MaterialsReagents

All chemicals and reagents used in this study were of analytical grade and stored at the recommended storage conditions during our investigations. Sodium bicarbonate, glucose, sodium phosphate, sodium citrate, sodium chloride, dithiothreitol (DTT), maltose, acetonitrile, Tris–HCl, glycine, sodium dodecyl sulfate (SDS), 2-aminobenzamide (2-AB), Borane-2-methylpyridine complex 95%, urea, dimethyl sulfoxide (DMSO), acetic acid and ammonium formate were purchased from Sigma-Aldrich (St. Louis, MO, USA), and stored at room temperature except DTT (stored at − 20 °C) and borane-2-methylpyrimidine complex 95% (stored at 4 °C). Organic solvents such as ethanol, methanol, and acetonitrile were of HPLC grade, purchased from Millipore Sigma (Burlington, MA, USA) and stored at room temperature in fire-proof cabinets. Nuclease free water, genomic DNA purification kit and L-glutamine were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Nuclease free water was stored at room temperature and glutamine at − 20 °C. RNase A and proteinase K from the DNA purification kit were stored at − 20 °C, and the cell lysis, elution, washing, and binding buffers were stored at room temperature. N-glycan sample preparation kit and 2x Luna Universal qPCR Master Mix were purchased from New England Biolabs (Ipswich, MA, USA). The master mix was stored at − 20 °C. Reagents from within the N-glycan sample preparation kit such as 10% NP-40, 10X denaturation buffer and 10X reaction buffer were stored at − 80 °C, and the PNGaseF enzyme at 4 °C. APTS (8-Aminopyrene-1,3,6-Trisulfonic Acid) labeling kit and peptide markers of pI 10, 9.5, 7, 5.5, and 4.1 were purchased from AB Sciex (Redwood City, CA, USA). APTS kit was stored at 4 °C and peptide markers at − 20 °C. Agencourt CleanSEQ magnetic beads were purchased from Beckman Coulter (Brea, CA, USA) and stored at 4 °C. Deionized and filtered Milli-Q water of 18.4 MΩ resistance from a Milli-Q purification system (EMD Millipore, Burlington, MA, USA) was used in the preparation of all aqueous buffers and reagents. APTS labeled N-glycan standards: G2FS2, Man 5, G0F, Man 9, G1F, Man 7, G2, Man 8, G2F, G0, G2FS1, G1, and 2-AB labeled N-glycans standards: G2FS2, Man 5, G0F, Man 9, G1F, Man 7, G2, Man 8, Man 6, G2F, G0, G2FS1, and G1 were purchased from Agilent Technologies (Santa Clara, CA, USA) and all glycan standards were stored at − 20 °C. Laemmli protein denaturation and loading buffer (2X) were purchased from Bio-rad (Hercules, CA, USA) and stored at 4 °C. Primer of sequence CCG ACT CGA GNN NNN NAT GTG G was purchased from IDT Technologies (Burlington, MA, USA) and stored at − 20 °C.

Cell line

A mammalian cell line of Chinese hamster ovary origin, CHO-K1, capable of producing a broadly neutralizing recombinant human monoclonal antibody against Human Immunodeficiency Virus (HIV), was developed by the Vaccine Research Center, NIAID, of the National Institutes of Health, Bethesda, MD, USA, and named VRC01. This cell line was shared with FDA for regulatory science and research under a material transfer agreement. In our studies, we have used this CHO-K1 cell line that produces an IgG1-κ monoclonal antibody known to broadly neutralize HIV strains (Li et al. 2011; Su et al. 2014).

Production of mAbs under different process intensitiesMammalian cell culture

A single vial from a working cell bank of CHO-K1 (VRC01, NIH) cell line was thawed and expanded in shake flask cultures in an incubator set to 37 °C that has 5% CO2 mixed air circulation. ActiPRO™ medium (Cytiva, Marlborough, MA, USA) was reconstituted as per manufacturer’s recommendation with sodium bicarbonate salt for buffer and 6 mM L-glutamine (Invitrogen, Carlsbad, CA, USA), adjusted to pH 7.05, then filter sterilized for cell culture and for production of the mAb. After the cells were expanded to ≥ 10E09, the cells were inoculated [(0.5 ± 0.15) × 10E06 · mL−1] to a reusable 5L glass vessel (~ 2L working volume) bioreactor equipped with stainless steel headplate and supports, operated with a DASGIP controller system (Eppendorf, Hamburg, Germany). The bioreactor run was aided with a perfusion system made up of Repligen ATF2 device and C24 rate controller (Repligen, Waltham, MA, USA). The ATF2 utilized a hollow fiber filter of 0.2 µm pore size and 0.1 m2 filter area. The bioreactor was equipped with an electric heat jacket and an in-line temperature probe to maintain temperature and set to operate at 37 °C, pH of 7.0 ± 0.1 and dissolved oxygen (DO) content of 30% of air saturation. An on-line pH probe (Mettler Toledo, Columbus, OH, USA) and a DO probe (Hamilton, Reno, NV, USA) were used to monitor and maintain pH and DO. The pH of medium in the reactor was maintained with a combination of sparged CO2 gas and 7.5% sodium bicarbonate buffer (Sigma Aldrich, St. Louis, MO, USA). The DO content was maintained by sparging oxygen and air, and, by stirring the tank with a pitched-blade impeller operated at 120 rpm. Perfusion was started on day 3 of the culture at the rate of half working vessel volume (1 L) per day. The perfusion rate was increased incrementally throughout the culture duration, and these rates are shown in Table 1. Perfusates were collected for analysis corresponding to three different process intensities such that the viable CHO cell densities were 15 ± 1, 20 ± 1 and 25 ± 1 × 10E06 cells/mL and finally the run was terminated after harvesting at 26 ± 1 × 10E06/mL cells.

Table 1 Perfusion Schedule and the viable cell density (VCD) of the bioreactor corresponding to the samples analyzedAntibody purification

Perfusates corresponding to different cell densities of bioreactor operation were first purified through a protein A column (Cytiva, Marlborough, MA, USA) to capture the mAb and to remove host cell proteins (HCP). A further cleaning up of the captured mAb was performed on an anion exchange resin (Diethylaminoethyl sepharose, DEAE) column of 5 mL size (Cytiva, Marlborough, MA, USA) to remove nucleic acid contamination, which gave a mAb purity ≥ 98%. For protein A capture, 20 mM sodium phosphate buffer pH 7.0 was used as the binding buffer, and 200 mM sodium citrate pH 3.0 as the elution buffer. For the DEAE column, 5 mM sodium phosphate buffer pH 7.0 was used as the binding buffer and 5 mM sodium phosphate buffer pH 7.0 containing 200 mM sodium chloride was used as elution buffer. Only one peak corresponding to the mAb was eluted during the gradient elution from the DEAE column (from 0 to 200 mM NaCl). Most of the mAb eluted by 75% of the gradient (≤ 150 mM NaCl) and no other peak was detected at 100% of elution buffer ran for another 8 column volumes (Figure S1). After each purification, an SDS-PAGE (sodium dodecyl sulfate–polyacrylamide gel electrophoresis) analysis was performed to ascertain the purity of mAb (Figure S2A). Additionally, we performed capillary electrophoresis with SDS (CE-SDS) analysis to determine the purity of the mAb protein after DEAE column purification.

Purification of light chain (LC) and heavy chain (HC) polypeptides from the mAb

First, we developed a size exclusion chromatography (SEC) method to separate and distinguish the LC and HC polypeptides of the VRC01 mAb. Purified mAb protein (1 mg) was buffer exchanged (twice, each time diluting tenfold v/v, in the 1x denaturing buffer) using Amicon 10 kDa filters (Millipore Sigma, St. Louis, MO, USA). The 1x denaturing buffer consisted of 2% SDS, 50 mM Tris–HCl, 1 mM DTT, 50 mM NaCl, pH 9.0. Then excess dithiothreitol (DTT) was added to make up to 2-, 5- and 10-mM concentration of DTT in 3 samples of analysis, respectively. These samples were incubated for 15 min on an Eppendorf heating block (50 °C, 400 rpm) cooled to room temperature and SEC was performed using a Superdex®200 Increase HiScale® 26/40 column on an AKTA avant 25 chromatography system fitted with a 5 mL super loop (Cytiva, Westborough, MA, USA). The column was pre-washed with Milli-Q water (18.2 MΩ), followed by pre-equilibration (1 CV) with running buffer consisted of 0.05% SDS, 50 mM Tris–HCl, 1 mM DTT, 50 mM NaCl, pH 9.0. Elution was performed at a flow rate of 3.0 mL/min for 212 mL and 2 mL fractions were collected into a deep well 96-well plate (at 6 °C). The fractions corresponding to the three distinct peaks (Figure S3) were concentrated separately using Amicon 10 kDa centrifugation filters (Millipore-Sigma). Protein concentrations were determined using Pierce™ BCA Assay (Thermo Scientific, Waltham, MA, USA) on a BioTek Synergy H1 plate reader (Agilent Technologies, Santa Clara, CA, USA). SDS-PAGE analysis was performed on the polypeptides belonging to the three peaks of SEC to ascertain the isolation of LC (~ 25 kD) and HC (~ 50 kD) polypeptides and the peak that contained the partially reduced and separated mAb protein (~ 75kD) (Figure S4). Based on the resolution, purification and the recovery (Figure S4 and Table S1), treatment with 5 mM DTT was adopted for isolating the LC and HC peptides of all the mAb samples of the current study. The isolated LC and HC polypeptides of the mAb samples were cold stored (− 20 °C) until further analysis of N-glycans by HPLC as described in the following sections.

Verification of DNA contamination in mAb by qPCR

The contamination level of the host cell nucleic acid content in the purified mAb was determined by qPCR analysis. The assay was based on a previous work (Kang et al. 2011). Briefly, a standard curve from 5 nanogram to 500 attogram of DNA was generated from a CHO-K1 cell DNA stock (IDT, Coralville, Iowa, USA). The stock DNA was produced from host cells and purified using a genomic DNA purification commercial kit (Thermo Scientific, Waltham, MA, USA). Then, real-time qPCR detection and quantification of target DNA sequences were performed on the SYBR®/FAM channel on an Applied Biosystems qPCR instrument (Thermo Scientific) using the Luna Universal qPCR Master Mix (NEB, Ipswich, MA, USA) (Figure S5).

Verification of mAb purity and Integrity by CE analysisCE-SDS analysis

All CE analyses were performed on PA800 Pharmaceutical Analysis Capillary Electrophoresis system (AB Sciex, Redwood City, CA, USA). CE analysis was conducted under reducing and non-reducing conditions as previously reported (Fratz-Berilla et al. 2017; Parhiz et al. 2019). Under reducing conditions, another comparative mAb (a biosimilar to Humira or Adalimumab obtained from the University of Massachusetts – Lowell) was also run to confirm the light chain glycosylation on the VRC01 mAb. CE-SDS under non-reducing conditions was performed to determine the mAb integrity.

cIEF analysis of charge variants and pI determination

We have performed capillary iso-electric focusing (cIEF) analysis as previously described (Parhiz et al. 2019) using a neutral capillary to determine the isoelectric point (pI) and charge variants. Briefly, first, cIEF assay conditions were standardized by performing sample preparation in 1M, 2M, and 3M urea containing cIEF gel. Based on the resolution and the sharpness of the peaks, 3M urea-cIEF gel was chosen for the analysis (Figure S6). Electro-focusing was performed using broad range ampholytes (pH 3–10; Pharmalyte™ carrier ampholytes by Cytiva). All charge variant peaks migrated within the marker pI of 9.5 and 7.0. Therefore, all samples were analyzed with the marker peptides of pI 9.5 and 7.0. Peaks were manually identified on the electropherograms within the two flanking pI markers (9.5 and 7.0). The migration time and corresponding peak area (%) values were obtained for all peaks (including the flanking pI markers) using 32 Karat software (AB Sciex). Migration times of each charge variant peak was converted into pI units using the migration time difference between the two flanking pI markers. The determined pI values were verified against a standard curve to agree within a standard deviation of ≤ 0.14 pI units among all the peaks analyzed. The standard curve was generated using peptide markers of pI 10, 9.5, 7, 5.5 and 4.1 (n = 20) (Figure S6).

Analysis of microheterogeneity of mAb produced under different process intensitiesN-linked glycan isolation

The N-glycans were isolated using a kit supplied by New England Biolabs (NEB, Ipswich, MA, USA). Samples containing 20 µg mAb were first denatured using 1x denaturation buffer for 10 min at 100 °C. Then cooled on ice for 1 min followed by release of the glycans by treating with PNGaseF enzyme. First, conditions for separation of glycans from the mAb were investigated by varying the amount of PNGaseF required and varying the treatment time. The mAb treatment with PNGaseF for 1 h or 2 h (Figure S7) was performed at 37 °C. As per the manufacturer-recommended protocol, we used the 1 × reaction buffer and 1% NP-40 in the de-glycosylation procedure. Then, SDS-PAGE analysis was performed to ensure that a downward shift in the mAb bands was observed to the expected molecular weight, indicative of complete deglycosylation.

Labeling N-glycans with APTS and CE analysis

N-glycans isolated from the mAb samples (250 µg protein/sample) were labeled with APTS (8-Aminopyrene-1,3,6-Trisulfonic Acid) using a kit supplied by AB Sciex (Redwood City, CA, USA) as per the manufacturer’s protocol. Maltose was used as an internal standard in all samples. For the dye and n-glycan reaction, the samples were incubated overnight at room temperature to decrease de-sialylation effect of higher temperatures. Following APTS labeling, the N-glycan samples were removed off excess dye using Agencourt CleanSEQ magnetic beads (Beckman Coulter, Brea, CA, USA) and 87.5% acetonitrile solvent. A magnetic rack for microcentrifuge tubes facilitated retention of beads in the Eppendorf vials for easy removal of solvents. First, the beads were treated with the solvent and the solvent was removed. Then, the samples were mixed with the beads and 87.5% acetonitrile solvent was added (at a volume ratio of 1: ≥ 7, sample: solvent) and mixed with the beads containing glycan sample. Placing the vials on the magnetic rack facilitated retention of beads and easy removal of solvent. Beads with glycan samples were washed two times with 87.5% acetonitrile solvent. Glycans were eluted from the beads in 20 µL of Milli-Q water. This process removed excess dye and gave better glycan profiles. APTS labeled N-glycan standards (Agilent Technologies, Santa Clara, CA, USA) were used to identify individual peaks (Figure S8). Glycans were analyzed as per the manufacturer’s recommendation using an nCHO capillary and glycan analysis kit (AB Sciex, Redwood City, CA, USA). Glycans were detected and quantified using a 488-nm laser induced fluorescence (LIF) detector on a PA800 Capillary Electrophoresis system (AB Sciex, Redwood City, CA, USA).

Labeling N-glycans with 2-AB and cleaning

The N-glycan samples isolated from the whole mAb samples as well as the LC and HC polypeptides were derivatized with 0.35 M 2-AB (Sigma Aldrich, St. Louis, MO, USA) and 1M Borane-2-methylpyridine complex 95% (Sigma Aldrich, St. Louis, MO, USA) dissolved in a mixture of 70% DMSO (Sigma Aldrich, Burlington, MA, USA) and 30% Acetic acid (Sigma Aldrich, Burlington, MA, USA), for 2.5 h at 65 °C. The labeled N-glycan samples were then cleaned to remove excess dye by using Agencourt CleanSEQ magnetic beads (Beckman Coulter, Brea, CA, USA). Briefly, 200 µL of thoroughly mixed magnetic beads were taken in 1.5 mL Eppendorf microcentrifuge tubes and bead separations were performed in a Permagen Labware (Permagen, Peabody, MA, USA) equipped with magnets to facilitate separation of beads from a solution. First, the magnetic beads were prepared as follows: 200 µL of thoroughly mixed magnetic beads were taken in microcentrifuge tubes and placed in the Permagen labware for 3 min, then the solvent was removed. The beads were then washed with HPLC grade acetonitrile (100%), 3 times with 3-min incubation intervals. The 2-AB labeled samples were diluted ninefold with 100% acetonitrile, mixed thoroughly with prepared magnetic beads and placed in the Permagen labware for 3 min. After removing the supernatants, the beads were washed with 96% acetonitrile for 3 times with 3-min incubation intervals. The beads were then finally washed with Milli-Q water to elute the 2-AB labeled N-glycans from the beads. The purified labeled N-glycans were dried for 2 h in a Speed Vac freeze-dryer (Thermo Scientific, Waltham, MA, USA) and resuspended in 15 µL of Milli-Q water and directly injected into an HPLC column.

N-linked glycan characterization by HPLC analysis

Analysis of cleaned 2-AB labeled N-glycan samples was performed using a previously published method (Sha et al. 2020) with a slight modification to accommodate more complex N-glycan profiles. This method involves a buffer A of pH 4.5 made up of 100 mM ammonium formate (Millipore Sigma, Burlington, MA, USA), and a buffer B of 100% acetonitrile (Millipore Sigma, Burlington, MA, USA). Separation of N-glycans was performed on a Glycan BEH Amide 150 mm column-130 A with 1.7 µm beads and dimensions of 2.1 mm × 150 mm (Waters, Milford, MA, USA). A guard column was also used (Acquity UPLC Glycan BEH Amide Vanguard column; Waters, Milford, MA, USA). The column was pre-equilibrated with 25% of buffer A at 60 °C column temperature. About 1 µL of the N-glycan sample was injected and elution was performed with the following conditions: (a) for 0–62.50 min, buffer A was varied from 25 to 40% at a flow rate of 0.3 mL/min; (b) for the next 7.5 min, the flow rate was reduced to 0.1 mL/min and the gradient increased from 40 to 100% of buffer A; (c) for the next 6 min, 100% buffer A was run to remove any strongly bound glycans; (d) the flow rate was then ramped up over 10 min from 0.1 to 0.2 mL/min while decreasing the percent of buffer A from 100 to 25%.

The column was pre-equilibrated to the starting condition, by gradually increasing the flow rate from 0.2 to 0.3 mL/min in 5 min and keeping the 25% of buffer A running for an additional 30 min before the next sample injection. After every 9 samples, a blank consisting of 25% ammonium formate and 75% acetonitrile was injected to ensure there was no sample carryover between injections. A standard mix consisting of 13 types of 2-AB labeled N-glycan standards was used and separated by the same HPLC procedure to identify the N-glycan peaks (Figure S9): G2FS2, Man 5, G0F, Man 9, G1F (G1F’), Man 7, G2, Man 8, Man 6, G2F, G0, G2FS1, and G1 (G1’). All the standards were purchased from Agilent Technologies (Santa, Clara, CA, USA). HPLC analysis was performed on an Agilent 1260 high pressure liquid chromatography system (Agilent, Santa Clara, CA, USA) consisting of Agilent 1260 UV detector (catalog no G4212B, serial no. DEAA301559), a fluorescence detector (catalog no.G1321B, serial no. DEABW04335), an autosampler (catalog no. G1329B, serial no. DEAB306274), pumps (catalog no. G1312B, serial no. DEABM02021), a degasser (catalog no. G1379B, serial no. JP02415768), a column thermostat (catalog no. G4212B, serial no. DEAA301559) and a chiller (catalog no. G1330B, serial no. DEBAK33683).

Statistical analyses

Quantitative differences in the composition of N-glycan types were evaluated by analysis of variance (ANOVA) of the integrated peak area (%) values normalized to total peak area of N-glycan types. The percent peak areas were determined for every glycan elution profile of CE analysis, using 32 Karat software (AB Sciex). Then, ANOVA was performed on peak area (%) values of the samples (n = 3 to 9) corresponding to each of the process intensity levels followed by comparison of means by Tukey–Kramer HSD. The relationship between N-glycan content and the viable cell density for different types of N-glycans was evaluated by regression analysis of variance (RANOVA) to ascertain the risk to the model and the risk to consider the slope ≠ 0. For these statistical analyses, we used the software JMP, version 16.0.0 (Cary, NC, USA).

Principal component analysis (PCA) on the data was performed to ascertain the association of process variables and the measured quality attributes of the mAb. The analysis with 3 components explained > 90% of the variability in the data. The relative distance between the parameters on the two component plots from the PCA analysis was used to generate a hierarchical clustering analysis (HCA) by a single linkage method. Thereafter, HCA clusters these parameters as groups to understand which parameters are closely related and which are distantly related. We performed clustering independently for relative distance of 0.01, 0.02, 0.03, 0.04, and 0.05 between the parameters. When the parameters were separated by shorter (0.01–0.03) relative distance, too many groups were observed. Conversely, when the parameters were separated by longer (> 0.05) relative distance, there was only one group, not distinguishing any parameters. Therefore, the parameters were clustered based on the relative distance of ≤ 0.05, which gave 3 distinct groups of parameters. We used a multivariate data analysis (MVDA) tool, SIMCA (Version 17, Sartorius, Cambridge, MA, USA) for these analyses.