Statistical Examination of the a and a + 1 Fragment Ions from 193 nm Ultraviolet Photodissociation Reveals Local Hydrogen Bonding Interactions
Dissociation of proteins and peptides by 193 nm ultraviolet photodissociation (UVPD) has gained momentum in proteomic studies because of the diversity of backbone fragments that are produced and subsequent unrivaled sequence coverage obtained by the approach. The pathways that form the basis for the production of particular ion types are not completely understood. In this study, a statistical approach is used to probe hydrogen atom elimination from a + 1 radical ions, and different extents of elimination are found to vary as a function of the identity of the C-terminal residue of the a product ions and the presence or absence of hydrogen bonds to the cleaved residue.
KeywordsUltraviolet photodissociation Tandem mass spectrometry a Ion Hydrogen-bonding
Advances in high throughput bottom-up proteomics approaches [1, 2] have been driven in part by improvements to established ion activation methods and developments of new methods [3, 4]. New activation techniques, such as electron-based [5, 6, 7, 8] and photoactivation methods [9, 10, 11], have resulted in the production of different types and distributions of diagnostic fragment ions than ones generated by conventional collisional activation methods. Ideally, for sequencing applications, fragmentation of peptides would be restricted to one or more of the three bonds of the peptide backbone (Cα–C, C–N, and N–Cα), leaving side-chains and modifications intact, aside from those that successfully differentiate isobaric leucine and isoleucine (w-type ions). These are the dominant cleavages promoted by collisional, electron-mediated, and photon-based activation for most peptides, and in silico-based database search algorithms have proven successful for identification of thousands of peptides in numerous proteomics applications. As a consequence of extensive studies involving examination of large populations of peptides, CID fragmentation is well understood and can be explained by the mobile proton and the pathways in competition models [12, 13], of which the latter largely encompasses the mobile proton model. These studies have uncovered several preferential backbone cleavages, typically observed at proline and aspartic acid, which can be used in a predictive manner to facilitate peptide sequencing [14, 15, 16, 17]. For example, upon collisional activation, proline-containing peptides exhibit enhanced cleavage of the amide bond N-terminal to the proline residues, a process that is particularly prominent for peptides in higher charge states . In contrast, preferential cleavage of the amide bond located C-terminal to acidic residues is exaggerated for peptides in low charge states [15, 16]. The statistical characterization of peptide fragmentation has been pursued in a number of other collisional activation studies [18, 19, 20, 21]. The inroads in deciphering the fragmentation of peptides upon collisional activation has spurred significant interest in understanding the underpinnings of other activation processes, including mechanistic aspects and the factors that contribute to variations in product ion abundances as a function of peptide length, sequence, and charge state.
UVPD is a relatively new activation approach that has rapidly developed in part because it yields unrivaled sequence coverage in large peptides and proteins [22, 23]. Activation by absorption of 193 nm ultraviolet photons promotes cleavage of all three backbone bonds, resulting in the formation of a and x (cleavage of the Cα–C bond), c and z (cleavage of the N–Cα bond), as well as b and y type ions. In addition to these six ion types, a + 1, x + 1, x + 2, and y – 2 ions are also prevalent due to various hydrogen atom migrations that can occur via radical-mediated pathways. Despite the promise of 193 nm UVPD, the mechanisms that govern the formation of these ions are not well understood. Consequently, algorithms and models that take advantage of 193 nm UVPD fragmentation are largely absent or underdeveloped.
The Reilly group has presented several elegant studies that examine UVPD mechanisms using fragmentation and modeling of small model peptides [24, 25, 26, 27, 28, 29, 30]. They have shown that activation of dipeptides by 157 nm UV radiation results in the elevation of an electron to a Rydberg orbital, which causes scission of the Cα–C bond by a Norrish type I mechanism to generate a + 1 ion and x + 1 radical ions depending on the location of an arginine residue [25, 26, 27]. Radical elimination via amide hydrogen or β-hydrogen elimination was shown to account for the a ions [25, 26, 27]. Here, we use lysN digestion to generate an ion series for a large number of peptides and use statistical analysis of the abundance of the a and a + 1 ions to explore hydrogen elimination from a+1 ions.
Materials and Methods
Methanol, DTT, iodoacetamide, hemoglobin, beta-lactoglobulin, carbonic anhydrase, myoglobin, lysozyme, albumin, beta-casein, and kappa-casein were obtained from Sigma-Aldrich (St. Louis, MO, USA). Guinea pig adrenocorticotropic hormone (ACTH 1-39, sequence SYSMEHFRWGKPVGKKRRPVKVYANGAEEESAEAFPLEF) was obtained from Bachem (Torrance, CA, USA). KGTDVLAWIRGCRL was obtained from ABI Scientific (Sterling, VA). Proteins lacking disulfide bonds were digested by incubation of LysN in 150 mM ammonium bicarbonate. The digestion was carried out using a 1:20 (m/m) ratio of enzyme to protein for 2 h at 55 C°. The resulting peptides were purified using Pierce (Grand Island, NY, USA) C18 spin columns, and the eluent was concentrated and resuspended in 5% acetonitrile for LC analysis. An E. coli lysate and proteins having one or more disulfide bonds were reduced using dithiothreitol (DTT) and acetylated using iodoacetamide prior to LysN digestion. The peptides included in the tabulation of a/a + 1 ions are summarized in Supplemental Table 1.
All peptides were analyzed on a Thermo Scientific Orbitrap Fusion mass spectrometer (San Jose, CA, USA) modified by addition of 193 nm excimer laser as described previously . In brief, an optical window was added to the manifold of the vacuum chamber at the back of the dual linear ion trap. To enhance overlap of the laser beam with the ion cloud and thus improve UVPD efficiency, the laser beam was focused using two plano-convex optical lenses. The pulsing of the laser was timed to the normal CID activation period in the high pressure region of the linear ion trap and the laser was fired at 500 Hz, which equates to 2 ms per pulse. Detection in the Orbitrap mass analyzer is on a longer timescale, requiring 100 ms or longer including ion transfer times. For the extensive peptide data acquisition needed for the statistical analysis of hydrogen atom elimination in a + 1 ions, liquid chromatography of LysN-digested peptides was performed using a Dionex nano-LC system coupled to the Orbitrap Fusion mass spectrometer. Peptides were separated on a 24 cm C18 column packed in-house and were eluted from the column using a 30 min gradient, during which the solvent composition was varied from 15% acetonitrile to 60% acetonitrile in water with 0.1% formic acid. Ions were generated by ESI via application of 1800 V to a solvent-voltage tee and were transmitted to the linear ion trap for UVPD. UVPD was performed using two pulses with 1.5 mJ delivered per pulse. Precursor and product ion spectra were acquired using the Orbitrap mass analyzer; a resolving power of 30 k was used for precursor ion spectra and a resolving power of 60 k was used for UVPD product ion spectra.
Peptide identifications were made manually from the single protein digests and using MaxQuant (v. 1.5) for the E. coli lysate, using a precursor tolerance of 10 ppm, a fragment tolerance 20 ppm, carbamidomethylation as a fixed modification, no variable modifications, and a false discovery rate of 0.01. The E. coli database was obtained from Uniprot. Determination of αa ([a]/([a] + [a + 1]) values for each peptide was undertaken by both manual and automated approaches. In the manual approach, theoretical isotope distributions for the fragment a ions were generated using Qualbrowser (Thermo Scientific, v. 2.2) and fit to the experimental a and a + 1 ion distributions using a solver function in Microsoft Excel. For the automated approach, mass spectra were converted to the mzML file format using MSconvert and a Python script based on the pymzML library was used to determine the abundances of all peaks belonging to a ion distributions . Using the optimization module of SciPy , the experimental abundances were fit to averagine a and a + 1 isotopic distributions generated with the fast Fourier transform method described by Palmblad et al . This provided an automated means to detect the peaks of the isotope distribution and correctly assign the abundances of the a and a+1 species. The algorithm can be found at: https://github.com/jarosenb/flask_ionratio_V2. The αa from the two methods was typically within ±0.05, and often less. The automated approach was used for the bulk of the fragments and manual interpretation used as needed for cleavages featuring an unusual standard deviation between replicates, the cause of which was typically overlapping isotope distributions from other ions.
Results and Discussion
Distributions of the Relative Abundances of an and an + 1 Ions Following Fragmentation C-Terminal to Alanine, Glycine, Leucine, Valine
From analysis of the UVPD mass spectrum of guinea pig adrenocorticotropic hormone (ACTH 1-39, sequence SYSMEHFRWGKPVGKKRRPVKVYANGAEEESAEAFPLEF), the relative abundances of an and an + 1 ions formed via cleavages of C–Cα backbone bonds were observed to vary with the identities of the C-terminal residue of the resulting a ions. Consequently, we applied a statistical approach to investigate the abundances of the a and a + 1 ions following UVPD fragmentation of a large series of peptides. The Reilly group used hydrogen-deuterium exchange (HDX) reactions of peptides to demonstrate that a ion formation upon 157 nm UVPD occurred via elimination of either the backbone amide hydrogen or the β-hydrogen of the C-terminal amino acid of the original a + 1 radical ion . Because some residues, such as Gly and Pro, lack either an amide hydrogen or side-chain hydrogens, a dependence was observed in the rates of the two hydrogen atom elimination reactions as a function of amino acid identity . In order to investigate the impact of hydrogen atom loss on the observed distribution of a and a + 1 fragment ions upon 193 nm UVPD, the portion of the a ions relative to the summed a and a + 1 abundances was tabulated as a function of the identity of the C-terminal amino acid of the a ion products from a series of 99 peptides obtained from an E. coli LysN digest; these peptides are listed in Supporting Information Table 1. Supporting Information Figure S1 shows the UVPD mass spectrum of peptide #1 from this dataset. Inclusion of each peptide into the dataset was contingent on several factors. First, the ability to quantify a and a + 1 abundances requires good isotopic resolution, a factor dependent on initial ion signal. Consequently, only the most abundant peptides were typically used. In order to avoid peptides with highly mobile protons, peptides having an over-abundance of charge relative to basic sites were not included to avoid problems arising from CID-type fragmentation associated with the presence of one or more mobile protons. Peptides having only a single basic residue were excluded in order to favor longer peptides with multiple basic sites. This resulted in all peptides having a charge state greater than or equal to two. By the same logic, only peptides with more than 13 residues were considered, as characteristic UVPD patterns (e.g., dominant production of a/x ions rather than conventional b/y ions) are only routinely observed for peptides of this size and larger . Finally, peptides with a basic amino acid at the C-terminus were excluded from the data set because of concerns that N-terminal product ions (such as a ions) would be suppressed to some extent and thereby obfuscate the data.
Upon assessment of the αa values for each amino acid and for specific peptide sequences, we observed that a reduction of the αa value commonly occurred when amino acids capable of hydrogen-bonding were found in positions adjacent to the C–Cα cleavage site along the peptide backbone. To explore this observation, the αa values were binned according to whether the amino acids flanking the terminal residue of the targeted a ion could engage in hydrogen-bonding (Asp, Glu, His, Lys, Asn, Gln, Arg, Ser, and Thr) or not (Ala, Phe, Gly, Ile, Leu, Met, Pro, Val, Cys, Tyr, and Trp). Note that Tyr and Trp have hydrogen-bonding side-chains, but cannot easily hydrogen-bond to the adjacent residue due to steric constraints. The segregated histograms for Leu, Val, Gly, and Ala based on hydrogen-bonding capability of the flanking residues are shown in the middle and bottom panels of Figure 1. Analogous histograms for the remaining 16 amino acids are shown in Supporting Information Figure S3. Comparison of each pair of segregated histograms suggests that the bimodal or asymmetric character of the original αa value histograms (Figure 1) can be largely accounted for by considering the influence of adjacent hydrogen-bonding residues. For example, the lower αa distribution is virtually eliminated in ULeuZ and UValZ sequences by controlling for aliphatic residues in the flanking positions (where U and Z are the flanking residues) and reduced by roughly 50% and 75% in UGlyZ and UAlaZ sequences, respectively. This suggests that hydrogen-bonding, either to the amide nitrogen or amide oxygen of the amino acid at which C-Cα cleavage occurs, results in a suppression of the hydrogen loss reaction that converts a + 1 to a ions. On average, this effect reduces the resulting αa value by 0.2–0.3.
It is interesting to note that many of the cleavages corresponding to the lower αa mode in Figure 1l were the consequence of AlaAla cleavages (i.e., the flanking amino acid was Ala, a non-hydrogen-bonding residue) or AlaPro or ProAla cleavages. It is possible that hydrogen-bonding from non-adjacent residues or the peptide backbone accounts for the reduction in the αa in these instances. Polyalanine has a high propensity to form helical secondary structures in the gas phase ; consequently, it is possible that small gas-phase helices form in the aliphatic regions having high Ala content. Similarly, four of the five instances of UGlyZ cleavages in which both U and Z were aliphatic and the observed αa was less than 0.1 had PheGlyAla or AlaGlyPhe motifs. Although the peptides in the present study were sprayed from denaturing conditions, there is substantial evidence that α-helices are unusually stable in the gas phase, and the enhanced hydrogen-bonding associated with the helical structure may result in the observed reduction in αa values [36, 37, 38, 39].
Hydrogen Atom Elimination from a + 1 Ions
Average and Standard Deviations of the αa Values of Each Distribution of Each Amino Acid Determined from GMM Fits to the αa Distributions (Using all Cleavages) of Each Amino Acid. The Number of Instances Included in Each Population is Listed in Parentheses After the Standard Deviation. The Amino Acids are Listed in Alphabetical Order Based on Their one Letter Code
Std Dev. (N)
Std. Dev. (N)
The majority of the amino acids, shown in Table 1, were best fit to bimodal distributions; however, there were several notable exceptions. Asp and Asn were best fit to unimodal distributions with an average αa of 0.49 and 0.49; however, separation of the histograms based on flanking hydrogen-bonding residues, shown in Supporting Information Figure S3, indicates the presence of lower αa distributions for both amino acids, centered approximately 0.2 below the higher αa distribution. Restriction of the algorithm to bimodal and higher order distributions resulted in the data being best fit to a bimodal distribution. As the bimodal distribution best reflects the histograms separated based on flanking hydrogen-bonding residues, Table 1 shows averages and standard deviations for the bimodal fit of Asp and Asn. A similar situation arose for Met, in which a unimodal fit was slightly more favorable than bimodal, and is likely a consequence of too few instances of Met cleavages, which is known to be a limitation for determination of bimodality by GMM, as inspection of the distribution suggests it is bimodal. Consequently, the Met data was also restricted to bimodal and higher order distributions, and Table 1 shows two means and standard deviations. His, Trp, and Pro were also best fit to unimodal distributions. Separation of the His and Trp data based on flanking residues also indicates unimodal distributions, although the number of instances of cleavage C-terminal to both residues may be too low to clearly define two distributions, if present. In contrast, Pro features more complexity, and it is possible that as many as four distributions are present. Again, however, the statistics are too limited to make any strong conclusions. It bears noting that Pro commonly features cis as well as trans isomerization of the amide bond, and the conformational consequences of these two isomerization states may contribute to the broad, ill-defined distributions. Iso and Thr were both best fit to trimodal distributions.
In order to illuminate the sources of hydrogen elimination in a + 1 ions in 193 nm UVPD, the αa values of Gly, Val, Leu, and Ala were examined. These four residues are aliphatic but feature differing numbers of β-hydrogens: 0, 1, 2, and 3, respectively. From inspection of Table 1 and Figure 1g, the average of the higher αa population of Gly is clearly both non-zero and significantly lower than the higher αa population of Val, Leu, or Ala. As Gly lacks a side-chain, the non-zero αa suggests that amide hydrogen elimination is an active pathway in 193 nm UVPD, as in 157 nm UVPD. This pathway is likely to occur for all a + 1 ions, regardless of the identity of the terminal residue. Interestingly, the Reilly group has shown a and a + 1 ion distributions for cleavages C-terminal to Gly [25, 42]; notably the αa for the a5 ion from RGGQGG was observed to be approximately 0.3, which is in excellent agreement with our results. Other cleavages, such as C-terminal to Phe, Glu, and others also seemed to be relatively consistent with the results from 193 nm UVPD . The difference between the average upper αa population of glycine and the average upper αa of Val, Leu, and Ala is considerable at 0.3–0.4, and indicates that amide hydrogen elimination is not the only route for conversion of a + 1 ions to a ions. The average of the higher αa populations for the Ala, Leu, and Val histograms are much more similar, 0.63, 0.70, and 0.75, respectively. ANOVA was carried out using the means, variances, and population weights from the GMM fits. The determined F-value of 28.85 is significantly greater than the 99% confidence critical value of 4.68, suggesting that the three populations are statistically different. The αa for Val, Leu, and Ala decreases as the number of β-hydrogens available for elimination increases. For example, Ala, which features the largest number of β-hydrogens, has the smallest αa, whereas Val features the largest αa. This is in direct contradiction of what would be expected for β-hydrogen elimination and suggests that one or more additional hydrogen elimination pathways may be present.
Akin to the observation from the Gly dataset that amide hydrogen elimination must be active in 193 nm UVPD, examination of dataset of Ala, which has a side-chain featuring exclusively β-hydrogens, is instructive. The mean of the higher αa distribution of the Gly data is 0.34, whereas the mean of the higher αa distribution of the Ala histogram is 0.63. It is clear from this difference that β-hydrogen elimination must occur and accounts for roughly half of the observed hydrogen elimination in Ala-terminated a ions. However, the previous comparison between the higher αa distributions of Ala, Leu, and Val suggests a third hydrogen transfer pathway also contributes to formation of a ions from a + 1 ions. Both Val and Leu have branched side-chains with six terminal hydrogens, respectively. Hence, the number of terminal hydrogens cannot explain the difference in the means of the αa values. Hydrogen atom elimination by the pathway shown in Scheme 1c requires a five-membered ring transition state for Val-terminated fragments and a six-membered ring transition state for Leu-terminated fragments. Given that five-membered rings feature faster formation kinetics, it is possible that the higher mean αa found for Val is a consequence of the kinetics of a pathway such as the one shown in Scheme 1c. Note that modeling is needed to confirm the details of this pathway. Interestingly, although this terminal hydrogen atom elimination pathway is relatively minor, contributing only about 15% to the total hydrogen atom elimination from Val side-chains, it has an important consequence. From inspection of Table 1, it is evident that relatively unique αa values are associated with each of the different amino acids. This is illustrated more clearly in bar graph format in Supporting Information Figure S7 (which shows each αa value for each amino acid) This semi-uniqueness of the αa may provide an avenue to pursue use of αa values as a validation strategy for de novo sequencing of 193 nm UVPD spectra. As there are currently no available de novo sequencing algorithms that are designed to handle the complexity of UVPD spectra, this may offer an opportunity to exploit the rich information that can be obtained from 193 nm UVPD while simultaneously taking advantage of the database-independent properties of a de novo approach.
To verify that a third elimination pathway contributes to a ion formation from radical elimination of a + 1 ions in 193 nm UVPD, ANOVA was performed on the higher αa population of Asp, Asn, and Leu, all of which feature identical branching structures but different terminal functional groups. The mean αa of the higher Asn and Asp populations are 0.60 and 0.59, respectively, and were not found to be significantly different from each other based on a Student’s t-test. However, ANOVA of the three populations [Leu (0.70), Asn (0.60), and Asp (0.59)] returned an F-value of 19.35, which is greater than the 99% confidence critical value of 4.70, suggesting that the average upper αa of Leu is significantly different from the average higher αa of Asp and Asn. Asp and Asn have similar terminal functional groups (amide versus a carboxylic acid) in which the one- to two-terminal hydrogen atoms present cannot undergo transfer by the pathway shown in Scheme 1c because Asp and Asn lack a γ-hydrogen. In contrast, the terminal methyl groups of the Leu side-chain feature six hydrogens, any of which can be transferred by a Scheme 1c-type pathway because of the available γ-hydrogen. The increased average αa of Leu relative to Asp and Asn can thus be accounted for by consideration of a third hydrogen elimination pathway from the terminal end of aliphatic resides, and supports a route such as the one shown in Scheme 1c. Note that additional modeling is needed to fully illuminate the source of the second hydrogen atom transfer. It is interesting to note that the mean αa values of the Asn and Asp datasets (higher αa populations) are less than that of the Ala population by approximately 0.05, suggesting that the third β-hydrogen atom of the Ala side-chain contributes approximately 0.05 to the observed hydrogen elimination.
It is interesting to consider the αa values in Table 1, particularly the higher αa distributions associated with the cleaving residue not being hydrogen-bonded, in the context of the stability of the a and a + 1 radical. In particular, the αa values for Tyr and Trp are higher than almost every other residue (0.86 and 0.84, respectively). Both of these residues are capable of amide and β-hydrogen elimination but cannot eliminate a terminal hydrogen because of the aromaticity of the side-chain. Loss of a β-hydrogen, however, results in the formation of an extended conjugation that likely stabilizes the a ion. This stabilization possibly drives the production of the a ion in fragments terminated by these two residues. In contrast, phenylalanine, which is also aromatic, features an αa value that is significantly lower (0.63). One possible explanation for this is that the phenylalanine radical is more stable and that the stability of the radical results in less favorable formation of the a ion product compared with Tyr and Trp.
Shifts in αa as a Function of Amide Hydrogen Bonding
In Figure 1, the abundance of the lower mode of the bimodal αa distributions from Ala, Val, and Leu was shown to be significantly reduced by considering only cleavages in which the adjacent residues were aliphatic (non-hydrogen-bonding). We therefore postulated that hydrogen-bonding to either the amide oxygen or amide nitrogen adjacent to the Cα–C cleavage site contributes to this reduction. From statistical analysis of the average higher αa values of Ala, Leu, Val, Gly, Asn, and Asp, hydrogen atom elimination is proposed to occur from the amide hydrogen, the β-hydrogen, and the terminal hydrogens of some amino acid side--chains. It is reasonable that hydrogen-bonding to the amide nitrogen could prevent amide hydrogen elimination to some extent, shifting the average αa value. However, it is also possible that hydrogen-bonding to the amide oxygen has an effect on the observed αa value. To explore whether amide nitrogen or amide oxygen (or both) hydrogen-bonding accounts for shifts in αa values, Δαa, the difference between the observed αa value for a given cleavage C-terminal to X in a sequence UXZ and the average αa value associated with the amino acid X, was calculated for all cleavages in which U and Z were Ser and Thr and X was Ala, Asp, Phe, Leu, Asn, Gln, and Val. These cleavages were selected because they featured bimodal αa distributions with relatively equal modes. Ser and Thr were chosen as the adjacent residues because they have short side-chains that restrict the hydrogen-bonding interactions in which they can engage. When Ser precedes X, for example, it naturally forms a stable hydrogen bond to the amide nitrogen. Hydrogen-bonding from a preceding Ser is also favorable to the amide oxygen of residue X. In contrast, hydrogen-bonding to the amide oxygen from a succeeding Ser is possible via the formation of a seven-membered ring; however, hydrogen-bonding to the amide nitrogen is not generally possible because of the orientation of the peptide backbone, which places the amide hydrogen 180° from the Ser OH group. Thus, preceding Ser/Thr residues can hydrogen-bond to the amide nitrogen or amide oxygen, whereas succeeding Ser/Thr residues can only hydrogen-bond to the amide oxygen.
Examination of the distribution in Figure 3c is useful to further elucidate the effects of hydrogen-bonding on Δαa. It is clear from Figure 3b that amide oxygen hydrogen-bonding results in a reduction of Δαa by approximately 0.3; however, the lower distribution in Figure 3a features a mean of only Δαa = –0.2. Because a preceding Ser/Thr can hydrogen-bond to either the amide NH or amide O, it is possible that this distribution represents an average Δαa, resulting from a –0.3 shift for all sequences with amide oxygen hydrogen-bonding and some other shift associated with amide nitrogen hydrogen-bonding. This assertion is supported by the histogram shown in Figure 3c, which features a broad distribution spanning from Δαa = –0.5 to 0.1. Inspection of this distribution and comparison to the widths of the distributions in Figure 3a and b suggests at least two or three distributions are present. GMM was unfortunately unsuccessful at fitting the data (it suggested a unimodal distribution), which is likely a consequence of too few instances and too much overlap of the apparent distributions. For this reason, Gaussian envelopes were manually fit to the distribution. Supporting Information Figure 9, 10, and 11 show histograms for four-, three-, and two-component fits. The sum of the variance between the fit and the experimental (Σχ2) was used to parameterize the fits, and the Σχ2 is reported for each plot. The four- and three-component fits by far outscored the two-component fit (Σχ2 of 27.0, 29.4, and 39.24, respectively), but were relatively similar to one another. However, because the four-component model best fits the data, the means and variances associated with that fit were used. The means associated with the four-component model are also the most consistent with the means found for the histogram data in Figure 3a and b. Using the four-component model, means were found at Δαa = 0.025, –0.095, –0.31, and –0.45. The populations at ≈ 0.0 and –0.3 are in agreement with the populations means observed for cleavages with a succeeding Ser/Thr. It is possible that the remaining two means, at Δαa ≈ –0.1 and –0.45, are the result of amide NH hydrogen-bonding and combined amide NH and O hydrogen-bonding, respectively. From Figure 3b, amide oxygen hydrogen-bonding is predicted to result in a –0.32 shift in αa; if amide NH hydrogen-bonding results in a reduction of αa by ≈0.1, the combined effect of amide NH and O hydrogen-bonding would be expected to be approximately –0.42, which is consistent with the –0.45 population only observed for cleavages with two adjacent hydrogen-bonding residues.
Examination of Hydrogen-Bonding Effect
The extent of atom elimination from a + 1 ions to generate a ions, presented as αa distributions, was statistically evaluated from 99 peptides from proteolytic digestion using LysN. Examination of the αa distributions suggests that a bimodality is present in the data, with one population representing more extensive hydrogen atom elimination and a second representing suppressed hydrogen elimination. Restriction of the amino acids adjacent to a given cleavage site to those with aliphatic side chains largely reduces and in some cases virtually eliminates the lower αa population associated with less extensive hydrogen atom elimination from the a + 1 radical ion. Restriction of adjacent residues to Ser/Thr further suggests that both amide oxygen and amide nitrogen hydrogen-bonding account for reduction observed in αa values, with amide oxygen hydrogen-bonding accounting for a shift of –0.3 in the αa value and amide NH hydrogen-bonding accounting for a shift of –0.1. It may be possible to use these shifts to probe the backbone hydrogen-bonding networks in native proteins. Further assessment of the αa distributions also suggests a third hydrogen elimination pathway is active, which results in mean αa values that are unique to the identity of the C-terminal amino acid of each a ion product. Moreover, the semi-unique mean αa values associated with the C-terminal amino acid of each a product ion may provide a means to assist de novo sequencing using 193 nm UVPD, an appealing approach that is inherently challenging because of the complex nature of the MS/MS spectra. Efforts are thus being undertaken to make use of the αa values to validate de novo methods.
Funding from the NSF (CHE-1402753), the Welch Foundation (F-1155), and NIH 1K12GM102745 (fellowship to L.M.) is acknowledged.
- 13.Paizs, B., Suhai, S.: Fragmentation pathways of protonated peptides. Mass Spectrom. Rev. 4(24), 508–548 (2005)Google Scholar
- 15.Gu, C., Tsaprailis, G., Breci, L., Wysocki, V.H.: Selective gas-phase cleavage at the peptide bond C-terminal to aspartic acid in fixed-charge derivatives of Asp-containing peptides. Anal. Chem. 72(23), 5804–5813 (2000)Google Scholar
- 22.Shaw, J.B., Li, W., Holden, D.D., Zhang, Y., Griep-Raming, J., Fellers, R.T., Early, B.P., Thomas, P.M., Kelleher, N.L., Brodbelt, J.S.: Complete protein characterization using top-down mass spectrometry and ultraviolet photodissociation. J. Am. Chem. Soc. 135, 12646–12651 (2013)CrossRefGoogle Scholar
- 41.Maia, H.L., Orrell, K.G., Rydon, H.N.: Evaluation of the thermodynamic parameters for the interconversion of the conformational isomers of two N-acylprolines by nuclear magnetic resonance spectroscopy. J. Chem. Soc. D. 19, 1209–1210 (1971)Google Scholar