This review outlines the application of time-resolved electrospray ionization mass spectrometry (TRESI-MS) and hydrogen-deuterium exchange (HDX) to study rapid, activity-linked conformational transitions in proteins. The method is implemented on a microfluidic chip which incorporates all sample-handling steps required for a ‘bottom-up’ HDX workflow: a capillary mixer for sub-second HDX labeling, a static mixer for HDX quenching, a microreactor for rapid protein digestion, and on-chip electrospray. By combining short HDX labeling pulses with rapid digestion, this approach provides a detailed characterization of the structural transitions that occur during protein folding, ligand binding, post-translational modification and catalytic turnover in enzymes. This broad spectrum of applications in areas largely inaccessible to conventional techniques means that microfluidics-enabled TRESI-MS/HDX is a unique and powerful approach for investigating the dynamic basis of protein function.
time-resolved electrospray ionization mass spectrometry
Conventional structural biology techniques, such as X-ray crystallography and structural NMR, have greatly advanced our understanding of protein structures, protein interactions and catalysis using static representations of typically low-energy-state conformers, but are generally ill-suited to the study of transient intermediates and excited species [1-4]. Analyzing short-lived intermediates of proteins as they transition between states, whether in protein folding, enzyme catalysis or protein–protein interactions, is a major challenge in analytical biochemistry. As a result, there is a substantial gap in our understanding of the dynamic processes that underlie protein-mediated biological function in the cell.
The occurrence of short-lived protein intermediates has been demonstrated by a variety of spectroscopic methods, including circular dichroism (CD) [5, 6], fluorescence-based assays  and NMR spectroscopy [8, 9]. However, these methods have restricted applicability with regard to performing kinetic experiments that provide structural and/or mechanistic insights. CD spectroscopy is a useful method for monitoring protein unfolding, or melting, allowing changes in the secondary and tertiary structures of proteins to be discerned, and, in certain cases, examination of protein–protein interactions [10, 11]. However, quantitative structural interpretation of the CD spectra is at best limited to an ‘aggregate’ measurement of secondary structure, and is reliant on calibration to known structures. Fluorescence-based methods, particularly Forster resonance energy transfer (FRET), have been widely applied in time-resolved studies of enzyme catalysis and protein–ligand interactions, and offer a large sensitivity range . However, these approaches provide limited structural insights, usually in the form of single inter-probe distance measurements, whose biological significance is somewhat reduced by the presence of the FRET probes themselves . Parameters obtained by the various analytical techniques have also enabled application of a vast number of computational tools for the theoretical study of protein dynamics . However, current in silico simulations are still greatly limited by processing power, with a trade-off between the accuracy of predictions and the complexity of the models and the built-in parameters [15, 16].
Of the various spectroscopic techniques applied in analytical biochemistry, NMR spectroscopy has made the largest contribution to studies of protein structure in solution. However, applications of NMR in kinetic studies of transient protein states are generally restricted by the relative insensitivity of the technique (requiring mid- to high micromolar concentrations), and the size limit of proteins that may be analyzed (typically, a subunit size of less than 70 kDa with use of transverse relaxation optimized spectroscopy) . Furthermore, performing time-resolved kinetic experiments in real-time with the application of NMR on the sub-second timescale is practically challenging, requiring complex pulse sequences with greatly diminished sensitivity.
Recent developments in biophysical NMR methods have allowed some insight into millisecond-timescale conformational dynamics of proteins. In particular, Carr–Purcell–Meiboom–Gill (CPMG) relaxation dispersion allows site-specific characterization of the kinetic and thermodynamic properties of nuclei on the microsecond/millisecond timescale as they sample different environments during conformational exchange . However, a continuous presence of the species/conformational states analyzed throughout the time window of the relaxation dispersion experiment is required, and these experiments are thus typically limited to systems at equilibrium . Furthermore, systems exchanging between more than two states are difficult to characterize by CPMG experiments .
Mass spectrometry with hydrogen/deuterium exchange (HDX) has recently emerged as a powerful tool in structural biology, especially in the context of high-throughput structural proteomics . Mass spectrometry offers the ability to perform millisecond-timescale measurements , in addition to high sensitivity and the potential for direct in-line coupling to a variety of liquid chromatography (LC) and microfluidic devices , allowing kinetic experiments to be implemented with ease [24, 25]. Electrospray ionization (ESI) and matrix-assisted laser desorption/ionization methods enable analysis of whole proteins of any size, and protein complexes larger than 1 MDa have been successfully analyzed [26, 27]. HDX experiments in conjunction with MS-based analysis permit simultaneous structural and kinetic analyses, enabling complex protein energy landscapes to be mapped in protein folding and enzyme catalysis studies [28-30].
Time-resolved electrospray ionization mass spectrometry (TRESI-MS) allows the detection of full-length, native proteins in kinetic experiments in which virtually all mass-distinguishable species may be monitored simultaneously [31-33]. In the study of enzyme catalysis, TRESI-MS has proven to be a remarkably enabling technology, with almost universal application (in principle) to characterization of catalytic reaction intermediates using native enzymes and substrates [31, 34]. This is a significant improvement over standard spectroscopic techniques that rely on colorigenic (chromogenic) or fluorogenic substrate analogs and methods utilizing radioactive substrates. Applied to protein dynamics, TRESI-MS, in combination with HDX, facilitates structural analysis of the dynamic regions of proteins undergoing conformational changes during biological activity [29, 35]. This review explores the unique capabilities of TRESI-MS/HDX in the study of rapid, functionally relevant conformational transitions in proteins, particularly in the challenging class of intrinsically disordered proteins and active enzymes.
Label-free analysis of protein structure by mass spectrometry
The intrinsic analytic capabilities of mass spectrometry are of limited utility on their own in the study of protein structure and dynamics. One direct method is based on the propensity of more unfolded proteins to acquire more charges in the process of electrospray ionization . The resulting distribution of multiply charged ions in the mass spectrum reflects the extent of protein unfolding, but this information is also limited to the overall 3D packing of the protein.
Ion mobility mass spectrometry is emerging as a label-free technique for structural analysis in the gas phase. A weak electrostatic field is applied across the ion drift space, influencing macromolecular ions to separate based on their drift velocities . Ions sample different possible geometric orientations as they proceed through the drift tube, and, in the presence of a neutral background gas, macromolecular ions with different sizes and geometries pass through the drift region at different velocities. The results are collisional cross sections that vary with the 3D organization of the ions . Therefore, ion mobility spectrometry provides information on the overall 3D packing of the protein and easily differentiates between folded and loosely folded states, but provides no true structural resolution.
Mass spectrometry with HDX
The advance that significantly enhanced MS-based analysis of protein dynamics is isotope incorporation by means of HDX . This approach has been extensively reviewed , and will be only briefly introduced here. HDX relies on the ability of hydrogen atoms to undergo exchange with deuterons from the solvent (or surrounding gas), resulting in incorporation of deuterium at accessible sites on the macromolecule. In solution, HDX is base-catalyzed (although general acid catalysis becomes dominant at pH < 2.6). The rate of HDX at individual sites is dependent on a multitude of factors, including the character of the functional group, accessibility to the exchange reagent, the degree of hydrogen bonding of the labile hydrogen, and the nature of deuterium donor, which dictates the exchange mechanism (i.e. salt bridge, tautomer, onium ion, flip-flop, and relay mechanisms) .
MS-based HDX measurements may be implemented ‘globally’ to give information on the overall flexibility of proteins (relevant to protein stability under specific conditions), or ‘locally’ to distinguish regions that exhibit varying degrees of protection from exchange due to structure. ‘Global’ measurements are made simply by exposing the protein to D2O for a variable period, followed by online or offline MS analysis to measure the increasing mass associated with deuterium uptake. To achieve spatial resolution (i.e. ‘local’ measurements), the HDX process may be quenched using low pH conditions (the minimum rate for HDX occurs at a pH of ~ 2.6), followed by enzymatic digestion by an acid protease and analysis of differentially labeled peptides. This is known as a ‘bottom-up’ workflow. Alternatively, non-ergodic fragmentation techniques such as electron capture dissociation may be used to fragment the labeled protein in the gas phase (known as a ‘top-down’ workflow). Either of these approaches achieves spatial resolution down to almost the single amino acid level [29, 30].
Time-resolved electrospray mass spectrometry with HDX
‘Time-resolved’ experiments employ rapid mixing set-ups to access short-timescale processes in (bio)chemical reactions, conventionally using optical detection. NMR is a particularly powerful means of measuring rapid conformational dynamics, with experimental approaches that include stopped-flow or quench-flow rapid mixing [42, 43] or equilibrium methods (e.g. CPMG relaxation dispersion) . The primary advantage of NMR-based measurements is that the data are (almost) always site-specific, which allows for semi-quantitative analyses of local structural stability. However, NMR has a number of inherent drawbacks that limit applicability to large proteins and dynamics on certain timescales. Lengthy data acquisition times in multi-dimensional NMR also represent a challenge in application to time-resolved measurements, particularly in the case of HDX at physiological pH, in which exchange processes often proceed to completion within a few seconds of exposure to D2O.
A number of devices have been developed to couple mass spectrometry with rapid mixing, including pulsed-flow, stopped-flow and quench-flow set-ups [45-47]. Wilson and Konermann  adapted continuous-flow rapid mixing for MS using a concentric capillary device with an automatically adjustable reaction chamber volume (Fig. 1A). Solutions are supplied to each of the capillaries at the inlet by infusion pumps, generating a continuous laminar flow. One of the key advantages of this apparatus is that the outer capillary may be used as the ion source, resulting in a minimum reaction volume of 0 and a dead time governed exclusively by the mixing volume (corresponding to ~ 8 nl). The approach requires a custom ‘front end’ (usually a straightforward modification of the commercial nanospray source) and application of a laminar flow-corrected analytical framework to extract accurate rate parameters from the data. A detailed treatment of the theory underlying capillary mixer data analysis is given in the original paper .
TRESI-MS with HDX provides a unique and powerful approach for performing continuous-flow HDX experiments outside the constrained time window accessible to studies employing the conventional LC set-up. The capacity of TRESI-MS/HDX to generate short (millisecond) labeling pulses confers the ability to analyze dynamics in weakly structured regions of proteins, such as molten globules, random coils and intrinsically disordered domains. The labeled products may be subsequently fragmented or digested to allow measurement of site-specific exchange [23, 30]. This is a significant improvement over conventional LC-based methods, which typically achieve minimum labeling times of 1 s at best, with a substantial delay between labeling and analysis . In the case of a ‘bottom-up’ approach, the outlet of the capillary mixer may be coupled to a microfluidic chip that handles subsequent sample processing (Fig. 1B). This includes introduction of a low pH solution for HDX quenching, and proteolysis in a pepsin-functionalized reaction chamber [23, 29].
Oxidative labeling is a complementary structural method that has also been linked to TRESI . This approach involves attachment of a (non-labile) covalent label to redox active amino acid side chains, eliminating the issue of back exchange, which may occur in HDX. In contrast to HDX, which is sensitive to hydrogen bonding and solvent access, oxidative labeling efficiency is linked specifically to solvent accessibility and is typically much lower resolution (as it can only be measured at redox active side chains). A thorough review of the topic has been published previously .
Global analysis of protein folding using TRESI-MS/HDX
Perhaps the most straightforward application of TRESI-MS/HDX is in the study of protein folding. The first such study was performed by Pan et al. , in which a pulse-labeling approach was used to identify a ‘hidden’ intermediate in the ubiquitin folding pathway that could only be detected by considering both the HDX and charge-state characteristics simultaneously. This folding intermediate had a ‘native-like’ most prevalent charge state of 5+, but acquired deuterium in a manner analogous to the unfolded species (Fig. 2A). As the folding reaction progressed, the 5+ peak transitioned from this intermediate to a species that acquired substantially less deuterium, indicating a more folded, native configuration. The authors were thereby able to propose a multi-step mechanism for ubiquitin folding (Fig. 2B), and to provide evidence for the common occurrence of folding intermediates, even for small proteins generally assumed to be ‘two-state’ folders.
TRESI-MS/HDX in the study of catalysis-linked dynamics in enzymes
The microfluidics-based rapid mixing device may be successfully applied to the study of conformational fluctuations of proteins under non-equilibrium conditions in the context of TRESI-MS/HDX experiments. This method permits conformational dynamics to be probed in a system that is inaccessible to study by alternative methods, such as relaxation-dispersion NMR on the millisecond timescale. Liuni et al.  utilized the rapid mixer to study conformational changes occurring during catalytic turnover in chymotrypsin. The rapid mixing device was used to generate sub-second pulses of HDX labeling in the pre-steady state of chymotrypsin-catalyzed hydrolysis of para-nitrophenyl acetate. In this experimental set-up, D2O and para-nitrophenyl acetate substrate were supplied simultaneously into the mixing chamber of the device. TRESI-MS permitted simultaneous detection and quantification of the free enzyme and an acyl–enzyme intermediate, while also allowing quantification of the rate and amplitude of global HDX that occurred in each of the species .
Significantly higher rates of deuterium uptake were detected in the acyl–enzyme intermediate compared to the free enzyme, suggesting that the acyl–enzyme intermediate experienced higher rates of conformational sampling, but the conformational space sampled by the enzyme and the acyl–enzyme intermediate was similar, as reflected by the comparable amplitudes of HDX observed between the two species (Fig. 3). In this instance, application of TRESI-MS/HDX yielded new mechanistic insights into the understanding of catalysis-linked dynamics, and provided evidence for a new model of conformer selection followed by intensified conformational searching during catalysis. An inherent advantage of MS-based methodology in kinetic and dynamic studies of catalysis lies in its applicability to nearly any enzyme–substrate system, without having to use modified or unnatural substrates.
Studying protein dynamics in weakly structured regions of intrinsically disordered proteins
A greater degree of characterization of conformational dynamics of proteins may be achieved by utilizing microfluidic device-based TRESI-MS/HDX to examine site-specific HDX rates. In order to accomplish this, a ‘bottom-up’ TRESI-MS/HDX workflow was implemented, which included an HDX quenching step and subsequent protein digestion, prior to ESI-MS. In the device introduced by Rob et al. , a capillary-based rapid mixer was incorporated onto a polymethyl methacrylate-based microfluidic chip that was etched with additional supply channels for a quenching solution and a microreactor chamber for enzymatic protein digestion (Fig. 1B). Aqueous acetic acid (4% v/v, pH 2.3) was used as the HDX quenching solution, rapidly decreasing the rate of exchange and back exchange to negligible levels. The microreactor chamber for protein digestion was filled with pepsin/agarose cross-linked beads immobilized onto the polymethyl methacrylate surface . Pepsin's broad proteolytic specificity generally provides good sequence coverage, but the experimental set-up is not restricted to the use of pepsin, and may utilize other peptidases that are active under acidic pH.
This integrated experimental set-up was validated by application to a number of model systems, specifically highlighting the utility of sub-second HDX labeling in examination of dynamics in weakly structured regions of protein. Utilizing a single 100 ms HDX pulse, Rob et al.  were able to derive a crude secondary structure profile of ubiquitin by analyzing deuterium uptake in the pepsin-generated peptide fragments. Similarly, by fitting the kinetics of segment-specific HDX of cytochrome c, the relative conformational flexibilities of loop regions were determined (Fig. 4). Loop flexibility was expressed as segment-averaged ‘protection factors’: a ratio of the calculated ‘intrinsic’ rate of exchange to the observed rate of exchange (which is attenuated by structure) . Experimentally determined protection factors were in good agreement with previously derived or predicted properties of the specific loops, and generally agreed with predictions based on the heme prosthetic group negatively affecting the dynamics of the regions it contacts, with those segments in closest contact with the heme group showing the least flexibility.
On the millisecond timescale, backbone amide HDX in strongly hydrogen-bonded secondary structures is negligible. It was therefore possible to judge the extent to which the solution structure deviates from the reported (crystal or NMR) structure by comparing the observed level of exchange with the level expected based on the number of loop residues in the peptide.
In yet another application of microfluidics-enabled ‘bottom up’ HDX, Rob et al.  observed a large, rapid change in conformational flexibility associated with substrate binding in the large tetrameric enzyme, 3-deoxy-d-arabino-heptulosonate-7-phosphate (DAHP) synthase . By comparing site-specific HDX kinetics of the free enzyme with the substrate-bound complex, the authors localized the changes in dynamics to a large region centered on the active site. However, the tetramer interface remained largely unchanged in the presence and absence of substrate, and this was supported by detection of primarily tetrameric DAHP synthase in the ‘native’ mass spectrum. Site-specific HDX analyses of DAHP synthase dynamics contributed an important mechanistic insight, suggesting that the residual tetramer structure in the molten globule-like substrate-free form may act as a template for the development of structure upon substrate binding.
TRESI-MS and intrinsically disordered proteins
One of the most exciting features of TRESI-MS/HDX is the ability to characterize residual structure in intrinsically disordered proteins . An example of this is a structural analysis of the neuronal Tau protein, which forms neurofibrillary tangles in Alzheimer's disease. By examining the disordered regions of the Tau protein under native (non-amyloidogenic) and hyperphosphorylatyed (amyloidogenic) conditions, regions that undergo substantial structural rearrangement during development of pathogenicity may be identified and characterized . Phosphorylation of the native Tau protein by glycogen synthase kinase 3β in vitro yields an amyloidogenic paired helical filament form, and subsequent TRESI-MS/HDX analysis of site-specific exchange rates between the two forms revealed striking differences in the rates of exchange, concentrated in the disordered regions and regions that fold into β-sheet structures in the native protein (unpublished results). This experimental system yields structural and kinetic insights that may greatly facilitate the development of new Tau aggregation inhibitors, as well as providing a platform for direct chemical library screens aimed at identifying new inhibitors.
Site-specific HDX measurements on the millisecond timescale enabled by microfluidics-coupled TRESI-MS are proving to be a versatile method for studying conformational dynamics in weakly structured protein regions and electron capture dissociations. This approach is useful in an array of applications, from high-throughput predictions of secondary structure and dynamics to characterization of weak hydrogen-bonding networks in flexible protein segments.
Conclusions and future directions
Coupling of ESI-MS to a microfluidic device that enables rapid mixing and provides an adjustable reaction chamber volume extends the reach of HDX analyses to weakly structured regions of proteins. The microfluidic device allows continuous-flow experiments to be performed, which greatly broadens the applicability of this methodology. In the study of enzyme kinetics, this device enables pre-steady state measurements to be performed with ease, offers high selectivity for reaction intermediates, and permits the study of natural substrates and native enzyme–substrate intermediates. In the study of protein dynamics, microfluidics-enabled TRESI-MS/HDX is capable of characterizing conformational changes that occur rapidly and frequently – conditions that are largely inaccessible to conventional techniques – with a spatial resolution of five amino acid residues on average.
Currently, site-specific HDX is limited to the analysis of a single structural species or an ensemble of species, which is the major limitation of this approach compared to the great deal of additional information it provides compared with global HDX experiments. This is because, in order to measure site-specific HDX rates of multiple resident structural states, it is necessary to map the detected peptide fragments to the originating species. Proteolytic ‘bottom-up’ work-up of HDX-labeled proteins is thus intrinsically limited to a study of a single structural species . Non-ergodic gas phase fragmentation techniques such as electron capture dissociation do not suffer from this limitation, and are thus potentially highly applicable to the study of multiple resident structural states by time-resolved, site-specific HDX [30, 54]. However, electron capture dissociation has its own limitations, particularly low sensitivity and a size limitation (presently ~ 18 kDa) .
Ultimately, TRESI-MS provides a powerful alternative for time-resolved studies of protein dynamics, revealing rapid, activity-linked conformational changes that occur in response to ligand binding, allosteric effects or post-translational modification. It is hoped that these new insights will translate into a much-improved understanding of protein function and the nature of pathogenic misfolding and aggregation in the cell.