Identification of ultramodified proteins using top-down tandem mass spectra

Abstract

Post-translational modifications (PTMs) play an important role in various biological processes through changing protein structure and function. Some ultramodified proteins (like histones) have multiple PTMs forming PTM patterns that define the functionality of a protein. While bottom-up mass spectrometry (MS) has been successful in identifying individual PTMs within short peptides, it is unable to identify PTM patterns spreading along entire proteins in a coordinated fashion. In contrast, top-down MS analyzes intact proteins and reveals PTM patterns along the entire proteins. However, while recent advances in instrumentation have made top-down MS accessible to many laboratories, most computational tools for top-down MS focus on proteins with few PTMs and are unable to identify complex PTM patterns. We propose a new algorithm, MS-Align-E, that identifies both expected and unexpected PTMs in ultramodified proteins. We demonstrate that MS-Align-E identifies many proteoforms of histone H4 and benchmark it against the currently accepted software tools.

Figure 1. Bottom-up MS lacks the ability to identify complex PTM patterns

(a) Two proteoforms of a protein with phosphorylation sites coexist in the sample (P represents phosphorylation). (b) Bottom-up MS identifies four peptides (shaded regions in the proteoforms) resulting in up to 4 putative proteoforms. However, it is unable to answer the question which of these putative proteoforms are present in the sample.

Figure 2. Spectral alignment

(a) A spectral alignment between the theoretical spectrum B = {0, 57, 144, 245, 302, 458, 559, 687} of a protein GSTGRTK and the theoretical spectrum B* = {0, 57, 304, 375, 432, 588, 659, 787} of a modified protein GS[+160]T[-30]GRT[-30]K. The path from the top left corner (source) to the bottom right corner (sink) represents the alignment of B and B* with three PTMs: +160 Da on the first S and −30 Da on the two T's. Diagonal and shift edges are shown in blue and red, respectively. The circles along the path denote the matching points in the alignment path. (b) A spectral alignment between a spectrum A = {0, 57, 375, 482, 588, 659, 787} generated from GS[+160]T[-30]GRT[-30]K and the theoretical spectrum B. Because mass 304 is missing in A, the PTM on the first S and the PTM on the first T are represented by a single shift edge (+130 Da) with a modification number 2. Another missing mass 432 in A results in replacing two consecutive diagonal edges by one diagonal edge. In addition, mass 482 is a noise mass. (c) A diagonal alignment between the spectrum A and the theoretical spectrum B (for a set of mass shifts S₁ = {−30, 160} and F = 3). The diagonal grid of A and B has 10 diagonal lines with offsets -90, -60, -30, 0, 100, 130, 160, 290, 320, and 480. The path from the source to the sink represents a diagonal alignment of spectrum A and protein B. The circles along the path denote diagonal points: blue ones have weight 1 and red ones have weight 0.

Figure 3. The histograms of numbers of expected PTM sites in 629 spectra identified from the histone H4 data set by MS-Align-E

(a) 457 spectra without unexpected PTM sites. (b) 172 spectra with a single unexpected PTM site.

Figure 4. Comparison of MS-TopDown and MS-Align-E

MS-TopDown and MS-Align-E were applied to align 1, 626 ETD spectra with the histone H4 protein with one unexpected PTM site, and the numbers of identified PrSMs and matched fragment ions are reported.

Figure 5. Comparison of the numbers of matched fragment ions of the PrSMs identified in the histone H4 data set by MS-Align-E and ProSightPC

(a) The “absolute mass” mode. A total of 1024 spectra were identified by both MS-Align-E and ProSightPC. For each of the 1024 spectra, the numbers M and P of matched fragment ions of the PrSMs identified by MS-Align-E and ProSightPC are reported and the difference M −P is computed. Of the 1024 spectra, MS-Align-E reported more matched fragment ions than ProSightPC for 312 spectra; ProSightPC reported a better proteoform than MS-Align-E for 135 spectra; and the two tools reported the same number of matched fragment ions for 577 spectra. (b) The “absolute mass” mode combined with the Δm mode. A total of 1011 spectra were identified by both MS-Align-E and ProSightPC. Of the 1011 spectra, MS-Align-E reported more matched fragment ions than ProSightPC for 60 spectra; ProSightPC reported more matched fragment ions than MS-Align-E for 740 spectra.

Figure 6. MS-Align-E and the “absolute mass” mode of ProSightPC reported two different proteoforms for the spectrum of scan number 2, 062 in histone H4 spectral data set

(a) The proteoform reported by MS-Align-E has 47 matched fragment ions. (b) The proteoform reported by ProSightPC has 30 matched fragment ions. The ‘]’ symbol right to the first methionine residue represents N-terminal methionine excision. Residues with PTMs are shown in red. AC, ME, 2M and 3M stand for acetylation, methylation, dimethylation, and trimethylation, respectively. Red lines represent matched fragment ions identified by only one tool; black lines represent matched fragment ions identified by two tools.