A monoclonal human antibody of the IgG1 subclass (IgG), kindly provided by Biogen (Cambridge, MA USA), was formulated at 65 mg/ml in 10 mM sodium citrate buffer containing 5% sucrose, pH 6.0. It was diluted in a same formulation buffer containing 10 mM sodium citrate with 5% sucrose, pH 6.0. 150 mM ammonium acetate (puriss p.a., for mass spectrometry, Fluka, Sigma-Aldrich Steinheim, Germany) was used as the buffer solution for mass spectrometric analysis. The formulations were filtered using 0.22 μm PES (polyethersulfone) low binding syringe-driven filter units (Millex^TM GP, Millipore, Ireland).

Sodium citrate dihydrate and glycine were from Merck (Merck KGaA, Darmstadt, Germany), NaH₂PO₄ dihydrate, Na₂HPO₄ dihydrate, tris(hydroxymethyl)aminomethane (Tris), β-mercaptoethanol and hydrochloric acid (HCl) were from Sigma (Sigma, Sigma-Aldrich Steinheim, Germany), sodium hydroxide from Boom (Boom BV. Meppel, The Netherlands), glycerol and sodium dodecyl sulfate (SDS) from Merck (Merck Darmstadt and Merck, Hohenbrunn, Germany), cesium iodide and sodium azide from Fluka (Fluka, Sigma-Aldrich Steinheim, Germany), bromophenol blue from Bio-Rad (Bio-Rad Hercules, USA) and 5% polyacrylamide tris-HCl ready gel, pre-stained broad range molecular weight markers, silver stain kit from Bio-Rad (Bio-Rad Veenendaal, The Netherlands).

Amicon 10 kDa molecular-weight cut-off filters were purchased from Millipore (Millipore SA, Molsheim, France). Deionized water was obtained from a MilliQ water purification system (Millipore, USA).

The sample under investigation was prepared by diluting the monoclonal antibody formulation with the 10 mM sodium citrate buffer containing 5% sucrose, pH 6.0 to a final concentration of 1 mg/ml IgG. To apply the pH stress, 1 M HCl was slowly added drop-wise with a pipette to the antibody solution to change the pH from 6.0 to 1.0. Then, 1 M NaOH was added to adjust the pH to 10.0. Finally, 1 M HCl was added again to adjust the pH back to 6.0. There was approximately 1 min waiting time between the pH shifts, while constant stirring at 500 rpm. No precipitation was observed after the pH stress treatments.

Unstressed and pH-stressed samples were analyzed with HP-SEC for the detection and separation of IgG oligomers. For this purpose, we used a TSK Gel 3000 SWXL column (300 × 7.8 mm, 5.0 μm particle size) with a TSK Gel 3000 pre-column (TOSOH Bioscience, Stuttgart, Germany) combined with a Thermo UV detector and a Gilson 234 Autoinjector. The separation was performed at a flow rate of 0.5 mL/min, and 300 μL of sample was injected. The elution buffer was composed of 100 mM sodium phosphate, 100 mM sodium sulfate, 0.05% (w/v) sodium azide at pH 7.2. The elution buffer was freshly prepared, filtered and degassed prior to use. Elution was monitored by UV detection (280 nm). Monomer, dimer, trimer/tetramer and high molecular weight oligomer fractions were collected in separate vials after UV detection.

All samples were dialyzed against 150 mM ammonium acetate at pH 6.0 using 10-kDa molecular-weight cut-off Slide-A-Lyzer dialysis cassettes (Thermo Fisher Scientific, Pierce Biotechnology, Rockford, USA). The dimer and trimer/tetramer solutions were further washed with 150 mM ammonium acetate buffer solution pH 6.0, using Amicon ultracentrifuge filters (10 kDa) prior to MS analysis. The sample was centrifuged at 14.000 x g for 15 min at 4°C. The same devices were used to finally concentrate the protein to about 5 mg/ml.

The obtained protein samples were measured with a modified Waters Micromass nano-ESI-TOF MS (positive ion mode). Needles were made from borosilicate glass capillaries (Kwik-Fil, World Precision Instruments, USA) on a P-97 puller (Sutter Instruments, USA), coated with a thin gold layer by using an Edwards Scancoat six Pirani 501 sputter coater (Edwards Laboratories, USA). To produce intact gas phase ions, the source was operated at an elevated pressure (~7 mbar). The created droplet size was 1 μm. The typical spray concentration of the proteins was 5 μM and the injected volume 2 μl. Mass spectra were recorded with a capillary voltage of 1.2 kV and a cone voltage of 150 V. The pressure in the time-of-flight was 2.7 × 10^-6 mbar. All spectra were calibrated using 25 mg/ml cesium iodide solution. Further details can be found elsewhere (34).

Approximately 24 μg of unstressed and pH-stressed IgG and about 3 μg of dimer (fraction 2) and high-molecular-weight oligomers (fraction 4) collected from HP-SEC were mixed with either non-reducing or reducing denaturing sample buffer at a 1:4 volume ratio in 0.5 ml reaction vials (Eppendorf, Hamburg, Germany). The non-reducing sample buffer contained 0.06 M tris(hydroxymethyl)aminomethane (Tris), 25% (v/v) glycerol, 2% (w/v) sodium dodecyl sulfate (SDS) and 0.1% (w/v) bromophenol blue, and the reducing sample buffer also contained 0.7 M β- mercaptoethanol. The samples were treated for 2 min at 98°C in an Eppendorf Thermomixer Confort (Hamburg, Germany) and loaded in a 5% polyacrylamide tris-HCl ready gel, alongside with pre-stained broad range molecular weight markers.

Gel electrophoresis was performed in a Bio-Rad Mini-PROTEAN 3 cell equipped with a Bio-Rad PowerPac 300 power supply (Bio-Rad, Veenendaal, The Netherlands). The electrophoresis buffer containing 0.3% (w/v) Tris, 1.44% (w/v) glycine and 0.1% SDS, pH 8.3. Separation was performed at a constant current of 100 V and at maximally 50 mA for about 1 hour. The gel was stained using the Bio-Rad silver stain kit.

IgG aggregates were prepared according to the procedure described in the “Materials and Methods” section. We initially compared unstressed IgG and pH-stressed IgG in formulation buffer (citrate buffer) with HP-SEC. As can be seen from the chromatograms in Fig. 1, it is evident that an applied pH stress on an IgG sample results in the formation of a number of additional peaks that are attributed to soluble IgG aggregates. The unstressed IgG was predominantly monomeric (peak at 17.7 min), with a minor peak at 14.8 min (Fig. 1, red line), whereas pH-stressed IgG was shown to contain substantial amounts of soluble aggregates (Fig. 1, black line), having distinct peaks at 15.1, and 14.0, 13.2 and 12.1 min. Presumably, these peaks belong to the dimer, trimer, tetramer and high-molecular-weight (HMW) oligomers of IgG, respectively. The observed slight shifts in retention times of unstressed and pH-stressed IgG monomers and dimers are indicative of different interactions with the column material, perhaps due to partial unfolding and increased hydrophobicity of these species. Fig. 1

Size-exclusion chromatograms of unstressed IgG (red) vs. pH-stressed IgG (black) in 10 mM citrate buffer pH 6.0.

While HP-SEC is ideal for initial screening of IgG oligomers, MS is far more suitable for accurate molecular weight determinations. ESI-MS produces multiply charged protein ions, resulting in a charge state distribution in the spectrum. From this distribution, the molecular weight of the protein can be determined using computer algorithms for the deconvolution of the charge envelope. However, ESI-MS sets stringent constraints on the sample buffer. Most buffer solutions used in structural biology and pharmaceutical formulations are not compatible with ESI-MS. The sample solution for ESI-MS needs to be free of any interfering charged species, such as salts and metal ions, as they may obstruct the protein ionization process (27,35). For this purpose, aqueous ammonium acetate solution has been introduced previously as an ESI-MS–compatible ‘volatile buffer.’ This volatile buffer easily desolvates, evaporation of solvents leading to the formation of gas-phased ions of protein molecules, and in the meantime the original quaternary structure of the protein is usually preserved (26,27).

Among many other approaches to identify the proteins by MS, we have followed a “top-down” method for the identification of intact protein aggregates (36). In this approach, the protein sample, dissolved in aqueous ammonium acetate buffer, is directly injected to ESI-TOF MS instrument without any chemical or enzymatic treatment. This results in a molecular mass of the intact protein and/or protein aggregates and their major post-translational modifications.

Figure 2 shows the mass spectra of unstressed IgG (top) and pH-stressed IgG (bottom). These spectra represent characteristic charge envelopes belonging to the protein species. Peaks distributed between m/z 5000 to 7000 represent the different charged states of monomeric IgG, both in unstressed and pH-stressed samples. At first glance the spectra are very similar. Detailed analysis, however, shows that the spectrum of unstressed IgG contains 99.8% monomer and 0.2% dimer (top panel, zoomed inset), whereas that of pH-stressed IgG shows 90.1% monomer, 9.5% dimer, and 0.4% trimer (Fig. 2, lower panel, zoomed inset). These percentages were calculated using the peak intensities of the entire charge state distribution for all species present. Fig. 2

ESI-TOF MS spectra of unstressed (top) and pH-stressed (bottom) IgG solutions, without HP-SEC separation. Samples are in 150 mM ammonium acetate buffer pH 6.0.

As can be seen in the previous section, the monomer signal is very abundant in pH-stressed IgG solution, and it suppresses the MS signals deriving from the other (higher oligomer) species in the solution. However, HP-SEC clearly showed the existence of other species. Therefore, chromatographic separation and purification of the individual fractions was performed. In this way, we can isolate and individually study the structural information and possible conformational variants within the dimers, trimers, and other oligomers.

HP-SEC was used to isolate the monomer, dimer, trimer/tetramer and HMW oligomer fractions of the stressed IgG mixture. The stability and purity of these fractions were tested by re-analyzing them with the same HP-SEC method (Fig. 3). Clearly, isolation of the individual species was successful: four fractions presumably corresponding to monomer, dimer, trimer/tetramer and HMW oligomers were obtained. The trimer and tetramer species were collected simultaneously because of the difficulty in separating the two species with this HP-SEC column. In addition, the results show that the separated products are largely irreversible as analysis of the individual fractions hardly shows other oligomeric forms or monomeric IgG in their respective chromatograms (Fig. 3). Fig. 3

Size-exclusion chromatograms of collected fractions of pH-stressed IgG; HMW: High-molecular-weight aggregates (black), trimers/tetramers (red), dimers (green) and monomer (blue). Spectra are not normalized.

SDS-PAGE is a well-established technique to investigate whether the aggregates formed were composed of covalently and/or non-covalently linked species (4,38).

In the literature, the simultaneous formation of covalent and non-covalent dimers has been reported for different IgG mAbs (18,39–41). In addition, it has been observed that some short-term stressed non-covalent antibody dimers can irreversibly convert into covalent dimers (18,42). The amount of non-covalent and covalent dimers formed by several IgG species depended on IgG type, solution conditions, and temperature (11).

We performed non-reducing and reducing SDS-PAGE on the unstressed IgG, pH-stressed IgG, isolated fraction 2 and fraction 4, as depicted in Fig. 6. Given that the amount of dimer was less abundant compared to monomer in both unstressed and pH-stressed IgG according to HP-SEC, an excessive amount of these samples was loaded in the slots of the gel in order to detect these dimers. This resulted in over-exposure of the monomer band at ~150 kDa in lanes 2 and 3 (Fig. 6), while the dimer bands were visible at ~300 kDa. In fact, the dimer bands under non-reducing conditions are rather similar in intensity between these two samples, which indicates that the majority of dimers were non-covalently linked. This was confirmed by SDS-PAGE analysis of the dimer fraction (Fig. 6, lane 4), where most of the dimers collected by HP-SEC fractionation had dissociated into monomers under denaturing conditions. For the HMW oligomer fraction 4, a band can be observed at the top of lane 5 and another one at ~150 kDa, which means that this fraction contained both covalent and non-covalent oligomers. The lack of any HMW bands on the gel run under reducing conditions (Fig. 6, lane 9) indicates that these covalent aggregates were disulfide mediated. Nevertheless, the pH-stressed IgG dimers were strong enough to remain intact during HP-SEC analysis, dialysis, and subsequent ESI-TOF MS analysis. Fig. 6

Silver-stained SDS-PAGE gel run under nonreducing (NR) and reducing (R) conditions. Lanes belong to standards (1), unstressed IgG (2 for NR, 6 for R), pH-stressed IgG (3 for NR, 7 for R), HP-SEC fraction 2 (4 for NR, 8 for R) and HP-SEC fraction 4 (5 for NR, 9 for R) of pH-stressed IgG (see Fig. 3 for HP-SEC profiles of the fractions).