Introduction

Tandem mass spectrometry is the key technology utilized in peptide sequencing and proteomics [1–3]. Typically, individual peptides are isolated prior to collisional activation and identification of sequence is subsequently achieved based on the detected mass-to-charge (m/z) ratios corresponding to the charged fragments and the precursor protonated peptide. Ideally, collision-induced dissociation (CID) initiates cleavage of the amide bonds to produce series of b n ions if the N-terminal fragment keeps the charge, y m ions if the C-terminal fragment keeps the charge, or a mixture of the two (for a peptide of length N = n + m) [4–6]. The relative abundance of the product ions depends on the peptide sequence, charge state, instrument type, and the specific conditions under which the experiments were performed [3, 5, 7–13]. Certain amino acid residues strongly enhance specific types of bond cleavage [10–12, 14–16]. Among this suite of residue-specific chemistries is that associated with proline, P (Scheme 1a). In the well-documented “proline effect” [17–24], a strong preference for amide bond cleavage N-terminal to proline residues is observed. This chemistry results in enhanced prevalence of the y m ion peak with the proline residue situated at its N-terminus with concomitant suppression of the complementary b n ion (Scheme 2). Proline is unique among the 20 commonly occurring amino acids in that it contains a secondary amine, in a 5-membered ring at its N-terminus (Scheme 1a). Consequently, once involved in an amide bond, it becomes a tertiary amine. Pipecolic acid, Pip, is the 6-membered ring congener of proline (Scheme 1b) and has its own residue-specific chemistry, the “pipecolic acid effect” [20]. Despite obvious structural similarity between proline and pipecolic acid, the two effects (residues) result in vastly different mass spectra of otherwise identical protonated peptide sequences [18, 20].

Scheme 1
scheme 1

Schematic of (a) proline, (b) pipecolic acid

Full size image
Scheme 2
scheme 2

Generic b2-y(N-2) mechanism of proline-containing peptides to form y(N-2) product ions

Full size image

Pipecolic acid has been shown to promote the dominant formation [18, 20] of specific b n peaks rather than y m peaks (Scheme 3). Furthermore, the amide bond that is cleaved differs in the pipecolic acid effect; the amide bond C-terminal to the pipecolic acid residue is broken. Raulfs et al. [20] demonstrated this phenomenon in their recent paper by utilizing combined experimental and theoretical comparisons of singly protonated pentapeptides, [AAXAA + H]+, where the identity of X was systematically varied [X = P, Pip, N-methylalanine, azetidine-2-carboxylic acid (the 4-membered ring analogue); A = alanine]. Parallel to earlier work [18], they showed that proline and azetidine-2-carboxylic acid behave similarly and give rise to dominant y 3 peaks [20], whereas pipecolic acid and N-methylalanine promote the dominant formation of b 3 peaks. Swapping the alanine residues for other aliphatic residues had minimal effect on this result [18, 20]. Furthermore, when peptides containing both the Pip and P residues were analyzed, the dominant products observed were always the b n peaks corresponding to the position of the Pip residue (i.e., for [APipAPA + H]+, the b 2 peak, for [AAPipPA + H]+ the b 3 peak, and for [APAPipA + H]+ the b 4 peak).

Scheme 3
scheme 3

Generic fragmentation mechanism of pipecolic-acid-containing protonated peptides: (a) formation of b 3 product ions, (b) proton transfer of the alanyl Cα H+ to form y(N-3) product ions (within the proton-bound dimer), (c) alternate proton transfer reaction from the oxazolone ring pipecolic acid Cα H+ to form y(N-3) product ions

Full size image

Additionally, Raulfs et al. [20] provided computational estimates of the proton affinities of the residues, which showed that Pip had a very similar proton affinity to P. Previously, the same corresponding author had analogous experimental findings [25]. Extrapolating on this basis [17, 20], the results could not be rationalized based solely on the basicity of the prolyl nitrogen as had previously been argued [18]. Consequently, Raulfs et al. [20] attributed the result to greater flexibility of the Pip 6-membered ring “which allows for peptides conformations that promote favorable transfer of the mobile proton to the amide C-terminal to the Pip/N-methylalanine residue.” This flexibility-based explanation essentially argues for a proton-transfer-limited reactivity, whereby the ability to protonate each specific amide nitrogen can be used as a surrogate for the barrier to the subsequent sequence ion formation. This is broadly consistent with the mobile proton model [15, 26, 27], and has been specifically argued for by Haeffner et al. [28]. Consequently, the specific question of what the transition structures look like and the product ion energies were not addressed explicitly.

These interesting findings [20] motivated the present computational investigation into the mechanisms of fragmentation of these related systems. Here, we provide a systematic computational study of the fragmentation chemistry of [AAXAo + H]+ peptides; where X = P/Pip and o = 0, 1, 2, 3 to test the flexibility hypothesis. We examine the relative energies of the critical transition structures, product ions, and neutrals as a function of peptide length to help explain the chemistry in play. Predictions based on the progression in leaving group size, composition, and proton affinities are also provided [24, 29–31].

Theoretical Methods

Density functional calculations of minima, and product ions and neutrals were performed with the Gaussian 09 suite of programs [32] at the M06-2X/6-31+G(d,p) level of theory [33, 34]. Multiple conformers of each protonation site were examined for each system. Our typical protonation site labeling convention is illustrated in Supplementary Scheme S1. Multiple transition structures (TSs) were calculated. Minima were confirmed by vibrational analysis (all real frequencies) and TSs were also examined in this manner (one imaginary frequency). The reaction pathway through the TSs was determined by intrinsic reaction coordinate (IRC) calculations with up to 18 steps in each direction. The terminating points of these calculations (one on product side, one on reactant side) were then optimized further to determine the exact minima connected by each specific reaction path. Estimates of the proton affinities of the leaving groups were determined as the difference between the zero-point energy-corrected M06-2X/6-31+G(d,p) total electronic energies (0 K) of the protonated and neutral forms of the potential b n and y m ions, respectively.

In response to a reviewer’s request, we added calculations on the [AA(D-Proline)AA + H]+ system to the analysis. In response to a second reviewer’s request, we provided additional calculations at the M06-2X/6-311++G(2d,p), B3LYP/6-31+G(d,p), and B3LYP/6-311++G(2d,p) levels of theory for the [AAXAA+H]+ systems.

Results and Discussion

Protonation Energetics

We calculated the potential energy surface of [AAXAo + H]+ peptides, where X = P/Pip and o = 0, 1, 2, 3. Our data are normalized to the lowest energy, all trans amide bond structure, labeled as the global minimum, GM, in each case. We note that in analogy with recent work from the Clemmer [35–37] and Paizs groups [24] that cis conformations are also potentially competitive (Supplementary Tables S1–S8). Our calculations predict the protonation site of the global minimum will vary as a function of peptide length, but that this is consistent between the P/Pip forms (Supplementary Tables S1–S14). For example, based on the M06-2X calculations, the global minimum for both [AAXAA + H]+ ions is protonated on the third oxygen, O3 (Supplementary Tables S1 and S2, Supplementary Figure S1). In some cases, a cis conformer is more energetically favorable than the corresponding trans global minimum. Again, this is consistent, irrespective of whether P/Pip is present. In agreement with the literature, protonation of an amide nitrogen requires additional energy [5, 8, 13, 21, 24, 28, 38–45]. A typical prerequisite to amide bond cleavage is protonation of the specific amide nitrogen that is to be cleaved, as this removes conjugation from the bond, reducing the bond order, and makes it easier to break [39, 41]. Our data do show that population of the third amide nitrogen site (N3, which is C-terminal to Pip/P) requires less energy (by >11 kJ mol–1) for pipecolic-acid-containing peptides than it does for the analogous proline forms. This supports the hypothesis of increased flexibility of pipecolic acid enabling population of these sites [20]. For example, for [AAPipAA + H]+ the reactive precursor structure protonated at N3 (71.9 kJ mol–1, Supplementary Table S2) can be readily populated via proton transfer from a proximal O3 protonated structure through a TS requiring only 89 kJ mol–1. Prior theory [45] and experiment [46] indicate that population of these sites will have substantially greater rate constants than those of subsequent higher energy amide bond cleavage barriers.

The evidence for the flexibility-based explanation being the sole cause of the difference in chemistry is contradicted by the additional finding that population of the second amide nitrogen site, N2, at P/Pip, is predicted to require much less energy than N3. If the amide bond cleavage TSs are rate-limiting, and follow directly from the amide nitrogen protonation energetics [28], then the b 2 -y N-2 pathway should be universally favored over the b 3 -y N-3 reaction. Consequently, there would be little difference in the spectra of protonated peptides containing P/Pip. The preponderance of experimental data [17, 20] and our calculations indicate this is not the case. Consequently, we need additional information on the relative energies of the competing TSs to determine if they are the cause of the differing reactivities.

Amide Bond Cleavage TSs

The mechanism of enhanced y m ion formation via the proline effect is illustrated in Scheme 2 [24]. The prevailing mechanism of fragmentation of pipecolic-acid-containing protonated peptides differs (Scheme 3a). While pipecolic-acid-containing protonated peptides typically generate b n ions, formation of the complementary y m ions is, in principle, also possible. This reaction requires abstraction of a non-mobile proton from the fixed charge N-terminal fragment. There are two likely candidate sites for this abstraction to occur from (Scheme 3b and c): the Cα proton of the preceding alanine residue or the fixed-charge oxazolone ring [47]. As neither proton is mobile, these are also potentially rate-limiting transition structures that might explain the lack of an y N-3 ion peak (Scheme 3b). First, however, we need to discuss the relative energetics of the amide bond cleavage reactions of [AAXAo + H]+ peptides as without this step occurring, discussion of proton-bound dimer chemistry is moot.

In agreement with the prior experimental [18, 20, 21, 24] and theoretical [20, 21, 24] findings, and the proline effect in general, our calculations predict the b 2 -y m amide bond cleavage TS to be consistently less energetically demanding for [AAPAo + H]+ peptides than the b 3 -y N-3 amide bond cleavage (Table 1, Supplementary Table S1). The degree to which this is the case varies, but is consistently >20 kJ mol–1. The general mechanism is described in Scheme 2 and an example TS is shown in Figure 1 for [AAPAA + H]+. Following this relatively facile amide bond cleavage, we would expect the high proton affinity [25] of the proline-terminated fragment (932, 957, 977, 989 kJ mol–1 for P, PA, PAA, PAAA, respectively, Table 2) relative to the neutral alanylalanine-oxazolone structure (912 kJ mol–1, see also [48]) to result in proton transfer from the oxazolone ring nitrogen to form the y m ion prior to complex separation. This process becomes increasingly likely as the size and, thus, proton affinity of the PAo fragment formed increases. If the b 3 -y N-3 amide bond cleavage reaction occurs at all (Supplementary Scheme S2, Figure 2a), the product distribution will again be affected by the proton affinities of the neutral forms of the fragments. However, for this to be relevant, an energetically feasible means of proton abstraction must be available to the C-terminal fragment (Supplementary Scheme S2, Supplementary Table S1, and Supplementary Figure S3) [47]. Our calculations indicate that the two potential proton abstraction reactions are equi-energetic with the amide bond cleavage barrier so they should not limit the reaction significantly. Poutsma and coworkers’ experiments on [AAPAA + H]+ show [20] a tiny b 3 peak (~1% relative abundance) consistent with the proton affinity of the b 3 neutral (≥966 kJ mol–1 Table 2) being substantially greater than AA. The product distribution should be much more even for the b 3 -y N-3 reaction of [AAPAAA + H]+ as AAA has a much more similar proton affinity (962 kJ mol–1, Table 2).

Table 1 Transition Structure Relative Energies (ΔE el+ZPE,0K (ΔG 298K)/kJ mol–1) of [AAXAo + H]+, where X = P/Pip and o = number of alanine residues for the b 2 –y (N-2) and b 3 –y (N-3) pathways
Full size table
Figure 1
figure 1

The b2-y3 transition structures at the M06-2X/6-31+G(d,p) level of theory; (a) [AAPAA+H]+, (b) [AAPipAA + H]+. Inset: standard projections illustrating the stereochemical consequences of protonation either below the ring as is favored in the P-containing analyte b2-y3 TSs: (a) R stereochemistry at the prolyl amide nitrogen, or above the ring as is favored in the Pip-containing analyte b2-y3 TSs: (b) S stereochemistry at the pipecolic acidyl amide nitrogen. Numbers indicate the priority of substituents in assigning configuration

Full size image
Table 2 Calculated Gas-Phase Proton Affinities of the Various Neutrals Present in Post Amide Bond Cleavage Proton-Bound Dimer
Full size table
Figure 2
figure 2

(a) b 3 -y 2 TS of [AAPAA + H]+ and (b) b 3 -y 2 TS of [AAPipAA + H]+ at M06-2X/6-31+G (d,p) level of theory. Critical bond lengths and dihedral angles are provided for contrast

Full size image

Our calculations are also in general agreement with the experimental data underlying the pipecolic acid effect. The M06-2X/6-31+G(d,p) transition structure calculations predict that the [AAPipAo + H]+ (o = 1–3) peptides should generally favor the b 3 -y N-3 amide bond cleavage over the b 2 -y m amide bond cleavage reaction (Table 1, Supplementary Table S2). For example, the [AAPipAA + H]+ system investigated by Poutsma and coworkers experimentally [20] produced b 3 and y 3 peaks in an approximately 4:1 ratio. The energy of the b 3 -y 2 amide bond cleavage TS [ΔEel+ZPE,0K (ΔG 298K) = 101.9 (104.8) kJ mol–1, Figure 2b) is lower than, but similar to, the b 2 -y 3 TS energy (107.8 (109.1) kJ mol–1, Figure 1b). The most facile subsequent proton abstraction reaction from the oxazolone ring (Scheme 3c, Table 2) is less energetically demanding (96 kJ mol–1) than the amide bond cleavage. However, the neutral b 3 oxazolone generated from this TS has substantially greater proton affinity (1052 kJ mol–1) than AA (923 kJ mol–1) which, consistent with the lack of any y 2 peak, makes this transformation very unlikely. The [AAPip + H]+ analyte is an exception in that it is predicted to yield a y 1 peak rather than a b3. Here, the b 2 -y 1 TS is > 25 kJ mol–1 lower than the b 3 -H 2 O TS. This is due to water being a comparatively poor leaving group [49], with limited hydrogen-bonding capability to stabilize the TS relative to alanine or polyalanine. Consequently, a y 1 ion, [Pip + H]+, is predicted to be formed and subsequently detected, consistent with pipecolic acid having greater proton affinity than neutral AA oxazolone (Table 2).

Proline versus Pipecolic Acid-Transition Structure Stereochemistry Differences

The lowest energy amide bond cleavage barriers for the pipecolic-acid-containing protonated peptides are consistently lower than their proline-containing congeners. This broadly agrees with Poutsma and co-workers’ [20] data on direct competition between P/Pip enhanced fragmentations in a single protonated peptide. These authors found that the b n peak with C-terminal Pip was always the base peak (i.e., the “pipecolic acid effect” appeared to be stronger than the “proline effect” under their experimental conditions). Broadly, the Pip residue’s increased size and flexibility enables superior stabilization of the key transition structures and intermediates necessary to form the requisite products. How does this happen?

Although the lowest energy b 3 -y N-3 TSs are quite similar, differences do exist. We utilize the [AAXAA + H]+ systems to illustrate this. Our density functional calculations predict two major areas of difference: (1) a longer critical amide bond length for the prolyl form (2.02 Å) versus the pipecolic acid form (1.97 Å), and (2) a rotation of the N-terminus of the oxazolone-ring, which is being formed (dihedral angle NP-C(O)A2-CA2-NA2 = 100°: NPip-C(O)A2-CA2-NA2 = 120°) resulting in a decreased degree of unfavorable steric interactions for [AAPipAA + H]+ (interaction distances increase 0.1–0.25 Å, Figure 2). Although these changes are relatively subtle, the differences in the corresponding b 2 -y N-2 TSs are much greater.

The switch from proline to pipecolic acid results in inversion of the stereochemistry of the critical b 2 -y N-2 TSs (Figure 1, Supplementary Figure S4). The necessity of considering stereochemistry arises from protonation of the ring nitrogen (N2) resulting in R or S stereochemistry depending on whether the added proton is “above or below” the ring (Supplementary Figure S4). This was recognized previously by Bleiholder et al. [24], who found the lowest energy critical amide bond cleavage b 3 -y 2 TS of the [AAAPA + H]+ peptide to be ~25 kJ mol–1 lower for the R stereo configuration. Consistent with this finding, our [AAPAo + H]+ systems also preferentially favor the R stereo configuration for the analogous b 2 -y N-2 TSs (Figure 1a, Supplementary Figure S4). In marked contrast, however, the [AAPipAo + H]+ congeners lowest energy TSs all have S stereochemistry (Figure 1b, Supplementary Figure S4). Why the difference? The bulkier, 6-membered ring of Pip is far less sterically hindered by interactions with the methyl group of the N-terminal alanine in the S configuration. Essentially, the ring is placed perpendicular to the planar C-terminal end of the protonated peptide and thus limits deleterious interactions with the alanine methyl group. The combination of the relative destabilization of the most competitive b 2 -y N-2 TSs coupled to the structural adjustments that stabilize the b 3 -y N-3 TSs provides the explanation for the change in major product from y N-2 peaks for the proline-containing systems to b 3 peaks for those containing a pipecolic acid residue (i.e., the pipecolic acid effect). As these structural effects are not independent and occur simultaneously, it’s not possible to single out one as the sole cause.

Product Energies and Fragmentation Regime: Why Do We Detect the Peaks We Detect?

It should be explicitly noted that the experimental data to which we are comparing our calculations was collected at energies substantially above the threshold for fragmentation. Consequently, the best measure of the reaction propensities we provide is the Gibbs free energy, ΔG 298K, as this incorporates both enthalpic and entropic contributions, which are important in this energy regime. If, instead, we were predicting the behavior at threshold (ΔE el+ZPE,0K), the enthalpic barrier (TS or products, whichever is higher) would be most relevant. This is particularly pertinent for the present systems, as at threshold, all of these reactions are product-limited. Consequently, at energies barely above threshold, we would expect a different experimental result: [AAPAA + H]+ and [AAPipAA + H]+ should both lead to predominant y 3 peaks, as the lowest energy thresholds (ΔE el+ZPE,0K) are 160.5 and 136.8 kJ mol–1, respectively, for neutral AA-oxazolone and y 3 + (Tables 3 and 4). As the degree of activation increased further above the threshold energy, the Gibbs free energy, ΔG, (TS or products, whichever is higher) becomes the most important quantity. It is important to recognize that the ΔG of the TS and products vary at different rates with temperature. This is because the combined entropy of two separated gas-phase species (one ion, one neutral here) adds up to a substantially larger entropy than that of a single gas-phase ion (like the pertinent TS), i.e., the ∆S term is large and positive for the products (~190 J mol–1, Tables 3 and 4), in comparison to the TS. So at very high temperature, very few dissociation reactions will be product-limited. Practically, this means that if the ΔG of the products is greater than that of the TS at threshold, as in the b 3 -y 2 pathway of [AAPipAA + H]+, this situation will invert at higher effective temperature, enabling b 3 ion production to become increasingly more competitive. The experiments analyzed here [17, 20] were performed under such conditions, thereby enabling the resulting spectra to favor the b 3 -y 2 pathway products for pipecolic-acid-containing systems (producing b 3 ions here).

Table 3 Summary Table of Separated Product Energies of [AAPAA + H]+ calculated at the M06-2X/6-31+G (d,p) Level of Theory
Full size table
Table 4 Summary Table of Separated Product Energies of [AAPipAA + H]+ Calculated at the M06-2X/6-31+G (d,p) Level of Theory
Full size table

Exploring Additional Stereochemical Differences with [AA(D-Proline)AA + H]+

Raulfs et al. [20] also examined the effect of substitution of a D-proline residue in place of the L-proline or L-pipecolic acid residues, [AA(D-Proline)AA + H]+. This system provided yet another type of dominant fragmentation chemistry, producing the b 4 peak as the base peak. At approximately a quarter of this abundance is the y 3 peak, followed by a barely discernible b 3 peak. So the typical “proline-effect” fragmentation chemistry (b 2 -y 3 pathway) is only the second most prevalent chemistry here. Consistent with this experimental finding, the lowest energy TS located was the b 4 -y 1 at 94.9 (101.4) kJ mol–1 at the M06-2X/6-31+G(d,p) level of theory. The b 2 -y 3 barrier was 122.5 (120.4) kJ mol–1 and finally the b 3 -y 2 at 137.9 (138.3) kJ mol–1. This is summarized in Supplementary Table S15 and the transition structures can be seen in Supplementary Figures S5 and S6. The b 4 ion product is also the most energetically favorable product based on our calculations (Supplementary Table S19), followed by the y 3 then the b 3 , which again is consistent with the experimental data.

Larger Basis Sets and Alternate Model Chemistries

In response to a reviewer request, we also performed additional calculations at the M06-2X/6-311++G(2d,p), B3LYP/6-31+G(d,p), and B3LYP/6-311++G(2d,p) levels of theory for the [AAXAA + H]+ systems. The findings of these calculations are generally consistent with the preceding explanation, so we have limited the discussion of these results to the present section and the Supporting Information. Increasing the M06-2X basis set size resulted in slightly lower TS barriers for both types of amide bond cleavage barrier (by ~10 and ~3.6 kJ mol–1 for the P and Pip congeners, respectively, Supplementary Tables S1, S2, S9, S12). Additionally, the larger basis set indicates that the b 2 -y N-2 TS for the [AAPipAA + H]+ system is much less entropically favorable than the b 3 -y N-3 TS. Although this finding is entirely consistent with the experimental result (supports b 3 ion formation over y 3 ), the magnitude of change is a little surprising, particularly as the M06-2X structures are essentially identical at the two levels of theory. So we have a difference in the description of the frequencies provided between the two levels of theory. In comparison, both sets of B3LYP values are similar to each other. Both B3LYP basis sets lead to the b 2 -y N-2 TS being more energetically and entropically demanding than the b 3 -y N-3 TS, consistent with the experiment. The magnitude of the entropic difference is smaller than for the M06-2X/6-311++G(2d,p) data though. The other difference observed with the B3LYP functional is switching of relative energies of the lowest energy protonation sites from O3 to O2. Although notable, this has minimal impact on the general description of the dissociation chemistry.

Conclusions

Our calculations indicate that proline and pipecolic-acid-containing protonated peptides should have differing product ion distributions under low-energy CID conditions. The reasons for this are: (1) the previously hypothesized increased flexibility of pipecolic acid [20] enabling increased stabilization of the b 3 -y N-3 amide bond cleavage transition structures relative to their proline-containing congeners; (2) a relative destabilization of the b 2 -y N-2 transition structures, which manifests as a required inversion in the stereochemistry of the b 2 -y N-2 transition structure from R (proline) to S (pipecolic acid). This is essentially due to the bulkier Pip side-chain imposing significant steric constraints. Additionally, we provide evidence for the further relevance of experimental energy regime when attempting to rationalize mass spectra and make predictions based on our calculations for related P/Pip systems’ likely product distributions.