Probing Protein Structure and Folding in the Gas Phase by Electron Capture Dissociation
The established methods for the study of atom-detailed protein structure in the condensed phases, X-ray crystallography and nuclear magnetic resonance spectroscopy, have recently been complemented by new techniques by which nearly or fully desolvated protein structures are probed in gas-phase experiments. Electron capture dissociation (ECD) is unique among these as it provides residue-specific, although indirect, structural information. In this Critical Insight article, we discuss the development of ECD for the structural probing of gaseous protein ions, its potential, and limitations.
KeywordsProtein structure Folding Gas phase Electron capture dissociation Mass spectrometry
Importantly, collisional activation or exposure to infrared (IR) photons of (M + 8H)7+• ions of ubiquitin gave no b and y fragments, which are typical products of slow (M + nH)n+ ion heating, but instead, c and z• fragments characteristic for ECD, whose site-specific relative abundances were substantially different from those in the ECD spectrum of (M + 8H)8+ ions . Vibrational activation of (M + nH)(n-1)+• ions can also yield c• and z fragments, presumably from hydrogen transfer between two complementary c and z• fragments that are still held together by noncovalent bonding [5, 6, 7]; c• and z ions generally disappear when the (M + nH)n+ ions are unfolded prior to or during reaction with electrons [7, 8]. In yet another study, ~90% dissociation of ubiquitin (M + 7H)6+• ions into c and z• (and c• and z) fragments was obtained by vibrational excitation at far lower energies (~10%) than those required for collisionally activated dissociation (CAD) or infrared multiphoton dissociation (IRMPD) of (M + 7H)7+ ions into b and y fragments, consistent with breaking of noncovalent instead of covalent bonds . This was interpreted as evidence for ECD data reflecting protein structure in the gas phase, in that the observation of separated c and z• fragments indicates the lack of noncovalent bonding between each other . Ever since, the unusual capability of ECD to break covalent N–Cα bonds while maintaining noncovalent bonds that account for the higher order structure of gaseous ions has been used to obtain information about peptide or protein structure in the absence of solvent [7, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25]. More recently, this approach has been extended to the study of noncovalently bound protein complexes [18, 26, 27, 28, 29, 30, 31].
Formation of ECD Fragments
Excess Energy in ECD
Is it possible that any excess energy in ECD [i.e., the energy from cation-electron charge recombination (CR) that is not consumed by N–Cα bond cleavage] causes secondary dissociation of the radical z• ions, or even changes in the higher order structure of the protein ions? At room temperature, ECD of linear proteins generally produces very similar yields of c and z• ions (Figure 1b), which shows that any excess energy must be significantly smaller than that required for secondary dissociation of the radical z• ions. For peptides and proteins that comprise cyclic backbone or disulfide bonded structures, secondary radical reactions can, nonetheless, proceed at room temperature [3, 20, 41], most likely driven by the energy released by breaking up the strained cyclic structures.
For energy minimized forms of the ubiquitin crystal structure in the absence of solvent, Williams and coworkers have conservatively estimated that capture of one and two electrons, with assumed CR energies of 6 and 12 eV, is associated with a temperature increase from ~20 to ~85 and ~140°C, respectively . Calculated CR energies for small (~130 to 1115 Da, 18 to 172 atoms) linear peptide ions with m/z values between ~100 and 555 and up to three charges did not exceed 6.6 eV, and generally increased with increasing charge . For example, the adiabatic CR energy for (AHDAL +2H)2+ ions with m/z 264, 5.53 eV, was about twice as high as that for (AHDAL + H)+ ions with m/z 526, 2.93 eV. Vertical CR energies were even smaller by an average factor of 0.8, in good agreement with the CR energy of 4.3 eV for (KYK + 2H)2+ ions at m/z 220 determined in nanocalorimetry experiments .
Furthermore, can we rationalize the observed linear correlation of CR energy with charge divided by the cube root of the number of atoms (Figure 2a)? Because ion volume should be roughly proportional to the number of atoms, the cube root of the number of atoms can be interpreted as the average value of the ion’s three dimensions. The linear correlation of CR energy with charge divided by the cube root of the number of atoms shown in Figure 2a is thus consistent with simple Coulomb interactions dominating cation-electron charge recombination in roughly spherical ion structures. From the fit in Figure 2a, adiabatic CR energies between 5.1 and 12.0 eV can be estimated for ubiquitin (M + nH)n+ ions with n = 5–13, which corresponds to ~1–3 meV or ~0.1–0.3 kJ/mol per vibrational degree of freedom (DOF, calculated as three times the number of atoms minus 6) as shown in Figure 2b. Nevertheless, this estimate for protein ion CR energies is far higher than that of 4–7 eV by Zubarev  who used a thermodynamic cycle that comprised apparent proton affinities of (M + nH)n+ ions .
However large or small, the CR energies may not be fully available for potential ion heating as part of it can be consumed by backbone cleavage. N–Cα bond dissociation energies between –0.92 and 1.63 eV were calculated for small radical peptide model systems , indicating that the excess energy in ECD can be substantially lower than the energy from charge recombination. Moreover, even the exoergic reactions with negative bond dissociation energies can be slowed by barriers, for which transition state energies between 0.02 and 0.99 eV were calculated . Computing CR and dissociation energies for peptide and protein (M + nH)(n-1)+• ions remains nonetheless challenging, not least because the exact charge locations and structures of most gaseous peptides and essentially all protein ions are unknown.
To illustrate the intricate interplay of charge location and structure in gaseous ions, three examples are given here. First, even for compounds as simple as para-aminobenzoic acid methyl ester, the preferred protonation site changes from the carbonyl oxygen to the amine nitrogen upon relatively small changes in chemical environment (i.e., by increasing the number of attached water molecules from 2 to 3 ). Translating this scenario to gaseous peptide or protein ions, even relatively small local structural changes such as sidechain reorientations, especially those involving hydrogen bonds, are likely to significantly affect the proton affinities of individual sites and, therefore, the CR energies. Second, ion mobility spectrometry (IMS) studies showed that simple crown ether binding can, depending on ion net charge, decrease or increase collision cross-sections of gaseous protein ions by more than 30% , indicating major structural changes caused by minor local binding events. Third, positioning of negative and positive charge on the N- and C-terminus of polyalanine peptides, respectively, stabilizes α-helical structures through favorable charge–helix dipole interactions, whereas polyalanine peptides with opposite charge locations adopt globular structures . The above findings exemplify the complexity of gaseous protein ions structures , which is further increased by the possibility of kinetic trapping of less stable structures from electrospray ionization (ESI) .
McLuckey and coworkers have recently investigated the effect of recombination energy by comparing ECD spectra of small peptides to those from electron transfer dissociation (ETD), in which the recombination energy should be smaller than in ECD by the electron affinity of the reagent used, 0.6 eV for azobenzene and 1.7 eV for 1,3-dinitrobenzene . Data interpretation was complicated by the fact that products from sequential electron transfer and H• loss could not be distinguished from products from competitive proton transfer. However, the partitioning among the various c, z• ion channels was remarkably similar in ECD and ETD for most of the model peptides studied, which would argue against any effect of charge recombination energy on peptide ion structure prior to backbone cleavage into c and z• ions.
Radiative cooling by emission of infrared photons can be a complex process with rates that generally increase with increasing internal energy of the ions (i.e., hotter ions cool faster). For example, Al4– clusters showed a decrease in radiation intensity from 40 eV/s at 1127°C to 1 eV/s at 227°C . Nevertheless, Dunbar pointed out that “often it will be a reasonable approximation to think of the cooling as exponential” such that the internal energy decay of a population of superthermal ions can be described in terms of first-order cooling rate constants . For small non-halogenated organic ions, infrared radiative cooling rate constants between 0.5 and 15 s–1 at energies between 0.3 and 4 eV have been measured , and those for singly protonated pentapeptide leucine enkephalin ions were deduced from experimental data as 30 s–1 at 165°C  and 7.5 ± 0.5 s–1 at ~25°C . The internal energy of (M + H)+ ions of leucine enkephalin at 25°C is ~1 eV , which can be increased to ~3 eV (at which the cooling rate constant of 7.5 ± 0.5 s–1 was determined)  in tandem MS experiments to overcome the activation energy of 1.09 ± 0.06 eV  for dissociation into typical products from slow ion heating  within a few milliseconds . From the fit in Figure 2a, CR energies of 2.90 and 5.09 eV can be estimated for electron attachment to (M + H)+ and (M + 2H)2+ ions of leucine enkephalin, respectively, which should be sufficient for dissociation into b and y fragments on a milliseconds timescale. Whereas the uncharged products from ECD of (M + H)+ ions cannot be directly detected, ECD of (M + 2H)2+ ions of leucine enkephalin in fact produced no c or z• but instead a, b, y and w ions . It has been suggested that b ions form after loss of H• from “hot” (M + 2H)+• ions, the probability for which depends on peptide composition, structure, and internal energy . Nevertheless, the formation of b ions in ECD of proteins and most peptides is rare, consistent with nanocalorimetry ECD experiments of partially hydrated (KYK + 2H)2+ ions, which gave both c and b ions, but c ion formation required only ~14% of the energy required for b ion formation .
Conversion of the CR energy, or parts of it, into IR radiation of course requires prior conversion into vibrational energy by intramolecular vibrational redistribution (IVR). It is generally agreed that electron capture involves higher-energy Rydberg states that undergo fast relaxation to lower energy Rydberg and then non-Rydberg (n < 3) electronic states , but whether this involves radiative or nonradiative relaxation by IVR, or both, is still unclear. Moreover, nothing is as yet known about radiative cooling of peptide or protein (M + nH)n+• ions by emission of higher-energy photons, but the possibility that radiation from electronically excited states cools the radical ions formed by electron capture cannot generally be excluded . Despite the gain in energy from charge recombination, the protein (M + nH)(n-1)+• ions from electron capture are sufficiently stable for isolation and subsequent dissociation into c and z• (or c• and z) fragments by IR laser radiation [9, 11, 20] at energies that are far lower than those required to dissociate (M + nH)n+ ions into b and y fragments. Possibly the radical (M + nH)(n-1)+• ions formed by electron attachment into a Rydberg orbital with subsequent transfer to π* or σ* orbitals  also cool by emission of higher-energy photons from electronically excited states, which raises the “old” question if ECD can be described as a nonergodic process or if it is better described as a statistical process [1, 73, 74]. Anyhow, when cooling rates are significantly faster than the rates of unfolding, any excess energy from charge recombination should have little effect on the protein structures probed by ECD, but this may not be the case for small peptides. For example, Turecek and coworkers have calculated a temperature of ~580°C for a doubly charged pentapeptide after electron capture without radiative relaxation .
Probing Protein Ion Structure
It is important to note that CAD differs markedly from ECD in this respect. For example, Clemmer and coworkers selected different conformers of ubiquitin (M + nH)n+ ions with n = 8–10 by ion mobility for subsequent dissociation by CAD, and found that the fragment mass spectra for different conformers of the same charge were identical within experimental error, which they attributed to rearrangement into structurally similar transition states prior to dissociation into b and y fragment ions . Because CAD relies solely on vibrational excitation that incrementally increases an ion’s internal energy until dissociation occurs, it is reasonable that noncovalent bonds are broken before covalent bonds, unless the latter are weaker than the former.
Apart from higher order structure, the site-specific c, z• ion yields from ECD are also affected by how positive charge is distributed and locally solvated within the protein ions , which gives rise to irregular (“jumpy”) features in plots of site-specific yield versus cleavage site, even when the protein ions are unfolded [9, 12, 19]. This is similar to steric and inductive effects of amino acid sidechains on hydrogen/deuterium exchange of protein backbone amides in solution, for which scalar correction factors have been established . Delineating the effects of charge location and solvation in ECD is less straightforward, but it was recently shown that the “jumpy” features are due to distinct patterns termed charge site spectra  that can be retained over a wide range of precursor ion charge states for a given protein sequence. Finally, because ECD requires the presence of positive charge, a lack of c and z• fragments can also be observed in protein regions that lack any basic residues, or in protein regions that are bridged by disulfide bonds .
In any case, structural probing by ECD is indirect, meaning that it relies on observing separated c and z• fragments from protein backbone cleavage to reveal the absence of noncovalent bonding between them. Information about the number, types, or strengths of intramolecular interactions that prevent fragment ion separation cannot directly be inferred from a single ECD spectrum. In other words, structural probing by ECD provides an imprint of unfolded structure that shows which regions are unfolded, but not how and to what extent the other regions are folded. This is similar to footprinting techniques in solution in which folded regions can be shielded from, for example, hydroxyl radical attack, but with the major difference that in ECD, a single noncovalent interaction between two non-adjacent residues could prevent separation of all c and z• fragments from backbone cleavage in between these residues, no matter how compact that region is. Nevertheless, ECD provides valuable structural information, especially when data obtained under different experimental conditions are compared with each other (Figure 5), and when it is used to probe protein unfolding or folding [9, 10, 21, 80].
Monitoring Protein Unfolding
Proteins can unfold for a number of different reasons, including solution pH, the presence of organic solvent or “denaturants” such as guanidinium hydrochloride, and elevated temperatures. As evidenced by ECD, the mere removal of solvent during ESI causes spontaneous unfolding of horse and tuna heart cytochromes c and ubiquitin, all of which have highly stable “native” folds in solution [21, 80]. The extent of unfolding in different regions of these proteins, however, was found to depend on the extent of stabilization by electrostatic interactions. Remarkably, the (M + 7H)7+ ions of the three-helix bundle protein KIX were sufficiently stabilized to fully preserve the native fold on a timescale of up to 4 s after transfer into the gas phase, and a gradual loss of higher order structure by sequential unraveling of the helices from their terminal ends was observed with increasing charge of the (M + nH)n+ ions . Thus, ECD can be used to monitor unfolding in solution, or to probe structural changes that occur during the transition from solution into the gas phase, but distinguishing between these two possibilities is not straightforward when the solution structure is not known.
Regardless of whether a compact protein ion structure resembles the solution fold or not, its thermal unfolding in the gas phase can be monitored by ECD. For ubiquitin (M + 6H)6+ to (M + 9H)9+ ions that were thermally equilibrated by blackbody infrared radiation for at least 40 s, the yield of separated c, z• fragments steadily increased with increasing temperature from 25 to 175°C . Importantly, a van’t Hoff analysis of the ECD data revealed a three-state process for ubiquitin (M + nH)n+ ion unfolding in the gas phase, with ΔH values generally decreasing with increasing ion net charge. However, the ΔH values for n = 7 were similar to those for n = 6 at temperatures of up to 100°C, and similar to those for n = 8 at higher temperatures, showing a stepwise decrease in ΔH from ~30 to ~10 kJ/mol, in contrast to the gradual decrease of ΔH with decreasing pH for ubiquitin unfolding in solution . Protein ions can also be unfolded by collisional activation prior to structural probing by ECD  or ETD , and although in such experiments carried out under multiple low-energy collisions it is usually not possible to determine ion temperatures and derive thermodynamic information, they can be used to establish the order of stability for different regions of a gaseous protein .
Monitoring Protein Folding
A major drawback of structural probing by ECD is that a single spectrum does not reveal how much structure is retained in regions framed by two residues that are still held together by noncovalent bonding. This means that if gaseous protein ions were to fold by first associating their N- and C-terminal residues, all fragment ions would disappear simultaneously, no matter how compact or extended the structure were of the residues in between. However, for the proteins studied so far, horse and tuna heart cytochromes c and ubiquitin, such behavior has not been observed. Instead, regions in which the site-specific yield of ECD fragments decreased with increasing time allowed for folding (indicating the formation of noncovalent interactions between the complementary c and z• fragments from backbone cleavage in that region) were generally separated by regions in which the yield did not change (no new interactions formed), decreased to a lesser extent (fewer interactions formed), or even increased (loss of interactions) [9, 80]. Moreover, for these proteins, folding in the gas phase is not only slower than folding in solution by several orders of magnitude [10, 12, 80, 83, 84], it is also dominated by nearest neighbor interactions instead of global structural changes, and driven by electrostatic instead of hydrophobic interactions .
Experimental Conditions for ECD
An important issue to be considered in ECD experiments for structural probing, as well as for sequencing applications, is secondary dissociation as a result of secondary (or multiple) electron capture. Because the cross-section for electron capture generally increases with the square of the net positive charge of an ion , the probability for secondary electron capture is generally higher for larger fragment ions that typically carry more charge than smaller fragments. Thus, under conditions of multiple electron capture, larger c and z• fragments are preferentially depleted, and only smaller fragments from backbone cleavage close to the protein termini are detected, which produces ECD spectra that can mistakenly be interpreted as reflecting noncovalent bonding in the center region of the protein. Such spectra are characterized by missing complementary ions, that is, only small c and z• fragments from cleavage close to the N and C terminus are detected, respectively. An example of such a spectrum can be found in Figure 1 of reference , and spectra recorded under conditions of single electron capture that show approximately equal yields of c and z• fragments at each cleavage site can be found in Figure 2 of reference .
Open Questions and Future Challenges
ECD of proteins does not provide direct structural information, but the observation of c and z• fragments immediately reveals the absence of noncovalent bonding between them. Moreover, when changes in ECD spectra are followed as the temperature or charge of the protein ions under study is increased, or time is allowed for folding, site-specific data on the structural changes involved can be obtained. A major challenge for comprehensive structural studies is the analysis of complex ECD spectra. Although automated algorithms for spectral interpretation that can correctly assign and quantify the majority (~90%) of signals in an ECD spectrum are available, additional manual interpretation is, in our experience, absolutely essential. Regarding the mechanism of protein backbone bond cleavage into c and complementary z• ions and the implications for structural probing, it remains to be seen if there are lower mass and higher charge limits within which the energy gained by electron capture does not cause significant unfolding, and the role of electronically excited states in radiative cooling needs to be addressed. Finally, because charge locations in peptide and protein ions from ESI are critical to the development of theoretical models of the ECD process, a better understanding of the electrospray process and the structural changes involved is vitally important.
ECD fragment ion yields are calculated as percentage values relative to all ECD products, considering that backbone dissociation of a parent ion gives a pair of complementary c and z• or a pair of a• and y ions, 100% = 0.5[c]. + 0.5[a•] + 0.5[z•] + 0.5[y] + [other products], in which other products are reduced molecular ions and products from loss of small neutral species from the latter. In cases where a• and y ions are of low abundance with correspondingly small signal-to-noise ratios, they can be excluded from the analysis.
This work was supported by the Austrian Science Fund (FWF): Y372 and P27347 to K.B.
- 3.Zubarev, R.A., Kruger, N.A., Fridriksson, E.K., Lewis, M.A., Horn, D.M., Carpenter, B.K., McLafferty, F.W.: Electron capture dissociation of gaseous multiply-charged proteins is favored at disulfide bonds and other sites of high hydrogen atom affinity. J. Am. Chem. Soc. 121, 2857–2862 (1999)CrossRefGoogle Scholar
- 8.Ledvina, A.R., McAlister, G.C., Gardner, M.W., Smith, S.I., Madsen, J.A., Schwartz, J.C., Stafford, G.C., Syka, J.E.P., Brodbelt, J.S., Coon, J.J.: Infrared photoactivation reduces peptide folding and hydrogen-atom migration following ETD tandem mass spectrometry. Angew. Chem. 48, 8526–8528 (2009)CrossRefGoogle Scholar
- 11.Oh, H., Breuker, K., Sze, S.K., Ge, Y., Carpenter, B.K., McLafferty, F.W.: Secondary and tertiary structures of gaseous protein ions characterized by electron capture dissociation mass spectrometry and photofragment spectroscopy. Proc. Natl. Acad. Sci. U.S.A. 99, 15863–15868 (2002)CrossRefGoogle Scholar
- 24.Le, T.N., Poully, J.C., Lecomte, F., Nieuwjaer, N., Manil, B., Desfrancois, C., Chirot, F., Lemoine, J., Dugourd, P., van der Rest, G., Gregoire, G.: Gas-phase structure of amyloid-beta (12–28) peptide investigated by infrared spectroscopy, electron capture dissociation, and ion mobility mass spectrometry. J. Am. Soc. Mass Spectrom. 24, 1937–1949 (2013)CrossRefGoogle Scholar
- 36.Ben Hamidane, H., He, H., Tsybin, O.Y., Emmett, M.R., Hendrickson, C.L., Marshall, A.G., Tsybin, Y.O.: Periodic sequence distribution of product ion abundances in electron capture dissociation of amphipathic peptides and proteins. J. Am. Soc. Mass Spectrom. 20, 1182–1192 (2009)CrossRefGoogle Scholar
- 44.Available at: http://webbook.nist.gov. Accessed 14 Jan 2015
- 56.Li, J.W., Taraszka, J.A., Counterman, A.E., Clemmer, D.E.: Influence of solvent composition and capillary temperature on the conformations of electrosprayed ions: unfolding of compact ubiquitin conformers from pseudonative and denatured solutions. Int. J. Mass Spectrom. 187, 37–47 (1999)CrossRefGoogle Scholar
- 58.Toker, Y., Aviv, O., Eritt, M., Rappaport, M.L., Heber, O., Schwalm, D., Zajfman, D.: Radiative cooling of Al4 – clusters. Phys. Rev. A 76, (2007)Google Scholar
- 64.Schnier, P.D., Price, W.D., Strittmatter, E.F., Williams, E.R.: Dissociation energetics and mechanisms of leucine enkephalin (M + H)+ and (2 M + X)+ ions (X = H, Li, Na, K, and Rb) measured by blackbody infrared radiative dissociation. J. Am. Soc. Mass Spectrom. 8, 771–780 (1997)CrossRefGoogle Scholar
- 67.Antoine, R., Broyer, M., Chamot-Rooke, J., Dedonder, C., Desfrancois, C., Dugourd, P., Gregoire, G., Jouvet, C., Onidas, D., Poulain, P., Tabarin, T., van der Rest, G.: Comparison of the fragmentation pattern induced by collisions, laser excitation and electron capture. Influence of the initial excitation. Rapid Commun. Mass Spectrom 20, 1648–1652 (2006)CrossRefGoogle Scholar
- 73.Lifshitz, C.: Intramolecular Vibrational Energy Distribution and Ergodicity of Biomolecular Dissociation. In: Principles of Mass Spectrometry Applied to Biomolecules. Wiley, Hoboken (2006)Google Scholar
- 75.Turecek, F., Chung, T.W., Moss, C.L., Wyer, J.A., Ehlerding, A., Holm, A.I.S., Zettergren, H., Nielsen, S.B., Hvelplund, P., Chamot-Rooke, J., Bythell, B., Paizs, B.: The histidine effect. Electron transfer and capture cause different dissociations and rearrangements of histidine peptide cation-radicals. J. Am. Chem. Soc. 132, 10728–10740 (2010)CrossRefGoogle Scholar
- 81.Breuker, K.: Protein Structure and Folding in the Gas Phase: Ubiquitin and Cytochrome c. In: Principles of Mass Spectrometry Applied to Biomolecules. Wiley, Hoboken (2006)Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.