- Top of page
- Experimental Section
Each of the two non-overlapping disulfide bonds of trypsin inhibitor (C39/C86 and C136/C145) has a basic residue as direct sequence neighbor of one of the cysteines involved in disulfide bonding (R38 and C39, K144 and C145). According to theoretical predictions,12 the positive charge on the neighboring basic residues should either allow direct electron attachment into the disulfide σ* orbitals, or electron attachment into Rydberg orbitals of the positively charged sites of R38 and K144, followed by fast (∼10−6 s) relaxation into lower-energy Rydberg levels suitable for electron transfer to disulfide σ* orbitals, with either scenario resulting in disulfide bond cleavage. However, even under conditions of capture of up to four electrons and vibrational excitation (up to 120 eV laboratory frame energy) for ion unfolding before electron capture, we observed no products indicating disulfide bond cleavage in ECD of (M+n H)n+ ions of trypsin inhibitor with n=9–12 (Figure 2 D and 2 E).
Moreover, gentle IR laser heating of (M+11 H)10+. ions from single-electron capture of (M+11 H)11+ ions of trypsin inhibitor separated c and z. fragments from cleavage at nearly all sites except those bridged by disulfide bonds (Figure 2 F). This data shows that capture of a single electron by (M+11 H)11+ ions of trypsin inhibitor does not bring about both disulfide and backbone bond cleavage. Instead, the data indicates virtually unselective backbone cleavage into c and z. fragments while fully preserving the C39/C86 and C136/C145 disulfide bonds.
A similar fragmentation pattern was observed in ECD of the more highly charged (M+n H)n+ ions of trypsin inhibitor with n=16–20 under conditions of gentle collisional activation before ECD (Figure 3 A). Only with more rigorous collisional activation before ECD did the number of separated c and z. fragments from backbone cleavage in disulfide-bridged regions increase significantly (Figure 3 B), but not nearly to the extent observed in unbridged regions. Apparently, collisional activation before ECD, and not electron capture, caused disulfide bond cleavage in some of the more highly charged (M+n H)n+ ions of trypsin inhibitor. In support of this hypothesis, are data from ETD of monoclonal antibodies by Tsybin and coworkers, who observed c and z. fragments mostly from backbone cleavage in regions not bridged by disulfide bonds.13
Insulin, which has three disulfide bonds but consists of two separate peptide chains, also shows evidence for disulfide bond cleavage as a result of vibrational excitation; ECD of its (M+5 H)5+ ions under conditions of capture of up to three electrons and without prior collisional activation gave no c and z. fragments from backbone cleavage in regions bridged by disulfide bonds (Figure 4 B), and no separated A or B chains, whereas collisional activation before ECD gave two products indicating disulfide bond cleavage (Figure 4 C). Likewise, the number of c, z. fragment ions from cleavage in disulfide-bridged regions increased with prior collisional activation of the (M+6 H)6+ ions of insulin (Figure 4 D and 4 E). A similar effect of vibrational excitation on disulfide bond cleavage was observed by McLuckey and coworkers, who found c and z. fragments from backbone cleavage in regions bridged by disulfide bonds and separated A or B chains in ETD of insulin (M+n H)n+ ions with n=3–6 only with collisional activation of the (M+n H)(n−1)+. ions.14
Like insulin, aprotinin has three disulfide bonds, and also comprises a similar number of residues (insulin: 51, and aprotinin: 58), but aprotinin consists of a single peptide chain. The three disulfide bonds (C5/C55, C14/C38, C30/C51) divide the protein backbone into regions bridged by up to three disulfide bonds; neighboring basic residues are K15 (C14), R39 (C38), and R53 (C51, C55). Without collisional activation and under conditions of capture of up to three electrons, ECD of the (M+6 H)6+ ions gave c and z. fragments from backbone cleavage in unbridged and singly, but not doubly or triply, bridged regions (Figure 5 B). With collisional activation before ECD, c and z. fragments from backbone cleavage in all regions were observed (Figure 5 C). Disulfide bond cleavage in the (M+7 H)7+ ions (Figure 5 D) required only about half the energy (28 eV) than that for the (M+6 H)6+ ions (60 eV), consistent with increasing Coulomb repulsion decreasing overall ion stability.
The data discussed so far show that collisional activation before ECD generally increases the number of c and z. fragments from backbone cleavage in disulfide-bridged regions, strongly suggesting that vibrational excitation can cleave protein disulfide bonds, at least to some extent. This finding is consistent with recent studies reporting disulfide bond cleavage, along with backbone cleavage into b and y fragments, in collisionally activated dissociation of even-electron (M+n H)n+ protein15 and peptide16 ions.
However, mere vibrational excitation can obviously not account for the observed .SH losses and apparently coinciding disulfide bond cleavages in ECD of insulin (Figure 4 D) and aprotinin (Figure 5 B) without prior collisional activation or subsequent IR laser heating. This data instead suggests that radical ion chemistry was involved in at least some of the disulfide bond cleavages. Concurrent disulfide bond cleavage and .SH loss was also observed in ETD of somatostatin, a 14 residue peptide with a disulfide bond between C3 and C14,16a and other small peptides comprising a single disulfide bond.17
If radical reactions can cleave disulfide bonds in ECD, are these affected by vibrational excitation? It was shown that ion heating before or after ECD can break noncovalent bonds in protein ions with more compact structures that could otherwise prevent separation of c and z. fragments from backbone cleavage,4, 11a, c, 18 consistent with the data for trypsin inhibitor (Figure 2 D–F). By contrast, heating of protein ions with elongated structures, for example, the (M+13 H)13+ ions of ubiquitin,19 showed no increase in yields of the even-electron c ions with increasing ion temperature from 25 to 125 °C,20 which corroborates the postulate of nonergodic dissociation in ECD.3 Yields of the complementary, radical z. ions (Scheme 1), however, were found to actually decrease with increasing ion temperature, which was attributed to their lower stability and higher reactivity.11a, 20–21 In other words, primary backbone cleavage into c and z. fragments, which is not affected by vibrational excitation,20 can be followed by secondary reactions that involve the radical z. ions. Evidence for secondary reactions of radical z. ions was also found in ECD of linear21 and cyclic peptides.22 These radical reactions are generally fast, proceeding on a sub-millisecond timescale,23 and—in striking contrast to primary backbone cleavage into c and z. fragments—become more efficient with increasing ion temperature.11a The increase in .SH loss from (M+6 H)5+. ions of aprotinin with increasing vibrational excitation (Figure 5 E) strongly suggests that the radical pathway resulting in disulfide bond cleavage involves secondary radical reactions of the z. ions from primary backbone cleavage. Note that the energy deposited per residue in IR laser heating of (M+6 H)5+. ions of aprotinin (Figure 5 E) was 2.4–6.1 times higher than that used for IR laser heating of (M+11 H)10+. ions of trypsin inhibitor (Figure 2 F). This qualitatively agrees with lower energy requirements for breaking of noncovalent bonds in trypsin inhibitor to separate c and z. fragments, and higher energy requirements for facilitating secondary radical reactions of z. ions—which involves breaking of covalent bonds—in aprotinin.
Secondary radical reactions can be part of radical ′cascade reactions′,22, 24 which provides a rationale for how IR laser heating of the (M+6 H)5+. ions of aprotinin, formed by capture of a single electron, cleaved up to three disulfide bonds and the protein backbone (Figure 5 D). Possibly these cascade reactions are preferentially interrupted at sites of higher radical stability25 such as residues with aromatic side chains, thereby increasing the relative number of c and z. fragments from backbone cleavage next to tyrosine and phenylalanine residues (Figures 4–4, 5, 6).
So why do we find strong evidence for secondary radical reactions in ECD of insulin and aprotinin, but not for ecotin and trypsin inhibitor? A major difference between these proteins is the number of cyclic regions formed by disulfide bonds normalized to the number of residues (Table 1). Ecotin and trypsin inhibitor have rather small ′densities′ of cyclic regions, 0.007 and 0.011, respectively, meaning that the gas phase structures of their (M+n H)n+ ions, which were electrosprayed from 50:50:1 water/methanol/acetic acid solutions at pH 2.5 to promote unfolding in solution, should be rather loose, although some higher order structure is indicated in some regions of the (M+n H)n+ ions of trypsin inhibitor with n=9–12 (Figure 2 D–F). On the other hand, the far higher density of cyclic regions of insulin (0.039) and aprotinin (0.052) necessitates far more densely packed and compact ion structures in which radical sites from primary backbone cleavage are far closer to sites for secondary reaction, including their disulfide bonds.
Table 1. Density of cyclic regions in proteins studied.
To test our idea of compact ion structures facilitating secondary radical reactions in ECD, we studied the small peptides K8- and R8-vasopressin with a density of cyclic regions of 0.111 (Table 1). In strong support of our hypothesis, ECD of their (M+2 H)2+ ions gave rather unselective backbone cleavage into c and z. fragments, even without prior collisional activation (Figure 7 A and 7 B). Note that NCα bond cleavage at site 6 does not separate c6 and the complementary z3. fragments, because the side chain of P7 is covalently bound to its amide nitrogen. The only other site for which no fragment ions were observed is site 5; possibly the radical z4. fragment was fully depleted by secondary reaction with the adjacent disulfide bond. Collisional activation (6 eV) prior to ECD gave virtually the same spectra and c, z. fragmentation patterns, suggesting that the secondary radical reactions proceeded to completion even without vibrational excitation; as illustrated in Figure 7 C for K8-vasopressin, the (M+2 H)+. ions from electron capture of (M+2 H)2+ ions were completely depleted by cleavage into c and z. fragments, and loss of H. and .SH.
We have studied here cyclic peptides and proteins, but protein digestion can also give disulfide-bonded, noncyclic peptide structures.6, 26 In a study by the McLuckey group, ETD of smaller peptides from tryptic digestion (up to 14 residues in total), consisting of two chains (of up to 9 residues) connected by a single disulfide bond, gave products from both backbone and disulfide bond cleavage, whose branching ratio varied substantially with the identity and location of positively charged sites within the peptide ions.26 Charge location is an important factor in determining ion structure and thus ion compactness, and although the structures of the peptides studied by McLuckey and coworkers are not known, an effect of charge location on the relative extent of disulfide bond cleavage is generally consistent with our hypothesis of compact ion structures facilitating secondary radical reactions that can lead to disulfide bond cleavage. In the 1999 study by the McLafferty group, ECD of a disulfide-bonded dimer (34 residues in total) of a 17-residue peptide gave products from both backbone and disulfide bond cleavage.6 Among the products from backbone cleavage, c fragments dominated over z. fragments, consistent with our proposal of secondary radical reactions of z. ions leading to disulfide bond cleavage. Higher relative abundances of c over z. fragments are also evident in most of the ETD spectra in reference,26 and the ECD spectrum of a small tryptic peptide (11 residues in total) showed only c but no z. fragments, along with separated peptide chains.27 In multistage dissociation experiments, Wu, Hancock, Karger, and coworkers utilized this ability of ETD to separate peptide chains for the characterization of disulfide-bonded therapeutic proteins.28 However, in a spectrum from ECD of (M+10 H)10+ ions of a far larger peptide consisting of two chains (A: 38 and B: 48 residues) connected by a single disulfide bond, McLafferty and coworkers found very few c and z. fragments, and primarily signals from reduced molecular ions and separated A and B chains.6 Possibly yet another mechanism for disulfide bond cleavage is operative in ECD of larger, noncyclic peptides.
- Top of page
- Experimental Section
It turns out that the question in the title “Does electron capture dissociation cleave protein disulfide bonds?” cannot be answered with a simple yes or no. We found unambiguous evidence for full preservation of disulfide bonds during electron capture dissociation (ECD), even for proteins that have basic residues next to cysteines involved in disulfide bonding, and under conditions of multiple electron capture.
However, we also found unambiguous evidence that ECD can cleave disulfide bonds, although not preferentially. Instead, our data support a mechanism of electron capture at protonated sites, resulting in backbone cleavage into c and z. fragment ions. This primary cleavage event can be followed by disulfide bond cleavage via secondary radical reactions, provided that the disulfide bond is in close proximity to a radical site from primary backbone cleavage. Apparently, this situation is favored in peptides or proteins with a high density of cyclic regions.
Vibrational excitation of the (M+n H)n+ ions before ECD can also cleave disulfide bonds, and it can promote secondary (but not primary) radical reactions in the (M+n H)(n−1)+. ions that may lead to disulfide bond cleavage. Unintended vibrational excitation can be caused by tuning electrostatic potentials in the source region of a mass spectrometer for high ion transmission, which may explain differences in ECD fragmentation patterns found by different experimenters. Our data also show that increased ion net charge generally resulted in more disulfide bond cleavage, consistent with an increased number of possible sites for primary backbone cleavage increasing the probability for secondary reaction at disulfide bonds; similar observations were recently made by Zhang and Loo.29
These rather intricate effects of vibrational excitation and secondary radical reactions make it nearly impossible to predict whether or not, or to what extent, disulfide bonds of any given protein are preserved during collisionally activated dissociation (CAD), electron capture dissociation (ECD), or electron transfer dissociation (ETD). For sequencing applications, peptide or protein disulfide bonds can be fully reduced in solution prior to analysis by mass spectrometry. Elucidation of protein disulfide bond patterns by CAD, ECD or ETD, however, can be seriously complicated by complex gas phase ion chemistry.
- Top of page
- Experimental Section
Experiments were performed on a 7 Tesla Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (Bruker, Austria) equipped with an electrospray ionization (ESI) source, a linear hexapole ion cell floated with argon gas for optional collisional activation, a hollow dispenser cathode for electron capture dissociation (ECD; electron energy<1 eV), and a CO2 laser (10.6 μm, 35 W) for IR activation in the ion cyclotron resonance (ICR) cell. Data reduction utilized SNAP2 (Bruker, Austria) and mMass software (version 3.11).30 Proteins and chemicals were purchased from Sigma–Aldrich (Austria). Amino acid sequences:
AESVQPLEKI APYPQAEKGM KRQVIQLTPQ EDESTLKVEL
LIGQTLEVDC NLHRLGGKLE NKTLEGWGYD YYVFDKVSSP
VSTMMACPDG KKEKKFVTAY LGDAGMLRYN SKLPIVVYTP
trypsin inhibitor (180 residues)
DFVLDNEGNP LENGGTYYIL SDITAFGGIR AAPTGNERCP
LTVVQSRNEL DKGIGTIISS PYRIRFIAEG HPLSLKFDSF
AVIMLCVGIP TEWSVVEDLP EGPAVKIGEN KDAMDGWFRL
ERVSDDEFNN YKLVFCPQQA EDDKCGDIGI SIDHDDGTRR
GIVEQCCTSI CSLYQLENYC N (chain A) and
FVNQHLCGSH LVEALYLVCG ERGFFYTPKT (chain B)
RPDFCLEPPY TGPCKARIIR YFYNAKAGLC QTFVYGGCRA
K8-vasopressin (9 residues) CYFQNCPKG
R8-vasopressin (9 residues) CYFQNCPRG