A Workflow for Validating Specific Amino Acid Footprinting Reagents for Protein HOS Elucidation

Abstract

Protein footprinting mass spectrometry probes protein higher order structure (HOS) and dynamics by labeling amino acid side-chains or backbone amides as a function of solvent accessibility. One category of footprinting uses residue-specific, irreversible covalent modifications, affording flexibility of sample processing for bottom-up analysis. Although several specific amino acid footprinting technologies are becoming established in structural proteomics, there remains a need to assess fundamental properties of new reagents before their application. Often, footprinting reagents are applied to complex or novel protein systems soon after their discovery, and sometimes without a thorough investigation of potential downsides of the reagent. In this work, we assemble and test a validation workflow that utilizes cyclic peptides and a model protein to characterize benzoyl fluoride (BF), a recently published, next-generation nucleophile footprinter. The workflow includes the characterization of potential side-chain reactive groups, reaction “quench” efficacies, reagent considerations and caveats (e.g., buffer pH), residue-specific kinetics compared to established reagents, and protein-wide characterization of modification sites with considerations for proteolysis. The proposed workflow serves as a starting point for improved footprinting reagent discovery, validation, and introduction, aspects of which we recommend before applying to unknown protein systems.

Graphical Abstract

INTRODUCTION

Protein footprinting mass spectrometry (MS) encompasses a variety of approaches wherein chemical reagents are employed to characterize changes in the solvent-accessible surface area (SASA) of a protein.^{1, 2} Their common feature is their sensitivity to changes in protein higher order structure (HOS) accompanying protein-protein interactions³, protein-ligand binding⁴, folding and unfolding⁵, mutations and truncations⁶, protein storage and age⁷, and remote conformational changes⁸. Footprinting reagents covalently modify—and thereby increase the mass—of solvent-accessible regions of a protein in physiologically relevant media, making the changes amenable to MS detection. With modern advances in liquid chromatography and tandem MS (LC-MS/MS) and the development of a variety of reagents, footprinting is increasingly robust and can be sensitive to even small changes in protein SASA¹.

Irreversible specific amino acid footprinting imparts “permanent” modifications of amino acid side-chain functional groups, whereas reversible covalent labeling MS (CL-MS), such as the widely applied hydrogen-deuterium exchange (HDX), necessitates special considerations to reduce back exchange and to inform at the residue level.⁹ The relative permanence of irreversible modifications and their ready implementation in any protein laboratory with access to MS makes this approach effective for digestion-resistant protein systems and allows for single residue resolution by using bottom-up analyses. Broad-based irreversible reagents (e.g., hydroxyl radicals used in fast photochemical oxidation of proteins (FPOP)¹⁰ and other platforms, carbene diradicals¹¹, and trifluoromethyl radicals¹² produced in source¹³ or via laser irradiation, plasma generation¹⁴, or synchrotron irradiation^{15, 16} are highly reactive on the millisecond and less time scales. Specific footprinters afford a more targeted approach by taking advantage of the varied functional groups within amino acid side-chains. This class of footprinters range from reagents targeting a single functional group, such as N-ethylmaleimide (NEM)—a highly specific cysteine footprinter^17-19,20—to reagents targeting one or more, usually nucleophilic, functional groups^21-27 (e.g., glycine ethyl ester (GEE)^{28, 29}, diethyl pyrocarbonate (DEPC), and recently introduced benzoyl fluoride (BF)³⁰). Footprinting one or more functional groups may be preferable when the biological question refers to a specific amino acid type. Reagent development for specific amino acid footprinting is an ongoing and constantly evolving process, requiring careful validation and an understanding of limitations. For example, although GEE footprinting is specific for two amino acids (Glu and Asp), it requires high molar equivalents of reagents to protein (~20,000:1); the label once installed is prone to acid hydrolysis. Footprinting using benzhydrazide (BHD) as an alternative to GEE overcomes some of these problems³¹. Similarly, we recently demonstrated 10x increased reactivity for the weakly nucleophilic OH group on Tyr using BF versus DEPC³⁰. Although this improvement in reactivity is promising, we should not overlook an extensive body of work about DEPC footprinting strategies, their advantages, and their limitations^21-27. Furthermore, complete evaluation of every footprinter may not be realistic prior to moving it forward for widespread adoption. Thus, we propose here a short list of evaluations to alert the community to reagent properties and concerns. Our proposed workflow provides a start for enhanced footprinting reagent discovery, validation, and introduction. We acknowledge that the list is not exhaustive for all investigations with potentially novel reagents. Although these experiments were performed by hand, they are apt candidates for automated specific amino acid footprinting, recently shown to expedite data collection while improving precision.³²

MATERIALS AND METHODS

Materials.

Unless otherwise indicated, all reagents were sourced from Millipore Sigma (St. Louis, MO) and used without further purification. Cyclic peptides cyclo(Arg-Gly-Asp-D-Tyr-Glu), cyclo(Arg-Ala-Asp-D-Phe-Lys), cyclo(Arg-Ala-Asp-D-Phe-Cys), cyclo(Arg-Ala-Asp-D-Tyr-Lys), cyclo(Arg-Ala-Asp-D-Tyr-Cys), and cyclo(Gly-Arg-Gly-Asp-D-Ser-Pro) (or c(RGDYE), c(RADFK), c(RADFC), c(RADYK), c(RADYC), and c(GRGDSP) respectively) (Peptides International, Louisville, KY) were dissolved in water and diluted in a phosphate buffered saline solution (1x PBS, pH 7.4 unless otherwise indicated) to a concentration of 10 μM. Bovine serum albumin (BSA) was dissolved and diluted in 1x PBS (pH 7.4) to a concentration of 100 μM. Benzoyl fluoride (TCI Chemicals, Tokyo, Japan) and diethyl pyrocarbonate were both diluted in acetonitrile.

Benzoylation and DEPC Modification (Peptides).

For initial characterization of residue-specific modification, cyclic peptides with 25 molar equivalents of BF (organic content ≤ 5% (v/v)) were incubated for 5 min at 25 °C with 200 rpm agitation, and the reaction quenched by adding 100 molar equivalents (with respect to the concentration of BF) of Tris. To evaluate the effectiveness of quench, 40 and 400 molar equivalents (with respect to the labeling reagent) of urea buffer, Tris buffer, cysteine buffer, and imidazole were added in separate experiments to the buffered cyclic peptide solutions; subsequently, 25 molar equivalents (with respect to the peptide) of BF or DEPC were added to the solutions and incubated at 37 °C with 200 rpm agitation for 1 h. The BF-modifications of cyclopeptides for kinetic evaluation were completed at various BF concentrations upon incubation for 2 min at 37 °C with 200 rpm agitation. Cysteine quench (~50 equivalents in excess of the highest BF concentration) was used to stop the reaction, and the samples were flash frozen for later LC-MS analysis. DEPC-modification of cyclic peptides for kinetic evaluation was completed at various DEPC concentrations and incubated for 2 min at 37 °C with 200 rpm agitation. Aqueous imidazole (Oakwood Chemical, Estill, SC) (~50 equivalents in excess of the highest DEPC concentration) was used to quench the reaction, and the samples were flash frozen for later LC-MS analysis.

Quenching with Zeba Spin Desalting Columns.

Zeba Spin Desalting Columns (7 k MWCO, Thermo Fisher, Waltham, MA) were prepared with 1x PBS in water/ACN (95:5 v/v). For a control evaluation, myoglobin was incubated in 10 molar equivalents of excess of DEPC and BF for approximately 1 minute at 25 °C with 200 rpm agitation prior to zeba desalting. For evaluation of the physical quench, samples were prepared with the same amount of DEPC and BF in 1x PBS in water/ACN (95:5 v/v). The footprinting reagents were first centrifuged with the Zeba^™ Spin Desalting Columns, and then incubated with 100 μM of myoglobin for approximately 1 minute at 25 °C with 200 rpm agitation.

LC-MS and LC-MS/MS Analyses (peptides).

For LC-MS analysis, the modified cyclic peptides were trapped, desalted, and eluted from a C18 trap (Opti-guard 1 mm C18 guard column, Millipore Sigma, St. Louis, MO) and analyzed by using a maXis 4G Q-ToF (Bruker Daltonics, Billerica, MA). LC-MS cyclopeptide data were compiled and analyzed by using Bruker Data Analysis 4.2.

For LC-MS/MS analyses of the modified cyclic peptides, they were trapped and desalted on a custom packed C18 trap column on a Dionex Ultimate 3000 (Thermo Fisher Scientific, Waltham, MA) and eluted into a Q Exactive Plus Orbitrap MS (Thermo Scientific, Waltham, MA) via nano-ESI, as described previously,¹² for MS/MS analyses of the cyclic peptides. Product-ion mass spectra were generated using XCaliber Qual Software (Thermo Fisher, Waltham, MA), and fragments were manually identified with < 20 ppm mass error.

Benzoylation and Digestion of BSA.

For reactivity validation of all 20 amino acids, BSA with 500 and 1000 molar equivalents of BF (organic content ≤ 5% (v/v)) were incubated for 10 min at 37 °C with 200 rpm agitation. Then 1 μL of formic acid was added to the samples, and the samples incubated at 25 °C with 0 rpm agitation for 15 min. To quench the reaction, acetone precipitation was performed by adding cold acetone (4x by volume) to the samples and then incubating them at −80 °C for 30 minutes. The samples were centrifuged at 12000 RCF for 10 min at 4 °C, and the supernatant was removed. The pellet was redissolved by addition at 10 μL of 8 M urea and 20 μL of 0.2% ProteaseMAX (Promega, Madison, WI). Addition of 1 μL of 0.5 M dithiothreitol and incubation for 20 min at 56 °C with 300 rpm agitation were used to reduce disulfide bonds, and free cysteines were capped with 2.7 μL of 0.55 M iodoacetamide for 15 min at room temperature in the dark. Samples were then diluted with 50 mM ammonium bicarbonate, and 1 μL of 1% ProteaseMAX was added. A 1:20 (w/w) ratio of sequencing grade Glu-C (Promega, Madison, WI) to protein was incubated at 37 °C for 4 h. The digestion was quenched with 1 μL of formic acid, and samples were flash frozen and stored at −80 °C for later LC-MS/MS analysis.

LC-MS/MS Analyses (BSA).

For LC-MS/MS analyses of BSA, peptides were trapped and desalted on a custom packed C18 trap desalting column on an Dionex Ultimate 3000 (Thermo Fisher Scientific, Waltham, MA) and eluted into a Q Exactive Plus Orbitrap MS (Thermo Fisher Scientific, Waltham, MA) via nano-ESI, as described earlier¹². LC-MS/MS analyses of modified peptides were conducted by using Byos^® (Protein Metrics, Cupertino, CA). Briefly, for both unmodified and modified BSA peptides, the search was conducted for fully specific C-terminal digestion at D and E residues and with three allowed missed cleavages. Precursor and fragment mass tolerances were fixed to ≤ 6 ppm and ≤ 20 ppm, respectively. A total of six modifications were allowed for each peptide, including carbamidomethyl modification of cysteine (from the cysteine capping) and benzoylation of all 20 amino acids. Each modified residue was verified manually to confirm the modification of specific amino acids. Those modifications without sufficient fragments to localize benzoylation to a single amino acid were excluded; many modifications were validated for multiply modified peptides.

RESULTS AND DISCUSSION

To characterize BF as a protein footprinter, we designed a series of experiments to probe various aspects of BF reactivity with amino acid side-chains.

(1) Identify the reagent’s amino acid side-chain specificity. The reagent’s reactivity as an organic chemical should adhere to known functional group specificity.

(2) Establish an appropriate quench for the modification. The use of a chemical or physical quench is critical for footprinting samples with precision and repeatability, especially for the most common footprinting design of two-state systems (e.g., comparing bound vs. unbound, mutant vs. wild type, etc.).

(3) Determine reagent-specific limitations that affect the reagent’s use. We found that it is important to determine potential challenges and downfalls early in the development process, as this increases the likelihood that other end users will be able to use the reagent in the future with minimal adjustment. For BF, we evaluated pH-dependence of nucleophilicity (because protonation affects side-chain nucleophilicity) by using cyclic peptides.

(4) Characterize the kinetics of BF modification of cyclic peptides with reference to an accepted reagent (in this case, DEPC).

(5) Survey potential BF-modifiable amino acids of a model protein for applications in footprinting-MS using bottom-up proteomics.

Single modifications on nucleophile-containing cyclic peptides

To probe the relative reactivities of BF with different amino acid side-chains, we selected cyclic peptides with one nucleophilic functional group identified to undergo footprinting. Their lack of termini limits possible reactive sites and allows for careful selection of sites of interest. These cyclic peptides, c(RGDYE), c(RADFK), and c(RADFC), undergo a single modification (i.e., gave a mass shift of 104.0262 u with < 5 ppm error, presumably by modifying OH, NH₂, SH, respectively) (Figure 1a-c, respectively). As expected, the relative abundances of the singly modified species in the mass spectra directly correlate with side-chain nucleophilicity (i.e., Cys > Lys > Tyr) ^{33, 34}.

Figure 1.

Mass spectra showing both unmodified and singly labeled (+104 amu) cyclic peptides, c(RGDYE), c(RADFK), and c(RADFC) (top of A-C, respectively). The highlighted area indicates the m/z selection window submitted to MS/MS to confirm nucleophile-specific benzoylation. MS/MS spectra for singly charged, modified, cyclopeptides, c(RGDYE), c(RADFK), and c(RADFC) (bottom of A-C, respectively), validating modification of the nucleophilic residue. Labelled peaks (red), correspond to key b-ions produced from fragmentation via ring-opening illustrating the benzoylation site. The fragmentation scheme shows the identity of the product ions (red). For full interpretation of other MS/MS product ions see Figure S2.

To confirm the benzoylation site, we analyzed the products by MS/MS. Cyclic peptide fragmentation requires ring opening and subsequent production of b-ions³⁵. Briefly, ions are labeled with a set of subscripts—for example, “bXYZ”—such that y and z are the single letter amino acid codes between which the peptide bond is initially broken (with new N- and C-termini formed, respectively), and x refers to the amino acid cleavage site number of the linear peptide product ion (all fragmentation patterns with the appropriate designations for cyclic peptides can be seen in Figure S1). From the MS/MS (Figure 1), we found that fragment ions are consistent with modification of only the nucleophilic site and no other sites (see Figure S2 for further identification of MS/MS product ions). Specifically, we observe neutral losses that reveal modification of Lys, Cys, and Tyr (for example, in Figure 1b, b4RK originates from a ring opening adjacent to the N-terminus of the nucleophilic amino acid residue and subsequent C-terminal loss of Lys-BF (i.e., Lys with a mass shift of +104.0262 u)). EICs for the modified species were monomodal, providing more evidence of a single population of modified cyclopeptide (Figure S3). We have recently shown that BF labeled tryptic and chymotryptic peptides undergo significant shifts in LC elution time due to increases hydrophobicity, and this phenomenon is recapitulated with these cyclopeptides.³⁰ Cyclopeptides with two adjacent nucleophilic residues, c(RADYC) and c(RADYK), yielded two peaks of m/z values corresponding to single and double additions of BF. MS/MS analysis of these doubly benzoylated peptides confirm Tyr, Lys, and Cys modification (Figure S4). The EIC of the doubly modified species is also consistent with a single population (Figure S3).

We also subjected the cyclic peptide c(GRGDSP) to incubation with BF and did not observe any modification. This finding is curious considering that we previously demonstrated that BF can modify serine on a linear peptide and on a protein, streptavidin, submitted to bottom-up MS³⁰. These findings, however, may indicate a limitation to using cyclic peptides (i.e., their inability to provide the microenvironment necessary to support weakly nucleophilic Ser/Thr labeling, as previously observed for DEPC reactivity^{36, 37}).

Role of pH and charging on BF-modification of Lys

While the reactivity of all reagents may not be sensitive to pH, the protonation state of a primary amine affects its nucleophilicity. At lower pH, BF-reactive Lys will be more protonated (less nucleophilic) and less reactive, which could have implications for Lys modification of a protein. Therefore, we labeled c(RADFK) with BF over a pH range (pH 6–8) (Figure 2A). Indeed, BF modification of c(RADFK) decreases from ~60% at pH 8 to ~5% at pH 6, indicating that lysine modification depends on pH. There may be potential applications of this reagent where pH changes do occur³⁸ as in endocytosis or on cancer cells; while this dependency does not preclude the application of BF modification to such systems, these results suggest caution in the interpretation of BF-labelling extents of Lys from such systems.

Figure 2.

pH-dependent (A) and BF-dependent (B, C) modification of c(RADFK) (C). BF modification extent of c(RADFK) shown as a function of buffer pH (A) indicates that BF modification is pH-dependent. Across a pH range of 6.0 - 8.0, Lys modification varies from 5 - 55%. Data in B and C show relative abundances of charge distribution across states, indicating that BF modification affects Lys proton affinity and an inverse correlation between BF modification and the 2+ charge state. All values taken from MS1 EICs.

BF-modification of Lys also affects the charge-state distribution in the ESI mass spectra of peptides and proteins. Benzoylation of lysine forms an amide that has a considerably lower proton affinity than an unmodified lysine. Despite that the preferred charge state of unmodified c(RADFK) is 2+, we observed that the favored charge state of BF-modified c(RADFK) is 1+ (Figure 2B, 2C).²⁷ Consequently, for protein mass spectra, BF-modified species will likely present at lower charge states. For characterization, a deconvoluted mass spectrum is preferred over working with one or a few charge states in intact protein analysis, and digestion products may have different charge states for bottom-up, potentially affecting MS/MS fragmentation efficiencies. Overall, while not applicable to all reagents, when the reagent is reactive with a charge carrying functional group (e.g. Lys, Arg, His, Asp, or Glu), we recommend characterization of reagent reactivity as a function of pH and noting the product-ion charge states to inform future applications. This step should encompass not just pH but any reagent-specific considerations with respect to the chemistry of the proposed reagent and relevant to a protein sample matrix.

Quench for BF and DEPC footprinting

We next evaluated the quench efficacy of four different chemical reagents and one physical quench. To test the efficacy of chemical quenching, various quench reagents at differing concentrations were added to the cyclopeptide sample prior to adding the labeling reagent; thus, the efficacy of the quench was indicated by a lack of cyclopeptide modification. We chose the chemical quench reagents based upon their nucleophilicity: urea is a carbamide and serves as a control; tris buffer contains four nucleophilic functional groups (i.e., three hydroxyl groups and a primary amine); cysteine contains a thiol; and imidazole is the recommended quencher for DEPC footprinting.²¹ Both low and high equivalents of urea buffer are ineffective at preventing BF and DEPC modification for each cyclic peptide, and the results indicate the modification extents expected for these labeling conditions (Figure 3A). Low equivalents of Tris buffer is also a poor quench for both BF and DEPC; however, the higher equivalents of Tris buffer does show some relative quenching capability for each cyclic peptide in comparison to the low Tris equivalents. Low and high equivalents of imidazole are more efficacious quenches for DEPC labeling, consistent with it being the recommended quench. Imidazole also appears to be significantly more effective at preventing BF-modification on the Lys-containing cyclic peptide, but not for the Tyr- and Cys-containing cyclic peptides. Lastly, with low and high equivalents of cysteine as the quench, almost no modification was detected on the cyclopeptides for BF and DEPC, proving to be the most effective quench for both reagents. Zhou and Vachet²³ have shown that free cysteine residues can cause label scrambling for DEPC. Assuming a similar behavior is possible with BF, using a cysteine buffer as a quench could potentially cause reversibility of some modifications on the protein, causing lower overall labeling. Furthermore, the quenching ability of Tris observed here has strong implications for the usage of Tris buffer in BF labeling experiments. Using Tris as a buffer system will impact the reactivity and amount of labeling on a peptide or protein. We recommend using a different buffer for labeling with BF.

Figure 3.

(A) BF (top) and DEPC (bottom) modification extent of cyclopeptides c(RADFC) (gold), c(RADFK) (green), and c(RGDYE) (purple) after initial incubation of the cyclopeptides in high (dark) and low (light) equivalents of quenching reagents urea, tris, imidazole, and aqueous cysteine (left-to-right, respectively) before adding the reagent. (B) BF (top) and DEPC (bottom) modification observed for Mb from the flowthrough of a zeba spin ultracentrifugation column (black); BF and DEPC was effectively retained on the column and thus Mb in the collection vial did not undergo BF and DEPC labelling. A control wherein the Mb was incubated with the reagents prior to the physical quench shows extensive BF and DEPC modification (orange).

To combat potential reversibility of a cysteine quench, we evaluated a physical, reagent-free method to quench the reaction. Zeba spin (7 kDa MWCO) desalting columns contain a resin that should rapidly remove small molecule footprinters from the reaction. We found that holo myoglobin (Mb) submitted to BF and DEPC incubation prior to Zeba spin desalting underwent high amounts of labeling (Figure 3B, orange). To test whether the resin retained footprinting reagent, effectively removing it from the reaction mixture (i.e., quenching the reaction), we submitted buffered solutions containing BF and DEPC to Zeba spin desalting and collected the desalted solution in a vial containing unmodified myoglobin. Myoglobin did not undergo any labeling by BF or DEPC, indicating the efficacy of this physical quench method for removing the reagent from solution; the flowthrough from the zeba spin column did not contain enough reagent to measurably modify the protein in the collection vial (Figure 3B, black). This discovery also promises that a simultaneous quench of DEPC/BF labeling and buffer exchange to an MS-compatible volatile buffer would work if the intended analysis is by native MS.

Kinetics of BF and DEPC modification for nucleophiles Tyr, Lys, and Cys

We next characterized the reactivity of BF with several nucleophilic side chains on cyclopeptides and compared their reactivities with those of DEPC. For protein analysis, modification should show unimodal second order kinetics if the structure is unperturbed^{39, 40}. The indication that protein HOS is perturbed upon footprinting is a sharp break from the rapid kinetics in their curves, revealing the conditions where the rate constant has changed. Cyclic peptides contain minimal structure and are less susceptible to structural perturbation, so we applied second order rate kinetics fitting to cyclic peptide BF modification to measure “intrinsic” reactivity with the nucleophile (Table S1). BF is found to be more reactive with Lys and Tyr when compared to DEPC, but it is less reactive with Cys (Figure 4). As reported previously³⁰, BF modifies hydroxyl-containing tyrosine at a rate ~10 times greater than DEPC. The modification extent of each peptide failed to reach 100% at all footprinting reagent concentraions (i.e., there is still unmodified cyclic peptides remaining after incubation with the highest concentration of BF and DEPC). Since cyclopeptides possess limited structure, we attribute the break in second order rate kinetics to insufficient incubation time to fully label 100 percent of the cyclic peptides (reaction rates are possibly diffusion limited). The usage of a Cys quench buffer may reverse some of the modifications especially on the Cys-containing cyclopeptide, with more modifications affording more potential for reversal observed in the leveling of the kinetic curve.⁴⁰ The apparent low extent of BF modification of c(RADFC) when using a Cys quench was unexpected, considering the extensive cysteine modification observed when using the Tris quench. Owing to the lower stability of the thioester and presence of excess free cysteine, the modified c(RADFC) may undergo benzoyl transfer to the free and abundant Cys quench (a transthioesterification). DEPC-labeled Cys does undergo this transfer,²³ and methylation of the Cys residues is recommended to avoid this problem. The protocol for DEPC modification implements an imidazole quench that is less likely for transthioesterification, but as noted above (Figure 3A) may not be fully quenching the reaction. Benzoyl transfer is more rapid for some amino acid residues, but both DEPC labeling and benzoylation can be reversible.⁴¹ A workaround for the problem of Cys scrambling is rapid removal of BF via centrifugal desalting columns, as shown above (Figure 3B) serving as a “physical” quench. An alternative if the biological question can only be solved by Cys footprinting is to use NEM, an established reagent for effective footprinting of Cys^{17, 18} and one that does not involve transthioesterification.

Figure 4.

Kinetics of BF (black) vs. DEPC (red) modification for Tyr, Lys, and Cys (A-C, respectively). The extracted k’s (μM⁻¹s⁻¹) for BF and DEPC modification of c(RADFC), c(RADFK), and c(RGDYE) are reported in the bottom left corner of each plot. The inset in C shows a different scale for c(RGDYE) modification. Figure C is adapted from Figure 1F in reference 30.

Identifying BF-modifiable residues of a model protein

Although the use of cyclopeptides has advantages as a start to determine the reactivity of particular residues, cyclopeptides are unlikely to span all amino acids unless specifically synthesized, and thus some residues may remain to be evaluated with the new footprinter. Furthermore, DEPC reactivity depends on the amino acid microenvironment such as the proximity hydrophobic residues,³⁹ which is not simulated using cyclic peptides. Our final step for evaluating novel labeling reagents is to validate the reactivity and specificity of the footprinter using a model protein that contains all canonical amino acids, high footprinting reagent concentrations, long reaction times, and bottom-up characterization of the modification sites. High labeling concentrations can induce structural perturbation of a protein, so lower concentrations of footprinting reagent should be used to ensure labeling on a natively folded protein structure; however, since we are interested in the reactivity of all 20 residues and not protein structure at this stage, we employed high concentrations of labeling reagent. We used bovine serum albumin (BSA) incubated with 500 and 1000 equivalents of BF for 10 min at 37 °C to obtain a survey of potential modifiable residues. Note that higher concentrations of BF and longer incubation times caused precipitation of BSA, likely owing to an increase in hydrophobicity imparted by benzoylation of amino acid side-chains.

Because benzoylation intrinsically alters amino acid side-chains, the protease should be chosen carefully for bottom-up analysis of BF-modified proteins. The addition of a benzoyl group to a particular residue may decrease the activity of a protease at a modified residue; these missed cleavages can decrease coverage and hinder relative quantification of modification sites. Considering preliminary results on cyclopeptides showing footprinting on Lys, Cys, and Tyr residues, a protease that preferentially cleaves at these residues is likely a poor choice. Although trypsin and chymotrypsin are two of the most common proteases for bottom-up proteomics, trypsin digestion at Lys and chymotrypsin digestion at Tyr (known BF-reactive residues) precludes their use for digestion of BF-modified proteins. Instead, Glu-C, another common protease that cleaves at the C-terminus of Asp and Glu residues, was chosen for this bottom-up analysis. Asp and Glu are potentially reactive with BF, but the product, an acid anhydride, is unstable at low pH. From the cyclopeptide analysis, no footprinting was observed on Asp residues, likely because acidification with formic acid for reversed phase LC-MS and LC-MS/MS analysis decomposed these modifications. Assuming similar behavior for both Asp and Glu side chains, 1 μL of formic acid was added after the incubation of BSA with BF to ensure reversal of any Asp and Glu benzoylation. Using this digestion protocol, we observed > 96% coverage of BF-labeled BSA, indicating that Glu-C is an appropriate protease for BF footprinted BSA (Figure S5).

To determine the reactivity and specificity of BF labeling of each residue, we allowed BF modification on all amino acids for the initial search of the data. Each unique peptide was verified manually to determine sites of BF modification; for “true-positive” footprinted residues, the fragmentation of the peptide must be sufficient that the modification can be localized to a particular amino acid(s) (Figure S6). Those peptides without MS/MS fragmentation pinpointing the exact site(s) of modification, where the residues could not confidently be determined to undergo BF labeling, were deemed uncertain or a false-positive. Neither uncertain nor false-positive modifications were included in our final analysis of overall modification.

We find that true-positive benzoylation occurs on Lys, Tyr, Ser, Thr, Cys, and the N-terminus of the protein (Table 1). These findings are consistent with the BF-modifiable nucleophilic groups identified using the cyclopeptides, with exception to c(GRGDSP), with modification on the protein likely due to microenvironment enhanced reactivity for Ser. In particular, BF is very reactive with Lys residues, as 90% of the Lys residues in BSA were footprinted. BF could potentially be a high-quality footprinting option if Lys residues are of high interest, such as in an investigation of protein-protein binding interactions where Lys residues are proposed to play a critical role. Although this survey of modifiable amino acids was generally successful, a similar experiment could be carried out by labelling the protein in a denaturant (if the denaturant is not reactive towards the labelling reagent) or using digested proteins with capped N-termini.³⁹ Finally, BF may afford additional uses such as hydrophobic core penetration, as observed for tetranitromethane^{42, 43}—a phenomenon that would be missed in the overlabeling experiment reported here.

Table 1.

Positively identified residues and percentage of possible residues of BSA modified via benzoylation.

Modified Residues	Count	Locations of Benzoylation on BSA	Total in BSA	Not Covered	% Modified
K	54	N-term, 4, 12, 20, 41, 51, 64, 76, 93, 106, 114, 116, 127, 131, 132, 136, 159, 173, 180, 211, 221, 232, 239, 242, 261, 273, 275, 280, 285, 294, 312, 316, 322, 350, 375, 377, 396, 431, 439, 465, 471, 474, 499, 504, 520, 523, 524, 533, 535, 537, 544, 556, 563, 573	61	1	90%
Y	8	139, 149, 155, 156, 331, 333, 451, 496	21	1	40%
S	9	5, 58, 65, 79, 286, 442, 453, 488, 577	32	1	29%
T	8	52, 68, 121, 238, 507, 539, 578, 580	34	1	24%
C	6	53, 62, 123, 278, 436, 460	35	5	20%

CONCLUSION

MS-based footprinting requires not only novel reagents with improved specificity, but also tests of rigor in the workflow. We have compiled this evaluation approach to validate BF as a next-generation nucleophile footprinter by demonstrating functional group specificity on cyclic peptides, establishing appropriate quenching reagents, discovering reagent-specific limitations (i.e., matrix effects), determining reaction kinetics compared to an established reagent, and evaluating the reactivity and specificity for all 20 amino acids. Using this workflow, we find that BF specifically reacts with Lys, Tyr, Ser, Thr, Cys, and the N-terminus; the extent of BF footprinting is affected by pH and buffer system (in particular, Tris buffer); a Cys-buffer quench appears to be most efficacious for stopping BF footprinting, but a physical quench is preferred to prevent reversibility. Before application of covalent labelling or footprinting to a novel protein system, we recommend system-specific analyses identifying appropriate concentrations of reagent to minimize structural perturbation. These initial assessments with reagents are vital to understand the capabilities and limitations of the footprinter to determine appropriate broad scope applications. Development and evaluation workflows such as these or others should be applied to both new and already established footprinters to better understand their properties and limitations. Typically, innovations and discoveries of novel protein footprinting reagents are incremental. Overall, this proposed suite of evaluations is a step towards standardization of novel footprinting reagent development, bringing footprinting-MS closer to the more widespread application of HDX-MS.