Hydrogen/Deuterium Exchange Mass Spectrometry (HDX-MS) to Characterize Monoclonal Antibody Structure
HDX is a protein labeling technique whereby accessible protons within the protein structure are replaced with deuterium through sample dilution in a deuterated buffer . The exchange with deuterium occurs for backbone peptide bond hydrogens and amino acid side chain hydrogens, the rate of which depends upon their structural connection and/or their location . The rate of exchange for different labile hydrogen atoms is also highly dependent upon pH . This pH dependence has been exploited in order to prevent deuterium back-exchange and migration; a pH of approximately 2.6, in conjunction with reduced temperatures (0 °C), is employed to allow sufficient time for MS measurements .
This technique was successfully coupled with a mass spectrometer (HDX-MS) in 1991 by Katta and Chait . The exchanged deuterons each have a relative mass increase of one mass unit, which can be easily detected using a mass spectrometer . The information obtainable from an HDX mass measurement experiment includes insights into conformational dynamics by observing the difference in proton exchange rates between the protein backbone and the aqueous solvent ; analysis of higher-order structure and dynamics of protein-based therapeutics [90, 96, 97] such as mAbs, both on their own and when interacting with target proteins or antigens ; and insights into the degree of glycosylation and the effects of the glycans upon structure . HDX-MS can also be used to locate conformational change in the tertiary fold , currently achievable at the peptide level . In addition, paratopes of mAbs and epitopes of antigens can be mapped [100, 101] enabling complementary pairings to be highlighted and proposed during new drug discovery . Figure 10.5 shows the basic workflow of a typical HDX-MS setup.
The information-rich nature and sensitivity of HDX-MS have positioned it as an emerging technique for the analysis of biologics. Houde et al.  performed a direct comparison between a glycosylated IgG1 with the deglycosylated equivalent. With use of limited proteolysis to generate Fab and Fc fragments, it was demonstrated that comprehensive deglycosylation affected the Fc CH2 domain significantly; however, the remainder of the protein was unaffected. This highlights how changes in glycosylation affect structures critical for Fc receptor binding and hence gives insights into the potential efficacy of an IgG1 mAb.
In 2013 Rose et al.  used HDX-MS to characterize the allosteric changes induced in the presence of a single point CH3 domain mutation. This was shown to have a knock-on effect upon the connected CH2 domains also. In
Figure 10.5 Workflow for a typical HDX-MS setup for the conformational dynamics analysis of monoclonal antibodies and other proteins. The mAb is combined with an equilibration buffer and then exposed to a deuterated labeling buffer for set time intervals. Experiments where no labeling occurs (f0) are used as a control to determine the rate of deuterium uptake. Following the specified time period, the labeled mAb solution is quenched using harsh conditions: 0 °C and pH 2.6. The quench buffer also contains reducing and denaturing reagents to reduce the disulfide bonds and aid pepsin digestion. The peptide fragments are then separated and analyzed via LC-MS.
order to improve the clinical efficacy of mAb therapeutics, mutations could be introduced into the CH3 domains to engineer CH2 domain N-glycosylation, such as afucosylation for enhanced ADCC.
Over recent years, Zhang et al have investigated mAb aggregation mechanisms  and the impact of chemical modification  and have developed a platform for automated data processing  all using HDX-MS. The aggregation mechanisms were demonstrated using two opposing extreme conditions: freeze-thaw cycles and heat denaturing . Their experiments were performed on the therapeutic mAb bevacizumab, and interestingly the aggregates that formed upon exposure to multiple freeze-thaw cycles consisted of several “native” structural monomers, with little difference observed between the LC and HC fragments of the intact mAb and the freeze-thaw-induced aggregates. This demonstrated the highly stable nature of the mAb toward freezing stress. However, significant disordering under thermal stress was observed, as confirmed by the significant increase in HDX within the LC of the Fab region during aggregation. Three peptides within the CDR became increasingly solvent protected, relative to the native state upon aggregation, implying that these regions were responsible for some of the intermolecular interactions within bevacizumab aggregates.
Aggregation of mAbs has also been explored by Iacob et al.  where monomers were compared to their naturally occurring corresponding dimers. Subtle changes in deuterium uptake were observed for one of the samples in the CH2 domain and close to the hinge region. Conclusions reported, based upon the HDX-MS data and other biophysical techniques, that the dimerization of the mAb (with changes to its structural dynamics) was due to an alternative dimerization pathway: domain swapping at the hinge region. For the mAb with no recorded exchange rate differences, it was proposed that the dimerization pathway predominantly involved CH2 amino acid side chains where the exchange rates were not measurable via HDX-MS.
Zhang et al. also used HDX-MS to detect local chemical modifications within IgG1 mAb therapeutics such as asparagine deamidation, methionine oxidation, and aspartic acid isomerization , in order to study their conformational impact. These modifications can ultimately affect the safety and efficacy of mAb therapeutics through changes in thermal stability, antigen binding capabilities, and Fc effector functions. For methionine oxidation, impact was observed close to the modification site and in some cases also affected the residues linking the CH2 and CH3 domains. Through the comparison of a glycosylated mAb and an aglycosylated mAb, the authors demonstrated the importance of N-glycans upon the effect of methionine oxidation; significantly greater conformational changes were observed for the deglycosylated mAb. This supported the idea that N-glycans increase the structural stability of the CH2 domain. Following on from this, it was demonstrated that aglycosylated mAbs also have decreased thermal stability and significantly increased levels of aggregation as compared to glycosylated mAbs. It was postulated that the effects upon the deglycosylated mAb conformation, following methionine oxidation, could be the structural requirements for changes toward thermal stability and aggregation. The effects of aspartic acid isomerization and asparagine deamidation upon the local mAb conformation were found to be minimal unless sufficient accumulation of succinimide intermediate was present. All of their HDX-MS observations were confirmed using structural modeling.
A combination of IM-MS and HDX-MS analyses has been reported by Edgeworth et al detailing the conformational and dynamic properties of IgG Fc-engineered variants with reduced thermal stability . Engineered to be active for reduced immune system recruitment and to increase serum half-life meant the samples exhibited increased thermodynamic destabilization. Using these techniques, the authors were able to show that the mutations effected a local destabilization within the CH2 domain (HDX-MS) without affecting the global conformation of the Fc region (IM-MS). Effected HDX-MS has shown that drug binding to form an ADC does not drastically affect the conformation of the stand-alone mAb . The authors compared interchain cysteine-linked IgG1 ADCs with the equivalent drug-free mAbs and observed 90% similarity between their HDX kinetics. The minor differences observed were a slight increase in the structural dynamics of the Fc region of the ADCs. Upon further testing, these differences were found to be attributable to the absence of intact interchain disulfide bonds and not the presence of the attached drug. Similarly, HDX-MS has been incorporated into comparability testing workflows between biosimilar candidates and originator products and revealed no difference between the exchange rate dynamics .
For a mAb therapeutic where a secondary effector function is part of the mechanism of action, receptor binding in vivo followed by induction of an immune response is required. As discussed previously, N-linked glycosylation in the Fc domain can affect the binding affinity between IgG and Fc receptors (FcR) and hence ADCC potency. Using HDX-MS, Jensen et al.  were able to show that analogous N-linked glycosylation present on neonatal FcRs (FcRn) did not affect IgG binding, implying that binding affinities are due solely to IgG glycosylation. Further HDX-MS experiments also revealed the expected conformational role for the Fab region in IgG1 mAbs with respect to receptor binding. Significant protection from HDX was observed for specific regions within the Fab arms upon receptor binding, indicating the presence of conformational interactions between the receptor, Fc and Fab regions. Generation of a 3D homology model for a full-length IgG1 in complex with FcRn confirmed the feasibility of interactions between the Fab and bound receptor.
Despite HDX-MS being well established, some limitations remain: one issue being the back-exchange of deuterium, as mentioned previously, and the effects that the quenching conditions may have upon protein structure, which could confound the results. This effect needs to be carefully considered for different protein systems . In addition, for antibody-antigen interactions when the binding stoichiometry is very heterogenic, the effect of binding on each component may be difficult to determine. The complexity of the analysis and in particular the data processing has held HDX-MS back from routine use . With the involvement of so many steps - deuterated incubation, proteolytic digestion, chromatographic separation, and MS identification, there is the requirement for an experienced analyst and comprehensive data analysis packages. In summary it is a powerful technique, but more akin to NMR in terms of the length of time needed to solve a protein structure, relative to an experiment based on MS alone.