrate is high. For example, a COOH-terminal extension of the PTP1B tyrosine phosphatase has been shown to be necessary and sufficient for targeting the enzyme to the cytoplasmic side of the endoplasmic reticulum (5). In other cases, noncatalytic segments of PTPases play a role in modulating enzyme activity in an allosteric fashion. Occupancy of both SH2 domains of the PTPase SH-PTP2 (also known as Syp, PTP1D, or PTP-2C) stimulates phosphatase activity (6,7). Thus, cellular SH-PTP2 activity and specificity is regulated by the interaction of the enzyme with phosphotyrosine-containing proteins that bind to its SH2 domains.
PTPase substrate specificity must also be mediated by intrinsic substrate specificities of the active site as well as structural features in the vicinity of the phosphorylated tyrosine residue. Using synthetic phosphotyrosine-containing peptides that correspond to natural phosphorylation sites in proteins, several groups have demonstrated that PTPases display a range of k cat /K m values for these relative short peptide substrates (8 -12). In fact, the k cat /K m values for some of the peptide substrates approach the efficiency limited by diffusion events, suggesting that short optimal phosphopeptide sequences may contain all the information that is needed for in vivo PTPase recognition (9,10).
The human PTP1B is the founding member of the PTPase family (13). The three-dimensional structure of the catalytic domain (residues 1-322) of PTP1B has been determined (14). The structural homolog of the human PTP1B from rat, PTP1, is one of the most extensively studied PTPases (9,10,(15)(16)(17)(18). The catalytic domain of PTP1 (residues 1-322) is 97% identical to the corresponding 322 residues of the human PTP1B. We have shown previously that amino acid residues flanking the phos-photyrosine are important for efficient PTP1 catalysis (Table I and Refs. 9, 10, and 17). For example, the k cat /K m value for the undecapeptide, EGFR 988 -998 (epidermal growth factor autophosphorylation site Tyr 992 , residues 988 -998) (Asp-Ala-Asp-Glu-pTyr-Leu-Ile-Pro-Gln-Gln-Gly) is 3220-fold higher than that of phosphotyrosine (Table I). We further demonstrated that a minimum of six amino acid residues are required for the most efficient PTP1 binding and catalysis. These include phosphotyrosine, four amino acid residues NH 2 -terminal and one amino acid residue COOH-terminal to the phosphotyrosine. Indeed, the hexapeptide, Asp-Ala-Asp-Glu-pTyr-Leu-NH 2 , is an excellent PTP1 substrate that exhibits a k cat /K m of 2.24 ϫ 10 7 M Ϫ1 s Ϫ1 (17). The recently solved crystal structure of the active site Cys 215 to Ser mutant of PTP1B complexed with the high-affinity substrate Asp-Ala-Asp-Glu-pTyr-Leu-NH 2 (19) reveals specific interactions between acidic side chains of the substrate and basic residues of the enzyme. However, kinetic studies have demonstrated that PTP1 can hydrolyze a variety of peptide substrates differing in sequence and length with almost equal k cat /K m values (9,10,17). For example, the peptide substrates Neu 546 -556 (Asp-Asn-Leu-Tyr-pTyr-Trp-Asp-Gln-Asn-Ser-Ser) and p60src 523-531 (Thr-Glu-Pro-Gln-pTyr-Gln-Pro-Gly-Glu) display kinetic parameters similar to those of EGFR 988 -998 (Asp-Ala-Asp-Glu-pTyr-Leu-Ile-Pro-Gln-Gln-Gly) ( Table I). These results cannot be easily explained by the observed interactions in the crystal structure of PTP1B complexed with Asp-Ala-Asp-Glu-pTyr-Leu-NH 2 and suggest that additional determinants for peptide substrate recognition by PTP1 must exist.
In order to investigate further the molecular mechanism of substrate recognition, we have in the present study carried out photoaffinity labeling experiments on the native form of PTP1 to identify region(s) of the enzyme involved in peptide substrate binding. The photoactive probe used was a phosphotyrosyl peptide derived from epidermal growth factor (EGFR 988 -998 ) autophosphorylation site Tyr 992 in which the Glu residue immediately NH 2 -terminal to the phosphotyrosine was replaced with the amino acid p-benzoylphenylalanine (Bpa) (20,21). This photoactive amino acid, like other benzophenone-type labels, will cross-link efficiently to a wide range of binding sites in proteins (22)(23)(24). Our results lead to the identification of an additional region of PTP1 which is involved in substrate recognition.

EXPERIMENTAL PROCEDURES
Materials-Potato acid phosphatase, L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin, and p-nitrophenyl phosphate (pNPP) were purchased from Sigma. ScintiVerse II scintillation fluid was obtained from Fisher. Microcon concentrators were purchased from Amicon. PTP1U323, a version of rat brain PTP1 in which a stop codon has been placed at amino acid residue 323, was expressed and purified to homogeneity as described previously (25). In this paper PTP1U323, which corresponds to the PTPase catalytic domain, will be referred to as PTP1. The peptides were synthesized by standard solid-phase methodology, purified, and characterized by mass spectrometry as described (26). The photoactive probe, Peptide 992N-Bpa (Ac-DADBpapY-LIPQQG), was a phosphotyrosyl peptide from EGFR 988 -998 in which the Glu residue immediately NH 2 -terminal to the phosphotyrosine was replaced with the amino acid Bpa. The specific activity of [ 3 H]Peptide 992N-Bpa, prepared by NH 2 -terminal acetylation with [ 3 H]acetic anhydride (23), was 23.2 mCi/mg.
Photoaffinity Labeling-Photocross-linking of PTP1 to the peptide 992N-Bpa was carried out at 4°C in a Rayonet RMR-500 photochemical reactor fitted with 8 RMR 3500-A lamps (Southern New England Ultraviolet Co., Hamden, CT) as described previously (23). The borosilicate glass tubes containing the samples were clamped 2 cm from the lamps. The peptide-based affinity labels (at various concentrations) and PTP (5 M) were combined in 50 mM Tris (pH 8.0) and photolyzed for up to 15 min. Aliquots were withdrawn after specific time intervals and diluted into an activity assay (see below).
Dephosphorylation of Peptide 992N-Bpa-The dephosphorylation of peptide 992N-Bpa (1.8 mM) by PTP1 (238 nM) was performed at 30°C in 100 mM sodium acetate (pH 5.5), 1 mM EDTA; the ionic strength of the buffer was adjusted to I ϭ 0.15 M using NaCl. Aliquots were removed at 1 min and 10 min and quenched by boiling for 5 min. The aliquots were analyzed on a ISCO 2350 HPLC system using a Vydac C18 column and mobile phases of (a) 0.1% (v/v) trifluoroacetic acid and (b) 75% (v/v) acetonitrile and 0.09% trifluoroacetic acid. The phosphorylated peptide 992N-Bpa and dephosphorylated peptide 992N-Bpa eluted at 26.9 and 28.1 min, respectively. The elution time of the dephosphorylated peptide was confirmed by incubating 184 mmol of peptide 992N-Bpa with 1.05 units of potato acid phosphatase in 50 mM sodium citrate, for 90 min at pH 5.0 and 37°C. The reaction was then analyzed by HPLC under the same conditions as described above.
To obtain kinetic parameters for the PTP1-catalyzed hydrolysis of the phosphorylated peptide 992N-Bpa, a continuous spectrophotometric assay described previously (9) was employed. This assay takes advantage of the difference in absorbance at 282 nm between phosphotyrosine and tyrosine and can be utilized to follow the complete time course of the enzyme catalyzed hydrolysis of phosphotyrosine containing peptide. The complete time course of the reaction can be fitted to the integrated form of the Michaelis-Menten equation (Equation 1) using a nonlinear least squares algorithm. All of the experiments were performed at 30°C in either pH 6.6 or pH 8.0 buffers. Buffers used were as follows: pH 6.6, 50 mM 3,3-dimethylglutarate, and pH 8.0, 50 mM Tris. Both buffer systems contained 1 mM EDTA and the ionic strength of the solutions were kept at 0.15 M using NaCl.
Enzyme Activity Assay-The PTP1 activity assay used to measure enzyme photoinactivation was performed at 30°C for 15 min. The reaction mixture (0.2 ml) contained 10 mM pNPP in 100 mM sodium acetate (pH 5.5), 1 mM EDTA with an ionic strength of 0.15 M (adjusted with NaCl). The reaction mixture was placed in a 30°C water bath for 5 min prior to the addition of PTP1 (40 nM). After 15 min, the reaction was quenched with 1 ml of 1 N NaOH. The amount of product (pnitrophenol) produced was measured from the absorbance at 405 nm, using a Beckman DU-7 spectrophotometer. The nonenzymatic hydrolysis of pNPP was corrected by measuring the increase in absorbance at 405 nm obtained in the absence of enzyme.
Identification of Modified Sites-Two preparative photocross-linking reactions of the Bpa containing peptide to PTP1 were performed under similar conditions. In the first reaction, 30 nmol of PTP1 was incubated with 300 nmol of peptide 992N Bpa (7.6 ϫ 10 4 dpm/nmol) in 50 mM Tris at pH 8.0 in a volume of 600 l and irradiated for 30 min at 4°C. The reaction was transferred to a Microcon-10 concentration unit (Amicon; molecular weight cutoff, 10,000) to remove the excess photoaffinity label. After washing, the recovered material was incubated with 1:100 (w/w) trypsin in 100 mM NH 4 HCO 3 for 24 h at 37°C. The only difference in the second photoaffinity labeling reaction was the amount of peptide 992N-Bpa (100 nmol) and enzyme (25 nmol) used in the experiment (final volume ϭ 350 l). Trypsin was removed by filtration through a Microcon-10 concentration unit. The unit was washed until Ϸ98% of the radioactivity applied to the Centricon-10 had passed through to the filtrate. Peptides in the filtrate were analyzed by reverse-phase HPLC by monitoring absorbance of the peptides at 220 nm. The first reaction was analyzed on a Vydac C18 column with mobile phases of (a) 0.1% (v/v) trifluoroacetic acid and (b) 75% (v/v) acetonitrile and 0.09% trifluoroacetic acid. A Vydac C4 column was used to analyze the second reaction. In each case, the peptides were eluted with a linear gradient of 2-98% solvent B in 105 min and a 1 ml/min flow rate. The radioactive peaks were identified using a Beckman scintillation counter. As a standard, unmodified PTP1 (4 ϫ 400 g) was incubated with trypsin under the conditions described above. Two of these digests were analyzed by HPLC on a C18 column and the remaining two were analyzed on a C4 column.
Radioactive peaks from the modified digests were concentrated in a Speedvac concentrator (Savant) and analyzed by Edman peptide sequencing using an Applied Biosystems model 475 A pulsed liquid protein sequencer. The samples were concentrated onto an Ultrafree device containing Immobilon-CD membranes (Millipore) by centrifugation at 800 rpm for 15 min. Precycling prior to normal sequencing cycles involved the following successive treatments: trifluoroacetic acid, dry, N-heptane, butyl chloride, dry, trimethylamine buffer, dry, phenylisothiocyanate in N-heptane, trimethylamine buffer, dry, N-heptane, ethyl acetate. For each peak sequenced, portions of the filtrate from the Ultrafree device, of the washes and of the precycling samples, were analyzed by scintillation counting. These samples contained negligible radioactivity (cpm-blank ϭ 10). At each sequencing cycle, one-half of the phenylthiohydantoin-derivative was identified by HPLC and the remainder was dissolved in ScintiVerse fluid and analyzed by scintillation counting.

RESULTS
Photoaffinity Labels-The p-benzoyl-Phe (Bpa)-containing peptide was synthesized using standard solid-phase methodologies and acetylated with acetic anhydride on the NH 2 terminus (26). The sequence of the peptide was based on the epidermal growth factor receptor (EGFR 988 -998 ) autophosphorylation site at position 992. This peptide has been shown previously to be an efficient substrate for PTP1 (10). NH 2 terminus acetylation does not affect the ability of the peptide to act as a substrate for PTP1 (17). The Bpa was positioned directly NH 2terminal to the phosphotyrosine (992N-Bpa). The sequence of the peptide is (Bpa in boldface): Ac-Asp-Ala-Asp-Bpa-pTyr-Leu-Ile-Pro-Gln-Gln-Gly. The photoreactivity of the purified peptide has been previously confirmed (26).
Photocross-linking to PTP1-PTP1 (5 M) was incubated with varying concentrations of peptide 992N-Bpa in 50 mM Tris (pH 8.0) at 4°C and irradiated at 350 nm. Aliquots were withdrawn at various times and assayed for enzyme activity. The photoinactivation of the enzyme proceeded in a time-and concentration-dependent manner (Fig. 1). After 1 min, approximately 50 and 86% of the enzyme's activity had been lost with the addition of 100 M and 500 M peptide, respectively (Fig.  1A). At the end of photolysis (after 15 min), less than 0.2% of the enzyme's activity remained in the samples containing Peptide 992N-Bpa (Fig. 1A). Throughout the duration of photolysis, 100% activity was maintained in the samples of PTP1 without Peptide 992N-Bpa, demonstrating that the enzyme alone is stable to photolysis conditions used.
Specificity of Photoaffinity Labeling-Protection experiments using a non-Bpa peptide were conducted to determine the specificity of the photolabeling. The photoaffinity labeling of PTP1 was performed using Peptide 992N-Bpa and a 30-fold molar excess of a non-Bpa-substituted peptide EGFR 988 -998 (Asp-Ala-Asp-Glu-pTyr-Leu-Ile-Pro-Gln-Gln-Gly). In this experiment, samples containing EGFR 988 -998 together with Peptide 992N-Bpa only lost 20% of the enzymatic activity after 15 min of irradiation, in contrast to the samples containing Peptide 992N-Bpa alone in which 94% of PTP1 activity was lost (Fig. 1B). The non-Bpa-substituted peptide partially protects the enzyme's active site from the irreversible incorporation of the photoaffinity label, suggesting that the photoaffinity labeling is directed to the active site of the phosphatase.
Further evidence supporting this binding of the Bpa-containing peptide to the active site was obtained by testing the peptide as a substrate. PTP1 (238 nM) was incubated with 1.8 mM Peptide 992N-Bpa under the activity assay conditions described above. Aliquots were withdrawn after 1 and 10 min of reaction and analyzed by high performance liquid chromatography. The phosphorylated peptide (Peptide 992N-Bpa) eluted at 26.9 min under these HPLC conditions (Fig. 2). The elution time of the dephosphorylated peptide 992N-Bpa was determined to be 28.1 min by incubating the peptide with potato acid phosphatase in 50 mM sodium citrate (pH 5.0) and analyzing on the HPLC (data not shown). After 10 min of reaction, the starting phosphotyrosyl peptide was converted to material with the HPLC mobility of authentic unphosphorylated peptide (Fig.  2). Thus, PTP1 completely dephosphorylates the peptide under these assay conditions, arguing that Peptide 992N-Bpa binds to the enzyme's active site.
In order to obtain kinetic parameters for the PTP1-catalyzed hydrolysis of phosphopeptide 992N-Bpa, we employed a continuous spectrophotometric assay described previously (9) to fol-low the dephosphorylation of tyrosine on the peptide. Values of k cat and K m for PTP1 using peptide 992N-Bpa at pH 6.6 and 30°C were 69.3 Ϯ 5.9 s Ϫ1 and 3.30 Ϯ 0.51 M, respectively. These values are similar to those for the peptide EGFR 988 -998 (Table I) and place peptide 992N-Bpa among the best substrates for PTP1. The kinetic parameters k cat and K m for peptide 992N-Bpa at pH 8.0 and 30°C were determined to be 23.4 Ϯ 1.9 s Ϫ1 and 9.20 Ϯ 0.75 M, respectively. Thus the concentrations of photoactive peptide 992N-Bpa utilized in the crosslinking experiments were well above that needed for saturation.
Sequencing of the Photocross-linked PTP1-In order to identify the residues on the enzyme modified by the Bpa-substituted peptide, two preparative cross-linking reactions were performed. The photoaffinity label was acetylated with [ 3 H]acetic anhydride on the NH 2 terminus and irradiated at 350 nm with the enzyme for 30 min at 4°C. The enzyme lost over 90% of its activity in each preparative experiment, indicating that crosslinking was as efficient as for the smaller scale reactions. The excess photoaffinity label was removed by centrifugation through a Microcon-10 concentration unit. After removal of the excess label, an aliquot of the sample was analyzed by liquid scintillation counting to determine the level of incorporation of photoaffinity label into protein. In each case approximately 0.9 mol of [ 3 H]Peptide 992N-Bpa were incorporated per mol of PTP1. (For example, in the first reaction Ϸ 2 ϫ 10 6 cpm were associated with the protein. The expected value based on 100% labeling and assuming a 1:1 stoichiometry of labeling would be 7.6 ϫ 10 4 cpm/nmol ϫ 30 nmol ϭ 2.2 ϫ 10 6 cpm).
To produce peptide fragments suitable for HPLC analysis, trypsin digestion of the cross-linked enzyme was carried out. The labeled enzyme was dissolved in 100 mM NH 4 HCO 3 and incubated with trypsin for 24 h at 37°C (conditions were determined by proteolysis of unmodified PTP1). A Microcon-10 concentration unit was then used to remove the trypsin from the reaction. The filtrates contained approximately 94% of the radioactivity; thus, the majority of the radioactivity was located within small (molecular weight Ͻ 10,000) peptide fragments. The first reaction was analyzed on an analytical Vydac C18 column (Fig. 3A). Fractions containing peaks of absorbance at 220 nm were collected and analyzed by scintillation counting after dissolving 5% of each sample in scintillation fluid. The major radioactive peak from this experiment (66% of radioactivity applied to column) eluted at 65.4 min (Peak A) (Fig. 3A). Minor radioactive peaks in the HPLC chromatogram (Ͻ10% of radioactivity) were present as well, possibly because multiple points of attachment into peptide chains are possible with benzophenones (24), and multiple stereoisomers may be formed. In addition, multiple peaks may be observed when a single cross-linked peptide exhibits differential interactions with an HPLC column (27). In addition to the labeled digest, two parallel unmodified digests of PTP1 were carried out under the same conditions. Both contained a major peak at 81.3 min not present in the modified digest (Peak B) (Fig. 3B). These results suggested that Peak B represented a tryptic fragment of PTP1, the mobility of which was altered upon cross-linking with Peptide 992N-Bpa. Peak B (Ϸ500 pmol, based on the peak height) and the radioactive peak A (Ϸ420 pmol, based on the specific activity) were concentrated in vacuo and analyzed by Edman peptide sequencing.
The second preparative cross-linking reaction was carried out under conditions similar to the first reaction. Again, Ͼ94% of the radioactivity was recovered in the Microcon-10 filtrate after trypsin digestion. In this case, radioactive tryptic fragments were separated by another method, reverse-phase HPLC using an analytical C4 column. The major peak of radioactivity in this experiment (63% of total radioactivity applied to column) eluted at 75.6 min (Peak C) (Fig. 3C). For comparison, a trypsin digest of unmodified PTP1 was carried out as described above and analyzed by HPLC on a C4 column. The peak at 75.6 min was substantially smaller in the unmodified digest, suggesting that Peak C arose by cross-linking of Peptide 992N-Bpa to a tryptic fragment (Fig. 3D). Peak C (Ϸ350 pmol, based on the specific activity) was concentrated and sequenced under the same conditions as Peaks A and B as described above.
The sequences of the peptides obtained in this manner are given in Table II. All three sequenced peptides correspond to the sequence of PTP1 in the region Ala 13 -Arg 24 . Sequencing terminated after the cycle corresponding to Arg 24 , consistent with the end of a tryptic fragment. For Peak B, which arose from the tryptic digest of native PTP1, an additional peptide was present at a lower amount. For Peaks A and C, a low recovery of phenylthiohydantoin-derivative (approximately  10% of the yield compared with the previous cycle) was observed in the 11th cycle (corresponding to Ile 23 ). These results suggest that the covalent modification took place at Ile 23 (see "Discussion"). At each sequencing cycle during the analysis of the radioactive peaks, one-half of the phenylthiohydantoinderivative was identified by HPLC and the remainder analyzed by scintillation counting. In no case was any radioactivity observed to be released from the sequencing filter in these measurements. Instead, the radioactivity remained associated with the filter (as judged by scintillation counting), even after complete sequencing of the peptide fragments. This result may reflect the hydrophobic nature of the cross-linked adduct between Ile 23 and benzoyl-Phe.

DISCUSSION
When bound to a protein, synthetic peptides containing the photoactive amino acid analog Bpa cross-link efficiently with nearby residues upon photoactivation (23). Benzophenone-type labels such as Bpa offer several advantages over other photoaffinity reagents (22)(23)(24). Benzophenones are activated at 350 nm, avoiding lower wavelengths which are potentially damaging to proteins. The photolytically generated intermediate reacts preferentially with C-H bonds, even in the presence of aqueous buffer; hence, normally unreactive amino acid side chains may be modified in a binding site. Finally, the crosslinked product is stable to proteolysis and to other manipulations involved in characterizing the labeled site. Because of these features of benzophenone photoaffinity labels, the yields of cross-linked products are often quite high, and they have been successfully applied in a wide variety of biochemical systems, including protein kinase active sites (21,23) and SH2 domains (26,28). Based on photochemical studies using model benzophenone-containing compounds, the reactive volume of Bpa may be approximated as a sphere with a radius of 3.1 Å centered on the ketone oxygen (24). Cross-linked regions of the target protein may be considered to be within this distance of the bound peptide at the point of photoactivation.
In the present study a good substrate for rat PTP1 (k cat /K m of the parent peptide EGFR 988 -998 ϭ 2.88 ϫ 10 7 M Ϫ1 s Ϫ1 ; (17)) has been substituted with Bpa at the position immediately NH 2terminal to phosphotyrosine. This position was chosen based on the results of Ala scanning experiments, which demonstrate that amino acid residues NH 2 -terminal to phosphotyrosine are important for binding and catalysis, especially at the Ϫ1 position (10). PTP1 was relatively tolerant of amino acid changes at the Ϫ1 position (relative to phosphotyrosine). Thus, a Glu to Ala substitution at this position increased K m by a factor of 4.7-fold and reduced k cat from 75.7 s Ϫ1 to 54.8 s Ϫ1 for PTP1. A comparable change produces a 126-fold drop in k cat /K m for the Yersinia PTPase (10). Although this may be due to the absence of the negatively charged glutamate, the dramatic drop in k cat /K m may also be a consequence of the inability of the alanine moiety to interact strongly with the enzyme surface in a favorable fashion. The negatively charged residues at the Ϫ1 position may be obligatory for peptide recognition by some PTPases but dispensable for others. Indeed, phosphopeptide 992N-Bpa is an excellent substrate for PTP1 with k cat /K m comparable with the parent peptide EGFR 988 -998 (Table I). This is a surprising result, since the molecular properties of Bpa differ drastically from Glu.
Because 992N-Bpa contains a phosphotyrosine residue and is a good substrate for PTP1, the photocross-linking experiments were conducted at pH 8.0 and at 4°C to reduce the rate of substrate hydrolysis (10,18). The conditions for the photoinactivation were comparable with those for the crystallization of the PTP1B C215S-phosphopeptide complex which were at pH 7.5 and 4°C (14,19). The pseudo first order rate constant for photoinactivation was 0.033 s Ϫ1 under these conditions. PTP1 catalysis is 5.8-fold slower at 3.5°C than at 30°C (18). Since k cat for 992N-Bpa hydrolysis was measured to be 23.4 s Ϫ1

FIG. 3. HPLC analysis of trypsin-digested PTP1.
Conditions for the digestions and for HPLC are given under "Experimental Procedures." Absorbance at 220 nm is shown on the y axis. Trace A, trypsin digestion of the complex between Peptide 992N-Bpa and PTP1 analyzed by C18 HPLC. The position of the major radioactive peak is indicated along with its elution time. Trace B, trypsin digestion of unmodified PTP1 analyzed by C18 HPLC. The position of the major peak, which is reduced in trace A is indicated along with its elution time. Elution times of two other peaks are given for comparison to trace A. Trace C, trypsin digestion of the complex between Peptide 992N-Bpa and PTP1 analyzed by C4 HPLC. The position of the major radioactive peak is indicated along with its elution time. Trace D, trypsin digestion of unmodified PTP1 analyzed by C4 HPLC.

TABLE II Sequenced peptides from native and cross-linked PTP1
For peaks A and C the yield of phenylthiohydantoin at the residue indicated in bold (corresponding to Ile 23 ) was approximately 10% that of the preceding residue. In peak B, which arose from the digest of native PTP1, the major and minor components were present at a ratio of approximately 2:1, based on yields of phenylthiohydantoins.
Peak Sequence A 13 Ala-Gly-Asn-Trp-Ala-Ala-Ile-Tyr-Gln-Asp-Ile-Arg 24 B 13 Ala-Gly-Asn-Trp-Ala-Ala-Ile-Tyr-Gln-Asp-Ile-Arg 24 (major) 170 Glu-Ile-Leu-His-Phe 174 (minor) C 13 Ala-Gly-Asn-Trp-Ala-Ala-Ile-Tyr-Gln-Asp-Ile-Arg 24 at pH 8.0 and 30°C, the value of k cat is estimated to be 4.0 s Ϫ1 at 4°C and pH 8. Thus, the rate for substrate turnover was 120-fold faster than the rate for photoinactivation. At pH 5.5, photoinactivation did not occur at a significant rate (data not shown). This was presumably because of the even faster dephosphorylation of the label (k cat ϭ 75 s Ϫ1 at pH 5.5 versus 23.4 s Ϫ1 at pH 8). When the phosphate group is removed from a phosphotyrosine-containing peptide, the resulting dephosphopeptide does not bind to PTPases (9). This may explain the low photocross-linking efficiency seen at pH 5.5. The Bpa-containing peptide cross-linked to a helical region in the NH 2 terminus of PTP1 in two separate experiments. The ␣-helix modified by Peptide 992N-Bpa (designated ␣2Ј in Barford et al. (14)) lies near the active site cleft of PTP1 (Fig. 4). However, this helix is situated on the opposite side of the cleft from Arg 47 , the residue in the Cys 215 3 Ser mutant of PTP1B that makes contact with the bound peptide Asp-Ala-Asp-Glu-pTyr-Leu-NH 2 at the Ϫ1 and Ϫ2 positions in the cocrystal structure (19). Arg 47 is located on a loop that connects ␣1 and ␤1 in the PTP1B structure. In the complexed structure, the NH 2 -terminal portion of the peptide adopts a twisted ␤-strand conformation, and amino acids at the Ϫ4 through ϩ1 positions make contact with the enzyme. The recognition pocket for phosphotyrosine represents the dominant driving force for peptide binding since phosphotyrosine contributes about 53% of the peptide solvent-accessible surface area (19). Invariant nonpolar residues in the PTPase catalytic domains (Tyr 46 , Val 49 , Phe 182 , Gln 262 ) form the binding site for the phenyl ring of phosphotyrosine. The phosphoryl group in phosphotyrosine is surrounded by residues corresponding to the PTPase signature motif (4). This suggests that the mechanism for phosphotyrosine recognition is similar among all PTPases. On the other hand, Arg 47 , Ile 23 , and the residues of helix ␣2Ј are not conserved among PTPases; specific recognition of substrates may be accomplished by residues which are variable. The side chain of Asp 48 in PTP1 forms two hydrogen bonds between the main chain nitrogens of phosphotyrosine and the ϩ1 residue, which are important for stabilizing the substrate peptide conformation. The guanidinium side chain of residue Arg 47 forms salt bridges with the side chains of Glu and Asp residues on the peptide (at positions Ϫ1 and Ϫ2, respectively), suggesting an important role in peptide recognition (19). Leucine at the ϩ1 position in the phosphotyrosyl peptide makes van der Waals contacts with Val 49 , Ile 219 , and Gln 262 of the enzyme. In that structure, the side chain of residue Arg 24 interacts with the carbonyl oxygen of the leucine residue at the ϩ1 position through a bound water molecule. No additional interactions between helix ␣2Ј and the bound peptide are present in the co-crystal structure (19).
Photocross-linking of Peptide 992N-Bpa to helix ␣2Ј implies that the peptide is bound to the enzyme active site in a different conformation than that observed for Asp-Ala-Asp-Glu-pTyr-Leu-amide in the crystal structure. Since the photoaffinity label-containing phosphopeptide acts as a good substrate for the enzyme, it must consequently be bound in the active site in a conformation that is compatible with PTP1 catalytic function. One possible alternative peptide binding mode would be if Peptide 992N-Bpa binds in the opposite orientation to that seen in the crystal structure, which could place the P-1 Bpa residue closer to Ile 23 . Replacement of the negative charge at position Ϫ1 in Peptide 992N-Bpa by a bulky hydrophobic Bpa residue may in fact reduce the interaction with the side chain of Arg 47 to such an extent that the mode of binding seen in the crystal structure is no longer energetically favored. Thus, interaction with helix ␣2Ј may be important in substrates of PTP1 which do not contain an acidic residue NH 2 -terminal to tyrosine. Interestingly, ␣2Ј in PTP1 is composed of amino acid residues 16 Trp-Ala-Ala-Ile-Tyr-Gln-Asp-Ile-Arg-His-Glu 26 which are fairly hydrophobic. Furthermore, in enzymes that are particularly dependent on an acidic residue at the Ϫ1 position (e.g. Yersinia phosphatase), helix ␣2Ј may not be an important determinant for peptide recognition. This hypothesis is supported by the fact that Peptide 992N-Bpa photocrosslinks only weakly to the Yersinia enzyme. 2 An alternative explanation for the results would be that 992N-Bpa binds in a similar fashion as observed by Jia et al. (19), but that the entire NH 2 terminus of PTP1, which contains the ␣2Ј helix, moves toward the bound peptide.
Although rat PTP1 and its human homolog PTP1B are localized to the cytoplasmic side of endoplasmic reticulum (5, 29), which may provide a level of regulation, in vitro they are not very specific PTPases because they dephosphorylate a wide variety of substrates. As shown in the three-dimensional structure of the catalytic domain of human PTP1B (14), the protein surface surrounding the catalytic cleft is relatively open and consists of a number of depressions and protrusions. This may allow numerous modes of peptide recognition and is consistent with the ability of PTP1 to hydrolyze a wide variety of phosphotyrosine-containing substrates with nearly equal efficiency (9, 10, 17; this study). It is reasonable to suggest, based on both kinetic and structural studies, that binding interactions between PTP1 and the side chain of tyrosine, the phosphate group, the main chain nitrogens of phosphotyrosine and the ϩ1 residue are conserved for all peptide substrates. The orientation of individual peptide/protein substrates may be different from each other and may be dictated by specific interactions between amino acid side chains in the vicinity of phosphotyrosine and residues near the enzyme active site cleft. Using photoaffinity labeling techniques, we have identified a unique mode of substrate recognition by PTP1, which is distinct from that observed in the crystal structure with a different peptide. These observations have implications for the design and development of PTPase inhibitors; they indicate that a systematic investigation of amino acid specificity at sites in close proximity to the phosphotyrosine may reveal sequences that are preferentially recognized by PTPases. FIG. 4. Crystal structure of PTP1 complexed with tungstate. A ribbon diagram of the PTP1 structure (14) was created using the program MOLSCRIPT (30). Tungstate ion is shown in Corey-Pauling-Koltun format. The positions of residues Arg 47 and Ile 23 (in helix ␣2Ј) are indicated.