Relative Specificities of Water and Ammonia Losses from Backbone Fragments in Collision-Activated Dissociation Mikhail M. Savitski,*,† Frank Kjeldsen,† Michael L. Nielsen, and Roman A. Zubarev Laboratory for Biological and Medical Mass Spectrometry, Uppsala University, Box 583, Uppsala S-751 23, Sweden Received March 6, 2007
Analysis of a database containing over 20 000 high-resolution collision-activation mass spectra of tryptic peptide dications was employed to study the relative specificity of neutral losses from backbone fragments. The high resolution of the FTMS instrument allowed for the first time the first isotope of the water loss and the monoisotope of the ammonia loss to be distinguished. Contrary to a popular belief, water losses from y′ ions are not specific enough to rely upon for detecting the presence of amino acids with oxygen in the side chains. At the same time, ammonia loss from b ions is sufficiently specific (>95%) to detect the presence of amino acids Gln, Asn, His, Lys, and Arg. This feature will be useful for de novo algorithms for high-resolution MS data. Clear trends were observed when the effect of amino acids proximate to the cleavage site on the rate of loss formation was studied. These trends turned out to be different for losses from b and y ions. Keywords: CAD • de novo sequencing • peptide identification • FTMS
Introduction Tandem mass spectrometry (MS/MS) of tryptic peptides is the basis of most proteomics studies.1-3 By far the most frequently used MS/MS technique is collision-activated dissociation, CAD. The purpose of CAD is to break peptide bonds to form N-terminal b and C-terminal y′ ions. Both de novo sequencing algorithms4-6 and database search engines, such as Mascot,7 preferentially use these backbone fragments for peptide sequence characterization. However, often enough b and y′ fragmentation is not extensive enough for reliable sequence identification. In a recent study, only ∼24% of the CAD MS/MS spectra of tryptic peptides that were identified by Mascot contained full sequence information, with an average sequence coverage of ∼80%.4 Thus, there is a need to involve other fragmentation reactions prominent in CAD, such as neutral losses of water and ammonia from b and y′ fragments. In a large-statistics study performed with a lowresolution ion trap (IT) mass spectrometer,8 ammonia loss (-17 Da) from b ions was found to give on average half the intensity of a median b ion and to appear in 55% of the cases. Ammonia losses from y′ ions were found 25% less abundant than from b ions, but still they appeared in 45% of the expected cases.8 Water losses affected 66% of the b ions and 51% of the y′ ions.8 Overall, water losses were 16% more frequent than ammonia losses, and their total intensity was 70% higher. Neutral losses can originate from both the side chains and the backbone.8,9 Besides, ammonia losses can occur from the * To whom correspondence should be addressed. +46 18 471 72 09. E-mail:
[email protected]. † These authors contributed equally to this work. 10.1021/pr070121z CCC: $37.00
2007 American Chemical Society
Phone/fax:
N-terminus and water losses from the C-terminus. Losses of water and ammonia possess a certain specificity. Ammonia losses from both b and y′ ions were found to be especially prominent if the corresponding ions contain Asn or Gln.8 A similar effect was found for His in y′ ions ending with Lys.8 On the other hand, the presence of His and Pro suppressed NH3 losses from b ions.8 The content of Pro had a similar, but smaller effect on ammonia losses from y′ ions.8 Water losses from b ions were encouraged by the presence of Asn and Gln (but even more so by Thr and Ser) and diminished by Pro and His.8 These features were similar to ammonia losses. For y′ ions containing His, water loss was found to be prominent, especially if the C-terminal amino acid was Lys.8 Thr and Ser were also found to facilitate water losses, but not substantially.8 The above findings of the large-statistics study have been supported by a number of other investigations, mainly on a limited range of molecules. Farrugia et al.10 found ammonia losses to stem from the side chains of Asn and Gln. Ballard and Gaskell studied singly charged peptides and found water losses from the C-terminus, acidic residues, side chains of Ser and Thr, and backbone amides.11 Recent efforts have also shown that it is not unusual to encounter particularly prominent neutral water and ammonia losses and the absence of traditional backbone fragments, b and y′, corresponding to these cleavages.12,13 These and other studies have revealed a wealth of information, but also left some important questions unresolved, since they were performed either with low-resolution instruments or on limited datasets. With a low-resolution mass spectrometer, distinguishing -18 Da water losses and -17 Da ammonia losses is impossible when both are present. This introduces an element of uncertainty in the influence of the amino acid Journal of Proteome Research 2007, 6, 2669-2673
2669
Published on Web 05/16/2007
research articles
Savitski et al.
Figure 1. (a) Mass profiles of y′ obtained by summing the MS/MS spectra of 20 000 peptides in the area of all possible y′ ion masses. The value 0 Da corresponds to the monoisotopic mass of the y′ ions. At -18.01 Da the water loss is located. The inset shows the resolved peaks of the ammonia loss and the first isotope of the water loss. (b) Same for all y′ ions that do not contain amino acids with oxygen in their side chains.
Figure 2. (a) Summed spectra of 15 000 peptides in the area of all possible b ion masses. The inset shows the resolved peaks of the ammonia loss and the first isotope of the water loss. (b) Same for all b ions that do not contain amino acids with oxygen in their side chains. (c) Same for all b ions that do not contain amino acids with nitrogen in their side chains. The inset shows the resolved peaks of the ammonia loss and the first isotope of the water loss.
content on the corresponding losses.8 Limited-dataset analysis allows one to get detailed insight into a particular situation, but does not highlight dominant trends. One of the unresolved important questions is the degree of specificity of neutral losses with respect to the presence of certain amino acids. If the specificity is high enough, these losses can serve as reliable indicators of the presence of these amino acids. If however the degree of specificity is low, the losses contain little information on the amino acid content. The Mascot search engine assumes the degree of specificity of some losses to be quite high. For b ions, the presence of ammonia losses is taken as an indicator of the presence of amino acids with nitrogen in the side chain, and for water losses from y′ ions the same specificity is assumed for the presence of oxygen atoms. However, the basis for such an assumption has not been thoroughly tested yet. Here we perform such a test using a database of 20 000 high-resolution CAD mass spectra of unique tryptic sequences.14-16
Results and Discussion The spectra were obtained on a 7 T Fourier transform (FT) mass spectrometer (LTQ FT, ThermoFisher Scientific, Bremen, Germany), with accuracies of the molecular mass determination of (5 ppm and the fragment mass of (20 mDa.14,17 Figure 1a shows the overall mass profile of y′ ions for all possible cleavage sites of the 20 000 peptides, including the small loss 2670
Journal of Proteome Research • Vol. 6, No. 7, 2007
region. Owing to the high resolution and excellent mass accuracy of FTMS, the contributions of NH3 and H2O losses are nicely separated, allowing one to distinguish between them in 95% of the cases. Since most tryptic peptide ions produce doubly charged species in electrospray ionization (ESI),18,19 whose fragmentation is dominated by singly charged b and y′ ions, we limited our study to these species. Instead of ion intensities, frequencies of ion appearance at a given mass relative to the theoretical mass of the y′ ions were counted. As follows from Figure 1a, applying this method provided realistically looking isotopic distributions. Overall analysis showed that the peaks corresponding to the water losses are present in 14.2% of the cases when the y′ ion is present. This value was obtained by dividing the abundance of the water loss peak in Figure 1a by the abundance of the y′ ion peak. The corresponding figure for ammonia losses is 5.8%. The corresponding values for b ions are 37.5% and 14.8%, respectively. The more frequent losses from b ions are in line with previous experimental studies.8 This effect is also expected on theoretical grounds since b ions are less stable than y′ ions.20 y′ Ions. Water loss is much more frequent from y′ ions than ammonia loss, and thus, it has superior analytical utility. To test the specificity of water losses from y′ species, the plot in Figure 1a drawn for all y′ ions was compared to the plot in Figure 1b made for y′ species comprising no amino acids with oxygen-containing side chains. Despite the absence of oxygen
NH3 and H2O Losses in CAD
Figure 3. CAD mass spectrum of the doubly protonated peptide LAVNMVPFPR.
in the side chains, a significant H2O loss peak is observed (Figure 1b). This loss must originate from either the C-terminus or the backbone of the y′ ion. Here peaks corresponding to the water losses are present in 7.9% of the cases when the y′ ion is present. Thus, the side chains of the amino acids are only responsible for 44% ((14.2 - 7.9)/14.2) of water losses from y′ ions. Consequently, the specificity of such losses for predicting whether a y′ fragment contains an amino acid with an oxygen in its side chain is inadequate. The insufficient specificity is contrary to the assumption made by, e.g., the Mascot search engine and some de novo sequencing programs. The specificity of the ammonia loss from y′ species cannot be tested in the same manner for tryptic peptides. Tryptic peptides almost always terminate with Arg or Lys and thus do not produce y′ ions without amino acids that contain nitrogen in their side chains. To circumvent this problem, we looked at a smaller (ca. 500 entries) set of peptides derived from the Glu-C digest. A total of 10% of all y′ ions had losses of ammonia. In contrast, only 0.2% of y′ ions which contained only amino acids without nitrogen in the side chains had ammonia losses (data not shown). This showed unambiguously that ammonia losses do not occur from y′ ions that only contain amino acids without nitrogen in their side chains. b Ions. Parts a-c of Figure 2 present plots made for all b ions and those without oxygen and nitrogen in their side chains, respectively. The water loss turned out to be unspecific: 13.4% of b ions without oxygen in the side chains contributed to the -H2O peak. On the other hand, only 1% of species without nitrogen gave -NH3 loss, which makes the ammonia loss from b ions much more specific than the -H2O loss. Since the “false positive” rate is below 5%, b-NH3 ions can be used as reliable indicators of the presence of nitrogencontaining amino acids Gln, Asn, His, Lys, and Arg. NH3 Loss from b in de Novo Sequencing. Figure 3 presents the CAD mass spectrum of 2+ ions of a peptide with a molecular mass of 1142.628 Da. Together with the ECD21 mass spectrum (not shown), these data allowed the de novo sequencing algorithm described by Savitski et al.4 to produce the following sequence: LAV[N/GG]MVPFPR. In this sequence, the unclear moment is the presence of N vs GG combination. Such a dilemma is frequent in a de novo sequencing task.4,6,22 It arises each time asparagine is met, since this amino acid is isomeric to two glycines. The problem is also aggravated by the fact that cleavage between two glycines is not favorable in CAD.15,23 Another type of confusion is Q vs G + A. Together, Asn and
research articles Gln account for 8.6% of the total amino acid abundance, meaning that an average tryptic peptide 10-12 residues long contains one such amino acid (N or Q). The solution to this problem is to examine the presence of the ammonia loss from the b ions. In Figure 3, NH3 loss is observed from the peak at m/z 529.280, which corresponds to b5 in the case of N and b6 in the case of GG. The presence of an -NH3 satellite to this peak speaks in favor of the first alternative. Indeed, the Mascot search identifies the spectrum in Figure 3 as the peptide LAVNMVPFPR from the protein tubulin 5-β. Together with the ECD spectrum, the score of this peptide is 87, above the threshold of 34. No reasonable alternative with GG instead of N is found in the database. Effect of Amino Acids Proximate to the Cleavage Site on the Rate of H2O and NH3 Losses. An important question is whether the proximity of a given type of an amino acid residue to the cleavage site has an influence on the rate of the losses. If this is the case, then the loss is more likely to be associated with the cleavage as opposed to the possibility of the loss occurring from the intact molecule prior to cleavage. The argument is that if the loss occurs prior to backbone fragmentation, then the subsequent fragmentation of the peptide species with the loss should give the same pattern as the fragmentation of the intact peptide. Consequently, the proximity of a specific type of an amino acid residue to the cleavage site should not have an influence on the rate of the losses. In Figure 4 the effect of proximal amino acids is shown for the 16 residues commonly encountered in the internal regions of tryptic peptides. AA denotes the amino acid of interest, XX denotes an arbitrary amino acid, and - denotes the cleavage site location. Here the statistics were accumulated in the same fashion as in ref 15. The analysis focused on the preferred fragmentation region of doubly charged tryptic peptides yn-3′-H2O/NH3, yn-4′H2O/NH3, yn-5′-H2O/NH3, b3-H2O/NH3, b4-H2O/NH3, and b5H2O/NH3, where n is the length of a given peptide.15 Briefly, the abundances of yn-3′-H2O/NH3, yn-4′-H2O/NH3, and yn-5′H2O/NH3 were normalized for each spectrum with respect to the most abundant of these three ions; the same was done for each b3-H2O/NH3, b4-H2O/NH3, and b5-H2O/NH3 triplet. The average abundances from all 20 000 spectra were calculated as a function of an amino acid in a given position relative to the cleavage site. It was observed (Figure 5) that the thus calculated average abundances correlated strongly with the average abundances of the y′ and b ions. Thus, the not normalized average abundance of y′-H2O/NH3 or b-H2O/NH3 is a poor indicator of whether the loss is preferential. Consequently, one has to take into account the average abundances of y′ and b ions. This was done as shown in Figure 5. A stringent criterion was used to determine which points were outliers (amino acids having a strong effect on losses). Outliers were considered to be the minimal number of points required to be removed to attain a >0.90 linear correlation with the remaining points. A straight line was fitted through these remaining points, and the distance of all the points from this line was used as a measure of the deviation in relative abundance. Deviation upward was recorded with a plus sign, while deviation downward was recorded with a minus sign. Recently, it has been reported that peptides with N-terminal Q show prominent losses of water and ammonia, especially in the presence of mobile protons.13 From Figure 4a,b it is clear that the same is true for y′ ions with N-terminal Q. Journal of Proteome Research • Vol. 6, No. 7, 2007 2671
research articles
Savitski et al.
Figure 4. Effect of 16 amino acid residues on the frequency of (a) water and (b) ammonia loss from y′ ions as well as of (c) water and (d) ammonia loss from b ions as a function of the residue position relative to the cleavage site. The value zero indicates that an average loss frequency is observed. The negative and positive values indicate decreased and enhanced frequency. The loss abundances in (a), (b) and (c), (d) are normalized to the y′ and b ion abundances, respectively.
The fact that a negative effect on ammonia loss is observed when Q and N are located N-terminal to the cleavage site (on the b ion) can be explained as a competition between losses from b and y′. If an ammonia loss occurs from the b ion, it is less likely to occur from the y′ ion. For the link between losses from the complementary b and y′ ions to act, at least some steps of the loss process must occur prior to the bond cleavage.
Figure 5. Relation between the relative abundances of y′ ions and y′-H2O ions. The removal of the two nonfilled points yields a linear correlation of >0.90 between the remaining points.
The presence of E, C-terminal to the cleavage site, strongly promotes water loss from the y′ ion. The result for E is expected due to the presence of a hydroxyl group in its side chain. As expected Q and N strongly promote the loss of ammonia, when they are located C-terminal to the cleavage site (Figure 4b). 2672
Journal of Proteome Research • Vol. 6, No. 7, 2007
Serine and threonine have the strongest effect on the water losses from b ions depending on their positions. The effect is either positive or negative. The fact that these residues play an important role is logical due to the presence of the hydroxyl group in their side chains. The reason why water losses are most pronounced when S or T is located at the next to last position in the b ions could be due to the formation of five- or six-membered rings between the side chains of serine and threonine and the backbone carbonyl oxygen of the b ion similar to the dehydration reaction described by Paizs et al.,9 Scheme 1. The involvement of the charge in Scheme 1 also explains why a similar trend is not observed for y′ ions, because there the charge is immobilized partially or entirely by the C-terminal lysine or arginine. N, Q, and H have the strongest effect on the ammonia losses from b ions. The fact that they have the strongest effect when
research articles
NH3 and H2O Losses in CAD Scheme 1
or Thr located in the next to last positions have pronounced water losses. The suggested explanation is the formation of fiveor six-membered rings through the process of cyclization.
Acknowledgment. This work was supported by the Knut and Alice Wallenberg Foundation and Wallenberg Consortium North (Grant WCN2003-UU/SLU-009 to R.A.Z. and an instrumental grant to R.A.Z. and Carol Nilsson) as well as the Swedish Research Council (Grants 621-2004-4897, 621-2002-5025, and 621-2003-4877 to R.A.Z.). Christopher Adams is acknowledged for insightful discussions. References (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
they are part of the y′ ion deserves more detailed investigation in the future.
Conclusions Study on a large database of high-quality mass spectra revealed that, contrary to popular belief, water losses from y′ ions are not specific enough to rely upon for detecting the presence of amino acids with oxygen in the side chains. At the same time, ammonia loss from b ions is sufficiently specific (>95%) to detect the presence of amino acids Gln, Asn, His, Lys, and Arg. The same is true for y′ ions from peptides generated using the Glu-C digest. y′ ions with N-terminal Gln exhibit very strong water and ammonia losses. b ions with Ser
(17) (18) (19) (20) (21) (22) (23)
Steen, H.; Mann, M. Nat. Rev. Mol. Cell Biol. 2004, 5, 699-711. Aebersold, R.; Mann, M. Nature 2003, 422, 198-207. Mann, M.; Jensen, O. N. Nat. Biotechnol. 2003, 21, 255-261. Savitski, M. M.; Nielsen, M. L.; Kjeldsen, F.; Zubarev, R. A. J. Proteome Res. 2005, 4, 2348-2354. Frank, A.; Savitski, M. M.; Nielsen, M. L.; Zubarev, R. A.; Pevzner, P. J. Proteome Res., in press. Frank, A.; Pevzner, P. Anal. Chem. 2005, 77, 964-973. Perkins, D. N.; Pappin, D. J. C.; Creasy, D. M.; Cottrell, J. S. Electrophoresis 1999, 20, 3551-3567. Tabb, D. L.; Smith, L. L.; Breci, L. A.; Wysocki, V. H.; Lin, D.; Yates, J. R. Anal. Chem. 2003, 75, 1155-1163. Paizs, B.; Suhai, S. Mass Spectrom. Rev. 2005, 24, 508-548. Farrugia, J. M.; O’Hair, R. A. J.; Reid, G. E. Int. J. Mass Spectrom. 2001, 210, 71-87. Ballard, K. D.; Gaskell, S. J. J. Am. Soc. Mass Spectrom. 1993, 4, 477-481. Mouls, L.; Aubagnac, J. L.; Martinez, J.; Enjalbal, C. J. Proteome Res., in press. Neta, P.; Pu, Q. L.; Kilpatrick, L.; Yang, X. Y.; Stein, S. E. J. Am. Soc. Mass Spectrom. 2007, 18, 27-36. Nielsen, M. L.; Savitski, M. M.; Zubarev, R. A. Mol. Cell. Proteomics 2005, 4, 835-845. Savitski, M. M.; Kjeldsen, F.; Nielsen, M. L.; Zubarev, R. A. Angew. Chem., Int. Ed. 2006, 45, 5301-5303. Savitski, M. M.; Nielsen, M. L.; Zubarev, R. A. Mol. Cell. Proteomics 2005, 4, 1180-1188. Savitski, M. M.; Nielsen, M. L.; Zubarev, R. A. Mol. Cell. Proteomics 2006, 5, 935-948. Fenn, J. B.; Mann, M.; Meng, C. K.; Wong, S. F.; Whitehouse, C. M. Science 1989, 246, 64-71. Wilm, M.; Mann, M. Anal. Chem. 1996, 68, 1-8. Aaserud, D. J.; Little, D. P.; Oconnor, P. B.; McLafferty, F. W. Rapid Commun. Mass Spectrom. 1995, 9, 871-876. Zubarev, R. A.; Kelleher, N. L.; McLafferty, F. W. J. Am. Chem. Soc. 1998, 120, 3265-3266. Frank, A.; Tanner, S.; Bafna, V.; Pevzner, P. J. Proteome Res. 2005, 4, 1287-1295. Huang, Y. Y.; Triscari, J. M.; Pasa-Tolic, L.; Anderson, G. A.; Lipton, M. S.; Smith, R. D.; Wysocki, V. H. J. Am. Chem. Soc. 2004, 126, 3034-3035.
PR070121Z
Journal of Proteome Research • Vol. 6, No. 7, 2007 2673
