Automated Identification and Quantification of Protein Phosphorylation Sites by LC/MS on a Hybrid Triple Quadrupole Linear Ion Trap Mass Spectrometer*

Complete phosphorylation mapping of protein kinases was successfully undertaken using an automated LC/MS/MS approach. This method uses the direct combination of triple quadrupole and ion trapping capabilities in a hybrid triple quadrupole linear ion trap to selectively identify and sequence phosphorylated peptides. In particular, the use of a precursor ion scan of m/z −79 in negative ion mode followed by an ion trap high resolution scan (an enhanced resolution scan) and a high sensitivity MS/MS scan (enhanced product ion scan) in positive mode is a very effective method for identifying phosphorylation sites in proteins at low femtomole levels. Coupling of this methodology with a stable isotope N-terminal labeling strategy using iTRAQ™ reagents enabled phosphorylation mapping and relative protein phosphorylation levels to be determined between the active and inactive forms of the protein kinase MAPKAPK-1 in the same LC/MS run.

Reversible protein phosphorylation is one of the most important post-translational modifications (PTMs) 1 with as much as one-third of all proteins in a mammalian cell being phosphorylated (1). Phosphate introduction occurs through a large group of protein kinases that are estimated to constitute between 2-5% of the expressed genome (2)(3)(4). Protein phosphorylation mediates most of the signal transduction pathways in eukaryotic cells, controls cell processes such as metabolism and transcription, and plays a role in intercellular communication. It is important to determine the exact site of protein phosphorylation to elucidate the physiological role of a particular phosphorylation event. Once phosphorylation sites are identified, it is extremely valuable to determine how these protein phosphorylation sites change in response to extracellular stimuli or intracellular events. Studying phosphorylation is very challenging in a highly complex protease digest mixture as individual phosphorylation sites are often only partially phosphorylated and therefore at very low concentrations.
Traditional ways to localize phosphorylation sites involve the use of 32 P to label the proteins under investigation. Protease digestion of the protein followed by HPLC separation of the peptide mixture and collection of the radioactive fractions can be coupled with Edman sequencing to determine the site of phosphorylation (5)(6)(7)(8). However, Edman sequencing has certain limitations (9,10) that have led to the rise of mass spectrometry as the preferred tool for peptide sequencing.
The ability of mass spectrometry to detect and characterize sites of phosphorylation is not without its problems. Signals of phosphopeptides in the MS spectra are, in most cases, suppressed by the relatively high concentration of non-phosphopeptides present in proteolytic digests of these samples. Even LC/MS of a single protein digest can result in phosphopeptides going undetected due to suppression from coeluting, more basic peptides. Therefore, a better approach to selectively identify and characterize phosphopeptides from digest mixtures is highly desirable. There have mainly been two approaches to overcome this issue: pre-enrichment of the phosphopeptides prior to MS analysis and selective detection in MS based on the unique fragmentation behavior of phosphopeptides.
The pre-enrichment strategy for phosphopeptides typically involves the use of ion exchange or immobilized metal ion affinity techniques, both on line (11) and off line (12,13). Both techniques have limitations in that not only phosphopeptides are isolated, but in the case of IMAC, acidic peptides are often also enriched. Methylation of the digest mixture prior to IMAC enrichment has been developed as an attempt to overcome this problem (14 -17). Even with these limitations both strategies have been successfully applied to large scale studies of human (11) and yeast (14) phosphoproteomes.
Selective detection in MS of serine, threonine, and tyrosine phosphorylation is possible by the use of a precursor ion scan of m/z Ϫ79 (PO 3 Ϫ ) on a triple quadrupole instrument (18 -20).
The resultant scan produces a mass spectrum that contains only the molecular ions of the phosphopeptides that are present in the sample. If the protein sequence is known, this information is often enough to identify the peptides that are phosphorylated in a simple mixture. To identify the site of phosphorylation or in cases were the protein sequence is not known, MS/MS is undertaken to obtain the amino acid sequence or the peptide and the site(s) of phosphorylation. However, sequencing information is difficult to obtain on a triple quadrupole mass spectrometer due to the limited sensitivity in full scan mode. This has led to the use of this precursor ion scan method in conjunction with fraction collection where phosphopeptide-containing fractions are then analyzed via a second round of mass spectrometry to sequence the phosphopeptides (19,20). The selective detection of tyrosine phosphorylated peptides has been reported by the use of precursor ion scanning in a quadrupole TOF mass spectrometer (21). Taking advantage of the high mass accuracy of TOF detection, the method utilizes the immonium ion of phosphotyrosine at m/z 216.043 for positive ion mode precursor ion scanning. The peptide sequence and site of phosphorylation is then determined by MS/MS sequencing on the same platform.
A new hybrid instrument has been discussed recently (22) that combines the capabilities of a triple quadrupole and an ion trap in a single platform through the use of novel triple quadrupole linear ion trap mass spectrometry technology. The system couples high selectivity triple quadrupole functionality such as precursor ion scanning with high performance linear ion trap performance for very sensitive full scan MS and MS/MS data. More importantly, the coupling of these two technologies allows for a variety of novel scan functions to be available that enable the automated investigation of phosphorylation sites in biological samples (23). In particular, it is now possible to couple the high selectivity and sensitivity of precursor ion scanning (for the selective detection of phosphorylated Ser, Thr, and Tyr) with the high sensitivity of an ion trap MS/MS mode (for the resultant phosphopeptide sequencing) on a single platform in an automated fashion.
Specifically when ions are detected in the highly selective precursor ion scan of m/z Ϫ79, the instrument switches from negative to positive ion mode (700-ms delay), and enhanced resolution (ER) scans are performed to obtain the accurate mass of the peptides. This is then followed by an MS/MS sequencing scan using the high sensitivity afforded by the linear ion trap mode. Such a measurement cycle takes less than 5 s to execute rendering this approach feasible for the analysis of complex mixtures even if several phosphopeptides eluted at the same time. To demonstrate the ability of this methodology to map multiple substoichiometric phosphorylation sites from isolated proteins, we mapped the in vitro phosphorylation sites on several protein kinases. This selective phosphorylated detection was then coupled with amine-reactive isobaric tagging reagents (24) (iTRAQ TM , Applied Biosystems, Foster City, CA) to obtain relative phosphorylation information between the active and inactive forms of mitogen-activated protein kinase-activated protein kinase-1 (MAPKAPK-1).

EXPERIMENTAL PROCEDURES
Samples were analyzed using nanoflow chromatography coupled to a hybrid triple quadrupole linear ion trap (4000 Q TRAPா system, Applied Biosystems) equipped with a Flow NanoSprayா source and a heated interface plate set to 150°C. The protein digests were separated using an LC Packings Integrated System (Dionex, Camberley, UK) consisting of a FAMOS TM microautosampler, Switchos TM microcolumn switching module, UltiMate TM micropump, a PepMap C 18 column (75 m ϫ 15 cm, LC Packings), and a mobile phase of acetonitrile, 0.1% formic acid with a 45-min gradient elution and a flow of 0.3 l/min. Samples were loaded onto a C 18 capillary trap (Michrom Bioresources, Auburn, CA) in 0.2% formic acid at a flow rate of 20 l/min. After ϳ5 min, the capillary cartridge was switched in line with the analytical column. Postcolumn addition of 80% propan-2-ol was used at a flow rate of ϳ100 nl/min from a syringe pump (Harvard Apparatus) connected to a microtee (Upchurch, Oak Harbor, WA) prior to the microion spray head to improve stability of the spray in negative ion mode.
Precursor ion scanning was performed over a mass range of m/z 400 -2000 at 500 amu/s (with Q1 set to low resolution and Q3 set to unit resolution) with an ion spray voltage of Ϫ2200 V applied to a Picotip FS360-50-15-N (New Objective, Woburn, MA) with ion spray gas (nitrogen). Precursors were collided in Q2 with a collision energy ramp of Ϫ65 to Ϫ110 V across the mass range. If a precursor of 79 was detected above a preset threshold value (1000 -5000 counts/s), the polarity was automatically switched to ϩ2300 V (after a 700-ms dwell at 0 V) with the same gas settings, and a positive ion enhanced resolution scan was performed at 250 amu/s to determine the charge state of the ion. Enhanced product ion scans (MS/MS) were performed at 4000 amu/s, and collision voltages were calculated automatically. Once this duty cycle was completed, the polarity was switched back to negative (after a 700-ms dwell at 0 V), and the cycle was repeated.
The protein kinases His 6 -GSK3␤ and GST-MAPKAPK-1 (expressed in Spodoptera frugiperda 9 cells) were prepared as described previously (25,26). For solution digests, 5 g of each protein kinase were denatured in 1% SDS, 10 mM DTT, and 50 mM ammonium bicarbonate at 90°C for 5 min. The proteins were then treated with 50 mM iodoacetamide for 30 min at room temperature and precipitated with 10 volumes of 20% (w/v) TCA (on ice for 10 min followed by centrifugation for 10 min at 13,000 ϫ g), and the pellets were washed with 0.5 ml of 10% TCA and then three times with 1 ml of water. The pellets were resuspended in 10 l of 8 M urea and then diluted 10-fold with 25 mM ammonium bicarbonate followed by the addition of 0.5 g of trypsin (Promega Trypsin Gold). Digests were performed at 30°C for 12 h, and the digests were diluted to 2 g/ml in 0.2% formic acid in water (v/v). iTRAQ reagent labeling was performed according to the protocol provided (Applied Biosystems). Essentially each digest was resuspended in 100 l of labeling buffer (0.25 M triethylammonium bicarbonate, 75% ethanol); the iTRAQ reagent was added, allowed to react for 30 min, and quenched with water; and the two digests were mixed.

RESULTS
The PTM discovery workflow described for the 4000 Q TRAP system relies on a stable electrospray beam in both negative and positive ion modes. With constant infusion of a synthetic phosphopeptide in negative ion mode, the electro-spray beam was noted to be relatively stable, especially if the organic content of the peptide solution was high. However, at low organic content, the beam was observed to be unstable, often due to a corona discharge from the nanospray tip. Because the workflow described here was to be applied to an on-line capillary reverse phase HPLC separation of phosphopeptides, it was logical to increase the organic content of the HPLC eluate with a postcolumn addition of organic solvent. This was achieved by the use of a second pump delivering organic solvent to a microtee prior to the connection to the nanospray tip. Various solvents were tested, and the solvent that provided the most stable spray was propan-2-ol.
By mixing 80% propan-2-ol, 20% water at 100 nl/min with the capillary HPLC eluate (300 nl/min) a final flow rate of 400 nl/min was achieved with a minimum organic content of 20% propan-2-ol.
To test the ultimate sensitivity of the method, a mixture of seven synthetic peptides, LRLSSpSSGRLR, KRRQIpSIRGIV, YPTAMpTSPR, IVADQpTPTPTR, RYPRPVpSVPPpSPSLSR, LLLRLpSENSG, and KFELLpTPPLSPSRRSG (where pS is phosphoserine and pT is phosphothreonine), were spiked in varying equimolar amounts into 500 fmol of an Escherichia coli ␤-galactosidase digest and then analyzed via LC/MS using this PTM discovery method. This represents a protein with phosphorylation at eight phosphorylation sites. With synthetic phosphopeptide standards spiked into the 500-fmol ␤-galactosidase digest at 100 fmol (20% phosphorylation level), all seven phosphopeptides could be easily detected and sequenced (data not shown). The same phosphopeptide mixture spiked in at 5 fmol (1% phosphorylation level) resulted in five of the seven peptides being detected (Fig. 1A). Two of the peptides, LRLSSpSSGRLR and KRRQIpSIRGIV, were detected with signals over 3000 counts (3 times the intensity requirement for polarity switching), but RYPRPVpSVPPp-SPSLSR and KFELLPpTPPLSPSRRSG could no longer be observed. The precursor ion scan from LRLSSpSSGRLR eluting at 22 min is shown in Fig. 1B demonstrating the good signal achievable at these low levels. The ER scan and EPI spectrum from LRLSSpSSGRLR is shown in Fig. 1, C and D, illustrating good sequence coverage is still possible at these low levels. This would suggest detection limits for this method in the high attomoles to low femtomoles.
The automated workflow of precursor ion scanning fol-lowed by positive ion enhanced resolution scans and enhanced product ion scans was applied to the analysis of a tryptic digest of glycogen synthase kinase-3␤. This PTM discovery workflow was carried out with a 45-min LC separation of a 500-fmol tryptic digest of this protein kinase, and a representative precursor of 79 scan, positive ion ER scan, and EPI scan for the phosphopeptide GEPNVSpYICSR (where pY is phosphotyrosine) are shown in Fig. 2 Table I. The identity of each peptide and phosphorylation site was determined by both a Mascot search of all the enhanced product ion scans as well as manual interpretation of the spectra. cps, counts/s.  Fig. 3.

Automated Protein Phosphorylation Site Analysis
To demonstrate the ability of this LC/MS-based method to map multiple substoichiometric phosphorylation sites, a study was undertaken to map the phosphorylation sites of MAP-KAPK-1 (also known as p90 Rsk-1) (28 -30). MAPKAPK-1 has a broad substrate specificity in vitro, phosphorylating proteins and peptides at serine residues that lie in Arg-Xaa-Arg-Xaa-Ser or Arg-Arg-Xaa-Ser motifs (30,31). However, few physiological substrates have been identified for this protein.  (Table I)  LGpSGPDGAEEIKR (Ser-307), and GFpSFVATGLMEDDSKPR (Ser-380).
Analysis was then performed using the specific phosphorylation scan in which multiple signals from phosphorylated peptides were observed. In all, 13 phosphopeptides were identified (Table I and Fig. 3) indicating that this method is compatible with relatively complex phosphopeptide mixtures. The three phosphopeptides identified in the initial LC/MS/MS study were observed together with DpSPGIPPSAGAHQLFR (Ser-363), pTPRDpSPGIPPSAGAHQLFR (Thr-359, Ser-363), AENGLLMpTPCYTANFVAPEVLKR (Thr-573), and KLPpSTTL (Ser-732). This is in agreement with a previous in vivo study in FIG. 4. The use of iTRAQ reagent to determine the relative phosphorylation between the active and inactive forms of the protein kinase MAPKAPK-1. The phosphorylated peptide LGpSGPDGAEEIKR (A) is observed to be 4-fold more abundant in the active form versus the inactive form (intensity of 116 versus 114) as shown in the expanded mass region 110 -120 Da. In contrast the unphosphorylated version LGSGPDGAEEIKR (B) is observed to have similar abundance between the active and inactive form as shown in the expanded mass region 110 -120 Da. cps, counts/s. which identification of the phosphorylation sites was performed using 32 P labeling in conjunction with Edman sequencing (28). Even more impressive was the finding of several new phosphopeptides: EIpSITHHVK (Ser-45), AGpSEK-ADPSHFELLK (Ser-54), ATQAPLHpSVVQQLHGK (Ser-402), HIFYSpTIDWNK (Thr-323), and AENGLLMpTPCYpTANFVA-PEVLKR (Thr-573 and Thr-577). The detection of previously unreported phosphorylation sites can be attributed to the fact that this analysis was performed on a recombinant form of the protein that was expressed in S. frugiperda 9 cells and activated in vitro as these sites were also not reported in a recent in vivo study using mass spectrometry (29).
Once phosphorylation sites are identified, it is extremely valuable to determine how these protein phosphorylation sites change in response to extracellular stimuli or an intracellular event. With this in mind, the need to enable relative quantification measurements of protein phosphorylation became apparent. This selective phosphorylation detection method was coupled to amine-reactive isobaric tagging reagents (iTRAQ reagents, Applied Biosystems) to obtain relative phosphorylation information between the active and inactive forms of MAPKAPK-1.
To be able to normalize the two protein samples for initial concentration, an initial run was performed in regular LC/ MS/MS mode. This was done to obtain an overall quantitation average for the active form versus the inactive form. A software program (ProQuant 1.4, Applied Biosystems) was used to identify and quantify the peptides. The average for the 16 non-phosphorylated peptides observed from MAPKAPK-1 was 0.86 Ϯ 0.25 indicating there was slightly less protein content in the active form protein preparation compared with the inactive form. LC/MS/MS analysis was then performed using the PTM discovery workflow to detect and quantitate the phosphopeptides from MAPKAPK-1 tryptic digests.
The MS/MS spectrum of the peptide eluting at 24.7 min (Fig. 4A) can easily be assigned to LGpSGPDGAEEIKR. In the spectral region 110 -120 Da (expanded region) two ions at 114 and 116 are observed originating from the inactive and active forms of the kinase. The 116 ion is ϳ4-fold more abundant (after normalization) than the 114 ion relating to the up-regulation of phosphorylation on serine 307 in the active form. In contrast, the non-phosphorylated form of the peptide is shown in Fig. 4B, and the expanded region clearly shows that the ratio of 114/116 is approximately one. All the previously discovered phosphorylated peptides were identified, and together with their relative expression ratios are shown in Table  II. In each case, phosphorylation was observed to be upregulated in the active form of the protein. Modest increases in phosphorylation levels were observed on Ser-221 and Ser-732 upon activation of the kinase (iTRAQ reagent ratios of 1.3 and 2.2, respectively; Table II) in agreement with the initial study, which also reported moderate increases of phosphorylation of these sites after activation (28). Phosphorylation was reported to be induced on Thr-359, Ser-363, and Ser-380 after activation, corresponding to iTRAQ reagent ratios of 5.7, 3.5, and 5.1, respectively. In three cases, TPRDpSPGIPPSA-GAHQLFR, AENGLLMpTPCYpTANFVAPEVLKR, and AEN-GLLMpTPCYTANFVAPEVLKR (Table II), no 114 peak could be observed, so it can be deduced that those phosphorylation sites are up-regulated, although no actual value can be assigned. It can be seen that the iTRAQ ratios are very consistent with previously reported results. Of the previously unreported phosphorylation sites, Ser-45, Ser-323, and Thr-577 showed only modest increases in phosphorylation after activation. However, Ser-403 (ATQAPLHpSVVQQLHGK) was significantly up-regulated after activation (iTRAQ reagent ratio of 5.9).

DISCUSSION
This PTM discovery method uses the direct combination of triple quadrupole and ion trapping capabilities in a hybrid triple quadrupole linear ion trap mass spectrometer to selectively identify and sequence phosphorylated peptides at low femtomole levels. The key to the success of this strategy is postcolumn addition of propan-2-ol to reduce the electrical discharge from the electrospray needle during polarity switching. It has been reported previously that methanol can be used for stabilizing negative ion electrospray mass spectrometry (32), and this idea has been included here in this workflow on the 4000 Q TRAP system. The introduction of a mixing microtee postcolumn did not affect the resolution of the chromatography dramatically but did dramatically enhance the stability of the electrospray beam during polarity switching. The major difference between performing this workflow on the 4000 Q TRAP system and the previously reported Q TRAP system (23) is the heated interface plate (33) and MicroIon-Sprayா source fitted to the 4000 Q TRAP system coupled with the make-up solvent on the overall stability and sensitivity of the method.
We have studied a wide array of synthetic phosphopeptides, recombinant phosphoproteins, transfected protein ki- nases (data not shown), and endogenous phosphoproteins isolated by immunoprecipitation (26). In all cases phosphopeptides were detected using the triple quadrupole precursor ion scan, and the phosphopeptide ion charge state and MS/MS spectra were accumulated using the linear ion trap scans automatically. This workflow was applied to both solution tryptic digests as well as in-gel tryptic digests of phosphoproteins, and the only sample enrichment that was used here was nanoflow reverse phase HPLC. The methodology described here is applicable to low nanogram amounts of phosphoproteins isolated by SDS-PAGE, although overall sensitivity is totally dependent upon the stoichiometry of phosphorylation at any given site. The selectivity of precursor ion scanning for phosphopeptides has been reported previously (20,21,23). However, there are a few cases where a precursor of 79 signal will be generated from non-phosphorylated peptides, most commonly from peptides containing cysteic acid and methionine sulfone. Therefore oxidation of both cysteine and methionine by performic acid should be avoided if this PTM discovery workflow is to be undertaken.
The phosphopeptides that will escape detection in the described workflow are those with a negative ion parent mass of less than m/z 400 and a mass greater than m/z 2000. It should be noted that during this study it was very rare that a phosphopeptide was detected in its triple charged state (in negative ion mode), which would mean the parent mass of large phosphopeptides (Ͼ40 residues) would be outside of the effective analysis m/z range of the 4000 Q TRAP system. To overcome this limitation, digesting the phosphoprotein with more than one protease should generate phosphopeptide ions within the effective mass range of the mass spectrometer. This has been proposed previously for the proteome wide study of protein expression and to assist in detecting PTMs in complex mixtures (34).
It should be noted that the described PTM discovery workflow does rely upon the same charge state for the phosphopeptide ions detected in both the negative ion (precursor ion scanning) and positive ion modes. This is because the enhanced resolution scan is performed over a narrow mass range using the mass detected by the precursor ion scan as a reference (22,23). Over the course of these studies it was noted that phosphopeptides detected by precursor ion scanning were, on the most part, detected as doubly charged ions (Figs. 2 and 3). However, in situations where the phosphopeptide contains internal basic residues, the most abundant positive ions are unlikely to be doubly charged but most probably triply or even quadruply charged. In these situations no ion signal is detected in the positive ion enhanced resolution scan, and hence no MS/MS spectral data will be acquired. The same is true for N-terminally acetylated phosphopeptides, which have been detected as doubly charged negative ions but singly charged positive ions (data not shown). To overcome this limitation a second LC/MS analysis should be performed. Using the masses detected by the precursor ion scan as a reference, multiple reaction monitoring using the neutral loss of H 3 PO 4 as a reporter (35) or an inclusion list with multiple charge states for each potential phosphopeptide can be performed.
The use of 32 P labeling has been the traditional method for detecting and then sequencing phosphopeptides from in vitro and in vivo labeled phosphoproteins (5-7, 27, 36). The power of this methodology is that all radiolabeled phosphopeptides can easily be detected, and only the radioactive fractions need to be characterized. Also the quantitation of the recovery of phosphopeptides from in-gel digestion and reverse phase HPLC can be performed, and losses can easily be identified. Although the technique is very time-consuming, it permits the detection and analysis of very large phosphopeptides that would otherwise not be detected by mass spectrometric methods alone (36). However, when this method is applied to in vivo labeling, the level of [ 32 P]orthophosphate used is extremely high (1 mCi/ml of medium), and many dishes of cells may need to be cultured in [ 32 P]orthophosphate to obtain enough labeled phosphoprotein for phosphopeptide identification (28,36). The PTM discovery method described here has been applied successfully to samples that would have traditionally been analyzed using the 32 P labeling route (26). It is envisaged that this methodology will eventually replace the use of 32 P labeling, but this non-radiolabeled workflow cannot report on all phosphopeptides. This methodology will not detect phosphopeptides that are too large for the quadrupole scan, and the method cannot report on phosphopeptides that have been lost during sample preparation. These are two areas of the methodology that will require further development.
Coupling of this methodology with a novel stable isotope N-terminal labeling strategy enabled both phosphorylation site identification and relative protein phosphorylation levels to be determined for the protein kinase MAPKAPK-1 in the same LC/MS run with good agreement with data reported previously (28,29). This iTRAQ reagent has been reported previously for the quantitation of protein expression (24) and cancer biomarkers (37) from either LC MALDI-TOF/TOF MS or electrospray LC/MS in positive ion mode. This is the first report of the use of this reagent in a targeted PTM discovery experiment on the 4000 Q TRAP system utilizing polarity switching and the full mass range sensitivity to be able to analyze and quantitate MS/MS reporter ions in the m/z 114 -117-Da range.
The ability to identify and quantitate phosphorylation sites in the same experiment could mean that the generation of phosphorylation site-specific antibodies may not be required to study protein phosphorylation in response to multiple stimuli. The workflow described is an attractive route to study the effect of three stimuli on the phosphorylation of multiple phosphorylation sites in a target phosphoprotein by using the multiplexed nature of the iTRAQ reagents. This quantitative phosphorylation site analysis has been applied to pheromone signaling in yeast using SILAC labeling and LC/MS after IMAC enrichment (38) as well as in human cells for Erk/p90 Rsk signaling (29), and it will be interesting to compare iTRAQ reagents and SILAC for the quantitation of protein phosphorylation in future experiments.