Direct-MS analysis of antibody-antigen complexes

Abstract

In recent decades, antibodies (Abs) have attracted the attention of academia and the biopharmaceutical industry due to their therapeutic properties and versatility in binding a vast spectrum of antigens. Different engineering strategies have been developed for optimizing Ab specificity, efficacy, affinity, stability and production, enabling systematic screening and analysis procedures for selecting lead candidates. This quality assessment is critical but usually demands time-consuming and labor-intensive purification procedures. Here, we harnessed the direct-mass spectrometry (direct-MS) approach, in which the analysis is carried out directly from the crude growth media, for the rapid, structural characterization of designed Abs. We demonstrate that properties such as stability, specificity and interactions with antigens can be defined, without the need for prior purification. This ty.

1. Introduction

Since the first commercial monoclonal antibody (Ab) was approved in 1986, Abs have become a powerful tool for treating various diseases. Out of the five main classes of Ab isotypes (IgA, IgD, IgE, IgG and IgM), monoclonal IgG form the vast majority of Abs in clinical and diagnostic applications¹. To date, 79 different monoclonal Abs are in clinical use². These Y-shaped heteromeric IgGs (150 kDa) consist of four heavily glycosylated polypeptide chains—two identical heavy chains (50 kDa) and two identical light chains (25 kDa). Considerable variability is allowed within the amino acids that populate the antigen-binding surfaces of IgGs, underlying their remarkable binding versatility³.

Given the integral role of Abs in modern research and medicine, it is important to develop high-throughput methods to structurally characterize them and measure their ability to bind target antigens⁴. Yet structural characterization of Abs can be very challenging, since they are highly heterogeneous proteins with a wide range of post-translational modifications (PTMs), including glycosylation⁵, oxidation, sulfonation, and deamination⁶. To probe Ab-antigen interactions, biophysical methods such as surface plasmon resonance⁷ or isothermal titration calorimetry⁸ can be used. Nuclear magnetic resonance⁹, X-ray crystallography¹⁰ and cryo-electron microscopy¹¹ provide structural details and information on antigen binding, but are hampered by low-throughput and homogeneous sample requirements.

Native MS is an emerging structural biology technique that is commonly used to study both the structural characteristics of Abs and their interactions with antigens¹². Unlike proteomic methods that involve the modification, denaturation or digestion of the sample prior to analysis, native MS measures intact assemblies that are ionized in their native states from a non-denaturing solution¹³. The method provides information on the composition, stoichiometry, subunit architecture and topological organization of the Ab complex. One of the advantages of native MS stems from its ability to detect all the sub-populations present in a heterogeneous mixture of proteins and complexes in a one-shot experiment^{14, 15}. This property is extremely important in the analysis of Abs that are known to be differentially modified by various PTMs^{16, 17}. Coupling ion mobility (IM) separation with native MS experiments further enhances the information obtained. As the time that it takes for an ion to travel through a tube densely filled with an inert gas depends not only on the mass but also on the shape of the analyzed protein complex, this knowledge provides insight into its packing and topology^18–20.

The last 20 years have seen the extensive use of native MS in the structural analysis of Abs and their complexes with antigens^{21, 22}. This includes the challenging characterization of N-linked glycosylation, the most abundant PTM in Abs²³. Native MS has also been applied to the analysis of the glycosylation patterns of therapeutic Abs^{24, 25}. Multiple other studies have applied it to investigating the interaction of Abs with different molecules, such as drugs and antigens^{12, 26–28}. Another viable option is characterizing the conformations and stability of Abs using IM-MS in combination with collision induced unfolding (CIU)²⁹. This approach was also used to determine the effect of different PTMs, drug binding and mutations on the overall stability of target Abs^{27, 30, 31}. However, all these studies were carried out with purified Abs, with substantial costs in time and labor invested in the purification process.

In this study, we demonstrate the ability of an emerging native MS approach named direct-MS to characterize the structural stability and interactions of recombinant Abs with their native antigens. The great advantage of this method is that it allows the in-depth structural analysis of secreted Abs directly from crude samples without the need for protein purification^32–35. Recombinant Abs are typically secreted from eukaryotic cell cultures, grown in defined media and quickly accumulate to become the major proteins in the culture. Taking advantage of this feature, we analyzed Abs from the crude growth medium with direct-MS, with the only required adjustment being the replacement of the medium with a volatile, MS-compatible solution³². We were able to demonstrate that the computationally optimized anti-vascular endothelial growth factor (VEGF) Ab G6^des13 is more stable than its parental Ab, G6. We also used direct-MS to characterize the interactions between the computationally optimized anti-lysozyme Ab D44.1^des and its target antigen. The designed versions improved Ab stability, expression levels and affinity, through the introduction of a few core mutations. These interesting case studies therefore demonstrate the ability of direct-MS to assess distinct Ab characteristics in the same experiment. To determine direct-MS’s utility in specificity analysis we monitored the capacity of the monoclonal Ab MCP21 to bind ortholog 20S proteasome complexes (20S is the ~720 kDa, 28-subunit, catalytic degradation machinery^36–38). Overall, we show that measurements of the crude culture medium of Ab-secreting cells provides in-depth characterization of different Abs. Our findings reflect the ability of direct-MS to be an integral part of Ab development workflows, considerably accelerating steps of Ab production, specificity, and stability.

2. Experimental Section

Antibody production

The variable regions of the heavy and light chains of the G6, G6^des13 and the D44.1^des chimeric Abs were cloned into p3BNC plasmids in frame with the human IgG secretion signal peptide. Suspension-grown HEK293F cells were grown in FreeStyle medium (Gibco), in a shaking incubator, at 37 °C, in a controlled environment of 8% CO₂, at 115 rpm. Plasmids encoding the light and heavy chains were co-transfected into cells at a cell density of 106 cells/mL, by linear 40 kDa polyethylenimine (PEI) (Polysciences), at 3 mg of PEI per 1 mg of plasmid (0.5 mg of the light chain plasmid and 0.5 mg of the heavy chain plasmid), per 1 L of culture. Growth media were collected after 5–7 days.

The MCP21 hybridoma (Sigma # 96030418), expressing a monoclonal IgG Ab against the human 20S proteasome subunit PSMA2, was grown at the Ab production unit of the Weizmann Institute of Science. Cells were cultured in a 250 ml flask, in high-glucose Dulbecco’s modified Eagle’s medium supplemented with 10% horse serum, penicillin-streptomycin, glutamine and sodium pyruvate (Biological Industries), according to the manufacturer’s instructions, in a humidified incubator, at 37 °C and in a controlled environment of 8% CO₂. For the production of Abs, 12x10⁷ cells were collected by centrifugation and transferred into 35 ml of DCCM-1 medium (Defined Cell Culture Medium, a high-protein, serum-free medium for hybridoma cell growth and monoclonal antibody production, Biological Industries, cat # 05-010-1), supplemented with penicillin-streptomycin, glutamine and sodium pyruvate. The cells were grown in a 35 ml miniPerm© bioreactor classic production module, topped with 400 ml of the same medium in the nutrient module (Greiner # 9600 1059), at 40 rpm, on a turning device (Greiner # 9600 1061). The nutrient medium was replaced every two days. The Ab-containing medium (~25 ml) was collected from the production module after one week.

20S proteasome purification

20S proteasomes from rat liver, HEK293 and yeast cells were purified as previously described³⁹.

Sample preparation

Samples were prepared for native MS experiments as previously described³⁴. Briefly, the growth media were cleared from cells by centrifugation at 600 g for 10 min. To avoid protein degradation, the HEK293F growth medium was mixed with 0.02% (w/v) sodium azide and 1 mM phenylmethylsulfonyl fluoride (PMSF). The culture media were further clarified by centrifugation at 10,000 g for 15 min to remove insoluble contaminants. Samples were then aliquoted, snap-frozen in liquid nitrogen and stored in −80°C. For CIU experiments, Abs were reduced for 4h at 37 °C using 20 mM tris(2-carboxyethyl)phosphine (TCEP) on the day of the measurement.

Prior to the native MS analysis, the media were buffer-exchanged twice into 1M ammonium acetate and a third time into 150 mM ammonium acetate, using Micro Biospin 6 columns (BioRad # 7326222). The different 20S proteasomes were buffer-exchanged once into 150 mM ammonium acetate³⁹ and diluted to 3 μM. For binding assays, the buffer-exchanged MCP21 medium was diluted 10-fold with 150 mM ammonium acetate and mixed 1:1 with 3 μM of the different proteasome samples. The anti-lysozyme D44^des growth medium was mixed with lysozyme (Sigma L6876) at a concentration ranging from 0.25–11 μM.

Native MS analysis

Native MS measurements were performed using three different instruments: a modified Q Exactive Plus Orbitrap EMR⁴⁰ (Thermo Fisher Scientific, Bremen, Germany), a UHMR Orbitrap (Thermo Fisher Scientific, Bremen, Germany) and a Synapt G1 HDMS (Waters MS Technologies, Manchester, UK), modified for optimal transmission of high-mass proteins⁴¹. Typically, 3 μL of solution was electrosprayed from gold-coated borosilicate capillaries prepared in-house as described⁴². All instruments were externally mass-calibrated using a cesium iodide solution at a concentration of 2 mg/mL.

Analyses of the D44.1^des Ab were performed on the Q Exactive Plus Orbitrap EMR using the following instrument parameters: capillary voltage 1.7 kV, inlet capillary was set to 180 °C, fore vacuum pressure 1.5 mbar, and trapping gas pressure 4.1, corresponding to a HV pressure of 1.2 × 10⁻⁴ mbar and a UHV pressure of 3.4 × 10⁻¹⁰ mbar. The source was operated at a constant energy of 2 V in the flatapole bias and interflatapole lens. The bent flatapole DC bias and gradient were set to 2.5 and 35 V, respectively, and the HCD cell was operated between 10–50 V. For tandem MS experiments, a single charge state was isolated using an isolation window of 20 m/z and further activated by elevating the HCD energy to 150 V.

MS measurements of the MCP21 Ab and the 20S proteasome were conducted using the UHMR Orbitrap, with the following instrument parameters: Capillary temperature 250 °C, capillary voltage 1.35 kV, fore vacuum pressure 1.6 mbar, trapping gas pressure 7, corresponding to a HV pressure of 2.6 × 10⁻⁴ mbar and a UHV pressure of 2.7 × 10⁻¹⁰ mbar. The detector m/z optimization and ion transfer target m/z were set to a high, desolvation voltage, −50 V. The bent flatapole DC bias and gradient were set to 2.0 and 30 V, respectively, and the HCD cell was operated at 50-70 V, unless indicated otherwise.

Ion mobility MS

IM-MS measurements were performed on the Synapt G1 HDMS instrument. Measurements of the G6 and G6^des13 Abs were preformed using the following parameters: Capillary voltage 1.2–1.7 kV, sampling cone 10-50 V, extraction cone 2-5 V, source temperature 25 °C, trap cell collision energy 10 V, transfer cell collision energy 10 V, and DC bias 18–22 V. Nitrogen was used as the IM-MS gas, at a flow rate of 22 mL/min and trap cell gas flow was set to 8 mL/min. The IM wave velocity was set to 250 m/s, and the wave height was set to 10, 11 or 12 V. For CCS measurements, T-wave calibration was performed as previously described, using Concanavalin A (103 kDa), alcohol dehydrogenase (143 kDa) and glutamate dehydrogenase (336 kDa) as calibrants⁴³. CCS values of the G6 and G6^des13 are reported for the three most intense charge states. For D44.1, the CCS is reported for the 23⁺ - 25⁺ charge states of the unbound Ab and for the 24⁺ - 26⁺ charge state of the fully bound Ab. Standard errors of the CCS measured values were calculated from three independent measurements. All instrumental parameters were kept constant during IM-MS data acquisition and T-wave calibration. For collision induced unfolding (CIU) experiments, proteins were activated in the trap region by increasing the trap voltage from 25 V to 220 V in steps of 5 V. CIU analyses were performed on the 21⁺, 22⁺ and 23⁺ charge states of the G6 and G6^des13 Abs, drift time values were extracted from the apex of each.

The instrumental settings for MCP21-20S proteasome complex measurements were adjusted for the measurement of high molecular weight complexes as follows: capillary voltage 1.2–1.4 kV, sampling cone 80 V, extraction cone 20 V, source temperature 25 °C, trap and transfer collision energies 50 V and 10 V, respectively, DC bias 20 V. Nitrogen was used as the IMS gas, at a flow rate of 22 mL/min. The gas flow in the trap was set to 10 mL/min, corresponding to a pressure of 9.35 × 10⁻² mbar in the trap cell. The IM wave velocity was set to 210 m/s, and the wave heights to 10-12 V. The backing pressure was set to 7 mbar.

For the CCS measurements of the MCP21-20S proteasome complexes, T-wave calibration was preformed using Concanavalin A, alcohol dehydrogenase, pyruvate kinase and GroEL. The analysis was performed for the most intense charge states of each specie: 20S proteasome (53⁺-61⁺), MCP21/20S (58⁺-64⁺), MCP21₂/20S (64⁺-69⁺), MCP21₂/20S₂ (lower m/z specie) (95⁺-104⁺), MCP21₂/20S₂ (higher m/z specie) (83⁺-91⁺). Errors represent standard deviation of the three independent experiments. The CCS measurements were performed at instrumental settings that allow the measurement of high molecular weight complexes, as detailed above. GroEL (802 kDa) was the largest utilized IM-MS calibrant used for CCS calibration. However, given that GroEL is significantly smaller than some of the Ab/20S proteasome complexes, the CCS values of the high molecular weight MCP21-proteasome complexes are only estimates.

Data Analysis

The spectra of the different G6 and G44.1 Abs were examined and analyzed using MassLynx software (Waters V4.2 SCN982, 2017). IM-MS data were analyzed using the PULSAR⁴⁴ and Driftscope™ HDMS™ V2.8 software (Waters, Hertfordshire, U.K.). Spectra acquired on the Orbitrap instruments were converted to MassLynx-compatible files using the DataBridge software (Waters). The masses of the different MCP21 Ab and 20S proteasome populations were determined by the computational suite UniDec v. 4.1.1⁴⁵. For each charge state series, deconvolution was performed in the relevant m/z range, covering all the charge states in the series. Mass ranges were set to 15,000–50,000 Da around the calculated mass of each species. Mass sampling was set to 10 Da. The peak detection range was set to 100–500 Da and the peak detection threshold to 0.1. Every measured mass in UniDec was manually inspected to ensure correct assignment. Reported masses are the average of at least 3 independent spectra. Standard deviations reflect the error of the measurements. For deconvolution analysis, the whole spectrum was included (11,000 – 22,000 m/z). Mass sampling was set to 100 Da, with a peak full width at half-max of 66. The peak detection range was set to 800 Da and the peak detection threshold to 0.052.

3. Results

Comparative gas-phase stability assessment of engineered and parental Abs performed directly from the growth medium of secreting human cells

The computational design of Abs with improved biophysical properties is an important goal in protein engineering⁴⁶. A multipoint mutant of the G6 Ab, which targets human VEGF, was computationally designed (G6^des13) using an automated method called AbLIFT, which optimizes Abs expression levels, stability and affinity⁴⁷. The G6^des13 variant comprises six mutations at the light-heavy chain interface of the variable domain to improve the contacts across this interface and backbone rigidity⁴⁷. We used direct-MS to characterize the differences between the parental and mutant Abs by recording their native IM-MS spectra directly from their crude growth media.

The measured mass of the parental and designed Abs were almost identical, 147,541 ± 72 Da and 147,543 ± 70 Da, respectively (Fig. S1A). Theoretical masses of the antibodies, according to their amino acid sequences, are 144,969 Da for the parental Ab and 144,943 Da for G6^des13. The mass difference between the theoretical and measured masses is attributed to two glycosylations, i.e. G0F (1,462 Da) and G0 minus GlcNac (1,113 Da). Typically, the computational design process leads to a significant improvement in expression levels of the designed antibodies⁵⁰. Indeed, the designed G6^des13 displayed a higher expression level, compared to its parental G6 Ab. This is reflected by the increased ratio between the peak heights of the Abs and the background protein LDH charge states, indicating a ratio of 0.6 and 1.8 for G6 and G6^des13, respectively (Fig. S1A). IM-MS measurements indicated that both Abs display similar (CCS) values, 7,381 ± 150 Ų for G6 and 7,333 ± 147 Ų for G6^des13 (Fig. S1B, D).

In order to compare the gas-phase stability of the two antibodies, we used collision induced unfolding (CIU) analysis. In this type of experiment, the energy in the collision cell is elevated in a stepwise manner, causing protein activation that may consequently induce conformational change. The collision voltage at which the transition between conformations occur, the mode of the transition and the arrival time distribution generate a characteristic unfolding trajectory of the protein²⁹. The CIU experiments were conducted after reduction of the two antibodies, in order to remove the structural constrains inflicted by the cysteine bonds.

IM-MS measurements revealed that the CCS values remained the same following reduction (7,322 ± 150 Ų for G6 and 7,370 ± 130 Ų for G6^des13), indicating that the overall shape of the Abs is sustained (Fig. S1C,D). Following CIU measurements, we extracted drift times values from the apex of each charge state (21⁺- 23⁺) in the different collision energies and plotted them as a function of collision energy (Fig. 1). The results indicated that during CIU activation, the Abs underwent two transitions, displaying three distinct CIU features. Both antibodies displayed similar drift times at low activation energies, differing from each other by 1.7%, 1.3% and 3.1%, for the 21⁺, 22⁺ and 23⁺ charge states, respectively. However, clear differences in drift time values were observed between the parental and designed Abs upon collision voltage elevation, wherein the parental Ab displayed larger drift time values. The gap between the two Abs increased in each transition, where in the second transition drift time, differences between the parental and designed Abs reached 6.0%, 5.4% and 6.9%, for the 21⁺, 22⁺ and 23⁺ charge states, respectively, indicating the higher sensitivity of the parental Ab to collision activation.

Figure 1 – CIU plots reflect the differences in gas-phase stability between the parental and designed G6 Abs.

The two Abs were reduced with TCEP and subjected to CIU analysis. Drift time values were extracted from the apex of each MS peak and plotted as a function of collision voltage. Data is shown for the 21⁺, 22⁺ and 23⁺ charge states. The overall unfolding pattern of the two Abs is similar and consists of two transition states. However, the increase in acceleration voltage leads to larger shifts in drift time values of the parental Ab in comparison the designed G6^des13, indicating increased gas-phase stability of the latter.

Taken together, the G6^des13 Ab exhibited higher tolerance to the applied collision energy, emphasizing its increased gas-phase stability in comparison to the parental Ab, which is in good agreement with our previous analysis of this Ab using thermal-melt experiments⁴⁷. Moreover, the measurements demonstrate that although both Abs are highly similar in mass and CCS, their CIU fingerprint is distinct, making it a useful tool in the Ab design process.

Characterizing antibody/antigen interactions by direct-MS

As Abs function by binding their target antigens, we next turned to characterizing Ab-antigen interaction⁴⁸. We examined the growth medium of HEK293F cells, expressing the anti-lysozyme designed variant D44.1^des (theoretical mass of 144,433 Da), upon spiking into the growth medium this Ab’s target antigen, purified commercial hen-egg lysozyme (theoretical mass of 14.3 kDa) (Fig. S3). We conducted a titration assay using increasing concentrations of lysozyme, ranging from 0.25 to 11 μM, and plotted the relative abundance of the bound lysozyme (Fig. 2A, B). At lysozyme concentrations lower than 5 μM, we identified a major charge state series with a calculated mass of 147,026 ± 1.9 Da corresponding to the unbound form of D44.1^des modified with two GF0 (1,462 Da) moieties. Another minor charge state series, with a measured mass of 147,643 ± 1.0 Da, corresponded to the addition of two fucose (2x146 Da) and two mannose (2x162 Da) moieties. At higher lysozyme concentrations, two additional series emerged, one with a calculated mass of 161,325 ± 3 Da, corresponding to a single-bound D44.1^des (D44.1^des/lysozyme₁), and a second with a mass of 175,635 ± 12 Da, consisting of double-bound antigen (D44.1^des/lysozyme₂) (Figs. 2A and S3B). At 10 μM, the major charge state corresponded to the fully saturated, double-bound Ab.

Figure 2 – Ab/antigen interactions can be probed in the context of the crude growth medium, as demonstrated for the anti-lysozyme D44.1^des Ab.

(A) Increasing concentrations of lysozyme were titrated into the D44.1^des growth medium, giving rise to the gradual occupation of the Ab’s binding sites. An extended view of the 6360–6540 m/z range is shown. The additional charge state series seen to the right of the D44.1^des peaks, correspond to a glycoform with a 620 Da mass shift, indicating the addition of two fucose and two mannose moieties. (B) Quantification of the relative abundance of the unbound (blue line), single-bound (orange line) and double-bound D44.1^des (grey line) Abs. The results are averaged over three independent measurements. Error bars represent standard deviations.

To validate the composition of the Ab/antigen₂ complex, we applied a tandem MS experiment⁴². The 26⁺ charge state of this charge series was isolated in the quadrupole mass filter using a 20 m/z window (Fig. S3C) and activated within the HCD cell at a collision voltage of 120 V (Fig. S3D). Upon HCD activation, the complex dissociated into its composing subunits: a monomer, reflecting the lysozyme, and an additional population corresponding to the single-bound D44.1^des (D44.1^des/lysozyme₁). Tandem-MS thus confirms that the isolated charge state is attributed to a D44.1^des/lysozyme₂ complex.

Next, we examined the D44.1^des/lysozyme interaction by IM-MS measurements of the free proteins and their bound states. Unlike the well-resolved charge states obtained in Fig. 2A, acquired on an Orbitrap instrument, the IM-MS data recorded on a Synapt G1 instrument, yielded broad peaks. Therefore, in order to unambiguously distinguish between the D44.1^des bound forms and ensure the measurement of a single homogenous population, we used a 10 μM lysozyme solution, which saturates the Ab binding sites. IM-MS measurements of the unbound D44.1^des (Fig. 3A) and the free lysozyme (Fig. 3B) gave CCS values of 7,303 ± 58 Ų and 1,434 ± 73 Ų, respectively. These values are in good agreement with previously calculated CCS values⁴⁹. IM-MS spectra of the D44.1^des/lysozyme₂ complex gave a CCS value of 8,348 ± 122 Ų. This value differs from the unbound D44.1^des by 1,045 Ų, reflecting the contribution of the bound lysozyme to the rotationally averaged projected area of D44.1^des.

Figure 3 – Revealing the Ab/antigen buried surface area of interaction in crude media.

IM-MS plots (left panel) and the extracted m/z spectra (right panel) of (A) D44.1^des, (B) lysozyme and (C) the saturated D44.1^des/lysozyme₂ complex. (D) High-resolution structure of the D44.1 Fab variant bound to lysozyme (left panel) and CCS values of free D44.1^des, lysozyme and their D44.1^des/lysozyme2 complex, as well as of buried surface area, as extracted from IM-MS measurements and PISA calculations using the crystal structure (1P2C) (right panel).

We then set out to determine the interfacial buried surface area between the D44.1^des Ab and its target antigen, lysozyme. This area reflects the surface area of D44.1^des occupied by lysozyme upon interaction^50–52. To address this issue, we subtracted the CCS value of the complex from the CCS sum of the free Abs and two free lysozyme proteins. The data revealed a buried surface area of 1,822 ± 153 Ų, which is within the standard size for interfaces burying 1,600 ± 400 Ų of protein surface⁵³. As a structure of D44.1^des bound to lysozyme is not available, we used instead the structure of another high-affinity D44.1-derived Ab (PDB entry 1P2C) and computed the buried surface area using the PISA algorithm⁵⁴. A calculated buried surface area value of 1,758 Ų was obtained, which is in close agreement with our experimental measurements, indicating that isolation and purification of produced Abs is not necessarily a prerequisite for an in-depth native mass spectrometry analysis. This is a major advantage given that structural properties, such as the CCS, can be measured very rapidly in comparison to traditional structural biology methods, such as nuclear magnetic resonance, X-ray, and electron microscopy.

Probing Ab specificity in a crude environment

Monoclonal Abs target a single epitope in the antigen, making them potentially extremely specific. Here, we used direct-MS to probe the binding specificity of a monoclonal Ab to a high-mass antigen directly from a hybridoma cell culture medium. Specifically, we used the growth medium of the MCP21 hybridoma, which secretes a monoclonal Ab against the PSMA2 subunit of the human 20S proteasome. The 20S proteasome is a large, ~720 kDa protein complex, which is one of the major degradation machineries in eukaryotic cells. It is composed of 28 subunits that are arranged in four rings that form a barrel-shaped structure³⁶.

Native MS measurements of the hybridoma growth medium revealed a highly resolved charge state series corresponding in mass to the MCP21 Ab (148,856 ± 22 Da) (Figs. 4A, S4A and Supplementary Table 1). Native MS of the purified 20S human proteasome yielded a major charge state distribution with a mass of 720,850 ± 615 Da, centered at 12,000 m/z (Fig. S4B). We then mixed the crude MCP21 hybridoma growth medium with the purified human 20S proteasome at a 1:1 ratio (v/v), reaching a final proteasome concentration of 1.5 μM. In addition to the unbound MCP21 and free 20S proteasome, a series of large species appeared at the high m/z region (Figs. 4A, S4C, S5, S6A and S7). We identified a population of MCP21 bound to a single 20S proteasome particle (MCP21/20S), whose measured mass was 869,270 ± 1,122 Da, centered at 13,000 m/z, and an additional population with a measured mass of 1,019,582 ± 1,204 Da, corresponding to one 20S proteasome bound to two MCP21 Abs (MCP21₂/20S). The spectrum further contained two additional charge state distributions, located between 15,000–18,000 m/z and 18,000–20,500 m/z. Remarkably, these two series were found to have an almost identical mass, corresponding to two MCP21 Abs bound to two 20S proteasome particles (MCP21₂/20S_2, 1,734,100 ± 2,565 Da and 1,735,380 ± 2,876 Da). Analysis using the UniDec software showed that these two high m/z populations indeed cluster together and correspond to the same complex (Fig. S5A), and bear the same mass after deconvolution (Fig S5B). Similarly, the charge by m/z and charge by mass plots (Fig S5C–D) which display the partitioning of the different ions across the m/z and mass ranges, clustered these two ion populations together.

Figure 4 -. The monoclonal MCP21 Ab is specific to human 20S proteasome and does not bind the rat and yeast proteasomes.

(A) Mass spectrum of the human 20S proteasome mixed with the MCP21 hybridoma growth medium, wherein multiple proteasome-Ab species are detected. In contrast, the rat (B) or yeast (C) 20S proteasomes did not interact with the MCP21 Ab, validating the specificity of the Ab towards human complexes. All measured masses are shown in Supplementary Table 1.

To further validate the identity of these two charge state series (Fig. S6A), we performed tandem MS experiments. We started by isolating the lower m/z population and activating it in the collision cell (Fig. S6B). At elevated HCD voltages (Fig. S6C), a new charge state series appeared, with a measured mass of 1,708,490 ± 271 Da, corresponding in mass to a complex composed of two MCP21 Abs bound to two 20S proteasome particles, but missing one α–subunit (MCP21₂/20S₂/Δα). Next, we isolated the higher m/z population and utilized a variety of approaches to induce dissociation of the complex. In particular, different concentrations of organic solvents were used and different fragmentation methods were employed (HCD, CID and SID) on multiple instrumental platforms. Despite all our attempts, we did not manage to dissociate this charge state series, indicating that this population is organized in a very stable and compact structure, as also evident from the lower number of charges that it bears (Figs. 4A, S4C, S5, S6A).

We therefore hypothesized that the presence of these two distinct populations bearing the same mass may correspond to two different MCP21/20S conformations. To confirm this assumption, we employed IM-MS and performed similar measurements on a Synapt G1 instrument (Fig. S7), followed by CCS calculations (Fig. S8). A CCS value of 407 and 377 nm² was obtained for the lower- and higher-m/z MCP21₂/20S₂ populations, respectively. This result, together with the tandem MS experiments, suggest that the appearance of the complex as two distinct charge state series is due to two conformationally distinct populations, in which the population that appears at the higher m/z region is more tightly packed.

To test the specificity of the MCP21 Ab, we incubated the MCP21 growth medium with similar amounts of two 20S proteasome orthologues, isolated from rat livers and yeast cells (Fig. 4B, C). Unlike the human proteasome, which bound the Ab, we observed only charge state series corresponding to the unbound MCP21 Ab and free rat and yeast 20S proteasomes (717,908 ± 136 Da and 732,758 ± 192 Da, respectively). We could not detect any interaction between MCP21 and the two orthologue proteasomes, in agreement with the Ab specification⁵⁵. In summary, our results validate MCP21’s specificity for the human 20S proteasome, confirming the ability of MCP21 to discriminate between orthologous proteasome complexes. Moreover, the data demonstrate the applicability of the direct-MS approach to the characterization of Ab/antigen interactions involving high-mass ligands, ones as high as ~2 MDa in total molecular weight.

Discussion

Abs are extensively used for research, diagnostics and therapy^{2, 56}. Regardless of their specific application, Abs often need to be engineered at the molecular level to modulate, decrease or increase certain biophysical properties. For instance, many Abs exhibit insufficient expression levels and/or low stability and affinity toward their antigens⁵⁷. The development of molecular and structural engineering and, more recently, in silico design enables significant improvements in recognition properties, modulated stability and specificity⁵⁸. As part of the Ab development process, it is necessary to experimentally validate the design. Here, we demonstrate a native-MS-based method for the rapid characterization of such designed Abs directly from the crude growth media. As we show, this method, termed direct-MS, can provide information on properties such as expression, stability, overall fold, specificity, and associations with antigens. Moreover, the method can also be advanced to include Ab identification using top-down sequencing^{32, 41}.

An important parameter in the development of Abs into research or medical tools is stability⁴⁷. Suboptimal stability can induce aggregation, result in a low yields, affect biological activity, lead to fast clearance from circulation, increase the immunogenicity of the molecule of interest and lead to high production costs^{48, 59}. Therefore, it is often required to assess the improved stability of designed Abs. Our comparative CIU analysis of Ab G6 and its engineered counterpart G6^des13 revealed a gain in stability of the designed form, in agreement with thermal denaturation and temperature of aggregation onset experiments⁴⁷. Thus, the higher apparent stability of G6^des13, as obtained by direct-MS, validated the design process.

The tradeoff between different Ab properties, in which improvements in one feature lead to deficits in others, is a critical concern during the development process⁶⁰. Here, we show that our methodology deciphers, on a single instrument, various tradeoff determinants, including stability, affinity and specificity. Direct-MS can also be used to characterize Ab interactions with antigens that vastly differ in size; in this study, lysozyme (14 kDa) and the 20S proteasome (720 kDa). And we also show that direct-MS can be employed to determine the buried interaction surface area between an Ab and its target. Integration of the direct-MS approach shown here with automated chromatography methods for online buffer exchange⁶¹ and/or size exclusion⁶² is expected to simplify the analysis, by reducing sample handling, i.e. the buffer exchange cycles and sample loading step. Such a configuration is expected to extend direct-MS analysis to throughput applications of Abs libraries. With the caution that on-line separation comes with limitations to data acquisition and interpretation, due to the narrow elution times that may restrict the fragmentation capacity or CIU measurements. Overall, regardless of the infusion method, direct-MS has the potential to promote biopharmaceutical characterization workflows and analysis of biosimilar Abs. Moreover, we anticipate that the ability to rapidly monitor the molecular determinants that mediate tradeoffs between various Ab properties will aid the future generation of optimized Abs for diverse applications.

Statement of significance of the study.

Therapeutic Abs have revolutionized treatment options for various diseases. Their diverse clinical applications have prompted their modulation and modification, laying the basis for Ab engineering. A critical step in the development process is quality assessment, which provides essential input for iterative redesign and optimization. However, this characterization step is usually carried out with purified Abs, which involves significant time and labor investment in product purification. Therefore, the ability to assess the quality of produced Abs without the need for prior purification offers a great advantage. Here, we demonstrate a native MS approach called direct-MS for the rapid characterization of intact secreted Abs directly from the crude growth medium. We show the applicability of the method for characterizing Ab-antigen interactions, stability and specificity. We anticipate that this efficient approach will reduce the time gap between production and characterization and accelerate the development of promising Abs for research and therapeutic application.

Abbreviations

Ab

Antibody

CCS

Collision cross section

CIU

Collision induced unfolding

EMR

Extended mass range

FAB

Fragment antigen-binding

HCD

Higher-energy collisional dissociation

PTM

Post-translation modification

IM

Ion mobility

MS

Mass spectrometry

MCP21

Mouse monoclonal proteasome 20S PSMA2 antibody

UHMR

Ultra-high mass range

VEGF

Vascular endothelial growth factor

WT

Wild type

6. Data availability

The data that support the findings of this study are available on request from the corresponding author.