Top-Down Characterization of an Intact Monoclonal Antibody using Activated Ion-Electron Transfer Dissociation

Abstract

Monoclonal antibodies (mAbs) are important therapeutic glycoproteins, but their large size and structural complexity make them difficult to rapidly characterize. Top-down mass spectrometry (MS) has potential to overcome challenges of other common approaches by minimizing sample preparation and preserving endogenous modifications. However, comprehensive mAb characterization requires generation of many, well-resolved fragments and remains challenging. While ETD retains modifications and cleaves disulfide bonds – making it attractive for mAb characterization – it can be less effective for precursors having high m/z values. Activated ion electron transfer dissociation (AI-ETD) uses concurrent infrared photoactivation to promote product ion generation and has proven effective in increasing sequence coverage of intact proteins. Here we present the first application of AI-ETD to mAb sequencing. For the standard NIST mAb we observe a high degree of complementarity between fragments generated using standard ETD with a short reaction time and AI-ETD with a long reaction time. Most importantly, AI-ETD reveals disulfide-bound regions that have been intractable, thus far, for sequencing with top-down MS. We conclude AI-ETD has the potential to rapidly and comprehensively analyze intact mAbs.

Graphical Abstract

Introduction

Monoclonal antibodies (mAbs) are effective therapies that target specific antigens to diagnose or treat a host of diseases from cancer to viral infections.^1–8 Their large size (~150 kDa), post-translational modifications (PTMs) and numerous disulfide bond linkages between and within the light and heavy chains necessitate extensive sequence and structure analysis.^{1, 9, 10} While tandem mass spectrometry (MS/MS) techniques such as shotgun proteomics provide comprehensive analysis for clinical mAb screening, improvements in mass spectrometry could reduce cost and time required to get the drugs to the patient.^11–21

Top-down mass spectrometry (TD MS) can provide robust and rapid characterization of intact biologics without digestion steps.^{22,23, 24} However, comprehensive sequence coverage remains challenging owing to difficulties in effectively dissociating these large molecules across the entire sequence. Beam-type collisional activation, i.e., higher-energy collisional dissociation (HCD),²⁵ or ion trap collisional activation dissociation (CAD),²⁶ preferentially cleaves the weakest bonds to generate spectra dominated by fragments produced only through a low number of the lowest energy dissociation pathways, providing incomplete information about a mAb.^{27, 28} Consequently, electron-based dissociation methods have enabled the most extensive sequencing of intact mAbs to date.^29–33 Electron-driven dissociation via ion-electron (i.e., electron-capture dissociation, ECD)³⁴ and ion-ion reactions (i.e., electron-transfer dissociation, ETD)³⁵ can generate extensive backbone cleavage while retaining modifications and simultaneously cleaving disulfide bonds, important factors for characterizing mAbs.^{29, 36–45}

While ETD presents several advantages, caveats exist. For example, Fornelli and co-authors observed that even averaging thousands of spectra and including multiple charge states and reaction times, they could only achieve mAb sequence coverage of 35% with ETD alone.^{18, 30, 31} Non-covalent interactions that occur across the gas-phase precursor ion are the main barrier that limits ETD. That is, following electron transfer backbone bonds may be cleaved; however these noncovalent interactions prevent the separation of the ETD-generated (c/z^•-type) fragments, i.e., electron transfer without dissociation (ETnoD).⁴⁶ ETnoD is especially pervasive for precursors with low charge density (e.g., m/z > 1,000), which permits a high degree of secondary structure, or a high number of disulfide bonds.^{44, 46, 47} Delivery of supplemental vibrational activation of the precursor ion population either during or after the electron transfer event can reduce ETnoD and boost ETD efficiency.^48–51,33 One way to accomplish this is to collisionally activate all ETD products using HCD after the ion-ion reaction (EThcD).^52–54 Unfortunately for mAb analysis, ETD alone provided more coverage than EThcD, although a combination of the two fragmentation methods enhanced the sequence coverage to approximately 31%.⁴⁶ This underwhelming performance by EThcD was largely attributed to its inability to effectively disrupt the secondary structure of the immunoglobulin-like domains.³¹

Activated ion ETD (AI-ETD)^{51, 55, 56} bombards the precursor ion population during ETD with infrared photons. These photons are tuned so that they provide optimal energy to vibrationally excite the precursor and disrupt the non-covalent interactions. We have shown AI-ETD to provide excellent performance for both large proteins (up to ~66 kDa)⁵⁶ and proteins rich in disulfide bonds.⁴⁷ Here we examined the utility of AI-ETD for sequencing of the intact NIST monoclonal antibody on a modified Fusion Lumos Orbitrap platform. With a significant quantity (>100 μg) of highly pure mAb in hand, direct infusion was used to rapidly screen different AI-ETD laser powers and ETD reaction times.³⁰ For AI-ETD, laser power and reaction time were varied generating distinct populations of fragment ions. For example, an increase of reaction times and laser powers revealed more fragments from disulfide-bound regions, suggesting that higher energy IR photons disrupted structures that stemmed from disulfide connectivity. Further, our results indicated that AI-ETD can provide substantially more information about the sequence of an intact mAb than ETD alone and that ETD and AI-ETD were complementary – especially when using different ion-ion reaction times. With this technique we achieved over 60% sequence coverage of the intact mAbs using AI-ETD for TD-MS.

Experimental Methods

All experiments were performed on an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, San Jose, CA) that has been previously modified with a Firestar T-100 Synrad 60-W CO₂ continuous wave laser (Mukiltwo, WA) for AI-ETD⁵⁷ (see Supporting Information for more details).

Results and Discussion

Owing to its well-characterized features,²⁸ we selected the NIST intact mAb standard to assess AI-ETD performance. Aiming to develop a comprehensive and fast approach we performed direct infusion of the NIST mAb standard. Shown in Figure S1, the major glycoforms of the NIST mAb were confirmed through intact mass analysis. These measurements provided a global overview of the antibody’s features, but therapeutic mAbs required unambiguous characterization of their PTMs and sequence. To achieve this, we selected the most abundant precursor population and tested AI-ETD performance by varying the reaction times until an apparent maximum sequence coverage was achieved for each laser power (Figure S2). AI-ETD robustly generated mAb fragments at moderate laser powers; at 12 and 18 W more than 300 products were assigned over broad reaction time ranges of 40 to 400 ms and 15 to 220 ms, respectively. At higher powers of 24 W and 30 W, the laser’s influence on the ion fragmentation was more prominent and the precursor ions likely fragmented multiple times generating unconventional product ions.

As previously reported^{31, 46, 58, 59}, the ETD reaction duration generated distinct spectra and AI-ETD recapitulated this trend(Figure S3). For instance, the spectrum resulting from a 5 ms ETD reaction provided large product ions with average charge states of 8+, across the whole spectrum (Figure 1A). With both longer ion-ion reaction times and irradiation at 18 W, charge-reduced products and products resulting from multiple electron transfer events distributed across the entire m/z range with average charge states of 4+ (Figure 1B). To better illustrate these differences, we magnified and annotated the region from 1,500 to 1,550 m/z from each of these tandem mass spectra. In the experiment using long reaction duration and AI-ETD at 18 W the resolving power was sufficient to delineate fragments of similar m/z, such as the light chain c₃₀ and heavy chain z^•₂₈ species, which were observed as well-resolved isotopic distributions (Figure S4). The main challenge of the prolonged reaction times was the presence of low intensity, internal fragments, a well-known caveat of overexposure during ETD fragmentation.^{60, 61} In contrast, short ETD reaction times preserve large, highly charged fragments that were difficult to deconvolute due to overlapping signals in greatly congested regions of the spectrum. Nevertheless, combining short ETD reaction times with longer AI-ETD reaction times resulted in almost 400 unique bonds broken and over 60% mAb sequence coverage. Future experiments using targeted approaches, e.g., MS³ or parallel ion parking with proton transfer reactions,^{58, 62–66} in combination with AI-ETD may make these internal regions more accessible while also improving deconvolution of highly charged fragments from shorter reaction times.

Figure 1.

Distinct patterns were evident in spectra generated using short and long reaction times. The spectra obtained from the accumulation of 200 scans at a resolving power of 240,000 at 200 m/z using A) ETD for 5 ms B) AI-ETD 18 W laser power for 120 ms. The peaks in the top spectra are colored according to their assigned charge states. Below, the annotated region from 1,500 to 1,550 m/z denotes the best matching ions for the heavy chain in blue and the light chain in red. NL = normalized level

The modifications generated from disulfide cleavage pathways revealed how the disulfide connectivity influenced the observed product ion populations. Since disulfides contributed to maintaining protein structure in solution, they promoted gas-phase structures that were conducive to ETnoD. AI-ETD disrupted the residual gas-phase structures after ETD cleaved the backbone and disulfide bond. For instance, the fragmentation patterns for the reaction time and the disulfide bond network were decipherable by tracking abundance of specific ion populations and cleavage sites at 18 W laser power (Figure S5). With the extended reaction times, b- and c-type ions dominated within the disulfide bound regions near the N-terminal regions of the heavy and light chains, while small y- and z•-type fragments dominated at the C-terminal region. Thus, the extended time provided the opportunity for specific reactions at the disulfide and backbone bond(s) and generated shorter fragments (<15 kDa). Products whose formation required more than two intra-disulfide cleavages, were rarely observed, and only captured in short reactions, as they were presumably lost as internal fragments with prolonged ETD exposure. The diversity of fragments from short and long reaction times prompted us to employ a pairwise comparison between different activation conditions. The results suggested that a greater number of complementary fragments were generated when AI-ETD was performed using different reaction times or laser powers, or when ETD alone was performed, relative to when other fragmentation techniques (Figure S6). Two conditions that afforded the highest number of cleavage sites for the light and heavy chains included one short and one long reaction time for each chain (Figure 2). More importantly, the number of unique bonds broken under each condition was greater than the number of overlapping bond cleavages between the two (Figure 2). Only 55 of 248 bonds (~22%) were broken under both conditions for the heavy chain, and 52 of 169 total bonds (~31%) were redundant between the two analyses of the light chain. The optimal conditions for the light chain involved higher vibrational energy at longer reaction times (AI-ETD at 18 W for 80 ms), while 5 ms ETD reaction was preferred for the heavy chain. Effective fragmentation of the light chain necessitates breakage of the interchain disulfide bonds and therefore, AI-ETD with longer reaction times and higher laser power was the most efficient at sequencing this region.

Figure 2.

The average signal (% TIC) of matched fragments obtained under the two conditions listed as plotted according to an amino acid residue position in the heavy (top) and (bottom) light chains. The intrachain disulfide bonds are shaded in light blue and the interchain disulfide bond is shown in pink. The Venn diagrams illustrate the number of distinct and overlapping bonds broken under each condition.

ETD is a charge-state dependent reaction,^{35, 60, 67} so additional fragment populations were expected from precursors with higher charge states. However, a pairwise comparison for different precursors showed that ETD provided considerably fewer unique cleavage sites, especially in comparison to AI-ETD (Figure S7). This supported that an invariable set of fragment ion populations consistently contributed to ETnoD products due to the disulfide bond network and stable gas-phase structures (Figure S8). The fragmentation patterns of ETD were highly dependent on the disulfide bonds, so they could validate disulfide connectivity and provide important structural information in future studies.

Beyond the amino acid sequence, glycan and glycosylation site information were also relevant for the antibody function and specificity.^68–72 Briefly, we successfully reconstructed several fragments with intact glycans using ETD and AI-ETD (Figure S9). Unfortunately, these low abundant fragments were insufficient to confidently localize the precise site of modification. The expected glycan-containing fragments had relatively large masses, between 20 to 53 kDa, so another experimental set-up may have been more conducive to glycan analysis. For instance, using a higher mass range and lowering the IRM pressure would have both transmitted and detected these ions more readily.

The complementarity determining region (CDR) was also an important region for sequencing the mAb because it engendered specificity by targeting distinct antigens.^{11, 73} In a single AI-ETD experiment (18 W, 100 ms), 78% of amino acids in this region were confirmed by observed product ions (Figure 3) suggesting that this technique alone can provide considerable insight into the sequence of this region.

Figure 3.

The fragment ion map for the NIST variable region for AI-ETD at 18 W, 100 ms.

Finally, we assessed the relative value of other fragmentation techniques for mAb sequencing. The highly abundant 48+ precursor was selected in all experiments, and when compared to ETD, EThcD, and HCD, AI-ETD at 18 W for 120 ms, yielded the highest total sequence coverage of 51% (Figure S10). The results with AI-ETD demonstrated that improvements in fragmentation techniques can provide considerable success for top-down sequencing of mAbs.

Conclusions

Rapid, comprehensive analysis of mAbs is a pressing need in biopharmaceutical research. TD-MS of intact mAbs expedites sample processing and data acquisition, and ETD-based methods have proven valuable for generating moderate sequence coverage of the intact mAb.^{29, 31, 46} Here we demonstrate that AI-ETD also provides myriad benefits for analysis of the intact mAb. Namely, AI-ETD improves fragmentation of disulfide-enclosed regions, yields a greater number of total sequencing ions, and as a supplemental activation method, considerably outperforms EThcD. Combination of multiple reaction times for a given precursor has been used to boost coverage in previous studies,^{24, 46} and it proves valuable here as well. However, the improved sequencing power of AI-ETD does not approach the nearly complete coverage from shotgun proteomics.^{16, 28} Major challenges remain for TD MS/MS involving the generation and analysis of complex product ion spectra. For the application of AI-ETD some of these issues are circumvented by restricting the m/z range to include well-resolved peaks and by limiting the fragment library to include well-known modifications. To that end, many product ions remain unexplained. Other important information is also missing including the localization of glycosylation site(s). While experimental MS/MS conditions could be optimized to detect large modified fragments, additional technological advances such as improved resolving power for heavy product ions^{74, 75} are required to confidently detect these ions. Other techniques to reduce spectral complexity including parallel ion parking with proton transfer reactions can increase the number of assigned peaks and the confidence in their identity and contribute to disulfide mapping. An additional factor that requires attention is the complexity and throughput required for mAb characterization. The direct infusion method used in this study is ideal for comparing MS/MS results with a highly pure mAb but is low throughput. The optimal MS/MS methods established here are adaptable to include online separation which would also increase the sensitivity and throughput of AI-ETD for characterizing mAbs. In addition to technological advancements, informatic tools are also required to confidently identify all product ions and completely profile mAb heterogeneity. With continued efforts, TD MS/MS using AI-ETD has the potential to rapidly and efficiently provide exhaustive structural information on therapeutic mAbs, including their sequence, disulfide linkages, and post-translational modifications.

Figure 4.

Number of disulfide cleavages for the observed fragments by EThcD at NCE 20% and AI-ETD at 18 W for 5 and 120 ms are plotted according to the amino acid residue position in the heavy and light chains. The intrachain disulfide bonds are shaded in light blue, and the interchain disulfide bond is shown in pink.