MALDI Peptide Mapping for Fast Analysis in Protein Footprinting

Abstract

Although protein footprinting results are commonly obtained by ESI-based LC-MS/MS, a more rapid-turnaround alternative approach is desirable to expand the scope of protein footprinting and facilitate routine analysis such as monitoring protein high order structure in quality control or checking epitope maps. Considering that MALDI is a faster procedure that can be easily adapted for high-throughput analysis, we explore here the feasibility of developing a MALDI-based analysis “portfolio” of bottom-up peptide mass mapping for footprinting. The approach was applied to several model proteins that were submitted to two footprinting strategies, FPOP and GEE labeling, and their performance was evaluated. We found adequate coverage that can be improved with automatic off-line separation and spotting, demonstrating the capability to footprint accurately protein conformational change, showing that MALDI may be useful for selected applications in protein footprinting.

Introduction

To understand protein high order structure and protein interactions with other biomolecules, protein footprinting is being developed and improved by inclusion of novel labeling reagents ^1-5, deeper insights into reaction mechanisms ⁶, optimized labeling^7-9, proteolysis ¹⁰, quantification ¹¹ strategies, and software support ^12-14. Combined with other orthogonal approaches such as X-ray crystallography, cryo-EM, immunoblotting, crosslinking, mutagenesis, and computational modeling, footprinting should play an indispensable role in understanding protein systems ^{11, 15-22}. Although data processing is now more rapid than in the early days of Fast Photochemical Oxidation of Protein (FPOP), for example, the turnaround in determination of a footprint has remained relatively slow despite the general advances in MS instrumentation.

One reason for slow turnaround is that electrospray ionization (ESI)-based LC-MS/MS, the currently dominant protein footprinting analysis approach, is limited in instrument turnaround principally by the HPLC gradient and re-equilibrium time for bottom-up ESI MS/MS. An alternative is matrix-assisted laser desorption ionization (MALDI). Although both MALDI and LC/ESI require blanks, each blank requires approximately 1 h for LC/ESI whereas they require a fraction of a minute for MALDI, lowering the overhead on the instrument. In addition, the stringent requirement for desalting in ESI analysis is reduced but not removed with MALDI. Although protein digestion is also slow and would be rate-limiting if done in series with the analysis, the sample preparation for both approaches is done in parallel, but the turnaround is considerably faster for MALDI.

We argue here that a higher throughput MS analysis method would promote broader adoptions of protein footprinting. For example, quality control in antibody development is needed to assure that an antibody binds at the same epitope, a time-consuming analysis with current footprinting approaches like HDX, FPOP, or specific- amino acid chemical modifications²³.

Although shortening the gradient time in LC-MS/MS at the cost of chromatographic resolution can help speed up analyses, ion suppression is minimized with good separation, both in MALDI and ESI²⁴, but for ESI some cleanup including desalting is done online, consuming instrument time, whereas for MALDI, it is executed offline.

MALDI may offer another solution. First, it can be easily adopted for automated high-throughput analysis ^25-28 because it requires less sample preparation than LC/MS and provides fast analysis time (a few seconds). Furthermore, MALDI peptide fingerprint mass mapping, which often involves no online LC or lengthy desalting pretreatments, is routine for analyzing protein digests ²⁹. Given the higher tolerance to contaminants that MALDI possesses over ESI ^30-32, MALDI promises to reduce the analysis time by removing or simplifying the online HPLC/desalting steps that occupy substantial time for LC-MS/MS. Forfeiting the LC also avoids purging carryover from the LC column between samples, usually by running lengthy blanks. Although the latter analysis can yield detailed solvent accessible surface area (SASA) information with high spatial resolution, a more rapid-turnaround approach to give “coarse-grained” results that do not require full coverage may suffice for some problems and be more time-efficient than LC/MS.

Even as one of the two “pillars” in modern protein MS analysis, MALDI plays a minor role at best in protein footprinting. It is often not recommended for reversible footprinting (i.e., hydrogen-deuterium exchange or HDX) owing to unavoidable and variable back exchange in the ionization.³³ Early on, several structural studies utilized MALDI to detect irreversibly labeled proteins at either the peptide or protein levels^34-37. In those studies, the footprinting reagents targeted specific residues, and conformational changes were determined simply based on whether the footprinted peptides appeared in the MALDI mass spectrum. Following those early studies, MALDI was replaced in footprinting by LC-MS/MS because the latter approach has high dynamic range, uses online separation with HPLC, affords routine and reliable MS/MS, and permits adequate quantification of labeled and unlabeled peptides. The questions facing MALDI are the ability to quantify peptides in protein footprinting, which has not been systematically evaluated to our knowledge, and the dynamic range for minor modifications.

Here we begin evaluation of a MALDI-based approach to increase the throughput of MS analysis for protein footprinting. We chose two protocols, one that includes no-LC and a second that uses an LC-MALDI spotting robot for off-line separation. Of course, incorporating an offline HPLC requires time for separation but does not slow instrument turnaround if samples to be collected, prepared, achieved, and batched before analysis, providing efficiency in the MS analysis and saving valuable spectrometer time. Furthermore, MALDI permits easy reanalysis. We tested methods with model protein systems submitted to either residue-specific footprinting or non-specific fast unspecific footprinting (e.g., FPOP). Our preliminary results suggest that a MALDI-based approach can increase the speed of analysis, while reporting protein SASA-based information that constitutes a footprint and indicate that a MALDI-based approach should be explored further. We are not comparing bioinformatics as we assume data processing and searching are available for both approaches. The focus of the research described here is whether instrument time can be saved in footprinting.

Materials and Methods

Materials:

MALDI matrices of α-cyano-4-hydroxycinnamic acid (CHCA) were purchased from Bruker Daltonics (Bremen, Germany). The ACTH 1-17, L-glutamine, L-methionine, catalase, hydrogen peroxide (H₂O₂), trifluoroacetic acid (TFA), formic acid (FA), phosphate buffer saline (PBS, 10 mM phosphate, 138 mM NaCl, 2.7 mM KCl), urea, dithiothreitol, iodoacetamide, 1-ethyl-3-[3-(dimethylamino)propyl] carbodiimide (EDC) and glycerol ethyl ester (GEE) were from Sigma Aldrich (St. Louis, MO). Trypsin was from Promega (Madison, WI). ZipTip^® C18 Pipette Tips (Silica, 15 μm, 200 Å pore size) is from Millipore.

Calibration:

The model peptide ACTH 1-17 was fully oxidized by incubating a 100 mM peptide solution with hydrogen peroxide at 1:1000 molar ratio for 6 h. The oxidation was quenched by adding 0.5 mL catalase (18 mM). To obtain multiple calibration curve points, the peptides were mixed to give 0%, 5%, 10%, 20%, 50%, 75%, 100% at a total concentration of 100 mM, before MS analysis.

FPOP:

The protein sample 10 μM solution in PBS was incubated with 20 mM H₂O₂ and 500 μM histidine before injection into the FPOP tubing. The flowrate was adjusted to give ~20% irradiation-excluded volume to minimize repetitive laser exposure. The laser beam was from a KrF excimer laser (GAM Laser Inc.), providing an excitation wavelength of 248 nm to yield hydroxyl radicals by H₂O₂ photolysis. After laser irradiation, the sample solution was collected in a tube containing 10 mM catalase and 20 mM Met to quench leftover H₂O₂ and prevent post-labeling oxidation. Samples of each state were subjected to FPOP in biological duplicate.

GEE footprinting:

For GEE footprinting, glycine ethyl ester (GEE), 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) stock solutions were prepared fresh in PBS buffer. GEE was added to each pre-equilibrated sample, followed by the addition of EDC. The final concentrations of protein, GEE, and EDC were 10 μM, 200 μM, 50 mM, respectively. The reaction was carried out for 1.0 h in biological duplicate, before quenching by adding equal volumes of 1 M ammonium acetate and desalting with a Zeba column (Thermo Scientific, Rockford, IL).

Proteolysis:

Urea (8 M) was added to dissolve the protein pellets. Trypsin was then added at an enzyme-to-protein ratio of 1:10 (w:w). Samples were incubated at 37 °C for 12 h. The digestion was quenched by adding formic acid to give a final concentration of 1% (by volume).

MALDI sample preparation:

For MALDI-based studies, a matrix solution of CHCA was prepared as a saturated solution (>10 mg/mL) in 70% aqueous acetonitrile. Protein solutions (10 μL) were desalted by using a Ziptip following the manufacturer’s instructions before mixing with 10 μL of matrix solution in an Eppendorf tube and vortexing for at least 60 sec. Aliquots (2 μL) of this mixture were spotted onto the MALDI stainless steel target plate and air-dried.

Mass Spectrometry:

A Bruker Daltonics UltrafleXtreme MALDI TOF/TOF mass spectrometer (Bremen, Germany) equipped with a Smartbeam-II laser (Nd:YAG, 355nm) was used for MALDI data acquisition. The instrument, equipped with a reflectron, was operated in the positive-ion mode. The Ion Source 1 was supplied with 20 kV acceleration voltage, and 17.65 kV was applied to Ion Source 2. The pulsed ion extraction (PIE) delay was set to 130 ns. The Smartbeam-II Laser parameters were: global attenuator offset, 20%; attenuator offset, 40%; attenuator range, 40%; the laser spot size was set to “3_Middle”. The laser was operated at 1000 Hz for data acquisition. The laser fluence was defined as 100% laser power minus laser beam attenuation that was controlled by the operator. For peptide mass mapping, analytes were ionized with laser fluences slightly higher than threshold to minimize peaks broadening. For LC-MALDI, mass spectra were acquired in a fully automated mode by Warp-LC (Bruker Daltonics).

For LC-MS/MS analysis, 5 μL digested sample was pre-concentrated on an Acclaim PepMap C18 column (Thermo Scientific, 100 μm × 2 cm, 5 μm, 100 Å) and desalted for 15 min before elution. A 20 cm custom packed C18 column (Magic, 75 μm × 15 cm, 5 μm, 8 Å) was then used to separate the analytes. Solvent A and B were the same as above. Peptides were eluted at a flow rate of 500 nL/min with the following gradient: 2% B from 0 to 10 min, then ramped to 20% in 47 min, ramped to 50% B in 29 min, increased to 98% in 10 min, held at 98% for 3 min, returned to 2% B in 1 min, and equilibrated at 2% B for 20 min. LC separation was directly coupled to online detection using a Q Exactive Plus (QE) hybrid quadrupole orbitrap mass spectrometer with a Nanospray Flex ion source (Thermo Fisher, Santa Clara, CA). The ten most-abundant ions in the mass spectrum were submitted to higher energy collision dissociation (HCD) for identification. The mass resolving power was 70,000 (m/z 400) for MS1 and 17,500 (m/z 400) for MS/MS.2.3.8

Automated HPLC fraction collection:

The automatic HPLC fraction collection/deposition was performed using a Proteineer fc robot (Bruker Daltonik, Germany) as a MALDI spotting device. The robot was connected via fused-silica capillaries to the HPLC outlet column. The robot was programmed to start the fraction collection with a delay of 20 min, with respect to the starting time of the LC run sequence. A volume of 700 μL MALDI matrix was aspirated into the needle approximately 15 min after the start of the run. An Anchor Chip 384 microtiter plate (Bruker Daltonics) with 0.6 μm anchor diameter was used as the MALDI target. The collection time of the LC eluent was set to 15 s per spot, including also the robot moving time. Concomitantly, the applied MALDI matrix was collected at the tip of the robot needle in continuously at a flow rate of 100 μL/h and subsequently deposited on the MALDI target. In this way, 0.12 μL of LC eluent was mixed with 0.42 μL of matrix solution at the tip of the Proteineer needle prior to spotting onto the Anchor Chip, thus basically following the “dried-droplet” sample preparation. Between consecutive runs, a rinsing step of the Proteineer needle with methanol/water (1:1) solution was carried out.

Data Processing:

The MALDI spectra were initially evaluated by using FlexAnalysis (Bruker Daltonics). In the LC-MALDI experiments, the spectra were combined by Warp-LC (Bruker Daltonics). The peak lists generated from FlexAnalysis (or from a compound list from Warp-LC) contain the monoisotopic masses. The MALDI-MS data were reduced to only monoisotopic masses, or in LC-MALDI case, plus retention time. From an in-silico digest, a peptide list with predicted monoisotopic m/z at +1 charge was transferred to Excel, the footprinting mass shift were entered, and a custom-built VBA (Figure 1) was run to yield the matching candidates for manual validation. The validation focused on monoisotopic mass mislabeling.

Figure 1. Peptide fragments analyzed by MALDI from FPOP footprinting of myoglobin.

(A) structure of apo-myoglobin; F marked in red forms a helix upon heme-binding. Adapted with permission from Ref. 1. (B) Mass spectra acquired of intact apo-myoglobin after FPOP footprinting. The multiple +16 signals are due to oxidations induced by FPOP. (C) the oxidation fractions for each tryptic peptide are based on their peak areas in the MALDI mass spectrum. The results acquired in the usual LC-MS/MS mode on a Thermo Q-Exactive Plus (QE) (apo-myoglobin and holo-myoglobin (with QE)) are provided as benchmarks. Mass, S/N, intensity, and area of the analyte peak signals were copied to an Excel worksheet (Microsoft, Redmond, WA) and analyzed through a custom-programed script (Virtual Basic for Application, VBA, in Excel) that automatically identified the target based on its theoretical monoisotopic mass (within 50 ppm) and calculated the modification fraction.

Results and Discussion

Quantification Calibration:

To determine whether MALDI can faithfully quantify modifications introduced by footprinting, we calibrated a mixture of oxidized peptides of ACTH 1-17 (Figure S1) where modification fraction is plotted versus the analytical fraction of the modified peptide over a range of unmodified to entirely modified.

We conducted the analysis with the model peptides in water or in the presence of 100 mM BSA digest to mimic the situation where only a few peptides in a mixture need analysis. The purpose is to test whether a complex sample environment affects MALDI quantification. This is suitable for the analysis done here but would need expanding for more complex analyses (e.g., for antibodies). Here the modification is induced by oxidation, and its fraction is calculated by eq. 1 in which MALDI MS provides the various signal intensities:

$$OxidationFraction=\frac{Intensit{y}_{oxidized}}{Itensit{y}_{oxidized}+Itensit{y}_{wild}}$$ Eq. 1

Eq. 1 is also used to calculate other types of modification. The oxidative modifications can be replaced with any other modifications that occur in the footprinting experiments.

For an aqueous sample, the curve shows excellent linearity with an R² of nearly 1 (Figure S2 A). Spiking with the BSA digest affects the curve’s linearity (R² = 0.945, Figure S2 B), although only in a minor way.

The dependence of oxidation fraction (OF) on the modified fraction in the sample is described by Eq. 2 ³⁸:

$$\lambda =\frac{(a-OF){\rm Y}}{OF(1-2{\rm Y})-[c-(a+c){\rm Y}]}$$ Eq. 2

where $\gamma =\frac{lonizationEfficency(IE{)}_{Modified}}{I{E}_{Wildtype}+I{E}_{Modifided}}$, a and c are constants. When $\gamma =0.5$, namely, $I{E}_{Wildtype}=I{E}_{Modifided}$, Eq. 2 can be reduced to a linear equation (Eq. 3):

$$OF=(c-a)\lambda +a$$ Eq. 3

The equation indicates that when the oxidized and the unmodified analytes have the same ionization efficiency, the calibration curve should be linear, an outcome that is consistent with our calibration curve analysis from the model peptides ACTH 1-17. Thus, we infer that the ionization efficiency is not significantly altered upon peptide oxidation although the presence of other peptides does complicate the situation by intensifying the difference in ionization efficiency before/after the modification. In summary, the good linearity of the calibration curve strongly suggests that MALDI can be used to quantify the modification induced by protein footprinting.

FPOP footprinting:

To evaluate the MALDI peptide mass mapping method, we chose myoglobin in its apo and holo states as a test case for FPOP footprinting. In MALDI mapping, each unmodified peptide and the modified counterpart were identified by their accurate monoisotopic mass, and the modification fractions were calculated from their signal intensities. Traditional LC-MS/MS analysis of the same samples was done by LC-MS using a Thermo QE orbitrap instrument for comparison.

From the FPOP-labeled myoglobin digest, seven peptides were identified by MALDI mass mapping, covering nearly half the protein sequence (Table 1). The MALDI sample preparation and the spectral acquisition took 20 mins. By way of comparison, the LC-MS/MS analysis on a QE required 2 h to finish an LC gradient and to record the mass spectra. We identified six additional modified peptides by the MALDI-based approach and found protein sequence coverage of approximately 75%, as compared to less than 50% by MALDI.

Table 1. Comparison of MALDI and LC-MS/MS analysis of oxidative modifications of myoglobin.

Peptides detected for footprinting of both the apo- and holo-states.

	MALDI	LC-MS/MS (QE)
# Peptides with oxidation	7	13
Peptide sequence coverage	46.4%	76.5%
Time	20 min	2 h

In MALDI, coverage and certainty are traded for speed and throughput even though the detection limit of MALDI does not lag far behind that of ESI when both methods are optimized.³⁹ MALDI suffers more from ion suppression effects when samples are complex, reducing the ionization efficiencies of most analytes. Consequently, the sensitivity in LC-free MALDI peptide mapping does not match that for the LC-MS/MS method, and some modification sites become barely detectable.

The 2-h gradient used in the LC-MS/MS run is designed to achieve the maximum peptide coverage in a general footprinting protocol used in our lab. To save time, this gradient can be shortened to improve the analysis throughput for the LCMS protocol. Even with shortening of that time, the gradient time is seldom shorter than 50 mins in our lab as determined by the chromatographic resolution, ion-suppression effects, and salt tolerance, as discussed in the introduction. Nevertheless, the time gap between MALDI and LC/ESI MS is still considerable.

The compromised lower coverage in the MALDI approach can be overlooked if the peptides that are of interest are well ionized and the sequence well covered as could be the case for protein-protein or protein-ligand interactions (e.g., epitope mapping) The outcome found here would be inadequate in basic discovery research, but the approach could provide sufficient coverage for many routine analyses where the few peptides that carry the needed information are well detected. If they are not, the digestion condition can be adjusted, or the mixture can be simplified by offline separation before the MALDI analysis (see below).

For myoglobin, the main structure difference between apo and holo is in the region 80-96 (Figure 1 A), where a helix exists for holo but is absent for apo. The helix should reduce the regions’ solvent assessable surface area, rendering it less accessible to the oxidative radicals utilized for footprinting in the FPOP. In both the ESI QE and MALDI approaches, we could see oxidized peptides covering the region of conformational change (peptide 78-96, 79-96). The oxidation levels of these peptides, however, are not dramatically different for the apo vs. holo forms (Figure 1 C). A possible explanation is that the iron in the heme promotes Fenton chemistry, generating additional ·OH radicals ⁴⁰ and producing modification not commensurate with the reduced SASA caused by the helix formation.

Most peptide modifications are quantified similarly by both MALDI and ESI analysis (on an orbitrap), but in some cases, considerable differences in relative abundances are observed; examples are peptides 119-133 and 134-139. For peptide 119-133, the oxidation fraction determined by MALDI is almost ten-times greater than that measured with QE orbitrap, whereas the situation is reversed in the case of peptide 134-139.

To explain in part this discrepancy, we consider some fundamental differences in the data acquisition of the two approaches. Since the LC separation is not on-line in MALDI, the mass spectra are more “congested” than those from LC-MS/MS, leading to more overlapping of the peptide signals and artificial increases in the signal intensity. As the previous calibration curve shows, the presence of more peptides also alters the analyte ionization efficiency in unpredictable ways.

For the on-line LC approach, the analyte signal often is split into different charge states, especially for large peptides whereas in MALDI, usually only singly charged peptides are observed.

Although MALDI is often not viewed as a preferred method for quantification because of its poor reproducibility⁴¹, the standard deviations from MALDI mass mapping are comparable to those measured for the QE orbitrap for FPOP-labeled myoglobin, and the two methods report similar modification trends for the apo and holo states, suggesting that trends captured by MALDI are trustworthy.

The major advantage of MALDI peptide mapping analysis is its high throughput. Any LC-MS/MS approach will use a long gradient relative to MALDI introduction, a re-equilibration of the analytical column, desalting online, and a blank to ensure all the analytes from last analysis had eluted entirely and the LC system is ready. By comparison, MALDI peptide mapping takes only seconds for integrating many spectral acquisitions and outputting an average. Even with desalting and preparing sample-matrix crystals, a single sample can be comfortably analyzed in minutes, and instrument time can be further decreased when multiple samples are deposited onto a single MALDI plate, giving a “multiplex” advantage.

GEE footprinting:

We evaluated a different labeling strategy using the Asp/Glu-specific GEE labeling on the apo and holo states of myoglobin, and the results were also analyzed by MALDI peptide mapping (Figure 2). The GEE modification fractions on peptides from apo-myoglobin are either smaller or on par with those from holo-myoglobin, except for two peptides, 78-96 and 79-96. Those two peptides cover the myoglobin conformational-change region; hence reduced modifications are occurring owing to the holo-myoglobin helix formation that reduces the SASA in this region.

Figure 2. Myoglobin footprinted by GEE labeling and analyzed by MALDI peptide mapping.

Significant modification fraction changes (p < 0.05) for 78-96 and 79-96, where the helix forms in the holo-myoglobin.

We strongly suggest when “coarse-grained” SASA information resolved at peptide-level is sufficient to solve a problem and the informative peptides are well-detected, MALDI peptide mapping can greatly increase the speed of protein footprinting, while still capturing conformational changes, at least for GEE footprinting.

LC-MALDI:

We showed that through MALDI direct peptide mass mapping, a protein footprint can be quantified to reflect SASA differences caused by a protein’s conformational change. The number of identified modified peptides and the sequence coverage they afford is lower than those obtained with the LC-MS/MS methods. Higher sensitivity is needed when the modified peptides from the region of interest do not reach the detection limit or when higher sequence coverage of the protein is needed in complete surface mapping. Therefore, the next step on method development is to increase the coverage and sensitivity of MALDI-based analysis approach to permit broader applications.

The limited coverage of MALDI peptide mass mapping (Table 2) as compared to that of the LC-MS/MS largely stems from MALDI suppression of components in complex samples owing to the lack of an LC separation ⁴². When complex samples (e.g., protein digests) are analyzed by MALDI, some peptides are preferentially ionized, resulting in biased detection owing to varying ionization efficiencies for those that do not “fly” well. Considering that most footprinting methods modify only a small fraction of the constituent peptides of a protein (i.e., the information is concentrated in only a few regions of the protein), it is imperative to adjust the digestion (e.g., change enzymes, vary pH, adjust chaotrope) to emphasize those peptides that are informative. This need is consonant with a goal to use the optimized method, originally set up using ESI/MS/MS, for routine detection when many experiments are needed to be a check on the biochemistry of the system (e.g., quality control).

Table 2. Comparison of sensitivities of direct peptide mapping, LC-MALDI, and LC-MS/MS(QE) for the analysis of GEE-labeled apo-calmodulin.

The LC-MALDI uses a gradient of 48 min, whereas the gradient time for LC-MS/MS is 57 min.

	MALDI	LC- MALDI	LC-MS/MS
# of peptide with modification	4	10	12
Peptide sequence coverage	26%	74%	78%

We improved the coverage by coupling automated chromatographic separation (RPLC) to reduce the complexity of the mixture and using a gradient for separation similar to that used in LC-MS/MS. We followed the separation with automatic matrix deposition and sample loading and evaluated its performance by analyzing GEE-footprinted apo-calmodulin as compared to the results from LC-MS/MS (Table 2). Although this strategy returns the LC to the overall analysis protocol, that analysis is offline and can be used to prepare many samples, preferably in an automated way, without significantly compromising more valuable instrument time.

In the absence of preseparation, we identified four modified peptides that cover only approximately one quarter of the protein’s sequence. With preLC separation, the coverage is considerably improved; the number of modified peptides increased to ten, covering three quarters of the protein sequence, a result that is comparable to that from LC-MS/MS. The result confirms that the low coverage of MALDI peptide mapping is mainly due to the complexity of the protein digest and can be improved through off-line LC separation.

We re-analyzed the GEE-labeled apo/holo-myoglobin with the LC-MALDI approach (Figure 3). With LC-separation, all seven previously identified peptides were seen, and six additional modified peptides were identified (1-16, 46-56, 48-63, 64-77, 146-153). Furthermore, significant modification fraction changes were again observed for the 78-96 and 79-96, consistent with the well-known conformational change induced by helix formation in this region of holo-myoglobin.

Figure 3. LC-MALDI analysis on GEE-labeled apo/holo-myoglobin.

The identified peptides cover 94% of the protein’s sequence. Significant modification change is observed (p < 0.05) for peptides 78-96 and 79-96, for which helix F is formed in the holo-myoglobin chain.

We noted that the reproducibility of the FPOP oxidation ratio measured by MALDI is better than the reproducibility of the GEE modification ratios in the Figures 2 and 3, as indicated by the smaller error bars from the biological duplicates in Figure 1. This is because FPOP is an online footprinting set up, whereas GEE labeling is an offline experiment. As already explained in the methods section, the footprinting reaction time for FPOP is accurately controlled by the apparatus flowrate and laser-pulse frequency. Therefore, the deviations in footprinting extents were inherently smaller for FPOP than for GEE reactions.

Conclusion

Overall, we found bottom-up MALDI to be promising for some applications as a high-throughput approach for obtaining coarse-grained footprinting information about selected regions of a protein. Although in the absence of instrumental strategies to give better coverage, standard MALDI has difficulties obtaining higher coverage than LC-MS/MS and may not be the method of choice for discovery. Instead, MALDI seems more appropriate for routine monitoring. We suggest that an investigator should first examine the full protein by using ESI/MS/MS to identify informative regions and then tune the digestion protocol to ensure informative regions are covered by MALDI.

The increase in speed of analysis should open possibilities for routine analysis by MS-based footprinting in cases where a binary answer is required. An example may be protein stress studies where the regions of conformation change are known from a detailed, high-coverage experiment by LC-MS/MS and only those regions need to be monitored routinely. Another example, as suggested earlier, may be epitope mapping where the epitope is known from LC-MS/MS footprinting, and it only needs to be confirmed in quality control. In these cases, full coverage is not required.

Future studies of MALDI-based protein footprinting should extend the approaches to more proteins, to larger, more complicated protein systems, and to improvements in reproducibility of quantification.