Expedited Evaluation of Conformational Stability-Heterogeneity Associations for Crude Polyclonal Antibodies in Response to Conjugate Vaccines

Abstract

Conjugate vaccines have been demonstrated to be a promising strategy for immunotherapeutic intervention in substance use disorder, wherein a hapten structurally similar to the target drug is conjugated to an immunogenic carrier protein. The antibodies generated following immunization with these species can provide long-lasting protection against overdose through sequestration of the abused drug in the periphery, which mitigates its ability to cross the blood-brain barrier. However, these antibodies exhibit a high degree of heterogeneity in structure. The resultant variations in chemical and structural compositions have not yet been clearly linked to the stability that directly affects their in vivo functional performance. In this work, we describe a rapid mass spectrometry-based analytical workflow capable of simultaneous and comprehensive interrogation of the carrier protein-dependent heterogeneity and stability of crude polyclonal antibodies in response to conjugate vaccines. Quantitative collision-induced unfolding-ion mobility-mass spectrometry with all-ion mode is adapted to rapidly assess the conformational heterogeneity and stability of crude serum antibodies collected from four different vaccine conditions, in an unprecedented manner, for the first time. A series of bottom-up proteomics and glycomics experiments were performed to reveal the driving force underlying these observed heterogeneities, including Fab backbone sequence variation and N-glycosylation alteration of the Fc region, and to establish its potential relationship to functional stability. Overall, this study not only presents a generally applicable workflow for fast assessment of crude antibody conformational stability and heterogeneity at the intact protein level based on raw signal without isotopically resolved datasets but also leverages carrier protein optimization as a simple solution to antibody quality control.

Graphical Abstract

Conjugate vaccines are a type of vaccine that is created by joining a smaller antigen with a larger protein molecule.¹ The large protein part of this vaccine acts as a carrier for the antigen and serves to magnify the immunological response to the vaccine. Conjugate vaccines were originally developed to immunize babies and children against certain bacterial infections. In recent years, the conjugate vaccine approach has been demonstrated to be a promising strategy for immunotherapy to address substance use disorders, wherein a synthetic hapten structurally similar to the target drug of abuse is conjugated to an immunogenic carrier protein.^{2, 3} When formulated with adjuvants and administered in vivo, the immunoconjugate should elicit serum IgG antibodies with the ability to sequester the target drug to prevent its entry into the brain, thereby acting as an immune-antagonist. Immunization with conjugate vaccines is a particularly promising approach compared to conventional medication since the antibodies thus generated have long half-lives, which allows for treatment on the timescale of months, far outlasting the metabolism and elimination of small-molecule therapeutics.⁴ Moreover, because immune pharmacotherapy primes the body’s own immune system to combat opioid effects, there are limited side effects.⁵

However, the performance of anti-opioid antibodies as overdose reversal agents is a function of affinity, and the effectiveness of a conjugate vaccine is highly dependent on the production of highly specific and potent drug-binding antibodies. Magnitude of antibody response to vaccination across individuals is highly variable, due in part to underlying differences in hapten-specific B cell populations prior to vaccination.⁶ Following vaccination, the generated antibodies also exhibit a relatively high degree of heterogeneity in structure due to numerous factors, including post-translational modifications, hapten structure, immunoconjugate identity,.^{7, 8} These characteristics have been reliably demonstrated to be modified depending on the identity of the immunoconjugate generated to induce an immune response,^{3, 9} but the relationship between vaccine design and resulting antibody properties remains unclear. Therefore, rapid characterization of structural heterogeneity across individuals, and comparison of heterogeneity, stability and performance of the crude antibodies boosted by vaccines with different carrier proteins and hapten conjugates are critical. The mechanistic understanding of the effective neutralizing antibody induction by conjugate vaccines will facilitate their design, production, and clinical use in immunotherapy.

A battery of analytical methods has been developed for the detailed and rapid characterization of therapeutic antibodies, including complementary chromatographic, spectroscopic, and spectrometric approaches.^10–12 Specifically, mass spectrometry (MS)-based workflows have been constructed for the detailed sequence and heterogeneity characterization of antibodies.^13–15 Of note, immunoglobulin G (IgG) accounts for the most abundant isoform amongst various types of human immunoglobulins, the origin of antibodies. As reported, different levels of IgG information ranging from sequence composition to higher-order structures could be obtained through different categories of MS-based analysis on various samples, e.g. intact IgG assemblies, intact IgG monomers, peptide fragments for IgG, and even stripped glycans of IgG.¹⁶ Specifically, native top-down MS has grown to be an indispensable tool for the comprehensive characterization of functional proteins, including IgG, at intact protein level.^{17, 18} To achieve the most attainable information, most native MS experiments highly rely on high-resolution, e.g. isotopically resolved, datasets to dissect the molecular heterogeneity of complex systems as heterogenous as antibodies.^{19, 20}

Meanwhile, ion mobility-mass spectrometry (IM-MS) is capable of separating gas-phase protein ions according to their size, shape, and charge (frequently converted to orientationally averaged collision cross section (CCS)) and can thus act as a tool to resolve ions that may be indistinguishable by mass alone or to achieve indispensable structural information at the intact protein level,^21–23 making it suitable for global heterogeneity characterization of intact antibodies. Collision-induced unfolding (CIU) has effectively enriched the IM-MS toolbox through potentially improved resolution in CCS and associated unfolding intermediate information for rapid protein structural and conformational analysis in the gas phase.^24–26 Many prior studies have successfully proven the power of the CIU-IM-MS regime in rapid quality assessment of monoclonal antibodies, mainly focusing on conformational stability and CCS variations in response to either extra physical stress or chemical treatments.^27–34 Hybrid experiments have identified key structural elements of IgG monoclonal antibodies, including glycosylation, disulfide bridge, domain exchange, and even batch-to-batch variations.^27–31 This work expands on those applications by assessing the structural heterogeneity of polyclonal antibody responses across biological replicates and the driving force behind this type of heterogeneity.

In this work, we combined two cutting-edge MS-based approaches, i.e., (1) expedited, quantitative CIU-IM-MS technique with all-ion mode, namely All Ion Unfolding (AIU),^{21, 35} and (2) bottom-up de novo sequencing and glycoproteomic profiling MS technique, for the conformational and compositional quantification of the heterogeneity and stability alteration of four groups of comparative crude polyclonal antibody (cpAb). Figure 1 illustrates the overall workflow for the comprehensive characterization of the carrier protein-dependent heterogeneity-stability relationship of cpAbs in response to four different vaccine conditions. Combining the data obtained at the AIU-equipped intact protein level and bottom-up proteomic level, we were able to better resolve the overall heterogeneity exhibited by cpAb generated by different carrier proteins and conjugate vaccines, and quantitatively profile them with satisfactory confidence. Based on our analysis, both the sequence variations in the antigen-binding domain (Fab) and the glycosylation at the CH2 domain (Fc) simultaneously contribute to the observed molecular micro-heterogeneity, which might be critical in affecting the conformational stability and binding efficacy towards haptens.

Fig. 1.

Overall workflow for comprehensive characterization of the carrier protein-dependent heterogeneity-stability relationship of crude antibodies in response to four different conjugate vaccines.

EXPERIMENTAL DETAILS

The key workflow was shown in Figure 1. Briefly, an expedited native IM-MS technique equipped with all-ion unfolding was combined with bottom-up glycoproteomics and de novo sequencing for comprehensive characterization of isolated antibodies. Detailed procedures for experiments were listed in Supporting Information.

RESULTS AND DISCUSSION

Inoculation and generation of polyclonal antibodies.

Keyhole limpet hemocyanin (KLH) is often used as a carrier protein because of its extreme immunogenicity and availability. However, due to its mullock origin and challenges in characterization arising from its multimeric nature, native KLH is not an ideal carrier. In contrast, CRM197, a genetically detoxified diphtheria toxin, is a recombinant protein of 58.4 kDa and is also highly immunogenic, making it a more tractable carrier for translational and clinical studies. Moreover, drug- and peptide-CRM197 conjugates can easily be molecularly characterized and identified with MS. Besides these two commonly used immunogenic carrier proteins, an exemplar hapten bioconjugate, CRM-NK, was applied for parallel studies. To generate an effective vaccine, the KLH or CRM197 immunogens were formulated alongside alum and CpG (100 μg protein, 50 μg CpG, 500 μg Alum), respectively. This formulation was administered intraperitoneally to Swiss Webster mice using a three-injection schedule on days 0, 14, and 28.

After that, as shown in Figure 1, a simple isolation step based on the Protein G column (see more details in Methods) was applied to obtain cpAbs from multiple mouse serum samples. This simple isolation process allowed for analysis of whether vaccine-induced changes to IgGs are readily observable even in the presence of non-vaccine associated circulating IgG populations. Then, the collected cpAb samples were first subjected to quality evaluation through ELISA-centered workflow. After that, de novo sequencing experiments and untargeted glycoproteomic profiling experiments were carried out on individual samples to examine the compositional heterogeneity in backbone sequence and glycosylation, respectively. Finally, the heterogeneity characterization was followed by conformational stability analysis via an expedited ion manipulation strategy, namely the all-ion unfolding (AIU) regime. These sets of experiments were aimed to tentatively establish the potential connectivity of conformational stability to compositional heterogeneity, in addition to vaccination functions.

Activity evaluation of polyclonal antibodies.

To functionally validate the antibodies after immunologic response, sera were collected from vaccinated mice on day 35 following the initial injection, and midpoint titers for serum antibodies were assessed using ELISA. Sera from each vaccine group was assessed against coated KLH or CRM. As shown in Figure 2, saline mock control without carrier protein exhibited little binding towards either KLH or CRM, whereas successful binding was observed for KLH or CRM vaccinated groups. Titer analysis suggested that the CRM vaccine produced a slightly more robust antibody response against CRM when compared with that of KLH antibody against KLH. A 2-fold higher midpoint titer was observed for CRM (Titer: 7133) versus CRM-NK (Titer: 3036), suggesting the compromise of selectivity towards the carrier protein itself after bioconjugation. Additionally, the capability of the CRM-NK antibody to recognize its small molecular antigen (i.e., ketamine) was validated in our previous report.³⁶ However, the basic principles underlying these observations remain unclear due to the lack of an effective and rapid method for comprehensive characterization of stability-heterogeneity associations for cpAbs in response to conjugate vaccines at the molecular scale.

Fig. 2.

Functional validation on antibody response after vaccination via ELISA. Measurement of optical density (OD) responses for sera against KLH or CRM as compared to a standardized positive control (1:1,000,000 horseradish peroxidase-conjugated donkey anti-mouse IgG) following vaccination of mice with saline mock (black), KLH vaccine (red), CRM vaccine (blue) or CRM-NK (green). Solid lines show the best non-linear fits for Log[dilution] vs % Response (R square: 0.99, 0.93, and 0.97 for KLH, CRM, and CRM-NK respectively). All data are shown as mean ± SEM (standard error of the mean).

Heterogeneity evaluation of polyclonal antibodies.

To this end, we first sought to characterize the molecular heterogeneity of these crude polyclonal antibodies, which were obtained by elution from a Protein G column and roughly verified by SDS-PAGE (Figure S1). In order to precisely evaluate the influence of carrier protein and chemical modification on the structural integrity and heterogeneity of the antibodies, a series of native MS experiments were performed on corresponding antibodies. Figure 3 shows the representative native MS spectra for standard IgG mAbs and isolated crude antibodies (anti-CRM, anti-CRM-NK, and anti-KLH) from different mice. Not surprisingly, we observed similar charge state distribution windows ranging from 24+ to 28+ as these were sprayed and ionized from the same buffer condition (e.g. 200 mM ammonium acetate), adding to their shared conformational distributions. Compared with standard polyclonal IgG, we found that the isolated crude antibodies (anti-CRM, anti-CRM-NK, and anti-KLH) exhibit peaks with a much broader half-width at half maximum (FWHM), indicating their heterogeneity in molecular composition. Notably, in addition to the primary peaks, a second antibody species (pink asterisks in Figure 3c) was easily observed in the anti-CRM-NK group, indicating that the CRM-NK group is even more heterogeneous compared to the CRM group, presumably linked to NK-hapten-induced vaccination process.

Fig. 3.

Representative native MS spectra for standard mAb (a) and isolated crude polyclonal antibodies (b-d, anti-CRM, anti-CRM-NK, and anti-KLH) from different mice. Pink asterisks in the anti-CRM-NK group indicate a second antibody species aside from the major ones as labeled, which contributes to its highest heterogeneity across all isolated antibodies. Note, five different trap voltages ranging from 10 V to 180 V were shown for each group of antibodies.

To dissect the driving force behind this observation, a series of bottom-up proteomic profiling experiments were carried out to examine the compositional heterogeneity across these crude antibodies. A total of 18 IgG-associated proteins (Table S1 & Figure 4a) were quantified from four groups of cpAb samples with mouse IgG database (version 20210830). As shown in Figure 4a, label-free quantification data from database-dependent bottom-up proteomics experiments confirmed the uniform expression and generation of total IgG proteins across different groups. Notably, quantification heatmap in Figure 4a clearly indicates the highest abundance of protein P01863, Ig heavy chain Fc region, across all quantified cpAb Ig proteins (Table S1). It is then suspected that the micro-heterogeneity for different cpAbs as shown in native MS experiments (Figure 3), should be related to two main factors, post-translational modification (PTM) and sequence variations in response to various conjugate vaccines.

Fig. 4.

Compositional heterogeneity characterization. (a) Overall IgG expression level comparison as quantified from bottom-up proteomics experiments (P01863 highlighted, detailed protein information is listed in Table S1). (b) N-glycosylation level comparison highlighting IgG protein P01863 @ Asn180 (often referred to Asn392 @ IgG heavy chain Fc region as shown in PDB structure of 5VAA). Hierarchical clustering of 12 quantified N-glycoform Z-score intensities across different groups were shown, with P01863 protein glycan clusters highlighted in blue rectangles. (c) Ranked plot of protein abundance difference among vaccine groups with ANOVA p-value versus protein rank. (d) Box plots showing expression levels of selected proteins across different conjugates. (e) Sequence variation comparison via de novo sequencing. Sequence coverage was shown in gray.

It is well known that glycosylation is one of the major PTMs of antibodies. As a result of IgG glycosylation-derived heterogeneity, it is reported to be tightly related to the immunogenicity, binding ability, maintaining solubility, and therapeutic efficacy of antibodies.^37–41 Therefore, to survey the potential effects of glycosylation on the differential degrees of observed molecular heterogeneity, a series of MS-based glycosylation profiling experiments were carried out, with a special emphasis on Fc region. As expected, the glycosylation site was identified in the CH2 region, Asn180 @ P01863 (Figure 4b). Complete N-glycosylation identifications were listed in Table S2, featuring 25 glycoforms from five different glycoproteins as identified at the tryptic glycopeptide level. For Fc region of IgG protein P01863, we quantified 12 types of N-glycans for site Asn180 on peptide EDYNSTLR, most of which contain one fucose and almost half with one or two NeuGc sialylation. The clustering heatmap (Figure 4b) for quantitative comparison of these N-glycosylation patterns indicates that overall N-glycans in the anti-CRM-NK groups are, generally, in lower level compared to that of the anti-CRM groups. In contrast, O-glycosylation profiling and quantification results (Figure S2 & Table S3) suggest comparable O-glycan expression between anti-CRM-NK groups and anti-CRM groups. Considering the above observation of higher heterogeneity for anti-CRM-NK groups directly from native MS experiments, we can conclude that the glycosylation level at the CH2 domain does not, or at least not solely, contribute to the carrier protein-responsive heterogeneity.

Antibody sequence information is crucial to the understanding of structural basis for antigen binding, which enables the use of antibodies as therapeutics and research tools. To further characterize the heterogeneity in response to a variety of vaccine treatments with differing carrier protein systems, we then moved onto the structural and compositional characterization of the antigen-binding fragment (Fab site), especially for the complementarity determining region (CDR). It is anticipated that the potential sequence mutation diversity in CDR should be closely related to the carrier protein-driven cpAb heterogeneity in response to conjugate vaccines. We performed a series of de novo sequencing experiments to examine the sequence variations for different anti-groups. Of all protein domains (Figure 4c) as sequenced from available tandem MS data, two regions showed significant differences across four groups, P01791 and P03987 (Figure 4d). Both P01791 and P03987 proteins belong to the heavy chain of IgG, but correspond to variable antigen-binding fragments (“VH”) and constant region (“CH”).

Not surprisingly, we observed significant levels of sequence mutation for the VH protein but only minor mutation levels for the CH protein (Figure 4e). For example, in protein P01791, various mutations were identified from a number of sites, including Glu35 with up to six mutants and Ala64 with up to five mutants. In contrast, for protein P03987, only a small portion of residues was identified with sequence mutations, showing a relatively constant and conserved sequence motif. Notably, the quantitation results in Figure 4d across four different vaccine groups clearly show much wider intensity distribution window for anti-CRM-NK groups compared to other groups including anti-CRM group, especially for the VH protein P01791, suggesting the highest heterogeneity in sequence composition for anti-CRM-NK groups. This observation shows good accordance with native MS results (Figure 3). Collectively, the molecular heterogeneity of IgG from cpAb in response to conjugate vaccines with varied carrier protein systems should be closely related to the sequence variation of CDR region in the heavy chain, although other PTMs like glycosylation @ CH2 domain can also affect the overall composition as well.

Stability and its associations with heterogeneity.

Following the comprehensive characterization of molecular heterogeneity, we then sought to evaluate the conformational stability of cpAbs. Gas-phase unfolding technique, namely CIU, has previously been well suited to gather the conformational and dynamic structural information of isolated IgG in a charge state-dependent data acquisition mode.^{23, 27–29} To further empower the unfolding technique in terms of data efficiency and signal-to-noise ratios, we employed an updated data acquisition and visualization package for CIU, recently published as the AIU method.^{21, 35} In addition to the first application of AIU to the crude antibody for this study, we inherit the merits of CIU and CIUSuite software tools developed by the Ruotolo group,^{42, 43} and proposed herein for the first time the use of unfolding fingerprints averaging from multiple biological replicates from different mice to perform quantitative comparisons (Figure 5), through the generated AIU50 values and RMSD values across different batches of cpAb samples. While averaged feature CCS values can deliver overall structural information for each identified unfolding intermediate, the AIU50 and RMSD values permit the direct readout of conformational stability and heterogeneity, respectively.³⁵

Fig. 5.

Workflow for expedited quantification of conformational stability and heterogeneity of cpAbs via the expedited unfolding IM-MS strategy. The conformational heterogeneity of IgG from various mice (n = 4) are quantitatively characterized by averaging the gas-phase unfolding fingerprints (feature CCS, CIU50 and transition number) and corresponding RMSD values from the unfolding difference plots.

As such, averaged IgG CIU fingerprints (with characteristic CCS, CIU50, and transition number) and corresponding unfolding difference plots (RMSD) calculated from biological replicates were obtained firstly in a charge-separated fashion and shown in Figures S3–6, for saline control, anti-CRM, anti-CRM-NK, and anti-KLH groups. The characteristic CCS distributions of unfolding intermediates at individual charge states were captured and the corresponding charge-dependent CIU50 values were calculated for all cpAbs. In accordance with the previous report,³⁵ different numbers of unfolding features were captured at different charges, generally with four to six features observed for charges 24+ to 28+ (Figures S3–6). Similarly, CCS and CIU50 values were also highly charged state-dependent and not straightforward to be compared across different charge states. Furthermore, the RMSD values calculated from replicates range from 13%-19%, 17%-20%, 17%-20% and 8%-10%, for saline control, anti-CRM, anti-CRM-NK and anti-KLH groups, respectively. Although we can draw some tentative conclusions on the diversity for different groups, it is still difficult to make a fair comparison as these RMSD values vary by charge states.

In contrast, when converting datasets into AIU mode (Figure 6), we can directly compare the unfolding behaviors across four cpAbs from a variety of mouse sera. In general, the features for the initial compact state (F1) among all the groups are very similar (Figure 6a & b), as revealed by the CCS values of 77 nm² with deviations of only 0.54% for different cpAb groups. Similarly, no significant differences were observed for other unfolding features as indicated by the accumulative CCS values for the first five features only with 0.43% deviation, although with 2.3% deviation for Feature 4 alone. Surprisingly, when AIU50 values as generated from AIU fingerprints were compared, significant differences can be clearly identified from varied groups of cpAbs (Figure 6c). The AIU50 values for the first transitions vary from 545 eV to 765 eV, showing a deviation of 15%. Although AIU50 values for second and third transitions hold only 2.3% and 6.4% deviations, the AIU50 values for fourth transitions between F4 and F5 (Figure 6a) highlight a significant deviation of 22% with an average value of 1989 eV. Notably, the AIU50 value of the fourth transition for anti-KLH cpAb (1694 eV) is lower than that of anti-CRM group (1897 eV) by ~12%, and the comparison between anti-CRM vs anti-CRM-NK (2618 eV) group produces a difference over 38%. It is also worth mentioning that the accumulative AIU50 values for the first four transitions reveal 9.8% deviation across four cpAb groups (Figure 6c), with higher values for anti-CRM-NK group (5907 eV) than other groups, including a value exceeding that of the anti-CRM group (5059 eV) by ~17%. Collectively, AIU-based evaluation is sensitive enough to differentiate these cpAb groups generated from different carrier proteins even with only subtle differences in feature CCSs, and the conformational stability can be unambiguously and quantitatively compared based on AIU50 values. Overall, these data indicate that the anti-CRM-NK group is more conformationally stable than the anti-CRM group.

Fig. 6.

Comprehensive assessment of conformational heterogeneity-stability associations of a series of cpAbs isolated from different mice (n = 4 for each group). (a) Representative AIU fingerprints for mouse antibodies under four different conditions. (b) Feature CCS distribution. (c) AIU50 distribution indicative of overall conformational stability. (d) AIU RMSD distribution indicative of overall antibody heterogeneity (the larger RMSD values, higher heterogeneity). Data are shown as mean ± SEM. Significance analysis was performed based on the ordinary one-way ANOVA method, ns, no significance; * p < 0.05; ** p < 0.01; **** p < 0.0001.

The observed diversity in conformational stability, as expected, can be further linked to their differences in heterogeneity. In this regard, we employ the RMSD values generated from unfolding fingerprint-based difference plots from different batches of vaccinated mice for a certain group of carrier proteins. The higher RMSD values thus generated, the higher heterogeneity for the cpAb of interest. As shown in Figure 6d, the average RMSD of the IgGs generated by CRM-NK is the highest, nearly 20%, indicating its highest heterogeneity among the four vaccinated groups. On the contrary, the KLH group, which has the lowest conformational stability as discussed above, seems to have the least heterogeneity, as proven by its smallest RMSD value of ~9.2% (Figure 6d). Notably, the stark differences in RMSD between KLH and other groups especially the CRM group, clearly indicate the difference in heterogeneity and the KLH seems to generate least vaccine effects in terms of heterogeneity. Moreover, all the vaccinated groups (CRM, CRM-NK, and KLH vaccines) present a significant difference in unfolding heterogeneity compared to the control group with mock vaccination. These observations imply that both carrier protein type and small molecule conjugation play key roles in the stability and heterogeneity of corresponding IgGs generated after inoculation. Notably, it seems to be more complicated and less reliable for the comparative CIU data with charge separated information (Figures S7/8) in drawing uniform conclusions in stability comparisons across these cpAbs with only minor differences.

Collectively, data from both AIU50-based conformational stability comparisons and RMSD-based overall unfolding fingerprint comparisons clearly demonstrated that AIU and CCS accumulation enables more structurally informative conformational comparisons, and comprehensive characterization of subtle structural differences of the biologically similar IgGs generated by different carrier proteins, either with or without chemical modification via conjugation.

CONCLUSIONS

In summary, we describe a comprehensive, integrative MS-based analytical workflow capable of simultaneous and rapid interrogation of the carrier protein-dependent heterogeneity and stability of antibodies in response to conjugate vaccines. Expedited unfolding-ion mobility-mass spectrometry is firstly adapted to quantitatively and rapidly dissect the conformational heterogeneity and stability of crude serum antibodies isolated from four different vaccine conditions. A series of bottom-up proteomics and glycomics experiments were performed to reveal the driving force underlying these observed heterogeneities, including Fab backbone sequence variation and N-glycosylation alteration of Fc region, establishing its tentative relationship to functional stability. Overall, this study not only presents a generally applicable workflow for fast assessment of antibody conformational stability and heterogeneity at the intact protein level based on the raw signal in the absence of isotopically resolved datasets, but also lays the foundation for carrier protein optimization and post-conjugation quality control assessments in the development of novel conjugate vaccines.