Influence of Primary Structure on Fragmentation of Native-Like Proteins by Ultraviolet Photodissociation

Abstract

Analysis of native-like protein structures in the gas phase via native mass spectrometry and auxiliary techniques has become a powerful tool for structural biology applications. In combination with ultraviolet photodissociation (UVPD), native top-down mass spectrometry informs backbone flexibility, topology, hydrogen bonding networks, and conformational changes in protein structure. Although it is known that the primary structure affects dissociation of peptides and proteins in the gas phase, its effect on the types and locations of backbone cleavages promoted by UVPD and concomitant influence on structural characterization of native-like proteins is not well understood. Here, trends in the fragmentation of native-like proteins were evaluated by tracking the propensity of 10 fragment types (a, a+1, b, c, x, x+1, y, y−1, Y, and z) in relation to primary structure in a native-top down UVPD data set encompassing >9600 fragment ions. Differing fragmentation trends are reported for the production of distinct fragment types, attributed to a combination of both direct dissociation pathways from excited electronic states and those surmised to involve intramolecular vibrational energy redistribution after internal conversion. The latter pathways were systematically evaluated to evince the role of proton mobility in the generation of “CID-like” fragments through UVPD, providing pertinent insight into the characterization of native-like proteins. Fragmentation trends presented here are envisioned to enhance analysis of the protein higher-order structure or augment scoring algorithms in the high-throughput analysis of intact proteins.

Graphical Abstract

INTRODUCTION

Transfer of native-like proteins into the gas phase via electrospray ionization (ESI) has enabled powerful and rapid structural biology analyses through native mass spectrometry (native MS).^1–4 In this technique, noncovalent interactions, particularly between protein subunits and ligand cofactors, are maintained as intact proteins are transferred from solution to the gas-phase environment, empowering subsequent interrogation of protein subunit stoichiometry, conformation, topology, and even ligand-binding or folding thermodynamics through MS measurements.^1–6 Combined with controlled disassembly and fragmentation by tandem mass spectrometry (MS/MS), dissection of native-like proteins or protein complexes releases constituent subcomplexes, subunits, and sequence fragments via a general strategy known as native top-down proteomics.^7–10 Contrary to traditional bottom-up proteomics, top-down methodologies provide greater characterization of proteoforms or distinct manifestations of a protein from a particular gene due to alternative splicing, mutations, or post-translational modifications, to enlighten a more comprehensive view of the proteome.^7–10 Integrated with native MS, top-down methodologies have enabled discovery, characterization, and direct measurement of protein complexes as they exist in the cell^11,12 while circumventing some of the challenges, such as molecular size constraints or prohibitive sample purity, that stifle traditional structural biology methodologies.^1–4

Dictating fragmentation pathways and consequently the degree of characterization, ion activation techniques are essential to the design of native top-down experiments. Most commonly, collision-induced dissociation (CID) is utilized; this technique entails slow heating of proteins through low-energy collisions with a neutral gas, typically resulting in protein unfolding, asymmetric charge partitioning, and concomitant ejection of highly charged monomers^13–18 when interrogating protein complexes.^19,20 Although dissociation via collisional activation informs subunit identity and complex stoichiometry, the limited degree of backbone cleavage and the occurrence of significant protein unfolding impairs comprehensive characterization of proteoforms and interrogation of natively folded structures. Applying alternative ion activation techniques yields distinct dissociation behavior that results in additional or alternative information. Electron-based dissociation (ExD) methods, for example, induce backbone cleavage events without disrupting protein structure, presenting an avenue to probe protein secondary and tertiary structure, ligand binding sites, and subunit interfaces.^2,20–22 The preservation of noncovalent interactions after electron activation suppresses the release of fragment ions from highly ordered or inaccessible regions of the protein, thus providing insight into structural motifs.^2,20–22 Furthermore, the sequence coverage obtained from ExD may be boosted even further by incorporating concurrent slow-heating methods to unfold regions of the protein during activation or to release fragment ions held together by noncovalent interactions, thus affording even greater characterization of proteoforms.^20,22 Surface-induced dissociation (SID) is a surface-based collisional activation technique, that contrary to CID, typically disassembles protein complexes into low-charge folded subunits, thus making it a powerful MS/MS method for exploring quaternary structure.²³ The subunits and subcomplexes produced from SID inform connectivity and stoichiometry to provide a robust overview of protein assemblies. Additionally, SID can reveal the relative strengths of binding interfaces²⁴ and, in combination with computational modeling, may predict protein architectures.^25,26 For each ion-activation technique, a thorough understanding of the mechanisms and factors that modulate fragmentation has been leveraged to expand the scope of information derived from the dissociation method and develop more predictive insight for new native MS applications.

Inducing fragmentation that reflects higher-order structure yields high sequence coverage and promotes subunit disassembly, ultraviolet photodissociation (UVPD) is an exemplary ion-activation technique that results in highly informative dissociation of native-like proteins.²⁷ Akin to some of the features of ExD methods, 193 nm UVPD cleaves covalent bonds in the protein backbone while preserving noncovalent interactions, as evidenced by the detection of holo product ions that retain bound ligands.²⁸ By capitalizing on this finding, ligand binding sites have been localized based on mapping the absence and presence of holo product ions.^29–32 Results from several studies have indicated that the propensity of backbone cleavages throughout the protein is mediated by the flexibility of the region, conformational changes, and engagement in noncovalent interactions, all of which modulate the release of detectable fragment ions.^29–38 UVPD has also been extended to the examination of multimeric protein assemblies, for which ejection of subcomplexes and monomeric subunits reveal subunit connectivity and stoichiometry.^28,33,39 In sum, native-top down UVPD is sensitive to quaternary, tertiary, and secondary structure, providing an intriguing opportunity to expand its application in the field of structural biology.

Although variations in protein fragmentation by UVPD have been attributed to a combination of primary, secondary, and tertiary structural features, the influence of side-chain identity and position of each amino acid of a protein has not been evaluated. The observation of unusually abundant product ions attributed to backbone cleavages C-terminal to phenylalanine residues and N-terminal to proline residues has been reported; however, these phenomena have not been systematically examined.^40,41 The most extensive evaluations of UVPD fragmentation patterns have examined cohorts of peptides and intact proteins under denaturing conditions, where protein higher-order structure is not expected to be conserved.^42,43 Under these conditions, preferential cleavages adjacent to aromatic residues and proline have been noted (using 213 nm photons in both studies).^42,43 Analogous studies of native-like proteins have not been undertaken for UVPD, and it is anticipated that there might be notable differences in dissociation patterns of native-like proteins and denatured proteins as reported for ExD and CID.^21,44 Identifying site-specific cleavages and determining the overall effect of primary structure on 193 nm UVPD would enable the distinction of fragments generated via preferential cleavages from those that are predominantly affected by higher-order structure and, ultimately, further enhance the degree of structural characterization afforded by UVPD in the context of the analysis of native-like proteins.

The benefits of thoroughly deciphering dissociation mechanisms have been exemplified by the landmark development of the mobile proton model to describe peptide and protein fragmentation via collision-induced dissociation.⁴⁵ Refined across decades, this model originated from elaborate examination of the impact of residue identity, gas-phase basicity, and charge sequestration on peptide fragmentation.^45–51 Not only does the mechanism propose a robust explanation of observed fragmentation patterns of peptides, but the model has also shaped the development of bioinformatics approaches and database searches relevant for mass spectrometry-based proteomics.^51,52 Recently, evaluation of fragmentation propensities of 10,000 fragment ions produced through CID on intact proteins informed the design of a scoring mechanism tailored for native protein analysis and resulted in improved identification and characterization.⁵³ Similar investigation of UVPD is envisioned to provide comparable dividends for enabling high-throughput strategies for native proteomics.⁵⁴

Here, the effects of primary structure on 193 nm UVPD of native-like proteins and protein complexes were evaluated through the analysis of >9600 fragment ions produced from 28 proteins. Correlations between residue position and residue identity were correlated with fragmentation frequency and propensity, respectively, with differing trends noted for each fragment ion type examined (a, a+1, b, c, x, x+1, y, y−1, Y, z). Fragmentation trends are discussed in the context of dissociation mechanisms proposed for 157 and 213 nm UVPD, as well as in relation to prior findings for ExD and CID. A notable relationship between b/y fragment ion abundances and the mobile proton model was discerned for UVPD. Establishing this trend enabled a more detailed rationalization of the fragmentation patterns observed for two proteins, adenylate kinase and calmodulin, by considering ligand-induced changes in proton mobility.

MATERIALS AND METHODS

UVPD Data Collection and Analysis.

Experimental data were collected from ongoing and previous native MS studies from the Brodbelt group utilizing a Thermo Scientific Orbitrap Elite mass spectrometer (Bremen, Germany) or Thermo Scientific Q Exactive Plus UHMR mass spectrometer (Bremen, Germany), each modified with a 193 nm excimer laser (Coherent, Santa Clara, CA) for UVPD as previously described.^55,56 Experimental details and data for chicken adenylate kinase,³¹ human C-reactive protein,³⁹ human hemoglobin,³⁹ human transthyretin tetramer,³⁹ Escherichia coli (E. coli) aquaporin Z (AqpZ),⁵⁷ E. coli ammonia channel (AmtB),⁵⁷ E. coli mechanoselective channel of large conductance (MscL),⁵⁷ Staphylococcus aureus staphylococcal nuclease,⁴⁰ human α-synuclein,⁴⁰ human apo calmodulin,⁴⁰ human transthyretin monomer,⁴⁰ equine cytochrome C,⁴⁰ equine myoglobin,⁴⁰ bovine ubiquitin,⁴⁰ human carbonic anhydrase II,⁵⁸ and human high-mobility group AT-hook 2⁵⁹ are reported in the respective references.

Mass spectra for bovine Cu/Zn superoxide dismutase from MP Biomedicals (Santa Clara, CA), bovine milk α-casein from Sigma-Aldrich (St. Louis, MO), bovine milk β-casein from Sigma-Aldrich (St. Louis, MO), bovine milk β-lactoglobulin A monomer from Sigma-Aldrich (St. Louis, MO), bovine erythrocytes carbonic anhydrase II from Sigma-Aldrich (St. Louis, MO), human erythrocytes carbonic anhydrase I from Sigma-Aldrich (St. Louis, MO), bovine pancreas trypsinogen from Sigma-Aldrich (St. Louis, MO), and Bacillus amylolique-faciens barnase and barstar were collected by preparing each protein at a 5–15 μM concentration of the highest oligomeric state in 100 mM ammonium acetate solution. Barnase•barstar heterodimer was prepared by mixing the subunit solutions in a 1:1 molar ratio. Calcium-bound calmodulin was prepared as previously described.³² Each solution was infused by ESI using Pd/Au coated static tips on the Orbitrap Elite mass spectrometer and the lowest charge state that could be observed in sufficient abundance during a less than 1000 ms accumulation time at an AGC target setting of 1E6 and isolation window width of 25 m/z was selected for MS/MS. UVPD mass spectra (1 pulse set at 1–3 mJ) were collected at a 240,000 resolution setting at m/z 400 by averaging 500 scans.

Mass spectra for rabbit muscle aldolase from Sigma-Aldrich (St. Louis, MO), yeast alcohol dehydrogenase from MP Biomedicals (Santa Clara, CA), bovine milk β-lactoglobulin A dimer from Sigma-Aldrich (St. Louis, MO), and jack bean concanavalin A from Sigma-Aldrich (St. Louis, MO) were collected by preparing each protein at a 10 μM concentration of the highest oligomeric state in 50–100 mM ammonium acetate solution. Each solution was infused by ESI using Pd/Au-coated static tips on the Orbitrap UHMR mass spectrometer. Charge states were isolated and subjected to UVPD (1 pulse at 0.5–3 mJ). Mass spectra were collected at 120,000 resolution at m/z 400, averaging 500 scans.

All UVPD mass spectra were deconvoluted using Xtract algorithm with a signal-to-noise ratio threshold of 3, fit factor of 44%, remainder of 25%, and max charge set to the precursor charge state. Resulting decharged and deisotoped monoisotopic masses were used for fragment assignments. Assignments for a, a+1, b, c, x, x+1, y, y−1, Y, and z ions were made using UV-POSIT.⁶⁰ Fragment ion intensities were normalized to the median deisotoped and decharged fragment intensity assignment for each protein, unless otherwise indicated. Subsequent analyses were performed using base R and the dplyr package prior to visualization using ggplot2.^61,62 Y refers to y fragments that are deficient by two hydrogen atoms. Protein isoelectric points were calculated using the ExPASy Compute pI/MW tool (https://web.expasy.org/compute_pi/).

Analysis of Pentapeptides.

AAAAA, PAAAA, KAAAA, and RAAAA were synthesized by Genscript (Piscataway, NJ) and prepared at 2 μM in 50% methanol, 0.1% formic acid prior to ESI on the Orbitrap Elite mass spectrometer. The [M + H]⁺ precursor ion was isolated with a 10 m/z window at an AGC target of 2E5 and subjected to either higher-energy collisional dissociation (HCD) or UVPD. Spectra were collected at a resolution setting of 120,000 at m/z 400, averaged over 50 scans, and deisotoped with the Xtract algorithm. Fragment total ion current (TIC) percentage was calculated from the deisotoped MS2 spectra using eq 1.

$$\text{Fragment TIC (\%)}=\frac{\sum \text{Fragment ion intensity}}{\sum \text{Total ion intensity}}\times 100\%$$ (1)

Fragment TIC percentages were monitored as the normalized collision energy (NCE) was varied for HCD, fit to a sigmoidal curve using Solver in Excel, and the NCE setting that produced a fragment TIC percentage of 50% (NCE₅₀) was determined. Triplicate analyses were performed for each peptide. E₅₀ values were graphed relative to gas-phase basicities reported by Amster et al.⁶³

Fragment ions from deconvoluted UVPD mass spectra for each peptide were classified as “CID-like” or “UVPD-unique”. Fragments were classified as “CID-like” if the UVPD fragments were also present within 0.01 Da in the respective HCD mass spectra of each peptide, while all other fragment ions were considered “UVPD-unique”. Fragment TIC percentages for each category of fragment ions were calculated using eq 1 for each of the peptides in triplicate.

Ubiquitin Carbamylation.

Ubiquitin was carbamylated as previously described.^40,64 Briefly, a 20 μM solution of ubiquitin in 100 mM ammonium bicarbonate, pH 8, was split into two aliquots, one for derivatization and one for control. Carbamylation was performed by incubating one aliquot with 8 M urea at 80 °C for 4 h. Prior to ESI on the Orbitrap Elite mass spectrometer, ubiquitin samples were buffer exchanged into 100 mM ammonium acetate and diluted to 10 μM. The 5+ charge state was isolated with a 25 m/z isolation window and subjected to UVPD set at 1 pulse at 3 mJ. Spectra at a resolution of 240,000 at m/z 200 were averaged across 500 scans, deconvoluted using the Xtract algorithm, and analyzed with UV-POSIT. Fragment abundances were compared across triplicate analyses of control and carbamylated ubiquitin using Welch’s t-test.

RESULTS AND DISCUSSION

Native Top-Down UVPD Data Set.

Previous native top-down 193 nm UVPD studies have noted the potential influence of protein sequence on fragmentation trends, including prevalent backbone cleavages C-terminal to phenylalanine, aspartate, and glutamate, as well as N-terminal to proline.^40,41,64 To provide a more comprehensive examination of the impact of primary structure on 193 nm UVPD of native-like proteins, a cohort of UVPD mass spectra were evaluated to determine both the effect of residue position and the residue identity on fragmentation of native-like proteins and protein complexes. In this study, the ramifications of secondary, tertiary, and quaternary structure were not integrated into the analysis in order to evaluate patterns based solely on absolute identities of the amino acids for a large collection of proteins in a variety of charge states. Other prior studies of native-like proteins by UVPD have considered the impact of higher-order structural features, mainly in the context of rationalizing changes in fragmentation propensities for pairs or sets of apo versus holo proteins.^27–38 An extensive statistical analysis of the influence of secondary, tertiary, and quaternary structural factors is underway and will be presented separately. Data in the present study was amassed from previous and new data sets collected using UVPD-enabled Orbitrap platforms based on the analysis of sequence fragments (i.e., N-terminal and C-terminal containing ions from backbone cleavages) from 38 protein molecular ions encompassing 28 proteins and different charge, ligation, and oligomeric states. The data set is presented in Table S1. Details regarding the protein identity, charge state, oligomer state, isoelectric point, subunit molecular weights, and ligand identity are included in Table S2. The PDB code used to examine structures of proteins in the data set are also included in Table S2. In total, 9654 sequence fragments were assigned in this data set when considering a, a+1, b, c, x, x+1, y, y−1, Y, and z type sequence fragments. Total numbers of fragments assigned to each type are tabulated in Table S3. The percentages of fragment assignments and abundances assigned to each fragment type are summarized in Figure 1. Although 10 fragment types were searched in each experiment, the probability of false fragment assignments is nonetheless expected to be low, as only one protein or protein complex of known identity was examined per UVPD experiment. To confirm this expectation, false discovery rates were estimated for a few proteins of varying lengths using an approach previously implemented to estimate the false discovery rate of internal fragment assignments by shifting the masses of product ions by 50 ppm from −500 ppm to 500 ppm (yielding 20 mass-shifted lists) and determining the number of fragments assigned using the mass-shifted data.^65,66 The FDR was calculated for each mass-shifted fragment list based on the number of fragment assignments identified in the mass-shifted set divided by the number of fragment assignments made for the fragment list prior to mass-shifting. Mass-shifting by 50 ppm from −500 to 500 ppm resulted in 20 mass-shifted fragment lists, for which FDR was averaged together to estimate the probability of false fragment assignments in the UVPD experiment. By searching for these 10 fragment types with a 10 ppm mass tolerance, the average FDR for ubiquitin (76 residues) was estimated as 1.32%. In the same way, FDR was estimated as 1.67% for myoglobin (153 residues) and 2.08% for carbonic anhydrase (259 residues), demonstrating the low propensity of false fragment assignments for proteins of multiple lengths. FDR for each mass-shifted data set of each of the three proteins is shown in Figure S1. Although internal fragments are expected and present in UVPD mass spectra,^67,68 only terminal fragments were considered in the present study owing to challenges with confidently assigning internal fragment ions that require high mass accuracy and special considerations to resolve assignment ambiguity.^65–70

Figure 1.

Percentage of (A) each fragment ion type and (B) total ion intensity in the entire native top-down UVPD data set based on analysis of 9654 sequence ions.

Although a and a+1 were both the most frequent and the most abundant sequence fragments, each of the other fragment types was identified at generally comparable frequencies and abundances. Suggesting a low selectivity for any fragment type, the percentages of fragment ions assigned to each type were within a narrow range of 8–13% relative to the total population of ions (Figure 1A). Interestingly, although moderate preferentiality for a and a+1 fragments is apparent based on abundance (Figure 1B), the abundances of the complementary C-terminal x and x+1 ions are not enhanced, suggesting differing fragmentation channels for formation of these N- and C-terminal fragments. A cursory comparison to fragmentation trends observed in a large-scale top-down study of denatured proteins using 213 nm UVPD⁴³ reveals an even more notable preference for a/a+1 ions and somewhat greater variations observed in the distribution of the other ion types than observed in the present study. For example, the abundances of b, c, and y ions were lower compared to x, x +1, y−1, and z ions.⁴³ However, it is unclear whether the variations between the 213 nm UVPD data set and the present 193 nm UVPD results are related to wavelength, laser frequency, and ion trap pressure used for UVPD or the fact that proteins are in low charge states consistent with native-like structures in the present study versus the higher charge states of the denatured proteins in the 213 nm UVPD study. The enhanced production of a and a+1 ions is undeniable in both UVPD studies. These differences and similarities merit more systematic comparison in a follow-up study.

Another notable feature of the present UVPD data set is the rather substantial abundance and frequency of c fragments which contrasts with previous UVPD analyses of denatured proteins where this fragment type is of low abundance using 193 nm photons^55,71,72 or 213 nm photons.⁴³ One possible explanation was offered in a native top-down 266 nm UVPD study which showed the prevalence of photoelectron-transfer dissociation, a pathway in which a zwitterionic structure loses an electron through electron photodetachment to produce a radical state that undergoes c/z fragmentation.⁷³ If a similar process is active for 193 nm UVPD, for which electron photodetachment is known to be predominant in the negative ion mode,⁷⁴ proteins retaining native-like zwitterionic salt bridges would be more predisposed to generating c/z fragments relative to denatured proteins. Examining the percentage of c fragment assignments per protein, membrane proteins (MscL, AmtB, and AqpZ) and hemoglobin subunit β (HBB) featured the highest percentages of c fragment assignments (30.7% for MscL, 28.7% for AmtB, 19.0% for AqpZ, and 28% for HBB). Interestingly, each of these proteins predominantly feature α-helical domains, with 62%, 77%, 76%, and 90% of MscL residues, AmtB residues, AqpZ residues, and HBB residues, participating in this secondary motif, respectively. Conversely, proteins with a high β-sheet content were found to yield a low percentage of c fragments, as only 0.7%, 1.0%, and 1.5% of fragments were attributed to c ions for C-reactive protein (37% β sheet), bovine carbonic anhydrase (29% β sheet), and transthyretin (48% β sheet), the latter of which featured the highest β-sheet content in the data set. To further examine this trend, c fragment percentages were plotted as a function of α-helical content and as a function of β-sheet content for proteins in the data set with known crystal structures to reveal a mild correlation between c fragments and structural motif, according to a Pearson correlation test (Figure S2). While α-helices featured a mild positive correlation with high c fragment abundances (Pearson coefficient = 0.64), β-sheets featured a mild negative correlation (Pearson coefficient = −0.54) presenting lower c fragment abundances as β-sheet content increased. Regardless of the specific dissociation mechanisms, it is apparent that a diverse set of fragment types are generated by 193 nm UVPD of native-like proteins, and as discussed in the following sections, each fragment type may encode a wealth of unique information that may be leveraged to characterize protein structure.

Effect of Residue Position on Backbone Cleavages.

To comprehensively characterize the effects of primary structure on native top-down UVPD, the influence of residue position and residue identity were evaluated. For characterization of proteins by MS/MS, it has often been observed that the majority of product ions originate from backbone cleavages occurring close to the N- and C-termini, yielding ample sequence coverage of the terminal sections but limited sequence coverage of the middle of the protein.^67,75–77 The limited sequence coverage of the middle sections of proteins tracks with protein size, becoming more pronounced for larger proteins. In the present study, the frequency of cleavages identified at different backbone positions was tracked based on the collection of fragment ions, then displayed in graphical form in Figure 2A in which the residue position of the backbone cleavage is mapped from the N-terminus to the C-terminus, and 0 and 1 represent the N- and C-terminus, respectively. All fragment ions are classified as either N-terminal (a, a+1, b, c, gray bars) or C-terminal (x, x+1, y, y−1, Y, z, gold bars) in Figure 2A. Significantly higher frequencies of backbone cleavages occur near the N- and C-termini, confirming the substantial impact of the sequence position and indicating a bias against the generation, survival, or detection of longer sequence fragments. There is a known mass-dependent bias in Orbitrap platforms that may partially explain the under-representation of larger mass fragment ions.⁷⁸ It must also be considered whether UVPD does not cleave the backbone as efficiently in the midsection of the protein, an explanation that we find improbable owing to the nature of the photoabsorption process. Alternatively, backbone cleavages in the midsection of the protein may occur with high efficiency, but the resulting fragment ions decompose into smaller ions or internal ions⁶⁷ or may produce ions that fall in highly congested regions of the mass spectra, rendering them ineffectively deconvoluted and unassignable. These latter explanations have generated considerable speculation^66–68,70 and motivated recent efforts to resolve them by the development of new search algorithms^70,76,79 and experimental strategies such as proton transfer reactions^80,81 and fragment ion protection.^75,82

Figure 2.

(A) Histogram of backbone cleavage positions (total of 9654 fragments) along the protein sequence (divided into 50 bins) for N-terminus (a, a+1, b, c) and C-terminus (x, x+1, y, y−1, Y, z) containing fragments shown in gray and gold plots, respectively. The relative backbone position extends from the N-terminus (0.00) to the C-terminus (1.00). (B) Histogram of backbone cleavages along the protein sequence by fragment type. N-Terminus and C-terminus containing fragments are shown in gray and gold plots, respectively. Table S3 shows the number of fragment ions assigned corresponding to backbone cleavages. The median relative position for N-terminal containing and C-terminal containing fragment ions are demarcated by a vertical dashed gray or gold line, respectively.

Interestingly, examining the distributions of backbone cleavage positions for distinct fragment types (a/a+1/x/x+1, b/y, c/z) reveals that the bias against larger fragment ions is not equally represented for each fragment type, as demonstrated in Figure 2B. Backbone cleavages leading to b/y ions are more tightly clustered near the termini, a preference less notable for c/z ions and even less significantly observed for a/a+1/x/x+1 ions. The magnitude of this selectivity is semi-quantitatively represented by the median backbone cleavage position, tabulated in Table S4, where a median value of 0.5 indicates an equal propensity for backbone cleavages throughout the protein and no bias for either terminus. Notably, the a/a+1/x/x+1 fragments, which are typically unique to UVPD, display distributions that tail far further into the middle of the protein sequence. This trend emphasizes an attribute of UVPD that may contribute to its delivery of high sequence coverages for proteins, especially in comparison to collisional activation methods that predominantly produce b/y sequence fragments. Again, this phenomenon highlights the concept that specific fragment types may present distinct advantages for structural elucidation; in this case, a/a+1/x/x +1 fragments contribute to sequence coverage deeper into the midsection of proteins in a manner crucial for more comprehensive characterization of primary structure.

Importantly, the differing trends for cleavage of the backbone relative to the N-/C-terminal proximity may also inform the distinct fragmentation channels responsible for cleavage of the C_α–C (a/x ions), C–N (b/y ions), or N–C_α (c/z ions) bonds. The fragmentation of intact proteins upon UV photoactivation is thought to occur through two primary pathways: direct cleavage of bonds from ions in excited electronic states on the femtosecond time scale, a feature unique to UVPD, and backbone cleavage after intramolecular vibrational energy redistribution (IVR) following internal conversion of the ion to the ground electronic state, a process that randomizes the internal energy.⁸³ Direct dissociation, without subsequent vibrational redistribution, portends that backbone cleavage frequency should occur evenly across the protein sequence,⁶⁷ an outcome consistent with the production of a/a+1/x/x+1 type fragments uniquely observed upon UV photoactivation. In contrast, the enhancement of backbone cleavages near the termini to produce b/y ions is consistent with two routes: (i) internal conversion and IVR of intact proteins, resulting in CID-like fragmentation that is modulated by protons, and (ii) IVR and secondary fragmentation of primary product ions generated with excess internal energy following their initial production by direct dissociation.

Examining the distribution of backbone cleavages that lead to hydrogen deficient y products, y−1 and Y (aka y-2), reveals median positions of 0.72 and 0.73, respectively (Table S4 and Figure S3). These values are comparable to those of x/x+1 ions (median of 0.75) and closer to 0.5 than those of conventional y ions (median of 0.86), suggesting that the production of these y−1 and Y fragments occurs via direct dissociation mechanisms rather than internal conversion/IVR-like standard y fragments. Although the mechanism of formation of y−1 ions has not been characterized in detail, 157 nm UVPD of proline-containing peptides has been proposed to generate Y fragments through photon-induced homolytic cleavage of C_α–C bonds that also produces complementary a+2 ions,⁸⁴ substantiating the postulate that hydrogen deficient y ions (i.e., y−1 and y−2) in the present study originate from direct dissociation pathways and not through other routes. Likewise, the similar median backbone cleavage position values of 0.27 and 0.22 for the formation of a and a+1 ions, respectively (Table S4 and Figure S4), offer further evidence that both are generated by direct dissociation of C–C_α bonds through Norrish type 1 processes, as previously proposed for 157 and 193 nm UVPD.^34,51,85,86 Minor differences between these two backbone cleavage position distributions may be attributed to the modulation of a versus a+1 abundances by secondary structural motifs, particularly hydrogen bonding that suppresses transfer of a hydrogen atom during cleavage of the C_α–C bond, as previously described.^34,85,86 A similar trend is observed in the median backbone cleavage positions for production of x/x+1 ions, values of 0.73 and 0.76, respectively (Table S4 and Figure S4) again suggesting that these fragments are generated through similar mechanisms with minor variations attributed to the presence or absence of hydrogen bonds local to the backbone cleavage site.

The distributions of backbone cleavage sites leading to c/z fragments and the resulting medians are intermediate between those observed for direct dissociation products (a/a+1/x/x+1) and CID-like products (b/y ions). The c/z ions produced upon UVPD likely originated from radical rearrangement and subsequent fragmentation according to mechanisms previously proposed for ExD.²¹ One prior study of 193 nm UVPD using an FT-ICR platform demonstrated that c/z fragments were produced through photodetachment of electrons from metal surfaces during laser irradiation and subsequent electron capture by the protein ions (photo-ECD).⁸⁷ Photoinduced electron transfer dissociation (PETD)⁷³ has been reported as a prominent process during 266 nm UVPD, and this pathway has not been explored in detail using 193 nm photons. At this time, the exact mechanisms of the c/z ions observed in 193 nm UVPD mass spectra remain uncharted.

Effect of Residue Identity on Backbone Cleavages.

Prior UVPD studies have noted the variation in the prevalence of backbone cleavages occurring adjacent to specific amino acids, such as N-terminal to Pro and C-terminal to amino acids with aromatic side-chains Phe and Tyr.^40,41,64 The former trend is often attributed to the well-known “proline effect”,^44,51,53 in which fragmentation adjacent to Pro with its unique tertiary amine side-chain is facilitated according to the mobile proton model, whereas the latter trend is attributed to enhanced UV photoabsorption and localized excitation of the aromatic side chains. Proline-directed cleavages should be more predominant for fragment ions that are generated through IVR and CID-like processes where proton migration is prominent. Backbone cleavages adjacent to amino acids possessing aromatic side-chains should be more prevalent for fragment ions produced by direct dissociation pathways where energy is not redistributed to locations remote from the site of photon absorption. To systematically examine the modulation of backbone cleavages by residue identity for UVPD of native-like proteins, the abundances of fragment ions (a, a+1, b, c, x, x +1, y, y−1, Y, z) originating from backbone cleavages occurring C-terminal (X|X′) and N-terminal (X|X′) to each amino acid were collectively summed and averaged for each residue. For example, a backbone cleavage between F and P (F|P) that generated a detectable fragment ion would be categorized as both C-terminal to F and N-terminal to P. A histogram charting backbone cleavages at each amino acid based on the evaluation of 9654 fragment ions from all the protein entries in Table S2 is shown in Figure S5. The histograms partitioned by fragment type are shown in Figures 3 and S6.

Figure 3.

Mean normalized intensity for fragment ions originating from backbone cleavages occurring N-terminal (X|X′ gray bars, fragment ions retaining the C-terminus) and C-terminal (X|X′ gold bars, fragment ions retaining the N-terminus) to each amino acid are displayed as gray bars and gold bars, respectively. Mean fragment intensity that would be achieved if backbone cleavages were uniformly distributed at every backbone position is shown as a horizontal dashed line. Each bar plot features distinct y-axis scales to accommodate the differences in abundances for each fragment type, illustrated in Figure 1B. The histograms are based on the number of backbone cleavages indicated in Table S3.

Trends associated with the proline effect and aromatic residues were first evaluated. Preferential cleavages N-terminal to Pro were noted for y ions and to a lesser extent, b ions, suggesting that these ions may occur via CID-like pathways despite originating from UVPD. However, enhanced abundances of cleavages N-terminal to proline were also noted for a/a+, y−, and Y ions which likely arise from direct dissociation pathways, indicating that Pro may play a special role in several types of mechanisms. In particular, y−1 and Y fragments were generated almost exclusively N-terminal to proline (Figure S6). In a past study that examined the fragmentation of Pro-containing peptides by 157 nm UVPD, cleavages adjacent to Pro resulted in a and a+2 ions through homolytic cleavage of the C_α–C bond that purportedly also results in complementary Y ions,⁸⁴ providing a possible mechanistic explanation for the observed N-terminal Pro cleavages often noted in 193 nm UVPD mass spectra. Notably, the same study also predicted dominant homolytic cleavage events C-terminal to Pro in a manner that results in abundant x ions, an outcome also observed in the present study. Thus, analysis of UVPD fragmentation trends suggests that the prevalence of backbone cleavages occurring N-terminal to Pro cleavage is only partially modulated by proton mobility and may have other origins related to excited electronic states of ions. This insight may be crucial for correlation of Pro cleavages with protein structural features, such as Pro cis/trans isomerization,^84,88 or may guide future attempts to mitigate Pro-specific cleavages in favor of nonspecific cleavages during UVPD.

Backbone cleavages occurring C-terminal to Phe and Tyr led to preferential production of a/a+1 and x/x+1 ions. These results suggest enhanced photoabsorption related to aromatic side-chains and are consistent with prior reports suggesting that a/a+1 and x/x+1 ions result from direct dissociation from ions in excited electronic states.^34,85,86,89 The enhancement of these site-specific cleavages observed in the trends for z ions similarly supports their production through direct dissociation pathways. Two prior investigations proposed conflicting mechanisms for the formation of z ions upon high energy activation, postulating that these fragments may result from homolytic cleavage of N–C_α bonds upon keV CID⁹⁰ or from neutral loss of NH₃ from y ions upon 266 nm UVPD.⁹¹ Both of these routes are feasible and are not readily differentiated in the present study.

Inspection of other preferential site-specific cleavages based on the resulting fragment ion type (a, a+1, b, c, x, x+1, y, y−1, Y, z) reveals additional patterns that further delineate the impact of primary structure on native top-down UVPD. Some preferentiality is also observed for both backbone cleavages adjacent to His for a/a+1 and c ions, as well as adjacent to Arg for x/x+1 and z ions. In addition, cleavages C-terminal to Asp are enhanced in the trends solely for b/y ions, and cleavages adjacent to His are increased exclusively for b ions, aligning with the mobile proton model often used to explain CID patterns of proteins.^44,53 Even though relationships between UVPD and the participation of the mobile proton model have been previously explored,^40,64,92 by partitioning the abundances of fragment ions according to fragment type and residue identity within the present UVPD data set, the mobile proton model can be pinpointed to primarily influence the production of b/y ions.

Additional preferential cleavages that are not readily accounted for by known mechanisms include preferential dissociation N-terminal to Phe for b ions and C-terminal to Tyr and Met for y ions, again highlighting a few of the uncharted consequences of primary structure on UVPD of proteins. Some enhancement of backbone cleavages occurring C-terminal to Tyr and Phe were observed for z ions; however, the overall signatures for c/z trends are generally distinct from those observed for b/y and a/x fragments, indicating the formation of these fragments through different types of pathways. Preferential cleavages C-terminal to Gly, Arg, His, Trp, and Gln and N-terminal to Lys, Arg, and Thr are noted for either c or z ions upon UVPD, partially aligning with a prior study where cleavages C-terminal to Glu, His, Asn, Gln, Arg, and Trp and N-terminal to Lys, Asp, Glu, and Asn were observed.⁹³ Nonetheless, compounding evidence demonstrating differing susceptibility of c/z ions to both residue position and residue identity upon UVPD in comparison to a/a+/x/x+ and b/y ions further substantiates the postulate that these fragment types are generated through a mechanism that is distinct from direct dissociation and IVR. Moreover, although the reason for some of these enhanced cleavage sites remains unclear, it is important to identify these site-specific cleavages that confirm the influence of primary structure on UVPD.

Considering the effects of primary structure within the context of complete protein structure, generalized linear models were developed and applied to provide some insight into the relative effects of primary structure and higher-order structure on protein fragmentation. Initially, a logistic regression (model 1, eq 2) was built for each fragment type to model the probability of detecting a fragment ion at a specific cleavage site using only predictors representative of primary structure. The predictors used include the identity of the amino acid N-terminal to the potential cleavage site (represented as a.a.n.), the identity of the amino acid C-terminal to the potential cleavage site (represented as a.a.c.), and the relative position of the cleavage site with protein length applied as an interactor to the relative position. A second model (model 2, eq 3) predicting the probability of detecting a fragment ion as a function of both primary structure (using the predictors delineated for model 1) and higher-order structure was also developed for comparison. Predictors representative of higher-order structure include b-factor, solvent accessible surface area, secondary structure, and oligomeric state (i.e., monomer or oligomer). Models 1 and 2 are represented in eqs 2 and 3, respectively.

$$\text{Fragment event}~\text{a}.\text{a}.\text{n}.+\text{a}.\text{a}.\text{c}.+\text{relative position}*\text{protein length}$$ (2)

$$\text{Fragment event}~\text{b factor}+\text{SASA}+\text{oligomeric state}+\text{a}.\text{a}.\text{n}.+\text{a}.\text{a}.\text{c}.+\text{relative position}*\text{protein length}$$ (3)

For the second model, higher-order structure had to be extracted from models in the protein data bank, and accordingly, only proteins for which appropriate PDB structures were available (indicated in Table S2) were used as the training data set. This data set featuring only proteins with known structures was used to develop logistical regressions for each fragment type using model 1 and model 2. The area under the curve (AUC) for receiver operating characteristic curves generated using model 1 and model 2 for each fragment type are tabulated in Table S5. Generally, using only primary structure as a predictor (model 1) resulted in curves with good AUC values in the range of 0.80–0.87 for each fragment type. Including higher-order structure (model 2) resulted in small increases (ranging from 0.01 to 0.05) in AUC that are most significant for x, x+1, and y−1 fragments, suggesting that higher-order structure holds a minor influence on protein fragmentation relative to primary structure. Thus, fragmentation patterns described in this manuscript can largely be attributed to primary structure, although we emphasize that higher-order structure does have an impact.

The comparative effects of primary and higher-order structure are further illustrated by applying these models to predict top-down UVPD mass spectra of native-like proteins. A new training data set including all proteins with a PDB structure except for ubiquitin was used to train logistic regressions for each fragment type, again using model 1 and model 2. For each resulting regression, a probability value for which the true negative rate is approximately 80% was selected. All data points above the probability cutoff were used to train a linear regression that models fragment intensity as a function of primary structure only, or as a function of primary structure and higher-order structure for each fragment type. The primary structure and higher-order structure of ubiquitin were applied to the logistic regression to determine if a fragment ion generated from cleavage of each backbone position is expected or not. If the probability value for each cleavage site was above that determined for the training set, the backbone is expected to cleave and thus applied to the linear model predicting fragment intensity. The summed predicted intensities of all fragment types corresponding to each cleavage site for ubiquitin are plotted in Figure S7, displaying a predicted trace built using model 1, a predicted trace built using model 2, and the trace for the experimental data for native top-down UVPD of ubiquitin. Pearson correlation coefficient comparing the trace built using primary structure only (model 1) to the experimental data show good correlation between the two traces with a coefficient of 0.47. Including higher-order structure (model 2) provides a somewhat better coefficient of 0.55, again indicating that primary structure by itself adequately explain much of the fragmentation observed upon UVPD, but that inclusion of higher-order structure is necessary to rationalize the more subtle variations in fragmentation trends. A thorough evaluation of the interplay between primary structure and higher-order structure, with a focus on the latter, merits further examination and will be examined in future studies.

Implications of Mobile Proton Model on Native Top-Down UVPD.

Re-evaluated and redefined across decades, the mobile proton model is a rigorously tested concept, producing consistent explanations for fragmentation of peptides observed by collision-induced dissociation.^{44,45,51–53,94} Broadly, the model proposes two major mechanistic routes, classified as charge-directed and charge-remote, for peptide dissociation based on charge site localization.⁵¹ Governing the predominance of either dissociation route is the “mobility” of ionizing protons, in which highly mobile protons promote charge-directed fragmentation, whereas peptides with sequestered protons (nonmobile) favor charge-remote fragmentation. For peptides in higher charge states in which the number of protons is greater than the number of highly basic residues (Arg, Lys), protons may occupy positions along the peptide backbone, weakening amide bonds that cleave through charge-directed routes upon collisional activation. This model also explains the “proline effect” noted upon CID, as the high proton affinity of the unique proline backbone amide favors charge sequestration, consequently promoting cleavage N-terminal to proline.^{45,46,49,51,52} For peptides in lower charge states, the ionizing protons are sequestered at side-chains with high proton affinity (i.e., Arg) and unable to migrate to facilitate backbone cleavage. In this scenario, fragmentation commonly occurs C-terminal to Asp, as acidic hydrogens on the carboxylic acid group mobilize to participate in the cleavage of the C-terminal amide bond prior to migration of more tightly sequestered protons.^45,50,51 We have already noted and discussed the probable influence of the mobile proton model on the production of b/y ions upon UVPD based on the effects of primary structure. We now take a closer look at the role of the mobile proton model in the context of UVPD of native proteins.

The gas-phase basicity of each amino acid side chain dictates the strength of charge sequestration and is thus a key feature that affects fragmentation under the mobile proton model. Indeed, side-chains with higher gas-phase basicity are expected to require more energy to mobilize a proton to the backbone, meaning more energy is required to induce fragmentation.⁹⁴ Via an energy-resolved ion activation study, Wysocki et al. demonstrated that the order of fragmentation efficiencies of a series of pentapeptides differing only in the identity of the N-terminal amino acid was inversely related to the gas-phase basicity of the unique N-terminal residue, substantiating assertions proposed by the mobile proton model.⁹⁴ Mirroring this series of experiments, we used 193 nm UVPD and HCD to examine the fragmentation efficiencies of four singly charged pentapeptides featuring unique N-terminal amino acids with distinct side-chain proton affinities. Fragment TIC percentage was calculated by dividing the sum intensity of all fragment ions by the sum intensity of all fragment ions and precursor ion in the MS/MS spectrum. This process was repeated at multiple collision energies (NCE values) for HCD to generate the energy-resolved mass spectrometry curves shown in Figure S8A from which the NCE₅₀ value (NCE level at which fragmentation efficiency is equal to 50%) was extrapolated (Figure S8B). In line with prior MS/MS experiments, the NCE₅₀ value was lowest for AAAAA upon HCD, followed by PAAAA, KAAAA, and RAAA, recapitulating the correlation between critical energy of dissociation and gas-phase basicity of the peptide. Obtaining NCE₅₀ values by UVPD is prohibited by the requirement for multiple laser pulses or very high laser energies, both of which contribute to secondary fragmentation confounded by multiphoton processes. However, fragment TIC percentages may be calculated using a single standard UVPD condition, 1 pulse at 3 mJ, for the four peptides as shown in Figure S9. The fragment ions originating from CID-type pathways (internal conversion and IVR, primarily b/y ions matching ones observed upon HCD) were segregated from those that arise from direct dissociation (e.g., a, c/z, Y ions), as summarized in Table S6. The resulting fragment percentages for CID-type ions and direct dissociation (Figure 4A) reveals that as gas-phase basicity of the peptide increases, fragment percentage declines for the production of CID-type ions, suggesting that these fragments are generated through charge-directed fragmentation. These results suggest the participation of mobile proton pathways in 193 nm UVPD of peptides. Unexpectedly, the data also revealed an increase in production of UVPD-unique fragments as gas-phase basicity increased from AAAAA to KAAAA, signifying a shift toward direct-dissociation pathways and highlighting the complexity of UVPD pathways.

Figure 4.

(A) UVPD (1 laser pulse at 3 mJ) fragment TIC percentage for custom pentapeptides with distinct amino acids at the N-terminus was calculated by considering all fragmentation events (gray bars), fragmentation events that occur for both HCD and UVPD (red bars), and fragmentation events uniquely generated by UVPD (blue bars). The error bars represent one standard deviation (n = 3). The gold dots denote the gas-phase basicity of the N-terminal amino acid reported in ref 57. (B) Percent of fragment ion intensity assigned to each fragment type for UVPD (3 mJ, 1 pulse) of unmodified ubiquitin (5+) and carbamylated ubiquitin (5+). Significance values were determined by Welch’s t test. An average of 264 and 178 fragments were assigned in the native and carbamylated spectra, respectively. The experimental data was initially collected and reported in ref 40.

This correlation can be extrapolated to inform dissociation routes for UVPD of intact proteins in native-like low charge states, as demonstrated for ubiquitin and carbamylated ubiquitin. The basicities of lysine side-chains of ubiquitin can be significantly decreased by converting them to carbamyl groups, resulting in significantly greater proton mobility (Figure S10).^40,64 UVPD was used to characterize the 5+ charge state of ubiquitin prior to and after carbamylation, in which 8 basic sites (7 Lys and the N-terminus) were carbamylated. Tracking each fragment ion type generated by UVPD reveals that upon carbamylation, the abundances of b/y ions is significantly increased (Figure 4B), validating our assessment that these fragments are produced at least in part by CID-like pathways modulated by mobile protons. At the same time, the abundances of a/a+1 ions, two prominent direct dissociation products, were greater for unmodified ubiquitin compared to carbamylated ubiquitin. Significant increases in abundance were also noted for c/z ions for carbamylated ubiquitin, an outcome that may be related to changes in protein structure or enhancement of radical-directed pathways upon carbamylation. Similar trends were observed in a prior UVPD study for calmodulin, myoglobin, cytochrome C, transthyretin, and ubiquitin in which an increase in the abundances of b/y ions was consistently observed after carbamylation.⁴⁰

In order to apply these fundamental principles to inform protein structural analyses, data for native proteins previously characterized by 193 nm UVPD were revisited to uncover new insight. Prior UVPD studies of native proteins focused on the patterns of a/x-type ions,^29–32 the ones most unique for UVPD and thought to best reflect variations in noncovalent interactions that are not disrupted during fast fragmentation of ions in excited electronic states. In one prior study, 193 nm UVPD was used to characterize changes in the fragmentation pattern of adenylate kinase (AK) throughout its catalytic cycle.³¹ Variations in the efficiencies of backbone cleavages leading to a/x ions throughout the protein were correlated with conformational changes resulting from substrate and cofactor binding.³¹ Reexamining this MS/MS data with a specific focus on the abundances of b/y fragments reveals that in comparing apo-AK to the AK•diadenosine tetraphosphate (AP4A) complex, enhancement of cleavages N-terminal to proline (Y|P), enhancement of nonspecific cleavages (L|K, V|G, A|T, V|F), and minor suppression of cleavages C-terminal to aspartate (D|A, D|T) are observed (Figure 5A), all of which indicate a change in mobile proton pathways, specifically an enhancement of charge-directed pathways. Examining the crystal structure of AK•AP4A (PDB 2C95) shows that interaction of the protein with the ligand is primarily driven by the formation of salt bridges between side-chains of R44, R97, R128, R132, R138, and R149 and phosphate groups of the ligand. Salt bridges were shown in a previous study to inhibit sequestration of protons by residues involved in salt bridges.⁹⁵ If Arg side-chains of AK participate in salt-bridges with AP4A, the Arg side-chains expected to tightly sequester protons in the apo form may be blocked from protonation in the holo AK•AP4A complex. Thus, ionizing protons would be localized to other less basic sites upon ESI of AK•AP4A, effectively shifting the protein from a less mobile protonation state (i.e., apo-AK) to a more mobile state (the holo complex). This repositioning of protons is consistent with the increase in charge-directed fragmentation observed for the AK•AP4A complex.

Figure 5.

(A) Suppression (blue) or enhancement (red) of the abundances of b/y ions produced upon UVPD upon binding of AP4A to adenylate kinase (AK) relative to apo AK are shown for specific backbone cleavage sites. Regions that showed no significant change (p-value >0.05 by Welch’s t test) are shown in gray. (B) Suppression (blue) or enhancement (red) of the abundances of b/y ions produced upon UVPD upon binding of 4Ca²⁺ ions to calmodulin relative to apo calmodulin are shown for specific backbone cleavage sites. Regions that showed no significant change (p-value >0.05 by Welch’s t test) are shown in gray. The experimental data was initially collected and reported in refs 31 and 32.

Variations in the prevalence of charge-mediated pathways are also observed for calmodulin upon binding to calcium ligands. The comparison of b/y ion abundances produced upon UVPD of apo and holo calmodulin bound to four calcium ions reveals an increase in charge-remote fragmentation for the holo protein. In this case, the intact mass of holo calmodulin indicates the incorporation of four calcium ions and removal of 6 hydrogen atoms compared to the apo protein, corresponding to (M+4Ca-6H) (see Figure S11). For this example, the 7+ charge state of both apo and holo calmodulin was subjected to UVPD. Upon UVPD, greater abundances of b/y fragments originated from backbone cleavages C-terminal to acidic side-chains (D|G, D|I, D|K, E|A) indicate a reduction in proton mobility (Figure 5B). The coordination of calcium by calmodulin is largely mediated by interactions between acidic side-chains (D20, D22, D24, E31, D56, D58, E67, D93, D95, D97, E104, D129, D131, D133, E140) and the metal cations.^32,96 The six hydrogen losses of (M+4Ca-6H) likely occur at acidic side-chains, thus resulting in deprotonated carboxylate side-chains. The increase in the prevalence of backbone cleavages adjacent to Asp and Glu via charge-remote pathways signifies a shift toward conditions in which the protons are less mobile, consistent with the proton-deficient calmodulin•4Ca complex.

CONCLUSIONS

The present study has deepened the understanding of UVPD of native-like proteins and enlightened the factors that influence fragmentation pathways by investigating the relationship between protein primary structure and the formation of CID-like versus UVPD-unique fragment ions. Tracking the frequency of backbone cleavages for a number of proteins revealed that not all fragment ion types generated by 193 nm UVPD (a, a+1, b, c, x, x+1, y, y−1, Y, z) were equally susceptible to protein primary structure. Whereas the b and y type fragment ions originating from backbone cleavages near the protein termini were most prevalent, the a, a+1, x, and x+1 fragment ions were much less susceptible to primary structure. These observations support a combination of direct dissociation pathways agnostic to the backbone cleavage site (a/x ions), and internal conversion/IVR pathways akin to CID pathways that exhibit preferential backbone cleavage sites. Our analysis of patterns suggests that both y−1 and Y products formed upon UVPD of native-like proteins occur via direct dissociation pathways. Examining the effect of side-chain identity revealed that b/y ions display preferential cleavage sites that are consistent with the tenets of the mobile proton model.

Higher-order structural elements also influence the fragmentation pathways of proteins, as previously noted for studies based on CAD, ETD, and UVPD. For example, α-helix and β-sheet motifs incorporate specific networks of hydrogen bonds between carbonyl and amide groups of the backbone that create localized folded regions of enhanced stability that may modulate fragmentation. A myriad of other noncovalent side-chain interactions contribute to three-dimensional tertiary structures of proteins, ultimately resulting in the greater prevalence of ionizable residues on the surfaces of the proteins. The critical role of salt bridges in mediating the preferential cleavage of covalent bonds versus disruption of noncovalent interactions has been described in detail.⁹⁷ Moreover, the intermolecular forces that participate in the genesis of tertiary and quaternary structures guide the distribution of charges and locations of charge sites upon disassembly and dissociation of proteins. Dissecting how each of these levels of structure influence the fragmentation pathways of proteins, while accounting for the significant effects of primary structure herein discussed, will be addressed through a future study incorporating statistical modeling of UVPD fragmentation as a function of protein structural features mined from crystal structures in the protein databank. Creating an integrated model poses an incredibly challenging task because of interdependence across the primary, secondary, tertiary, and quaternary structure of a protein system, but will be addressed and discerned through the construction of models that consider protein structure at each level.

Ultimately, the findings presented in this investigation will guide future mechanistic and structural biology studies utilizing UVPD for MS/MS characterization of proteins and protein complexes. Although b/y ions are not the predominant ones produced by UVPD of intact proteins, re-evaluation of these CID-like ions produced by UVPD revealed new insight into factors that modulate their abundances. Our findings suggests that while fragment ions unique to UVPD were not susceptible to mobile proton pathways, fragments that were generated by both CID and UVPD (i.e., b/y fragments) followed trends predicted by the mobile proton model. Examples derived from reanalysis of adenylate kinase and calmodulin complexes affirmed that b/y ions provide useful insight about proton mobility and how it is influenced by ligand binding. Systematic evaluation of the dissociation trends for other proteins should capitalize on all the ion types created by UVPD, not solely the a/x ions most unique for UVPD. A more robust understanding of the impact of primary structure on UVPD spectra of proteins may further advance the determination of higher-order structure as well as inform scoring algorithms to assist strategies for high-throughput native analyses.