Optimization of TripleTOF spectral simulation and library searching for confident localization of phosphorylation sites

Abstract

Tandem mass spectrometry (MS/MS) has been used in analysis of proteins and their post-translational modifications. A recently developed data analysis method, which simulates MS/MS spectra of phosphopeptides and performs spectral library searching using SpectraST, facilitates confident localization of phosphorylation sites. However, its performance has been evaluated only on MS/MS spectra acquired using Orbitrap HCD mass spectrometers so far. In this study, we have investigated whether this approach would be applicable to another type of mass spectrometers, and optimized the simulation and search conditions to achieve sensitive and confident site localization. Synthetic phosphopeptides and enriched K562 cell phosphopeptides were analyzed using a TripleTOF 6600 mass spectrometer before and after enzymatic dephosphorylation. Dephosphorylated peptides identified by X!Tandem database searching were subjected to spectral simulation of all possible single phosphorylations using SimPhospho software. Phosphopeptides were identified and localized by SpectraST searching against a library of the simulated spectra. Although no synthetic phosphopeptide was localized at 1% false localization rate under the previous conditions, optimization of the spectral simulation and search conditions for the TripleTOF datasets achieved the localization and improved the sensitivity. Furthermore, the optimized conditions enabled sensitive localization of K562 phosphopeptides at 1% false discovery and localization rates. These results suggest that accurate phosphopeptide simulation of TripleTOF MS/MS spectra is possible and the simulated spectral libraries can be used in SpectraST searching for confident localization of phosphorylation sites.

Fig 1. Simulation of single phosphorylation on beam-type CID spectra.

From nonphosphorylated fragment ions observed in beam-type CID spectra of enzymatically dephosphorylated peptides, such as a, b, and y-series ions and their ammonia and water neutral loss ions (NL-A and -W), SimPhospho predicts serine/threonine phosphorylated ions (pS/pT intact ions) and their phosphoric acid neutral loss ions (NL-P ions) by a +80 Da shift and a -18 Da shift, respectively. Note, fragment ions with the combination of NL-A and NL-W were not taken into consideration. Tyrosine phosphorylated ions (pY intact ions) are predicted from the nonphosphorylated ions only by a +80 Da shift. In our previous study [39], intensities of pS/pT intact ions, NL-P ions, and pY intact ions relative to the nonphosphorylated ions were set to 10%, 100%, and 100%, respectively.

Fig 2. Experimental scheme of phosphopeptide simulation on TripleTOF CID spectra.

SimPhospho enables spectral simulation of all possible single phosphorylations. A spectral library of simulated phosphopeptides can be used in SpectraST searching. pS: phosphoserine, pT: phosphothreonine, and pY: phosphotyrosine.

Fig 3. Phosphopeptide simulation on TripleTOF CID spectra.

Phosphopeptides were subjected to enzymatic dephosphorylation, TripleTOF CID analysis, and then spectral simulation of single phosphorylations (refer to Fig 2). (A) A representative SpectraST spectral match of a synthetic phosphopeptide (lower spectrum; SGAQASSTPLpSPTR, pS: phosphoserine) with a simulated phosphopeptide (upper spectrum) is shown. Serine-phosphorylated fragment ions (pS intact ions) and their phosphoric acid neutral loss ions (NL-P ions) were predicted from nonphosphorylated ions with 10% and 100% intensities, respectively (refer to Fig 1). (B) The same match is shown, but with the peptide simulated using 50%-50% intensities. (C and D) Fragment ions at m/z 830–1030 in the spectral matches A and B are shown, respectively. In the synthetic phosphopeptide spectra, y-series ions matched to those of simulated phosphopeptides are shown in red.

Fig 4. Phosphorylation site localization on synthetic phosphopeptides by simulated spectral library searching.

(A) The synthetic phosphopeptides analyzed by TripleTOF CID were searched by SpectraST against the simulated spectral library and the results were sorted by one of the SpectraST scores, F-value or HOM1 deltadot. (B) FLRs as a function of F-value are shown. (C) FLRs as a function of HOM1 deltadot score are shown.

Fig 5. Optimization of simulation conditions for localizing phosphorylation sites on synthetic phosphopeptides.

Different simulation conditions were used for creating TripleTOF spectral libraries from the dephosphorylated synthetic peptides. Four different intensities were used for pS/pT intact ions, NL-P ions, and pY intact ions: 10%-100%-100%, 50%-100%-100%, 50%-50%-50%, and 50%-50%-100% (refer to Fig 3 and [39]). Those ions were predicted from nonphosphorylated fragment ions: a, b, y, and their ammonia (A) and water (W) NL ions. Four different NL conditions were used: NL-PAW, NL-PA, NL-PW, and NL-P. Against the 16 libraries, the synthetic phosphopeptides were searched by SpectraST with scoring versions 4 and 5.

Fig 6. Phosphorylation site localization on K562 phosphopeptides by simulated spectral library searching and MaxQuant searching.

From the K562 phosphopeptides analyzed by TripleTOF CID, singly phosphorylated peptides were identified at 1% FDR by SpectraST with the simulated spectral library and by MaxQuant, followed by applying 1% FLR cutoff. Three simulated spectral libraries generated under the default condition (10%-100%-100% intensities, NL-PAW) and the optimized conditions (50%-50%-50% and 50%-50%-100% intensities, NL-P; refer to Fig 5) were used for SpectraST searching with scoring versions 4 and 5.

Fig 7. Sequence and site agreements of simulated spectral library searching and MaxQuant searching on K562 phosphopeptides.

From the singly phosphorylated K562 peptides analyzed by TripleTOF CID, spectral matches shared by SpectraST searching against the simulated spectral library under the optimized condition (50%-50%-100% intensities, NL-P, versions 4 scoring) and MaxQuant searching were extracted at 1% FDR and 1% FLR (see Fig 6). Out of those, sequences and phosphorylation sites agreed by the two searches were counted.