Evaluation of the utility of neutral-loss-dependent MS3 strategies in large-scale phosphorylation analysis

Abstract

Phosphopeptide identification and site determination are major challenges in biomedical MS. Both are affected by frequent and often overwhelming losses of phosphoric acid in ion trap CID fragmentation spectra. These losses are thought to translate into reduced intensities of sequence informative ions and a general decline in the quality of MS/MS spectra. To address this issue, several methods have been proposed, which rely on extended fragmentation schemes including collecting MS3 scans from neutral loss-containing ions and multi-stage activation to further fragment these same ions. Here, we have evaluated the utility of these methods in the context of a large-scale phosphopeptide analysis strategy with current instrumentation capable of accurate precursor mass determination. Remarkably, we found that MS3-based schemes did not increase the overall number of confidently identified peptides and had only limited value in site localization. We conclude that the collection of MS3 or pseudo-MS3 scans in large-scale proteomics studies is not worthwhile when high-mass accuracy instrumentation is used.

Figure 1

(A) Neutral loss fragment ion frequency in ion trap CID MS2 spectra from identified phosphopeptides. Proteolyzed yeast protein (10 mg) was separated by SCX chromatography. Eleven collected fractions were subjected to phosphopeptide enrichment using IMAC. Each fraction was analyzed by LC-MS/MS using the TOP10 method and spectra were assigned using SEQUEST. From 19 087 identified phosphopeptides, the intensity rank (Nth) of the peak corresponding to neutral loss of phosphoric acid (−49 for 2+, −32.7 for 3+ and −24.5 for 4+) was plotted in relation to the charge of the precursor ion. Neutral loss was a very common event in IT spectra. The lability of the phosphate group was charge dependent, being more stable at higher charge states. (B) Scheme for the scan cycles in the TOP10 MS2 method: on each cycle, one MS scan was followed by ten MS2 scans. (C) Scheme for the scan cycles in the data-dependent neutral loss MS3 (DDNLMS3) method: one MS scan was followed by ten MS2 scans. EachMS2 was interrogated by the presence of a neutral loss peak and by its intensity rank in the spectrum. If both conditions were satisfied, an MS3 scan was triggered. (D) Scheme for the scan cycles in the pseudo MS3 method: one MS scan was followed by ten MS2 scans. Each MS2 was collected only after additional activation of fragment ions at −49, −32.7 and −24.5 mass units from the parent ion, corresponding to neutral losses of phosphoric acid. (E–G) Examples of spectra corresponding to the same doubly charged ion for the phosphopeptide, VIS*QDALQHFR, from GTPase Nog2, (E) an MS2 spectrum, (F) an MS3 spectrum, (G) a pseudo MS3 spectrum. Note different y-axis scale for a better representation of fragment ion populations.

Figure 2

Scan collection and scan cycle schemes for MS2 and MS3 methods. (A) Number of MSn scans collected and (B) number of scan cycles for the TOP10 MS2 (MS2), DDNLMS3 (MS3), and pseudo MS3 (MS3′) methods. Values represent the mean ± SD of triplicate analyses of a single SCX fraction (fraction #5) for each method. MS2 or pseudo MS3: dark, MS3: pale grey. (C) Average cycle times for the data acquisition methods used in this study. All scan times were calculated directly from the acquired data for each method. Ion accumulation times and scan times for full MS scan, as well as total scan cycle times were calculated from cycles were ten dependent MS2 scans, ten MS2 and three MS3 scans or ten pseudo MS3 were collected, which corresponded to the median numbers of scans per cycle for each method, respectively. Ion accumulation periods are shown in black, while analysis periods are shown in grey scale for the different scan types.

Figure 3

Comparison of MS2 and MS3 spectra. (A) The percentage of all the sequence b- and y-type ions that were matched in the spectra is plotted for those MS2–MS3 pairs that produced the same peptide sequence (n = 2439). The mean fraction of total ions matched was 0.42 for MS2 spectra and 0.38 for MS3 spectra (P = 3 · 10⁻²⁸, paired T-test). (B) TIC for all the MS2–MS3 pairs (n = 5117) plotted on a log10 scale. The data points can be fitted to a line with a slope of 1 and an offset 0.8, meaning that only 15% of the total signal from the MS2 spectra is channeled into the MS3 spectra. (C–D) Comparison of XCorr scores between MS2 and MS3 spectra pairs assigned to the same sequence. (C) Cumulative frequency of the difference of XCorr values [XCorr(MS3)-XCorr(MS2)], grouped by charge (z). Xcorr values were clearly superior for MS2 spectra of peptide ions with charge states of 3+ and 4+ and similar for 2+. (D) Scatter plot of XCorr values for peptide ions with charge 2+ (n = 1171). Most of these peptides (94%) were singly phosphorylated.

Figure 4

Peptide spectral matches of MS3 and their preceding MS2 scans for a triplicate analysis a complex phosphopeptide mixture. All three runs used the DDNLMS3 method. As both spectra in each pairwere answers to the same precursor ion species, theywere treated as a unit. (A) All MS2–MS3 pairs (n = 5117) were considered. The first filtering criterion was both spectra (MS2 and MS3) yielding the same peptide sequence (n = 2480, 48%). In this pool only 9 decoy hits were identified, thus the FDR was <1%. The remaining pairs were classified based on passing filtering criteria to establish a 1% FDR only considering MS2 spectra (n = 651, 13%), only MS3 spectra (n = 150, 3%), passing neither (n = 1832, 36%) or passing both with different sequences (at least one is a false positive, n = 4, 0.1%). (B) Only MS2–MS3 pairs that identified the same sequence and passed the mass accuracy filterwere considered (n = 2439, no reverse hits). This dataset should be entirely composed of correct sequence matches. To these spectra, we applied the Xcorr and dCn’ score values according to <1% FDR combined with a mass accuracy filter (“MMA”) or not (“no MMA”). Not using mass accuracy information simulates low mass accuracy phosphopeptide identification. See Section 2 for scoring and mass accuracy filtering criteria. Peptides were classified into four categories: both MS2 and MS3 passing the filtering criteria, only MS2, only MS3 and none passing. Although obtaining the same sequence for both spectra is an excellent filter by itself, in the presence of mass accuracy information, 95% of these MS2 spectra still passed filtering criteria for a 1% FDR.

Figure 5

Phosphopeptide identifications per analysis at <1% FDR using each of the three methods. Mean ± SD values for triplicate analyses are shown. (MS2 = TOP10 MS2 method, MS3 = DDNLMS3 method, MS3′ = pseudo MS3 method, MMA: searches at 50 ppm precursor mass tolerance, using mass accuracy information as a filtering criteria, no MMA: searches at 2.0 Da precursor mass tolerance, not using mass tolerance as a filter). (A) Total (redundant) phosphopeptide spectral matches. (B) Unique (non-redundant) phosphopeptide sequences. In order to be conservative, unique phosphopeptides were calculated at sequence level without considering different phosphorylation site localizations. (C) Success rate in phosphopeptide and unmodified peptide identification for each method. Values are calculated based on the total number of MSn scans collected. Details for different scan types are given in the MS3 method. (D) Overlap in phosphopeptide identification between the three methods. For simplicity, overlaps were calculated at sequence level without considering different phosphorylation site localizations. Each circle represents the combined triplicate analyses for each of the methods. A total of 2041 non-redundant phosphopeptides were identified in the SCX fraction #5 used in this study.

Figure 6

Site localization comparison for MS2–MS3 pairs. (A) Scatter plot of each MS2–MS3 same-sequencematching pair. Rarely does an MS2 spectrum receive a score <19, which is rescued by a score >19 in the MS3 spectrum (pale grey bottom-right box; 9%). (B) Distribution of binned Ascore values for MS2 and MS3 spectra. (C) Cumulative distribution of unique phosphopeptides matches passing as a function of Ascore values. On their own, MS2 spectra produce relatively higher Ascore values than MS3.