Introduction

Chemical cross-linking combined with mass spectrometry is commonly utilized to study topological features and interactions of proteins by establishing structural distance constraints [1,2,3,4,5]. This rapidly emerging technique has unveiled structures and interactions of protein complexes [6,7,8,9,10,11,12] and networks [13, 14]. Nonetheless, interpretation of MS/MS spectra of cross-links is universally recognized as challenging due to the quadratically augmented computational search space and the unequal fragmentation efficiency of the two linked peptides [14, 15]. Confusion of isobaric cross-link and dead-end species, incomplete fragmentation of cross-linked products as well as low mass accuracy fragment ion data also lead to ambiguous matches, and it has been argued that reported cross-links are often incorrect [16]. Approaches that provide more definitive identifications of cross-links are needed.

Cleavable cross-linkers were introduced to reduce some of the above problems through the formation of diagnostic fragmentation patterns, such as reporter ions and peak doublets associated with constituent peptides [17]. A number of cross-linkers cleavable by collision-induced dissociation (CID) have been introduced, such as protein interaction reporters (PIR) [18, 19], disuccinimidyl-succinamyl-aspartyl-proline (SuDP) with its derivatives [20,21,22], BuUrBu (DSBU) [23] with its derivatives [24], disuccinimidyl sulfoxide (DSSO) [25] with its derivatives [26, 27], DC4 [28], and cyanurbiotindipropionylsuccinimide (CBDPS) [29]. All of these incorporate functional groups that form labile sites along the spacer arms of cross-linkers. In a common workflow with cleavable linkers, characteristic cross-link fragments are observed in MS2 spectra. In some cases, such as with BuUrBu, sequences of the cross-linked peptides are identified from backbone fragments in MS2 spectra [30]. In others, such as with DSSO, intact peptide peaks from MS2 spectra are further fragmented [25].

Cross-linkers cleavable by electron transfer dissociation (ETD) have received considerably less attention. ETD is an activation method that leaves many CID labile sites intact and is relatively indifferent to the sequences of peptides [31]. These features make ETD widely implemented in top-down proteomics [32,33,34] and in characterizing post-translational modifications (PTMs) [34,35,36]. Interestingly, ETD sometimes outperforms CID and higher energy collisional dissociation (HCD) for peptide identification of highly charged or low m/z precursors [37, 38]. A recent study with disuccinimidyl suberate (DSS) and bis(sulfosuccinimidyl)suberate (BS3) cross-links showed encouraging results: ETD with supplementary activation of HCD (EThcD) led to the best sequence coverage for highly charged cross-links [39]. Cross-linkers with ETD-cleavable features might enhance the confidence of cross-link identifications. However, only a few ETD-cleavable cross-linkers have been reported. Bis-arylhydrazone (BAH) cross-linked peptides cleave at the N-N hydrazone bond yielding one radical peptide and one even-electron peptide. MS3 experiments on these two peaks generate nearly complete sequence coverage [40]. Based on this, an ETD and CID dual cleavable cross-linker, DUCCT, was introduced and has been applied to standard peptides and proteins [41]. Another ETD-cleavable cross-linker, 1,3-diformyl-5-ethynylbenzene (DEB), forms cross-links through reductive amination. DEB cross-linked peptides maintain the two protonation sites of amino moieties and yield diagnostic ions that reveal the intact unmodified constituent peptides [42]. These encouraging developments suggest that ETD-cleavable cross-linkers with different sizes, polarities, reactive chemistries, and cleavage propensities may find useful applications.

While carboxyl-reactive cross-linkers [43, 44] and heterobifunctional cross-linkers [45] have been reported, most cross-linkers react with primary amines of lysine residues and protein N-termini. With few exceptions [42, 46, 47], nearly all reported amine-reactive cleavable cross-linkers that have demonstrated efficacy with proteins at physiological conditions (pH ≈ 7.4) utilize N-hydroxysuccinimide (NHS) ester reactive groups. A disadvantage they share is that these cross-linkers acylate primary amines, and therefore neutralize their positive charge. This may result in disruptions of the electrostatic properties of proteins and reduction of the charge states of cross-linked peptides. The latter would not facilitate their ETD fragmentation.

Previously, we developed diethyl suberthioimidate (DEST) [48], a cross-linker that reacts with primary amines to form amidino linkages that are highly basic and preserve the charge on amines at physiological pH. DEST has a simple six-carbon spacer arm and thioimidate groups at both ends. The structures of DEST and its cross-linking product as well as the most abundant hydrolyzed dead end [48, 49] are depicted in Scheme 1. Its structure is derived from S-methyl thioacetimidate (SMTA), a peptide and protein labeling reagent comprising one thioimidate reactive group [50,51,52,53,54]. Since proteins labeled by SMTA maintain their functions [55] and native structures [56,57,58], DEST cross-linking is unlikely to perturb structure. Also, the two positive charges retained on the amidino linkages should enable more effective enrichment of DEST cross-links with strong cation exchange chromatography (SCX) [12] compared to cross-links of NHS-ester cross-linkers that lack these positive charges. With DEST, we have probed the native structure of model protein cytochrome c [48], the E. coli ribosome [12], and established the location of E. coli ribosomal protein S1 [11]. Meanwhile, another amidinating cross-linker, dimethyl suberimidate (DMS), also reacts with primary amines to form the same amidine groups. However, as an imidoester rather than a thioimidoester, DMS requires a strongly alkaline environment (pH > 9) to maximize its cross-linking efficiency [59]. It forms almost no cross-links at physiological pH [60] compromising studies of native protein structures. Nevertheless, when the reaction is done at high pH, DMS reportedly produces more highly charged cross-links than the acylating reagent DSS [61]. In addition to their improved SCX enrichment efficiency, the highly charged DEST cross-links are attractive candidates for fragmentation using ETD.

Scheme 1
scheme 1

DEST and its major reaction products

Full size image

In this work, we investigated the ETD fragmentation patterns of DEST cross-linked peptides and dead-end peptides. Exploiting their unique diagnostic features, we developed an ETD MS2-based approach to facilitate identifications of DEST cross-links and enable on-the-fly differentiation between cross-links and dead ends. HCD MS3 experiments on individual peptides from cross-links were automatically implemented to validate identifications from ETD MS2 spectra. This approach was demonstrated in model studies of cytochrome c and ubiquitin. These simple systems were selected to observe cross-link fragmentation propensities because the identities of cross-links are virtually unambiguous. The simplicity of the sample also results in spectra with particularly good signal-to-noise ratios. Because the fragmentation of cross-linked peptides is independent of the complexity of their source, cross-links derived from other biological systems yield spectra that display similar ion types and fragmentation patterns.

Experimental

Chemicals and Reagents

Equine cytochrome c, bovine ubiquitin, proteomics grade trypsin, Tris hydrochloride, Tris base, suberonitrile, and ethanethiol were purchased from Sigma-Aldrich (St. Louis, MO). Anhydrous diethyl ether, Optima® grade water, and Optima® grade acetonitrile (ACN) were ordered from Fisher Scientific (Hampton, NH). Dichloromethane and sulfuric acid were procured from EMD Millipore (Burlington, MA). Sodium chloride and ammonium bicarbonate were supplied by Avantor Performance Materials (Center Valley, PA). Formic acid (FA) was provided by Acros Organics (Waltham, MA).

Protein Cross-linking Experiments

DEST was synthesized in-house as previously described and discussed in Supporting Information [48]. Five micromolar equine cytochrome c and 10 μM bovine ubiquitin were reacted with DEST at molar ratios of 1:100 and 1:50 respectively in 50 mM Tris buffer (pH 7.4). After 4 h, the reactions were quenched with 100 mM ammonium bicarbonate buffer (pH 7.8). Centrifugal filters with 3 kDa molecular weight cutoff (Millipore, Burlington, MA) were used to clear hydrolyzed DEST and exchange the products to 50 mM ammonium bicarbonate buffer (pH 7.8).

Digestion of Cross-linked Proteins

Proteomics grade trypsin was added to DEST cross-linked protein solutions at a 1:20 (w/w) ratio of protease to substrates in 50 mM ammonium bicarbonate buffer. Samples were incubated for 18 h at 37 °C with constant rotation after which FA was added to a concentration of 1%. Digests were dried in a SpeedVac and stored at – 20 °C.

Nano-HPLC/Nano-ESI MS2 or MS3 Analysis

Tryptic digests were analyzed with an Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific, Waltham, MA) coupled with an EASY-nLC 1200 liquid chromatograph (Thermo Fisher Scientific, Waltham, MA). Between 0.1 and 0.5 μg of tryptic digest was loaded onto a C18 trap column with mobile phase A (0.1% FA in water), and then separated by an Acclaim PepMap RSLC C18 column (75 μm× 25 cm, 2 μm, 100 Å) (Thermo Fisher Scientific, Waltham, MA) with a 30-min or 60-min gradient from 2 to 50% mobile phase B (0.1% FA, 80% ACN, and 19.9% water) at 300 nL/min. Eluting species were ionized by nano-electrospray with a spray voltage of 1900 V. Full scan spectra were recorded at a resolution of 120 K.

For MS2 analysis, the data-dependent mode was utilized to fragment the most abundant peaks by CID with a 35% normalized collision energy or ETD with 40 or 50 ms reaction time. The cycle time was set at 5 s. Precursor ions in charge states of 3+ to 7+ were subjected to fragmentation with an isolation window of 2.0 m/z, a maximum ion injection time of 100 ms and AGC of 5E5. All MS2 spectra were recorded in an orbitrap in one microscan with a resolution of 30 K.

ETD fragment ions with a mass spacing of 169.1453 Da were further selected in data-dependent mode and activated by HCD in an ion-routing multipole (HCD cell) with 35% collision energy. The precursor isolation window was set at 2.5 m/z. A maximum ion injection time of 200 ms and AGC of 5E5 were applied. Product masses were analyzed in a linear ion trap with one microscan.

Data Analysis

Raw data files were converted to *.mgf files by Proteome Discoverer 2.1 software (Thermo Fisher Scientific, Waltham, MA). Precursors, charge reduced precursors and neutral loss fragments within 18.5 Da of the precursors were removed from all ETD MS2 mass lists. All spectra were interpreted with an in-house program that computes metrics such as precursor mass errors, numbers of peaks matched for each constituent peptide, percentages of ion intensities matched to each constituent peptide, and scores for tentative identifications. ETD and CID MS2 spectra were interpreted as arising from cross-links, dead ends, loop links, and regular peptides against a target database of equine cytochrome c or bovine ubiquitin concatenated with two different decoy databases: one containing ten randomized versions of the target protein sequence and a second one containing 55 E. coli ribosomal proteins. All MS2 spectra were searched with precursor mass tolerance of 5 ppm and fragment mass tolerance of 10 ppm. Up to three missed cleavages were considered, and oxidation of methionine was allowed as a variable modification. MS3 spectra of ETD product ions were matched to tentative sequences that had been derived from MS2 spectra. All MS3 spectra were searched with a mass tolerance of 0.2 Da, which is compliant with the lower mass accuracy of ion traps. All cross-link identifications were manually checked to verify that more than 50% of the total fragment ion intensity was assigned. For cross-links matched from ETD spectra, a minimum of two mass pair peaks was required. For cross-links found in CID spectra, the percentage of peaks matched to tentative identification was required to exceed 35%, and a minimum of two peaks was required to match fragments derived from each constituent peptide. The above criteria led to FDR < 1% with both ETD and CID approaches. All cross-links satisfying these criteria are tabulated in Tables S1 and S2 of Supporting Information. Most of the data are significantly better than the minimal criteria just mentioned. For example, on average, 82.7% of ETD fragment ion intensities were assignable and 83.6% (123 out of 147) of the spectra contained all four mass pair peaks. Likewise, 83.8% of the CID fragment ion intensities were assignable. In plotted mass spectra, the most intense peak in each isotope cluster is displayed. The precursor ions presented in the spectra were not considered in calculated numbers of matched peaks, percentages of fragment ion intensities matched, and scores.

Results and Discussion

Electron Transfer Dissociation of DEST Cross-linked Peptides

Cytochrome c and ubiquitin were each cross-linked with DEST. Following tryptic digestion, the resulting products were analyzed. For example, a quadruply charged 394.7404 m/z ion from cytochrome c digest was fragmented by ETD and yielded the spectrum shown in Figure 1a. Green peaks result from fragmentation of the longer peptide, that is referred to as peptide α; blue peaks are fragments of peptide β. Precursors, charge reduced precursors, and those with neutral losses of ammonia or water, are in red. Unassignable peaks are in black. This cross-link was identified as MIFAGIKK―KYIPGTK. The identity of the former peptide is established from consecutive z•1 to z•7, z + 12 to z + 16, and c2 to c7 ions as well as y1 and a + 17 ions. The latter is identified from z•1 to z•3, z•5, z•6, c2, and c4 to c6 ions in addition to y1, y4, a + 15, and a + 16 ions. In addition to these backbone fragment ions, four intense peaks in the spectrum can be assigned to Pα-NH21+, Pα + L + NH32+, Pβ-NH21+, and Pβ + L + NH32+, where Pα and Pβ refer to the two intact peptides and L represents the DEST linker. As depicted in Scheme 2, these ions are formed through cleavage of the bond between the amidino group and the adjacent carbon. Their large intensities indicate that ETD favors this cleavage site, consistent with previous observations that amidine groups are readily protonated during ESI [50, 62]. It appears that the protonated amidino groups preferentially attract electrons and are neutralized to form hypervalent species. This induces the homolytic cleavage of the adjacent N-C bonds to form P-NH2•, a radical ion of one peptide, and P + L + NH3, an even-electron species of the other peptide. Two factors favor this dissociation over the formation of conventional c- and z•-ions. First, since the amidino groups are protonated, the migration of a hydrogen atom (H•) is not necessary to initiate this N-C bond dissociation, whereas it would be required to produce backbone c- and z•-ions [63, 64]. Second, the even-electron product ion P + L + NH3 has stable amidine resonance structures that backbone c ions lack.

Figure 1
figure 1

MS2 spectra of the cytochrome c cross-link MIFAGIKK―KYIPGTK4+ fragmented by (a) ETD and (b) CID. To display the relative intensities of smaller peaks, the CID spectrum is also plotted with the y63+ ion off scale in (c). Green peaks are fragments of peptide α; blue peaks are fragments of peptide β. Precursors, charge reduced precursors, and those with neutral losses of ammonia (asterisk) or water (prime) are in red. Unassignable peaks are in black. a + 1, c + 1, and z + 1 ions are represented as A, C, and Z (capitalized) ions; a tilde denotes addition of water. Subscripts following M represent the numbers of hydrogen atoms abstracted from the precursor ion after electron capture

Full size image
Scheme 2
scheme 2

Preferential cleavages induced by ETD. Structures in green belong to peptide α, and those in blue are associated with peptide β. DEST is in black

Full size image

In the above example, it is likely that the two amidine groups sequester two protons, and each peptide retains one. When the proton on the amidine group associated with peptide α is neutralized, it forms the singly charged Pα-NH2• and the doubly charged Pβ + L + NH3. Likewise, when the proton on the amidine group associated with peptide β is neutralized, it produces singly charged Pβ-NH2• and doubly charged Pα + L + NH3. In their singly charged states, Pα-NH2• and Pα + L + NH3 differ in composition by C8N4H17, which corresponds to 169.1453 Da. Pβ-NH2• and Pβ + L + NH3 have the same mass spacing. Components of mass pairs in other charge states, such as singly charged Pα + L + NH3 and Pβ + L + NH3 also appear in Figure 1a. These could arise either from peptides of cross-links protonated in other ways or from some charge reduction process [31]. Additionally, other mass pair-related fragments such as Pα + L + NH3 + 11 + • and Pβ + L + NH3 + 11 + • also appear. These apparently result from P + L + NH32+ ions capturing an electron or P + L + NH31+ ions picking up a hydrogen atom. Both processes are common in ETD [65]. The occurrence of mass pairs enhances the confidence of peptide mass determination and overall cross-link assignments.

The CID spectrum displayed in Figure 1b is strikingly different. The dominance of just a single y63+ ion fragment formed by cleavage of peptide α is evident. To better visualize other less intense fragment ions, the y63+ peak is plotted off scale in Figure 1c. Peptide α is identified based on y2 and y4 to y7 ions along with a2, b2 to b4, and b7 ions; the sequence of peptide β is determined by y4 to y6, b2, and b3 ions. Almost all of the fragments associated with peptide β are extremely low in abundance, on the order of 1% of the base peak. If the experiment were conducted on a less sensitive instrument or performed with more complex protein samples, these fragments might not be detected at all. The inequivalent fragmentation of two constituent peptides, as displayed in this spectrum, is a common phenomenon when cross-links are collisionally dissociated and often leads to ambiguous identifications [15]. It is noteworthy that the Pβ, Pα + L, and Pβ + L peptide ions are generated from peptide/linker cleavages. This observation is consistent with our previous study with singly charged DEST cross-linked peptides showing that the amidino bonds between the linker and peptides are somewhat labile [66].

Figure 2a presents an example of a lower charged cross-link, the ETD MS2 spectrum of the triply charged cytochrome c cross-link KATNE―GKK. Peptide α is confidently identified by successive c1 to c4 ions in addition to a + 14 ions. Neutral losses of CO2 (43.9898 Da) and CH3CONH2 (59.0371 Da) from charge reduced precursors are associated with side chain cleavages of glutamic acid and asparagine respectively. Peptide β is matched to GKK based on c2, z•2, z + 12, and a + 12 ions. Since peptide β is so short, the protein from which it derived would be difficult to establish in most cross-linking experiments; however, this sample contains only cytochrome c. Mass pairs of Pα-NH21+/Pα + L + NH31+ and Pβ-NH21+/Pβ + L + NH31+ are all observed as intense peaks, demonstrating that in the 3+ charge state cleavages at amidino groups are still favored. Note that the cN-1 ions of both peptides (c42+ of peptide α and c22+ of peptide β) are highly abundant. This is consistent with observations of Wysocki and coworkers that cN-1 ions are favored in ETD when peptides are in lower charge states (2+ and 3+) regardless of the composition of the C-terminal residues [67].

Figure 2
figure 2

MS2 spectra of the cytochrome c cross-link KATNE―GKK3+ at 343.8729 m/z fragmented by (a) ETD and (b) CID

Full size image

The CID MS2 spectrum of the same cross-link, shown in Figure 2b, displays intense b42+ peptide backbone fragments. The y1, y2, a1-NH3, b2, b3, and b4 ions match peptide α, KATNE. One side chain loss of HCONH2 (45.0215 Da) from asparagine of the precursor is observed. Peptide β yields only cleavages between the last two residues that result in y1, b2-NH3, and b2 + H2O ions so its identification as GKK is less confident.

As an example of the ETD-induced preferential cleavage of a highly charged ion, Figure 3a displays a 6+, 573.6473 m/z cross-link from cytochrome c. Peptide α is identified as GITWKEETLMEYLENPKK based on z•1, z•2, z•4 to z•14, z•16, y1, y7 to y11, y11, c3 to c7, c9 to c13, c17, and c + 1•14 as well as a + 14, a + 17, and a + 111 ions. Fragment ions c2 to c8, z•1, z•3, z•4, z•8, z + 12 to z + 15, z + 17, and y1 together cover the entire sequence of peptide β, EDLIAYLKK. Note that fragments of peptides α and β have comparable intensities. Four intense peaks are assignable to mass pairs of Pα-NH22+/Pα + L + NH33+ and Pβ-NH21+/Pβ + L + NH32+. This example demonstrates that ETD cleaves a large DEST cross-link in a high-charge state to form mass pairs related to its constituent peptides. Along with these four intense peaks, several similar peaks in different charge states are also assignable. The appearance of multiple mass pairs provides additional confidence in peptide mass assignments.

Figure 3
figure 3

MS2 spectra of the cytochrome c cross-link GITWKEETLMEYLENPKK―EDLIAYLKK6+ fragmented by (a) ETD and (b) CID

Full size image

The CID spectrum of this crosslink, displayed in Figure 3b, shows that collisional activation predominantly dissociates the backbone of peptide α. Based on y3 to y11, y16, b7 to b12, a9, and a10 ions, many of which also lose ammonia and/or water, peptide α is identified. In contrast, peptide β yields less abundant fragments a4-H2O, b4-H2O, and b5-H2O. ETD yields a more confident identification of this peptide than CID.

Note that typical of ETD, precursor ions are not fragmented completely. However, more highly charged cross-links produced by DEST display higher dissociation efficiencies and more abundant peptide mass pairs and backbone fragments. An example demonstrating this is depicted in Figure S2 of Supporting Information. In this case, when the DEST cross-link TLSDYNIQKESTLHLVLR―LIFAGKQLEDGR from ubiquitin is in the 5+ charge state, precursor and charge reduced precursors are predominant though mass pairs are still among the next most intense peaks. As the charge state of this cross-link increases to 6+ and then 7+, the precursors’ associated peaks become less intense while mass pairs and backbone fragments become more abundant.

Sequencing Individual Peptides of a Cross-link by MS3 Experiments

In many cases, cross-links of DEST can be identified through their ETD MS2 spectra using mass pairs to define the masses of peptides α and β and backbone fragments to determine their sequences. However, ambiguous identifications containing isobaric or near-isobaric peptides can still arise when inadequate peptide backbone fragments appear in the ETD MS2 spectrum or when searching a large proteome database. Further activation of ETD product ions would help to clarify ambiguous identifications or validate tentative identifications. Relatively abundant P-NH2 and P + L + NH3 mass pair ions are ideal precursors for MS3 experiments. For example, a quadruply charged cross-link precursor ion at 418.7495 m/z from cytochrome c was activated by ETD and the resulting spectrum is displayed in Figure 4a. The most intense four peaks match two sets of mass pairs and establish the masses of peptides α and β. Peptide α is identified as acetyl-GDVEKGK based on z•1 to z•4, z•6, z + 13 to z + 16, y1, c3 to c6, c + 16, a + 15, and a + 16 ions; peptide β is assigned to KIFVQK from z•1 to z•5, y1, c1 to c5, c + 15, and a + 15 ions. These assignments were next confirmed with MS3 experiments. P-NH2 and P + L + NH3 ions were automatically selected for collisional activation by the instrument based on their mass spacing of 169.1453 Da.

Figure 4
figure 4

(a) ETD MS2 spectrum of cross-link acetyl-GDVEKGK―KIFVQK4+ from cytochrome c. HCD MS3 spectra derived from ETD product ions (b) Pα-NH21+, (c) Pα + L + NH32+, (d) Pβ-NH21+, and (e) Pβ + L + NH32+

Full size image

The HCD MS3 spectrum of Pα-NH2 in Figure 4b yielded backbone fragments of y1, y2, z•2, z + 13, z•6, a2, b2, b3, and b4, that definitively validate the sequence of acetyl-GDVEK(-NH2)GK. The peak at 715 Da involves a neutral loss of C3H7 (43.1 Da or 43.055 Da measured by the orbitrap) from the precursor ion. This corresponds to a partial loss of the side chain from cross-linked lysine that had previously lost NH2• in the ETD process and is consistent with side chain fragmentation of activated z•-ions [68]. Since these lysine residues are prone to further loss from their side chain, analogous backbone fragments such as y3-43 and y4-43 also appear. Significantly, products of radical-driven fragmentation processes, such as z•2, z + 13, z•6, v3, and v4, are also observed in this spectrum, consistent with P-NH2• being a radical ion [68].

The doubly charged partner of Pα-NH2•, Pα + L + NH32+, was also collisionally activated and yielded the spectrum shown in Figure 4c. A full series of y-type ions from y1 to y6 along with b2 to b6, a2, a3, and a6 ions validate the sequence of peptide α. The appearance of a-, b-, y-, and ammonia-loss ions is compliant with previous observations following collisional activation of c ions [68]. The base peak at 154.1 Da corresponds to L+ NH3 that is generated by cleavage between the amidino group and the peptide. Note that this spectrum does not contain radical-derived fragments, consistent with the Pα + L + NH3 ion being an even-electron species.

Figure 4d, the HCD MS3 spectrum of Pβ-NH2•, contains a full series of y ions from y1 to y5, a strong a2, along with b2, b3, and several internal ions, such as IF, FV, and VQ. These peaks confirm the composition of peptide β. Fragments generated through radical-driven processes, such as v3, v4, and w5 ions are also observed. Similar to Pα-NH2• ion, this Pβ-NH2• ion also produces the unique neutral loss of 43.1 Da from the precursor, observed as a peak at 703 m/z.

Figure 4e shows the HCD MS3 spectrum of Pβ + L + NH32+. Consecutive y1 to y5, a2 to a5, and b2 to b5 ions confirm the sequence of peptide β. The base peak at 154.1 m/z again corresponds to L + NH3. As expected, no radical ions are detected.

The four MS3 spectra together demonstrate that the product ions obtained by fragmentating P-NH2• and P + L + NH3 ions provide sufficient information to validate the compositions of the two peptides. P-NH2• ions dissociate in different ways from their even-electron P + L + NH3 counterparts, and combination of the two spectra should improve peptide and cross-link identifications [69,70,71,72].

Electron Transfer Dissociation of DEST Dead-end Peptides

Dead ends are by-products of cross-linking reactions that form when one end of the cross-linker reacts with a protein and the other end reacts with small molecules from the solution, such as water or a buffer component. Dead ends not only augment the search space but also complicate identifications when cross-links are isobaric with dead ends. Diagnostic ions that could distinguish dead ends from cross-links would improve the confidence of cross-link identifications. An ETD MS2 spectrum of a DEST dead end from ubiquitin is displayed in Figure 5. The triply charged precursor ion at 500.9559 m/z is confidently identified as LIFAGKQLEDGR with a 154.1106 Da hydrolyzed dead end. Numerous z•-ions along with z + 18 to z + 111, consecutive c5 to c11, y6 to y8, y10, as well as a + 16 ion support this identification. The intense singly charged ion at 172.1444 m/z corresponds to hydrolyzed linker+NH3 (C8N3OH18) and serves as a dead-end reporter ion. Its complementary ion, the singly charged peptide-NH2• is also abundant. Its formation mechanism, analogous to that in Scheme 2, is depicted in Figure S3. In addition, reaction of one end of DEST with a peptide and the other end with free ammonium generates a dead end that adds 153.1266 Da (C8N3H15) to the peptide. This type of dead end yields an abundant ion at 171.1604 m/z (C8N4H19) (Figure S4) through the same mechanism as shown in Figure S3.

Figure 5
figure 5

ETD MS2 spectrum of the DEST hydrolyzed dead-end LIFAGK (+ 154.1106 Da) QLEDGR3+ from ubiquitin

Full size image

The unique masses and high abundance of 172.1444 and 171.1604 m/z peaks make them ideal reporter ions for dead ends. The distinctive features of reporter ions for dead ends and mass pairs for cross-links offer a simple but reliable means to differentiate dead ends from cross-links. When the labile sites of cleavable cross-links are incorporated along the spacer arm, as with most reported cleavable cross-linkers, dead ends and cross-links are endowed with identical labile sites. Consequently, dead ends and cross-links preferentially cleave at the same sites and often produce mass pairs with identical spacings. Since MS3 peak selection is often based on predefined mass differences [73], MS3 experiments can be triggered by dead ends. In contrast, the DEST spacer arm remains intact during ETD activation so dead ends and cross-links display different ion products. This feature should significantly reduce the chance of dead ends being fragmented in MS3 experiments.

Comparison of Backbone Cleavage Induced by CID and ETD

An important distinguishing characteristic of ETD spectra of DEST cross-links is that they contain numerous backbone cleavages in addition to the amidino cleavages of the DEST-peptide linkages. In contrast with some CID cleavable cross-linkers, these backbone cleavages enable at least tentative identification of peptides in the MS2 experiments. It is interesting to compare the relative proclivities of CID and ETD for inducing backbone cleavages in DEST cross-links. To do this, we considered all confidently identified cross-links from cytochrome c and ubiquitin tabulated in Tables S1 and S2 of Supporting Information except for those with constituent peptides of equal length. The average lengths of longer peptide α and shorter peptide β are represented as black bars in Figure 6. It is evident that the lengths of peptides α and β are very similar in both sets of data. The red bars display the percentage of fragment ion intensity that is associated with cleavage of the two peptides in the two experiments. Intensities of characteristic ions, such as P and P + L yielded by CID as well as P-NH2• and P + L + NH3 formed with ETD are counted towards their corresponding peptides. It is clear that the relative fragmentation of peptides α and β by ETD is simply proportional to their lengths indicating that the average likelihood of a backbone bond being fragmented by ETD is similar in both longer and shorter peptides. In contrast, CID cleaves peptide β less efficiently than peptide α. This is probably because longer peptides have larger collision cross sections and the energy gained during activation may not be effectively transferred to the smaller peptide. The distribution of fragment intensities assigned to peptides α and β also appears to be wider in CID than with ETD as indicated by error bars representing one standard deviation. It is noteworthy that CID backbone cleavage propensities can be quite sequence dependent [74,75,76] and this can make it harder to identify peptides of cross-links. This is exemplified in Figure 1b, where the CID favored cleavage between I and F leads to a2, b2, and y63+ ions, the latter being particularly intense, but most other backbone fragments are much lower in abundance. Since positive charges repel each other, the protons on a DEST cross-link are likely to be distributed over both peptides. Therefore, it is not surprising to see comparable backbone fragmentation induced by ETD. This characteristic should improve the confidence of identifying both peptides involved in a cross-link, which is why cross-links of DEST can often be confidently identified based on just ETD MS2 spectra. The above statistical analysis was conducted using cytochrome c and ubiquitin cross-links, and it could be improved with a larger study. In comparison to prevalent workflows with other cleavable cross-linkers that often require the combination of one MS2 spectrum with multiple MS3 spectra to identify a cross-link, ETD fragmentation of DEST cross-links yields both mass pairs and backbone cleavages that enable their identification.

Figure 6
figure 6

Comparison between peptide length (black bars) and the percentage of total fragment intensity associated with cleavage of each peptide (red bars) for CID and ETD experiments using the DEST cross-linker. Standard deviations are indicated by error bars

Full size image

Key Attributes of ETD Strategy with DEST Cross-links

The results presented above demonstrate that when fragmented with ETD, DEST cross-links yield peptide mass pairs along with backbone fragments while dead ends form reporter ions. This allows us to propose a strategy for arriving at confident cross-linked peptide identifications. As depicted in Scheme 3, precursors are initially fragmented by ETD. In the resulting MS2 spectra, cross-links are diagnosed by mass pairs and identified through additional peptide backbone fragments. Dead ends are discernable from cross-links on-the-fly due to their unique reporter ions and the absence of mass pairs. Detection of a dead-end reporter ion along with the absence of two sets of mass pairs provides a simple yet effective means to discriminate dead ends from cross-links. To validate MS2 identifications of cross-links, ETD mass pairs, product ions with mass spacings of 169.1453 Da, are fragmented in MS3 experiments. Dissociation of either the P-NH2• or the P + L + NH3 ion of a mass pair may be sufficient, although having the complementary data from both ions is generally advantageous.

Scheme 3
scheme 3

Unique ETD fragment patterns of DEST cross-links and dead ends

Full size image

Applications of the DEST Cross-linking Strategy to Model Proteins

Previously, SCX enrichment and fractionation have proven to be effective tools to identify more DEST cross-links in complex samples. The two positive charges retained on the amidine groups certainly facilitate this process [12]. Due to the simplicity of the samples in this study, SCX enrichment was not employed. All cross-links of cytochrome c and ubiquitin confidently identified are listed in Tables S1 and S2 respectively. Comparisons of unique cross-linking pairs identified by ETD and CID approaches are displayed in Tables S3 and S4. The numbers of unique DEST cross-links observed in cytochrome c and ubiquitin are comparable to those that have been reported in previous studies with DSSO [25] and DUCCT [41]. A few of the cross-linking sites are different (Table S5), possibly due to differences in cross-linker polarities. The distances between Cα of each linked residue pair were calculated based on their crystal structures, and all are less than 24 Å, as expected for cross-linkers of this length.

Conclusions

We have demonstrated a novel ETD-induced cleavage mechanism associated with amidine groups that efficiently cleaves DEST cross-links. Cross-links of DEST preferentially dissociate at the amidine groups yielding mass pairs P-NH2•/P + L + NH3 of both constituent peptides. Simultaneously, backbones of the two peptides cleave. As a result, DEST cross-links can often be identified from single ETD MS2 spectra. Additionally, dead ends generate strong reporter ions through the same mechanism. The substantially different diagnostic characteristics of cross-links and dead ends enable them to be distinguished. On-the-fly MS3 experiments on peptide mass pairs can be used to validate sequences tentatively identified in MS2 spectra. P-NH2• radical ions and their even-electron P + L + NH3 counterparts dissociate to yield somewhat different products. This DEST cross-linking workflow provides a simple but reliable approach to confidently identify cross-links and potentially enable high-throughput experiments with protein complexes or networks.