Ion mobility spectrometry (IMS) is increasingly used as an added dimension in mass spectrometry analyses, providing gas-phase separations of ionized analytes based upon their collision cross sections [1]. IMS is now commonly used in structural biology for low-resolution analysis of protein structure, conformational dynamics, and complex topology [2]. Differential mobility spectrometry (DMS) is an IMS approach that has found a niche in ion pre-filtration of small molecules in particular [3, 4]. In DMS, ions are ‘pre-selected’ for transfer into the mass spectrometer based on the difference between their high-field and low-field mobilities ΔK. To do this, an asymmetric rf voltage (separation voltage) is applied perpendicular to the flow of ions, resulting in a net lateral displacement and an ‘unstable’ trajectory. For each ΔK, it is possible to apply a specific DC offset called the ‘compensation voltage” (CV) that will nullify lateral displacement, allowing a stable trajectory through the DMS instrument [3, 5].

One characteristic of this approach that is distinct from conventional drift tube or traveling-wave IMS is that the separation parameter ΔK is determined by a host of factors rather than principally by collision cross section. In particular, interactions with transport gases within the DMS instrument can have a substantial impact on ΔK. This can be used to great advantage, particularly for small molecules, where optimization of the transport gas by addition of volatile chemical modifiers can allow for high-resolution separation even of species with identical collision cross-sections [69]. On the other hand, these ‘clustering’ effects and others associated with the high field, high temperature environment within the DMS apparatus make quantitative determinations (of collision cross-section, for example) challenging even for relatively small molecules and effectively impossible for macromolecules. For large proteins with significant molecular dipoles, a particular problem is field-induced alignment, which results in bimodal CV profiles for unimodal structural distributions [10, 11]. As a result, there has been little effort to apply DMS to large biomolecules [12] and no success in linking gas-phase structural insights acquired by DMS to the liquid-phase conformational ensembles of proteins [1315].

Gas-phase HDX is a structure-dependent labeling technique in which exposed and relatively acidic protons (e.g., hydroxyl, thiol, amino, and amido hydrogens) undergo exchange with deuterons from surrounding gas [16]. In contrast to liquid-phase HDX, which is principally a base-catalyzed process, the reaction mechanisms underlying gas-phase exchange are less well understood [17, 18]. Nonetheless, gas-phase HDX is viewed as a sensitive, qualitative probe of gas-phase protein structure [19], and it has been combined with various ion mobility techniques, including commercially available traveling wave instruments [20]. It has been recognized previously that gas-phase HDX and FAIMS could be combined to provide an orthogonal characterization of gas-phase protein structure; however, the resulting profiles were uninterpretable in the context of solution structure [14, 21]. In particular, the profiles indicated a multitude of co-populated conformations with substantially different CVs even for proteins with a well-defined native fold in solution, suggesting that much of the conformational heterogeneity was generated during FAIMS separation [14, 22]. Thus, FAIMS-HDX was demonstrated as a useful tool for investigating gas-phase conformational heterogeneity, but was also shown to have poor correlation to liquid-phase structure.

Separation conditions in DMS are similar enough to those of FAIMS that one could predict a similar degree of solution structure preservation (i.e., little to none); however, liquid-phase structural information might still be retained if the set of conformations that become populated during DMS depend on the solution structure. Here we apply DMS-HDX to a set of model proteins in order to determine if experimental conditions can be found under which liquid-phase structure is reflected in DMS-HDX profiles. Proteins were selected to exhibit the full-range of solution-structure characteristics, including native Tau (a 46 kDa intrinsically disordered protein [23]), hyperphosphorylated Tau (partially disordered [24]), native holo-myoglobin (17 kDa), and cytochrome c (12 kDa) in their solution-folded and solution-unfolded states. These proteins were investigated under a range of instrumental conditions using a planar DMS-TOF instrument modified to allow addition of HDX reagents in the modifier gas and/or in the throttle gas (at the DMS/MS interface). Ultimately, conditions were found under which DMS-HDX provides a reliable assay for global structural properties of intact proteins in solution based on the relationship between CV and deuterium uptake during transit through the DMS.



Deuterated ammonium hydroxide (ND4OD), deuterium oxide (D2O), deuterated methanol (CH3OD), methanol (CH3OH), ammonium hydroxide (NHzOH), isopropyl β-D-1-thiogalactopyranoside (IPTG), 2-(N-morpholino)ethanesulfonic acid (MES), magnesium chloride (MgCl2), dithiothreitol (DTT), phenylmethanesulfonylfluoride (PMSF), 10× halt protease inhibitor cocktail, and sodium chloride (NaCl) were purchased from Fisher Scientific (Waltham, MA, USA). Horse heart cytochrome c was purchased from Sigma Aldrich (St. Louis, MO, USA). Glycogen synthase kinase 3β (GSK3β), adenine triphosphoate (ATP) were purchased from New England Biolabs (Ipswitch, MA, USA). The pET-29b/tau plasmid was purchased from Addgene (Cambridge, MA, USA).

Protein Preparation and Phosphorylation

Tau was purified as previously described using strong cation exchange on AKTA FPLC purifier [25]. Briefly, pET-29b plasmid containing full length htau40 isoform was transformed into E. coli BL21 for protein expression. Bacteria were grown at 37 °C for 3 h to an optical density of 0.5–0.7 at 600 nm absorbance. Subsequently, the cultures were induced with IPTG for an additional 3.5 h for protein overexpression. Cells were then pelleted and resuspended in resuspension buffer: 20 mM MES, 0.2 mM MgCl2, 5 mM DTT, 1 mM PMSF, and 1× of halt protease inhibitor cocktail (1 mM AEBSF• HCl, 80 nM aprotinin, 5 μM bestatin, 1.5 μM E-64, 2 μM leupeptin, 1 μM pepstatin A), pH 6.8. Lysates were sonicated on ice for 20 min (15s on/30 s off) followed by incubating in boiling water bath for 20 min. The boiled lysates were then pelleted at 40,000 × g for 1 h at 4 °C. The supernatant containing tau was dialyzed three times for a day against cation exchange loading buffer (Buffer A): 20 mM MES, 50 mM NaC1,1 mM MgC12, 2 mM DTT, 1 mM PMSF, pH 6.8. The buffer exchanged tau was loaded onto the purifier and purified using strong cation exchange column: SP sepharose FF. Unbound proteins were washed out with at least 5 column volumes (CV) of Buffer A and tau protein was eluted using cation exchange elution buffer (Buffer B): 20 mM MES, 1 M NaC1, 1 mM MgC12, 2 mM DTT, 1 mM PMSF, pH 6.8 with step gradient of 10% in 2 CV, 15% in 2 CV, 20% in 3 CV, 25% in 3 CV, 30% in 3 CV, 35% in 3 CV, 40% in 3 CV, and 100% in 6 CV. The purified tau was then pooled and concentrated by 10 kDa MWCO vivaspin to final volume of 3 mL. The protein concentration was determined by bicinchoninic acid (BCA) assay. The protein stock was then aliquoted and stored at – 80 °C.

Tau phosphorylation was carried out by a procedure described previously with slight modification [26]. One hundred μM tau was incubated with 500 units of GSK3β in the presence of 3 mM ATP at 30 °C for 30 h. The concentrated tau was buffer exchanged in 50 mM ammonium acetate overnight and was further filtered through 10 kDa MWCO vivaspin concentrators to remove peptides from protein degradation overnight. Native cytochrome c was prepared by dissolving commercially supplied, salt-free lyophilized powder in ddH2O containing 50 mM NH4Ac, pH 6.8. Unfolded cytochrome c was prepared by dissolving lyophilized protein (as above) into ddH2O with 1% acetic acid (pH 2.6). Final concentrations were 20 μM for tau/phosphotau and 5 μM for cytochrome c.

DMS Conditions

DMS experiments were performed on a quadrupole time-of-flight mass spectrometer (specifications similar to reference [27]) equipped with a planar DMS cell that was incorporated between the mass spectrometer’s sampling orifice and the electrospray source (Figure 1) [3]. Various modifier gases including HDX reagents were injected into the transport gas via the ‘curtain gas’ line (total flow 25 L/min with nitrogen) and/or supplied through the ‘throttle gas’ line via bubbler. ESI probe voltage was maintained at 4800 V with source temperature maintained at 37 °C. Nebulizer and auxiliary gas pressure were maintained at 10 and 80 psi, respectively, for tau and phosphotau, whereas for cytochrome c it was 10 and 30, respectively. DMS temperature was maintained at 275 °C. A SV of 3800 V was maintained throughout the experiments for all proteins and CV was scanned in 0.03 V increments. All samples were infused at 7 uL/min.

Figure 1
figure 1

Schematic cross-section of the DMS cell, depicting the relevant components and gas flows. Adapted with permission from reference [3]


To perform gas-phase HDX after separation with DMS, D2O or ND4OD was introduced in the resolving (throttle) gas line by bubbling nitrogen through a volume of either deuterating reagent at the interface between the DMS and TOF-MS [28]. In experiments where HDX during transit through the DMS cell was desired, 1:1 ND4OD:CH3OD (generated via bubbler) was infused at 1.5% (v/v) in the curtain gas as a chemical modifier together with nitrogen [28]. For HDX in the DMS cell, it was also important to qualify the degree of mobility changes caused by ion/molecule clustering between the protein ions and HDX molecules. The degree of such clustering was evaluated by conducting the same experiment using non-deuterated modifiers, and comparing any mobility changes when the HDX analogues were employed. In these experiments, all other instrumental conditions were held constant at levels designed to achieve maximum DMS separation power (see DMS Conditions above).

Data Analysis

Data were analyzed using PeakView ver. 2.1. The mass spectra were averaged over windows of 1.0 CV for tau and phosphotau and 0.5 CV were averaged for cytochrome c. For example, for tau CV = 6 V is the average of 5.5 to 6.5 V, whereas for cytochrome c, CV of 10.25 V is the average of 10 to 10.5 V. All of the data were smoothed for better data analysis and reliability. The centroid of the protein peaks were chosen and were subtracted from nondeuterated counterpart for HDX uptake. Mass spectra with low intensity or poor quality were omitted from the analysis. The error bars represent one standard deviation determined from triplicate measurements.

Results and Discussion

DMS profiles of Tau and Hyperphosphorylated Tau

Differential mobility profiles for proteins with well-defined native structures, such as cytochrome c and myoglobin, have been obtained previously using FAIMS [10, 14]. However, Tau represents a new and particularly interesting analyte as it is an intrinsically disordered protein (IDP) with no well-defined secondary or tertiary structural elements under native conditions [29, 30]. In the context of the current study, Tau provides a model for an ‘unfolded’ protein that can be examined under identical solvent conditions as proteins with well-defined native states. Figure 2a depicts a DMS-MS spectrum of native Tau. The unfolded character of Tau is illustrated here both by the broad, high charge distribution and the exceptionally broad DMS profiles associated with each peak. Pendular alignment effects [10, 12], which are expected for a 46 kDa protein, are observed in the bimodal appearance of the CV profile (Figure 2a, top panel). All proteins including Tau were subjected to extensive optimization of DMS separation, principally in the selection of modifier gas (Supplementary Figure S1). Ultimately, it was determined that the best modifier gas for DMS separation that could double as an HDX reagent was 1:1 NH4OH:CH3OH.

Figure 2
figure 2

Heatmap DMS-MS profiles of Tau protein and hyperphosphorylated Tau protein. (a) The DMS-MS heatmap profile of native Tau, with the m/z-combined DMS profile shown above and the DMS-combined m/z profile on the right. Inset: The high-charge, low CV region where changes in the DMS profile are observed upon hyperphosphorylation by GSK-3β. (b) The DMS-MS heatmap profile of hyperphosphorylated Tau, with m/z-averaged DMS profile shown above and the DMS-combined m/z profile shown on the right. Inset: The high-charge, low CV region of the plot, which exhibits peak splitting upon hyperphosphorylation of Tau. Black arrows indicate the centroids of the two CV peaks on the heatmap

Hyperphosphorylation of Tau by kinase GSK-3β induces a pathogenic shift in its conformational ensemble that involves a general ‘lengthening’ of the protein and the development of several pseudo-stable structural nodes [24]. Treatment of Tau with GSK-3β/ATP for 30 h results in an ensemble of phosphorylation states, normally distributed around a mean of 8 with σ = 1. In the context of the current study, hyperphosphorylated Tau serves as a model for a partially-folded protein that can be examined under ‘native’ solvent conditions. Figure 2b represents the full DMS-MS profile for hyperphosphorylated Tau. Aalthough broadly similar to the native profile, the charge distribution has shifted to favor lower charge states and a few new features have emerged in the ionogram, particularly in the high-charge/low CV region (Figure 2, insets). Splitting of the low CV peak is consistent with the appearance of new structural features. However, it is equally possible that the changes in the DMS spectrum are purely a consequence of differing clustering behavior resulting from chemical modification (phosphorylation) of the protein. This degree of open interpretation is illustrative of the challenge of drawing conclusions about protein structure (gas phase or liquid phase) based on CV profiles alone. With the inclusion of gas-phase HDX, however, DMS separation can be linked to information that is specifically structure-dependent.

Optimization of DMS-HDX Parameters

In the planar DMS setup shown in Figure 1, a transport gas (typically N2) containing the modifier gas mix is introduced at the DMS inlet [31, 32]. A second ‘throttle’ gas can be introduced at the interface between the DMS and the mass spectrometer for resolution enhancement [3, 28]. In principle, both of these inlets could be used to provide HDX reagents to the ion beam. Adding HDX reagents to the transport gas would allow the labeling reaction to occur during DMS separation (i.e., ‘in transit’), whereas injecting HDX reagents into the throttle gas would ‘pulse label’ species that are exiting the DMS. Initially, implementation of HDX in the throttle gas seemed the most appropriate approach, since it would provide structural insights into species exiting the DMS at particular CVs without the convolved effects of simultaneous ion/molecule clustering and gas-phase HDX during separation. To achieve this, ND4OD or D2O vapor were added through the throttle gas line via a bubbler. As expected, the more basic ND4OD was by far the more reactive HDX reagent [33], providing an average uptake of 70 deuterium on 6+ ‘native’ cytochrome c compared with 15 for D2O, and ND4OD was therefore selected as the labelling reagent.

However, the HDX versus CV profiles that were generated by this approach exhibited a poor correlation between CV and deuterium uptake, and the profiles of adjacent charge states were wildly different (Figure 3a). These observations are consistent with the findings of a handful of studies that combine FAIMS separation with HDX, in which no reliable correlation between gas phase HDX and CV-delineated gas-phase conformations was observed [14, 21, 22]. However, more importantly for our purpose, there appeared to be no connection between throttle gas DMS-HDX profiles and the solution structure of the protein; unfolded, partially folded. and ‘native’ proteins exhibited equally uninterpretable profiles (Supplementary Figure S2). Thus, it is clear that gas-phase conformations exiting the DMS retain no information about their solution structure origins, neither directly (as retention of some elements of liquid-phase structure) nor indirectly (e.g., as an impact on the set of gas-phase conformations that are adopted). Consequently, in our setup, throttle gas HDX cannot be used as a probe of liquid-phase protein structure.

Figure 3
figure 3

Mass spectra of cytochrome c with DMS-HDX profiles for dominant charge-states. (a) Solution-folded cytochrome c with throttle gas HDX. Insets: DMS-HDX profiles (filled circles) were spline fit (solid line) due to a lack of consistent correlation between HDX and CV. (b) Solution-folded cytochrome c with transport gas HDX. Insets: DMS-HDX profiles (filled circles) were fit to a simple linear expression (solid line). (c) Solution-unfolded cytochrome c with transport gas HDX. Insets: DMS-HDX profiles (filled circles) were fit to a simple linear expression (solid line)

As an alternative approach, a 1:1 mixture of ND4OD and MeOD was added at 1.5% (v/v) to the transport gas line as both HDX reagent and modifier gas. With this approach, HDX occurs over a period of roughly 16 ms (at the transport gas flow-rate used) as ions transit through the DMS, considerably longer than the sub-ms ‘pulse’ provided by throttle-gas labeling. The result was a slight decrease in overall charge, a substantial increase in deuterium uptake (average 110 deuterium on 6+ cytochrome c), and an HDX versus CV profile showing a positive linear correlation between CV and uptake (Figure 3b). Fitting the data to a linear expression yielded slopes of 1.1 × 10−2, 1.0 × 10−2, and 1.0 × 10−2 for the 5+, 6+, and 7+ charge states, respectively (R2 min = 0.95).

These profiles represent a ‘mixed’ result in terms of their relationship to solution-folded cytochrome c. On the one hand, the wide range of CVs over which cytochrome c ions are transmitted indicates that a relatively broad set of gas-phase conformations have been populated. Thus, the protein has not retained a well-defined, ‘solution-like’ fold during DMS. On the other hand, and in contrast to the FAIMS and the throttle-gas DMS-HDX profiles described earlier, these HDX versus CV profiles are reproducible between charge states and are interpretable, at least in the sense that conformers with higher CV values acquire more deuterium. Similarly ‘sloped’ profiles were measured for solution-folded myoglobin (Supplementary Figure S3). Should this positive correlation between CV and deuterium uptake be unique to solution-folded proteins, then planar DMS with transport-gas HDX could provide at least a coarse-grained probe of protein structure in solution.

Solution Structure Analysis by DMS-HDX

To determine if DMS with transport gas HDX could reproducibly generate unique profiles for folded, unfolded, and partially folded proteins, we acquired transport gas DMS-HDX data for a set of proteins with varying structural characteristics. A tabular overview of the model proteins used for this study and their HDX characteristics is provided in Table 1.

Table 1 An Overview of the Proteins Used in This Study and Their HDX Characteristics

The profiles shown in Figure 3b were acquired under ‘native-like’ solvent conditions (50 mM NH4Ac, pH 6.8) from a charge state distribution similar to that normally observed for the folded protein [34] (though somewhat lower charge on average due to charge stripping – see below). When electrosprayed from denaturing solvent (pH 2.4 using acetic acid), cytochrome c typically generates a broad, high charge state distribution centered on ~14+ [34]. However, in our DMS experiments, the solution-unfolded protein exhibited extensive charge stripping due to the basic ND4OD modifier gas, resulting in a charge state distribution centered on 5+. Charge stripping was observed for all proteins when the modifier gas included ND4OD, but was much more pronounced for solution-unfolded species.

In spite of the similar charge-state distribution, solution-unfolded cytochrome c (Figure 3c) generates substantially different DMS-HDX profiles compared with solution-folded cytochrome c. Specifically, for all dominant charge states, the solution-unfolded profiles are essentially flat, with slopes of 1.0 × 10−3 (5+), 1.2 × 10−3 (6+), and 3.5 × 10−3 (7+), R2 min = 0.92. A reasonable physical interpretation of these results is that while solution-folded proteins populate a set of gas-phase conformers that are HDX-distinct, the set of gas-phase conformers populated by solution-unfolded proteins as they traverse the DMS are essentially HDX equivalent. Thus, solution conformation, while not necessarily retained through DMS separation, controls the set of gas-phase conformers that become populated. This explanation is consistent with the observations of Wright and co-workers, who reported that proteins retained a ‘memory’ of their original solution structures through substantial gas-phase manipulations associated with kinetic energy loss cross-sectional measurements [35].

Although the solution-folded and solution-unfolded DMS-HDX profiles are clearly different, it could be argued that the profiles are an artifact from ionization under substantially different solvent conditions. To explore this, we implemented transport gas DMS-HDX on Tau, a protein that is intrinsically disordered under native-like solvent conditions. Here as well, under solvent conditions identical to those used to generate ‘native’ (sloped) cytochrome c and myoglobin profiles, the Tau DMS-HDX profile was essentially flat (Figure 4a), suggesting that the ‘flat’ profile is characteristic of solution-unfolded proteins and not an artifact of ionization conditions. As proposed in recent work from Beveridge and co-workers [36], the set of gas phase structures adopted during IMS may depend heavily on the ESI mechanism, with charged residue ionization of folded proteins producing a relatively compact set of conformations and chain ejection of unfolded proteins favoring an elongated structural ensemble (and thus HDX equivalence) [37, 38].

Figure 4
figure 4

Mass spectra of tau with DMS-HDX profiles from selected charge-states. (a) Intrinsically disordered tau under native solvent conditions. Insets: DMS-HDX profiles were fit to a simple linear expression (solid line). (b) Partially-ordered, hyperphosphorylated tau under native solvent conditions. Insets: DMS-HDX profiles were split at the inflection point (−10 CV, determined by maximizing R2 over the global data set) and fit using two linear expressions (solid lines)

In order to investigate the DMS-HDX behavior of a partially-folded protein under native solution conditions, we used GSK-3β hyperphosphorylated Tau. Previous work has shown that hyperphosphorylated Tau is substantially more structurally constrained compared with native Tau, with a mix of disordered regions and small structural nodes [24, 39]. When this form of Tau is subjected to ‘in-transit’ DMS-HDX analysis, the profiles have a ‘biphasic’ appearance (Figure 4b). This cannot be a saturation effect, as the maximum number of exchanges at 635 is well below the theoretical maximum of 760. Moreover, saturation was not observed for myoglobin, which exhibited a much higher deuterium uptake relative to its theoretical maximum (see Table 1). Thus, these ‘biphasic’ profiles indicate a mixture of exchange equivalent and nonequivalent states, with nonequivalent states populating the CV = −20 to −10 region (observed as a positive slope), and largely equivalent states populating compensation voltages above CV = −10 (observed as a comparably flat region). Should these mixed profiles be characteristic of partially folded species, one could envisage a more fine-grained measure of liquid-phase structural stability for sets of similar proteins (e.g., for point mutants or different post-translational modifications) based on the relative contributions of exchange-equivalent and nonequivalent states.


DMS is an IMS technology with underexplored potential for intact protein studies. In this work, we applied a planar DMS-HDX approach to characterize the liquid-phase conformational states of folded, unfolded, and partially folded proteins. Our results support the notion that liquid-phase structural information can be preserved through DMS separation (even if actual solution structures are lost), and lay out experimental conditions under which liquid-phase structure characterization can and cannot be achieved in a planar DMS system. Although the structural insights obtained in the current study are coarse-grained in the sense that they distinguish only large, global folding characteristics, this work provides a foundation for more fine-tuned analyses aimed at rapid measurements of protein stability. For instance, the occurrence of ‘mixed’ profiles like those observed for hyperphosphorylated Tau could indicate the presence of partially-folded states that are associated with aggregation. Fine measurements of relative conformational stability in closely related native proteins (e.g., for point mutants, or differently post-translationally modified species) may also be possible using the HDX versus CV slope. Ultimately, such an assay could prove to be an invaluable tool for biopharmaceuticals analysis, among other applications.