Covalent Labeling with an α, β Unsaturated Carbonyl Scaffold for Studying Protein Structure and Interactions by Mass Spectrometry

Abstract

A new covalent labeling (CL) reagent based on an α, β-unsaturated carbonyl scaffold has been developed for studying protein structure and protein-protein interactions when coupled with mass spectrometry. We show that this new reagent scaffold can react with up to 13 different types of residues on protein surfaces, thereby providing excellent structural resolution. To illustrate the value of this reagent scaffold, it is used to identify the residues involved in the protein-protein interface that is formed upon Zn(II) binding to the protein β-2-microglobulin. The modular design of the α, β-unsaturated carbonyl scaffold allows facile variation of the functional groups, enabling labeling kinetics and selectivity to be tuned. Moreover, by introducing isotopically enriched functional groups into the reagent structure, labeling sites can be more easily identified by MS and MS/MS. Overall, this reagent scaffold should be a valuable CL reagent for protein higher order structure characterization by MS.

Graphical Abstract

A new covalent labeling reagent scaffold with broad and tunable reactivity can be combined with mass spectrometry for protein higher order structure analysis.

INTRODUCTION

Reliable and versatile methods that can analyze protein higher order structure (HOS) are essential for understanding protein activities and investigating biological machinery. Nuclear magnetic resonance (NMR) spectroscopy, X-ray crystallography (X-ray) and cryo-electron microscopy (cryo-EM) have been the primary tools used for protein HOS analysis due to the high resolution information they provide.^1–3 These techniques, however, are limited by the relatively high sample amounts and purities required in addition to the extensive analysis times that limit their throughput. In comparison, mass spectrometry (MS) is emerging as a valuable tool for protein HOS analysis because it requires less sample and has fewer purity constraints, while also providing higher throughput.^4–14

Because MS measures the mass-to-charge ratio (m/z) of ions, obtaining structural information about a protein in solution requires encoding the desired structural information into the mass of the protein. Hydrogen deuterium exchange (HDX) is a popular means of encoding such structural information for MS detection. In HDX, exchangeable backbone amide hydrogens are replaced with deuterium at rates that depend on protein backbone structure, dynamics, and hydrogen bonding.^{12–13,15–18} Back exchange and deuterium scrambling, though, can limit structural information or lead to ambiguous data. Another common MS-based method to encode structure information is chemical cross-linking, which uses bi-functional chemical reagents to bridge different residues with certain distance constraints, thereby providing insight into the structural arrangement of residues that are distant in primary sequence.^{10–11,19–21} Cross-linking usually produces complicated datasets that can be difficult to analyze by MS, and the best cross-linking reagents tend to react with a limited set of residues. Covalent labeling (CL) methods that utilize non-specific labeling reagents or amino acid-specific reagents to modify amino acid side chains are beginning to emerge as complementary methods to HDX and cross-linking approaches.^22–27 These CL methods use reagents that react with solvent-accessible side chains and are typically most useful for identifying changes in the surface topology of a protein as caused by protein binding or protein HOS changes. CL together with MS can pinpoint specific amino acids that are involved in interaction sites by comparing labeling extents of the same residue under different conditions (e.g. bound version vs. unbound version). The irreversible side chain modifications and straightforward data analyses make CL-MS an excellent method for probing interfaces in protein complexes.

Numerous CL reagents that are capable of reacting with specific amino acids have been developed and used with MS to analyze protein structure, including N-hydroxysuccinimide (NHS) derivatives, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide/glycine ethyl ester (EDC/GEE), dimethyl(2-hydroxy-5-nitrobenzyl)sulfonium bromide (HNSB), N-alkylmaleimides and many more.^22,28–34 Even so, due to the specificities of these reagents, multiple reagents are often necessary to more fully map protein structure, making such an approach more time and labor intensive. Non-specific labeling reagents, such as hydroxyl radicals and carbenes, overcome the problems with specific reagents as they are capable of modifying numerous residues simultaneously.^{25–27,35–37} Production of hydroxyl radicals or carbenes, however, requires specialized equipment, such as synchrotron sources or lasers, which cause these approaches to be less broadly accessible. Furthermore, due to the high reactivity of radicals, these methods usually generate many different modification products, making data interpretation challenging. Our group has been interested in developing and applying reagents that retain the simplicity of amino-acid specific reagents (e.g. NHS derivatives) while at the same time reacting with a broader set of residues. To this end, we have explored diethylpyrocarbonate (DEPC) as a CL reagent because it labels various nucleophilic residues such as Lys, His, Tyr, Ser and Thr simultaneously, enabling a typical structural resolution of 8-12 Å.^38–39 Using this reagent, we have successfully studied protein HOS and protein complexes.^40–44 Despite the utility of DEPC, its amino acid coverage and chemical tunability could still be improved upon, so new CL reagents that retain its ease of use are of interest.

To expand the number of residues that can be labeled and to introduce tunability and other attributes, we have designed and studied a new class of CL reagents having an α, β-unsaturated carbonyl scaffold. In this work, we describe how these new reagents can modify a broad set of nucleophilic residues on the protein surface, rendering them excellent probes for studying protein structure and interactions. When coupled with MS, we have successfully used these new reagents to investigate protein-protein interactions. We also demonstrate the ability to incorporate isotopic tags for improved confidence and efficiency in detecting modified peptides. Moreover, by varying the functional groups on the reagent, labeling kinetics and selectivity can be tuned to uncover subtle protein structural differences.

MATERIALS AND METHODS

CL reagent synthesis and characterization

The synthetic schemes and characterization details for the reagents can be found in the Supporting Information (SI) (Schemes S1–S10), and additional synthetic details can be found in previous work.⁴⁵ All of the synthesized reagents are water soluble and are strong alkylating agents, so they must be handled with care.

Materials

The proteins used in this study included human β-2-microglobulin (β2m), horse heart myoglobin (Myo), bovine carbonic anhydrase (bCA), and bovine β-lactoglobulin b (βlgb), and their sources are indicated in the SI. Other buffer materials and reagents are also described in the SI.

β2m dimer preparation

β2m was dissolved in 25 mM MOPS, 150 mM potassium acetate and 500 mM urea at pH 7.4, and then zinc suclfate was added to the protein solution at a ratio of 4:1 Zn(II):protein. The final zinc(II) and protein concentrations were 800 and 200 μM, respectively. The protein solution was then incubated at 37 °C for 1 h. Based on previous studies^46–47 and size-exclusion chromatography results (see the SI), β2m is known to form a stable dimer under these conditions.

Protein covalent labelling

Stock solutions of the CL reagents were prepared in water. Covalent labeling reactions were typically performed for 10 min at room temperature with 100 μM protein and various molar excesses of CL reagents in 50 mM pH 8.0 phosphate buffer. The molar excesses of the CL reagents were 23, 20, 20, and 10 for bCA, βlgb, β2m and Myo, respectively, unless otherwise stated. For the kinetics experiments, a 20-fold excess of each CL reagent was used. For intact protein analyses, the reaction mixtures were injected immediately after the 10 min reaction period into an HPLC to remove excess CL reagents and phosphate salts and were analyzed by ESI-MS.

HPLC separations and mass spectrometry for intact protein analysis

To quench the CL reaction of intact proteins and remove excess CL reagents and buffer salts, a Thermo Scientific Ultimate 3000 HPLC system (Thermo Scientific, Tewksbury, MA) with an OPTI-TRAP C4 reverse phase column (1 × 8 mm) was used. The protein was eluted using an acetonitrile gradient that increases from 1 to 99% over 12 min at a flow rate of 0.2 mL/min. The labeled protein was collected for proteolytic digestion or intact protein MS characterization. Size-exclusion chromatography (SEC) was used to purify the β2m dimer after formation, and the details of this separation can be found in the SI.

Mass spectral analyses of the HPLC separated intact protein samples from the covalent labeling experiments were acquired on a Bruker AmaZon (Billerica, MA) quadrupole ion trap mass spectrometer equipped with an electrospray ionization source. The electrospray needle voltage was kept at 4 kV, and the capillary temperature was set to 250 °C.

Proteolytic digestion

The labeled protein samples were first buffer-exchanged using 10k NMWL Amicon Ultra centrifugal filters (Millipore, Burlington, MA) with 100 mM triethylamine acetate (pH 8.0) to adjust the pH for proteolytic digestion. For β2m and Myo, the samples were incubated with 10% (v/v) acetonitrile at 50 °C for 45 min to denature the protein. For bCA, the samples were incubated with 8 M urea to denature the protein, and this urea concentration was reduced to 1 M before enzymatic digestion. To reduce the disulfide bonds in β2m, TCEP in water was added at a protein:TCEP molar ratio of 1:20, and the sample was incubated at room temperature for 10 min. To alkylate the reduced cysteines, iodoacetamide in water was added at a protein:iodoacetamide molar ratio of 1:80, and the sample was incubated in the dark at room temperature for 30 min. The denatured, reduced, and alkylated protein samples were then digested with trypsin or chymotrypsin at an enzyme: substrate ratio of 1:10. After 4 h of digestion at 37 °C, the enzyme was separated from the mixture by centrifugation using a 10k NMWL Microcon filter (Millipore, Burlington, MA). The filtrate was then analyzed by LC-MS and LCMS/MS.

HPLC separations and mass spectrometry for digested proteins

To analyze the digests from the CL experiments, a Thermo Scientific EASY-nLC 1000 liquid chromatography system (Thermo Scientific, Tewksbury, MA) with an Acclaim PepMap RSLC C18 reverse phase column (75 μM × 15 cm, 2 μm particle size) from Thermo Scientific (Tewksbury, MA) was used. To achieve efficient separation of the proteolytic peptides, a gradient was used in which the %B (0.1% formic acid in acetonitrile) was increased from 0% to 40% over 45 min. The column was then flushed by increasing to 95% B over 15 min. The column was then cleaned at 95% B for another 20 min. A flow rate of 300 nL/min was used throughout the run.

LC-MS and LC-MS/MS analyses of protein proteolytic fragments were conducted on a Thermo Orbitrap Fusion Tribrid (Tewksbury, MA) mass spectrometer. The electrospray ionization source was typically operated at a needle voltage of 2100 V, and the ion transfer tube temperature was set to 300 °C. Tandem mass spectra were collected using CID with a normalized collision energy of 35%. Due to the large number of detectable peaks, an exclusion limit of 60 s was applied after five spectra had been collected for any given peak. The resolution of the Orbitrap was set to 60000. Peptides were analyzed and identified using the Thermo Proteome Discoverer 2.2 software or via manual analysis. Search details are provided in the SI.

Solvent accessibility calculation

The online software GetArea was used to calculate the solvent accessible surface area (SASA) of residues on β2m (pdb: 1JNJ), Myo (pdb: 1DWR), and bCA (pdb: 1V9E). A probe radius of 1.4 Å was used. Residues having SASA above 20% were considered as solvent accessible.

RESULTS AND DISCUSSION

The α, β-unsaturated carbonyl (ABUC) scaffold that was investigated as a CL reagent was designed and synthesized as a Michael acceptor to readily react with nucleophilic residues on a protein’s surface (Figure 1). Nucleophilic groups in residue side chains are particularly reactive with the CL reagent when a good leaving group is present at the R₂ position. For example, a NEt₃⁺ functional group at the R₂ position leads to a triethylamine leaving group upon Michael addition at the allyl position. When R₁ is a methyl group, the resulting mass shift is 98 Da as confirmed via LC-MS measurements of intact proteins (Figure S1). It should be noted that because the labeling reaction happens via nucleophilic attack by residues in proteins, buffer systems that have components with weak nucleophiles should be chosen. Buffers that contain Tris, ammonium groups, or other nucleophilic functional groups should be avoided.

Figure 1.

Nucleophilic residues (i.e. -Nu:) on the proteins surface can react with the CL reagent to generate a covalent adduct that can then be identified by MS and MS/MS.

To identify the types of nucleophilic residues that are reactive with the ABUC molecules, four different model proteins, namely bovine carbonic anhydrase (bCA), β-2-microglobulin (β2m), myoglobin (Myo) and β-lactoglobulin b (βLGb) were reacted with 20 equivalents of the reagent bearing -CH₃ as R₁ and -NEt₃⁺ as R₂ (Figure 1). After proteolytic digestion and LC-MS/MS analyses of the labeled proteins, we detect labeling on relatively strong nucleophilic residues such as Cys, His, and Lys, as expected. For instance, solvent exposed Lys42 in Myo is readily labeled as indicated by CID of the peptide TGHPETLEKFDKF (Figure S2a). Example tandem mass spectra of labeled Cys and His peptides from the studied proteins are also found in the SI (Figure S2b–c). Surprisingly, weak nucleophilic residues such as Ser, Thr, Tyr, Asp, Glu, Arg, Trp, Asn, Gln and Met are also found to be modified by the ABUC reagent, as exemplified by MS/MS of the HPGDFGADAQGAMTK peptide from Myo that shows labeling at Asp123 (Figure S2d). Additional examples of tandem mass spectra of peptides with labels on other weak nucleophiles can also be found in the SI (Figure S2e–j). We speculate that good reactivity with such a wide range of residues is due to both the good leaving group on this reagent and the stability of the resulting product, which is not prone to hydrolysis after labeling. This resistance to hydrolysis is in contrast to DEPC labeling, whose Ser, Thr, and Tyr products have been shown to be subject to hydrolysis after labeling.⁴⁸ Overall, the 13 different types of amino acid residues that can be labeled by this new reagent account for about 55% of the sequence of the average protein, indicating that it should provide excellent structural resolution. Moreover, among these labeled residues, Tyr, Arg and Trp are frequent ‘hot spots’ at protein-protein interfaces,⁴⁹ so that this new reagent should be an excellent probe of protein-protein interfaces.

Because probing protein-protein interactions via CL relies on monitoring the solvent accessibilities of the protein residues that are involved at the interfaces, we evaluated whether those residues labeled by the ABUC reagent are primarily solvent accessible ones. Based on previous CL work,^23,38,44 proteins are likely to maintain their HOS when the total labeling extent is below 1.4 labels/protein. We decided to carry out labeling reactions so that the proteins have 1.1 – 1.2 labels on average to minimize protein structural perturbations. At this level of labeling, any given protein molecule will be reacted only at one site, thereby preventing any subsequent labeling that would necessarily occur on a structurally-perturbed protein, but the ensemble of protein molecules will have modifications at many different residues. Upon reaction of bCA, β2m, and Myo with the reagent shown in Figure 1, we find that the vast majority of the labeled residues are on the surface of the protein (Figure 2). Space-filling models show that the residues labeled in bCA, β2m and Myo cover much of the surface of the protein (Figure 2a–c). In addition, greater than 90% of the labeled residues have SASA values above 20%, which typically qualifies them as solvent accessible, and over 60% of the labeled residues are highly exposed with SASA values of greater than 50% (Figure 2d). Of the 10 residues having SASA values < 20%, the vast majority are Lys or His that have high reactivities or are residues with SASA values within a few percent of 20%. A more quantitative relationship between reactivity and residue SASA is not possible with the current data set because of peptide-dependent differences in ionization efficiencies between modified and unmodified peptides. Future work will assess this relationship in greater detail.

Figure 2.

Space-filling models for (a) bovine carbonic anhydrase (bCA), (b) β-2-microglobulin (β2m) and (c) Myoglobin (Myo). The residues labeled on each protein are indicated in orange. (d) Bar graph showing that most residues that are labeled by the reagent have high solvent accessible surface areas (SASA).

With good evidence that this reagent reacts with solvent exposed residues, we examined its ability to identify the residues present at the interface of a protein dimer. We chose the Zn(II)-induced dimer of the protein β2m as a testbed for this reagent. β2m forms amyloid fibrils in the presence of Cu(II), and the structure of the pre-amyloid dimer formed upon Cu(II) binding is known.⁴⁰ In contrast, Zn(II) enables the formation of a very stable dimer, but it does not induce amyloid fibril formation.^46–47 A comparison of the modification levels for the apo β2m monomer and SEC-purified Zn(II)-induced dimer (Figure S3) after reaction with the reagent described above shows that ten residues are found to undergo statistically significant differences in modification extents (Figure 3a). Four of these residues undergo increases in reactivity and six of them undergo decreases in reactivity. The sites of these changes can be mapped onto the structure of the β2m monomer with decreases shown in blue and increases shown in red (Figure 3b). His51 and a cluster of residues near the C-terminal end of β2m undergo decreases in their extents of modification. His51 has previously been identified as the Zn(II) binding site in the protein, explaining this residue’s decreased modification extent when Zn(II) is present.^47,50 The other residues, including S11/R12, H13, K19, K41, and K75, are clustered near the C-terminus or ‘tail’ of the protein, suggesting that the Zn(II)-induced dimer is formed via a ‘tail-to-tail’ interaction of two monomers. This interface contrasts sharply with the Cu(II)-induced β2m dimer, which forms a side-to-side interaction (Figure S4).⁴⁰ This distinct interface in the Zn(II)-induced dimer perhaps explains why Zn(II) does not allow β2m amyloid formation to occur, whereas Cu(II) does.

Figure 3.

(a) Residue modification levels after reactions of the reagent in Figure 1 with the β2m monomer (green) and Zn(II)-induced β2m dimer (orange). The bar with an arrow indicates a statistically significant change at a 95% confidence interval. (b) Sites of modification level changes mapped onto the monomeric structure of β2m. The labeling decreases (in blue) are clustered around the C-terminal ‘tail’ of β2m, with the exception of His51, which is the known binding site of Zn(II).^47,50 The increases in labeling (in red) are also indicated.

Another feature of the ABUC reagent is that the R₁ and R₂ groups can be varied to introduce new reactivity or new capability for CL. For instance, the CL reaction kinetics can be tuned by varying the functional groups at R₁ and R₂ (Figure 4a). With a more electron withdrawing group as R₁ and a better leaving group as R₂, the reaction can be accelerated, whereas the reaction is slowed when R₁ is an electron donating group and R₂ is a poor leaving group. As an example, Reagent 2, which has a very good leaving group, reacts to yield 1.2 labels per protein during a 4 min reaction, whereas Reagent 6, which has a relatively weak leaving group in the R₂ position, requires a 32 h reaction to achieve the same labeling extent (Figure 4b). Reagent 1, which has both a good leaving group at R₂ and a good electron withdrawing group at R₁, achieves a labeling extent of 1.2 labels/protein after only a 5 sec reaction (Figure 4c).

Figure 4.

(a) Chemical structures of different CL reagents with functional group variations. (b) Covalent labeling reaction kinetics of 100 μM Myoglobin with 2 mM different CL reagents respectively in 50 mM pH 8.0 phosphate buffer. (c) Zoom-in of the 0-0.5 h region of plot (b). Error bars represent the standard deviations from three experimental replicates.

Tuning the reaction kinetics can allow for the tuning of the labeling selectivity to some extent. Comparing the Myo labeling results from reactions with Reagent 2 and Reagent 6, we find differences in the reactivity at certain residues. As a particularly compelling example, we observed that His113 reacts quite distinctly with these two reagents. The ISDAIIHVL peptide from Myo that contains His113 gives rise to two isomeric products due to reactions on the δN and εN of His113, which can be separated by LC and confirmed by tandem MS (Figure 5 and Figure S5). Interestingly, Reagent 2 reacts about equally with both nitrogens on the side chain, whereas the Reagent 6 reacts three times more extensively with only one of the nitrogens on this residue (Figure 5a). An analysis of the micro-environment surrounding His113 in Myo reveals that one of the nitrogens is more solvent exposed than the other due to an adjacent Arg side chain (Figure 5b). Such reaction selectivity might arise from the interplay between the dynamics of the protein around His113 and the relatively slow reactivity of Reagent 6. Perhaps Reagent 6’s low reactivity causes it to primarily label the protein conformational state that has the exposed nitrogen on His113, whereas Reagent 2 reacts rapidly enough to label both states, leading to almost equivalent reactivity at both nitrogens on His113. Analogous atom-specific reactivity on the side chains of Phe residues was recently observed in hydroxyl radical labeling of wide-type and ΔN6 β2m.⁵¹

Figure 5.

(a) Extracted ion chromatograms (XIC) of both unmodified and modified ISDAIIHVL peptides containing His113 in myoglobin. Labeling reaction with Reagent 2 for 4 min (top two chromatograms). Labeling reaction with Reagent 6 for 32 h (bottom two chromatograms). Peak areas from XIC were used to calculate the ratio between the two modified peptides in the digests mixture. (b) A cartoon representation of the micro-environment of His113 in myoglobin in PDB ID: 1DWR, where His113 interacts with neaby Arg31, which causes the partial burial of one of the nitrogens on the His113 side chain.

The R₁ group in the reagent can also be isotopically enriched with deuterium instead of hydrogen on the ethyl group (Figure 6a), and if both the normal and deuterated form are simultaneously reacted with proteins under identical conditions, the labeled peptides after digestion can be readily identified by the characteristic multiplet of peaks (Δm/z = 1.67 for a +3 ion; Δm/z = 2.5 for a +2 ion; or Δm/z = 5.0 for a +1 ion) with equal ion abundances (Figure 6b). Moreover, simultaneous tandem MS of both covalently labeled peptides (i.e. the normal and deuterium enriched peptide) can generate tandem mass spectra that make it easier to identify the covalently labeled residue in the peptide.⁵² For example, upon MS/MS of the labeled peptide IARRHPYFL, all the b ions appear as doublets, and all the y ions appear as singlets, indicating that the labeling site is on the N-terminus (Figure 7). Given the fact that both the normal and deuterated reagents are chemically identical, we envisage using these set of reagents to perform relative labeling quantification, which would facilitate CL analyses of proteins under two or more conditions.

Figure 6.

(a) Chemical structures of the normal and deuterated forms of the indicated CL reagent. (b) Mass spectra of the labeled IARRHPYFL peptide when it is simultaneously reacted with both reagents from (a). A characteristic 5 mass unit difference (Δm/z = 1.67 for a +3 ion or Δm/z = 2.50 for a +2 ion) allows the labeled peptide to be readily identified.

Figure 7.

Simultaneous tandem mass spectrometry of the normal and deuterated forms of the covalently labeled IARRHPYFL peptide, where the N-terminal modification is easily identified from all the doublet b ions and singlet y ions. Reagents 8 and 9 (see structures in Figure 6) were used, and the mass shifts that occur with these reagents are 112 Da and 117 Da, respectively.

CONCLUSIONS

We have developed and applied a new CL reagent based on an α, β unsaturated carbonyl (ABUC) scaffold that can react with 13 different types of residues on protein surfaces. These residues range from very nucleophilic residues, such as Cys, His, and Lys, to much less nucleophilic residues such as Asp, Glu, Asn, and Gln. This broad reactivity with surface exposed residues enables probing of protein-protein interactions as demonstrated for the Zn(II)-induced β2m dimer. The new reagent can also be easily functionalized allowing for outstanding structural tunability. Labeling kinetics and selectivity can be controlled in a predictable way by varying functional groups on the reagent. In addition, by incorporating isotopes into the reagent structure, labeling sites during MS and MS/MS analyses can be identified in straightforward way. One could envision modifying the ABUC reagent to install functional groups that enable selective enrichment of modified proteins or peptides from protein digests, extending the applicability of CL-MS to more complex mixtures such as cells. Moreover, because the products of the reaction with the ABUC reagents still contain an α, β unsaturated carbonyl group, these reagents could also be considered for cross-linking applications; although no evidence of such chemistry was found in this work. Future work will investigate many of these potential applications and will further evaluate the reaction kinetics of the ABUC reagents with different amino acid residues to more fully understand the reaction scope of these reagents. Overall, this new CL reagent can be a potentially powerful tool for studying protein HOS and protein interactions because of its modularity, broad reactivity, and ease of use.