Proton Mobility Frag

RAPID COMMUNICATIONS IN MASS SPECTROMETRY Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and main fragmentation pathways of protonated lysylglycine Istva´n Pa´l Csonka1, Be´la Paizs1*, Gyo¨rgy Lendvay2 and Sa´ndor Suhai1 1

Department of Molecular Biophysics, German Cancer Research Center, Im Neuenheimer Feld 280, D-69120 Heidelberg, Germany Institute of Chemistry, Hungarian Academy of Sciences, P.O. Box 17, H-1525 Budapest, Hungary

2

Received 27 March 2001; Revised 17 June 2001; Accepted 18 June 2001

Theoretical model calculations were performed to validate the `mobile proton' model for protonated lysylglycine (KG). Detailed scans carried out at various quantum chemical levels of the potential energy surface (PES) of protonated KG resulted in a large number of minima belonging to various protonation sites and conformers. Transition structures corresponding to proton transfer reactions between different protonation sites were determined, to obtain some energetic and structural insight into the atomic details of these processes. The rate coefficients of the proton transfer reactions between the isomers were calculated using the Rice-Ramsperger-Kassel-Marcus (RRKM) method in order to obtain a quantitative measure of the time-scale of these processes. Our results clearly indicate that the added proton is less mobile for protonated KG than for peptides lacking a basic amino acid residue. However, the energy needed to reach the energetically less favorable but±from the point of view of backbone fragmentation±critical amide nitrogen protonation sites is available in tandem mass spectrometers operated under low-energy collision conditions. Using the results of our scan of the PES of protonated KG, the dissociation pathways corresponding to the main fragmentation channels for protonated KG were also determined. Such pathways include loss of ammonia and formation of a protonated a-amino-e-caprolactam. The results of our theoretical modeling, which revealed all the atomic details of these processes, are in agreement with the available experimental results. Copyright # 2001 John Wiley & Sons, Ltd.

Protonated peptides activated under low-energy conditions dissociate mainly by charge-directed fragmentation along their backbone.1±5 In many cases these reactions result in structurally informative bn and Yn@ ion series.6±10 (bn and Yn@ ions are formed by charge retention on the N- or C-termini, respectively.) It is well known from the results of molecular orbital calculations3±5 that protonation on the amide nitrogen leads to considerable weakening of the amide bond. On the other hand, protonation on the amide oxygen makes the amide bonds even stronger than those in the neutral counterparts. It is also well known that protonation on the amide nitrogen is thermodynamically unfavored compared to other protonation sites like the amide oxygens, the Nterminal amino group, or basic amino acid (AA) side chains such as arginine and lysine. In short, from the point of view of decomposition, protonation on the amide nitrogen is favorable, while from the thermodynamic point of view this site is not the most favored one. To explain why decomposition still occurs, the `mobile proton' model1,3,4,11±20 was introduced, which proposes that the proton(s) added to a peptide migrate(s) amongst protonation sites prior to fragmentation provided that they are not sequestered by a *Correspondence to: B. Paizs, Department of Molecular Biophysics, German Cancer Research Center. Im Neuenheimer Feld 280, D-69120 Heidelberg, Germany. E-mail: [email protected] Contract/grant sponsor: Hungarian Scienti®c Research Fund; Contract/grant number: OKTA T22824. DOI:10.1002/rcm.388

basic AA side chain. The mobile proton model has been verified experimentally by using deuterium-labelling techniques13,15,17 which indicated strong H/D mixing prior to collisionally induced dissociation (CID) of MD‡ ions of a number of small peptides. In a very early study, Tsang and Harrison11 showed that facile H/D mixing occurs in the D2 and CD4 chemical ionization of AAs. In a recent study Harrison and Yalcin17 suggested that the proton added to non-basic AAs, and to peptides containing no basic AAs, samples all positions bearing labile hydrogens prior to fragmentation of the protonated species. Also, Wysocki and co-workers4,14,16 have demonstrated that the relative positions of fragmentation efficiency curves obtained by electrospray ionization in combination with surface induced dissociation (ESI-SID) depends on the amino acid composition (absence or presence and type of a basic residue) and on the sequence and the size of the peptide investigated. This dependence could be firmly rationalized based on the `mobile proton' model, providing further support for the underlying mechanistic considerations. In our previous ab initio and RRKM studies19,20 we investigated the rate of proton transfer and isomerisation processes (transitions determining the `proton traffic' of a protonated species) in protonated N-formylglycineamide, glycylglycine, and N-formylglycyl glycineamide, using pure theoretical methods. These calculations started with a careful investigation of the potential energy surfaces (PESs) of the above species to determine minima corresponding to various Copyright # 2001 John Wiley & Sons, Ltd.

1458 I. P. Csonka et al.

Scheme 1. Various protonated forms of KG. Since the positive charge is delocalized and the bond orders are not integer numbers, the valence bond limiting structures as drawn are only approximate representations.

protonation sites and conformers, and transition structures which describe possible transitions between these minima. Because all the processes involved in the `proton migration' are unimolecular in nature, one can estimate the corresponding rate constants by using the RRKM formalism. In the RRKM calculations we applied the results of quantum chemical calculations such as energetics, vibrational frequencies, etc. The most important conclusions of these studies are as follows. The global minima on the PESs of protonated Nformylglycineamide and glycylglycine are the isomers protonated at the amide oxygen and terminal amino nitrogen, respectively. The relative energies of the minima corresponding to the other isomers are higher, but the differences are not so large that they could prevent formation of isomers protonated at the amide oxygens (protonated GG) and amide nitrogens, in mass spectrometers operated under the most common conditions. Detailed analysis of the saddle points separating various minima on the PES indicate that low-energy pathways exist which connect the global minimum with potential wells corresponding to isomers protonated at the amide nitrogen. RRKM calculations suggested that transitions occurring on these pathways are fast and take place easily on the time-scale of commonly used spectrometers. These proton transfer (PT) reactions are usually very fast as soon as the internal energy of the ion exceeds the corresponding threshold energies. Furthermore, in most cases, the rates of the PT reactions and of the conformational transitions around flexible bonds are comparable. In the case of a longer peptide chain lacking strongly basic amino acid residues, the most possible route for proton migration along the backbone involves proton movement between adjacent C=O-H¼O=C bridges. Our studies indicated that the rate of proton transfer along the backbone of the peptide chain by proton hops between the oxygens of adjacent amide bonds is also fast. Overall, our theoretical Copyright # 2001 John Wiley & Sons, Ltd.

modeling of proton migration in protonated peptides lends strong support to the mechanistic considerations involved in the `mobile proton' model. As a side project we examined the role of various fivemembered ring structures in the fragmentation of protonated N-formylglycineamide (loss of ammonia) and glycylglycine (loss of water). In these studies we found that the relative energies of the species containing five-membered rings are close to those of the amide-nitrogen-protonated species. Furthermore, the barriers to dissociation initiated from the five-membered ring structures are high (the corresponding unimolecular rate constants are small), indicating that these species are not involved in backbone fragmentation of the protonated peptides investigated. On the other hand, we have presented strong theoretical evidence5,21 which suggests that backbone fragmentation indeed involves amide-nitrogen-protonated isomers. The available experimental information on proton migration in lysine-containing peptides is quite limited. Wysocki and co-workers investigated4,16 fragmentation efficiency curves determined by ESI-SID for various peptides containing lysine, such as leucine-enkephalin analogs, Ax oligopeptides substituted by K, etc. These authors found that the peptides lacking any basic AAs fragment more easily than those containing either K or R. Also, the lysine-containing peptides fragment more easily than those containing R. In the language of the `mobile proton' model this means that the energy required to produce the fragmenting (reactive) amide-N-protonated species increases for the series of peptides containing no basic AA, a lysine, and an arginine side chain, respectively. Fragmentation of protonated lysine and small peptides containing lysine was studied by Harrison and co-workers.22,23 Protonated lysine fragments almost exclusively by loss of ammonia under both metastable and low-energy CID conditions. Studies of protonated [a-15N]lysine showed that Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

1459

Table 1. Calculated total and relative energies (with ZPE correction) of the investigated conformers of protonated KG at the B3LYP/631G(d) level of theory and the main structure determining interactions. The relative energies are measured from the global minimum of protonated KG (E01) Species E01 E02 E03 E10 E11 E12 E16 E17 E18 E19 E20 A01 A02 A03 A10 A13 A15 A20 A21 A22 O01 O02 O03 O08 O14 O15 O16 I01 I02 I03 D01 D02 D03 D04 D06 D10 D15 D16 D17 D18 D19

Energy (a. u.)

Rel. Ezpe (kcal/mol)

Interactions

Conformers protonated on the e-amino group 0.0 Ne-H¼Oamide, Ne-H'¼Na, Namide-H¼Ocarboxyl; delocalization 0.1 Ne-H¼Oamide, Ne-H'¼Ocarboxyl=O, Namide-H¼Na 0.2 Ne-H¼Oamide, Ne-H'¼Ocarboxyl=O, Namide-H¼Na 3.3 Ne-H¼Oamide, Namide-H¼Na, Namide-H¼Ocarboxyl=O; delocalization 3.5 Ne-H¼Oamide, Namide-H¼Ocarboxyl=O; delocalization 3.5 Ne-H¼Oamide, Namide-H¼Na, Namide-H'¼Ocarboxyl=O; delocalization 4.9 Ne-H¼Na, Na-H¼Oamide, Namide-H¼Ocarboxyl; delocalization 5.3 Ne-H¼Na, Na-H¼Oamide, Namide-H¼Ocarboxyl=O; delocalization 7.2 Ne-H¼Oamid, Namid-H¼Ocarboxyl=O; delocalization 8.7 Ne-H¼Na, Na-H¼Oamid, Namid-H¼Ocarboxyl OH; delocalization 11.2 Ne-H¼Na, Na-H¼Oamid, Namid-H¼Ocarboxyl OH; delocalization Conformers protonated on the a-amino group 705.440327 3.5 Na-H¼Ne, Na-H'¼Oamid, Namid-H¼Ocarboxyl=O; delocalization 705.439854 3.5 Na-H¼Ne, Na-H'¼Oamid, Namid-H¼Ocarboxyl=O; delocalization 705.439418 3.7 Na-H¼Ne, Na-H'¼Oamid, Namid-H¼Ocarboxyl=O; delocalization 705.432689 7.7 Na-H¼Ne, Na-H'¼Oamid, Namid-H¼Ocarboxyl OH; delocalization 705.430559 9.4 Na-H¼Ne, Na-H'¼Oamid, Namid-H¼Ocarboxyl OH; delocalization 705.427518 11.4 Na-H¼Oamid, Namid-H¼Ne, Namid-H¼Ocarboxyl=O; delocalization 705.423985 14.6 Namid-H¼Ne, Na-H¼ Ocarboxyl=O, Na-H'¼Oamid 705.411258 21.8 Namid-H¼Ne, Na-H¼ Ocarboxyl=O 705.402464 27.3 Ocarboxyl OH-H¼Ne, Na-H¼ Ocarboxyl=O Conformers protonated on the amide oxygen 705.426279 11.3 Namid-H¼Ne, Oamid-H¼Na, Ne-H¼Ocarboxyl=O,Namid-H¼Ocarboxyl=O 705.416834 17.5 Namid-H¼Ne, Oamid-H¼Na, Ne-H¼Ocarboxyl OH,Namid-H¼Ocarboxyl OH 705.414935 18.8 Namid-H¼Na, Oamid-H¼Ocarboxyl=O, Na-H¼Ne 705.410184 21.2 Oamid-H¼Na, Na-H¼Ne, Namid-H¼Ocarboxyl=O 705.401433 26.6 Namid-H¼Na, Oamid-H¼Ocarboxyl=O 705.401361 27.0 Oamid-H¼Ocarboxyl OH, Namid-H¼Ne, Na-H¼ Oamid 705.400817 27.3 Namid-H¼Ne, Oamid-H¼Ocarboxyl OH Imine-type conformers, protonated on the e-amino group and on the amide oxygen 705.426910 11.5 Ne-H¼Nimine, Ne-H'¼ Ocarboxyl=O, O(amid)-H¼Na, delocalization 705.413858 19.9 Ne-H¼Na, Ne-H'¼Ocarboxyl=O 705.408663 23.1 Ne-H¼Nimine, Ne-H'¼ Ocarboxyl=O, Na-H¼O(amid), delocalization Conformers protonated on the amide nitrogen Ê 705.408436 22.5 Namid-H¼Ne, Namid-H'¼Na, Na-H¼Ocarboxyl=O;Camid-Ocarboxyl=O 2.812 A Ê 705.406218 24.1 Namid-H¼Ne, Namid-H'¼Na, Na-H¼Ocarboxyl=O;Camid-Ocarboxyl=O 2.839 A 705.395373 30.8 Ocarboxyl OH-H¼Ne, Namid-H¼Na, Namid-H'¼Ocarboxyl=O Ê 705.394236 31.4 Namid-H¼Na, Na-H¼Ne; Camid-Ocarboxyl=O 2.747 A 705.446754 705.447191 705.446779 705.440298 705.441123 705.440202 705.437334 705.436847 705.434768 705.429751 705.426675

705.393793 705.388727 705.388367 705.385700 705.386395 705.385779 705.386396

31.8 34.9 35.7 36.6 36.7 36.7 36.7

Namid-H¼Na, Namid-H'¼Ocarboxyl=O, Na-H¼Ne, Ê) Namid-H¼Na, Namid-H'¼Ocarboxyl=O, (Camid-Ne, 3.713 A Ê Namid-H¼Na, Namid-H'¼Ocarboxyl=O, Camid-Ne, 2.665 A Ê) Namid-H¼Na, Namid-H'¼Ocarboxyl=O, (Camid-Ne, 3.827 A Ê Namid-H¼Na, Namid-H'¼Ocarboxyl=O, Camid-Ne, 1.999 A Namid-H¼Na, Namid-H'¼Ocarboxyl-OH, Na-H¼Ne, Ê Namid-H¼Na, Namid-H'¼Ocarboxyl=O, Camid-Ne, 2.704 A

the expelled ammonia specifically carried away the nitrogen of the side chain in both unimolecular (i.e., collison-free) and collision-induced fragmentation. Harrison and co-workers suggested that the ion formed by ammonia loss from protonated K is protonated pipecolic acid. Amongst the investigated peptide derivatives, protonated KG was found to have main dissociation pathways leading to loss of ammonia and formation of an ion with m/z 129. Although the m/z 129 ion is nominally an acylium ion, its metastable ion characteristics and CID spectrum suggest that this ion is indeed a protonated a-amino-e-caprolactam. The formation of the m/z 129 ion was found to be characteristic for other lysine-containing peptides as well. Copyright # 2001 John Wiley & Sons, Ltd.

In the present work we extend our theoretical modeling of proton migration in peptides to include lysylglycine, which is one of the simplest peptides containing a basic amino acid. Our goal here is to validate the mobile proton model using a combined ab initio/RRKM methodology for peptides containing a basic amino acid. Furthermore, a scan of the PES of protonated KG enables us to calculate the energetics and kinetics of the main fragmentation channels of this ion. In the remainder of this paper, after summarizing the methods used, we present the results pertaining to the proton transfer in protonated lysylglycine (KG), followed by the results related to the mechanism of decomposition via ammonia loss and a-amino-e-caprolactam formation. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1460 I. P. Csonka et al.

COMPUTATIONAL DETAILS In the case of molecules which contain many flexible torsional degrees of internal freedom, generating the proper input geometries is at least as important as the choice of the employed theoretical method. In the last few years we worked out an efficient search engine20,21,24 which can explore the low-energy structures of protonated peptides in a rather efficient way. Our strategy involves molecular dynamics simulations and subsequent geometry optimizations at both low (initial ab initio scan) and higher levels (final ab initio scan) of theory. The whole system is highly automated using scripts and small programs. The search applied for protonated KG resulted in species protonated at the side-chain e-amine group (E, Scheme 1), species protonated at the a-amine group (A, Scheme 1), species protonated at the carbonyl oxygen of the amide bond (O, Scheme 1), species protonated at the nitrogen atom of the amide group (D, Scheme 1), and species which contain an imine bond instead of the K-G amide bond (I, Scheme 1). The ab inito calculations were carried out at the HF/3-21G level of theory for the initial scan of the PES of protonated KG. In our previous articles19,20 we showed that the results of RRKM calculations depend much less on the level of theory employed for the quantum chemical calculations than the calculated relative energies of the conformers do. This observation allows us to use a relatively inexpensive quantum chemical level, so that the final ab initio scan of the PES of protonated KG was carried out at the B3LYP/631G(d) level of theory in the present work. The ab initio calculations were carried out using the Gaussian 98 program suite.25 After checking the structures obtained at the B3LYP/631G(d) level to be true minima by calculating harmonic frequencies, we located transition states (TSs) which connect different protonation sites or correspond to the main fragmentation pathways of protonated KG. The TSs found were checked to be first-order saddle points by calculating harmonic frequencies. Furthermore, in some cases, we applied IRC (intrinsic reaction coordinate) calculations to check which minima are connected by a given TS. The zero point energy (ZPE) corrections were calculated from the harmonic frequencies with no scaling factor. Using the results of the aforementioned ab initio calculations, the rate coefficients for the transitions between the minima on the PES were calculated using the RRKM method,26,27 over a grid of energies up to 120 kcal/mol of internal energy. We used the same method for each type of transition (proton transfer or fragmentation transition). No tunneling correction was included for two reasons: (1) in most cases the rates are very high as soon as the energy is above the reaction threshold; and (2) the accuracy of a few kcal mol 1 for the barrier height does not warrant the attempt to achieve the accuracy addressed by a tunneling correction.

Figure 1. Equilibrium structure of E01, the global minimum of protonated lysylglycine.

naming scheme as follows. The first character in the acronym refers to the protonation site (Scheme 1). In the names, the letters E, A, O, and D denote isomers protonated on the eamine N atom, on the a-amine N, on the amide oxygen and on the N atom involved in the amide bond, respectively. The I conformers contain imine bonds instead of the amide bond connecting the K and G units. The second and third characters, 01, 02, 03, etc., denote respectively the most stable, second most stable, etc., investigated minimum for a given structural isomer. For example, E01 represents the most stable species among the conformers in the series protonated on the e-amine group of the lysine residue. The notations of the proton transfer TSs are constructed from those of the corresponding minima, with the lower minimum occupying the first place. For example, E10_D01 represents a TS connecting E10 and D01, where E10 has lower energy than D01. The calculated total and relative energies of selected conformers of protonated KG are collected in Table 1. The relative energies are measured from the global minimum of the PES of protonated KG, isomer E01 in our nomenclature, and are corrected for ZPE. The main stabilizing interactions of the conformers are also given in Table 1. Our search for the minima on the PES of protonated KG resulted in a very large number (approximately 200) of various species. This is due to the numerous possible protonation sites of the neutral molecule and also to the many flexible rotational degrees of freedom such as those of the lysine side chain. It is evident that a detailed description and visualization of such a tremendous amount of information is not possible. Therefore, we describe only those structures in detail which are

RESULTS AND DISCUSSION Conformers of protonated KG

The conformers of protonated KG are identified using the Copyright # 2001 John Wiley & Sons, Ltd.

Figure 2. Equilibrium structure of A01. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

1461

Figure 3. Equilibrium structure of O01.

energetically favored and, accordingly, the data presented in tables are restricted to those isomers which are important from the point of view of any (PT or fragmentation) reaction. The global minimum on the PES of protonated KG is one of the conformers of the e-amine protonated isomer, namely, E01 (Fig. 1). In this conformer the positively charged e-amine group is stabilized by H-bonds, Ne-H¼Oamide and NeH'¼Na, and the hydrogen of the amide bond is involved in a Namide-H¼Ocarbonyl interaction. Furthermore, in E01, the glycine residue is planar, indicating a possible conjugation over this part of the molecule. The calculated energies of the next two conformers in the E-series are almost the same as that of E01 (Table 1). In E02 the positively charged e-amine group is stabilized by two strong H-bonds, one of them involves the oxygen of the amide-bond (Oamide), while the other involves the carbonyl oxygen of the C-terminal carboxyl group (Ocarbonyl). A third H-bond is present between the hydrogen of the amide bond and the a-amine group (Na). The structure of E03 is very similar to that of E02. The most stable conformer amongst the Ax structures, A01 (Fig. 2), is 3.5 kcal/mol less stable than E01. In this species the positively charged a-amine group is stabilized by Hbonds Na-H¼Ne and Na-H'¼Oamide. Furthermore, the hydrogen of the amide bond is involved in the NamideH¼Ocarbonyl H-bond, and there is a possible conjugation over the glycine residue due to the planarity of this part of

Figure 4. Equilibrium structure of I01. Copyright # 2001 John Wiley & Sons, Ltd.

Figure 5. Equilibrium structure of D01, the starting point of the fragmentation reaction over TS_B1_B.

the molecule. In the next two conformers of the Ax series, A02 and A03, the same interactions are present as in A01, and, consequently, they have only a slightly higher relative energies (3.5 and 3.7 kcal/mol, respectively). In our opinion the higher energies of the Ax structures relative to those of Ex can mainly be attributed to the larger proton affinity of the eamino group, but the larger ring strain of the rings created by H-bonds in the Ax-structures also plays some role in the relative stability of the Ex and Ax species. Although both protonated amine groups can be stabilized by two strong Hbonds, the e-amine group is positioned at the end of a long and flexible alkyl chain while the a-amine group is connected to the a-carbon of the relatively rigid amide bond. This causes a larger ring strain for the Ax species especially in the case of the H-bonds involving Oamide. For example, in the Ex-structures, one of the H-bonds is involved in a ninemembered floppy ring while a similar interaction for the Ax structures induces formation of a five-membered ring. The different bonding pattern has a significant impact on the geometric parameters of the respective Na-H¼Oamide HÊ bonds. For example, while the H¼Oamide distance is 1.701 A

and the Ne-H¼Oamide angle is 156.11 ° in E01, at the same Ê and Na-H¼Oamide is 121.33 ° in time H¼Oamide is 1.862 A A01. These features clearly indicate that the Na-H¼Oamide interaction is weaker in A01 than in E01. In general, the energy ranges of the ten most stable conformers of Ax and of Ex overlap. The equilibrium energy of the most stable conformer of KG protonated at the carbonyl oxygen, O01 (Fig. 3), is 11.3 kcal/mol above E01. Its structure is very similar to that of O1 of protonated GG.20 In both cases the protonated Oamide is stabilized by a H-bond involving Na and there is also a possibility for a conjugation over the backbone of the molecules due to their planarity. For O01 (protonated KG) Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1462 I. P. Csonka et al.

proton donor. In conclusion, it is not the case that Ox structures are destabilized, but that the Ax/Ex structures are overstabilized in KG compared to the case of GG, due to the larger extent of intramolecular solvation of the positively charged center. The O02 conformer possesses the same main features as O01, although its glycine moiety is not planar. In contrast, the structure of O03 is quite different from the former Ox structures. This species is stabilized by OamideH¼OcarboxylO, Namide-H¼Na and Na-H¼Ne H-bonds. When scanning the conformational space of protonated KG we found some minima (Ix, Scheme 1) which contain imine bonds instead of the K-G amide one. The Ix structures can be considered as tautomers of the Ex structures (Scheme 1). All the Ix species were obtained from geometry optimizations started from structures protonated on the amide oxygen. In some cases when the e-amino group was close to the hydrogen of the amide nitrogen, spontaneous transfer of this proton to the basic e-amino group occurred. In this way the former protonated amide oxygen became a hydroxyl group, and these species are protonated at the eamine group. The relative energies of structures of this type are comparable to those of the Ox structures. In fact, the relative energy of I01 (Fig. 4) is only 11.5 kcal/mol, which is almost as low as that of O01. In I01 the positively charged eamine group is stabilized by H-bonds, Ne-H¼Nimine and NeH'¼Ocarbonyl, while the H-bond O(amide)-H¼Na and the possible conjugation over the planar backbone of the molecule provide further stabilization for the molecule. In conformer I02 the H-bonds Ne-H¼Na and Ne-H'¼OcarboxylO imply stabilization for the positively charged e-amine group, while there is no H-bond involving the hydroxyl group formed from the amide oxygen and also there is no possibility for the conjugation since the backbone of the molecule is not planar. This could be the reason for its 19.9 kcal/mol relative energy which is much higher than that of I01. Conformer I03 has 23.1 kcal/mol relative energy, and the main interactions are Ne-H¼Nimine, Ne-H¼Ocarboxyl=O, Na-H¼O(amide) and a possible conjugation over the planar backbone of the molecule.

Table 2. Calculated absolute and relative energies (with ZPE correction) of the investigated PT TSs of protonated KG calculated at the B3LYP/6-31G(d) level of theory. The relative energies are with respect to the global minimum of protonated KG (E01) TS A01_E16 A04_E17 A10_E18 A13_E20 E02_O14 A15_O01 A13_O08 I01_O01 I03_O16 E10_D01 E12_D02 A21_D01 A20_D02 A22_D03

Energy (a. u.)

Rel. Ezpe (kcal/mol)

705.436901 705.436382 705.429682 705.426546 705.401422 705.422469 705.410100 705.423424 705.397879 705.408079 705.405789 705.396606 705.396137 705.387146

3.4 3.8 7.5 9.8 26.5 11.8 20.1 11.1 26.9 21.1 22.7 27.9 28.2 34.0

the hydrogen of the amide group is involved in a H-bond with the e-amine group. This means that the O01 conformer of KG is stabilized by one more intramolecular interaction than the O1 of GG. In spite of this extra stabilizing effect, the relative energy of O1 of protonated GG with respect to the most stable a-amine-protonated isomer is 1.2 kcal/mol,20 while that of O01 is 7.8 kcal/mol. An investigation of the structures of the A isomers of protonated KG and GG shows the reason why O01 seems to be less stable: there is efficient extra stabilization of the A isomers of protonated KG with respect to those of GG. While in GG the Ax structures (including A1, the global minimum) contain only one strong H-bond involving the positively charged amine group, the lowest energy Ex and Ax structures of KG are stabilized by two strong H-bonds. In contrast, the Ox structures cannot take advantage of the larger number of H-bond acceptor groups in KG compared to the case of GG, because the protonated Oamide can participate only in one H-bond as a

Table 3. ZPE-corrected barrier heights in kcal/mol of the PT reactions relative to the connected lower (left columns) and higher (right columns) energy minimum and unimolecular rate constants for reaction starting from the respective minima calculated at selected internal energies by the RRKM method. (Note that lower estimates for the rate are given for processes with a ‘negative’ barrier, see text for details.) Lower minimum A01 A04 A10 A13 E02 O01 A13 O01 I03 E10 E12 A21 A20 A22 a

Barrier

log(k1ev)

log(k32)

Log(k2ev)

TS

Barrier

log(k1ev)

log(k32)

log(k2ev)

Higher minimum

0.1 0.1 0.2 0.4 26.4 0.5 10.7 0.2 3.8 17.9 19.2 6.1 13.6 6.6

v.f.a v.f.a v.f.a v.f.a Ð 12.2 3.7 v.f.a Ð Ð Ð Ð Ð Ð

v.f.a v.f.a v.f.a v.f.a Ð 12.4 7.6 v.f.a 9.7 3.1 1.7 7.3 3.8 Ð

v.f.a v.f.a v.f.a v.f.a 3.5 12.5 9.4 v.f.a 11.3 5.6 4.5 9.5 7.3 9.3

A01_E16 A04_E17 A10_E18 A13_E20 E02_O14 O01_A15 A13_O08 O01_I01 I03_O16 E10_D01 E12_D02 A21_D01 A20_D02 A22_D03

1.5 1.6 1.2 1.3 0.1 0.4 1.1 0.4 0.4 1.3 1.4 5.4 4.1 3.2

v.f.a v.f.a v.f.a v.f.a Ð 12.2 v.f.a v.f.a Ð v.f.a Ð Ð Ð Ð

v.f.a v.f.a v.f.a v.f.a v.f.a 12.3 v.f.a v.f.a v.f.a v.f.a v.f.a 8.2 9.1 Ð

v.f.a v.f.a v.f.a v.f.a v.f.a 12.4 v.f.a v.f.a v.f.a v.f.a v.f.a 10.3 10.9 10.7

E16 E17 E18 E20 O14 A15 O08 I01 O16 D01 D02 D01 D02 D03

Very fast.

Copyright # 2001 John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

The lowest energy species protonated on the amide N, D01 (Fig. 5), has 22.5 kcal/mol energy relative to E01. This species is stabilized by the Namide-H¼Na and Namide-H¼Ne H-bonds and charge-transfer interaction between Ocarbonyl and the protonated amide bond. The main difference between the most stable amide-nitrogen-protonated isomers of GG and KG is that the latter is stabilized by an extra Hbond between the protonated amide nitrogen and the eamine group. Otherwise, the Dx structures of these two species are very similar to each other. The additional stabilizing H-bond is probably the reason why, compared to the Ox structures, D01 of KG has lower energy than the corresponding D1 species of GG (11.2 kcal/mol vs. 15.7 kcal/mol). However, as compared to the global minimum, the relative energy of D01 is larger than that of the similar D1 species of GG at 16.9 kcal/mol.20 The D02 conformer has a similar structure to D01, and accordingly its relative energy is just a little higher at 24.0 kcal/mol. The other Dx structures have much higher relative energies. This finding can be explained by the fact that in those structures a positively charged amide nitrogen is stabilized by weaker Hbonds than those of D01 and D02. For example, conformer D03 has 30.8 kcal/mol relative energy, and in this structure the amide nitrogen is involved in the H-bonds Namide-H¼Na and Namide-H¼Ocarbonyl (an Ocarboxyl OH-H¼Ne H-bond is also present).

Proton mobility in protonated lysylglycine

In our previous studies19,20 on protonated N-formylglycinamide, N-formylglycyl glycinamide and diglycine, we found that most of the internal rotations connecting different conformers of a given protonated structure are very fast. They occur on a much shorter time-scale than the MS experiments even at relatively low internal energy values. Therefore, during the evaluation of proton mobility for protonated KG, we focused our attention on the PT reactions and did not deal in detail with the very large number of internal rotations connecting the different conformers of a given protonated structure. The calculated total and relative energies of the investigated PT transition states (TS) of protonated KG are collected in Table 2. The barrier heights are measured from the global minimum of the PES of protonated KG, isomer E01 in our nomenclature. The ZPEcorrected barrier heights relative to the respective minima are collected in Table 3, together with the rate constants calculated using the RRKM method at internal energies of 1eV (23.06 kcal/mol), 2eV (46.12 kcal/mol), as well as at 32.1 kcal/mol, which is the threshold energy of the lowestenergy fragmentation pathway of protonated KG (see below). The data presented in Table 3, however, require a caveat. Our combined DFT/RRKM model of proton transfer reactions assumes that the investigated protonation sites are separated by substantial barriers, i.e. the relative energy of any TS is higher than those of either of the minima connected by that TS. In such a case one could apply ab initio data to obtain kinetic information on the given process by using the RRKM formalism which provides internal-energydependent unimolecular rate constants. The potential surface of protonated KG involves a number of potential Copyright # 2001 John Wiley & Sons, Ltd.

1463

minima which, formally, prove not to be minima when the ZPE of the vibrational modes is also taken into account. This means that the region of the configuration space corresponding to such isomers is not a basin in which the system can be trapped. Instead, formally it can be considered as a shoulder or small plateau on the wall of a deeper well. We think, however, that as the minima were unequivocally detected by the minimum search, it is better to consider them as distinct minima when their energy differs significantly because it seems to be reasonable that such structures exist, and in addition the vibrational ZPE, which is conventionally calculated from separable harmonic modes, is not accurate enough to guarantee full accuracy. (However, more accurate calculation of ZPE by including factors like anharmonicity, mode-mode coupling etc., is far beyond the reach of any theoretical method for a molecule of this size). It is a question, then, of how one can calculate the rate of a reaction to and from such a poorly defined isomer: the quantum chemical calculation yields a `negative barrier' for such reactions, which is clearly nonsense. We distinguish two types of such cases: 1 The two minima on the PES have similar relative energies and the relative energy of the TS connecting them becomes smaller when ZPE correction is applied than those of the minima themselves. In this case the proton is shared by the two protonation sites and essentially oscillates between the bridgehead atoms in a shallow potential energy valley instead of a double well potential corresponding to two clearly separated isomers. This means that the concept of reaction rate does not apply to the virtual reactions. When the PES of a protonated peptide is represented by many such valleys, the most important transitions are internal rotations connecting these shallow valleys along perpendicular directions. Our previous studies19,20 showed that these internal rotations around flexible bonds are fast. 2 One of the corresponding minima is at a much lower energy level than the other, and the relative energy of the TS connecting them becomes smaller than that of the higher minimum after the ZPE correction is applied. This situation corresponds to the physical picture in which the proton is shared by the two protonation sites, but it is more localized on the lower energy site and can be found only with a small probability in the proximity of the higher energy site. Being a fast process, an internal rotation can occur while the proton is close to the higher energy protonation site and this process can result in a structure in which the proton is really localized on that site. It is also possible that a real chemical reaction can occur while the shared proton stays close to the higher energy protonation site, which is usually more reactive than the low-energy species. As we assume here that the higher energy minimum exists, but we obtain from the quantum chemical calculation a `negative barrier height', we can only estimate the rate. We assume that, although the TS after ZPE correction does not correspond to an energetic barrier, it still can present a `bottleneck' in phase-space on the way from one protonation site to another. As the ZPEcorrected barrier is always just a little below the energy of the higher minimum (and the difference is actually in the Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1464 I. P. Csonka et al.

order of the accuracy of the calculation itself), in the RRKM calculations we consider the energy of the TS to be equal to that of the higher minimum. This means that reaction from the higher energy minimum is extremely fast while a reasonable estimate for the rate of the reverse process can be obtained. After thus describing the underlying theoretical background, we turn to the detailed discussion of proton traffic in protonated KG. The most favored protonation sites of protonated KG are the e- and the a-amino groups, as discussed in the previous section. Although the relative energies of the Ax structures tend to be slightly higher then those of the Ex structures, the differences are not very large. We found some low barriers corresponding to TSs A01_E16, A04_E17, A10_E18, A13_E20 which, in principle, connect the Ax and Ex basins, but each proves to disappear after the ZPE correction is applied. This means that the proton motion in the molecule is so fast that there seems to be no prolonged existence of either of these isomers. In other words, the Ax to Ex and the reverse reactions are very fast even at low internal energies. As we showed earlier,20 the Ox structures play an important role in the mobility of the added proton, since most probably the amide oxygens are heavily involved in the transfer of the proton along the peptide chain. There is a TS connecting the Ex/Ax manifold of minima with the Oprotonated isomers: passage through TS A15_O01 is extremely fast in both directions even at low internal energies. We located another barrier (E02_O14) which connects E02 with a high-lying Ox conformer (a shoulder in terms of the previous discussion), but the corresponding transition is quite slow according to the results of the RRKM calculations and cannot be considered to be an efficient pathway. It is possible that there are further fast channels connecting the Ex/Ax and the Ox conformers. The lowest-lying conformers of the Ix and Ox isomers (both at around 11.5 kcal/mol above E01) are connected by TS I01_O01. The barrier to this reaction disappears after correcting for ZPE indicating a very fast transition for which we did not carry out RRKM calculations. Other Ix_Ox transitions start at significantly larger energies (for example, TS I03_O16 is at 26.9 kcal/mol relative energy), and become fast only at higher internal energies. Considering both the kinetic and energetic information available for the Ix species, one can conclude that formation of these species is very likely in mass spectrometers operated under the most common conditions. As mentioned above, fragmentation of backbone amide bonds requires protonation of the corresponding amide nitrogen. Therefore, PT reactions connecting the Dx valleys with others play an important role in formation of the reactive configuration for these decompositions. Our theoretical modeling suggests that direct Ex $ Dx reactions have a large activation energy and are quite slow even at large internal energies. For example, E10_D01 connects the Ex and Dx wells at 24.5 kcal/mol relative energy, and the rate constant is as small as 4  105 s 1 even at 2 eV internal energy. The transition between the Ax and Dx structures is more facile, mainly due to the higher internal energy of the Copyright # 2001 John Wiley & Sons, Ltd.

respective Ax structures. The Ax → Dx reactions became relatively fast at less than 1.5 eV. For example, the relative energy of the A21_D01 TS is 27.9 kcal/mol above E01 but only 6.1 kcal/mol above A21, so that the rate constant is 2  107 s 1 already at 32.1 kcal/mol internal energy. Comparing the accessibility of the amide nitrogen protonation site in protonated GG and KG, one notes that transfer of the added proton to this site needs significantly larger energies, and is correspondingly much slower for the case of KG. This is clearly due to the presence of a basic side-chain group capable of forming low-energy amino-protonated isomers in KG, which makes the relative energy of the amide-nitrogenprotonated species rather high compared to the analogous protonated GG. This has a significant effect on the fragmentation of lysine-containing peptides, making these molecules more resistant to backbone fragmentation compared to other non-basic peptides.

Main fragmentation pathways of protonated KG

Protonated KG fragments mainly on two dissociation pathways on the metastable ion time scale.23 These two pathways are also important for other lysine derivatives and lysinecontaining peptides. The first reaction is loss of ammonia (57% of the total ion signal in the metastable ion fragmentation spectrum), while the second one is formation of an ion with m/z 129 (40% of the total ion signal). Interestingly, neither the a1 ion (NH2‡=CH-(CH2)4-NH2 immonium ion) nor the y1 ion (protonated glycine) are formed during lowenergy dissociation of protonated KG, although these types of fragment ions are prominent products in the mass spectra of other dipeptides.28±30 15 N-Labeling experiments22 on the fragmentation of protonated lysine have shown that the ammonia lost specifically involves the nitrogen of the side chain. This reaction produces an ion with m/z 130, and this ion fragments further resulting in an ion with m/z 84. Based on the stoichiometry and the kinetic energy release (KER) of this second fragmentation (m/z 130 → 84), Yalcin and Harrison suggested that the ion with m/z 130 in the mass spectrum of protonated lysine and other lysine derivatives may be a protonated pipecolic acid23 (Scheme 2). The calculated absolute and relative energies of the structures occurring on the fragmentation paths leading to loss of ammonia and formation of an ion at m/z 129 from protonated KG are collected in Tables 4 and 5, and the results are visualized in Fig. 6. The calculated rate constants obtained by RRKM theory are shown in Table 6 and in Fig. 7. Based on the experimental observation of Dookeran et al.22 on protonated lysine, we assumed that ammonia loss from

Scheme 2. Protonated pipecolic acid, the supposed product ion of the ammonia loss from the side chain of protonated lysine. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

1465

Table 4. Calculated total and relative energies (including ZPE correction) of the TSs and products (molecular complexes) of the investigated fragmentation reactions of protonated KG calculated at the B3LYP/6-31G(d) level of theory. The relative energies are measured to the global minimum of protonated KG (E01) TS

Energy (a. u.)

TS_NH3_Na TS_NH3_O

705.391923 705.38536

TS_B1_A TS_B1_B

705.382850 705.382319

TS_A1Y1

705.358446

Rel. Ezpe (kcal/mol)

Complex

Ammonia loss 32.1 CO_NH3_Na 35.9 CO_NH3_O b1 ion formation (caprolactam formations) 39.3 CO_B1_A 39.4 CO_B1_B `a1-y1' pathway 51.0 CO_A1Y1

Energy (a. u.)

Rel. Ezpe (kcal/mol)

705.434911 705.414198

5.9 17.6

705.420364 705.420103

15.8 15.9

705.391440

29.0

Table 5. Calculated absolute and relative (kcal/mol, including ZPE correction) energies of the neutral and ionic fragmentation products of protonated KG resp. complex

a.u

CO_NH3_O

Prot. pipecolic acid der. 648.872782 Prot. 7-membered ring 648.852710

CO_B1_A CO_B1_B

a-amino-e-caprolactam 420.976601a 420.976601a

CO_NH3_Na

a

a.u

Rel. to E01

Ammonia loss NH3 56.547948 NH3 56.547948 b1-ion formation Glycine 284.420042 284.420042

Rel. to the complex

Rel. to the resp. TS

13.3

7.4

18.8

25.0

7.4

11.0

29.6 29.6

13.8 13.7

9.8 9.8

Protonated at the amine group.

Table 6. ZPE-corrected barriers of the TSs of the investigated fragmentation reactions relative to the reactant minimum and unimolecular rate constants at selected internal energies calculated by the RRKM method Minimum

TS

E11 E03

TS_NH3_Na TS_NH3_O

D17 D01

TS_B1_A TS_B1_B

D01

TS_A1Y1

Barrier

Ammonia loss 28.6 2.7 38.5 Ð b1-ion formation (caprolactam formations) 2.6 10.1 16.9 3.6 `a1-y1' pathway 28.5 Ð

protonated KG involves the e-amino group. That is, our search for various minima and transition structures occurring during loss of ammonia from protonated KG was initiated from species protonated at the side chain of the lysine residue (Ex structures). The most straightforward possibility would be a direct cleavage of the Ce-Ne bond, which results in a carbocation. The calculations showed, however, that this process requires a substantial amount of energy (more than 60 kcal/mol relative to the global minimum E01). In addition, the carbocation structure is not stable: it is transformed to isomers with a cyclic structure, without the need to pass a barrier, in intramolecular reactions in which one of the nucleophilic groups of the molecule (for example a-amine or amide O) attacks the carbocation center and forms the ring. This led us to assume that bond rupture between Ce and Ne atoms and attack of a suitable nucleophile on the positive center occur in a concerted manner (SN2-type reaction, Scheme 3). There are four nucleophilic groups in protonated KG, which can be Copyright # 2001 John Wiley & Sons, Ltd.

Log(k2eV)

Log(k3eV)

Log(k4eV)

6.3 4.0

8.0 6.1

11.1 7.5

11.3 8.8

4.5

7.7

involved in such a reaction: (a) the a-amino group, (b) the amide O, (c) the carbonyl oxygen of the carboxyl group, and (d) the hydroxyl oxygen of the carboxyl group. According to our calculations, the last two cases can be ruled out as important processes, because the TSs and the products of these reactions have very high relative energies (>55 kcal/ mol). Hence we will consider only those cases where the aamino group or the amide oxygen substitutes Ne, resulting in

Scheme 3. Ammonia loss from the protonated lysine side chain through an SN2-type reaction. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1466 I. P. Csonka et al.

Figure 6. Calculated energetics of the investigated fragmentation reactions of protonated KG.

Figure 7. RRKM rate constants of the investigated fragmentation reactions of protonated KG. Copyright # 2001 John Wiley & Sons, Ltd.

Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

1467

Scheme 4. SN2-type substitution reaction on Ce involving the a-amino group resulting in a protonated pipecolic acid derivative and ammonia.

Scheme 5. SN2-type substitution reaction on Ce involving the amide O resulting in a protonated seven-membered ring (imidocaprolactone) derivative and ammonia.

The transition structures obtained for these pathways, denoted as TS_NH3_Na for path a and TS_NH3_O for path b, are shown in Figs 8 and 9, respectively. In the case of TS_NH3_Na the nitrogen of the a-amino group attacks the Ce atom. As seen in Fig. 8, the Ne±Ce distance is lengthened Ê ) while N1 gets close to Ce (2.140 A Ê ) and the CeH2 (2.067 A group becomes planar. In the case of TS_NH3_O the amide oxygen attacks the Ce center (Fig. 9). Similarly to the TS on Ê ) and planarity of path a, cleavage of the Ce-N bond (2.133 A

a molecular complex formed by the leaving ammonia and a pipecolic acid derivative (path a, Scheme 4) or a sevenmembered ring derivative (path b, Scheme 5), respectively.

Figure 10. Equilibrium structure of CO_NH3_Na.

Figure 8. Transition structure TS_NH3_Na.

Figure 9. Transition structure TS_NH3_O. Copyright # 2001 John Wiley & Sons, Ltd.

Figure 11. Equilibrium structure CO_NH3_O. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1468 I. P. Csonka et al.

the CeH2 group show direct evidence for an intramolecular SN2-type reaction. Also, the short distance between atoms Ê ) indicates that formation of the sevenOamide and Ce (1.993 A membered ring is nearly completed. According to an investigation of the geometries on the product wing of the IRC, the products of these reactions are molecular complexes formed by ammonia and a pipecolic acid derivative (CO_NH3_Na, Fig. 10) along path a, and an aminocaprolactone imide (CO_NH3_O, Fig. 11) along path b. The relative energy (5.9 kcal/mol) of CO_NH3_Na is lower by approximately 12 kcal/mol than that of CO_NH3_O (17.6 kcal/mol). This difference is due to the fact that CO_NH3_Na is a pipecolic acid derivative protonated at a secondary amine. On the other hand, in CO_NH3_O, protonation occurs on an imide nitrogen. The relative energies of TS_NH3_Na and TS_NH3_O are very close to each other at 32.1 and 35.9 kcal/mol, respectively. However, the unimolecular rate constants calculated by RRKM theory show that reaction via path a is faster, indicating that formation of a pipecolic acid derivative is more probable in the mass spectrometer. Both reactions are relatively slow and reach the ms characteristic time at approximately 3 and 4 eV internal energy, respectively. An experimental confirmation of the theoretical prediction can be based on the difference of the products of paths a and b: further fragmentation of the pipecolic acid derivative will be different from that of the caprolactone derivative. This means that an MS/MS experiment on the fragment ion of the ammonia loss can decide which path dominates the ammonia loss of lysine derivatives, or both can play a role in this process. Path a corresponds to the fragmentation pathway suggested by Yalcin et al., namely, that the ion at m/z 130 is a pipecolic acid derivative. On the other hand, reactions similar to that including TS TS_NH3_O can also play a major role in the substantial ammonia loss observed for lysine derivatives lacking the aamino group, such as Ac-Lys-OH (11% in metastable fragmentation23) or Ac-Lys-OMe (32% in metastable fragmentation23). This latter type of reaction can also be involved in the ammonia loss of such peptides, where the lysine residue is not positioned at the N-terminus but somewhere in the chain or at the C-terminus (for example a peptide resulting from a tryptic digest). The second main fragmentation pathway of protonated KG results in a b1 ion (m/z 129) (40% relative intensity in the metastable ion spectrum23). It is well known that aaminoacylium ions are unstable and exothermically eliminate CO.31 In fact, there is only one known example for an detectable b1 ion, namely, that formed from an N-terminal methionine residue of peptides, which is supposed to have a cyclic structure involving the S atom of the methionine side chain.32 On the other hand, bn (n 2) ions of larger peptides are known to contain a protonated oxazolone-type ring.5,31 Taking into account this information, and the fact that the m/z 129 ion derived from protonated KG is stable, one can expect that it has a structure other than an acylium ion. Yalcin and Harrison23 showed that the metastable fragmentation of the ion with m/z 129, and also the kinetic energy release values measured from the metastable peaks, are very similar to those of protonated a-amino-e-caprolactam. ThereCopyright # 2001 John Wiley & Sons, Ltd.

Scheme 6. b1-ion formation resulting in a molecular complex of neutral glycine and protonated a-amino-e-caprolactam.

fore, these authors concluded that the ion with m/z 129 is an a-amino-e-caprolactam formed by a ring-closure coincident with the expulsion of a neutral glycine23 (Scheme 6). Since substantial evidence exists5,29,30 that cleavage of amide bonds requires protonation on the amide nitrogen, we initiated our search for the pathway leading to formation of the protonated a-amino-e-caprolactam from the Dx species. These calculations resulted in two TSs, namely TS_B1_A (initiated from D01) and TS_B1_B (initiated from D17) shown in Figs 12 and 13, respectively. Protonation on the amide nitrogen makes the carbon of the amide bond even more positive than the neutral counterpart. In this way, this

Figure 12. Transition structure TS_B1_A.

Figure 13. Transition structure TS_B1_B. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

Figure 14. Equilibrium structure of D17, the reactant of the fragmentation reaction over TS_B1_A.

carbon becomes rather electrophilic and can be attacked by nearby nucleophilic groups. In a similar situation most of the protonated dipeptides29,30 fragment via concerted cleavages of the amide and the Ca-Camide bonds, since there is no nucleophilic center in the proximity of the protonated amide bond which could stabilize the large partial positive charge by ring formation. However, in the case of protonated KG, the e-amino group is close enough to the protonated amide bond, and can be involved in such a stabilization (for example see Figs 5 and 14). The analysis of the IRCs shows that reactions via TS_B1_A and TS_B1_B start from two different isomers, D01 and D17, respectively, and result in two ion-molecule complexes of a caprolactam derivative and glycine, CO_B1_A and CO_B1_B, respectively (Figs 15 and 16, respectively). It is very interesting that, although the structures of CO_B1_A and CO_B1_B are quite different, their relative energies are very similar at 15.8 and 15.9 kcal/ mol, respectively. On the other hand, the energies of the initial conformers are significantly different, i.e. D17 is 14.2 kcal/mol less stable than D01. The relative energies of TSs TS_B1_A and TS_B1_B are essentially the same at 39.3 and 39.4 kcal/mol, respectively. This, however, does not mean equal reaction rates. If one assumes that the reaction starts from the initial isomers

Figure 15. Equilibrium structure of CO_B1_A. Copyright # 2001 John Wiley & Sons, Ltd.

1469

Figure 16. Equilibrium structure of CO_B1_B.

described above, a substantial difference is observed between the unimolecular rate constants calculated for the two pathways. For the entire energy interval investigated the pathway including TS_B1_B is much faster than that including TS_B1_A. This difference can be explained by the fact that, while the e-amino group is heavily involved in the stabilization of the protonated amide bond for D01 (Fig. 5), this is not the case for D17 (Fig. 14), where the e-amino group is located above the plane of the amide bond taking part in stabilization by a weaker charge-transfer interaction. One has to note, however, that the relative energy of D17 is definitely higher than that of D01. Therefore, it is very likely that the population of the D17 isomer is much lower than that of D01 in mass spectrometers operated under usual conditions. On the other hand, if the fast fragmentation reaction empties this state but the PT reactions can feed it from other isomers, i.e. if the fragmentation is rate determining, then this pathway may be an efficient channel for the formation of a caprolactam derivative. This may happen at high excitation energies. If, however, re-population of D17 is relatively slow, as is indicated in the calculations presented in the previous section, the resulting rate of reaction is slower so that the ammonia-loss reaction discussed above can successfully compete with caprolactam formation and result in comparable abundance ratios. Interestingly, the y1 ion (protonated glycine) is not present in the metastable ion spectrum of protonated KG.23 In recent papers we investigated the mechanism of y1-ion formation from protonated GG29 and other dipeptides.30 A new pathway (`a1-y1') was proposed which leads to integrated formation of a1 and y1 ions, the ratio of which depends on the composition and the energy distribution of the fragmenting species for a particular dipeptide. The first TS in this

Scheme 7. First step of the ‘a1-y1’ fragmentation pathway of dipeptides.29,30 Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1470 I. P. Csonka et al.

Scheme 8. Proton transfers in molecular complexes CO_B1_A.

mechanism corresponds to concerted cleavage of the amide and the Ca-Camide bonds resulting in a trimer of a protonated imine, the C-terminal amino acid and CO (Scheme 7). After loss of CO, a proton-bound dimer of the a1 ion and the Cterminal amino acid is formed. Under low-energy conditions the lifetime of this dimer is long enough that numerous proton transfers can take place between the imine and amino acid in the dimer. There is a pathway through which the dimer can dissociate without passing a barrier in the next step (depending on the energy of the ion, the PA of its monomers, etc.) to form either a1 or y1 ions. We explored the initial step of this `a1-y1' pathway in the case of protonated KG as well. The calculation revealed that, on the reaction path starting from the D01 structure, the corresponding saddle point, TS_A1Y1, with a relative energy of 51.0 kcal/mol, is the highest barrier among all the TSs we found to lead to fragmentation. The product of this reaction also has a high relative energy of 29.0 kcal/mol (Table 4, Fig. 6). Furthermore, according to the RRKM calculations, this is the slowest fragmentation reaction at low internal energies, and it approaches the rate of the concurrent reactions only at very high internal energies (Table 6, Fig. 7). We did not explore further this complicated reaction pathway, as the key steps are presumably similar to those found in the case of protonated GG.29 It is clear that the formation of the b1 ion in the case of protonated KG is favored compared to the `a1-y1' pathway at low internal energies, totally depleting the a1 and y1 channels. However, a generalization of the `a1-y1' mechanism to the case of protonated KG, where the complex formed in the fragmentation reaction containing a protonated aamino-e-caprolactam instead of the a1 ion of other peptides, leads one to expect that y1-ion formation could also take place via proton transfer from the amide-nitrogen-protonated a-amino-e-caprolactam to the neutral glycine in CO_B1_A and CO_B1_B (Figs 15 and 16, respectively). At first sight this process seems to be very favorable because an amide N protonation site, based on experience on the relative stability of isomers of dipeptides, is expected to be less favorable than the amino group of G. However, a closer Copyright # 2001 John Wiley & Sons, Ltd.

look at the structures CO_B1_A and CO_B1_B (Figs 15 and 16, respectively) reveals that there are serious reasons for the a-amino-e-caprolactam to remain protonated. First, the proton transfer to glycine in CO_B1_A and CO_B1_B would result in a trans-amide bond in the a-amino-e-caprolactam, which causes a high ring strain in the seven-membered ring. Second, the other proton of the protonated amide nitrogen can transfer to the a-amino group of the a-amino-ecaprolactam itself, resulting in a cis-amide bond causing much less ring strain (Scheme 8). Our calculations indicate that decomposition of CO_B1_A to form the amide-nitrogen-protonated a-amino-e-caprolactam plus G requires 28.0 kcal/mol energy, while decomposition to trans-a-amino-e-caprolactam plus protonated G requires 34.6 kcal/mol. It is striking that the third pathway, i.e. formation of the amino-protonated a-amino-e-caprolactam plus G, requires only 13.9 kcal/mol energy. This energetic argument explains why y1-ion formation does not take place for protonated KG under metastable conditions. However, it is possible that a protonated dipeptide KX, having a C-terminus amino-acid residue with higher proton affinity than that of glycine, could fragment to produce y1 ions. For example, the proton affinity of valine can be large enough to compete with the aamino site of the a-amino-e-caprolactam. An MS experiment on protonated lysylvaline (KV) would be decisive on this question. Finally, we compare the energetics (Fig. 6) and kinetics (Fig. 7) of the main fragmentation pathways of protonated KG leading to ammonia loss and formation of the protonated a-amino-e-caprolactam (b1 ion). As can be seen in Fig. 6, the pathways leading to loss of ammonia are energetically favored compared to those resulting in formation of protonated a-amino-e-caprolactam (b1 ion). On the other hand, unimolecular rate constants calculated using RRKM theory formally prefer formation of the a-amino-e-caprolactam (b1 ion) (Fig. 7). However, pathways leading to the formation of the b1 ion are initiated from Dx conformers, while ammonia loss occurs from the Ex isomers. In protonated KG the PT processes re-populating the Dx isomers are relatively slow so that, at low excitation, this may be the rate-determining step. Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

Proton mobility and fragmentation pathways of protonated Lys-Gly

Scheme 9. Proton traffic in protonated lysylglycine involving the most important protonation sites (E, A, O and D, see text). Numbers set in roman denote the energies of the minima, while numbers in italic denote the height of the barriers between them (all values calculated at the B3LYP/6-31G(d) level of theory and are ZPE-corrected). Bold letters are the acronyms for the given protonation sites in our nomenclature.

In the present case we do not have all the barriers on all possible paths connecting all isomers that lead from the Ex/Ax basin to the Dx basin, so that we cannot provide a numerical estimate of the relative importance of ammonia loss and caprolactam derivative (b1 ion) formation. However, the energetic and kinetic data obtained in our calculations do provide a qualitative understanding of the main fragmentation pathways for protonated KG.

CONCLUSIONS The theoretical model of the dynamical processes in protonated KG (Scheme 9) presented in this work involves a detailed understanding of the potential surface of the ion based on density functional theory (B3LYP/6-31G(d)) and estimation of the time-scale and relative rates of the isomerization and fragmentation processes using RRKM calculations. We found five structural isomers in which the proton is located on the amino nitrogen of the e-position (E), the a-position (A), on the carbonyl oxygen of the carboxyl group (O), on the N atom participating in the amide bond (D), as well as a tautomer of E (I). The structures and energies of numerous conformers of these were determined. The E and A isomers were found to be the most stable, with relatively small potential barriers separating the potential wells of their conformers as well as connecting the E and A basins. The O, D, and I isomers are in higher regions of the potential surface, and in many cases are dynamically not well defined because the potential barriers separating them from the deep A and E minima are very low. Our DFT/ RRKM modeling showed that the accessibility of the reactive amide-nitrogen-protonated species is less pronounced for protonated KG than for other peptides lacking basic residues. This observation is in line with the experimental results showing that lysine-containing peptides fragment less easily than protonated peptides lacking any basic AAs. Copyright # 2001 John Wiley & Sons, Ltd.

1471

The fragmentation reactions of protonated KG take place in very interesting complex, concerted processes. Direct chain breakage was found to be much less favored than the concerted processes resulting in cyclic products. One group of pathways leads to the loss of ammonia from the e-amino group and formation of either a protonated pipecolic acid derivative or a protonated imido derivative of an a-amino-elactone. The other group of decomposition pathways starts from the amide-nitrogen-protonated isomer, and results in the formation of a glycine molecule and a protonated aminocaprolactam (b1-ion). Interestingly, the `a1-y1' fragmentation pathway, which is important in protonated GG, does not play a role in the fragmentation of protonated KG. The theoretical calculations firmly identify the structures of the possible fragmentation products, and make possible the qualitative understanding of the metastable ion spectrum of protonated KG.

Acknowledgements

Financial support of this work by the Hungarian Scientific Research Fund (G.L, OTKA T 22824) is gratefully acknowledged.

REFERENCES 1. Alexander AJ, Thibault P, Boyd RK. Rapid Commun. Mass Spectrom. 1989; 3: 30. 2. Ballard KD, Gaskell SJ. Int. J. Mass Spectrom. Ion Processes 1991; 111: 173.  , Wysocki VH, Mayer I. J. Am. Soc. Mass Spectrom. 3. Somogyi A 1994; 5: 704.  , Dongre AR, Wysocki VH. Anal. 4. McCormack AL, Somogyi A Chem. 1993; 65: 2859. 5. Paizs B, Lendvay G, VeÂkey K, Suhai S. Rapid Commun. Mass Spectrom. 1999; 13: 525. 6. Hunt DF, Yates III JR, Shabanowitz J, Winston S, Hauer CR. Proc. Natl. Acad. Sci. USA 1986; 83: 6233. 7. Biemann K, Martin S. Mass Spectrom. Rev. 1987; 6: 75. 8. Biemann K. Biomed. Environ. Mass Spectrom. 1988; 16: 99. 9. Biemann K. In Mass Spectrometry Methods in Enzymology, vol. 193; J. A. McCloskey, Ed.; Academic Press: San Diego, CA, 1990; chaps. 18,25. 10. Biemann K. In Biological Mass Spectrometry Present and Future; T. Matsuo, R. M. Caprioli, M. L. Gross, Y. Seyama, Eds.; Wiley: New York, 1993. 11. Tsang CW, Harrison AG. J. Am. Chem. Soc. 1976; 98: 1301. 12. Johnson RS, Martin SA, Biemann K. Int. J. Mass Spectrom. Ion Processes 1988; 86: 137. 13. Mueller DR, Eckersley M, Richter W. Org. Mass Spectrom. 1988; 23: 217.  , Wysocki VH. J. Am. Chem. 14. Jones JL, Dongre AR, Somogyi A Soc. 1994; 116: 8368. 15. Johnson RS, Krylov D, Walsh KA. J. Mass Spectrom. 1995; 30: 386.  , Wysocki VH. J. Am. Chem. 16. Dongre AR, Jones JL, Somogyi A Soc. 1996; 118: 8365. 17. Harrison AG, Yalcin T. Int. J. Mass Spectrom. Ion Processes 1997; 165: 339. 18. Vaisar T, Urban J. J. Mass Spectrom. 1998; 33: 505. 19. Csonka IP, Paizs B, Lendvay G, Suhai S. Rapid. Commun. Mass Spectrom. 2000; 14: 417. 20. Paizs B, Csonka IP, Lendvay G, Suhai S. Rapid Commun. Mass Spectrom. 2001; 15: 637. 21. Paizs B, Suhai S. under publication. 22. Dookeran NN, Yalcin T, Harrison AG. J. Mass. Spectrom. 1996; 31: 500. 23. Yalcin T, Harrison AG. J. Mass. Spectrom. 1996; 31: 1237. 24. Paizs B, Szlavik Z, Suhai S, Hargittai B, Hruby VJ, Somogyi  . under publication. A 25. Gaussian 98, Revision A.5, Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, Zakrzewski VG, Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472

1472 I. P. Csonka et al. Montgomery JA Jr., Stratmann RE, Burant JC, Dapprich S, È, Millam JM, Daniels AD, Kudin KN, Strain MC, Farkas O Tomasi J, Barone V, Cossi M, Cammi R, Mennucci B, Pomelli C, Adamo C, Clifford S, Ochterski J, Petersson GA, Ayala PY, Cui Q, Morokuma K, Malick DK, Rabuck AD, Raghavachari K, Foresman JB, Cioslowski J, Ortiz JV, Stefanov BB, Liu G, Liashenko A, Piskorz P, KomaÂromi I, Gomperts R, Martin RL, Fox DJ, Keith T, Al-Laham MA, Peng CY, Nanayakkara A, Gonzalez C, Challacombe M, Gill PMW, Johnson B, Chen W, Wong MW, Andres JL, Gonzalez C, Head-Gordon M, Replogle ES. and Pople JA. Gaussian, Inc., Pittsburgh PA, 1998.

Copyright # 2001 John Wiley & Sons, Ltd.

26. Holbrook KA, Pilling MJ, Robertson SH. Unimolecular Reactions, Wiley: Chichester, 1996. 27. Forst W. Theory of Unimolecular Reactions, Academic Press: New York, 1973. 28. Cordero MM, Houser JJ, Wesdemiotis C. Anal. Chem. 1993; 65: 1594. 29. Paizs B, Suhai S. Rapid. Commun. Mass Spectrom. 2001; 15: 651. 30. Paizs B, Csonka IP, SchnoÈlzer M, Suhai S. under publication. 31. Yalcin T, Khouw C, Csizmadia IG, Peterson MR, Harrison AG. J. Am. Soc. Mass Spectrom., 1995; 6: 1165. 32. Tu Y-P, Harrison AG. Rapid. Commun. Mass Spectrom. 1998; 12: 849.

Rapid Commun. Mass Spectrom. 2001; 15: 1457±1472