Identification by Electrospray Ionization Mass Spectrometry of the Sites of Tyrosine Phosphorylation Induced in Activated Jurkat T Cells on the Protein Tyrosine Kinase ZAP-70*

We have developed a rapid and sensitive two capillary-column chromatography and mass spectrometry-based method for the determination of protein phosphorylation sites following recovery of individual phosphopep- tides from two-dimensional phosphopeptide maps. With a standard phosphopeptide, we demonstrate detection sensitivity of at least 250 fmol for this system. We applied this technique to the analysis of in vitro sites of tyrosine phosphorylation induced on the T cell-specific protein tyrosine kinase ZAP-70 in the absence and presence of ~56'"~. We show that ZAP-70 has a primary autophosphorylation site at Tyr-292, with a secondary site at Tyr-126. We also show additional phosphorylation at Tyr-69, Tyr-178, Tyr-492, and Tyr-493 upon the addition of the pro- tein tyrosine kinase, ~ 5 6 " ~ . By comparative two-dimen-sional phosphopeptide mapping, we show that ZAP-70 isolated from Jurkat T cells also autophosphorylates at Tyr-292 and "-126. Similar analysis of s2P-labeled Jurkat cells stimulated with anti-T cell receptor antibodies reveals Tyr-492 and Tyr-493 mass spectrometry (ESI-MS). of the affinity of phosphopeptides for immobilized Fe3+ ions linking microbore immobilized metal affinity chromatography (IMAC) and high performance liquid chromato~aphy (HPLC) series, with on-line detection of eluted peptides by ESI-MS. this demonstrated detection of a phosphopeptide with sensitivity of of using determine in vitro in vivo

W e have developed a rapid and sensitive two capillarycolumn chromatography and mass spectrometry-based method for the determination of protein phosphorylation sites following recovery of individual phosphopeptides from two-dimensional phosphopeptide maps. With a standard phosphopeptide, we demonstrate detection sensitivity of at least 250 fmol for this system. We applied this technique to the analysis of in vitro sites of tyrosine phosphorylation induced on the T cell-specific protein tyrosine kinase ZAP-70 in the absence and presence of ~5 6 ' "~. We show that ZAP-70 has a primary autophosphorylation site at Tyr-292, with a secondary site at Tyr-126.

isolated from Jurkat T cells also autophosphorylates at
Tyr-292 and "-126. Similar analysis of s2P-labeled Jurkat cells stimulated with anti-T cell receptor antibodies reveals Tyr-492 and Tyr-493 as the principal sites of T cell antigen receptor-induced tyrosine phosphorylation, with additional phosphorylation at the Tyr-292, but not the Tyr-126 autophosphorylation site. The high degree of sensitivity achieved with this technology should greatly facilitate the direct biochemical determination of inducible protein phosphorylation events, an experimental strategy that until now has been both time consuming and difficult.
In recent years, it has become clear that almost all cellular responses to external stimuli, hormonal, chemical, or otherwise, are to a large part regulated on an intracellular level by reversible protein modification. Of particular importance is the role of protein phosphorylatioddephosphorylation events, which are carried out by protein kinases and phosphatases, respectively. I t is now known that many of the critical early phosphorylation events that occur within a cell following stimulation are performed by protein tyrosine kinases (PTK1.l The importance of these molecules in regulating cellular function, especially in the regulation of cell proliferation, is now well established and is confirmed by the oncogenic potential of mutation of a number of PTKs (for recent reviews, see Refs. In T cells, a number of PTKs are known to play significant roles in the generation of a response to the engagement of the T cell antigen receptor (TCR). These include members of the src family of PTKs, in particular ~5 6 ' '~ and p 5 P , along with ZAP-70, a member of the syk family of PTKs (for reviews on TCR signaling and T cell PTKs, see Refs. 4-7). ~5 6 ' "~ and p5@ have been extensively studied in a number of model and transgenic cellular systems, and at least one or both have been shown to be vital components in multiple TCR-mediated signaling pathways, including positive and negative selection during T cell development (8)(9)(10)(11), TCR-induced proliferation (10, 121, interleukin-2 production (13,14), as well as TCR-mediated killing (15,16). The activation of both ~5 6 ' "~ and p 5 9 is itself regulated via reversible tyrosine phosphorylation at a conserved C-terminal tyrosine residue, most likely involving the PTK p5OCsk (17)(18)(19) and protein tyrosine phosphatase CD45 (20)(21)(22). In addition, ~5 6 "~ may play a role in interleukin-2-mediated T cell responses through its interaction with the interleukin-2 receptor (23-26).
On the other hand, relatively little is known about the role of ZAP-70 in the TCR response. Following receptor engagement, a number of TCR subunits, in particular the TCRS and C D~E become multiply tyrosine-phosphorylated on a conserved motif, sometimes referred to as a tyrosine-based activation motif (TAM) or antigen recognition activation motif, occurring in a number of lymphocyte receptor complexes. (For recent reviews on the role of TAMs in TCR signaling, see Refs. [5][6][7]. It is known that ZAP-70 binds specifically to the tyrosine-phosphorylated TAMs of the TCRS and C D~E subunits (27)(28)(29)(30)(31)(32). This event requires the two SH2 domains of ZAF"70 and the presence of two phosphotyrosine (pTyr) residues in each TAM (30, 32,33). SH2 domains are conserved protein structures, known to mediate protein-protein interactions through their binding to pTyr-containing sequences (for an overview of SH2 domain function, see Ref. 34).
its association with the phosphorylated form o f TCRS (27, 29-32, 35)2 and CD3e subunits (31, 32,35). Recent studies have shown that ZAP-70, like ~56"' and p59"", is vital to the induction of a full TCR-mediated response, including the induction of tyrosine phosphorylation events, through its interaction with the phosphorylated forms of TCR subunits. These studies include the analysis of immunodeficient individuals lacking ZAP-70 expression (36"38), or in cells in which the binding of ZAP-70 to phosphorylated TCRS is blocked.2 However, since little is currently known about the regulation of ZAP-70 activity and the nature of its potential substrates, the mechanism by which it plays such a critical role in TCR-mediated signaling remains unclear. Since it is now clear that many protein kinases are regulated via reversible tyrosine phosphorylation (for example, all members of the src family), the dete~ination of the TCR-induced sites of tyrosine phosphorylation on ZAP-70 should be an important first step in answering the above questions.
Current techniques for the analysis of protein phosphorylation rely largely on two-dimensional (2D) phosphopeptide mapping of isolated proteins which have been phosphorylated either via in vitro kinase assays, or in vivo following metabolic labeling of cells with radiolabeled phosphate. However, due to the very small quantities of most of the phosphoproteins of interest present in living cells and the generally low stoichiometry of protein phospho~lation, direct biochemical determination of phosphorylation sites inducible in vivo is not presently a viable experimental approach. In order to circumvent such limitations, we combined and adapted a number of current techniques in column chromato~aphy and electrospray ionization mass spectrometry (ESI-MS). We took advantage of the affinity of phosphopeptides for immobilized Fe3+ ions (39) by linking microbore immobilized metal affinity chromatography (IMAC) and high performance liquid chromato~aphy (HPLC) columns in series, with on-line detection of eluted peptides by ESI-MS. With this system, we demonstrated detection of a synthetic phosphopeptide with sensitivity of at least 250 fmol of injected peptide. By using recom~inantly expressed and isolated ZAP-70 and p56lCk, we were able to recover tryptic phosphopeptides from 2D peptide maps and determine the sites of ZAP-70 autophosphorylation and those induced on ZAP-70 by ~56'"' in vitro. The generation of tryptic phosphopeptide maps of ZAP-70 isolated from stimulated Jurkat T cells following metabolic labeling, and their comparison with our in vitro generated data thus allowed us to determine the sites of tyrosine phosphorylation induced on ZAP-70 following TCR engagement on the basis of the co-migration of in vitro and in vivo derived phosphopeptides.
Since many o f the i m p o~n t phosphoprotein components of cell signaling pathways have now been identified, the availability of expressed forms of these proteins means that such an approach for the determination of the in vivo phosphorylation sites on such molecules should be a more rapid approach than by mutagenic analysis of all potentia1 phosphorylation sites (ZAP-70 itself contains over 30 tyrosine residues). Additionally, by its very nature, direct biochemical determination of phos-pho~lation sites is more reliable than indirect analysis via mutagenesis. The availability of rapid, sensitive, and direct approaches for the determination of inducible phosphorylation sites on ZAP-70 and other proteins of interest in cell signaling pathways should thus greatly facilitate investigation of the specific roles played by these modifications via site-directed mutagenesis.
R. L. Wange, N. Isakov, T. R. Burke, Jr., A. Otaka, P. P. Roller, J. D. Watts, R. Aebersold, and L. E. Samelson, submitted for publication. EXPERIMENTAL PROCEDURES Chemicals and Reagents-All laboratory chemicals were from Fisher Scientific and of appropriate purity, unless otherwise stated. Solvents for 2D peptide mapping were from British Drug House and were of HPLC grade or better, trifluoroacetic acid was from Applied Biosystems. 18-megaohm water was purified on a Barnstead NanoPure system. Nitrocellulose membranes were from Schleicher & Schuell; RPMI 1640 culture media were from the Terry Fox Laboratories (Vancouver, BC); fetal calf serum from Life Technologies, Inc. Reagents and standards for gel electrophoresis were purchased from Bio-Rad; phenylmethylsulfonyl fluoride (PMSF) from Boehringer Mannheim; 6-mercaptoethanol fmm Millipore; 32P-labeled reagents from ICN Biomedicals; ATP, poly-~n y l p~o l i d o n e -4 0 , protein G-Sepharose, 6-methylaspa~ic acid, soybean trypsin inhibitor, and aprotinin from Sigma; sodium orthovanadate was obtained from British Drug House; dithiothreitol from Calbiochem; sequencing grade trypsin from Promega; cellulose thin layer chromato~aphy plates and x-ray film from Kodak.
Synthetic peptides were generously provided by Dr. I. Clark-Lewis (Biomedical Research Centre, University of British Columbia). Phospho-<-beads were prepared essentially as described elsewhere (29) and were generously provided by Dr. P. Orchansky (Depa~ment of Microbiology, University of British Columbia). The monoclonal antibody (mAb) OKT3 was prepared as described elsewhere (40), and 9ElO mAbs were prepared as ascites fluid.
~~c~o I~A C Column ~hro~a~ography-A 50-pm inner diameter x 360-pm outer diameter x 20-cm long fused silica capillary (Polymiero Tech. Inc., Phoenix, A Z ) was inserted 5-10 mm into a 250-pm inner diameter x 1.59-mm outer diameter x 6-cm long piece of Teflon tubing (Mandel Seientific Company Ltd., Guelph, Ontario), which was connected with standard fittings to a 1 4 syringe filled with 100 of a 50% slurry of chelating Sepharose Fast Flow (Pharmaeia Biotech Inc.) in 20% ethanol. Since the gel bead diameter was larger than the outlet capillary inner diameter, no terminating frit had to be used at the outlet of the microIMAC column.
A p p r o~m a~l y 3 cm of the Teflon tubing was Nted manually with chelating Sepharose beads as monitored under a stereo microscope (final column volume, -1.5 pl). After disconnection from the syringe, the Teflon tubing was cut 5-10 mm above the packed beads and a 5-cm long piece of fused silica capillary (Wpm inner diameter x 360-w outer diameter) was inserted to close the open end of the column. Fig. 1 shows schematics of the instrumentation described below. The assembled microIMAC column was connected to a Rheodyne (Cotati, CA) 8125 injector equipped with a 5-1.11 sample loop and washed with water delivered by a Harvard apparatus (South Natick, MA) syringe pump at a flow rate of 5 pymin for 10 min. The column was activated with five injections of 5 p1 of 30 mM FeC1, solution at 1-min intervals, then washed for another 10 min with water. The water was similarly replaced with 0.1 M acetic acid, and the column washed with the same for at least 10 min at a flow rate of 5 pl/min. Prior to the initial use, the microIMAC column was washed by injecting 5 pl of 0.1% ammonium acetate, pH 8, containing 50 m~ Na,HPO, (elution buffer).  (Fig. 1B). Bound phosphopeptides were then eluted by injecting 5 pl of elution buffer, the IMAC column eluate filling the 7-pl sample loop (Fig. u9 ) prior to delivery on-line to the HPLC-ESI-MS system (see Fig. IC). The exact timing for the valve switching was determined using a 32P-labeled phosphopeptide and collecting 1-p1 fractions following the injection of 5 p l of elution buffer onto the microIMAC column. 80-90% of the radioactivity was reproducibly recovered in fractions 2-6, thus for the above column dimensions and flow rates, a time delay of 96 s (8 p1) was used to allow eluted phosphopeptide(s) to fill the 7"pl loop prior to switching the valve to apply the phosphopeptide to the HPLC column (Fig. 1C). Any remaining sample bound to the micro-IMAC column was washed out by injecting another 5 pl of elution buffer onto the column and washing to waste, while the HPLC column gradient was developed. By switching the solvent back from acetic acid to water and then to water/ethanol(4:1 v/v), the microIMAC column could be stored at 4 O C for 2-3 weeks without a reduction in performance. For re-use, the microlMAC column was preconditioned as described above for a new column, except that a 10-min wash with elution buffer was performed prior to the coIumn activation.
Narrow Bore HPLC Column Ch.romatography-The instrumentation used and HPLC-ESI-MS analyses performed were a modification of IMAC and HPLC columns were connected on-line to a triple quadrupole electrospray mass spectrometer and data station as shown. Full details of the system are given under "Experimental Procedures." Expanded views of the switching valve from Panel A (*) used to link the microIMAC column to the HPLC system are shown. B, valve position for IMAC column loading, washing, and subsequent elution of bound phosphopeptides into the sample loop. C , valve position for loading contents of the sample loop onto the HPLC system and subsequent development of HPLC column. those described elsewhere.3 Fig. L4 gives a schematic summary of the instrumentation used in these studies.
A 320-pm inner diameter x 15-cm long C,, reversed phase HPLC column (Micro-Tech Scientific, Sunnyvale, CA) was connected to port 4 of a Rheodyne 7000 valve, and by a fused silica capillary (50-pm inner diameter x 150-pm outer diameter) to the electrospray probe tip of a PE Sciex (Thornhill, Ontario) APIIII triple quadrupole mass spectrometer. Valve port 5 was connected with PEEK-tubing (0.005-inch inner diameter) (Upchurch Scientific Inc., Oak Harbor, WA) to the gradient pump system of a Michrom Ultrafast Microprotein Analyzer (Michrom Bio-Resources Inc., Pleasanton, CA). Chromatography solvents were 0.05% trifluoroacetic acid, 2 4 acetonitrile in H,O (solvent A), and 0.045% trifluoroacetic acid, 80% acetonitrile in H,O (solvent B). Following a 10-min isocratic wash in solvent A, the HPLC column was developed with a gradient from 0 to 50% solvent B over 15 min a t a flow rate of 5 pVmin. A pre-column flow-split was installed to reduce the flow rate from 100 pVmin generated by the pump system to the 5 pVmin required for the capillary column.
For analyses which required the microIMAC column to precede the HPLC column, the microIMAC outlet capillary was connected to port 1 of the switching valve with port 2 as exit to the waste, with a 7-pl sample loop installed between ports 3 and 6 (see Fig. 1, R and C ) . In the IMAC column load position of the switching valve, ports 1-6, 2-3, and 4-5 were connected, allowing the eluate from the syringe pump-driven low pressure microIMAC column to pass through the sample loop to waste (Fig. IR), while the HPLC column remained connected to the high pressure pump system. Switching the valve (port connection 3-4 and 5-6) thus allowed the contents of the sample loop to be applied to the HPLC column, while the microIMAC system maintained its flow through the port 1-2 connection (Fig. 1C). The valve was switched back to its starting configuration after 5 min to reduce void volumes and time delays on the HPLC system. Mass Spectrometric and Data Analyses-ESI-MS analyses were performed on a PE Sciex APLTII triple quadrupole mass spectrometer, equipped with a pneumatically assisted ESI source (ion spray). The mass spectrometer scanned repetitively over a mass to charge ratio Computer software provided with the mass spectrometer permits mass spectra to be displayed for any observed peak of ion detection events, allowed background subtraction, as well as extraction of a defined input mass or mass range from the data set. A deconvolution algorithm was used to calculate peptide masses based on computer matching of observed signals with the predicted m / z values for the various possible charge states of the same peptide, and to perform theoretical fragmentation of any input protein sequence, listing all possible fragments along with their predicted charged mass values (M + H*).
'Thodimensional Phosphopeptide Mapping-Following SDS-polyacrylamide gel electrophoresis (PAGE), phosphoproteins were either digested in situ following Coomassie Blue staining, destaining, and partial vacuum drying of excised gel slices, or immobilized on a membrane following transfer to nitrocellulose and blocking in 1% polyvinylpyrrolidone-40, 100 m M acetic acid for 1 h a t 37 "C (41). Proteolytic digestion was overnight a t 37 "C in 150-200 pl of 1% ammonium bicarbonate, pH 8.3, with 1 pg of trypsin added to each sample. Eluted peptides were recovered in the supernatant, Cerenkov counted in a Packard 2200CA Tri-Carb liquid scintillation counter to quantitate peptide recovery, and dried under vacuum. Samples were then resuspended in a minimal volume of water and spotted onto (20 x 10 cm) cellulose TLC plates. These were electrophoresed in the first dimension (20 cm) in water/ acetic acidpyridine (89:10:1, v/v) a t 10 "C and 1000 V for 110 min, dried in air, and developed in the second dimension (10 cm) in water/pyridine/ butan-I-ol/acetic acid (34:30:30:6, v/v). The plates were extensively dried in air, and phosphopeptides subsequently visualized by autoradiography.
Phosphopeptides required for further analysis were recovered from the cellulose TLC plate by carefully removing the cellulose matrix containing the phosphopeptide spot of interest from the plastic support with a surgical blade. The matrix was resuspended in 150-200 p1 of water/acetonitrile (4:1, v/v) and samples placed in a water bath sonicator for -15 min to break up the cellulose. Following centrifugation, phosphopeptide supernatants were removed and the recovery quantitated by Cerenkov counting the supernatants and residual cellulose pellets. These counts were also used to estimate the quantity of each phosphopeptide present prior to IMAC-HPLC-ESI-MS analyses. Samples were reduced to a minimal volume in a Speedvac concentrator to facilitate easy loading onto the microIMAC column.
Cloning and Expression of DU"70 in a Baculovirus System-The generation of the ZAP-70-con~ining baculovirus transfer vector will be described elsewhere: Briefly, a full length cDNA of the human ZAP-70 gene was inserted into the baculovirus transfer vector, pVL1393 (Invitrogen, San Diego, CAI. The W -7 0 cDNA also had an epitope tag derived from the human c-Myc protein (SMEQKLISEEDLN) which is recognized by the mAb 9ElO (42) added at the C terminus. DH5a Escherichia coli cells were transformed with the ligation product, and clones were screened for the appropriate 11.7-kilobase plasmid containing a 1.9-kilobase insert that could be excised by digestion with BamHI and EcoRI. One of the positive clones, designated pVL1393-ZAPmyc-6, was used to generate plasmid for baculovirus expression by Invitrogen. ZAP-70 in vitro phosphorylation assays were performed by the addition of 20 pl of 50 mu Tris, pH 7.5,20 m~ MnCI,, 0.1% Nonidet P-40,250 p~ ATP, with ~Y~P I A T P added to either 880 d p~p m o l (for preparative 2D peptide mapping) or 8800 dpdpmol (for analytical 2D peptide mapping). When required, recombinantly expressed and purified ~5 6~' * (43) was also added in three 5-pl aliquots at 20-min intervals. Phosphorylation reactions were performed for 1 h at 30 "C with frequent mixing. Reactions were terminated with the addition of SDS-PAGE sample loading buffer, boiled for 5 min and run on 10% SDS gels. 2D phosphopeptide mapping was performed in situ in the (wet) gel as described above.
In Va'tro P h o s p h o r y~~~o n ofJurkat ZAP-70"Jurkat cells were maintained in RPMI 1640, supplemented to 10% with fetal calf serum, 200 m M L-glutamine, and 25 p~ f3-me~aptoethanol. Cells were spun down and lysed at 1 x 107/ml in ice-cold 50 m M Tris, pH 7.5, 150 m M NaCl, 5 m M NaF, 2 m M EDTA, 0.5% Triton X-100,l m M Na,VO,, 1 m M Na,MoO,, 1 m M PMSF, 10 pdml soybean trypsin inhibitor, 5 pgimt aprotinin. Insoluble matter was removed by centrifugation, and supernatants were precipitated at 4 "C for 4 h with 10 pl of a 50% slurry of agarose beads covalently coupled to either phosphorylated synthetic TCRy (residues 52-1641, non-phosphorylated TCRC (residues 52-1641, or uncoupled beads alone (for a description of the generation of these affmity beads; see Ref. 29). Precipitates were subsequently washed four times each with 1 ml of the above lysis buffer. In uitro phosphorylation assays were performed on the material still bound to the beads for 30 min at 30 "C with frequent mixing in 25 m M Tris, pH 7.5, 10 m M MnCI,, 1 ATP, with [ySPlATP added to 0.5 pCilpmo1. Reactions were terminated with the addition of SDS-PAGE sample loading buffer, boiled for 5 min, and run on 10% SDS gels. 2D phosphopeptide mapping was performed as described above following transfer of the gel to nitrocellulose.
Jurkat Cell Labeling and Stimulation-Cells were spun down and resuspended at 1.6 x 107/ml in phosphate-free RPMI 1640, supplemented t o 10% with dialyzed fetal calf serum, and aliquoted into three 1-ml samples. To one sample was added 10 mCi of (carrier free) 32plabeled orthophosphate, to the other two, 5 mCi each were added. Cells were incubated for 1.5 h at 37 "C. To two samples (one 5- aprotinin. Insoluble matter was removed by centrifugation, and supernatants were precipitated with non-phospho~lated or p~o s p h o~l a t e d TCRC beads as described above (the n o n -p h o s p h o~l a~ TCRC beads were added only to the stimulated, 5 mCi-labeled sample). Precipitates were subsequently washed four times each with 1 mi of the above lysis buffer, boiled for 5 min following the addition of 100 pl 1 x SDS-PAGE sample buffer, and run on 10% SDS gels. 2D phosphopeptide mapping was performed as described above following transfer of the gel-separated proteins to nitrocellulose.

RESULTS
We have previously described a method for the recovery of phosphopeptides from 2D phosphopeptide maps and their analysis by microbore HPLC with on-line detection by electrospray ionization mass spectrometry (HPLC-ESI-MS).3 In those studies, we were analyzing phosphorylated synthetic peptides and thus had an abundant supply of material. We found that the lower sensitivity limit for this system was about 10-20 pmol of injected peptide due to background signal interference, most likely originating from the cellulose TLC plate-coating material or the solvent system used during the chromatography dimension? To develop a technique more suited to the analysis of phosphopeptide maps derived from phosphoproteins isolated by immunoprecipitation from cultured cell lines, an improvement in sensitivity was required.
Earlier studies had shown that phosphopeptides have an affinity for immobilized Fe3* ions (39). However, attempts to couple Fe"' IMAC columns with MS detection resulted in the simultaneous detection of all phosphopeptide species eluted from the column, with poor detection sensitivity (10-20 pmol lower limit) (44). Also in these studies, the optimal conditions for the elution of phosphopeptides from the IMAC column and their detection by ESI-MS are not compatible, leading to a lower detection sensitivity due to the need to compromise. Under these conditions, the IMAC column also rapidly deteriorates, making its frequent replacement necessary. We overcame these limitations by linking an HPLC column in series following a microIMAC column. This allowed both the removal of some o f the background signal resulting from the IMAC elution conditions, as well as giving separation of multiple phosphopeptides. The schematic in Fig. L4 shows the instrumentation used in these studies. The flow to the IMAC column was delivered by a mechanical syringe drive to provide a low pressure slow flow rate and to facilitate the frequent changes of buffers required for delivery to the IMAC column. Following loading and extensive washing of the IhUC column, the phosphopeptide was eluted with 0.1% ammonium acetate, pH 8, 50 m M Na2HP0,, and a switching valve was used to inject the sample onto the HPLC system (Fig. 1, B and C ) . Following a 10-min isocratic wash, a 15-min acetonitrile gradient was used to elute bound material from the HPLC column. Sample detection was by on-line ESI-MS.
The IhUC-~PLC-ESI-MS system was calibrated by loading different amounts of a synthetic phosphopeptide based on the platelet-derived growth factor receptor (PDGFR) (residues 851-863) (45). Fig. 2A shows the extracted total ion current (TIC) in the 500-2000 m / z range observed in the time window where the peptide eluted (-18 min) following a 250-fmol loading (-400 pg). The data analysis software scales the data set to the largest peak observed, the peak intensities in this case being relative to the large injection peak caused by the phosphate salt in the IMAC elution buffer (not shown). Fig. 2B shows the m i z spectrum observed within the indicated peak  (45) was injected onto the system as outlined in Fig. 1, the peptide concentration having being determined by quantitative amino acid compositional analysis (not shown). Bound material was eluted from the IMAC column with sodium phosphate, loaded onto the HPLC column, and subsequently eluted with a short acetonitrile gradient. A, the TIC observed within the mlz range of 500 to 2000 was extracted @om the intact data set, and the region of the gradient where the peptide eluted from the HPLC (12-22 min) was expanded. Peak intensities are displayed relative to the largest peak in the entire data set, the injection peak (not shown), and are expressed as a per-  Fig. 2B is the number of detection events recorded for the largest signal peak (in this case a t 776.0). By plotting this number of detection events against sample loading (Fig. 2C), we observed a linear response in the 50 fmol to 5 pmol sample range, suggesting that systematic loss of phosphopeptide due to adsorption to the experimental system i s not a problem. As is clear from Fig. 2 4 , the peak of interest is well resolved above the background. Data analysis of the later eluting peak seen in Fig. 2 3 (-20-21 min) revealed a scattering of mainly low mass signals, and thus clearly did not contain a significant peptide signal (data not shown).
For the analysis of unknown phosphopeptides (i.e. unknown mass) it is important to be able to identify the "correct" peak in the TIC data for further analysis. The PDGFR phosphopeptide was a dominant peak in the 12-22 min time window in two 250-fmol loadings, and in one of two 125-fmol loadings (data not shown). Thus the IMAC-HPLC-ESI-MS system currently has an effective lower sensitivity limit of -125 fmol for a phosp h o~p t i d e of unknown mass. For a peptide of known mass (where the data set can be extracted for the predicted mlz values), we could easily detect 50 fmol of injected PDGFR peptide signal above background mass detection levels ( Fig. 2C and data not shown).
We then compared the IMAC-HPLC-ESI-MS system with the HPLC-ESI-MS system alone to test whether it was capable of improving the detection of phosphopeptides recovered from 2D phosphopeptide maps. For this we used a phosphopeptide derived from the in uitro phosphorylation of ZAP-70 by ~5 6 ' '~ recovered from a 2D phosphopeptide map (described below). due to the characteristic "hump" of observed contaminants resulting from the 2D mapping procedure. Fig, 3 8 shows the significant improvement in the nonspecific mass background for an estimated 5.0 pmol loading of the same peptide upon inclusion of the IMAC column in chromatography system. The inset again gives the mass spectrum observed within the indicated peak. As can be seen from Fig. 3 4 the signal peak for a 5-pmol loading is well resolved above the noise levels (signal to noise -15:1), thus the detection sensitivity of phosphopeptides eluted from TLC plates appears comparable to that observed with the PDGFR standard peptide. Thus the IMAC-HPLC-ESI-MS was clearly suitable for the analysis of the low picomole quantities of phosphopeptide recoverable from 2D phosphopeptide maps. We elected to apply our IMAC-HPLC-ESI-MS technology to the determination of sites of tyrosine phosphorylation inducible on ZAP-70. Since we had both ~56''' and ZAP-70 recombinantly expressed and a T cell line (Jurkat) which expresses high levels of both FTKs, we hoped to be able to identify induced phosphorylation sites on ZAP-70 in uitro as described above, and try to match these with those produced in Jurkat cells following metabolic 32P-labeling, stimulation with anti-TCR antibodies, and isolation of ZAP-70.
Baculovirus-expressed ZAP-70 was precipitated by way of a C-terminal affinity tag recognized by the 9E10 anti-c-myc d b . Immune complex kinase assays were performed either in the

. Comparison of HPLC-ESI-MS and IMAC-HPLC-ESI-MS.
A single phosphopeptide resulting from the in vitro phosphorylation of expressed ZAP-70 by expressed p56"' following trypsin digestion and elution from a 2D phosphopeptide map, as described under "Experimental Procedures" was analyzed on both systems. Cerenkov counting was used to estimate quoted pmol values from the known ATP specific activity used in the kinase reaction. A, an estimated 16.7-pmol loading of sample onto a previously described HPLC-ESI absence or presence of purified, baculovirus-expressed p561ck (43). Samples were analyzed by SDS-PAGE, and phosphoproteins identified by autoradiography of the (wet) gel. As can be seen in Fig. 4A, the phosphorylation of ZAP-70 increases on the addition of ~5 6 ' '~. Bands were digested in situ with trypsin and recovered peptides analyzed by 2D phosphopeptide mapping. Fig. 4B shows that the expressed ZAP-70 produces a principal autophosphorylation spot (spot A ) and two lesser autophosphorylation spots (spots B and C). When p56lCk was added to the kinase reaction prior to phosphopeptide mapping (Fig. 4C), a number of additional spots were observed (spots D to G), most likely due to the phosphorylation of ZAP-70 by p56lCk, or possibly by p56lCk-induced ZAP-70 autophosphorylation. 2D phosphopeptide mapping of ~5 6 ' '~ phosphorylated in the presence of ZAP-70 (Fig. 4A, lane 2) was found to be identical to that for autophosphorylated p56lCk, suggesting that ~5 6 "~ is not a substrate for ZAP-70 (data not shown). Cerenkov counting of samples before and after digestion consistently gave a recovery of 80-90% of the phosphopeptides into solution (data not shown), suggesting that most, if not all major phosphorylation sites are represented in Fig. 4, B and C. Phosphoamino acid analyses of each spot confirmed all as containing exclusively pTyr (data not shown). Spot G was analyzed by IMAC-HPLC-ESI-MS as described above, the data set being shown in Fig.   3B. The singly charged ion (M + H+ = 1247.0) and doubly charged ion ((M + 2H)2+ = 624.5) correspond to a phosphopeptide mass of 1246.0, which matches the predicted mass of 1166.6 for a partial tryptic fragment of ZAP-70 corresponding to residues 176-186 (KLYSGAQTDGK) plus an additional 80 mass units due to the phosphate ester group.
The remaining spots (A-F) marked in Fig. 4C were analyzed in the same way, and these data are summarized in Table I. These analyses revealed that two of the autophosphorylation spots (A and B ) corresponded to autophosphorylation at the same residue (Tyr-292) being derived from partial digestion at Calculated peptide masses for a theoretical tryptic digest of the human ZAP-70 protein, corresponding to the nonphosphorylated peptide ZAP-70 residue nos. and inferred peptide sequences thus determined are given. Spot C gives only a singly charged species a t a n m l z = 632, being too small to observe additional charged species. Since only one charge state The observed mass for spot E indicates the addition of only a single phosphate ester group. Since the peptide has 2 tyrosine residues, we were observed peptide mass must be 80 mass units greater than the calculated mass to allow for the addition of a single phosphate ester group. masses.
was detected, this assignment and matching with the indicated ZAP-70 fragment was done manually.
unable to determine which was phosphorylated, or what the ratio of the two possible peptide species was.
Arg-282lArg-283, a tryptic artifact we had previously observed while mapping phosphorylation sites in TCRi.3 To confirm the assignments indicated in Table I, based on mass analysis alone, peptides were synthesized corresponding to regions of ZAP-70 containing Tyr-126, Tyr-292, and Tyr-492-Tyr-493, along with surrounding tryptic cleavage sites. These were phosphorylated in vitro with p56lck, digested with trypsin and analyzed on 2D phosphopeptide maps. By mixing these peptides, following tryptic digestion, with the corresponding peptide eluted from a 2D map of ZAP-70 phosphorylated in vitro with ~5 6 ' "~ ( Fig. 4C) and re-running the 2D analyses, the appearance in each case of a single spot for both A and C confirmed the assignments indicated in Table I (not shown). In the case of spots E and F , a partial phosphorylation of a synthetic peptide encompassing residues 482-498 of ZAP-70 was performed. Following tryptic digestion, the peptides apparently corresponding to the singly (spot E ) and doubly (spot F ) phosphorylated form were then separated on a 2D map and eluted prior to mixing with the ZAP-70-derived phosphopeptides from Fig. 4C and re-analysis. Once again, the appearance in both cases of single spots for both E and F confirmed the assignments indicated in Table I (data not shown). We next investigated the autophosphorylation of ZAP-70 isolated from Jurkat T cells to test whether it would produce the same autophosphorylation pattern as the expressed ZAP-70. We used the affinity of ZAP-70 for the phosphorylated form of TCRi as a means of isolating ZAP-70 from the cell extract. We phosphorylated a synthetic TCRi peptide corresponding to the entire cytoplasmic domain (residues 52-164) in vitro with ~5 6 "~ and coupled this to agarose beads (pi-beads). We have previously determined that these beads are capable of precipitating ZAP-70 from T cell lysates (29). Jurkat cell lysates were thus precipitated with pt-beads, an in vitro kinase assay was performed on the beads, and phosphoproteins were separated by SDS-PAGE and visualized by autoradiography following transfer to nitrocellulose. Fig. 5A shows the precipitation of a phos- The ZAP-70 band from Fig. 5A, lane 1, was cleaved with trypsin on the nitrocellulose membrane, and eluted peptides analyzed by 2D phosphopeptide mapping (Fig. 5B ). By comparison with Fig. 4L3, it is clear that the ZAP-70 precipitated from Jurkat cells by binding to pi-beads autophosphorylates in the same manner as the baculovirus-expressed ZAP-70 precipitated through a C-terminal affinity tag. Scintillation counting of the membrane before and after trypsin digestion showed a -90% recovery of counts into solution (not shown), ruling out the possibility of the selective loss of a major autophosphorylation peptide to the membrane. As before, to confirm "-292 as the major autophosphorylation site of Jurkat ZAF"70, spot A from Fig. 5B was eluted from the TLC plate and mixed with a synthetic peptide centered around Tyr-292 of ZAP-70, which had been phosphorylated in vitro with p56" and digested with trypsin. The observation of a single spot following 2D phosphopeptide mapping confirmed this assignment (not shown).
We next investigated sites of phosphorylation induced on ZAP-70 in Jurkat cells following stimulation with the anti-

+
CD3e mAb, OKT3. Jurkat cells were metabolically labeled with 32P-labeled phosphate, and ZAP-70 again precipitated with p(beads. Immunoblotting of such precipitates with anti-pTyr antibodies (4G10) had shown that the p?'yr content of ZAP-70 was undetectable in unstimulated cells and was maximal a t -2 min post-OKT3 stimulation (data not shown). Following cell labeling, stimulation and precipitation with p(-beads, phosphoproteins were visualized by autoradiography following SDS-PAGE and transfer to nitrocellulose. As shown in Fig. 6A, upon OKT3 stimulation, there is an increase in the phosphorylation of a 70-kDa protein (lane 2) which is not apparent when stimulated labeled cells were precipitated with the non-phosphorylated TCR (-beads (lane 3).
The phosphorylated 70-kDa band indicated in Fig. 6A (lane 2) was subjected to 2D phosphopeptide mapping following trypsin digestion in situ on the nitrocellulose (Fig. 6B). By comparison with Fig. 4C, it appears that the major ZAP-70 autophosphorylation, spot A (Tyr-292) is present, consistent with the reported activation of ZAP-70 following TCR engagement.' However, the major induced spot is the doubly phospho-  (Fig. 6, C-E). It was also thought that one of the minor spots visible in Fig. 6B migrating toward the negative electrode (marked x and y ) might correspond to spot G (Fig. 4 0 , thus these two spots were eluted and mixed with equal quantities each of spot G and re-run on 2D phosphopeptide maps. Fig. 6F shows the 2D map observed for spot y (Fig. 6B) mixed with spot G (Fig. 4 0 . While the spots y and G are clearly very similar in mobility, their failure to co-migrate indicates that it is unlikely that spot y corresponds to i n vivo phosphorylation of ZAP-70 at Tyr-178. Similar analysis of spot x also gave a different mobility to spot G (not shown). While the identities of these minor spots ( x and y ) remain unknown, they may in fact not be derived from ZAP-70. As can be seen from Fig. 6 A , the extensive co-precipitation of other phosphoproteins with ZAP-70 suggests that spots x and y could be derived from a -70-kDa phosphoprotein which is co-precipitating with ZAP-70 and the p(-beads. Attempts to further reduce the co-precipitating phosphoproteins were unsuccessful, suggesting that they may be interacting specifically with the phospho-( or ZAP-70 (data not shown).
From the comparison of in vitro and in vivo produced data, we can thus conclude that ZAP-70 has a primary autophosphorylation site at Tyr-292, with a minor in vitro autophosphorylation site a t Tyr-126. Additionally, from the analysis of ZAP-70 from labeled stimulated Jurkat cells, we observe induced phosphorylation of ZAP-70 a t Tyr-492 and Tyr-493, with additional phosphorylation again at Tyr-292.

Phosphorylated residue
In vitro autophosphorylation In vitro + p561ck In vivo + OKT3

DISCUSSION
Due to growing interest into the role of reversible protein modification in the regulation of a wide range of cellular events and responses, especially in the initiation of cell proliferation, the need to devise rapid and sensitive methodologies for the determination of the sites and nature of such modifications has also grown. Of particular importance is the role played by protein phosphorylation and dephosphorylation events, Conventional approaches for the analysis of protein phosphorylation usually rely on 2D phosphopeptide mapping (46), phosphoamino acid analysis and anti-pTyr immunoblotting. While these techniques are very sensitive, they do not reveal the actual phosphorylation sites. Unfortunately, the quantities of many phosphoproteins of interest obtainable from tissue samples or cultured cells are also too small, and may be phosphorylated to too low a stoichiomet~ to permit direct biochemical analysis of the phosphorylation sites.
Until now, the determination of in uivo phosphorylation sites could only be achieved by making predictions of likely phosphorylation sites. These can then be investigated by expressing mutant molecules in cell lines and looking for a change in phosphorylation patterns, or by synthesizing phosphopeptides corresponding to predicted proteolytic fragments and testing for their co-migration in a 2D peptide map with those produced following proteolytic cleavage of proteins isolated from the cells. Both methods are time consuming, may miss important phosphorylation sites, and in the case of mutagenic analysis, may lead to misinterpretation of data if the mutation affects some other function of the protein. Additionally, if there is co-ordinate phosphorylation at several sites, as in the case of the insulin receptor (471, then such mutagenic analyses will yield misleading results. Taking advantage of the wide availability in expressed form of many of the phosphoproteins known to play roles in cell signaling pathways, we devised a method both rapid and sensitive enough to be of use for the direct biochemical determination of protein phosphorylation sites. The IMAC-HPLC-ESI-MS system involved a two-step chromatography system (immobilized Fe3+ affinity and CIS HPLC connected in series) with on-line detection by ESI-MS (Fig. 1). With a standard phosphopeptide we demonstrated sensitivity down to a t least 250 fmol of injected sample (Fig. 2). By utilizing this two capillary column system, we have achieved significant improvements in detection sensitivity over previous single column strategies (44): with the added benefits of the elimination of background signals and sample separation as a direct result of the on-line connection of the two chromatography systems (see Fig. 3). The level of sensitivity demonstrated here is within the range appropriate for the analysis of peptide samples recovered from 2D phosphopeptide maps following protein labeling and isolation.
We applied this technology to the analysis of the T cell specific F' TK ZAP-70, which is known to be critical in mediating TCR response and to be inducibly p h o s p h o~l a~d upon TCR engagement (27,(29)(30)(31)351. ' The availability of expressed forms of both W -7 0 and p56' ' ' allowed us to directly determine the sites of ZAP-70 autophosphorylation, as well as a number of additional tyrosine phosphorylation sites induced on ZAP-70 upon addition of p56Ick ( Fig. 4 and Table I). Comparative 2D phosphopeptide mapping of ZAP-70 isolated from Jurkat T cells allowed us to confirm the major autophosphorylation site of T cell ZAP-70 as Tyr-292 (Fig. 5 ) and identify Tyr-492 and Tyr-493 as the two principal sites of phosphorylation induced following TCR engagement (Fig. 6). The schematic shown in Fig. 7 indicates the location of all the major phosphorylation sites we observed on ZAP-70 and indicates the conditions under which they have so far been induced. ~nte~stingly, synthetic peptides modeled on ZAP-70 around Tyr-126, "yr-292 and "yr-492-Qr-493 were found to not be substrates for the expressed ZAP-70 in vitro. 6 The role of the tyrosine phosphorylation of ZAP-70 during T cell activation is still unclear. Mutation of Tyr-292, Tyr-492, and Tyr-493 should now permit these questions to be addressed by the overexpression of such ZAP-70 mutants in T cell lines, or their expression in cell systems lacking ZAP-70 expression. Recent studies have shown that the tyrosine phosphorylation of ZAP-70, its association with the phosphorylated forms of the TCRS and C D~E subunits and catalytic activation are interdependent (27,29,30,32). 2 The observation that pF~6'"~ can induce phosphorylation of ZAP-70 at the principal sites observed in uiuo is made more interesting by the recent observation that ZAP-70 and p5SECk associate in an SH2-dependent manner (48).
We have also found that the two kinases can associate in vitro, and in a phosphory~ation-dependent manner.7 Additional~y, since the tyrosine phosphorylation of the TCRS requires functional ~5 6 ' '~ (31,49), these findings taken together would support a model whereby ~5 6 ' "~ (and/or possibly p 5 P ) is required for TCR-induced phosphorylation of Tyr-492 and Tyr-493 of ZAP-70 in uivo. However, the mechanism for ZAP-70 activation still remains unresolved, but is most likely regulated via either tyrosine phosphorylation, or its interaction with the phosphorylated TCRS or p56'", or any combination of the three. Due to the complexity of the probIem, It thus seems apparent that much careful work has yet to be done. The identification of the in uiuo sites of TCR-induced phospho~lation of ZAP-70 is thus an important first step in this process.
Unlike src-family kinases, the ZAP-70 autophosphorylation site does not lie within the kinase domain, instead being between the second SH2 domain and the kinase domain (see   . One could speculate that this region may function as a molecular hinge to alter the spatial relationship between the kinase domain and the SH2 domains, dependent on the phosphorylation state of Tyr-292. Alternatively, it may serve as a binding site for other SH2 domain-containing proteins, such as p56lCk, in the same manner as do the tyrosine phosphorylation sites between the two kinase domains of receptor PTKs, such as the PDGFR (50). However, the TCR-induced phosphorylation sites at Tyr-492 and Tyr-493 do lie within the kinase domain of ZAP-70 (see Fig. 7). There is a degree of sequence homology between the kinase domains of src-family, syk and ZAP-70 PTKs (27) and, in fact, these phosphorylation sites in ZAP-70 line up with the conserved WAR motif, common to all srcfamily members, found at the autophosphorylation site of all src kinases. Thus it is reasonable t o suppose that src-family kinases such as ~5 6 ' "~ might be able to induce phosphorylation at one or more of these residues. While src-family PTKs are regulated via reversible phosphorylation at a C-terminal tyrosine residue, ZAP-70 contains no analogous sequence. It is possible that phosphorylation of Tyr-492 and Tyr-493, or autophosphorylation at Tyr-292 of ZAF"70 play roles in its catalytic activation. They may also contribute to the ability of ZAF"70 to associate with TCRC or p56"'. Once again, the identification of these phosphorylation sites should permit these hypotheses to be tested. Unfortunately, our initial attempts to characterize the role of tyrosine phosphorylation of ZAP-70 on its in vitro activity have so far been unsuccessful since we have yet t o identify a ZAP-70-specific substrate, and are unable to remove ~5 6 "~ from ZAP-70 following an in vitro kinase reaction due to the strong interaction of the two.
It is clear that the determination of the mechanism by which ZAP-70 plays such a critical role in T cell activation will be difficult. However, the direct biochemical determination of in vivo sites of induced tyrosine phosphorylation of ZAP-70 will greatly facilitate these efforts. We believe that the IMAC-HPLC-ESI-MS technology will prove a valuable tool in not only addressing these questions, but a wide range of related problems in the area of intracellular signaling. Its high sensitivity makes the technology compatible with the analysis of proteins isolated by means of immunoprecipitation and SDS-PAGE. Mass spectrometric analysis is also compatible with the analysis of serinephosphorylated peptides: and by its nature, is also suitable for the determination of protein modifications other than phosphorylation (e.g. methylation, acetylation etc.) on the basis of the defined mass differences induced on modified peptides, In fact, the method can be directly interfaced with tandem MS techniques for the sequencing of phosphopeptides, or verification of phosphorylation sites by collision-induced fragmentation of the phosphate ester bond and the resulting elimination of 80 mass units (51-53). Finally, without direct biochemical analysis ofthe molecules involved in cell signaling pathways, the mechanism by which they create the desired response will remain elusive. Without the understanding of such mechanisms, the development of treatment for diseases resulting from dysfunctional cell signaling pathways, based on biochemical intervention, will remain severely hindered.
Research Centre: I. Clark-Lewis and P. Owen for peptide syntheses, and