H. Jörnvall, Division of Physiological Chemistry I, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, S-17177 Stockholm, Sweden Fax: +46 8 33 74 62 Tel: +46 8 524 87702 E-mail: Hans.Jornvall@ki.se
The study of protein structure and function has evolved to become a leading discipline in the biophysical sciences. Although it is not yet possible to determine 3D protein structures from MS data alone, multiple MS-based techniques can be combined to obtain structural and functional data that are complementary to classical protein structure information obtained from NMR or X-ray crystallography. Monitoring gas-phase interactions of noncovalent complexes yields information on binding constants, complex stability, and the nature of interactions. Ion mobility MS and chemical crosslinking strategies can be applied to probe the architecture of macromolecular assemblies and protein–ligand complexes. MS analysis of hydrogen–deuterium exchange can be used to determine the localization of secondary structure elements, binding sites and conformational dynamics of proteins in solution. This minireview focuses first on new strategies that combine these techniques to gain insights into protein structure and function. Using one such strategy, we then demonstrate how a novel hydrogen–deuterium exchange microfluidics tool can be used online with an ESI mass spectrometer to monitor regional accessibility in a peptide, as exemplified with amyloid-β peptide 1–40.
With the advent of (soft) ionization techniques, MS has evolved as a tool for studying proteins in their folded state. Specifically, the development of nondenaturing ESI-MS in 1989  led to the observation that protein complexes could be analyzed without the destruction of noncovalent interactions [2,3]. Although it has been long disputed whether the absence of solvent after the transfer into the gas phase has an impact on protein structure, recent studies indicate that proteins can retain their native fold for a limited time span during MS analysis before adopting a gas-phase conformation .
Steinberg et al.  investigated the behavior of cytochrome c during desolvation, using molecular dynamics simulations, and reported that, whereas side chain interactions were lost almost immediately, the backbone was able to retain its fold on a nanosecond timescale after solvent removal. This set of data is supported by electron capture dissociation (ECD) studies of cytochrome c under native conditions . Additionally, Badman et al.  compared the drift times of cytochrome c molecules in an ion mobility MS instrument after ion trapping from 10 ms to 10 s. The data indicate that, up to a trap time of 60 ms, only one compact conformer can be observed, whose cross-section corresponds to that calculated for the native cytochrome c structure adjusted for minor unfolding events. For trap times > 60 ms, a single compact population disappears, and a total of four populations with different drift times are observed, representing four distinct folding states that appear to be elongated and partially unfolded structures. Recently, Badman et al. have extended this approach, and compared the drift times of cytochrome c ions at different injection voltages and ion–ion reactions in a 3D trap . Both approaches were found to alter ion cross-sections, ranging from tightly folded (low injection voltage and charge states reduced by ion–ion reactions) to more unfolded (high injection voltage and higher charge states) conformations.
A different approach to the investigation of protein structure after desolvation was presented by Breuker et al.,  who subjected ionized proteins trapped for variable times to ECD. The effect of trapping time on the relative stability of the protein backbone was then determined by analysis of the c-ion and z′-ion yields from ECD of the protein at different charge states. Even at long trapping times (≥ 4 s), the results show a striking correlation between ECD-resistant segments and the location of helices, salt bridges and hydrogen bonds in the solution structure of the protein. Higher charge states exhibited more complete ECD fragmentation, indicating gradual unfolding with increasing precursor ion charge. This indicates that electrostatic interactions can compensate for the relative weakening of hydrophobic interactions after desolvation into the gas phase, and allow proteins to retain a native-like fold.
The observation that proteins can retain a near-native fold even in the gas phase has led to the concept that protein–ligand interactions can be studied with nondenaturing ESI-MS. Evidence of intact protein–ligand complexes was found in 1991 [2,3]. However, these reports have led to discussions of whether the observed complexes represent natural interactions under solution conditions [9,10]. On the other hand, the binding preferences of proteins interacting with hydrophobic ligands have successfully been determined with ESI-MS [11,12].
Three types of information have been extracted from ESI-MS-based studies of protein complexes: binding energies, complex stoichiometries, and complex architectures.
However, a significant body of work also indicates that gas-phase binding energies determined from ESI-MS experiments do not always reflect complex behavior in solution . Direct quantification of protein–ligand complexes has proven difficult, because of the in-source decay during desolvation, which artificially lowers the yield of complexes observed in ESI-MS. The group of Klassen has addressed this problem by developing a competition-based assay with a specific interaction partner as a reference . The reference ligand forms a stable complex with the protein that can be readily detected with ESI-MS. The relative abundance of protein–reference ligand complexes allows indirect determination of the binding affinity of the competing ligand, even if this complex is not well preserved in the gas phase. This approach, however, requires detailed knowledge of reference binding partners.
A different approach to detect protein–ligand interactions by MS was taken by Przybylski et al., who developed an online bioaffinity instrument combining surface acoustic wave (SAW) technology with ESI-MS . Briefly, a protein is immobilized on a SAW chip, over which a mixture of potential ligands is pumped. The outlet of a SAW instrument is connected to an ESI-MS instrument via an interface for desalting and concentration. With this setup, binding partners can be ‘fished out’ from the ligand mixture and subjected to identification and direct affinity determination on the basis of the SAW binding curves obtained with standard ESI-MS technology.
Architecture of protein complexes
Until recently, the characterization of protein complex architecture by MS was largely focused on determination of stoichiometries. The groups of Robinson and Benesch have successfully used time-resolved and temperature-controlled tandem-MS to determine the variable stoichiometries of small heat shock protein (sHSP) complexes . It was observed that, with increasing temperatures, sHSP assemblies shift from a dodecameric state to monomers and dimers, as well as large assemblies containing from 13 to 20 monomers. In the presence of an unfolded ‘client’ protein, protein–sHSP complexes with diverse stoichiometries are found that could represent the key to a wide range of chaperone activities.
Another MS-based technology for probing the architecture of protein complexes utilizes ion mobility MS. Here, the drift time of a molecule through a chamber filled with an inert buffer gas is used as a measure of the cross-sectional size of the molecule, which is added to intensity and m/z as a third dimension. Cross-sectional size has successfully been employed as a constraint in modeling of the quaternary structure of large protein assemblies [16,17]. To obtain a yet more detailed picture, Robinson et al. have extended this approach to integrate ion mobility MS data with electron microscopy data . In their study using ion mobility cross-sectional measurements as constraints, topological models of macromolecular GroEL assemblies were fitted into low-resolution structures obtained from electron microscopy.
However, MS of native complexes, just like any other technique, has limitations. Even minor sample heterogeneity can influence interpretations and conclusions about protein complex stoichiometry significantly: in a model study with bacteriophage genome packaging proteins, it was shown that the presence of a truncated version of one of the components gave rise to protein assemblies with multiple compositions that were not representative of the native complex .
To avoid the limitations caused by altered protein structure in the gas phase, solution-phase crosslinking has emerged as a tool for MS analysis of native protein and complex structures, as presented by Leitner et al. . Their approach employs crosslinkers labeled with stable isotopes. After crosslinking, the protein is subjected to enzymatic digestion and LC-MS. The resulting crosslinked peptides are of low abundance, but can be identified by their unique isotope pattern with the xquest search engine and reference spectra of unlabeled material . The combination of crosslinkers with different lengths allows the generation of distance constraints between putative crosslinking sites that can be employed in molecular modeling approaches for protein assemblies as large as the proteasome [22,23].
Hydrogen–deuterium exchange (HDX) MS (HDX-MS)
Protein structure dynamics monitored by HDX
HDX combined with MS measures the replacement of protein backbone amide hydrogens with deuterium, the rate of which is dependent on the strength of hydrogen bonding and solvent exposure, yielding information on local protein structure and dynamics. For a review on the history and underlying principles of HDX-MS, see . Theoretical and methodological considerations are described in . By combination of peptic digestion under quenching conditions and HPLC with MS analysis, information on local deuterium exchange rates can be obtained.
Our group has successfully applied a combination of ESI-MS and HDX-MS to monitor structural changes in the N-terminal domain of major ampullate spidroin 1 (MaSp1 NT), a regulatory domain from spider silk , which regulates silk assembly in a pH-dependent manner . Using nondenaturing ESI-MS, we found that pH-dependent structural changes cause dimerization driven by electrostatic forces, and HDX-MS revealed a global tightening of the protein fold and protection of peptides at the dimer interface from deuterium exchange (Fig. 1). These processes are critically dependent on conserved protonatable residues at the dimer interface. The results were subsequently confirmed by the report of an NMR structure , as well as by fluorescence spectroscopy and pull-down studies , illustrating the power of combined MS-based approaches.
Top-down analysis of HDX
The resolution of HDX-MS is limited by the efficiency of the peptic digestion and separation . However, there is also a need for greater resolution in the HDX profiles. Collision-induced dissociation is not suitable for determination of backbone deuteration. By designing model peptides that are selectively deuterated at either the C-terminal or the N-terminal side, Rand et al.  showed that collision-induced dissociation of labeled peptides leads to complete loss of selective labeling, owing to collision-induced scrambling of the backbone amides. To address this problem, ECD and electron transfer dissociation have recently been employed to obtain deuteration profiles from reverse-phase purified peptic peptides at single-residue resolution  and from intact proteins by the use of continuous infusion in the labeling solution . Both approaches yield high-resolution HDX-MS data by stepwise sequencing of the labeled protein.
A novel microfluidic tool for online HDX to study protein folding
Conventional HDX approaches require coordinated pipetting involving both the labeling and quenching steps, followed by freezing of the protein sample. The labeling involves dilution of the protein into a deuterated buffer followed by further dilution with the quenching solution. Coordinate pipetting is a potential source of error, and sample dilution necessitates the use of concentrated protein solutions (usually in the range of ≥ 1 mg mL−1). To increase accuracy when short incubation times are used, specially adapted autosampler systems , as well as the previously mentioned flow-based HDX-MS system with online sample dilution in deuterated buffer , have been employed. To eliminate the need for protein dilution in a flow-based system, we have contributed to the development of a dual-channel microfluidic cell (HDX cell) for continuous sample labeling. This HDX cell has two channels, separated by a semipermeable, ion-selective membrane for online deuteration of proteins and peptides without dilution of the sample in deuterated solvent  (Fig. 2A). The deuterium content in the sample flow channel can be adjusted through the flow rate. This determines the degree of sample deuteration and can be adjusted from minimal addition of deuterium up to nearly complete exchange.
We have applied this online deuteration cell to monitor the aggregation states of peptides, and here we summarize the results obtained with amyloid-β peptide 1–40 (Aβ40), a 40-residue peptide that can serve as a model for peptide aggregation associated with Alzheimer’s disease. Aβ40 forms amyloid fibrils within hours when incubated at room temperature. Using the HDX cell setup, we found that fibrillated Aβ40 incorporated significantly less deuterium than freshly dissolved, nonaggregated Aβ40 (Fig. 2B), which is in agreement with previous studies . This demonstrates that the HDX cell can be applied to automated MS-based screening of Aβ40 aggregation. The HDX cell can also be combined with top-down fragmentation of labeled proteins to obtain single-residue-resolution HDX data without side chain deuteration or loss of cleavage site amides, an approach that is currently under investigation in our laboratory. The origin of our HDX cell concept is our previous development of an electrocapture device with conductive membranes, leading to protein concentration , purification  and successive online attachment  to the present routine tool for HDX applications. An online microfluidic digestion and electrospray chip was recently described by Wilson et al. . An interesting approach would be to combine the HDX cell with such additional components to create a complete microfluidic ‘lab-on-a-chip’ unit.
Protein structure analysis still relies heavily on NMR and X-ray crystallographic data. However, protein size and heterogeneous assemblies limit the availability of NMR for protein studies, and interpretations of flexible loop structures and conformational dynamics are limited with crystallographic methods. Knowledge of these regions that have limited accessibility with either method is often crucial for understanding protein function. MS can then be used as a complementary technique. It can be applied to the analysis of very large or heterogeneous protein assemblies, and, combined with HDX, it can identify flexible segments in protein structures or ligand-binding regions, and can monitor the conformational dynamics in functional transitions. Because of the robust nature of the MS techniques, they have proven useful where other types of structural analysis have been difficult. One example is demonstrated here by use of an online microfluidic HDX cell system to monitor aggregation of Aβ40. With the commercial availability of ion mobility mass spectrometers, improved resolution, large mass range, and suitable ionization and labeling methods, it is likely that the MS approaches described here will become standard tools in protein structure analysis.
This work was supported by grants from the Swedish Research Council (grants 3532 and 10371) and Karolinska Institutet.