12468
J. Phys. Chem. B 2008, 112, 12468–12478
The Effect of the Secondary Structure on Dissociation of Peptide Radical Cations: Fragmentation of Angiotensin III and Its Analogues Zhibo Yang,† Corey Lam,‡ Ivan K. Chu,‡ and Julia Laskin*,† Fundamental Sciences DiVision, Pacific Northwest National Laboratory, Richland, Washington 99352, and Department of Chemistry, The UniVersity of Hong Kong, Hong Kong, China ReceiVed: June 13, 2008
Fragmentation of protonated RVYIHPF and RVYIHPF-OMe and the corresponding radical cations was studied using time- and collision energy-resolved surface-induced dissociation (SID) in a Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS) specially equipped to perform SID experiments. Peptide radical cations were produced by gas-phase fragmentation of CoIII(salen)-peptide complexes. Both the energetics and the mechanisms of dissociation of even-electron and odd-electron angiotensin III ions are quite different. Protonated molecules are much more stable toward fragmentation than the corresponding radical cations. RRKM modeling of the experimental data suggests that this stability is largely attributed to differences in threshold energies for dissociation, while activation entropies are very similar. Detailed analysis of the experimental data obtained for radical cations demonstrated the presence of two distinct structures separated by a high free-energy barrier. The two families of structures were ascribed to the canonical and zwitterionic forms of the radical cations produced in our experiments. Introduction Radical reactions play an important role in biological processes including enzymatic reactions, photosynthesis, and radiative damage. Radical damage of peptide and proteins induced by UV photolysis, interaction with common oxidants, or electron transfer reactions is responsible for numerous oxidative processes within cells, resulting in several human diseases, carcinogenesis, and aging.1,2 Once formed, the radical site is transferred within the protein structure, and the damage is propagated via a variety of pathways including electron transfer, hydrogen abstraction, addition, rearrangement, fragmentation, or substitution. Similar pathways determine gas-phase fragmentation of oddelectron peptide and protein ions in mass spectrometry extensively used both for top-down and for bottom-up characterization of proteins.3,4 Different types of gaseous odd-electron species including [M + nH](n-1)+ · , [M + H]2+ · , M+ · , and [M nH](n-1)- · ions can be produced by photoionization,5 capture of low-energy electrons by multiply protonated precursors,6 electron detachment from multiply deprotonated ions,7 electron transfer between transition metal complexes and analyte molecules during CID of the corresponding ternary complexes,8-12 or through free-radical-initiated reactions.13,14 Fragmentation of hydrogen-rich radical cations, [M + nH](n-1)+ · , is dominated by the cleavage of N-CR bonds, while collision-induced dissociation (CID) results in facile cleavages of amide bonds. Improved sequence coverage for peptides and proteins is often obtained by combining CID with electron capture dissociation (ECD)15,16 and the related technique of electron transfer dissociation (ETD).17,18 It has been suggested that distonic ions, in which the charge site is separated from the radical site, play an important role in the dissociation of M+ · ions.8,19,20 Comparison between the * Corresponding author. E-mail:
[email protected]. † Pacific Northwest National Laboratory. ‡ The University of Hong Kong.
dissociation patterns obtained for the M+ · , [M - 2H]- · , and [M + H]2+ · peptide ions demonstrated that the dissociation of odd-electron peptide ions is often determined by charge-remote radical-driven fragmentation processes.21 For example, many side chain losses and some backbone fragmentation of oddelectron peptide ions occur in both positive and negative modes, suggesting that the charged site is not involved in these fragmentation pathways. Radical-driven fragmentation has been also observed in UV photodissociation of protonated and deprotonated peptides22-24 and in dissociation of peptides cationized with copper and other transition metals.25-28 Fundamental understanding of the energetics and mechanisms of fragmentation of peptide radical cations will provide important insights on a broad range of dissociation pathways of gas-phase biomolecules that involve radical chemistry. Gas-phase fragmentation of protonated peptide molecules has been extensively studied phenomenologically,29-33 and the energetics and dynamics of fragmentation have been determined for a number of model systems.34,35 However, little is known about the energetics of dissociation of odd-electron peptide ions. We have recently reported a first accurate measurement of the energetics and dynamics of dissociation of an R-radical (DRVG · IHPF+) for which the initial location of the radical site is well-defined.36 The R-radical was produced by in-source fragmentation of the [CuII(terpy)DRVYIHPF]2+ complex. We demonstrated that fragmentation of this ion is dominated by bond cleavages that are remote from the initial position of the radical site and that the dissociation rates of all of the decomposition channels observed in our surface-induced dissociation (SID) experiments were adequately described by the RRKM theory. Here, we present a comparison between the energetics and dynamics of dissociation of radical cations and protonated molecules of angiotensin III (RVYIHPF) and its analogue methylated at the C-terminus (RVYIHPF-OMe). Protonated species are produced using electrospray ionization (ESI), while radical cations are formed by in-source fragmentation of ternary
10.1021/jp805226x CCC: $40.75 2008 American Chemical Society Published on Web 09/09/2008
Dissociation of Peptide Radical Cations complexes of peptides with CoIII(salen)+ and CuII(terpy)2+ ions. Fragmentation energetics of mass-selected ions was studied using time- and collision energy-resolved SID experiments combined with RRKM modeling of the experimental data. We have previously shown that fast ion activation by collision with a surface combined with the long and variable time scale of a Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS) is perfectly suited for studying the energetics and dynamics of peptide fragmentation.34,35 In this study, we demonstrate for the first time that two different conformations of radical cations of angiotensin analogues separated by a high free energy barrier for isomerization are produced in our experiments. Dissociation of the different conformers is characterized by dramatically different threshold energies and activation entropies. The results are rationalized using density functional theory (DFT) calculations. Experimental Section SID experiments were conducted on a specially fabricated 6T FT-ICR mass spectrometer described in detail elsewhere.37 The SID target is introduced through a vacuum interlock assembly and is positioned at the rear trapping plate of the ICR cell. Ions are electrosprayed, at atmospheric pressure, and transferred into the vacuum system via an electrodynamic ion funnel.38 Two quadrupoles following the ion funnel provide collisional focusing and mass selection of the ion of interest. Collisional octopole held at elevated pressure (about (2-5) × 10-3 Torr) is used for accumulation of mass-selected ions and collisional relaxation of any internal energy possessed by ions generated by electrospray ionization prior to their injection into the ICR cell. Radical cations, M+ · , were produced through in-source fragmentation of the corresponding positively charged [CoIII(salen)M]+ or [CuII(terpy)M]2+ complexes, while their protonated counterparts were generated directly in the electrospray source. Mass-selected M+ · or [M + H]+ ions were accumulated for 0.2-5 s, extracted from the accumulation octopole, and transferred into the ICR cell. For SID experiments, ions were allowed to collide with the surface. Scattered ions were captured by raising the potentials on the front and rear trapping plates of the ICR cell by 10-20 V. Time-resolved mass spectra were acquired by varying the delay between the gated trapping and the excitation/detection event (the reaction delay) from 1 ms to 1 s. Immediately following the fragmentation delay, ions were excited through a broadband chirp and detected. The collision energy is defined by the difference between the potential applied to the accumulation quadrupole and the potential applied to the rear trapping plate and the SID target. The self-assembled monolayer surface of 1-dodecanethiol (HSAM) was prepared on a single gold {111} crystal (Monocrystals, Richmond Heights, OH) using a standard procedure. The target was cleaned in a UV cleaner (model 135500, Boekel Industries Inc., Feasterville, PA) for 10 min and allowed to stand in a solution of 1-dodecanethiol for 10 h. The target was removed from the SAM solution and washed ultrasonically in ethanol for 10 min to remove extra layers. Chemicals. All chemicals and reagents were commercially available (Sigma-Aldrich, St. Louis, MO; Bachem, King of Prussia, PA). Angiotensin III (RVYIHPF) was purchased from Sigma/Aldrich (St. Louis, MO). Angiotensin III methylated at the C-terminus (RVYIHPF-OMe) was synthesized via acidcatalyzed esterification of the C-terminal carboxyl group. A 2 M solution of hydrochloric acid in methanol was prepared by dropwise addition of 800 µL of acetyl chloride into 5 mL of
J. Phys. Chem. B, Vol. 112, No. 39, 2008 12469 anhydrous methanol followed by 5 min stirring at room temperature. Ten milligrams of Angiotensin III was dissolved in 1 mL of this solution, and the mixture was stirred for 3 h at room temperature. The resulting solution was dried using SC250DDA Speedvac Plus (Thermo Electron Corp., Waltham, MA).8e The methylated peptide was mixed with metal complexes in each experiment without further purification. The copper(II)-terpy complex (terpy ) 2,2′:6′,2′′-terpyridine) was synthesized according to the experimental procedure described by Henke et al.39 Synthesis of N,N′-ethylenebis(salicylideneaminato) and metal(III)-salen complexes [salen ) N,N′-ethylenebis(salicylideneaminato)] followed the procedure described elsewhere.40 Samples typically were comprised of 600 µM metal complex and 50 µM peptide in a 50:50 (v:v) water/ methanol solution. A syringe pump (Cole Parmer, Vernon Hills, IL) was used for direct infusion of the electrospray samples at flow rates ranging from 20 to 50 µL/h. Theoretical Calculations. Molecular mechanics modeling was performed on an SGI Onyx 3200 workstation running Insight II/Discover (97.0, Accelrys Software Inc., San Diego, CA). Preliminary structures of the neutral and protonated RVYIHPF were built using the Biopolymer builder of Insight II (Biosym Technologies, San Diego, CA). Both initial and final structures were energy minimized using the CFF91 force field41 and quasi-Newton-Raphson (VA09A) minimization algorithm.42 To randomly generate multiple conformations, molecular dynamics (MD) of the minimized structures was performed at 1000 K in a vacuum with a 1.0 fs time step. Conformations were saved at 2 ps intervals over a 1 ns dynamics run, and each structure was then minimized. Density functional theory (DFT) calculations were carried out using NWChem (version 5.0) developed and distributed by the Pacific Northwest National Laboratory (PNNL).43 Extensible Computational Chemistry Environment (ECCE) developed at PNNL44 was used to set up calculations and visualize the results. Preliminary geometry optimization was performed at the B3LYP/3-21G level of theory. Final geometries and single-point energies were obtained by subsequent optimization of these structures at the B3LYP/6-31G(d) level of theory. The barrier for the loss of p-quinomethide was obtained using the B3LYP/ 6-311++G(2d,2p)//B3LYP/6-31++G** level of theory. RRKM Modeling. Time-dependent survival curves (SCs) were constructed from experimental mass spectra by plotting the relative abundance of the corresponding ion as a function of collision energy for each delay time. SCs were modeled using an RRKM-based approach as described previously.45,46 (1) The microcanonical rate coefficient k(E) is calculated as a function of internal energy using the microcanonical RRKM/ QET expression.47 (2) The survival probability of the precursor ion as a function of the internal energy of the precursor ion and the experimental observation time (tr), F(E, tr), was calculated from the rateenergy k(E) dependency taking into account radiative decay of the excited ion population. (3) The internal energy deposition function was described by the following analytical expression:
(
1 (E - ∆) P(E,Ecoll) ) (E - ∆)l exp C f(Ecoll)
)
(1)
where l and ∆ are parameters, C ) Γ(l + 1)[f(Ecoll)]l+1 is a normalization factor, and f(Ecoll) has the form:
f(Ecoll) ) A2Ecoll2 + A1Ecoll + A0
(2)
where A0, A1, and A2 are parameters, and Ecoll is the collision energy.
12470 J. Phys. Chem. B, Vol. 112, No. 39, 2008
Yang et al.
Figure 1. Survival curves (SCs) for M+ · (filled symbols) and MH+ ions (open symbols) of RVYIHPF (triangles) and RVYIHPF-OMe (circles) on the HSAM surface at reaction delays of (a) 1 ms and (b) 1 s.
We have shown previously that this analytical form for the collisional energy deposition function has enough flexibility to reproduce experimental fragmentation efficiency curves obtained using both gas-phase collisional activation and SID.34b,d,45 (4) The overall signal intensity at a given collision energy, I(Ecoll), was obtained using eq 3: ∞
Ιi(Ecoll) )
∫ F(E,t)P(E,Ecoll) dE
(3)
0
Vibrational frequencies of the peptide were adopted from our previous study.48 Vibrational frequencies for the transition state were estimated by removing one C-N stretch (reaction coordinate) from the parent ion frequencies as well as scaling all frequencies in the range of 500-1000 cm-1 to obtain the best fit with experimental data. SCs were constructed using the above procedure and compared to the experimental data. The internal energy deposition function was determined by fitting the experimental SCs of the singly protonated RVYIHPF, for which the dissociation parameters are known from our previous study,48 and kept the same for all reaction times. The fitting parameters included the critical energy and activation entropy for dissociation of the precursor ion. They were varied until the best fit to experimental SCs was obtained. Results In this study, we compared the energetics and dynamics of unimolecular fragmentation of the singly protonated angiotensin III (RVYIHPF) and radical cations of angiotensin III produced by CID of ternary complexes of this peptide with CuII(terpy)2+ and CoIII(salen)+. It has been suggested that peptides containing basic arginine residue can be bound to the metal complex as zwitterions, in which the basic residue is protonated and the carboxylate group of the C-terminus is deprotonated.19,20 Alternatively, peptide can exist in its canonical form in the complex. It is plausible that different gas-phase structures of peptide radical cations could be generated depending on the structure and the mode of binding of the peptide to the metal complex. This question is addressed by comparing the stability of radical cations of angiotensin III produced by in-source fragmentation of ternary complexes between the peptide with CuII(terpy)2+ and CoIII(salen)+. Methylation of the C-terminal carboxylate group prevents the formation of the zwitterionic structure involving the deprotonated C-terminus of the peptide. In this article, we compared fragmentation of angiotensin III and its methylated form (RVYIHPF-OMe) to distinguish between zwitterionic and canonical structures of peptide cations. Time- and collision energy-resolved SID experiments were
Figure 2. SID spectra obtained at 1 s delay time for (a) [M + H]+ ion of RVYIHPF, 90 eV; (b) [M + H]+ ion of RVYIHPF-OMe, 90 eV; (c) M+ · ion of RVYIHPF, 50 eV; and (d) M+ · ion of RVYIHPF-OMe, 65 eV.
performed for singly protonated species and radical cations of angiontesin III, 1, and its analogue methylated at the C-terminus, 2, using HSAM as a target. Relative Stability and Fragmentation Behavior. Figure 1 compares the survival curves (SCs) obtained for radical cations and [M + H]+ ions of 1 and 2 at reaction delays of 1 ms (Figure 1a) and 1 s (Figure 1b). In general, protonated molecules are more stable toward fragmentation than the corresponding radical cations. The SCs obtained for M+ · ions of 1 and 2 are shifted by 48-52 eV and 31-40 eV, respectively, toward lower collision energies relative to the SCs of the corresponding [M + H]+ ions. Methylation does not affect the relative stability of the protonated molecule but increases the relative stability of the radical cation toward fragmentation. Fragmentation spectra obtained for [M + H]+ and M+ · ions of the same peptide are also very different. Representative SID spectra obtained for the four species at reaction delay of 1 s are shown in Figure 2. All spectra correspond to ca. 90% fragmentation efficiency of the precursor ion. Clearly, methylation has only a minor effect on the fragmentation behavior of these ions. SID spectra of [M + H]+ ions are dominated by N-terminal fragments resulting from backbone cleavages with the proton remaining on the arginine residue (e.g., bn, bn - NH3, an -
Dissociation of Peptide Radical Cations
J. Phys. Chem. B, Vol. 112, No. 39, 2008 12471
TABLE 1: Normalized Abundance of SID Fragments of the RVYIHPF and RVYIHPF-OMe Radical Cations Obtained at Collision Energies Corresponding to ca. 90% Fragmentation of the Precursor Ion and the Reaction Delay of 1 s RVYHPF (50 eV) RVYIHPF-OMe (65 eV) fragment
m/z
normalized abundance
a1 + [a2 - H]+ · a2 + a3 + [a4 - 106]+a [a5 - 106 - C4H8]+ z4 + · a4 + [a5-106]+ a 5+ [M - 106 - C3H8N3 · ]+ [M - 106 - C4H8]+ · [M - 106 - CO2]+ · [M - CO2 - C4H9N3]+ · [M - 106 - C2H5 · ]+ [M - CO2 - C3H8N3 · ]+ [M - 106]+ · [M - C3H8N3 · ]+ [M - COOH · ]+ [M - CO2]+ · [M - C2H5 · ]+ [M - NH3]+ · M+ ·
129.1140 227.1746 228.1824 391.2458 398.2880 479.2843 497.2638 504.3298 535.3469 641.3887 738.3940 768.4032 780.4760 787.4381 795.4267 800.4459 824.4658 844.4358 885.5099 886.5178 901.4685 913.481 930.5076
2.9 5.8 18.1 6.9 10.1 2.9 1.1 7.5 54.4 83.4 1.8 6.4 16.3 3.7 4.9 1.2 100.0 5.1 2.4 23.8 0.3 3.2 66.8
a
m/z
normalized abundance
752.4096 782.4188
4.5 14 50.9 13.5 35.3 13.4 1.8 7.5 100.0 58.9 11.0 7.9
809.4423
24.2
838.4814 858.4514
48.0 13.1
915.4841 927.4966 944.5232
3.6 3.4 49.0
511.2795
106 ) p-quinomethide (C7H6O, 106.0418).
NH3 ion), some internal fragments, and immonium ions.48 In contrast, very different fragmentation behavior is observed for the corresponding radical cations. Specifically, SID spectra of odd-electron ions are dominated by losses of small molecules from peptide side chains and the formation of an ions through CR-C bond cleavage.21 The most abundant neutral loss observed for both peptides corresponds to the loss of p-quinomethide (C7H6O, 106.0418) from the tyrosine side chain. Differences between fragmentation patterns of 1 and 2 are attributed to the presence or absence of the C-terminal carboxyl group. Specifically, while loss of CO2 is one of the major fragmentation pathways for the radical cation of 1, this channel and its sequential dissociation pathways are completely blocked by methylation of the carboxyl group. These observations are in agreement with the previous study by O’Hair and co-workers that compared the fragmentation behavior of radical cations of di- and tripeptides and their methyl ester analogues.49 Detailed comparison of SID spectra obtained for the radical cations of 1 and 2 is shown in Table 1. In addition to the loss of CO2 (for 1) and p-quinomethide (for 1 and 2), SID spectra display parallel and consecutive losses of C2H5 · (29.0391, Ile), C4H9N3 (99.0796, Arg), C3H8N3 · (86.0718, Arg), C4H8 (56.0626, Ile), and the C-terminal COOH · (for 1). CR-C bond cleavages are observed along the entire sequence excluding the bond between proline and phenylalanine. In addition, abundant [a5 - 106]+, [a4 - 106]+, and [a5 - 106 - C4H8]+ ions from [M - 106]+ · are observed in the spectra. Most of the backbone fragments are even-electron ions. However, two odd-electron backbone fragments, [a2 - H]+ · and z4+ · , are also present in the spectra. The latter is formed by N-CR bond cleavage of the tyrosine residue that is the most likely location of the radical site for both peptides.50 Collision energy-resolved fragmentation efficiency curves (FECs) obtained for major fragments of the radical cation of 1 are shown in Figure 3. Losses of CO2 and p-quinomethide (106) are the lowest-energy pathways. The a5 ion formation is
characterized by somewhat higher appearance energy, suggesting that this fragment could be formed either directly from the precursor ion or by consecutive fragmentation of the M+ · CO2 fragment ion. Fragments observed at higher collision energies are attributed to dissociation of the primary product ions that results in additional loss of 106 from the M+ · - CO2 fragment ion, and formation of the a5 - 106, a4 - 106, a3, a2, and M+ · - 106 - C2H5 · ions. Comparison of the collision energy-resolved data obtained for 1 and 2 is shown in Figure S1 of the Supporting Information. Methylation eliminates the formation of M+ · - CO2 and [M COOH]+ ions and their subsequent fragments. Perfect overlap between the M+ · - 106 curves obtained for both precursor ions suggests that the energetics and dynamics of this pathway is not affected by methylation. Formation of several secondary fragments including a2 - H, a2, a4 - 106, and a5 - 106 is not affected by methylation, suggesting that they are most likely formed by subsequent dissociation of the M+ · - 106 ion. In contrast, FECs of the a1, a3, and a4 ions are shifted toward lower energies for 1, suggesting that at least a fraction of these products is formed consecutively from the M+ · - CO2 fragment. The a5 fragment of 1 can be formed either directly from the precursor ion or from the M+ · - CO2 ion. In contrast, primary fragmentation of the M+ · ion is the only pathway for the formation of the a5 ion from the radical cation of 2. Because similar FECs were obtained for this fragment formed from both precursor ions, we conclude that the a5 ion is a primary fragment of the vibrationally excited M+ · precursor ion. TFECs of the precursor ion of 1 and its primary fragments, M+ · - CO2, M+ · - 106, and a5, are shown in Figure 4. Kinetically hindered loss of CO2 from the radical cation of 1 is consistent with the kinetics of the CO2 loss from the R-radical DRVG · IHPF reported in our recent study.36 Interestingly, the formation of M+ · - 106 and the a5 fragment from 1 follows much faster kinetics, suggesting that the loss of p-quinonethide and the CR-C bond cleavage do not require significant rearrangement of the precursor ion. RRKM Modeling Results. The experimental SCs were modeled using the approach described earlier. The decay of the precursor ion was modeled using two rate constants corresponding to the slow (statistical) decay and fast dissociation by shattering.51 The shattering component is modeled assuming that the ion fragments instantaneously after reaching certain threshold energy. Shattering was important only for modeling of the SCs obtained for protonated molecules that fragment at fairly high collision energies; fragmentation of radical cations studied in this work is not affected by shattering. Radiative decay of the internally excited precursor ion was taken into account in the modeling. The energy deposition function (EDF) was obtained from the best fit of the time-resolved SCs of the protonated RVYIHPF using the dissociation threshold of 1.62 eV reported in our previous study.48 Parameters obtained for [M + H]+ ions of 1 and 2 are summarized in Table 2. Very similar dissociation energies and activation entropies are obtained for the total decomposition of the protonated species. The efficiency of translational to vibrational energy transfer (TfV) for the HSAM surface of 10.9% obtained from the modeling is in good agreement with our previous study52 and results reported by other groups.53 SCs obtained for the radical cations of 1 and 2 were modeled using slightly broader EDF with similar most probable TfV transfer efficiency of 10.7%. Initially, the same two decay-rate model was used to calculate the breakdown graph for 1 and 2. However, we found that this model cannot adequately describe
12472 J. Phys. Chem. B, Vol. 112, No. 39, 2008
Yang et al.
Figure 3. Collision energy-resolved SID data for the most abundant fragment ions of the radical cation of RVYIHPF (1) at 1 s reaction delay.
Figure 4. TFECs obtained for the (a) M+ · , (b) M+ · - CO2, (c) M+ · - 106, and (d) a5 ions of RVYIHPF. The results are shown for delay times of 1 ms (9), 5 ms (O), 50 ms (4), and 1 s (b).
TABLE 2: Comparison of the Dissociation Parameters Obtained for [M + H]+ Ions of RVYIHPF and RVYIHPF-OMea E0, eV ∆Sq, cal/(mol K) A, s-1 Efast, eV TfV transfer (%) krad, s-1 DOF
RVYIHPF
RVYIHPF-OMe
1.62 -3.9 4 × 1012 11.7 10.9 19.3 396
1.64 -7.0 7 × 1011 11.5 10.9 20.1 405
a E0 is the threshold energy, ∆Sq is the entropy change for the transition state, A is the pre-exponential factor at 450 K, Efast is the threshold for shattering, TfV transfer (%) is the percentage of the ion’s kinetic energy converted to internal energy upon collision, krad is the radiative decay rate, and DOF is the number of vibrational degrees of freedom.
the experimental results obtained at different reaction times and over a range of collision energies. Introducing additional decay channels with very different dissociation parameters did not improve the quality of the fit. Clearly, the experimental data cannot be described using the model in which several products are formed from a single precursor. We then constructed a twostate model54 shown in Figure 5, in which the reactant exists in two isomeric forms with a significant isomerization barrier. Figure 6 shows the results of the RRKM modeling of timeresolved SCs for the radical cations of 1 and 2 using the twostate model. Good agreement between experimental and simulated results suggests that both radical cations exist in two
Figure 5. Schematic drawing of the potential energy surface describing the two-state reaction.
distinctly different isomeric forms. Furthermore, we found that isomerization rates are much slower than dissociation rates, suggesting that there is a significant free energy barrier for isomerization. It follows that the two populations of structures of M+ · ions of 1 and 2 do not interconvert at internal energies and reaction delays times sampled in our SID experiments. The fraction of ions contributing to each population and the corresponding dissociation parameters are summarized in Table 3. It was not possible to unambiguously determine the difference in the heats of formation, ∆∆H, of the structures contributing to the two populations based on the modeling of the experimental data. Threshold energies listed in Table 3 were obtained assuming ∆∆H ) 0. It should be noted that qualitative trends
Dissociation of Peptide Radical Cations
J. Phys. Chem. B, Vol. 112, No. 39, 2008 12473
Figure 6. RRKM modeling of the time- and energy-resolved survival curves for the radical cations of (a) RVYIHPF and (b) RVYIHPF-OMe for reaction delays of 1 ms (9), 5 ms ([), 10 ms (O), 50 ms (+), 100 ms (2), and 1 s (0).
TABLE 3: Dissociation Parameters for Radical Cations of RVYIHPF and RVYIHPF-OMe precursor complex
[CoIII(salen)M]+ M ) RVYIHPFOMe
[CuII(terpy)M]2+ M ) RVYIHPF
peptide
RVYIHPF
% ions E0, eV ∆S q, eua A
64 1.02 -9.5 2 × 1011
Population 1 47 1.16 0.5 3 × 1013
69 0.99 -6.3 1 × 1012
% ions E0, eV ∆S q, eu A
36 2.3 54 1 × 1025
Population 2 53 2.4 55 2 × 1025
31 2.1 54 1 × 1025
a eu ( entropy unit ( cal/(mol K); activation entropies and pre-exponential factors at 450 K. The estimated uncertainties are (7% for the threshold energies and (3 eu for the activation entropies.
in threshold energies and activation entropies discussed below are not significantly affected by this assumption. The difference between threshold energies and activation entropies obtained for the two populations of ions is striking; fragmentation of the population 1 is associated with low threshold energies and negative activation entropies, while dissociation of the population 2 is characterized by a high threshold and large positive activation entropy with the corresponding pre-exponential factor greater than 1025 s-1. Large preexponential factors obtained for dissociation of population 2 are indicative of loosening of numerous hydrogen bonds in the course of fragmentation. In contrast, 12-15 orders of magnitude lower pre-exponential factors and slightly negative entropy effects characterize fragmentation of population 1. Similar dissociation parameters and somewhat different abundances of the two populations were obtained for M+ · ions of 1 produced from [CoIII(salen)M]+ or [CuII(terpy)M]2+ complexes. Theoretical Results. DFT calculations were used to examine possible structures of neutral molecules and radical cations of RVYIHPF (1). Because the force fields for odd-electron species do not exist, the starting structures for DFT calculations were obtained from MD simulations of the neutral (both canonical and zwitterionic) and protonated peptide mimicking the canonical and the zwitterionic structures of the radical cation. Twelve low-energy structures representing the most populated conformations of the neutral and protonated species of 1 were selected out of 2000 optimized structures. Structures of radical cations were obtained by removing an electron for a starting neutral
structure or a hydrogen atom from the protonated species. Subsequent DFT optimization was performed at the B3LYP/ 3-21G level of theory followed by optimization at the B3LYP/ 6-31G(d) level of theory of the conformations representing the lowest-energy structures of the canonical, zwitterionic, bizwitterionic, and capatodative radical cations discussed later in this article. Because of the complexity of the system and the lack of force field potentials for odd-electron species, it is impossible and impractical to fully explore the potential energy surface for the radical cation. As a result, the structures of radical cations obtained from DFT calculations may not represent the lowestenergy species. However, we are confident that we obtained reasonably low-energy representative structures that are important for qualitative discussion of the experimental results. Depending on the distance between the arginine (R) and tyrosine (Y) residues, low-energy conformers of the neutral peptide shown in Figure S2 of the Supporting Information can be generally grouped into three categories characterized by small (1.9-2.6 Å), intermediate (4.0-8.7 Å), and large (>10 Å) separation between the side chains of these residues. The most stable structures of the neutral peptide were obtained for small and large separation between R and Y. Both low-energy structures are zwitterions. When R and Y residues are close to each other, the zwitterionic structure is formed by proton transfer from Y to R and stabilized by strong hydrogen bonding between the side chains of these two residues, the C-terminal carboxyl group, and the carbonyl group of the last amide bond. The second stable zwitterionic structure is formed by deprotonation of the C-terminal carboxyl group and stabilized by a strong interaction between the COO- and the protonated guanidine group of R. In addition, a strong hydrogen bond is formed between the imidazole ring of histidine and the phenolic hydrogen of Y. The third canonical family of structures with intermediate separation between R and Y is about 15 kcal/mol less stable than the zwitterionic species. Representative low-energy structures of the radical cation of 1 are shown in Figure 7. Ionization of structures with short distance between Y and R results in the formation of a zwitterionic radical cation (a) with the charge localized on R and the spin density on Y. When a canonical neutral species is selected as a starting structure, ionization is followed by a spontaneous proton transfer from the hydroxyl group of phenol to the guanidino group of arginine and formation of a zwitterion. Structure (a) is stabilized by hydrogen bonding between the protonated arginine side chain, the C-terminal carboxyl group, and the phenol ring. Another 2.7 kcal/mol more stable zwitterionic structure (b) is obtained when the protonated guanidino group is solvated by 5 carbonyl oxygens and the N-terminal
12474 J. Phys. Chem. B, Vol. 112, No. 39, 2008
Yang et al.
Figure 7. B3LYP/6-31G(d) optimized low-energy structures of the two zwitterionic structures of the radical cation of RVYIHPF with (a) short and (b) intermediate distance between R and Y; (c) double zwitterionic structure; (d) canonical structure; and (e) captodatively stabilized N-terminal R-radical. Peptide backbone is shown as a colored thick line; side chains are shown as thin lines. Relative energies do not include the ZPE correction.
amino group, while the phenol ring interacts with the C-terminal carboxyl group. Figure 7c shows an interesting structure of the radical cation that is formed by ionization of neutral species, in which Y is located close to the histidine residue. In this bizwitterionic structure, the deprotonated phenyl ring interacts with the protonated imidazole ring, while the deprotonated C-terminus interacts with the protonated guanidino group. Canonical structures of the radical cation of 1 are generally 30-60 kcal/mol less stable than the zwitterionic structures. A representative canonical structure obtained from DFT calculations is shown in Figure 7d. In this structure, both the charge and the radical are localized on the guanidino group of R. This group is solvated by three backbone carbonyl groups and hydrogen bonded to the imidazole ring. The COOH group forms a strong hydrogen bond with the last backbone carbonyl oxygen and with the imidazole ring. The most stable canonical structure is obtained by hydrogen transfer from the N-terminal R-carbon to the guanidino group. The stability of the N-terminal R-radical is attributed to the strong captodative stabilization of the radical site by the electron-donating NH2 group and the electronwithdrawing CONH group.55-57 The relative stability of the captodatively stabilized canonical structure shown in Figure 7e is similar to the stability of the low-energy zwitterionic structures. Discussion Our results demonstrate that both the energetics and the mechanisms of dissociation of even-electron and odd-electron angiotensin ions are quite different. Specifically, protonated molecules are much more stable toward fragmentation than the corresponding radical cations. Because dissociation of the protonated angiotensin analogs is driven by the basic arginine residue, methylation has no effect on the energetics, dynamics, and fragmentation pathways observed for these species. In contrast, methylation of the C-terminus eliminates one of the major dissociation channels of the radical cation that undergoes a facile loss of the CO2 group followed by sequential fragmentation of the M+ · - CO2 fragment ion. As a result, the methylated radical cation is energetically more stable than its unmodified counterpart. Finally, we found that while fragmentation of protonated species is adequately described using a single population of vibrationally excited ions, SCs obtained for peptide radical cations could be modeled only using two nonintercon-
verting populations of isomeric structures. Such bimodal behavior has never been observed in SID studies of evenelectron peptide ions and is particularly surprising for ions produced by in-source fragmentation that facilitates proton and hydrogen atom migration through energetic collisions with the background gas. Energetics and Dynamics of Dissociation. In our previous study of SID of protonated angiotensin analogues, we found that in the absence of acidic residues gas-phase fragmentation of these ions was characterized by fairly high threshold energies.48 The high thresholds (>1.6 eV) determined in our experiments were qualitatively rationalized using the “mobile proton” model58-61 according to which the dissociation energy of singly protonated peptides increases when arginine is present in the sequence. We suggested that the principal mechanism of fragmentation of angiotensin analogues involves loss of NH3 from the protonated precursor followed by fast mobilization of the ionizing proton and facile fragmentation of the primary MH+ - NH3 fragment ion.49 The primary fragmentation pathway requires proton transfer to the guanidino group of the arginine residue. It follows that the initial proton transfer step in dissociation of angiotensin III analogues is associated with ca. 1.6 eV threshold energy. In this study, we examined the effect of C-terminal methylation on the stability of the radical cation of angiotensin III and its protonated form. Identical dissociation parameters obtained for protonated 1 and 2 (Table 2) suggest that the C-terminal carboxyl group is not involved in dissociation of RVYIHPF. Similarly, dissociation of population 2 of the radical cation (Table 3) is not affected by methylation. In contrast, methylation of the C-terminal carboxyl group has a measurable effect both on the energetics and on the dynamics of dissociation of the radical cation. Specifically, methylation eliminates the loss of the CO2 from the radical cation, the reaction channel characterized by very slow kinetics and low dissociation threshold (Figure 4). As a result, the threshold energy for the total decomposition of the radical cation increases by 0.14 eV and the activation entropy increases by 10 eu upon methylation (Table 3). Low threshold energy of 1.02 eV obtained for the radical cation of 1 mainly reflects the barrier height for the loss of CO2 from this ion. This value is only ca. 0.2 eV lower than the barrier for the loss of CO2 from a much smaller GW+ · ion reported by Siu and co-workers.56
Dissociation of Peptide Radical Cations
Figure 8. Potential energy surface for the elimination of p-quinomethide (C7H6O, 106.0418) from the radical tyrosine side chain obtained at the B3LYP/6-311++G(2d,2p)//B3LYP/6-31++G** level of theory with the zero-point energy (ZPE) correction. The values are in kcal/ mol.
Gas-phase fragmentation of peptide radical cations can be initiated either by the radical site or by the charge site.19,20,56,62 It has been demonstrated that similar fragmentation patterns are obtained for even- and odd-electron ions when dissociation of odd-electron species is initiated by the proton transfer processes. In contrast, radical-driven fragmentation usually produces different types of fragments including facile side chain losses that are rarely observed for protonated peptides.8,19-21,49,56 Different fragmentation behavior observed for even- and oddelectron ions of 1 and 2 (Figure 2) indicates that dissociation of odd-electron ions studied in this work is most likely driven by the radical site. Comparison between the energetics and dynamics of fragmentation of population 1 of radical cations and the protonated species of 1 and 2 shown in Tables 2 and 3 provides further support for this conclusion. Specifically, it demonstrates that lower stability of the radical cations mainly results from lower dissociation thresholds and not from changes in activation entropies. The threshold energy is 0.6 and 0.48 eV lower for radical cations of 1 and 2 as compared to the corresponding protonated species, while activation entropies are quite similar. Because dissociation energies obtained for population 1 of radical cations are much lower than the threshold energy of ca. 1.6 eV required for proton mobilization from the arginine side chain of 1 and 2, we conclude that fragmentation of these M+ · ions is not driven by the mobile proton. Because the loss of p-quinomethide from the radical cation of 2 is the major primary decomposition pathway, the dissociation threshold of 1.16 eV obtained from the RRKM modeling of the corresponding survival curves largely reflects the barrier for this reaction channel. DFT calculations were performed for a model radical of N-methyl tyrosine methylamide (NMTMA) representing a small portion of the peptide backbone with the tyrosine side chain. The potential energy surface for the loss of p-quinomethide from NMTMA is shown in Figure 8. The reaction barrier of 19.3 kcal/mol (0.84 eV) is slightly lower than the dissociation threshold for the loss of p-quinomethide from the radical cation of 2 determined in this study. This is not surprising because fragmentation of a peptide ion requires breaking of several hydrogen bonds resulting in a somewhat higher dissociation threshold. In contrast, a much higher dissociation barrier of 58.2 kcal/mol (2.52 eV) was obtained for the loss of p-quinomethide from the deprotonated NMTMA,
J. Phys. Chem. B, Vol. 112, No. 39, 2008 12475 suggesting that the 106 loss is predominantly observed from distonic structures, in which the radical site is located on the tyrosine side chain. The activation entropy for the loss of p-quinomethide from NMTMA of -3 eu at 450 K was obtained using the calculated vibrational frequencies of the radical and the transition state. The calculated entropy effect is in good agreement with the experimental value of 0.5 eu (Table 3) obtained for the radical cation of 2. It should be noted that the correspondence between the threshold energy and the activation entropy for the loss of p-quinomethide from the neutral radical of NMTMA and from the radical cation of 2 is consistent with the charge-remote radical-driven nature of this dissociation pathway proposed in our recent study.21 Fragmentation of the second population characterized by very high threshold energy and activation entropy becomes significant at collision energies above 40 eV. FECs of several abundant fragments including a4 - 106, a3, and a2 ions formed at collision energies above 40 eV do not show any time dependence. These fragments could be ascribed to the kinetically favored fragmentation of population 2, for which no time dependence can be observed on the time scale of our FT-ICR experiment. However, these product ions could be also formed by consecutive fragmentation of the M+ · - 106 primary fragment. It follows that competition between consecutive fragmentation of primary fragments and formation of primary products of population 2 makes it difficult to unambiguously determine characteristic fragments of the fast fragmenting population of radical cations. The Origin of the Two Populations. It is interesting to discuss the origin of the two noninterconverting populations of the angiotensin III radical cations. In our recent study, we demonstrated that the presence of the tyrosine residue is essential for the formation of the radical cation from ternary complexes of angiotensin analogues with [CoIII(salen)]+. Substitution of the tyrosine (Y) with glycine in the RVIYHPF sequence reduces the yield of the radical cation formation by a factor of 200. We suggested that Y is involved in the binding between the peptide and the [CoIII(salen)]+ complex and that the radical cation is mainly formed by the electron transfer from the tyrosine side chain axially coordinated to the metal center. It follows that for this class of peptides the radical site is initially located on the phenol ring of tyrosine. Several groups suggested two possible modes of binding of neutral peptides in ternary transition metal complexes: one, in which the canonical form of the neutral peptide molecule is bound to the metal complex, and another one, in which peptide exists in its zwitterionic form and is bound to the metal complex via the deprotonated C-terminal carboxyl group. Electron transfer from the deprotonated carboxyl group to the metal complex results in formation of the radical cation with the radical located on the C-terminus followed by fast fragmentation of the unstable carboxyl radical.19,20,62 This pathway results in formation of metastable radical cations recently described by Siu and co-workers.63 Specifically, they demonstrated the formation of two isomeric structures of radical cations of histidine, lysine, and arginine formed by CID of several [CuII(L)(amino acid)]2+ complexes. The type I ions were stable on the time scale of the experiment, while the type II ions were metastable and fragmented by the loss of CO2 prior to detection. The two types of isomeric structures are produced by ionization of the canonical and zwitterionic forms of the amino acid following electron transfer to the metal complex. The zwitterionic form of the amino acid, in which the C-terminal carboxyl group is deprotonated and the side chain of the basic amino acid is protonated, was responsible for the metastable structures
12476 J. Phys. Chem. B, Vol. 112, No. 39, 2008 of radical cations, while the stable ions were attributed to the canonical form of the amino acid. It was also reported that the branching ratio of the two types of structures was dependent on the auxiliary ligand. In this work, we examined the role of the C-terminal carboxyl group on the structures and energetics of the radical cations of angiotensin III by comparing fragmentation of RVYIHPF and its analogue methylated at the C-terminus. Methylation of the C-terminal carboxyl group prevents the formation of zwitterionic peptide structures involving the deprotonated C-terminus. However, we found that the presence of the two populations of isomeric structures of angiotensin radical cations is not affected by methylation, suggesting that the isomeric families of structures observed in our experiments are not produced from the zwitterionic species with deprotonated C-terminus. Metastable radical cations described by Siu and co-workers63 are most likely produced in the ion source of our instrument, but because they are very fragile they fragment prior to entering the massresolving quadrupole and are not transmitted through the quadrupole mass filter. In our experiments, we did not observe any M - CO2 fragments originating from the accumulation stage, suggesting that metastable ions are either stabilized by collisions with the background gas or fragment very quickly before they reach the quadrupole mass filter. Kinetically hindered loss of CO2 shown in Figure 4b is also inconsistent with the presence of metastable ions because their fragmentation should follow fast kinetics. Zwitterionic structures of angiotensin peptides can be also formed as a result of proton transfer between the acidic tyrosine and the basic arginine side chain.8a,20,27,64 Phenolic hydrogen is the second most acidic site in these peptides; the gas-phase acidity of phenol is only 0.5-1 kcal/mol higher than the gasphase acidity of acetic acid.65 While it is commonly assumed that peptide is bound through the deprotonated tyrosine side chain to the metal complex, DFT calculations summarized earlier demonstrated that ionization of 1 results in spontaneous transfer of the acidic proton of the tyrosine side chain to the basic arginine or histidine residues, suggesting that deprotonation of the tyrosine residue can occur following the electron transfer from the peptide to the metal complex. It is interesting to note that oxidation of tyrosine in biological systems is accompanied by loss of a proton. This process, a key step in the activation of Photosystem II toward water oxidation,66 is attributed to a dramatic increase in the acidity of phenolic hydrogen following oxidation. The pKa value for tyrosine changes from 10 for the neutral molecule to pKa ) -2 for the radical. Theoretical calculations of the structures of the neutral RVYIHPF and its radical cation confirm that proton-coupled electron transfer also occurs in the gas phase. Computational results shown in Figure 8 suggest that loss of p-quinomethide is a facile low-energy dissociation pathway for odd-electron peptide ions with the radical site delocalized over the phenol ring of the tyrosine side chain. DFT calculations of the structures of the radical cation of 1 demonstrated that for all zwitterionic structures of 1 the charge is localized on R and the spin density on Y. It follows that all possible zwitterionic structures of 1 should readily lose the tyrosine side chain with the threshold energy comparable to the dissociation threshold of population 1 derived from the experimental data. In contrast, for canonical radical cations shown in Figure 7, the radical site is localized on the guanidino group for structure (d) and on the R-carbon of the first residue for the captodative radical (e). Loss of p-quinomethide from canonical radicals requires abstraction of the phenolic hydrogen followed by the CR-Cβ bond cleavage.
Yang et al. Clearly, this dissociation pathway is a higher-energy process for canonical radicals as compared to zwitterionic structures of 1. Finally, we note that only one population of structures was observed in our previous study of the fragmentation of the DRVG · IHPF+ R-radical cation36 produced by the loss of p-quinomethide (106) from the radical cation of DRVYIHPF, while bimodal SCs were observed for the radical cation of DRVYIHPF (unpublished results). This can be rationalized assuming that loss of the tyrosine side chain occurs only from one family of structures and not from another. Because fragmentation of population 2 characterized by a high threshold energy and high positive entropy becomes important only at collision energies in excess of 35 eV while loss of pquinomethide occurs at much lower collision energies (