Characterizing thermal transitions of IgG with mass spectrometry

Abstract

Variable temperature electrospray ionization (ESI) is coupled with mass spectrometry techniques in order to investigate structural transitions of monoclonal antibody immunoglobulin G (IgG) in a 100 mM ammonium acetate (pH 7.0) solution from 26 to 70 °C. At 26 °C, the mass spectrum for intact IgG shows six charge states +22 to +26. Upon increasing the solution temperature the fraction of low-charge states decreases and new, higher-charge state ions are observed. Upon analysis it appears that heating the solution aids in desolvation of the intact IgG precursor. Above, ~50 °C a cleavage event between the light and heavy chains is observed. An analysis of the kinetics for this processes at different temperatures yields transition state thermochemistry of ΔH^‡ = 95 ± 10 kJ·mol⁻¹, ΔS^‡ = 8 ± 1 J·mol⁻¹·K⁻¹, and ΔG‡ = 92 ± 11 kJ·mol⁻¹. The mechanism for light chain dissociation appears to involve disulfide bond scrambling that ultimately results in a non-native Cys¹⁹⁹-Cys²¹⁷ disulfide bond in the light chain product. Above ~70 °C, we are unable to produce a stable ESI signal. The loss of signal is ascribed to aggregation that is primarily associated with the remaining portion of the antibody after having lost the light chain.

Introduction

The immunoglobulin G antibody (IgG) is a ~147 kDa protein in the immune system that is involved in antigen recognition and binding.[1] This molecule is often visualized by the ‘Y’ shaped diagram shown in Scheme I. As shown, IgG is comprised of a dimer of heterodimers (the heavy and light chains). The heterodimers are linked by two disulfide bonds. The light and heavy chains of each heterodimer are bound by a single disulfide bond. Together these regions create a highly-specific antigen binding pocket called the F_AB portion of the molecule that is critical for immune response.[2] In recent years, numerous monoclonal antibodies with therapeutic value have been introduced.[3-6] Because of this, an understanding of the structures and stabilities of these molecules is of fundamental importance.

Scheme 1:

IgG schematic showing light chain (red), heavy chain (blue), the antigen binding region (F_AB) and the crystallizable region (F_C). Cys or disulfide bonded cysteines are shown as black lines. Cys residues discussed in main text are labeled. Single free cysteine is shown on each light chain.

Although calorimetric studies of antibodies are routinely carried out in the development and testing of new therapeutic antibodies these methods provide information about only the stability of the ensemble average. That is, the structural change is observed as a two-state cooperative transition, and little is known about the nature of the configurations and mechanisms leading to denatured states.[7, 8] In the work presented below, we investigate the stability of IgG using a new, variable temperature electrospray ionization (vT-ESI) source coupled with mass-spectrometry (MS) measurements.[9, 10] At elevated temperatures (above 50 °C) the MS measurements reveal that the light chain of IgG dissociates, through a mechanism that involves scrambling of the disulfide bonds, resulting in the formation of a light-chain product that adopts non-native Cys¹⁹⁹-Cys²¹⁷ and Cys⁹¹-Cys¹⁴⁰ disulfide bonds. From an Arrhenius analysis of the kinetics of dissociation at varying temperatures we derive transition state thermochemistry for dissociation process. This thermochemistry is discussed.

The present work builds on a number of new MS-based measurements that are being developed with the aim of understanding structures and structural transitions of biomolecules in solution as well as the gas phase. In the last decade ‘native ESI’ has enabled the study of large complexes.[11-16] Analyses of biomolecular conformations from solutions of varying composition and temperature now have an extensive history.[17-20] Differences in structures found under varying solution conditions can be investigated with a range of reaction chemistries and techniques, including: isotopic hydrogen-deuterium (H-D) exchange;[21-26] fast photochemical oxidation of proteins;[27-30] chemical cross-linking[31-33] and other residue-specific modifications;[34, 35] and ion mobility measurements.[36-46] Once ionized, an array of physical and chemical methods can be used to investigate biomolecular structure in vacuo. These include, low-energy and high-energy collisions with buffer gasses[47-52] and surfaces;[53-55] photodissociation techniques;[56-59] measurements of collision cross sections with many new ion mobility methods;[60-67] ion-molecule reactions, including proton-transfer[68-71] and H-D exchange reactions;[72-75] ion-ion reactions;[76-78] as well as electronic and vibrational spectroscopies.[79-82]

Experimental

Variable-temperature electrospray ionization.

The solution temperature of the ESI emitter, is controlled using a home-built variable temperature electrospray ionization emitter.[9] This interface holds a borosilicate glass ESI emitter which has been pulled to a narrow inner dia. of ~1 to 5 μm using a Flaming/Brown P-97 pipette tip puller (Sutter Instruments, Novato, CA, U.S.A.) The emitter interface is made of a thermally conductive ceramic block that is resistively-heated using a cartridge heater. An ESI voltage of 0.7 to 1.0 kV is applied to a platinum wire that is inserted into the back of the emitter, making electrical connection with the solution. The temperature is measured (to a precision of ± 0.5 °C) using a K-type thermocouple, that is inserted into the ceramic block.

Instrumentation.

Initial experiments were performed on Waters Synapt G2 and remaining kinetics experiments were preformed Waters Synapt G2s instrument. Both instruments were used in with the source interlocks overridden to accommodate the vT-ESI source.[9, 83] Source pressures and voltages that minimize ion activation were used for initial studies, summarized below.[84, 85] Backing pressures were increased to ~ 8 mbar using Speedivalve (Edwards, Burgess Hill, UK). Gas control were optimized to minimize ion activation and increase ion transmission; including flow rates in the source (20.9 mL/min), trap (10 mL/min), helium cell (180 mL/min) and IMS cell (90 mL/min). We additionally optimized and used instrument settings; including sampling cone voltage (62 V), extraction cone (1.6 V), source temperature (50 °C), cone gas (10 L/h), flow gas (0.6 bar), purge gas (100 L/hr), trap collision energy (12.6 collision energy). Finally, TriWave DC voltages were also optimized: entrance voltage (3.2 V), bias (45.3 V), trap DC (1.7 V) and exit voltage (1.3 V). We note that there are different source configurations between the Synapt G2 and the Synapt G2S that can contribute to charge state shifts, presumably due to collisional activation. This may be the origin of the charge state shift between the two sets of mass spectra presented. The analysis presented below uses only the time-of-flight mass analyzer. That is, the quadrupole is fixed to transmit all ions. The data were originally acquired as nested ion mobility mass spectra, however only the mass dimension was used for this analysis as there were no resolvable changes in drift time distributions for these species.

Analysis of the data.

Each dataset was collected by systematically increasing the temperature from 26 to 70 °C. Samples were allowed to incubate at each temperature for at least three minutes. After this time the mass spectra recorded at low solution temperatures (26 to 45 °C) do not appear to change for an extended period and it appears that we have reached an equilibrium (on this timescale). At higher temperatures the antibody dissociates. Kinetics experiments of this process were carried out by increasing the temperature to a set point, and then collecting a series of three minute acquisitions until spray was lost. Unidec[86] (Oxford, UK) was used for deconvolution of native charge state distributions. Data from both experiments were exported using TWIM extract (University of Michigan, Ann Arbor, MI) and processed using Origin2018 (OriginLab Corporation, Northampton, MA, U.S.A.). Kinetic data were fit using a first order reaction rate of formation. Measured rate constants, k, were plotted as a function of the temperature in an Arrhenius plot and fit linearly to obtain transition state chemistry.

Sample preparation.

Immunoglobulin G (IgG1, universal antibody standard, human, ≥ 90% purity) was purchased from Sigma Aldrich (St. Louis, MO, U.S.A.). IgG (1 mg) was resolubilized in 100 mM ammonium acetate solution (pH 7.0, 500 μL). Sample was buffer exchanged using a spin concentrator (molecular weight cut off = 30,000 Da, 100 to 500 μL × 3, Millipore Sigma, Burlington, MA, U.S.A.). IgG was brought to a final concentration of 80 μM in 100 mM ammonium acetate (Sigma Aldrich, St. Louis, MO, U.S.A).

Results and discussion

Changes in ESI charge state distribution of IgG precursor with temperature.

Figure 1 shows representative mass spectra for intact IgG recorded upon electrospraying a 100 mM ammonium acetate solution at 28, 32, 42 and 52 °C. At low temperatures (28 to 42 °C) a narrow distribution of IgG charge states from +22 to +26 and centered at +24 is observed. As the solution temperature is increased beyond this point, a new peak corresponding to the +27 species is observed and the +25 and +26 species increase in relative abundance. When the peaks in the mass spectrum are examined more carefully (see the inset in Figure 1) we find that at low temperatures peaks are broader and extend to higher masses. As the solution temperature is increased each charge state becomes noticeably shaper and the distribution of charge states changes. From ~28 to 42 °C the abundances of the lower-charge state +22 and +23 species decrease with increasing temperature; in this same temperature region the populations of +24 through +27 also increase. Above 42 °C, the +24 species decreases in abundance and the +25 species is favored. This ion reaches a maximum abundance at ~54 °C and decreases above this temperature. The +26 and +27 continue to increase until ~70 °C, where we no longer maintain a stable ion signal. At this temperature, the clear solution becomes turbid due to the formation of insoluble aggregates,[87, 88] likely originating from IgG unfolding. The slight shift in charge state observed here is similar to the that seen when tetrameric concanavalin A is heated, which was coupled to a structural change.[89] While we do not observe a change in the collision cross sections for IgG, there have been reports that IgG undergoes structural changes between 25 and 70 °C under acidic pH.[90, 91] In these cases, highly charged MS peaks emerged at elevated temperatures, indicating that the protein had unfolded. We also find that the peaks decrease in width at elevated temperatures, suggesting that IgG emerges from hot electrospray droplets with fewer species adducted.

Figure 1:

Mass spectrum showing the charge state shift observed for IgG charge states 22 to 27 at 28, 32, 42, and 52 °C (left). Inset shows +22 species at 28 and 52 °C. The relative abundance plotted as a function of temperature for charge states +22, +23, +24, +25, +26 and +27. Above ~70°C spray stability is lost.

High-temperature dissociation of the light chain.

At low temperatures IgG remains stable for long times. We monitored the mass spectra of a solution at 45 °C for up to ~15 hours, and it shows no measureable change. At elevated temperatures IgG is known to dissociate by loss of the light chain.[92, 93] Figure 2 shows mass spectra acquired after incubating solutions at 65 °C for 3, 18 or 42 min. At relatively short incubation times (3 min.) the mass spectrum is dominated by peaks associated with the IgG precursor. At longer times, (18 min. as shown in Figure 2), the relative abundances of the IgG peaks decrease and a new, well-defined set of peaks m/z < 3000 are observed. These peaks increase in magnitude with increasing incubation time. The theoretical molecular weight of the light chain species is 22,942 Da. Using this value, we determine that the major peaks correspond to +9 through +13 charge states of the light chain. Once the charge states are assigned our experimental measurement yields m = 22,943 ± 1 Da, in close agreement with the theoretical value. It is interesting that we do not observe the complementary remaining IgG fragment, which should have m ~ 124 kDa, although aggregation of this species is known to occur rapidly.[88, 92]

Figure 2:

Mass spectrum showing formation of light chain charge states +9 through +13 after incubation at 65 °C for 3, 18, and 42 min.

Mechanism of light chain dissociation.

Before describing the kinetics experiments, we first present a possible mechanism for the covalent bond cleavage leading to the release of the light chain species. Dissociation of the light chain from the heavy chain must involve cleavage of the Cys²¹⁷-Cys²²⁴ disulfide bond, between the two chains (Scheme 1). It is established that during the dissociation of the light chain, disulfide bonds can scramble.[92, 94, 95] The precursor IgG antibody used in our study has a single free Cys⁹¹ residue on each of the light chains. As we think about thermodynamic considerations we see that disulfide scrambling upon dissociation of the light chain could stabilize the products. That is, if only the native Cys²²⁴-Cys²¹⁷ bond between the heavy and light chains were cleaved, three free, unbound Cys residues would be available (Cys²²⁴ on the heavy chain, and Cys²¹⁷ and Cys⁹¹ on the light chain). Consider one scenario. After the Cys²²⁴-Cys²¹⁷ bond is cleaved the freed light chain might refold and in doing so scramble its disulfide bonds in order to stabilize this product. For example, if the Cys¹⁹⁹-Cys¹⁴⁰ disulfide bond were also to cleave, we might form a non-native Cys¹⁹⁹-Cys²¹⁷ linkage. In this case the newly freed Cys¹⁴⁰ residue could form a second disulfide bond with the Cys⁹¹ residue and the resulting covalent Cys¹⁴⁰-Cys⁹¹ bond would further stabilize the light chain product. With this change, the only free Cys residue is located on the heavy chain (Cys²²⁴) and the light chain is no longer covalently linked to the heavy chain. Overall, this process is a disulfide mixing[96, 97] step with nearby Cys residues both making and breaking covalent bonds, a process that has been described previously for light-chain dissociation from IgG under mildly denaturing conditions.[92, 98-101] It is likely that at elevated temperatures a similar disulfide bond scrambling will occur.

With this idea in mind, we carried out studies to identify the location of the non-native disulfide bond. After incubation at 65 °C for 90 min. we alkylated the free Cys residues with iodoacetimide. The products of antibody dissociation were then proteolytically digested and the tryptic peptides that were formed were analyzed using a combination of chromatographic separations with MS detection (see supporting information). An analysis of the cross-linked peptides provides evidence for the non-native Cys¹⁹⁹-Cys²¹⁷ disulfide bond. Although we anticipate that the Cys⁹¹-Cys¹⁴⁰ disulfide bond should also stabilize this fragment, we did not detect this cross linked peptide in our analysis. We note that it would be difficult for us to fragment this species with our experiment due to its size (60 amino acids); so, the dearth of experimental information does not rule it out entirely. Overall, this result supports the idea that the mechanism for light-chain dissociation involves disulfide bond scrambling.

Kinetics measurements at varying temperatures.

We next carried out a series of kinetics experiments at specified 57, 60, 62, and 65 °C using the vT-ESI source. Because we do not observe the IgG fragment that complements the light chain, we report kinetics based on only the increase in the light chain signal and the decrease of the intact IgG precursor (see supporting information for details). As mentioned above, it is interesting that we do not observe the complementary heavy-chain fragment. We suspect that this species aggregates soon after the dissociation process. At our longest times at each temperature, where the dissociation of the intact IgG precursor approaches completion, the ESI signal is lost, consistent with this idea, and the experiment is terminated.

Examples of kinetics data recorded at several temperatures (60, 62, and 65 °C) are shown in Figure 3 for the increase in the light-chain abundance with increasing time (the 57 °C data are not included because the figure appears crowded; but, an example for these data can be found in the supporting information). Examination of the kinetics shows that this process follows a simple, first order reaction rate, and can be modeled with equation 1,

$$B_t=1-I_0\times e^{-kt}$$ (1)

where B_t is the intensity of the light chain signal at time t, I₀ is the final light chain signal upon reaction completion, and k is the rate constant.

Figure 3:

Light chain signal monitored over time at 60, 62, and 65 °C, as open squares, triangles, and circles, repectively. Change in light chain abundance with time is shown modeled with first order kinetics (black lines).

Transition state thermochemistry associated with formation of the light chain.

The rate constants obtained from the kinetics experiments shown above were used to generate the Arrhenius plot in Figure 4. This plot yields a pre-exponential factor, A = 4.33 ± 0.8 × 10¹³ s⁻¹ and activation energy, E_a = 96 ± 28 kJ·mol⁻¹. This value of A indicates that accessing the transition state is very efficient – occurring near the vibrational frequency that is expected for a simple bond cleavage of a small molecule. We suggest that this may indicate that the transition state involves a very localized motion associated with cleavage of the native Cys²²⁴-Cys²¹⁷ bond in the intact precursor IgG. These values can be converted into transition state thermochemistry, yielding: ΔG^‡ = 92 ± 11 kJ‡mol⁻¹, ΔH^‡ = 95 ± 10 kJ·mol⁻¹, and ΔS^‡ = 8 ± 1 J·mol·K⁻¹. The large enthalpic barrier is consistent with the cleavage of a covalent disulfide bond. Furthermore, the relatively small entropy change at the transition state suggests little change in structure. Overall, this thermochemistry is consistent with a sequential process. First, the native Cys²²⁴-Cys²¹⁷ bond in the IgG precursor is cleaved. Upon cleavage the light-chain fragment refolds such that at least one (and possibly two) new non-native disulfide bridges (the scrambled Cys²¹⁷-Cys¹⁹⁹ bond that was detected, and the Cys¹⁴⁰-Cys⁹¹ bond that we anticipate could be formed, but was not directly detected) are formed, stabilizing the light chain product. The complementary heavy chain product of dissociation, having the single reduced free Cys²²⁴ rapidly aggregates, and is not detected in our experiments.

Figure 4:

Arrhenius plot showing triplicate reaction rate of formation versus inverse temperature. Error bars represent the standard deviation of the triplicate analysis.

Conclusions

Variable temperature ESI and mass spectrometry have been used to investigate thermal transitions in the IgG antibody. It is found that IgG dissociates through loss of the light chain, a process that involves disulfide bond scrambling. Kinetics studies at multiple temperatures were used to determine transition state thermochemistry of ΔH^‡ = 95 ± 10 kJ·mol⁻¹, ΔS^‡ = 8 ± 1 J·mol⁻¹K⁻¹, and ΔG^‡ = 92 ± 11 kJ·mol⁻¹.