Enhancing Cysteine Chemoproteomic Coverage through Systematic Assessment of Click Chemistry Product Fragmentation

Abstract

Mass spectrometry-based chemoproteomics has enabled functional analysis and small molecule screening at thousands of cysteine residues in parallel. Widely adopted chemoproteomic sample preparation workflows rely on the use of pan cysteine-reactive probes such as iodoacetamide alkyne combined with biotinylation via copper-catalyzed azide–alkyne cycloaddition (CuAAC) or “click chemistry” for cysteine capture. Despite considerable advances in both sample preparation and analytical platforms, current techniques only sample a small fraction of all cysteines encoded in the human proteome. Extending the recently introduced labile mode of the MSFragger search engine, here we report an in-depth analysis of cysteine biotinylation via click chemistry (CBCC) reagent gas-phase fragmentation during MS/MS analysis. We find that CBCC conjugates produce both known and novel diagnostic fragments and peptide remainder ions. Among these species, we identified a candidate signature ion for CBCC peptides, the cyclic oxonium-biotin fragment ion that is generated upon fragmentation of the N(triazole)─C(alkyl) bond. Guided by our empirical comparison of fragmentation patterns of six CBCC reagent combinations, we achieved enhanced coverage of cysteine-labeled peptides. Implementation of labile searches afforded unique PSMs and provides a roadmap for the utility of such searches in enhancing chemoproteomic peptide coverage.

Graphical Abstract

INTRODUCTION

Chemoproteomics has become a workhorse technology for functional biology and drug discovery efforts, enabling target deconvolution for a number of bioactive molecules, clinical candidates, and even drugs.^1-4 Most sample preparation workflows rely on the same general strategy. Cells or cellular lysates are first labeled with chemical probes that incorporate an electrophilic moiety or a photoactivatable group for irreversible modification of the probe-binding proteins. Probe-labeled proteomes are then conjugated via bioorthogonal copper-catalyzed azide–alkyne cycloaddition (CuAAC) or ’click chemistry’ to biotin or desthiobiotin capture reagents (typically azide-or alkyne-modified).^5-7 The use of bioorthogonal chemistry for protein and peptide capture, in contrast with probes derivatized directly with biotin, offers the advantages of decreased probe size and increased cell permeability.^8,9 After enrichment on avidin resin and proteolytic digest, labeled peptides are then identified through liquid chromatography—tandem mass spectrometry (LC-MS/MS) analysis.

Cysteine-reactive probes are particularly favored for chemical probe development campaigns, given the unique chemistry of the cysteine thiol¹⁰ and its important roles in functional and therapeutically relevant targets.^11,12 Chemoproteomics studies using cysteine-reactive probes such as iodoacetamide alkyne have demonstrated that the proteome harbors thousands of potential ligandable cysteines.^13,14 A central challenge of these studies is that they still only sample a small fraction of all cysteines in the human proteome.¹⁵ Closing this gap, which can be ascribed to a host of factors spanning peptide abundance, length, and ionization efficiency, among others, requires the development of optimized chemoproteomic sample preparation and data analysis workflows. Supporting this premise, our recent work revealed that using a SP3 cleanup method combined with on-line sample fractionation could achieve 5.5-fold increased coverage compared to our prior methods.¹⁵

One aspect of global cysteine chemoproteomics workflows that remains unexplored is the impact of reagent gas-phase fragmentation during MS/MS analysis and whether such fragmentation might impact data analysis or coverage of labeled peptides. The general inertness of the triazole formed during click chemistry to gas-phase reactions has been widely documented.^16-18 In one notable exception to this paradigm, collision-induced dissociation (CID) of triazole-modified peptides has been found to afford nucleophilic displacement of the N3 nitrogen in the 1,2,3-triazole ring.^19,20 Whether and to what extent 1,2,3-triazoles show additional gas-phase fragments remains unexplored.

In contrast with the seeming gas-phase inertness of the triazole moiety, the MS/MS fragmentation of biotin conjugates has been widely documented, including the application of signature reporter ions for the identification of biotinylated peptides. These studies have extended to the fragmentation of biotin–lysine conjugates, biotin–phenol modified tyrosine residues (e.g., conjugates formed during APEX proximity labeling²¹), and biotin conjugates formed during covalent modification of cysteine,²² serine,²³ and tyrosine residues.^24,25 Dehydrobiotin (m/z 227.0845) is a diagnostic ion shared across most biotin fragmentation studies.

The variable modification approach employed by search algorithms (e.g., SEQUEST,²⁶ ProLuCID²⁷) widely used for chemoproteomics data analysis looks for fragment ions that contain intact biotin conjugates. Some algorithms (e.g., CoMEt²⁸) allow for neutral losses to be specified. Open search is an alternative approach in which the peptide mass is determined by matching fragment ions without knowledge of the precursor mass.^29,30 For labile modifications, open searches, together with the more focused mass offset³¹ or multinotch³² searches, enable spectral matching for a larger proportion of observed fragmentation ions. Offset searches of glycoproteomics datasets have been found to significantly increase the number of successfully annotated spectra.³³

Here, we report the discovery that gas-phase fragmentation of triazole-and biotin-modified peptides affords characteristic fragment ions, including previously reported and newly identified species. Using open and mass offset search strategies, we conduct an in-depth analysis of the relative intensities and specificity of these signature fragments. By varying the nature of the labeling reagent, including linker length and position of the triazole, we achieved improved coverage of labeled peptides. Collectively, our study demonstrates the utility of labile ion and mass offset search strategies in the analysis of chemoproteomics datasets and reveals the ubiquity of triazole fragmentation in MS/MS analysis of chemically modified peptides.

EXPERIMENTAL SECTION

Proteomic Sample Preparation.

Samples were prepared as reported previously.¹⁵

Liquid Chromatography—Tandem Mass Spectrometry (LC-MS/MS) Analysis.

The samples were analyzed by liquid chromatography—tandem mass spectrometry using a Thermo Scientific Orbitrap Eclipse Tribrid mass spectrometer or coupled with a High-Field Asymmetric-Waveform Ion Mobility Spectrometry (FAIMS) Interface.

Protein and Peptide Identification.

Raw data collected by LC-MS/MS were searched with MSFragger (v3.3) and FragPipe (v17.1). The MS raw files and search results have been deposited to the ProteomeXchange Consortium³⁴ via the PRIDE partner repository³⁵ with the dataset identifier PXD028853 and PDX030737. File details are listed in Table S6.

Additional experimental details can be found in the Supporting Information.

RESULTS

Identification and Verification of Oxonium-Biotin as the Major Fragmentation Product of Cysteine Biotinylation via Click Chemistry (CBCC).

During our prior chemoproteomics analysis of the cysteinome¹⁵ following the cysteine biotinylation via click chemistry (CBCC) (sample preparation workflow shown in Figure 1A), we observed that nearly all MS/MS spectra displayed signals at m/z 227.085 and m/z 284.143 (Figure 1B). The m/z 227.085 fragment ion matches closely with the calculated exact mass of dehydrobiotin F1, which has been previously reported as a signature ion for biotin-modified peptides.^22,24,25,34 While the m/z 284.143 fragment ion has not, to our knowledge, been previously reported as a signature ion associated with biotin, close inspection of potential labile bonds in the biotin-triazole product M1 suggested that this signature ion likely results from fragmentation at the N(triazole)─C(alkyl) bond (N6─C5) together with cyclization to afford cyclic oxonium species F2, termed here oxonium-biotin (Figure 1C).

Figure 1.

(A) Workflow for cysteine chemoproteomic identification. Cysteines are capped with the pan cysteine-reactive probe iodoacetamide alkyne (IAA or 1), followed by copper-catalyzed azide–alkyne cycloaddition (CuAAC or click chemistry) conjugation to biotin-azide 2, SP3 sample cleanup, tryptic digest, neutravidin enrichment, and LC-MS/MS analysis. (B) Representative MS/MS spectrum of biotinylated cysteine peptide. (C) Scheme of potential fragmentations of the click product and the resulting fragment ions. The figure shows blue-labeled remainder modification of the biotinylated cysteines and red-labeled fragment ions. (D) Representative MS3 spectrum of biotinylated cysteine b/y ions. MS3 analysis was performed on the b6 ion of a representative cysteine biotinylated peptide ITGC(+463.23656)ASPGK. The MS2 spectrum is in blue and MS3 spectrum of the biotinylated b6 ion is in red. (E) Representative MS/MS spectrum of cysteine peptide labeled with biotin-d-azide 3.

As this fragmentation of biotin-triazole conjugate is, to our knowledge, unprecedented, we next sought to further vet the identity of the m/z 284.143 fragment ion. MS3 analysis of biotinylated peptides revealed that the dehydrobiotin and biotin-oxonium ions were the only major fragment ions produced from CBCC, and both ions were observed with high relative intensity (Figure 1D). Consistent with prior MS3 studies, for example, those that rely on isobaric tags,³⁵ these data support the fact that these fragment ions are likely derived from the cysteine biotin modification.

To further pinpoint the source of m/z 284.143 fragment ion, we synthesized and applied a deuterated isotopologue 3 of our biotin-azide capture reagent (Figure S1) to chemoproteomic analysis of cysteine-containing peptides. Consistent with the mass shift afforded by the –d6 modification, MS/MS analysis revealed the production of a new characteristic fragment ion with m/z 290.180 (Figure 1E). A comparison of samples labeled with the “light” 2 and “heavy” 3 biotin reagents revealed comparable PSMs and unique peptides (Figure S2A). For samples labeled with 1:1 heavy and light biotin-azide tags, quantification of the relative intensities of labeled peptides using FragPipe’s IonQuant³⁶ revealed a median log₂(H/L) ratio close to zero with minimal variance across peptides quantified, which indicated equal labeling efficiency for both reagents (Figure S2B). Consistent with prior reports,³⁷ we do observe a minor (~3 s) average retention time shift for peptides labeled with the deuterated reagent 3 (Figure S2C).

In-Depth Analysis of the Fragmentation Products of CBCC.

Having pinpointed the likely nature of the m/z 284.143 fragment ion, we next expanded our analysis of the CBCC fragmentation products with the overarching objective of more fully deciphering how chemoproteomic-labeled peptides behave in the gas phase. Inspired by recent advances in the identification of peptides with labile modifications, we opted to test whether the mass offset search of the MSFragger search engine^33,38 could be extended to the analysis of CBCC fragmentation productions.

We first extended PTM-Shepherd to identify biotinylation-specific spectral features corresponding to fragments of the biotinylation modification (diagnostic ions) or partially fragmented biotinylation remaining on the peptide (diagnostic peptide remainder masses) (Figure 2A). For biotinylated PSMs, spectra had all possible diagnostic features calculated and aggregated into a common histogram. For diagnostic ions, ions were inserted into the histogram at their m/z observed in the experimental spectrum. For peptide remainder masses, the distance between every ion in the experimental spectrum and the theoretical, unmodified peptide mass was calculated and inserted into the common histogram. Recurring features were identified based on bin height. The recurring features were then quantified across modified and unmodified spectra, and their specificity for biotinylation was assessed via comparisons to the unmodified PSMs.

Figure 2.

(A) Workflow for the identification of diagnostic ions (red) and peptide remainder masses (blue). All potential diagnostic ions (top) and peptide remainder masses (bottom) were calculated using an extended version of PTM-Shepherd (left panel). Recurring features (light blue dash boxes) across the spectra were identified by aggregating diagnostic features (middle panel). Recurring feature intensities were then extracted from modified and unmodified spectra, and statistically significant differences in feature intensities between modified and unmodified peptides were identified (right panel). (B) Representative MS/MS spectra of biotinylated cysteine peptides with annotations of fragment ions (F1–F5), peptide remainder ions (M2, M3), as well as b/y ions with different cysteine modifications. (C) Frequency distribution and (D) relative intensities of signature fragment ions and peptide remainder ions. Experiments were performed in triplicate. Bar plots display mean values across replicates, and box plots display minimum, first quartile (Q1), median, third quartile (Q3), and maximum values of the sample (similarly hereinafter). All data can be found in Table S1.

Several diagnostic fragment ions and peptide remainder masses were identified, including m/z 227.085 (dehydrobiotin), m/z 284.143 (oxonium-biotin), m/z 424.249, m/z 327.185, and m/z 301.169 as fragment ions and +180.101, +152.095, and +463.237 (intact modification) as peptide mass shifts. Based on the diagnostic masses and structure of the modification, several fragmentation pathways were proposed (Scheme 1). We expected that N6─C5 bond cleavage and cyclization with C1 carbonyl oxygen of the M1 ion would afford biotin oxonium ion F2 and peptide remainder ions M2 and M3, with M3 produced by an additional loss of nitrogen gas from M2.³⁹ M1 would also fragment to produce both the aforementioned dehydrobiotin F1 and diagnostic fragmentation F3, which is generated by cleavage of the C12─N13 bond. While still somewhat speculative, we anticipate that F4 and F5 form as a result of cleavage of the C9─C8 bond to form putative intermediate I-1 followed by a gain of 4 hydrogen atoms and loss of nitrogen and acetylene gas, respectively. Manual annotation of MS2 spectra of representative peptides using an integrative proteomics data viewer (PDV)⁴⁰ (Figures 2B and S3) identified spectra corresponding to fragments F1–F5 (red ions), peptide remainder ions M2 (green), M3 (blue) as peptide ions with a diagnostic mass shift and b/y ions with corresponding cysteine modifications.

Scheme 1. Fragment Ions and Remainder Modification on Cysteines from the Fragmentation of Cysteine Biotinylation with IAA 1 and Biotin-Azide 2 via Click Chemistry^a.

^aBlue-labeled remainder modification of the biotinylated cysteines and red-labeled fragment ions.

Looking across our entire labile ion search dataset (Table S1), we next analyzed the frequency distribution of all of the identified fragment ions. Unsurprisingly, frequency analysis showed that F2 oxonium-biotin ion is the most common fragment ion generated from LC-MS/MS analysis of CBCC samples, with nearly 100% occurrence in all MS2 scans. The F1 and F3 fragment ions were also ubiquitous, appearing in >80% of MS2 scans (Figure 2C). In contrast, M1 (p + 463.237)-, M2 (p + 180.101)-, and M3 (p + 152.095)-modified peptide remainder ions are only found in ~25% of all MS2 scans. The relatively modest frequency distribution of modified peptide precursor ions can be rationalized in part by only partial fragmentation of the biotin modification, natural low abundance of unfragmented peptide backbone on MS2 level, together with the inherent ambiguity and challenges associated with accurate annotation of tandem mass spectra for modified peptides.⁴¹

As demonstrated by our representative annotated spectra (Figure 2B and S3), an inspection of MS2 scans for representative peptides revealed that the oxonium-biotin F2 fragment ion was almost always the dominant ion, with 100% relative intensity. To more rigorously quantify the relative intensities of each characteristic fragmentation ion, we established a customized PTM-Shepherd workflow in FragPipe that reports the relative intensities for all of the fragment and peptide remainder ions of biotinylated cysteine peptides (See Labile-PTMShepherd-sample.workflow). Consistent with our observations from manual inspection of the spectra, dataset-wide quantitation revealed a 100% median intensity for the oxonium-biotin ion. All other fragment and peptide remainder ions were found to have substantially reduced median of relative intensities spanning 0–7.6% (Figure 2D and Table S1), which is consistent with their low frequency of detection. The dominance of the biotin-oxonium F2 ion in nearly all spectra makes it an intriguing candidate signature ion for CBCC-labeled peptides. We next subjected CBCC peptide samples to a collision energy ramping experiment varying normalized collision energy (NCE) from NCE = 20 to NCE = 60 for higher-energy C-trap dissociation (HCD) LC-MS/MS experiments (Figure S4 and Table 1). We found that the greatest number of biotinylated PSMs and unique peptides were identified at NCE = 30 and the lowest number of peptides at NCE = 60 (Table 1). Analysis of the relative intensities of the fragment and peptide remainder ions by labile search revealed a dramatic increase in the intensity of diagnostic ions, including m/z 227.085 and m/z 284.143 with ramping of collision energy in HCD modes, while other fragment ions (F3–F5) remained relatively low intensity. In contrast, the intensities of signature peptide precursor ions decreased as collision energy was increased, consistent with over-fragmentation of precursor ions (Figure S4).

Table 1.

Number of Biotinylated PSMs and Peptides Identified in HCD Fragmentation Mode with Varying NCE

NCE (%)	20	25	30	35	40	50	60
PSMs	16 570	16 664	17 214	15 412	14 105	5618	953
peptides	11 882	11 757	12 037	11 155	10 366	4602	874

Comparison of ion-trap-based collision-induced dissociation (CID) to HCD revealed 30% fewer IDs for CID (Table S2) together with much less pronounced NCE-dependent changes in relative ion intensities (Figure S5). Based on these findings, we selected HCD at 30% NCE as optimal for subsequent experiments.

Assessment of Specificity of the Fragmentation Products.

High specificity is essential for a fragment ion to serve as a signature or diagnostic ion for a specific chemical modification. To test the specificity of each observed fragment ion for precursor ion modification state, we next quantified the frequency distribution of the fragment and peptide remainder ions for PSMs harboring or lacking cysteine modifications (with or without (w/o) mass shift, respectively; Figure 3A,B). This comparison of biotin-labeled and unlabeled PSMs revealed high and statistically significant specificity for all three signature peptide precursor ions (M1, M2, and M3; Figure 3A-Gray), with the frequency of detection of these ions near 0% for unlabeled PSMs—for samples that had been enriched on neutravidin resin (Figure 3A), the unlabeled peptide subset constitutes non-specific, unmodified peptides carried through the neutravidin enrichment. We observe more moderate, albeit still statistically significant, specificity for all fragment ions (F1, F3, F4, and F5; Figure 3A-Gray), with detection frequencies of ~50% for PSMs lacking the biotin mass shift compared with almost 100% for biotin-modified PSMs. In contrast, the F2 ion was identified close to 100% in both modified and unmodified PSMs. The generally modest to low specificity observed for the fragment ions can likely be rationalized by co-isolation of the modified and unmodified cysteine peptides as precursor ions. Upon formation, the fragment ions lose peptide sequence specificity and therefore can be matched to both modified and unmodified co-isolated peptides. In contrast, peptide remainder ions retain sequence information and therefore maintain specificity despite co-isolation with unmodified precursors. This model is analogous to the findings that quantification of complementary peptide ions offers improved accuracy compared with that afforded by reporter ions generated by peptides labeled with TMTPro reagents.⁴²

Figure 3.

Frequency distribution of fragment and peptide remainder ions in PSMs with or without cysteine modifications in (A) neutravidin enriched samples or (B) unenriched samples with a narrower isolation window (0.5 Da compared with our standard isolation window 1.6 Da) and with or without FAIMS. (C) Relative intensity of oxonium-biotin in PSMs with or without cysteine modifications with narrower isolation window, with FAIMS, with both, or without. Samples were enriched on neutravidin resin prior to analysis. (D) Peptide length analysis of cysteine peptides identified with high- or low-intensity fragment ion F2. Statistical significance was calculated with unpaired Student’s t-tests with equal variance, *p < 0.05, **p < 0.01, ***p < 0.001, and NS p > 0.05. Experiments were performed in triplicate. All data can be found in Table S3.

By decreasing the proportion of CBCC-labeled peptides in the MS sample and therefore the frequency of modified and unmodified peptide co-isolation, we expected that we could achieve increased specificity of fragment ions for CBCC modified peptides. Therefore, we next extended the fragment ion frequency analysis to unenriched (preneutravidin enrichment) trypsin digested CBCC-labeled lysates (Figure 3B). As expected, an increase in specificity for all fragment ions was observed for the unenriched samples, as indicated by the decrease in detection frequency compared with the neutravidin enriched samples, as shown in Figure 3A. Surprisingly, even in these unenriched samples, the specificity of the F2 ion remained modest, with a frequency of detection in unlabeled peptide PSMs of >80%.

Given the observed high intensity of the oxonium-biotin F2 fragment ion for PSMs lacking a biotin modification, both for neutravidin enriched and unenriched samples, we next sought to assess whether the specificity of fragment ions, particularly the F2 fragment ion, could be improved by modifications to our data acquisition method. On-line sample fractionation using a FAIMS ion mobility device has been found to increase sample quantitation at the MS2 level, affording decreased ratio compression for TMT samples.⁴³ Therefore, we next applied FAIMS ion mobility to CBCC-labeled samples and, as described above, compared the frequency distribution of each ion for biotinylated vs nonbiotinylated precursor ions. We find that the use of FAIMS improves the specificity of all fragment ions (Figure S6). Similarly, the use of a narrower precursor isolation window (0.5 Da compared with our standard isolation window of 1.6 Da) also afforded improved specificity.

By combining both FAIMS and the narrower precursor isolation window, the frequency of detection of fragment ions F1, F3, F4, F5 dropped to <20% for neutravidin enriched samples (Figure 3A-Red) and to close to 0% for unenriched samples (Figure 3B). For the F2 oxonium-biotin, while the frequency of detection for unlabeled peptides remained relatively high (~80% for neutravidin enriched samples and ~20% for unenriched samples, Figure 3A,B, respectively), median ion intensity analysis (Figure 3C) revealed that addition of FAIMS, the narrow precursor isolation window, and, most significantly, the combination of both FAIMS and the narrower precursor isolation window reduced the median intensity of the F2 ion to 20%. Collectively, this analysis points to the potential utility of the F2 ion as a characteristic ion for CBCC peptide identification, with particular relevance in the analysis of unenriched samples using both a FAIMS device and a narrow precursor isolation window. With the implementation of an intensity threshold cutoff of ~40%, we expect that this ion will also prove diagnostic of biotinylation state for neutravidin enriched samples.

We next assessed whether specific peptide features correlated with F2 ion formation. We classified the identified cysteine peptides into two groups—peptides with high (100%) or low (<100%) relative intensity of F2. Neither peptide charge state nor amino acid content was found to correlate with F2 intensity (Figure S7). In contrast, peptide length was found to be significantly related to ion intensity (Figure 3D), with an observed mean length of 15 amino acids for high-intensity groups and 19 amino acids for low-intensity groups.

Investigation of the Fragmentation of CBCC Peptides Labeled with Different Reagent Combinations.

To investigate the generalizability of our findings across chemoproteomics workflows, we next expanded our analysis to include additional labeling reagent combinations (Scheme 2 and Table 2), with the expectation that each reagent pair would afford a unique fragment profile. CBCC samples were prepared and analyzed using three reagent combinations with different iodoacetamide alkyne and biotin-azide probes—A (IAA 1 + Biotin-azide 2), B (IAA 1 + Biotin-d-azide 3), and C (Phenyliodoacetamide alkyne (PIAA) 4 + Biotin-azide 2)⁴⁴ (Scheme 2). Labile ion search revealed that the observed frequency distribution of peptide remainder and fragment ions for each reagent combination matched with the expected fragmentation pattern (Table 2, Scheme 2, with expected ions highlighted in red and blue in Table 2). The F2 ion was detected for both reagent combinations prepared using biotin-azide reagent 2.

Scheme 2. Major Fragmentation Anticipated in Different Reagent Combinations: (A) IAA + Biotin-Azide; (B) IAA + Biotin-d-Azide; (C) PIAA + Biotin-Azide; (D) IAAz + Biotin-Alkyne; (E) IAA + Biotin-C4-Azide; and (F) IAA + N-Ac-Val-Azido-Lys-Lys(biotin)-Gly-COOH Tetrapeptide Reagent^a.

^aBlue-labeled peptide remainder modification of the biotinylated cysteines and red-labeled fragment ions.

Table 2.

Frequency Distribution of Signature Fragment Ions and Peptide Remainder Ions Using Different Labeling Reagent Combinations to Generate CBCC Modified PSMs

		m/z
	Reagent Combination	p+152.095 M3	p+172.064 M4	p+98.060 M5	284.143 F2	290.180 F6	297.138 F7	298.158 F8	430.212 F10
A	IAA + Biotin-azide	25.44	1.42	1.70	100.00	0.02	0.06	0.60	0.70
B	IAA + Biotin-D-azide	30.75	1.85	2.36	1.22	100.00	0.09	0.67	0.60
C	PIAA + Biotin-azide	1.03	42.52	1.33	100.00	0.03	0.15	0.44	0.39
D	IAAz + Biotin-alkyne	4.84	1.78	22.05	1.50	0.12	12.89	0.60	0.43
E	IAA + Biotin-C4-azide	23.20	1.83	1.75	2.11	0.00	0.04	99.66	0.49
F	IAA + tetrapeptide	12.90	1.56	1.79	3.05	0.00	0.04	0.36	99.99

The specificity of expected ions for all reagent combinations was observed to be high. For example, for the reagent combination B, the F2 ion was only observed in 1.22% of all PSMs and, instead, the F6 (m/z 290.180, Oxonium-d-biotin) was detected in 100% of PSMs. Similarly, both reagent combinations A and B afforded the expected p + 152.095 peptide remainder ion with a detection frequency of 25 and 29%, respectively. For combination C, the p+152.095 ion was only detected in 1% of all scans and, instead, peptide remainder ion p + 172.064 was detected in 42% of all PSMs, which matches the phenyl azirine M4 resulting from fragmentation of the N(triazole)─C(alkyl) bond of the click product. As with A and B, for C, low specificity was observed for the F1 and F2 ions, with detection in most PSMs that lacked the biotin modification (Figure S8A).

Given the seeming exquisite selectivity of the identified fragment ions and different fragmentation patterns for reagent combinations A–C, as shown in Table 2, we next expanded our panel of reagents to include three additional reagent pairs (D (IAAz 5 + Biotin-alkyne 6), E (IAA 1 + Biotin-C4-azide 7), and F (IAA 1 + N-Ac-Val-Azido-Lys-Lys(biotin)-Gly-COOH tetrapeptide reagent 8)), with the goal of further probing the nature of the fragment ions and determining whether structural modifications to the reagents would impact the identification and production of fragments detected. Using reagent combinations D and E, we tested whether changes to the length of the C3 alkyl chain and orientation of the triazole would impact the frequency or intensity of the fragment and peptide remainder ions. For reagent combination D, in which the triazole orientation is reversed, we observed the formation of p + 98.060 peptide remainder ion and m/z 297.138 fragment ion (Table 2), which are consistent with our predicted reagent fragmentation to form 1,3-oxazonium M5 peptide remainder ion and biotin-azirinium F7 fragment ion (Scheme 2).

We observed a striking decrease in the detection rate of oxonium species formation with the reagent combination D when compared with our standard combination A (compare 100% for F2 species with 22% for M5). We find that these reagent differences extend to the frequency of ion detection and relative intensities of observed ions (Figure S8A,B). We suspect that the swapped azide–alkyne configuration results in less favorable fragmentation of the N(triazole)─C(alkyl) bond and/or decreased gas-phase stability of the afforded product ions.

Extension of the alkyl linker by one carbon in reagent combination E afforded production of m/z 298.158 fragment ion, which is consistent with the structure of F8 oxonium-C4-biotin (Scheme 2). While the frequency of detection of F8 was comparable to that observed for F2 (Table 2, 99.7 vs 100%, respectively), the mean relative intensity of F8 ions was substantially decreased to 29.1% (Figure S8B). Given the general instability of primary carbocations, we expect that this ion also cyclizes as proposed for the F2 fragment ion, although we cannot exclude that the decreased favorability of seven-membered ring formation might preclude cyclization. The biotin-C4-azide 7 labeling experiment also afforded improved specificity for peptide modification state when compared to the biotin-azide 2 labeled samples, which was expected and was consistent with the less favorable fragmentation of the longer alkyl linker. Further highlighting the positive impact of the longer alkyl linker, reagent combination E also afforded improved specificity for peptide modification state when compared to A. The improved specificity was indicated by the decreased frequency of detection of signature ions F1, F8, and M3 in PSMs without biotin modifications (Figures 4 and S8A). This increase was further enhanced (2-fold) with the use of FAIMS and a narrower precursor isolation window (Figure 4). Improved specificity was also observed for the F7 ions, albeit at a substantially attenuated frequency of detection (Figure 4).

Figure 4.

Frequency distribution of fragment and peptide remainder ions in PSMs with or without cysteine modifications in cysteine-enriched samples with narrower isolation window and with (w/) or without (w/o) FAIMS. Statistical significance was calculated with unpaired Student’s t-tests, *p < 0.05, **p < 0.01, ***p < 0.001, and NS p > 0.05. Experiments were performed in triplicate. Data for reagent combination A is reproduced from Figure 3 for clarity. All data can be found in Table S4.

Next, we investigated whether the fragmentation patterns observed for our more minimalist reagent combinations A–E would extend to a larger labeling reagent—as many chemoproteomic studies employ relatively elaborated isotopically labeled peptide-based reagents, we were optimistic that such an investigation could help inform the performance of such reagent combinations. Analysis of samples labeled with reagent combination F (IAA 1 + N-Ac-Val-Azido-Lys-Lys(biotin)-Gly-COOH tetrapeptide reagent 8) revealed a number of characteristic fragment ions produced by the reagent 8 (proposed fragmentation is shown in Figure S9), including known and novel species. In addition to the formation of dehydrobiotin, we also detected a m/z 310.158 (F9) fragment ion, which we assigned as ImKbiotin-NH3,³⁴ as characterized previously for biotinylated peptides. Additionally, we detected a fragment ion of m/z 430.21187 (F10), which we assigned as likely corresponding to the Lys(biotin)-Gly dipeptide and fragment ion m/z 682.35926 (F11), corresponding to the tetrapeptide formed from C─N triazole bond fragmentation. Together with these fragment ions, we also identified peptide remainder ions p + 449.275 (M7), p + 433.256 (M8), p + 180.101 (M2), and p + 152.095 (M3), which match with the proposed fragmentation scheme. As a demonstration of the complexity of the fragmentation pattern observed for this larger biotin reagent, we also identified several ions that appeared diagnostic of the peptide labeling state as indicated by their presence in labeled PSMs but did not readily match with any fragmentation pattern tested, including m/z 243.08, m/z 412.20, m/z 464.26, and m/z 491.14.

Despite the complex fragmentation of the large biotin reagent, we observed high reagent-specificity of the observed fragment ions, comparable to that of the more minimalist biotin labels. Exemplifying this, F10 was detected in 99.99% of PSMs using reagent combination F and less than 1% of all scans in other reagent combinations (Table 2). The specificity of F10 for modified cysteine peptides was comparable to F8 and was improved with the use of FAIMS and a narrower precursor isolation window as expected (Figure 4). Higher mean intensity of F10 was observed at ~60% compared to ~30% for F8 (Figure S8B), possibly stemming from F10’s increased size and propensity to fragment.

Together, these ions provide further evidence of the gas-phase lability of triazole modifications, including that this fragmentation is generalizable to multiple combinations of reagents of varying linker lengths and triazole configurations, including for larger peptide-derived labeling reagents.

Improving the Coverage of CBCC Peptides Using Optimal Reagent Combinations and Labile Search.

Given the value in achieving high coverage in chemoproteomics experiments, our next step was to determine whether we could enhance the detection of PSMs and unique peptides through judicious reagent selection and implementation of labile search. We first quantified the mean PSMs and unique CBCC peptides for each reagent combination. Consistent with a model where increased formation of these diagnostic fragment ions can decrease coverage, we were pleased to observe that the simple modifications made to both reagent combinations D and E afforded a ~10% increase in PSMs and a ~9% increase in unique peptides (Figure 5) when compared with the standard reagent pair A. In contrast, fragmentation-prone combination C decreased the PSMs by ~10% and the unique peptides by ~11%. For larger peptide-based reagent 8, this decrease was even more pronounced, with only 66.0% PSMs and 61.8% peptides detected. The result supports the general dogma of the field that smaller labeling reagents are likely preferable when compared to bulkier tags.

Figure 5.

PSMs and peptides identified in CBCC with different labeling reagents. Statistical significance was calculated with unpaired Student’s t-tests with equal variance, *p < 0.05, **p < 0.01, and NS p > 0.05. Experiments were performed in triplicate. All data can be found in Table S5.

We also extended this analysis to include the fraction of peptides that were uniquely identified by one reagent combination. We were surprised to observe a relatively modest overlap between the combinations tested (Figure S10). While we acknowledge the semi-stochastic nature of data-dependent acquisition (DDA) proteomic data acquisition and the resulting problem of “missing” data may in part rationalize some of these differences, collectively we expect that these data point to the possible utility of varying CBCC labeling reagents as a strategy to improve detection of certain low abundance or otherwise tough-to-detect peptides.

With our successful identification of a panel of labile ions derived from the CBCC modifications using the labile MSFragger search options, we next asked whether such labile ion searches would also increase the coverage of CBCC peptides. Due to the high abundance of the oxonium-biotin species in all MS2 scans, we initially expected that, as has been reported for analysis of glycopeptide datasets,³³ such searches would similarly boost the coverage for CBCC. Curiously, we did not observe such an increase (Figure S11A) and instead found that labile search consistently underperformed, with only ~68% the number of PSMs and peptides compared with our established closed-search method. We suspect that this decrease in coverage observed for both reagents for the labile search may stem from the relative gas-phase stability of many of the ions detected, despite the observation of intense diagnostic ions. Partial fragmentation of the modification on multiple peptide fragment peaks can be concentrated into a single diagnostic peak, resulting in high relative intensity despite much of the modification remaining intact.

Comparison of closed vs labile ion searches revealed that nearly all cysteines identified in the labile search are shared with the closed search, with a modest ~3% unique cysteines only identified by a labile search for reagent combination A (Figure S11B). Extension of these analyses to the tetrapeptide reagent 8 revealed similar findings with closed search outperforming the labile search with ~4% unique cysteines only identified by labile search (Figure S11A,C). While these gains remain modest, this finding points to the possible utility of labile search in the analysis of more complex chemoproteomic labeling reagents, particularly for those with substituents prone to fragmentation.

DISCUSSION

In this study, we performed a detailed characterization of the gas-phase fragmentation products generated from cysteine biotinylation via click chemistry (CBCC) chemoproteomic analysis. Manual inspection of CBCC MS/MS spectra first revealed two fragment ions, the well-characterized dehydrobiotin (m/z 227.085) and a new species m/z 284.143. Using a biotin isotopologue alongside MS3 analysis, we then interrogated the nature of the m/z 284.143 fragment ion and proposed it to be the oxonium-biotin species F2, which forms as a result of fragmentation at the N(triazole)─C(alkyl) bond together with cyclization to afford the cyclic oxonium species. This finding goes against the generally held assumption that triazoles are relatively inert to gas-phase chemistry. There is some uncertainty surrounding the formal nature of the m/z 284.143 fragment ion as either cyclic oxazonium or linear carbocation. However, the general instability of primary carbocations, together with the favorability of the six-member ring formation, supports the oxonium-biotin structure as the most likely nature of the fragment ion.

We next leveraged the labile search features of the MSFragger algorithm together and extended the PTM-Shepherd module to identify additional fragments and peptide remainder ions associated with the CBCC-labeled peptides. Consistent with our proposed fragmentation of the N6─C5 bond in the clicked conjugate, we also identified peptide remainder ions that correspond to peptides retaining the triazole M2 and the corresponding azirine species M3, afforded by loss of nitrogen. An analysis of the specificity of these ions for the CBCC peptide modification state revealed a striking difference between the fragment ions and the peptide remainder ions, with all peptide remainder ions only identified in PSMs for biotinylated peptides. Overall, peptide remainder ions were identified at a lower frequency and with lower relative intensity compared to the fragment ions.

A narrower isolation window minimizes the co-isolation of modified and unmodified peptides, which, together with the use of FAIMS, successfully improved the specificity of all fragment ions, including oxonium-biotin and dehydrobiotin. The improvement in specificity was observed particularly for unenriched samples, which contain lower abundance of the biotinylated species. When compared to the oxonium-biotin species F2, several of the other peptide remainder (M2, M3) and fragment ions (F1, F3, F4, F5) showed considerably greater specificity for the peptide biotinylation state. However, given the moderate to low detection frequency of such ions, they are likely suboptimal as diagnostic species. Distinguished by its high intensity and ubiquity, we expect that with an appropriate intensity threshold, the oxonium-biotin species could serve as a diagnostic ion for CBCC peptides.

The extension of these analyses to five additional reagent pairs revealed several striking results. First, we found that the production of the oxonium-biotin species was not restricted to our IAA probe and instead was produced by both alkyl and aryl iodoacetamide probes, with apparently increased favorability of fragmentation for the aryl probe likely due to the pi conjugation stabilizing the peptide remainder ion. The ion’s generalizability points to its utility for a wide variety of clickable probes, including more advanced drug-like scaffolds. Our reagent panel also revealed that a ~10% increase in PSMs could be achieved together with increased specificity of the oxonium-biotin species by small and easily implementable modifications to the reagents themselves.

Alongside providing an improved fundamental understanding of how chemoproteomics samples behave in the gas phase, our study offers several added benefits. First, we present the low-cost synthesis of an isotopically labeled pair of biotin-azide reagents, which compares favorably to the cost and complexity of established isotopically labeled reagents, both azide-containing and those that feature cysteine-reactive electrophiles.^11,13,45,46 Our demonstration of the labile ion search features built into FragPipe should also provide a generalizable computational platform for others interested in leveraging fragmentation of chemoproteomics samples. Exemplifying the utility of such studies, gas-phase fragmentation of cysteines modified by covalent drugs, such as ibrutinib, has been leveraged to improve the identification of labeled cysteine residues.^47,48 In addition, fragmentation of sulfonyl-triazole probes has been harnessed for site-of-labeling studies.⁴⁹ When combined with custom isobaric data analysis algorithms, these modifications should also provide an avenue to improve MS2-level quantification of peptide labeling. We anticipate that realizing the full potential of labile search algorithms in chemoproteomic applications may depend on advances in these algorithms to take full advantage of partial fragmentation, such as that observed for the CBCC reagents described here.