Impact of Protonation Sites on Collision-Induced Dissociation-MS/MS Using CIDMD Quantum Chemistry Modeling

Abstract

Protonation is the most frequent adduct found in positive electrospray ionization collision-induced mass spectra (CID-MS/MS). In a parallel report Lee, J. et al. J. Chem. Inf. Model. 2024, 10.1021/acs.jcim.4c00760, we developed a quantum chemistry framework to predict mass spectra by collision-induced dissociation molecular dynamics (CIDMD). As different protonation sites affect fragmentation pathways of a given molecule, the accuracy of predicting tandem mass spectra by CIDMD ultimately depends on the choice of its protomers. To investigate the impact of molecular protonation sites on MS/MS spectra, we compared CIDMD-predicted spectra to all available experimental MS/MS spectra by similarity matching. We probed 10 molecules with a total of 43 protomers, the largest study to date, including organic acids (sorbic acid, citramalic acid, itaconic acid, mesaconic acid, citraconic acid, and taurine) as well as aromatic amines including uracil, aniline, bufotenine, and psilocin. We demonstrated how different protomers can converge different fragmentation pathways to the same fragment ions but also may explain the presence of different fragment ions in experimental MS/MS spectra. For the first time, we used in silico MS/MS predictions to test the impact of solvents on proton affinities, comparing the gas phase and a mixture of acetonitrile/water (1:1). We also extended applications of in silico MS/MS predictions to investigate the impact of protonation sites on the energy barriers of isomerization between protomers via proton transfer. Despite our initial hypothesis that the thermodynamically most stable protomer should give the best match to the experiment, we found only weak inverse relationships between the calculated proton affinities and corresponding entropy similarities of experimental and CIDMD-predicted MS/MS spectra. CIDMD-predicted mechanistic details of fragmentation reaction pathways revealed a clear preference for specific protomer forms for several molecules. Overall, however, proton affinity was not a good predictor corresponding to the predicted CIDMD spectra. For example, for uracil, only one protomer predicted all experimental MS/MS fragment ions, but this protomer had neither the highest proton affinity nor the best MS/MS match score. Instead of proton affinity, the transfer of protons during the electrospray process from the initial protonation site (i.e., mobile proton model) better explains the differences between the thermodynamic rationale and experimental data. Protomers that undergo fragmentation with lower energy barriers have greater contributions to experimental MS/MS spectra than their thermodynamic Boltzmann populations would suggest. Hence, in silico predictions still need to calculate MS/MS spectra for multiple protomers, as the extent of distributions cannot be readily predicted.

Graphical Abstract

1. INTRODUCTION

Electrospray is the most popular ionization method linking liquid chromatography with mass spectrometry, using electrical energy as voltage differences to transfer molecular ions from solution into the gas phase.¹ Mechanistic theories postulate that electrospray starts with a solvated ion that is then transferred as solvated ions into the gas phase (ion evaporation, charge residue, and chain ejection models) before ultimately reaching the mass spectrometer as isolated ions. Protonation of organic molecules is the most often observed ion in positive electrospray ionization. Intuitively, protonation occurs at the most basic site on the target molecule. However, the actual regiospecific protonation sites (protomers) may change depending on the interaction with solvent molecules that induce liquid-phase characteristics, which may present different acid/base equilibria or stability from gas-phase characteristics.^2,3 The current consensus is that multiple protomers and/or tautomers can be produced by the ESI process.^4,5 Therefore, different protomers often contribute to different product ion formations, yielding different mass spectral patterns. Furthermore, electrospray source parameters such as the probe position, the sampling cone voltage, the solvent composition at the time of eluting from the chromatographic column, and analyte concentrations may exert pronounced effects to determine the gas phase-preferred protonation sites of molecular adducts.^6-8 Hence, multiple parameters may define the distribution of protomers under experimental conditions.⁹ The impact of the different parameters on protomer formation is actively debated.^10-15

Protonation sites define fragmentation pathways within the collision cell of mass spectrometers that yield the final tandem mass spectra.^5,16-19 Protonation affects which bonds will be elongated or weakened or possibly already broken before the molecular ion enters the collision cell of the mass spectrometer (“in-source fragmentation”). If there were a single predominant protomer within the chemical environment of the mass spectrometer and if we could predict that specific protonation site, then, MS/MS predictions might become less challenging. Interestingly, some MS/MS product ions may be formed by different fragmentation pathways from more than one protomer ion species.²⁰ It has also been shown that the ionization energy itself may change the population of protomers, for example, for aniline.⁷ Hence, current MS/MS spectra must be predicted on all possible protomers in CIDMD software, resulting in a high computational cost as well as statistical uncertainty in MS/MS predictions. An accurate prediction of the prevalent protomers within a specific experimental condition (pH, solvents, ionization energies, buffer salts) would be very helpful to utilize CIDMD for large-scale MS/MS predictions of molecules that cannot be purchased or manufactured easily.

Protonated molecules are considered to fragment by charge migration (to another atom) or charge retention (at the same atom), which are empirical rules created to assign fragments in mass spectra.^21,22 However, these rules cannot completely explain many fragmentation patterns.^21,23 In addition, the “mobile proton” model was established in peptides to explain the difference of locations of proton adducts before and after electrospray ionization. For example, Wysocki et al. proposed a mechanism for how protonation of aspartic acid and histidine residues can lead to enhanced peptide bond breakage for CID- or surface-induced dissociation-MS/MS.^24,25 Here, initially, protons are localized at the most basic sites in molecules, for instance, the N-terminus of a peptide or the side chain of arginine, lysine, and histidine. During the electrospray ionization process, protons are then transferred to less-basic sites to a thermodynamically higher energy state. This mechanism leads to charge-directed cleavages at less-basic sites in protonated peptides and has been studied both experimentally and by theoretical calculations.^24,25 However, it is difficult to accurately predict the location to where the metastable “activated proton” migrates from the most basic site and what the proportion of different protomers might be within the cloud of molecular ions during and after the electrospray process.

In our parallel study presenting CIDMD (collision-induced dissociation molecular dynamics),²⁶ we selected multiple protonation sites to perform CIDMD and generated mass spectra of 12 small molecules. We showed that computational chemistry methods can be used to predict theoretical CID-MS/MS mass spectra with high similarity to the experimental data.²⁶ We confirmed that protomers produced different distributions of product ions and hence distinctive theoretical CIDMD mass spectra and therefore also impacted the similarity between predicted and experimental MS/MS spectra.²⁶

We present here the largest and most detailed study to date that investigated the impact of the protonation site on the predicted CIDMD mass spectra of metabolites and their comparisons to experimental CID-MS/MS spectra. First, we used more molecules and more protomers than previous investigations to obtain a more comprehensive understanding of matching predicted to experimental MS/MS spectra. Second, we calculated the proton affinity for all protomers in the gas phase (i.e., the molecule in vacuum) in comparison to solution phases; a 1:1 mixture of water and acetonitrile was used to determine whether the gas- or liquid-phase proton affinity of a protonation site was correlated with the quality of the predicted mass spectrum of the corresponding protomer. Third, we considered proton migration as an alternate hypothesis because we did not find strong positive correlations between theoretical proton affinities and experimental MS/MS similarity scores. Here, we aimed at specifying the location of mobile proton migrations (including migration routes and activation energies) to trace back the prevalent protomers for the 10 small-molecule test cases presented in this study.

2. MATERIALS AND METHODS

2.1. Preparing the Molecular Ion.

A brief description of the CIDMD framework and analysis package is shown in Figure 1. Molecular structures with specific protonation sites were manually constructed using Avogadro software.²⁷ The geometries of molecular structures were then optimized with the (u)B3LYP density functional²⁸ and the 6-31G** basis set²⁹ using TeraChem.³⁰

Figure 1.

Overall procedure of the CIDMD framework and analysis package.

2.2. Setting Up the CIDMD System.

We designed CIDMD to simulate the collision-induced dissociation process that occurs during CID-MS/MS experiments. CIDMD is based on simulating the collision dynamics between the molecular ion and an inert gas molecule (called the collider). In the initial condition of the simulation, the inertial frame was chosen such that the molecular ion center of mass is initially at rest (called the MD frame) and the initial positions and velocities of the collider atoms were chosen such that they form a straight line that intersects with the molecular envelope of the molecular ion. For more details of the CIDMD framework procedure, please see our parallel paper published in this issue [REF inserted during the publication process]. The collider atom is initially positioned 5 Å away from the molecular envelope. During the CIDMD, the collider gas atom moves at a constant velocity toward the molecular ion and initiates the collision. Multiple CIDMD simulations were performed varying the initial coordinates and velocities of the collider atom. All CIDMD was performed with unrestricted B3LYP/6-31G* in the microcanonical (NVE) ensemble with a time step of 1.0 fs utilizing TeraChem. The total allowed time frame for this CIDMD process was 1 ps for the majority of the simulations, but in some cases, it was allowed to proceed for a longer time scale, up to 50 ps. Finally, the simulation trajectories were processed with the analysis package that we developed.

2.3. Analyses.

The CIDMD analysis package was developed to determine the accurate mass of the monoisotope fragments up to 10 mDa to allow for mass errors in the experimental MS/MS spectra. Models also determined the charges of molecular fragments formed in the simulation trajectories. The frequencies of the observed ions were converted into the simulated mass spectra. The bond order time series method³¹ was employed for reaction detection and employed to identify fragments during CIDMD simulations. By taking the sum of the Mulliken populations over the atoms in the fragment, we estimated the charges on the fragments at each time step. Fragments were considered to be charged at average charge values >0.70. Details and full abilities of the CIDMD analysis package are described in our parallel paper.²⁶ We evaluated simulated CIDMD mass spectra by calculating entropy similarity scores compared to experimental MS/MS spectra in the NIST20 library as follows³²

$$\text{entropy similarity}=1-\frac{2\times S_{\text{AB}}-S_{\mathrm{A}}-S_{\mathrm{B}}}{\text{ln}4}$$ (1)

where $S=-{\sum }_{\mathrm{p}=\mathrm{i}}^n{I}_p^{w}ln{I}_p^w$ and for $I^w\{\begin{array}{c}w=1(S\ge 3)\hfill \\ w=0.25+S\times 0.25{(S<3)}^{.}\hfill \end{array}\phantom{\}}$

While entropy similarity scores are a good numerical representation of how similar two mass spectra are, the use of a numerical score can obscure interesting details in pairwise spectral comparison in MS/MS spectra. Graphical comparisons of mass spectra were performed to judge the presence and absence of fragment ions.

Frequency analyses were performed at the same level of theory as the initial optimization with DFT (u)B3LYP/6-31G** using TeraChem. Any molecular system simulating the liquid phase was calculated with COSMO (conductor-like screening model), a continuum solvation model. To model the solvent mixture of acetonitrile/water (1:1), the dielectric constant $\epsilon$ was set to 56.13. To calculate the dielectric constant of the heterogeneous mixture, $\epsilon$ is as follows³³

$$\epsilon ={[({\epsilon }_2^{1∕3}-{\epsilon }_1^{1∕3})v_2+{\epsilon }_1^{1∕3}]}^3$$ (2)

where ${\epsilon }_1$ and ${\epsilon }_2$ are two dielectric constants of pure respective solvents and $v_2$ is the volume of the second solvent in the mixture.

We calculated the proton affinity of each protonation site for molecules to determine the position of the proton under biochemical standard conditions. For the reaction $A+H^{+}\to A{H}^{+}$, the proton affinity (PA) of the reaction is $\text{PA}(A)=-\Delta H=-\Delta E+RT=-\Delta E_{\text{elec}}-\Delta \text{ZPVE}+\Delta E_{\text{vib}}(T)+C$, where $H$ is the enthalpy of the reaction, $\Delta E_{\text{elec}}$, $\Delta \text{ZPVE}$, and $\Delta E_{\text{vib}}(T)$ are the differences in electronic ground-state energy, zero-point vibrational energy, and thermal vibrational energy between $A{H}^{+}$ and $A$, respectively, and $C$ is the correction for the translational rotational energy change, assuming classical behavior. Proton affinity is then calculated as

$$\text{PA}(A)=-\Delta E_{\text{elec}}-\Delta \text{ZPVE}+\Delta E_{\text{vib}}(T)+\frac{5}{2}RT$$ (3)

To find the minimum-energy reaction paths for proton migrations, we employed the nudged-elastic band (NEB) method. Geometric software³⁴ was used to create the input chains of 50 structures on an interpolated path including the initial and final coordinates of protonated atoms. Geometric software³⁴ engaged the Psi4 program³⁵ to perform NEB with optimizations at each step with B3LYP/6-31G*. The NEB was applied to the intermediate structures between each set of initial and final states with a spring constant of 0.1 using 100 iterations. Transition-state structures were optimized using Geometric software.³⁴ Finally, intrinsic reaction coordinate (IRC) calculation was performed using Gaussian³⁶ to confirm the transition states to find the activation energies for proton migration reaction pathways.

3. RESULTS

To better understand the impact of protonation sites on the prediction of collision-induced mass spectra, fragmentation pathways, and mechanisms of mobile proton migrations, we selected the same small molecules as in our parallel report.²⁶ These compounds can be classified into four categories: (1) small cyclic molecules (<150 Da) uracil, dihydroxy pyrimidine, aniline, and ribono-γ-lactone, (2) noncyclic small acids at <131 Da itaconic acid, citraconic acid, mesaconic acid, and sorbic acid, (3) noncyclic larger molecules (>150 Da) taurine and citramalic acid, and (4) cyclic larger molecules (>150 Da) bufotenine and psilocin.

3.1. Proton Affinity Calculations.

We first calculated the proton affinity in the gas phase and acetonitrile/water mixture (1:1) for each molecule and for all heteroatom protonation sites (Figure 2). Most of the proton affinity values ranged from 170 to 270 kcal/mol. As expected, the calculated proton affinities for solvents were about 40–100 kcal/mol higher than in the gas phase. Similarly, we observed higher proton affinity values for the carbonyl protonation in comparison to those for the hydroxyl protonation. When comparing the trans- and cis-isomers mesaconic acid and citraconic acid, we found that citraconic acid protomers #1 and #3 (Figure 2) produced much lower matching CIDMD mass spectra compared to the corresponding protomers in the trans-isomer mesaconic acid (SI Figure 1). This finding indicates a lower probability of the carbonyl groups being protonated in citraconic acid during the experimental CID process than in mesaconic acid. Surprisingly, all protomers of the positional isomer itaconic acid, including protonated hydroxyl groups, yielded high entropy similarity to the experimental MS/MS spectrum despite the difference in proton affinity values for hydroxyl- and carbonyl-oxygen atoms. Likewise, we found surprising differences between proton affinity and MS/MS experimental similarity for taurine. Here, the CIDMD simulation for the protonated amino group (with the highest proton affinity) gave lower similarity scores compared to the carbonyl- or hydroxyl-protonated taurine protomers (Figure 2). Similar trends exist within the protomers of bufotenine and psilocin.

Figure 2.

Scatter plots of calculated proton affinity and entropy similarity scores. In each plot, the molecular structure is shown. The numbers in black next to heteroatoms represent the protonation sites. The color blue represents the proton affinity values calculated in the acetonitrile and water 1:1 mixture ($\epsilon =56.13$), whereas red represents the values calculated in the gas phase.

We therefore expanded this comparison for all molecules and all protomers, correlating the calculated proton affinity values and their respective entropy similarity scores to the experimental MS/MS spectra (Table 1). For the four compounds citraconic acid, sorbic acid, taurine, and citramalic acid, we found strong negative correlations. Counterintuitively, protomers with the highest proton affinity values yielded the lowest entropy similarity scores (Figure 2, Table 1). Indeed, for the other molecules studied here, bufotenine, psilocin, itaconic acid, and mesaconic acid, variable correlations were found between −0.4 and +0.8 for MS/MS similarity and proton affinity (Table 1). This could indicate that molecular ions may exist in different protomer states during CID fragmentation, possibly even at different proportions of all protomers. Still, it is difficult to state that carbonyl protomers are completely absent; rather, it is likely that different protomer proportions exist simultaneously for each molecule.

Table 1.

Spearman-Rank $R$ Calculations Correlating Predicted Proton Affinities of all Protomers in Gas (Vacuum) Against Experimental Entropy Similarity MS/MS Matching Scores and Resp. Proton Affinities in the Water/Acetonitrile Solvent Phase versus Entropy Similarity

	proton affinity (gas) to similarity score		proton affinity (water/ acetonitrile) to similarity score
molecules	Spearman’s $R$	P value	Spearman’s $R$	P value
psilocin	0.50	0.667	0.50	0.667
bufotenine	0.50	0.667	0.50	0.667
citramalic acid	−0.80	0.104	−0.87	0.054
taurine	−1.00	0.0	−1.00	0.0
itaconic acid	0.80	0.2	−0.40	0.6
mesaconic acid	0.40	0.6	0.40	0.6
citraconic acid	−1.00	0.0	−0.80	0.2
sorbic acid	−1.00	0.0	−1.00	0.0

How can we explain this difference in protomer states? First, one can argue about differences in distances between hydroxyl and carbonyl groups. For citraconic acid, both carboxyl groups are oriented toward each other via the fixed cis-double bond. In comparison, the trans-double bond in mesaconic acid inhibits any interaction between the two carboxyl groups. Hence, for isolated carboxyl groups (as also found in sorbic acid), the carbonyl and hydroxyl oxygen atoms cannot be truly distinguished under experimental conditions, especially under solvation (and protonation/deprotonation equilibria). However, for citraconic acid, the interaction of the two carboxyl groups may favor a double protonation of an oxygen atom (also known as a protonation of a hydroxyl moiety) in comparison to adding protons to both oxygens (also known as a protonation of a carbonyl moiety).

Another explanation may be found for the large difference in correlation scores for itaconic acid under gas-phase and solvation-phase protonation states (Table 1) as well as for the obvious difference in MS/MS similarities and proton affinities for taurine (Table 1, Figure 2). Under typical experimental conditions such as for electrospray MS/MS spectra in the NIST20 database,³⁷ molecules are dissolved in acetonitrile and water (50:50, v/v), acidified with formic acid. Due to energy input under the high temperatures in electrospray ionization (often 250–350 °C) and high voltage differences (typically >3 kV), energy distributions may make higher energy states more probable relative to their Boltzmann populations, favoring proton migration. This observation supports the “mobile proton” hypothesis, which says that a molecule is initially protonated at the most basic site, but protons may migrate to other sites that are less basic. This process may certainly be assisted by water molecules that form adducts with the target molecules during the electrospray process rather than migrating protons along carbon chains.

Unfortunately, the lack of clear correlations between proton affinities and MS/MS entropy similarity scores means that all protomers must be considered in CIDMD predictions. It also means that additional positional and water coordination effects need to be considered for calculating the most likely protomers. Although the proton affinities for the gas and liquid phases showed major differences, the change in phase did not change the proton affinity rank ordering of protomers or the correlation with the experiment. This result supports the notion that solvation effects have minor contributions compared to positional isomer effects, as observed for citraconic and mesaconic acid.

3.2. Mechanistic Details of Fragmentation Reaction Pathways.

Next, we studied fragmentation pathways to link fragment ions back to the initial protomers. If these pathways were specific in comparison to other protomers, then, one might be able to glean information about the relative abundance of different protomers in a precursor ion population. We selected specific cases that we hypothesized might be specifically suitable. Protonation sites may impact structural characteristics such as bond elongation or weakening. Such characteristics might favor certain fragments or fragmentation pathways. The indole derivatives bufotenine and psilocin structures include three heteroatoms each, raising the possibility that any of these three protomers could exist under experimental MS/MS conditions. For both bufotenine and psilocin, we found that protomer site #3 at the dimethyl amine moiety exerted the highest proton affinity values; however, in both cases, the protomer #3 did only yield medium entropy similarity scores when compared to the experimental MS/MS spectrum (Figure 3A, SI Figure 2). While the lowest proton affinity value at protonation site #1, the hydroxyl group, gave also the lowest MS/MS similarity score, it was in fact protonation site #2 (at the indole nitrogen) that had only median proton affinity but by far the highest MS/MS similarity score (Figure 3A).

Figure 3.

(A) Relationship between proton affinity and CIDMD entropy similarity scores for all bufotenine heteroatom protomers. Numbers 1–3 refer to protomer positions given in structures 3C. (B) Comparison of the HCD bufotenine experimental mass spectrum (top) at a collision energy of 8 eV with the CIDMD-predicted mass spectra of the three bufotenine heteroatom protomers (lower panels). (C) Main fragmentation pathways for the bufotenine protomers #1, #2, and #3. Green check marks: product ions observed in the experimental spectra. Red cross-mark: product ion not observed experimentally.

When we simulated CIDMD spectra for all three protomers for bufotenine and psilocin, we observed three distinctive mass spectra with three clearly different similarity scores (Figure 3A,B). For instance, for bufotenine, protomer 1 yielded two major ions at m/z = 58 and m/z = 187 (Figure 3B). CIDMD predicted an initial water loss if bufotenine was protonated at the hydroxyl group (Figure 3C); however, this predicted ion m/z 187.124 was not detected in the experimental data. While a secondary fragmentation might open the indole ring structure as neutral loss, producing the trimethyl amide ion m/z 58.648, this product ion could be more easily explained by a direct neutral loss from the intact bufotenine structure from protonation site #2 at the indole nitrogen (Figure 3C). Conversely, protonation at the most proton-affine site #3, the alkylamine nitrogen, does not show a pathway toward producing m/z 58.648. Hence, this pathway can be excluded. If we assume that protomer #3 might be the dominant cation before the electrospray process, it is clearly not the case when undergoing the CID fragmentation. Again, the data support the notion of a “mobile proton” mechanism from protomer #3 to protomer #2 to mechanistically explain the experimental fragment ions. Psilocin mechanistic investigations showed very similar trends (SI Figure 3). Just like bufotenine, psilocin can be protonated at three different sites, and again, only a protonation at the indole-ring nitrogen can sufficiently explain the experimentally observed fragment ions.

We then tested if we could distinguish the most likely fragmentation mechanisms for the two oxygen atoms of a simple aliphatic carboxylic acid, sorbic acid (Figure 4). Here, two different protomers can be defined, either protonation at the hydroxyl moiety or the carbonyl moiety. While both oxygens bear similar proton affinities in both the gas and solution phases, we found a significant difference in entropy similarity scores even when trying to fit the best-matching experimental MS/MS spectrum of 805 (for the hydroxyl protonation CID, at 16 eV collision energy) and 722 (for the carbonyl protonation CID, at 7 eV collision energy), Figure 4. Fragmentation pathway analysis for the protomers showed that both protomers undergo the exact same fragmentation pathways that lose water and carbon monoxide, resulting in a m/z of 67.054, 85.064, and 95.049 ions (Figure 4B). Nevertheless, protonation at the carbonyl group showed lower entropy similarity due to the differences in the ratio of predicted ions 95/85, meaning the relative likelihood of losing a water molecule (to produce m/z 95) or carbon monoxide (producing m/z 85). Hence, the protonated hydroxyl group is more likely to be the dominantly populated protomer for sorbic acid when undergoing CID fragmentation.

Figure 4.

Sorbic acid protomers. (A) Comparison of the CIDMD-predicted mass spectra to the experimental spectra for two sorbic acid protomer #1 (left) and protomer #2 (right). (B) CIDMD-predicted fragmentation pathways demonstrating CO and H2O losses for both protomers.

We continued our mechanistic studies by detailing the different fragmentation pathways for 13 possible protomers in uracil in both of its tautomeric structures. Fragmentation of uracil is widely debated.^38-42 It consists of a small conjugated system, and its protonated sites highly affect the dissociation patterns. Under experimental conditions, the main CID pathways of the protonated uracil molecular ion can be easily determined by visualizing breakdown curves of experimental mass spectra under 10–35 eV collision energies (Figure 5). The molecular ion of uracil m/z 113 remains the base peak until >17 eV collision energies, indicating the large stability of aromatic compounds compared to aliphatic structures like sorbic acid. With increasing collision energies, fragments m/z 96 and 70 become prevalent ions. In addition, m/z values of 68 and 53 increase to up to 30–40% bp intensity at 35 eV energy (Figure 5). Initial neutral losses of H₂O, NH₃, and HNCO were experimentally verified at collision energies of <15 eV,⁴⁰ yielding fragment ions at m/z 95, 96, and 70, respectively (Figure 5). Under higher energy collisions, m/z 96 continues to lose CO, while m/z 70 fragments a neutral loss of NH₃ (Figure 5).

Figure 5.

Energy-dependent fragmentation of uracil and its protomers. Abundance of experimental positive-mode MS/MS product ions in the NIST20 library for increasing MS/MS collision energies.

A total of 18 stable protomers and tautomers of uracil may be present as molecular ions.³⁸ We included six of these protomers for detailed mechanistic studies (Figure 6B, top panel) based on lower ground-state energy values. These six protomers may therefore be more stable and are therefore likely to exist in higher proportions in experimental MS/MS fragmentations.³⁸ We added an additional lower ground-state energy protomer (Figure 6B, top panel, UL7), plus 6 higher ground-state energy protomers (Figure 6A, top panel). We then calculated proton affinities (denoted as PA) in gas for all 13 test protomers (Figure 6A, B). Experimental proton affinity values have been reported for uracil from 853 to 899 kJ/mol (203.9–214.9 kcal/mol), depending on the different protonation sites.^38,43 QM calculations yielded 203.9 kcal/mol³⁸ for one of the protomers, close to experimental values.

Figure 6.

CIDMD fragmentation of uracil and its protomers. (A) Entropy similarity scores, proton affinities, structures, CIDMD product ion abundances, and MS/MS similarity scores of six higher ground-state energy uracil protomers (UH1–6). (B) Entropy similarity scores, proton affinities, structures, CIDMD product ion abundances, and MS/MS similarity scores of seven lower ground-state energy uracil protomers (UL1–7). (C) Extracted CIDMD fragmentation reaction pathways from the CIDMD simulations of uracil lower energy protomer UL7. (D) Head-to-tail graph of CIDMD-predicted and experimental mass spectra at 20 eV collision energy. Correctly predicted fragment ions are annotated.

We found here proton affinity values for the six selected high-ground-state energy uracil protomers (UH) to be ranging from 167.5 to 203.4 kcal/mol and 187.08–206.2 kcal/mol for the seven low-ground-state energy uracil protomers (UL). Our CIDMD calculations did not yield statistically significant different entropy similarity scores for UH and UL uracil protomers, with UH protomer recordings at 716 ± 117 entropy similarity scores (compared to the experimental MS/ MS spectrum) and 685 ± 167 similarity scores for UL protomers. Interestingly, the three protomers UL2, UL3, and UH4 that were the closest to the reported proton affinity of 203.9 kcal/mol yielded worse similarity to the experimental data (Figure 6A,B, S4) than protomers with higher or lower proton affinities, reaffirming our earlier observation that proton affinity alone is a poor predictor of MS/MS similarities. We therefore investigated each fragmentation pathway for each protomer. One of the most prevalent fragment ions, m/z 70.039, was well-predicted by all protomers, whereas m/z 96.001 was only found in four low-energy UL protomers (Figure 6B, UL4-UL7) that all had double protonation ring nitrogens that then underwent a neutral loss of NH₃ to yield m/z 96.001. Among all tested protomers, only UL7 was predicted to undergo all important fragmentation reactions that matched all product ions in its corresponding best-match experimental mass spectrum (Figure 6C). For UL7, we found that its local minimum energy conformer favored breaking the bond between the ketone carbon (C-4) and the doubly protonated nitrogen (N-3) (Figure 6C). Hence, this UL7 conformer allowed uracil to undergo fragmentation reactions that led not only to the loss of H₂O and HNCO but also to NH₃ and NH₂COH losses, yielding m/z 96.001 and 68.014 (Figure 6). These two fragmentation reactions were found only for uracil protomers with positively charged ring nitrogens. If we assume that initial protonation occurs at the oxygen atoms that have higher calculated proton affinities, our results support the mobile proton theory as our CIDMD predictions indicate uracil N-protomers UL4, 5, 6, and 7 as most likely species to yield the experimental MS/MS spectra.

3.3. Mobile Proton Migration.

We calculated energy profiles to test how protons may have migrated across a molecule backbone from initial high-proton affinity sites to less basic sites before fragmenting in the MS/MS collision cell. In the literature, different protonation sites have been debated vigorously for aniline.^7,9,17,44,45 Aniline may be protonated either at the amino nitrogen or at the para position of the ring-carbon (C_para) (Figure 7, far bottom left and right). Our proton affinity calculation showed that C_para-protonated aniline has a proton affinity energy of 212.21 kcal/mol, whereas the nitrogen-protonated aniline has an energy of 210.10 kcal/mol. Hence, the calculated proton affinity difference is only 2.11 kcal/mol between these two protomers. We calculated 1.99 kcal/mol lower ground-state free energy for the C_para-protonation site compared to nitrogen protonation. Counterintuitively, these calculations indicate that C_para is the thermodynamically favored structure, meaning that a C_para-protonated aniline must convert to the N-protonation state before it can undergo a neutral loss of ammonia [M+H-NH₃]⁺. A previous study⁹ reported that the fragmentation energy was identical for both protomers, suggesting that the same bond was dissociated.

Figure 7.

Proton migration and MS/MS fragmentation for aniline. (A) Transition-state calculations for a mobile proton transfer pathway from Cpara-protonation to nitrogen protonation. (B) Detailed proton migration (yellow) from C_para to nitrogen-protonated aniline, followed by a fragmentation reaction pathway of the nitrogen-protonated aniline. (C) Experimental MS/MS breakdown curves by collision energies using nitrogen or argon as collision gases, for triple-quadrupole (QQQ) or high-collision dissociation (HCD) Orbitrap mass spectrometers.

When calculating the transition-state energy needed to move the proton from the C_para-protonation site to the aniline-nitrogen site, we found an energy difference of 60.78 kcal/mol (2.6 eV) (Figure 7). Experimentally, at low ESI cone voltages of 10 eV, the nitrogen-protonated aniline is dominant, whereas C_para-protonated aniline is prevalent at higher cone voltages (100 eV).⁷ Hence, proton transfer between two protomer sites is energetically feasible because the molecule may take up energy during the electrospray ionization. Our calculation showed that the proton from the C_para position sequentially transfers to the C_meta, C_ortho, C─N, and finally nitrogen of the amine group (Figure 7b). One intermediate structure was found at the C_ortho position. Energy differences between the C_ortho and the C_para position were only 4 kcal/mol. Two transition states were found at the C_meta position (16.1 kcal/mol higher than that of C_para) and at the C-1 atom adjacent to nitrogen (48.15 kcal/mol higher than that of C_para). Overall, our results indicate that proton migration may be possible across the backbone of aromatic structures like aniline, although of course, other mechanisms like water-assisted migrations cannot be excluded.

MS/MS fragmentations depend not only on the protomer state but also on instrument types, collision energies, and even the collider gas. When we contrasted the experimental fragmentation profiles of nitrogen and argon collider gases for high-resolution Orbitrap data (HCD) and low-resolution triple-quadrupole instruments (QQQ), we found that in all cases, m/z 77 was produced as the first and abundant fragment ion (Figure 7C). This ion can only be explained by a neutral loss of ammonia from the protonated aniline, showing in all cases that the N-protomer must dominate in experimental conditions. Hence, even these different experimental conditions support a mobile proton hypothesis, assuming that at least some proportion of the initial protomer state is found at the lower energy C_para-position.

4. DISCUSSION

Identifying metabolites is a foundation of metabolomic studies. In untargeted chemical screening studies, most detected chemical compounds remain unidentified, likely due to the enormous chemical space of combined metabolic reactions and exposome diversity. Because matching experimental mass spectra to library spectra is the foremost method for compound annotations, increasing the size and chemical complexity of MS/MS libraries is key to improving annotation coverage in untargeted MS analyses. We here tested our new CID-MS/MS workflow to predict spectra from the first principles,²⁶ using known compounds to benchmark the efficiency of this tool while explaining specific mechanisms, in particular, protonation states.

We found that different protomers of molecules may result in widely different CIDMD-predicted MS/MS spectra. When we first embarked on this study, we expected to find a positive correlation between the proton affinity and similarity scores of different isomers because we thought that the fragmentation simulations that started from isomers with higher thermodynamic probability would be the most realistic. The results of our simulations diverged significantly from the initial hypothesis. Indeed, in more than half of the cases we studied, we found an inverse correlation. Potentially, these unexpected results might be explained by the limited simulation time of CIDMD fragmentation trajectories at 1 ps, compared to experimental times in MS/MS, which are commonly reported to extend to milliseconds,^46,47 a million times longer than our simulation experiments. If we had the capability to reach longer simulation times, it may be possible to directly observe the pre-equilibration of the protonation states in the molecular dynamics trajectories. However, because proton transfer pathways can be readily guessed from chemical intuition, it may be computationally more practical to study mobile proton transfers using conventional potential energy surface exploration such as the nudged-elastic band method (NEB).

For bufotenine, psilocin, taurine, and aniline, we could not find direct fragmentation pathways starting from the thermodynamically most stable protonation state and leading to all of the experimentally observed fragment ions. In all of these cases, a less-stable protonation state was predicted to have direct fragmentation pathways, leading to the experimentally observed peaks. This is consistent with the mobile proton hypothesis, which states that the protonation state of the molecular ion could fluctuate prior to fragmentation. Our calculations of the proton transfer activation energy in aniline further corroborate this hypothesis, as we found that the proton transfer pathway is energetically accessible at CID energy scales. Solvent water molecules are known to play a catalytic role in proton transfer between protonation sites,⁴⁸ and this may facilitate proton migration if any water molecules are still associated with the molecular ion immediately prior to the collision event.

Our confirmation of the mobile proton hypothesis suggests that it is not important to start simulations from the thermodynamically most stable protonation state. Instead, one should consider the perspective that the protonation states are in a pseudothermodynamic equilibrium, and the most likely fragmentation pathway should minimize the maximum free energy along the pathway. This most likely pathway could proceed from a protonation state whose free energy may be tens of kilocalories per mole above the thermodynamically most stable one. Interestingly, the result of proton migration during electrospray may result in various protomer populations of a target molecule. Some molecules, like uracil, taurine, and aniline, appeared to have a singular dominant protomer, while MS/MS spectra of other molecules such as sorbic acid, bufotenine, and psilocin are best explained by originating from more than one dominant protomer. Recently, a computational study on electrospray MS/MS on serine protomers⁴⁹ compared them in gas, condensed, and microsolvated liquid phases. Starting from the idea that the zwitterionic form of serine is more stable in water, whereas the neutral form of serine may be most stable in the gas phase, the authors found that the number of water molecules present in microsolvated phases affected the tautomeric form of serine.⁴⁹ As little as two water molecules can stabilize the zwitterion state of serine given the presence of additional salt cations, resulting in a low energy difference of the zwitterion serine of only 1.3 kj/mol higher than the minimum energy structure.⁴⁹ Hence, not only protomers but also zwitterionic states should be considered in MS/MS simulations.

Proton migration may occur during the transition from the condensed (aqueous) state to the gas phase during the electrospray process. As per the standard model, electrospray droplets rapidly decrease in size, with an intermediate state for which only a few molecules surround a molecular ion. This process may explain how protonation states of molecular ions are influenced by local environments, as shown for serine in gas-phase microsolvation.⁴⁹ Distributions of protomer populations have also been shown to depend on multiple components in electrospray solvents such as voltage, methanol/acetonitrile differences, or basicity.^7,15,50 Under these rapid changes in environmental conditions of molecular ions, it is possible for protomers to experience kinetic trapping that results in specific protonation states,^7,15,50 which could not be explained by thermodynamic stability in either the gas or liquid phase. Hence, we may need to understand the electrospray process and concomitant formation of protomers as a nonequilibrium system. Consequently, different energy levels impact the accurate prediction of protomer distributions. We therefore calculated ground-state electronic energy, proton affinity, and free energy levels to predict the best protomer sites of the molecules studied here. While we found different contributions for these energy levels to predict protonation sites, no single approach resulted in clear pathways toward defining specific protonation distributions in this type of nonequilibrium experiment.

5. CONCLUSIONS

We report here CIDMD models to study the impact of protonation sites on in silico predicted MS/MS spectra for 10 small molecules protonated at 43 different 43 sites, many more than previous reports. We also comprehensively matched all in silico spectra to all available experimental MS/MS spectra, to test for protomers and fragmentation pathways that may be present in some experimental conditions but not in others. We can clearly confirm that different protomers may lead to highly different CIDMD-predicted MS/MS spectra, sometimes converging to the same product ions by using different fragmentation pathways. Unexpectedly, however, thermodynamic proton affinity calculations did not readily explain which protomer might be present under the experimental MS/MS conditions. Any single thermodynamically favored protomer is likely insufficient to explain the experimental MS/MS spectra. Instead, MS/MS match scores showed a weak correlation to proton affinities. Energy differences during proton migrations are in the low-eV range, which are similar to energies found during the electrospray process. Due to the extensive energy input under experimental conditions in mass spectrometry, protons may migrate from thermodynamically favored sites to other atoms, probably in a mixture of different protonation sites. For complex fragmentation patterns like uracil MS/MS, we found that only one protomer predicted all experimentally detected ions but that this protomer also yielded false-positive fragments and an overall medium MS/MS similarity score. Hence, even in such cases, experimental MS/MS spectra cannot be predicted from single-molecule protomer species. We therefore conclude that (1) our studies suggest that the protomer with the lowest overall free energy along the fragmentation pathway may produce more accurate simulated mass spectra from short CIDMD simulations and (2) in complex cases, such as for uracil, the experimental spectra clearly cannot be explained by simulations started from a single protomer. Other approaches such as machine learning or semiempirical quantum chemistry models may complement density functional theory-based MS/MS predictions to keep computational times manageable.

ABBREVIATIONS

CID

collision-induced dissociation

LC

liquid chromatography

MS/MS

tandem mass spectrometry

ESI

electrospray ionization

MD

molecular dynamics

CIDMD

collision-induced dissociation molecular dynamics

KE

kinetic energy

QTOF

quadrupole time-of-flight

HCD

high-energy collision dissociation

NIST

National Institution of Standard and Technology

CE

collision energy

Data Availability Statement

CIDMD mass spectra are freely available at the MassBank of North America, https://massbank.us, accessed on 04 September 2023. All in-house scripts that were created for this research project are freely available at https://github.com/jesilee/.