Introduction

Therapeutic monoclonal antibodies (mAb) have achieved remarkable success in the treatment of various diseases, owing to their exceptional specificity, efficacy, stability, and safety profiles1,2,3. Recently, combinations of two or more mAbs as cocktail treatments have attracted significant interest due to their improved therapeutic outcomes, such as enhanced efficacy from interacting with different targets (or different epitopes within a target) and reduced resistance from pathogens or cancer cells4,5,6. Additionally, the mAb cocktail co-formulation approach, when compared to sequential or co-administration approaches, provides benefits in reduced risk of medical error1 and improved patient compliance and convenience4. However, despite these advantages, mAb cocktails also introduce additional challenges in clinical and regulatory procedures, process developments, and analytical quality control7,8,9. A key challenge in developing mAb cocktail co-formulation is identifying a common formulation condition that ensures stability and controls the variant profiles of all constituent mAbs in the drug product. Among these variant forms, heterodimers-formed through intermolecular interactions between co-formulated mAbs- are particularly important to monitor, as they are unique to the co-formulated drug product9. Notably, protein aggregates have been reported to potentially mediate adverse events, including increased immunogenicity, reduced product efficacy, and heightened hypersensitivity10,11,12. Therefore, heterodimer is considered as a potential critical quality attribute and must be rigorously characterized and monitored throughout the development of therapeutic mAb cocktail products8.

In recent years, a variety of technologies and methodologies have been developed to support size variant analysis in the development of protein therapeutics. A few commonly applied techniques include sedimentation velocity analytical ultracentrifugation (SV-AUC)13,14, asymmetrical flow field-flow fractionation (AF4)15,16, and size exclusion chromatography (SEC)17,18, owing to their excellent resolving power for size variants and quantitative performance. In particular, SEC with ultraviolet (UV) detection is routinely employed as a batch release assay to monitor the levels and profiles of high molecular weight (HMW) species in therapeutic mAb products19,20. These separation techniques, however, encounter significant challenges when analyzing the dimer species present in mAb cocktails. This difficulty arises from the highly heterogeneous dimer compositions (i.e., homodimers and heterodimers) in the cocktail samples and their similar biophysical properties, which impede efficient analytical separation. In addition, commonly used detection techniques such as UV, fluorescence (FLR), and multi-angle light scattering (MALS)17,18,21,22 lack the specificity required to differentiate various hetero- and homodimer species, preventing their individual quantitation.

Over the past few years, online coupling of SEC with direct MS detection under native conditions (nSEC-MS) has become increasingly popular for studying mAb HMW species due to its high sensitivity, specificity, and capability of providing structural insights23,24,25. Using MS-compatible mobile phases (e.g., ammonium acetate) that preserve the native conformations and inter-molecular interactions of mAb dimers, nSEC-MS is particularly effective in enabling simultaneous identification and quantitation of multiple dimer species from a mAb cocktail sample through accurate mass measurements. In a previous report17, we first demonstrated the feasibility of applying nSEC-MS to differentiate a heterodimer from two homodimers in a two-mAb cocktail sample, achieving their respective quantitation. However, a more comprehensive evaluation of the method’s validity, sensitivity, and general applicability is necessary before it can be widely adopted for heterodimer analysis to support mAb cocktail co-formulation development. In particular, factors that could challenge the method’s detection capability, such as the low abundances of heterodimers (commonly found in unstressed samples) and the presence of dimers with similar molecular weights (common to three-mAb cocktails), should be carefully investigated and addressed. An alternative approach, multiple dimensional liquid chromatography coupled to MS detection (mD-LC-MS/MS), was recently introduced by Kuhne et al.26 to detect covalently associated heterodimers in co-formulated mAb cocktails. This method involves fractionating covalent dimers using reversed-phase (RP) LC (1st dimension), followed by online trypsin digestion (2nd dimension) and bottom-up analysis via RPLC-MS/MS (3rd dimension). Subsequently, crosslinked dipeptides consisting of unique peptides from each mAb were used as surrogates to compare the relative abundances of each hetero- and homodimer. This method, however, cannot provide a complete assessment of the dimer species from the mAb cocktail samples, as the non-covalent dimers cannot be preserved and analyzed under the denaturing chromatography conditions. Therefore, there remains a significant unmet need for advanced analytical methods capable of characterizing the full spectrum of heterogeneous dimer species in co-formulated mAb cocktails, which are crucial for their successful development.

In this study, we report the development and evaluation of an nSEC-MS method and associated analytical strategies for heterodimer identification and quantitation in co-formulated mAb cocktails. First, the suitability of using an MS-compatible SEC mobile phase was assessed by comparing its quantitative performance against conventional SEC-UV methods, which use non-volatile buffers, across a large set of mAb samples. Further, several strategies, including native deglycosylation treatment, low-resolution (LR) MS data acquisition, and a high-throughput SEC-MS method, were developed to enhance method sensitivity and throughput. Lastly, to address the increasing complexity of the three-mAb cocktail samples- where mass detection alone was insufficient to resolve all dimers- an immunodepletion-assisted SEC-MS method was developed to interpret the convoluted dimer spectra and achieve individual quantitation of all six hetero- and homodimers. Together, our findings demonstrated that the developed nSEC-MS method, along with these analytical strategies, provided a highly effective solution for the identification and quantitation of heterodimers in co-formulated mAb cocktails. This approach holds great promise for supporting the development of the next-generation mAb cocktail therapies.

Results and discussion

Quantitative performance of nSEC-UV/MS method using an MS-compatible mobile phase

SEC separates biomolecules based on their hydrodynamic radii. When coupled with UV detection, SEC-UV analysis is widely used as a quality control (QC) method to monitor HMW contents in therapeutic mAbs. Conventional SEC methods utilize mobile phases containing high concentrations of nonvolatile salts, such as sodium chloride27 or sodium perchlorate28, to minimize the undesired protein-matrix interactions (e.g., electrostatic interaction) and ensure good quantitative recovery19. Recently, online native MS detection coupled with SEC-UV (nSEC-UV/MS) analysis has emerged as a valuable tool for therapeutic mAb characterization by allowing sensitive and unambiguous identification of low-abundance size variants. In particular, with additional mass-based separation, co-eluting dimers from the co-formulated mAb cocktail samples could potentially be differentiated and quantitated individually. However, since online MS detection requires the use of volatile salts in the SEC mobile phase (e.g., ammonium acetate)19,29,30, this may introduce changes in protein retention and recovery behavior compared to conventional SEC buffers. Therefore, its quantitative performance should be thoroughly evaluated.

To facilitate nSEC-UV/MS analysis, a previously established platform (Scheme S1) was adopted, which uses an analytical-scale SEC column, a mobile phase of 150 mM ammonium acetate (pH 6.8), and nanospray ionization (NSI) for sensitive MS detection17,31. Notably, the use of a relatively high concentration of ammonium acetate in the mobile phase was crucial for achieving good protein recovery from SEC analysis, as lower concentrations may lead to increased electrostatic interaction between mAb molecules and the column matrix32. Subsequently, the adoption of NSI for MS analysis was essential to mitigate the ion suppression effect from ammonium salt and achieve sensitive MS detection. Using 80 mAb cocktail samples as test subjects, UV-based quantitation of total dimer abundances was performed by the nSEC-UV/MS method (mobile phase: 150 mM ammonium acetate) and the results were compared to those obtained by conventional SEC-UV methods used in quality control (mobile phase: 0.5-1 M of sodium chloride or sodium perchlorate, per molecule optimization). As shown in Figure S1, the two methods exhibited a strong linear correlation for total dimer abundances (dimer%), with a coefficient of determination (R²) of 0.9508 and a slope of 1.074 (a slope of 1 indicates identical datasets). Additionally, Pearson correlation analysis (performed by GraphPad Prism) demonstrated a high Pearson correlation coefficient (Pearson r) of 0.9751 between the two datasets. Overall, these results suggested that the developed nSEC-UV/MS method, despite using a different mobile phase, achieved good quantitative consistency with conventional SEC-UV methods across a large set of mAb samples. Therefore, it is concluded that the nSEC-UV/MS method can be reliably implemented for dimer detection and quantitation during mAb cocktail development, offering results that align well with those from conventional SEC-UV methods commonly used in process/formulation development and quality control.

Native deglycosylation improves heterodimer analysis by nSEC-MS

Both IgG1 and IgG4, the most commonly used subclasses in therapeutic mAbs, feature N-linked glycans at the conserved N-glycosylation site N297 (EU numbering system) in the heavy chain CH2 domain33. This N-glycosylation can exhibit both macro- (i.e., the N-glycan occupancy) and micro-heterogeneities (i.e., the N-glycan structure), which contribute to undesired mass heterogeneity of mAbs during MS analysis, resulting in increased spectral complexity and decreased detection sensitivity. This mass heterogeneity can pose significant challenges when analyzing dimer species in mAb cocktail samples, as the different hetero- and homodimers are not only present at low abundances but also frequently exhibit similar molecular weights, leading to overlapping mass profiles. To demonstrate the impact of N-glycosylation, a two-mAb fixed-dose co-formulation (FDC) sample (consisting of Ab1 and Ab2) aged at 5 °C for 3 months was subjected to nSEC-MS analysis to obtain the mass spectra of the dimers (Fig. 1A). Both the raw and deconvoluted mass spectra revealed three distinct dimer populations: the Ab1 homodimer, Ab1/2 heterodimer, and Ab2 homodimer, consistent with their predicted masses (Table S1). However, due to the mass heterogeneity from N-linked glycans, each dimer exhibited a broad cluster of mass peaks, leading to significant mass overlaps between the heterodimer and homodimers. Notably, as the heterodimer’s molecular weight fell between those of the two homodimers, its mass profile cannot be clearly resolved from either homodimer, rendering MS-based quantitation challenging. To tackle this challenge, PNGase F-mediated deglycosylation treatment was explored to remove the Fc N-glycans and convert various mAb glycoforms into a single deglycosylated population. To preserve the non-covalent dimers, the reaction was conducted under mild conditions without any denaturants or surfactants (45 °C for 1 h, pH 7.5). This treatment resulted in efficient removal of the Fc N-glycans, as evidenced by the decreased mass and reduced spectral heterogeneity in both monomer (Figure S2) and dimer mass spectra (Fig. 1). Despite the presence of satellite peaks in the mass spectra- attributed to post-translational modifications (PTMs) such as unprocessed C-terminal Lys and Lys glycation- the deglycosylated dimer species exhibited distinct mass profiles, enabling clear resolution and unambiguous peak integration for individual dimer quantitation. In addition to the enhanced mass separation, a concomitant improvement in MS intensity, signal-to-noise ratio (S/N), and spectral quality were also observed for the deglycosylated sample, presumably due to the signal consolidation from various mAb glycoforms following N-glycan removal. This further improved the method sensitivity, which was particularly critical for samples containing very low levels of dimers.

Fig. 1
figure 1

Raw mass spectra and the corresponding deconvoluted mass spectra of the dimer species from a two-mAb cocktail sample (A) without and (B) with PNGase F treatment. The cocktail sample was composed of 60 mg/mL Ab1 and 60 mg/mL Ab2 and stored at 5 °C for 3 months. The zoomed-in view of the charge state z = + 35 in each raw mass spectrum is shown as an inset.

Full size image

Although the deglycosylation reaction was conducted at elevated temperatures, we hypothesized that the formation of artifactual heterodimer during this process should be negligible, considering the short incubation time (i.e., 1 h) as well as the low mAb concentrations (≤ 10 mg/mL) used in this reaction. To assess the potential impact of this treatment on artifactual heterodimer formation, the individual Ab1 and Ab2 formulated drug substance (FDS) samples were mixed immediately before the deglycosylation treatment and then analyzed using nSEC-MS. As expected, no apparent mass peak corresponding to the heterodimer was observed in the freshly prepared Ab1 and Ab2 mixture sample, suggesting the selected deglycosylation conditions were sufficiently mild to avoid artifactual dimer formation (Figure S3A). For comparison, a positive control experiment was conducted using the aforementioned FDC sample of Ab1 and Ab2 (5 °C for 3 months), which successfully detected the heterodimer species (Figure S3A). Further, a similar study was performed using a three-mAb FDC sample and achieved the same conclusion (Figure S3B). In this case, the freshly mixed sample of individually formulated Ab3, Ab4, and Ab5 FDS (mixed at a 1:1:1 ratio) was compared against the corresponding FDC sample for dimer detection following deglycosylation treatment. Consistently, heterodimers were detected only in the FDC sample and were absent in the freshly mixed sample, further supporting the conclusion that the deglycosylation treatment did not introduce artifactual dimer formation. These findings demonstrate that the native deglycosylation treatment does not affect heterodimer quantitation and can be reliably implemented in subsequent dimer analyses.

Characterization of the heterodimer growth in FDC stability studies by nSEC-MS

To demonstrate its applicability in mAb FDC development, the nSEC-UV/MS method was implemented to monitor the heterodimer growth during stability studies of an FDC sample. The FDC sample, composed of 1:1 mixed Ab1 and Ab2, was placed under different storage/thermal stress conditions (5 °C, 25 °C, and 40 °C) for extended periods of up to 6 months. At designated time points, aliquots were collected, treated with native deglycosylation, and analyzed using nSEC-UV/MS. For quantitation of individual dimers, the integrated peak area of each dimer from the deconvoluted mass spectra (Fig. 2 insets) was initially used to estimate its relative abundance (i.e., homo- and heterodimer distribution) within the dimer population. This was combined with the UV-measured total dimer percentage to calculate the abundance of each individual hetero- or homodimer in the sample. Using this strategy, the heterodimer abundance was quantified for each sample, and its growth over time was plotted under different thermal conditions (Fig. 2). As expected, the heterodimer abundance increased over the course of stability studies. A positive correlation was observed between the rate of heterodimer growth and the storage/stress temperature: the fastest growth occurred at 40 °C (an average of 0.27% per month), moderate growth at 25 °C (an average of 0.08% per month), and the slowest growth at 5 °C (an average of 0.03% per month). Interestingly, by comparing the relative distribution of hetero- and homodimers in the deconvoluted mass spectra, it was revealed that the relative abundances of heterodimer were significantly higher in the stressed samples (e.g., 25 °C at 6-month and 40 °C at 3-month) than in the control sample (T0) (Fig. 2 insets). This suggests a greater propensity for heterodimer formation compared to the homodimers under thermal stress conditions. These findings highlighted the importance of monitoring heterodimer formation during mAb cocktail co-formulation development and demonstrated nSEC-UV/MS as a valuable tool to study this critical quality attribute.

Fig. 2
figure 2

The growth of the heterodimer abundance (%) in a two-mAb cocktail sample under different storage/stress temperatures. The cocktail sample was composed of 60 mg/mL Ab1 and 60 mg/mL Ab2. The insets show the distributions of hetero- and homodimers at T0 and different storage/thermal stress end points.

Full size image

Heterodimer analysis by High-Throughput SEC-MS (HT-SEC-MS)

Formulation development for mAb cocktail drug products is an intricate process and often involves comprehensive evaluation of the relationships between the formulation components (e.g., buffer, pH, excipients, etc.) and the product attributes through extensive stability studies (e.g., different temperatures, low/high pH, freeze-thaw, light exposure, agitation, etc.). These studies often produce a large number of samples, necessitating the use of high-throughput analytical methods. For nSEC-MS-based heterodimer analysis, while the deglycosylation reaction can be automated and performed in a multiplexed fashion, the nSEC-MS analysis itself remains a speed-limiting step, taking 20 min per sample. To address this limitation, we investigated strategies to reduce the analysis time by using a shorter SEC column (15 cm instead of 30 cm) and increasing the flow rate (0.6 mL/min compared to the standard 0.2 mL/min). Notably, the adoption of the previously established nLC-MS platform, which utilizes a split-flow design and a multi-nozzle emitter, was pivotal in accommodating the higher flow rate while maintaining sensitive MS measurements31.

Using a stressed FDC sample composed of Ab1 and Ab2 (25 °C, 20 months), the performance of this high-throughput nSEC-MS method (HT-SEC-MS) was evaluated and compared with the regular nSEC-MS method. As shown in Fig. 3, the HT-SEC-MS method completed the analysis in just 4 min, achieving a 5-fold improvement in throughput compared to the regular nSEC-MS method, which required 20 min per sample. Further, the two methods demonstrated comparable separation performance, consistently resolving four groups of UV peaks corresponding to the dimeric, monomeric, and low molecular weight (LMW) species of Ab1 and Ab2. The LMW species detected were identified as truncated mAb species (i.e., mAb losing an Fab arm and the Fab fragment) resulting from protein backbone clippings at the upper hinge region, which is common for IgG1 molecules under prolonged thermal stress conditions (25 °C, 20 months)34. Most importantly, the HT-SEC-MS method demonstrated baseline separation between the dimer and the monomer, which was essential for obtaining accurate UV-based quantitation of the total dimer and achieving sensitive MS measurements. Interestingly, the regular nSEC-MS method displayed partial separation of the Ab1- and Ab2-related size variants (e.g., split peaks for dimer, monomer, and LMWs), likely driven by the different secondary interactions between the mAb molecules and the column matrix due to their distinct surface properties. In contrast, this molecule-specific separation was not observed in the HT-SEC-MS method, suggesting less prominent protein-matrix interactions under the HT-SEC-MS conditions. The reduction in secondary interactions is likely attributed to the smaller surface area of the column matrix (due to the shorter column) and the increased mobile phase-to-stationary phase volume ratio (from the higher flow rate). It is worth noting that eluting different dimer species within a single confined UV peak is beneficial for quantitative analysis. It not only simplifies the integration of mass spectra across the entire dimer elution peak for deconvolution, but also minimizes ionization discrepancies, as all hetero- and homodimer species are all ionized under identical solvent/MS conditions.

Fig. 3
figure 3

Comparison of HT-SEC-MS method and regular nSEC-MS method in analyzing a stressed FDC sample composed of 60 mg/mL Ab1 and 60 mg/mL Ab2 (25 °C, 20 months). The inset shows the deconvoluted mass spectrum of the dimer peak and MS-based quantiation of the dimer distribution from each method.

Full size image

In addition to chromatography resolution, quantitative performance between the regular nSEC-MS and HT-SEC-MS methods was also evaluated. To this end, both methods were assessed for UV-based quantitation of the total dimer and MS-based quantitation of the dimer distribution within the dimer peak. Using the UV peak areas, the total dimer abundances (dimer%) in the FDC sample was found to be identical between the two methods, measured at 1.8%. Meanwhile, using the integrated MS peak areas from the deconvoluted dimer mass spectra, the relative distributions of Ab1 homodimer, Ab1/2 heterodimer, and Ab2 homodimer were also determined to be highly comparable between the HT-SEC-MS and the regular nSEC-MS methods (18.4%, 32.6%, and 49.0%, respectively compared to 16.1%, 31.9%, and 52.0%, respectively) (Fig. 3). Together, these results demonstrate that the HT-SEC-MS method delivers comparable quantitative performance to the regular nSEC-MS method, highlighting its potential to achieve both accurate and high-throughput quantitation of heterodimers to support mAb cocktail co-formulation studies.

Low-Resolution nSEC-MS (LR-SEC-MS) improves heterodimer detection sensitivity

As heterodimers are formed after the mixing of individual mAb samples, their relative abundances are usually low in the FDC samples under normal storage or mild stress conditions. In addition, the absolute abundances of the heterodimers are also highly dependent on the concentrations of the individual mAbs. For example, intravenous (IV) compatibility studies, which are often conducted during formulation development to support mAb cocktail co-administration, typically involve diluting and mixing individual mAbs at low concentration ranges (e.g., < 1 mg/mL)35,36. These low concentrations generally lead to very low levels of heterodimers, presenting significant challenges for MS detection. Therefore, strategies that can effectively improve the MS detection sensitivity are highly desirable for successful detection and quantitation of the low-abundance heterodimers, particularly in unstressed FDC samples and IV compatibility samples.

To improve the MS detection of dimer species, we explored a strategy known as “proteoform integration”, which was first reported by Nagornov et al. to enhance the drug-to-antibody ratio (DAR) measurement of antibody-oligonucleotide conjugates37. In this strategy, the Orbitrap-based MS data acquisition was conducted under low-resolution settings so that the related proteoforms from a single DAR species were consolidated, leading to significantly simplified spectra and an improved signal-to-noise ratio (S/N). In the case of mAb dimer species, despite deglycosylation treatment, multiple proteoforms persist due to the presence of various PTMs. Therefore, we hypothesize that a similar strategy may also benefit the MS detection of hetero- and homodimer species in mAb cocktail samples. Previously, we demonstrated that a mass resolution setting of 12,500 (at m/z = 200) was appropriate for typical native LC-MS analysis of mAbs, as it allowed for baseline resolution of common glycoforms (∆mass = 162 Da) for mAb monomers while providing excellent sensitivity for mAb variants (e.g., ~ 0.1% for mAb dimer)31. To determine if a lower resolution setting can further improve dimer detection, we compared nSEC-MS analyses at two resolution settings (R = 6,250 and R = 12,500) using a two-mAb cocktail sample. This mAb cocktail sample was co-formulated with 40 mg/mL Ab6 and 50 mg/mL Ab7 and was stored at −80 °C prior to the analysis. SEC-UV measurements estimated the total dimer abundance in this sample at 0.66%, which was further characterized by MS to consist of three dimers: Ab6 homodimer, Ab6/7 heterodimer, and Ab7 homodimer. From the native mass spectra, it was clear that both MS resolution settings led to the unambiguous detection and differentiation of the three dimers (Fig. 4A, left panels). At R = 12,500, each dimer exhibited multiple peak features corresponding to different dimer proteoforms, such as the unmodified form and those with various PTMs (e.g., unprocessed C-terminal Lys and Lys glycation). While these peak features offer insights into the heterogeneity of each dimer species, they complicate the dimer distribution measurement. In contrast, at R = 6,250, these proteoform-related peaks were absent, and each charge state displayed a single, unresolved m/z peak for each dimer species (Fig. 4A, insets). Considering that both resolution settings produced well-resolved mass spectra, MS-based relative quantitation of these three dimers should not be affected under these two conditions. Indeed, using the integrated mass peak areas from the deconvoluted mass spectra, the dimer distributions were measured similarly at R = 12,500 (26.4%, 15.5%, and 58.1%, respectively) and at R = 6,250 (25.7%, 18.0%, and 56.3%, respectively) (Fig. 4A, right panels). Importantly, a concomitant improvement in signal intensity and S/N ratio was observed for both the raw mass spectrum and the deconvoluted mass spectrum acquired at R = 6,250 compared to those at R = 12,500, highlighting the benefit of proteoform integration for enhancing MS detection sensitivity.

Fig. 4
figure 4

Comparison of raw mass spectra (left) and deconvoluted mass spectra (right) obtained at R = 12,500 and R = 6,250 for dimer species from (A) a two-mAb cocktail sample composed of 40 mg/mL Ab6 and 50 mg/mL Ab7 and (B) an IV compatibility sample composed of 1 mg/mL Ab6 and 5 mg/mL Ab7. The zoomed-in view of the charge state z = + 36 in each raw mass spectrum is shown as an inset. MS-based quantitation of the dimer distribution is indicated in each deconvoluted mass spectrum.

Full size image

This dimer proteoform integration strategy, as achieved by the LR-SEC-MS method, is particularly valuable for analyzing low-abundance dimer species in mAb cocktail samples. To showcase its effectiveness in improving dimer detection and quantitation, this method was subsequently utilized to study the heterodimer formation in an IV compatibility sample consisting of Ab6 and Ab7. Mimicking the co-administration process, this sample was prepared by mixing individual Ab6 and Ab7 drug products in IV diluent at concentrations of 1 mg/mL and 5 mg/mL, respectively. The mixture was stored at 5 °C for 24 h followed by room temperature storage overnight. Using SEC-UV analysis, the total dimer abundance in this sample was measured at 0.36%; however, only a very small fraction was anticipated to arise from Ab6/7 heterodimer, due to the lower concentration of Ab6. Indeed, native mass spectra of the dimer peak revealed that Ab7 homodimer dominated (> 75% relative abundance), while the relative abundances of the other two species (Ab6 homodimer and Ab6/7 heterodimer) were significantly lower (Fig. 4B). Notably, at R = 12,500, the raw mass spectrum showed a poor S/N ratio for the charge states of both Ab6 homodimer and Ab6/7 heterodimer. This is problematic as it may lead to under-representation of these species in the deconvoluted mass spectra, thus compromising the accuracy of MS-based quantitation (Fig. 4B, top panels). In contrast, through dimer proteoform integration, the LR-SEC-MS method (R = 6,250) led to a significant improvement in both overall signal intensity and S/N ratio for all three dimer species. This enhancement enabled confident spectral deconvolution (Fig. 4B, bottom panels), leading to more reliable quantitation of the dimer distribution. In this example, the relative abundance of Ab6/7 heterodimer within the total dimer peak was measured at 9.6% at R = 6,250, compared to 6.9% at R = 12,500. The latter measurement likely underestimated the heterodimer due to its poor detectability. Given that the total dimer abundance in this sample was only 0.36%, the absolute abundance of Ab6/7 heterodimer was estimated at just 0.035% (equivalent to 3.5 ng, based on a total injection amount of 10 µg), highlighting the excellent dynamic range/sensitivity achieved by this method.

Differentiation and quantitation of dimer species in Three-mAb cocktails by Immunodepletion-Assisted SEC-MS (ID-SEC-MS)

To achieve the desired therapeutic outcome, antibody cocktail products sometimes contain more than two mAbs38. For instance, REGN-EB3 (Inmazeb), the first Food and Drug Administration (FDA) approved therapeutic for treating EBOV infection, contains three different mAbs each targeting different epitopes of the virus39,40. In such a three-mAb system, heterodimer characterization is expected to become more challenging due to the increased complexity arising from the many possible dimer species present in the mixture samples (e.g., either during co-formulation development or during IV compatibility assessment for co-administration). For example, while a two-mAb cocktail sample may contain up to three dimer species (i.e., two homodimers and one heterodimer), a three-mAb cocktail sample could potentially contain six different dimer species (i.e., three homodimers and three heterodimers). The larger number of dimer species is expected to pose a significant challenge for their differentiation by mass-based measurement. This is particularly true considering the similar molecular weights of mAb molecules (i.e., ~ 150,000 Da) and the broad mass profiles of each dimer (i.e., 500-1,000 Da) due to the presence of multiple proteoforms. In many cases, two or more dimer species within the same mixture sample may exhibit overlapping mass profiles, rendering MS-based differentiation and quantitation impossible.

To cope with the challenges associated with dimer analysis in three-mAb cocktail samples, we developed an immunodepletion-assisted SEC-MS (ID-SEC-MS) workflow. This method leverages resins coated with an anti-drug Ab that specifically binds to one mAb component in the cocktail, enabling the selective removal of multiple dimer species containing that mAb. By simplifying the dimer complexity in the sample, this approach facilitates MS-based differentiation and quantitation of the remaining dimer species. Furthermore, combining the results from the untreated sample allows for back-calculation of the abundances of the removed dimer species (Fig. 5A). To illustrate this workflow, a three-mAb cocktail sample (60 mg/mL each of Ab8, Ab9, and Ab10, stored at 25 °C for 1 month) was analyzed. In the initial dimer mass spectrum, due to mass profile overlaps between dimers with close molecular weights, the six possible dimer species can only be resolved into four distinct mass groups for relative distribution quantitation, including (1) Ab8 homodimer and Ab8/9 heterodimer (27%), (2) Ab9 homodimer (55%), (3) Ab8/10 heterodimer and Ab9/10 heterodimer (9%), and (4) Ab10 homodimer (10%) (Fig. 5B, top panel). To resolve the contributions of individual dimers within Group 1 and Group 3, immunodepletion using immobilized anti-Ab9 antibody (A9A) was applied to the cocktail sample. This treatment selectively removed all Ab9-containing species, including Ab8/9 heterodimer (Group 1) and Ab9/10 heterodimer (Group 3), thereby eliminating their interference. This enabled accurate quantitation of the remaining dimers: Ab8 homodimer in Group 1 and Ab8/10 heterodimer in Group 3. As expected, the nSEC-UV/MS analysis demonstrated complete removal of Ab9 monomer from the immunodepleted cocktail sample (Figure S4). Additionally, the UV chromatogram showed notable changes in the dimer profile and reduced abundances following immunodepletion, presumably resulting from the removal of Ab9-containing dimers (Figure S4). This was subsequently confirmed by MS analysis of the dimer peak, which showed a greatly simplified spectrum with the remaining three dimers- Ab8 homodimer, Ab8/10 heterodimer, and Ab10 homodimer- well resolved from each other (Fig. 5B, bottom panel). Using the mass peak areas, the abundance ratios between these three dimer species can be readily established. As Ab10 homodimer (10%) was reliably quantitated in the untreated sample without interference, it can be used as a reference to deduce the abundances of the other two dimers. Specifically, using the measured ratios of [Ab8 homodimer]-to-[Ab10 homodimer] and [Ab8/10 heterodimer]-to-[Ab10 homodimer] from the immunodepleted cocktail sample, the relative abundances of Ab8 homodimer and Ab8/10 heterodimer in the untreated cocktail sample were deduced to be 16% and 5%, respectively (Fig. 5B). Subsequently, through simple subtraction, the relative abundances of Ab8/9 heterodimer and Ab9/10 heterodimer were back-calculated to be 11% (i.e., 27%−16%) and 4% (i.e., 9%−5%), respectively. Collectively, the individual dimer abundances of all six dimer species in this three-mAb cocktail sample can be determined by this immunodepletion-assisted nSEC-MS (ID-SEC-MS) method (Fig. 5B). Lastly, this ID-SEC-MS method may also be applicable to mAb cocktails comprising of more than three mAbs. However, multiple rounds of immunodepletion may be needed to disentangle individual dimer contributions in more complex mixtures.

Fig. 5
figure 5

(A) The schematic of the immunodepletion-assisted SEC-MS (ID-SEC-MS) workflow for dimer quantitation for a stressed three-mAb cocktail sample. The three-mAb cocktail sample was composed of 60 mg/mL Ab8, 60 mg/mL Ab9, and 60 mg/mL Ab10 (25 °C, 1 month). (B) Deconvoluted mass spectra of the dimers in the untreated cocktail sample (top) and immunodepleted cocktail sample using anti-Ab9 antibody (bottom). The relative abundances of all six dimer species were obtained and listed.

Full size image

Conclusion

Antibody cocktails offer unique therapeutic benefits compared to antibody monotherapies; however, they also present distinct analytical challenges. Heterodimers, formed via inter-molecular interactions between two different mAbs upon co-formulation or co-administration, are important quality attributes to monitor during the development of mAb cocktails. In this study, we report the development and evaluation of a broadly applicable nSEC-MS method for highly effective analysis of dimers in mAb cocktails. Various sample treatment strategies were developed to facilitate dimer differentiation and quantitation in both two-mAb and three-mAb cocktails. In addition, key parameters in SEC separation and MS data acquisition were explored to enhance sensitivity and throughput.

First, the quantitative performance of the nSEC-UV/MS method using an MS-compatible mobile phase (150 mM ammonium acetate) was evaluated, which demonstrated results highly comparable to those obtained with conventional SEC mobile phases. Secondly, a mild deglycosylation step was introduced to reduce mass heterogeneity from Fc N-glycans, significantly improving the differentiation and quantitation of hetero- and homodimers without inducing artifactual dimer formation. This workflow was applied to investigate heterodimer growth in a two-mAb cocktail subjected to varying storage/thermal stress conditions, revealing a direct correlation between heterodimer growth rate and stress levels. Furthermore, to accommodate the large sample size from mAb co-formulation studies, an HT-SEC-MS method was developed, offering a 5-fold increase in throughput compared to the regular method while maintaining comparable quantitative performance. Additionally, a strategy of integrating dimer proteoforms through low-resolution MS data acquisition proved effective in enhancing the detection of mAb dimers. Lastly, to cope with the increasingly complex dimer mixtures, such as those present in three-mAb cocktails, an immunodepletion-assisted strategy was developed. This method was successfully applied to a three-mAb cocktail sample, enabling relative quantitation of all six dimer species despite overlapping mass profiles.

Our research demonstrated that the newly developed nSEC-MS method, along with these analytical strategies, enables sensitive and high-throughput identification and quantitation of hetero- and homodimers in mAb cocktail samples. To the best of our knowledge, this method represents the first analytical approach capable of providing unambiguous and sensitive analysis of heterodimers directly from the mAb cocktail DP samples, making it highly useful to support mAb cocktail therapy development. Notably, like other MS-based relative quantitation strategies (e.g., quantitation of post-translation modifications or drug-to-antibody ratios in ADCs), reliable measurement of dimer distributions within mAb cocktail samples assumes similar ionization efficiency of different dimer species. While this assumption is generally valid due to the similar nature and size of mAb dimers, caution is advised in specific cases. Additionally, while the immunodepletion strategy effectively addresses dimer differentiation challenges when mass separation is insufficient, its applicability is constrained by the availability of specialized reagents capable of highly specific interactions with the target mAb. Lastly, it is important to note that the method developed and evaluated in this study are currently intended for extended product characterization, which supplements SEC-UV-based quality control assays by providing detailed dimer information. Further method validation is required before extending its utility into quality control environment.

Materials and methods

Chemicals and reagents

All mAbs (Ab1-Ab10), anti-drug antibody (against Ab9), and the fixed dose co-formulation (FDC) products were produced at Regeneron Pharmaceuticals, Inc. (Tarrytown, NY). Ammonium acetate and dimethyl sulfoxide (DMSO) were purchased from Sigma-Aldrich (St. Louis, MO). EZ-Link™ NHS-Biotin, Pierce™ High Capacity Streptavidin Agarose, Pierce™ micro-spin columns, formic acid (FA), and UltraPure 1 M Tris-HCl buffer, pH 7.5 were purchased from Thermo Fisher Scientific (Waltham, MA). PNGase F was purchased from New England Biolabs (Ipswich, MA).

Deglycosylation of co-formulated antibodies (Abs)

Deglycosylation treatment was performed by mixing each mAb cocktail sample with PNGase F (1 IUB milliunit per 10 µg of protein) and incubating at 45 °C for 1 h in 50 mM Tris-HCl buffer (pH 7.5) to remove the Fc N-glycans. After the deglycosylation treatment, samples were centrifuged at 14,000 g for 5 min. The supernatants were then transferred to HPLC vials and ready for native SEC-MS analysis.

Native SEC-UV-MS analysis

Native SEC chromatography was performed on an UltiMate 3000 UPLC (Thermo Fisher Scientific, Bremen, Germany) equipped with a Waters Acquity BEH protein SEC column (4.6 × 300 mm, 1.7 μm, 200 Å; Milford, MA) with the column compartment set to 30 °C. An injection amount of 6 to 20 µg of the mAb cocktail samples was used for each analysis. SEC separation was achieved using an isocratic mobile phase containing 150 mM ammonium acetate (pH 6.8) operated at a flow rate of 0.2 mL/min. The post-column flow was split into a microflow (< 10 µL/min) for nano-electrospray ionization MS detection and a remaining high flow for UV detection (280 nm). A Thermo Q-Exactive UHMR mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a microflow-nanospray electrospray ionization (MnESI) source and a M3 multinozzle emitter (Newomics, Berkeley, CA) was used for MS data acquisition. MS resolution settings of 12,500 or 6,250 (at m/z = 200) were used for data acquisition. The scheme of the nSEC-UV/MS platform interface is depicted in Scheme S1. Further details on the experimental setup and instrument parameters can be found in a previous publication31.

For the high-throughput SEC-MS (HT-SEC-MS) method, a 150 mm Acquity BEH protein SEC column (4.6 × 150 mm, 1.7 μm, 200 Å; Milford, MA) was used with an increased mobile phase flow rate at 0.6 mL/min for mAb size variant separation. All other operation parameters were kept the same as regular SEC-MS methods described above.

Immunodepletion using anti-drug antibody for the three-mAb cocktail sample consisting of Ab8, Ab9, and Ab10

The anti-Ab9 Ab (A9A) sample was biotinylated by mixing the 10 mM NHS-biotin solution and the 10 mg/mL A9A solution at 12 : 1 molar ratio followed by incubating at room temperature for 30 min. The reaction mixture was subsequently buffer exchanged into 100 mM Tris-HCl (pH 7.5) solution using an Amicon 30 kDa MWCO ultra centrifugal filter (MilliporeSigma, Burlington, MA). The final protein concentration of biotinylated anti-Ab9 Ab was determined by Nanodrop 2000 spectrophotometer (ThermoFisher Scientific, Waltham, MA USA) at 280 nm. Immobilization of A9A was then prepared using a published protocol41 in a microspin column. Specifically, 2.5 nmole of the biotinylated A9A in 50 µL 100 mM Tris-HCl was mixed with 50 µL of high capacity streptavidin resin (Thermo Fisher Scientific, Rockford, IL, equivalent to 100 µL of resin slurry) in an empty microspin column and incubated for 15 min under room temperature. Subsequently, the A9A-immobilized resin was washed three times, each time with 200 µL of Tris-HCl buffer (100 mM, pH 7.5) and centrifuged at 3,000 g for 1 min to remove the excess buffer from the resin. For immunodepletion experiment, 20 µL of the cocktail sample composed of Ab8, Ab9, and Ab10 (10 mg/mL) was added to the A9A-immobilized resin in the microspin column and incubated at room temperature for 45 min. Following incubation, the supernatant (flow-through) fraction was collected by centrifuging the microspin column at 3000 g for 1 min. The collected flow-through sample, together with one aliquot of the untreated cocktail sample, were then subjected to native deglycosylation followed by nSEC-MS analysis.

Data analysis

The UV peak area was integrated using Xcalibur software (Thermo Fisher Scientific, Waltham, MA USA). Deconvoluted MS spectra were generated using IntactMass™ software (Protein Metrics, Boston, MA USA). To calculate the total dimer abundance in each sample, the UV-based peak area of the dimer was compared to the combined peak areas of both dimer and monomer peaks. The relative abundance (distribution) of each dimer species within the total dimer peak of a mAb cocktail sample was calculated using its integrated mass peak area and compared against the summed mass peak areas of all dimers. Lastly, the absolute abundance of each dimer in the cocktail sample was calculated by multiplying the UV-based total dimer abundance with the MS-determined relative abundance of each dimer.