A Surface Plasmon Resonance‐Based Integrated Assay for Quantification and Glycosylation Characterization of Monoclonal Antibodies in Crude Heterogeneous Samples

ABSTRACT

The rise in cancer, autoimmune, inflammatory, and infectious diseases in recent decades has led to a surge in the development of monoclonal antibodies (mAbs) therapies, now the most widely used family of biologics. To meet the growing global demand, biopharmaceutical industries are intensifying their production processes. One approach to achieve more efficient production of effective mAbs is to develop tools for real‐time quality monitoring. Specifically, the glycosylation profile of mAbs must be closely monitored, since it greatly impacts their therapeutic efficacy and innocuity, making it a critical quality attribute. In this study, we developed a surface plasmon resonance‐based integrated assay allowing for the simultaneous quantification and glycosylation characterization of mAbs in crude samples, hence permitting the at‐line analysis of bioreactor cell cultures. Thanks to the high specificity of the interaction between biosensor surface‐bound protein A and the Fc region of mAbs, we quantified crude IgG samples under mass transport limitations. Next, by flowing running buffer on the surface, impurities contained in the mAbs samples were washed away from the biosensor surface, allowing subsequent recording of the kinetics between the captured mAbs and injected FcγRII receptors. Of interest, with this strategy, we were able to quantify terminal galactosylation and core fucosylation of IgG lots, two important glycan modifications for mAb efficacy.

1. Introduction

Since 2017, two out of every five newly approved drugs are biotherapeutics; those are expected to rise from 17% to 41% of the marketed pharmaceutical products to represent a global market of 445 billion USD by 2028 (Dequier 2023; Lyu et al. 2022). Of all classes of biotherapeutics, monoclonal antibodies (mAbs) are the most widely used, especially with the market entry of biosimilars and the constantly growing demand (Gherghescu and Delgado‐Charro 2021; Jyothilekshmi and Jayaprakash 2021). Indeed, the number of patients in the principal targeted medical indications of mAbs, namely cancer, autoimmune and infectious diseases, is exploding. For instance, it is estimated that the incidence of cancers worldwide will increase by 77% between 2022 and 2050 (WHO 2024).

Most therapeutic mAbs are IgG type immunoglobulins. Their Fab regions recognize antigens while their Fc regions interact with the immune effector cells which express Fcγ receptors (FcγRs) on their cell surface (Boune et al. 2020). During their synthesis, IgGs undergo posttranslational modifications, of which N‐glycosylation is one of the most important. N‐glycosylation occurs on the CH2 domain of the Fc region, more precisely on the asparagine residue at position 297 (Subedi and Barb 2016). The glycosylation motif depends on numerous factors, including host cell specificities and cell culture conditions such as pH, temperature and composition of the culture media (Hossler et al. 2009). Due to the variability inherent to living cells and the complexity of their network of enzymatic reactions, IgG glycosylation at the end of a production process is heterogeneous, even within the same batch (Sha et al. 2016).

Heterogeneous glycosylation is critical from a therapeutic perspective since it influences the immunogenicity, stability and capacity to initiate immune responses of IgGs (Ghaderi et al. 2010; Goetze et al. 2011). For example, IgGs presenting afucosylated N‐glycans exhibit a stronger interaction with FcγRIIIA present at the surface of natural killer (NK) cells, and thus can initiate antibody‐dependent cell‐mediated cytotoxicity (ADCC) more effectively (Falconer et al. 2018; Karampatzakis et al. 2021; Yamane‐Ohnuki et al. 2004). In another case, terminally galactosylated glycans promote IgG binding to C1q complexes, thus playing a beneficial role in complement‐dependent cytotoxicity (CDC) (Nguyen et al. 2023). Human wild‐type IgGs predominantly carry biantennary complex glycans, harboring varied amounts of core fucose, galactose, high mannose, N‐acetylglucosamine and terminal N‐acetylneuraminic acid or sialic acid (Cobb 2019).

Biomanufacturers must demonstrate that the glycosylation of the IgGs they produce is consistent from one production lot to another. In the 2000s, the Food and Drug Administration (FDA) published guidelines on Quality by Design (QbD) and Process Analytical Technology (PAT) (Food and Drug Administration 2004, 2009). QbD seeks to guarantee final product quality by incorporating thorough knowledge of the product and process from the beginning of product development, rather than relying only on final quality testing. PAT is a systematic approach that aims to design and control the production process through real‐time measurements of Critical Process Parameters (CPP), which are linked to Critical Quality Attributes (CQAs). Hence, applying those strategies to the case of IgGs would consist in monitoring and maintaining the N‐glycosylation state constant. To achieve this, the operation parameters of the bioreactor may be adjusted each time the CQAs are measured. Recently, efforts have been made to make high performance liquid chromatography (Chemmalil et al. 2025; Gyorgypal and Chundawat 2022; Tharmalingam et al. 2015) and mass spectrometry (Chi et al. 2020; Song et al. 2021) PAT‐compatible to measure complete glycan profiles of antibody during their production. As summarized in Table 1, despite their multi‐step procedures and demanding maintenance, some of these techniques are now applied in biopharmaceutical industries (Song et al. 2021; Tharmalingam et al. 2015).

Table 1.

Summary of advantages and drawbacks of the main chromatography‐ and mass spectrometry‐based tools for the at‐line monitoring of the glycosylation profile of IgGs.

Analytical tool	Multi‐step?	Reagent demanding?	Monitoring of all major glycans?	Reference
RPLC‐MS	Yes	Yes	Yes	Song et al. (2021)
HILIC‐UPLC	Yes	Yes	Yes	Tharmalingam et al. (2015)
HILIC‐HPLC	Yes	Yes	Yes	Gyorgypal and Chundawat (2022)
RPLC‐UPLC	Yes	Yes	No (sialic acid only)	Chemmalil et al. (2025)
RPLC‐MS	Yes	Yes	Yes	Chi et al. (2020)
SPR assay	No	No	No (terminal galactose and core fucose only)	This study

Surface plasmon resonance (SPR) biosensors measure real‐time protein‐protein interactions and do not require large volumes of samples or large amounts of biological materials nor any molecular label, making them valuable for real‐time bioprocess monitoring and quality assurance (Gaudreault et al. 2021). Our group previously harnessed a SPR biosensor to a bioreactor (Chavane et al. 2008; Jacquemart et al. 2008) to quantify IgG concentration during their production. Characterizing IgG glycosylation with an SPR tool during the cell culture stage would allow monitoring of batch‐to‐batch repeatability and biosimilar equivalency, in accordance with QbD and PAT principles. Using SPR, we recently identified FcγRIIA and FcγRIIB as terminal‐galactose and core‐fucose discriminating receptors for IgG glycan characterization, respectively (Gaudreault et al. 2024).

In this manuscript, we describe an SPR assay that is suitable for the evaluation of two key IgG glycosylation attributes in crude cell‐culture samples. In this configuration, crude samples are first injected over a protein A functionalized SPR surface, which enables IgG quantification thanks to the highly specific interaction between the two species. Additionally, the protein A functionalized surface behaves effectively as an affinity chromatography unit, separating IgG from contaminant proteins and effectively bypassing the need for prior upstream sample purification. With the IgG captured at the surface, a second injection, this time of FcγRIIA/B, is performed. The kinetics of the resulting IgG‐FcγR interaction enable the elucidation of the relative proportion of core‐fucose and terminal galactose on the IgG glycans. Critically, these injections are sequential and provide both quantification and glycosylation profiles of crude samples in a short time frame and with minimal user input. Hence, the evaluation of IgG concentration, terminal galactosylation and core fucosylation levels could be monitored directly from culture samples taken from the bioreactor, thus opening the door to its application in the biopharmaceutical industry.

2. Methodology

2.1. Chemical Reagents and Biological Material

Sodium acetate (#S2889) and ethanolamine hydrochloride (#E6133) were purchased from Sigma‐Aldrich. N‐hydroxysuccinimide (NHS, #24500), N‐ethyl‐N’‐(3‐dimethylaminopropyl) carbodiimide (EDC, #22980), glycine (#BR100354) and HBS‐EP buffer (0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% v/v surfactant P20, #BR100669) were purchased from Cytiva. Acetate and HBS‐EP buffers were prepared with ultrapure MilliQ water (Millipore Gradient A10 purification system) and filtered with 0.22 μm polyether sulfone membranes. Protein A (#P6031) was purchased from Sigma‐Aldrich as a lyophilized powder and then solubilized to 1 mg/mL in 10 mM phosphate buffered saline (PBS).

2.2. Plasmids Construction

pTT5® cDNA vectors encoding Trastuzumab (TZM) light chain (LC) and TZM heavy chain (HC), cDNA vectors encoding human glycosyltransferase β1,4‐galactosyltransferase 1 (GT) or cDNA encoding FcγRIIA and FcγRIIB were constructed as described previously (Cambay et al. 2019; Raymond et al. 2015; Zhang et al. 2009). For Rituximab (RTX) production, LC and HC cDNAs were carried on the same plasmid, derived from the pTT109 or pTT221 vectors (Joubert et al. 2022).

2.3. Cell Culture

Transient gene expression in CHO^55E1 cells was performed to synthesize FcγRs and wild‐type TZM (TZM‐WT), following a previously published protocol (Stuible et al. 2021). Briefly, the culture medium used is a proprietary formulation, supplemented with 4 mM L‐glutamine (Cytiva). Two days before transfection, CHO^55E1 cells were seeded at 1 × 10⁶ cells/mL and incubated in shake flasks (Corning Life Sciences, 120 rpm agitation, 37°C and 5% CO₂ concentration). When the cell density reached approximately 8 × 10⁶ cells/mL, transfection was performed. Herein, cells were diluted with 25% fresh medium and dimethylacetamide was added up to 0.083% (v/v). The transfection was performed using PEI‐Max (Polysciences) at a DNA:PEI (polyethylenimine) ratio of 1:7 (w:w). The final transfected DNA concentration in the cell culture medium was 1.4 μg/mL, and its composition was a combination of 85% (w/w) pTT5‐HC and pTT5‐LC (ratio HC:LC of 1:1) or FcγR constructs, 10% pTT‐Bcl‐XL (antiapoptotic effector) and 5% pTT‐GFP (for transfection efficiency measurement). One day post transfection, each flask was supplemented with an anticlumping supplement (1:500 dilution) (Irvine Scientific) and Feed 4 (2.5% v/v) (Irvine Scientific). Then, cell cultures were moved to a different incubator, at 32°C. 5 days posttransfection, the cultures were supplemented once again with Feed 4 (5% v/v) and glucose to keep its concentration over 17 mM. Cell culture supernatants were collected 6 or 7 days after transfection.

Production of galactosylated TZM (TZM‐GT) was achieved by transient expression, co‐transfecting the human GT pTT5 vector with the TZM light and heavy chains pTT5 vectors. Furthermore, the FUT8 knock‐out cell line was derived from CHO^55E1 using CRISPR/Cas9 as described elsewhere (Koyuturk et al. 2022) and used for transient expression of afucosylated TZM (TZM‐AF).

To produce RTX (RTX‐1, RTX‐2, RTX‐3 and RTX‐4), a cell line (CHO^BRI/rcTA) allowing the inducible expression of recombinant proteins based on the cumate gene switch, was used. Clonal CHO^BRI/rcTA cell lines were generated and cultured under fed‐batch conditions, and the expression of RTX was induced by the addition of 2 µg/mL of cumate (Ark Pharm Inc) in the culture media. Additionally, each clonal cell line expressed various levels of an anti‐FUT8 intrabody to inhibit FUT8‐mediated core‐fucosylation (Joubert et al. 2022).

2.4. Protein Purification

Cell cultures containing secreted FcγRs were harvested and centrifuged at 3300 g for 20 min. Supernatants were filtered (0.45 μm) and the clarified supernatants were purified by immobilized metal ion affinity chromatography (IMAC) with 1 mL Ni Sepharose Excel resin (Cytiva) as described elsewhere (Cambay et al. 2019). The equilibration was performed with a buffer recommended by the manufacturer (50 mM NaPO4 pH 7.8, 300 mM NaCl). The sample was injected at a flowrate of 3 mL/min. The washing and elution steps were performed using the same buffer supplemented with 10 mM and 300 mM imidazole, respectively. The collected fractions containing the FcγRs were pooled and transferred to PBS using a desalting NAP‐25 column (Cytiva). A concentration with Amicon Ultra‐4 10 K centrifugal filter units (Millipore) was conducted before further purification. The last stage, a size exclusion chromatography (SEC) using a Superdex75 10/300 GL column (Cytiva), included equilibration with PBS, injection of 0.5 mL of each receptor at a flowrate of 1 mL/min and elution with PBS. Purified FcγRs without aggregates were obtained.

Secreted TZM lots (TZM‐WT, TZM‐GT and TZM‐AF) were harvested and purified following a previously described protocol (Cambay et al. 2019). Their purification included cell culture medium centrifugation at 3300 g and a filtration step with 0.45 µm‐pore membranes. Next, protein A affinity chromatography was performed using a 2 mL MabSelect SuRe column (Cytiva). The equilibration was made with PBS and the clarified supernatant containing TZM was injected at a flowrate of 1 mL/min. The column was washed with PBS and the antibodies were eluted with 100 mM citrate buffer at pH 3. The eluate containing TZM was transferred into PBS with a desalting NAP‐25 column (Cytiva). A concentration with Amicon Ultra‐4 30 K centrifugal filter units (Millipore) was conducted before a SEC using a Superdex200 column (Cytiva). This last step included equilibration with PBS, injection of 5 mL of IgGs at a flowrate of 1.5 mL/min and elution with PBS. Purified TZM lots without aggregates were obtained.

Similarly, RTX lots (RTX‐1, RTX‐2, RTX‐3 and RTX‐4) were harvested, centrifuged at 1500 g for 10 min and filtered through 0.45 μm hydrophilic, low binding protein filter (Millipore). Supernatants were purified by affinity chromatography, using 1 mL HiTrap MabSelectSuRe columns (Cytiva). PBS was used for equilibration and washing, 0.1 M citrate pH 3.0 was used for elution, and 1 M HEPES was used for neutralization. Antibodies were then desalted into PBS using Zeba spin desalting columns (Thermo Fisher Scientific). Finally, ultrahigh performance liquid chromatography‐size exclusion chromatography (UPLC‐SEC) was used to assess the purity and aggregation levels of all antibodies (Joubert et al. 2022).

A Nanodrop spectrophotometer (Thermo Fisher Scientific) was used to quantify the proteins of all purified samples by absorbance at 280 nm. All protein lots were sterilized by filtration through 0.22 µm filters, aliquoted and stored at ‐80°C.

2.5. Glycan Analysis by HILIC‐UPLC‐Fluorescence

To release glycans from the IgGs, a 2 min denaturation at 80°C was performed before using the Rapid PNGase F (New England BioLabs, #P0710) at 50°C for 10 min as described in a previously published protocol (Koyuturk et al. 2022). Deglycosylated IgGs and Rapid PNGase F were separated from the glycans thanks to a solid phase extraction, using either a 50 or 250 mg Discovery glycan SPE column (Millipore‐Sigma). The isolated glycans were dried under vacuum, labeled with 2‐aminobenzamide (2‐AB) (Sigma‐Aldrich, #PP0520) and further purified (Bigge et al. 1995).

The 2‐AB labeled samples were treated with a2‐3,6,8 neuraminidase and a1‐2,3,6 mannosidase (New England Biolabs, #P0768S) to identify peaks containing sialic acid and high‐mannose glycoforms. Peaks were calibrated with a 2‐AB labeled dextran ladder standard (Waters, #186006841) and compared to Glucose Units (GU) values in a database of previously ran samples of known composition (Zhao et al. 2018). Finally, glycans were analyzed by Hydrophilic Interaction Chromatography and Ultra Performance Liquid Chromatography (HILIC‐UPLC) with fluorescence detection using an Acquity UPLC wide‐pore glycoprotein BEH amide (Waters, #176003702) and a Vanguard glycoprotein BEH amide guard column (Waters, #176003699). The column was set at a temperature of 60°C, the flowrate was fixed to 0.5 mL/min and glycans were eluted with 50 mM ammonium formate at pH 4.4 (mobile phase A) and 100% acetonitrile (mobile phase B). The gradient of mobile phase B differed between analysis of the three TZM lots, the RXT‐1 and RTX‐2 lots, and the RTX‐3 and RTX‐4 lots (see Supporting Information S1: Figures S1–S3).

2.6. SPR Experiments and Analysis

All SPR experiments described in this manuscript were performed on a Biacore T100 biosensor using a CM5 sensor chip (Cytiva) and HBS‐EP as running buffer. This biosensor makes use of the geometry proposed by Kretschmann and Raether (De Crescenzo et al. 2008; Homola 2003). The temperature was set to either 25°C (surface preparation) or 5°C (IgG‐Protein A and IgG‐FcγR interaction measurements). The data acquisition rate was 10 Hz.

In this study, we employed an assay configuration that involves capture of IgGs (ligands) to covalently immobilized protein A on the biosensor surface. A subsequent injection of FcγRs (analytes) allows measurement of the IgG‐FcγR interaction.

2.7. Protein A Covalent Immobilization

Protein A was covalently grafted on the biosensor surface by amide coupling chemistry inspired by a previously published protocol (Navratilova et al. 2007). First, two of the four channels of the CM5 chip (the mock and assay channels) were primed three times with running buffer, cleaned and hydrated with two consecutive 10 s injections of 100 mM HCl, 50 mM NaOH and SDS 0.5% (w/v) at a flowrate of 100 µL/min. Next, the carboxyl groups of the dextran matrix were activated by injecting a 1:1 (v/v) mixture of 0.4 M EDC and 0.1 M NHS for 14 min at a flow rate of 20 μL/min. A 50 µL PBS solution containing 50 µg of protein A was then dissolved in 550 µL of 100 mM acetate buffer at pH 4.5 and injected at 20 μL/min until near exhaustion of the volume to saturate the surface. This step was performed on only one of the two channels available on the CM5 chip (the assay channel). Ethanolamine (1 M, pH 8.5) was then injected in the two channels during 15 min at a flowrate of 20 µL/min to deactivate the remaining carboxyl sites. Finally, two 20 s pulses of 10 mM glycine at pH 1.5 were performed at 100 µL/min to remove all unreacted species that remained near the surfaces. A final protein A immobilization density of approximately 4000 Resonance Units (RU, where 1 RU is equivalent to approximately 1 pg/mm² of immobilized protein) was achieved with this procedure. Ample amounts of running buffer were then injected in the two channels to ensure the biosensor was well equilibrated for the rest of the experiments.

2.8. Calibration for IgG Quantification

The calibration for IgG quantification was made under mass transport limitations (MTL). To this end, TZM diluted in HBS‐EP or affinity chromatography flowthrough at 10 different known concentrations ranging from 0.5 to 50 nM was injected in triplicate at a flowrate of 5 µL/min. The linearity range extended from 0.5 nM to 50 nM, indicating the dynamic range of the quantification part of the assay.

2.9. Preparation of TZM Crude Samples of Known Concentrations

To mimic clarified supernatants, meaning IgG samples containing contaminant proteins and culture medium components, we prepared various crude samples. Purified TZM and RTX lots were diluted to varying concentrations into affinity chromatography flowthrough. The flowthrough had been collected during IgG purification from the harvest volume collected at Day 14 of antibody production. A further dilution in HBS‐EP was performed to bring the crude samples concentrations in the dynamic range of the quantification part of the SPR assay (0.5–50 nM). With this approach, we thus mimicked cell culture samples that would be obtained by sampling a bioreactor, after cells removal.

2.10. Sequential SPR Injections for Quantification, Purification and Study of the IgG‐FcγR Interaction

The first step was the injection of 20 nM IgG in the two channels of the SPR sensor chip, the high‐density protein A surface and the reference surface, at a flowrate of 5 µL/min for 60 s and at a temperature of 5°C. In the second step, running buffer was injected during 100 s at a flowrate of 30 µL/min to wash the surface. The purpose was twofold: first, it served to purify the sample in situ, by specifically capturing the IgGs in a configuration that can be qualified of miniature protein A affinity chromatography unit, thus eliminating potential nonspecific interactions; second, it enables quantification of IgGs as the injection occurs under MTL conditions thanks to the low flowrate of IgG injection and the high density of immobilized protein A.

The third step then involved the injection of either FcγRIIA or FcγRIIB diluted in HBS‐EP at a flowrate of 30 µL/min for 45 s. This was followed by a 1 min buffer injection at a flowrate of 30 µL/min to record the IgG‐FcγR dissociation. This final step enabled characterization of the glycoform distribution in the sample via the area under the curve (AUC) method.

The protein A surface was then regenerated by two regeneration buffer injections (glycine 10 mM, pH 1.5, at 100 µL/min for 20 s each).

This cycle was repeated for every SPR sensorgram recorded in this study. Cycles with a blank analyte injection (running buffer, HBS‐EP) were also performed to allow double referencing (Myszka 1999).

2.11. Equilibrium Constant

FcγRIIA or FcγRIIB solutions of known concentrations (C_A in M) were injected until an equilibrium plateau was reached (45s). This process was repeated with a series of 5 concentrations (10, 30, 60, 100 and 300 nM) of FcγRIIA and FcγRIIB. The plateau values were recorded (R_EQ , in RU) and the steady‐state model was used to determine the thermodynamic constant of dissociation (K_D in M) of the IgG–FcγRIIA/B interaction:

$$R_{\mathit{EQ}}=\frac{R_{\max }{C}_A}{K_D+C_A},$$

With R_max corresponding to the theoretical SPR signal (in RU) that would be reached at surface saturation and K_D the apparent thermodynamic dissociation constant (in M) of the IgG–FcγRIIA/B interaction. K_D and R _max were identified using an in‐house MATLAB program via optimization, given the series of known C_A values and the series of measured R_EQ values. The latter were estimated by taking the average of the signals between 1 and 3 s before the end of the association phase of the sensorgrams. K_D and R _max were fitted once per TZM‐FcγR combination, considering all triplicates of each TZM concentration (15 data points in total). The standard error of the fitted K_D was approximated via the inverse Hessian matrix, and the appropriate Student t quantile was applied to obtain 95% confidence intervals.

2.12. Normalization of SPR Sensorgrams for AUC Calculation

Sensorgrams corresponding to FcγRIIA/B injections were normalized in two different ways. First, the sensorgrams were translated to 0 RU at the end of the dissociation phase by subtracting the average of the last 25 s of the signal from the whole sensorgram. Hence, a potential bias consisting in a signal converging to a non‐zero response at the end of the sensorgram (causes of this phenomenon could be detachment of protein A or IgGs, temperature variation or slight changes in the running buffer) was avoided. Second, the signal at the end of the association phase was fixed to 100% by taking the average signal between 42 and 44 s (the association phase lasting 45 s).

Since poor fits are obtained when using a 1:1 model to study the FcγR‐IgG interaction, an alternative calculation method was used. Many complex parametric dissociation models consist in calculating a sum of exponential decays, which can be challenging to fit to noisy data. Hence, we propose to use the AUC, a nonparametric approach, to study the FcγR‐IgG interaction. The AUC for the dissociation phase of each normalized SPR sensorgram was calculated by numerical integration of the signal. To this end, the computation made use of the trapezoidal method (“cumtrapz” function in MATLAB). The signal of the first 0.5 s of the dissociation phase was excluded from the calculation to discard potential biases hailing from bulk shift artefacts (Gaudreault et al. 2024).

3. Results

3.1. Sequential SPR Assay—Overall Strategy

Figure 1 shows a schematic representation of the sequential SPR assay developed in this study. After covalently grafting approximately 4000 RU of protein A on an SPR surface, the assay consists of three steps. In step 1, IgGs (ligand), crude or purified, are injected at a low flowrate to induce mass transport limitation. In these conditions, the recorded SPR signal is linear, with a slope that is proportional to the IgG concentration, allowing quantification thanks to a previously generated calibration curve. In step 2, running buffer is flowed on the surface, to wash out non‐IgG proteins. At this stage, the IgGs retained at the surface have been effectively purified by affinity separation. In step 3, purified FcγRIIA/B (analyte) is injected over the captured IgGs to characterize the IgG‐FcγRIIA/B interaction in a purified environment. This part of the sensorgram is used to calculate either the K_D or the area under the curve (AUC) during the dissociation phase.

Figure 1.

Schematic representation of the different steps and the typical sensorgram recorded with the sequential assay. The sensorgram starts with a baseline signal that is representative of the presence of protein A on the sensor surface (step 1). The density of immobilized protein A is high so that the system is mass transport limited. After IgG injection, buffer is flowed on the surface (step 2), washing any impurities present on the sensor surface. This allows measurement of the IgG‐FcγR interaction in a clean, purified setting when FcγRs are subsequently injected under kinetic limiting conditions (i.e., at higher flow rate) during step 3.

The experimental temperature is set to 5°C throughout all the steps of the sequential assay to observe significant differences between the SPR signals related to the different IgG glycoforms interacting with a given Fcγ receptor.

The same steps are simultaneously performed on the reference channel, which is devoid of protein A. We complete the double‐reference procedure by subtracting signals obtained by replacing step 1 or step 3 with injections of running buffer, one at a time. Note that we verified that the interaction between FcγRIIA/B and protein A was negligible (data not shown), thus confirming that a channel free of protein A is sufficient for referencing purposes.

3.2. Elucidation of the N‐Linked Glycans Nature and Proportion by HILIC‐UPLC

Using glycoengineered cell lines and transient gene expression, we produced three lots of TZM. Using inducible stable cell pools expressing an anti‐FUT8 intrabody, we also produced four lots of RTX. Each lot displayed various degrees of galactosylation and fucosylation and contained full‐length antibodies (see Supporting Information S1: Figures S4–S7).

We then precisely determined the glycosylation profile of all IgG lots by analyzing them by HILIC‐UPLC. Resulting chromatograms are included in Supporting Information S1: Figures S1–S3, and the peak assignment for all lots in Supporting Information S1: Tables S1–S3. The retention times, expressed in glucose units (GU), showed a variability ranging from 0.01 to 0.18 which is less than 2.5% of the GU value of all glycan chain peaks. The nature and proportion of N‐linked glycans in each lot are summarized in Table 2. TZM‐WT showed a high fucosylation level, with more than 90% of the antibodies containing a core fucose. Antibodies in the TZM‐GT lot were also highly fucosylated (75%), but they contained much more galactose (68.3%) and high mannose (20.9%) than TZM‐WT. The slight reduction in the fucosylation level of TZM‐GT could be explained by an enzymatic competition between the fucosyltransferase and the galactosyltransferase as both enzymes catalyze the transfer of their specific glycan in the Golgi apparatus (Hossler et al. 2009). The presence of bisecting N‐acetylglucosamine (GlcNAc) among the few unassigned peaks is another hypothesis. Indeed, bisecting GlcNAc is often associated to galactosylation and is known to inhibit core fucosylation by steric hindrance (Golay et al. 2022). As expected, and already reported by Gaudreault et al. TZM‐AF presented no fucosylation at all since it was produced in FUT8 knock‐out cells. Each RTX lot presented a different fucosylation level, extending from 4.1% for RTX‐1 to 59.1% for RTX‐3, depending on the anti‐FUT8 intrabody expression level. Antibodies in the RTX‐2 and RTX‐3 lots were similarly galactosylated (11.7% and 11.5% respectively) as well as RTX‐1 and RTX‐4 (34.0% and 33.6% respectively). Finally, the high‐mannose proportion appeared to be almost ten times higher in RTX‐3 (14.7%) than in RTX‐4 (1.7%). Data concerning the glycans analysis by HILIC‐UPLC of RTX‐1 and RTX‐2 have already been published by Joubert et al. in 2022. None of our IgG lots were significantly sialylated.

Table 2.

Summary of the N‐linked glycans analysis by HILIC‐UPLC for each TZM and RTX lot. The proportions of glycoforms harboring a core fucose or at least one terminal galactose, are reported for each IgG lot. The glycan chains with multi‐antennary mannose constructs are reported as high‐mannose glycoforms.

		Sample
		TZM‐WT1	TZM‐GT1	TZM‐AF1	RTX‐12	RTX‐22	RTX‐3	RTX‐4
N‐linked glycan	Fucose (%)	91.6	75.0	0.0	4.1	19.1	59.1	37.5
	Galactose (%)	27.3	68.3	11.5	34.0	11.7	11.5	33.6
	High‐mannose (%)	3.5	20.9	1.6	6.4	5.0	14.7	1.7
	Unidentified (%)	1.5	1.6	0.8	3.9	5.3	6.0	2.6

Note: ¹Gaudreault et al. 2024; ²Joubert et al. 2022.

3.3. Quantification of IgGs in Crude Samples by SPR

We calibrated the quantification part of the assay (step 1 in Figure 1) by injecting a series of 10 known TZM‐AF concentrations over the protein A‐decorated sensor surface. The calibration solutions were generated by diluting a purified TZM‐AF stock solution (at 4.56 g/L in PBS ‐ approximately 31,000 nM). The dilutions were performed either with HBS‐EP (Figure 2A) or with the affinity chromatography flowthrough we collected during TZM purification (Figure 2B). The flowthrough was confirmed to be devoid of IgGs, as a negligible signal was obtained when injecting it on the protein A surface (data not shown).

Figure 2.

Sensorgrams resulting from the injections of 10 concentrations of purified TZM‐AF (A) or crude TZM‐AF (B) on a protein A‐decorated biosensor surface under mass transport limitation conditions. Calibration was performed by calculating the slope of each sensorgram between 5 and 55 s for the solutions of purified (C) and crude TZM‐AF (D). Predicted TZM concentration with respect to actual concentration for different samples (E). The concentrations were estimated using the calibration curve obtained with purified samples (C). Crude stocks of different TZM‐AF concentrations (0.25, 0.4, 0.6 and 0.8 g/L) were prepared and further diluted in HBS‐EP to bring them between 0.5 and 50 nM.

For IgG quantification, we measured the slope of each sensorgram between 5 and 55 s post antibody injection (Figure 2A,B). We observed a linear relationship between the TZM concentration (in nM) and the measured SPR signal slope (in RU/s) with excellent R ² values (Figure 2C‐D) and negligible standard deviations (less than 1.2% of the value of the slope for each concentration). We found very similar trends, whether the TZM was in a HBS‐EP buffer or in crude sample. Indeed, the relative difference between the two calibration slopes is approximately 10%, as reported in Figure 2. We concluded that the amount of protein A we immobilized was sufficient to reach the mass transport limitation conditions required for the IgG capture step of the assay.

For TZM diluted in the affinity chromatography flowthrough, the bulk effects corresponding to the beginning and the end of the sample injections were not eliminated completely by the double‐reference procedure (Figure 2B, note that the reference solution was HBS‐EP). However, there was no alteration of the quantification performances of the assay, as this bulk effect did not alter the slope of the signals.

Towards the end of the culture, the antibody concentration is expected be much higher than the dynamic range of our quantification assay (50 nM and below), hence the samples would need to be diluted in biosensor running buffer. To test the robustness of the quantification part of our assay for the monitoring of IgGs during their production, we emulated samples that would have been taken from a bioreactor at different time points by creating TZM stocks diluted in chromatography flowthrough at various concentrations (0.25, 0.4, 0.6 and 0.8 g/L). Those samples were then brought between 0.5 and 50 nM by further dilution with HBS‐EP. We then estimated these crude TZM concentrations using the calibration curve obtained in Figure 2C (i.e., generated with the TZM stock solution in HBS‐EP buffer).

Figure 2E compares the estimated TZM concentrations to the actual concentrations that were pipetted. The results indicated that the more diluted the crude IgG sample is, the greater the uncertainty of quantification. Indeed, points corresponding to 0.8 g/L IgGs in crude samples were closer to the red central line compared to other crude stocks. Thus, cell culture samples collected towards the beginning of the IgG production would require less dilution in HBS‐EP buffer, given their lower titer in IgGs, and would in turn be less accurately quantified. In any case, the maximal relative deviation was beneath 25%.

In all cases presented in this study, the reference solution was HBS‐EP. We found that using a blank solution corresponding to an HBS‐EP‐to‐flowthrough ratio matching that of the diluted IgG samples significantly reduced the quantification error. This strategy is however not practical in an at‐line bioprocess monitoring framework. In addition, given that the glycosylation differences only have minimal effects on molecular weight and diffusion of IgGs, all glycoforms can be quantified with the calibration generated with TZM‐AF. This statement was verified by injecting other IgG lots and confirming the applicability of the obtained calibration (data not shown). In summary, we concluded that our assay can be used to accurately quantify IgGs contained in cell culture samples, whether they be purified or not. The only requirement is a reference IgG solution that was previously quantified with an orthogonal method for the calibration. Rather than the slope of the signal, we found that using the SPR signal value after IgG injection (e.g., at 90 s on Figure 2A) led to similar precision and was potentially less computationally demanding.

3.4. Efficient In Situ Purification and IgG‐FcγR Interaction Measurements

To demonstrate the purification efficacy of step 2 of the sequential assay (Figure 1), the thermodynamic dissociation constant (K_D ) between purified or crude TZM solutions and the FcγRIIA/B injected in step 3 were calculated and compared.

We performed sequential assay experiments by first injecting TZM samples at 20 nM (step 1), and then injecting a solution of FcγRIIA/B (step 3). The procedure was repeated for 5 concentrations of FcγRIIA and FcγRIIB (10, 30, 60, 100 and 300 nM), and for 4 TZM lots: purified TZM‐AF and TZM‐GT as well as crude TZM‐AF and TZM‐GT (diluted in chromatography flowthrough only). For FcγRIIA and FcγRIIB, the injection time was fixed at 45 s, long enough to reach pseudo‐equilibrium. The signal corresponding to the injection of 300 nM FcγRIIA and FcγRIIB reached approximately 40 RU whereas the capture of 20 nM TZM generated a signal of approximately 400 RU (see Figure 2). This observation can be explained by the difference of molecular weight between FcγRs and IgGs (the first being about 5 times smaller than the second) and by the random orientation of the amine coupling used to graft the protein A, rendering some surface‐bound IgGs sterically unavailable for binding to FcγRs.

We calculated the resulting K_D in each case. The K_D values for crude samples were very similar to those obtained with purified TZM, with overlaying confidence intervals. This confirmed that the washing procedure (step 2) performed by injecting running buffer allowed the removal of impurities from the crude samples, in turn permitting the subsequent recording of signals corresponding to the IgG‐FcγR interactions only. All the K_D values are summarized in Table 3, and the corresponding sensorgrams are shown in Figure 3.

Table 3.

Calculated K_D (nM) for the IgG‐FcγR interactions for two glycoforms (TZM‐AF and TZM‐GT) and two receptors (FcγRIIA and FcγRIIB). Crude samples were obtained by diluting stock TZM solutions using affinity chromatography flowthrough. The confidence intervals on the K_D values were calculated by computing their standard error and applying the appropriate Student t quantile to obtain a 95% confidence interval.

		FcγRIIA	FcγRIIB
TZM‐AF	Purified	250 ± 50	180 ± 10
	Crude	200 ± 50	170 ± 30
TZM‐GT	Purified	180 ± 20	260 ± 40
	Crude	170 ± 30	220 ± 40

Figure 3.

SPR sensorgrams recorded at 5°C by injecting 20 nM of TZM‐AF and TZM‐GT in a purified solution (left) or in crude samples (right) followed by a series of five concentrations (10, 30, 60, 100 and 300 nM) of FcγRIIA (top) and FcγRIIB (bottom). The corresponding steady‐state (pseudo‐equilibrium) fits are presented under each set of sensorgrams. The plateau values were used to calculate the thermodynamic dissociation constants (K_D ) for each receptor concentration.

We also compared the K_D values obtained with our sequential assay to those obtained with a conventional assay with captured Fcγ receptor (ligand), and TZM injected in solution (analyte). The K_D values were very close in the case of FcγRIIA, and within twofold with FcγRIIB (Gaudreault et al. 2024). Thus, the sequential assay configuration showed its potential for in situ purification of crude samples and subsequent unbiased IgG‐FcγR interaction measurement.

3.5. Characterization of the Glycosylation Profile With the AUC

We then normalized the sensorgrams recorded with 20 nM purified TZM lots and 100 nM FcγRIIA/B that were presented in Figure 3 in two ways: first by translating the signal at the end of the dissociation phase to 0 RU, and second by normalizing the response to 100% at the end of the association phase. The normalized sensorgrams are shown in Figure 4. We observed that the more galactose a TZM lot harbored, the slower its dissociation was with FcγRIIA, differentiating it from the others. We also observed that the signal measured during the dissociation phase of FcγRIIB in presence of TZM‐AF (without fucose) was distinct from the signal measured in the presence of other TZM lots (TZM‐GT and TZM‐WT, both highly fucosylated). Those qualitative differences, function of the glycosylation profile of the antibodies, were only observable thanks to the low experimental temperature (5°C), which slowed the interaction kinetics. Indeed, at 25°C, the interaction kinetics of the IgG‐FcγRIIA/B interaction are extremely fast (sensorgrams with a ‘box‐like’ rectangular shape) and thus show little glycoform‐dependant differences. Performing the experiments at 5°C greatly facilitated the discrimination since a low temperature slows kinetics and amplifies the differences of SPR signals obtained for different IgG glycoforms interacting with the same receptor.

Figure 4.

Overlay of normalized SPR sensorgrams recorded at 5°C and illustrating the kinetic interactions of 20 nM of TZM‐AF, TZM‐GT or TZM‐WT with 100 nM of FcγRIIA (A) and FcγRIIB (B) in the sequential assay. Insets show a zoom‐in‐view of the dissociation phase of the normalized sensorgrams.

We then calculated the AUC of the dissociation phase for each normalized sensorgram. In agreement with observations from our previous work (Gaudreault et al. 2024), the AUC values were strongly correlated to the glycosylation profile of the antibodies, providing a quantitative characterization. This observation highlights the relevance of this non‐parametric calculation method as it allows inferences on IgG glycosylation while avoiding the use of ill‐fitting models. Moreover, a linear relationship between the proportion of galactosylated IgG captured on the sensor surface and the calculated AUC was highlighted when the receptor used for the kinetic interaction measurement was FcγRIIA. A similar relationship was brought to light between the preponderance of captured fucosylated IgGs and the AUC when the receptor used for the kinetic interaction measurement was FcγRIIB. The resulting linear regressions and equations are shown in Figure 5A,B. The AUCs shown on this figure were obtained with two of our TZM lots (TZM‐AF and TZM‐GT) and the four RTX lots (RTX‐1, RTX‐2, RTX‐3 and RTX‐4) as well as pairwise mixtures of these lots (AF‐GT mixtures were composed of 25%, 50%, and 75% of TZM‐AF, the remainder being TZM‐GT; RTX mixtures contained equal parts of the two RTX lots that composed them).

Figure 5.

Linear regressions between the AUC and the galactosylation percentage of the captured TZM and RTX when injecting FcγRIIA (A) and between the AUC and the fucosylation percentage of the captured TZM and RTX when injecting FcγRIIB (B). Predicted proportions of terminal galactosylation (C) and core fucosylation (D) of crude TZM and RTX lots with respect to their actual proportions measured with HILIC‐UPLC chromatography. The proportions were estimated with the linear regression equations shown in (A) and (B). The AUC values were computed from the normalized dissociation phase of SPR sensorgrams recorded at 5°C by injecting 20 nM of IgG (TZM‐AF, TZM‐GT, TZM mixtures, RTX‐1, RTX‐2, RTX‐3, RTX‐4, or RTX mixtures) followed by 100 nM of FcγRIIA or FcγRIIB. AF‐GT mixtures were composed of 25%, 50%, and 75% of TZM‐AF, the remainder being TZM‐GT. RTX mixtures contained equal parts of the two RTX lots that composed them. Error bars were obtained by calculating the standard deviation of the AUC of the three replicates of each sample (A, B) or the standard deviation of the predicted % galactosylation (C) or % fucosylation (D) of three replicates.

We verified the robustness of the assay by using the linear relationship found in Figure 5A,B to predict the terminal galactosylation and core fucosylation of crude TZM and RTX samples, and comparing them with HILIC measurements, as reported in Figure 5C,D. The terminal galactose and core fucose levels predictions were found to be very accurate with very small mean absolute deviations from target values of 4.2% and 6.5%, respectively. Furthermore, the prediction accuracy was very similar for both TZM (mean absolute deviations from target values of 4.7% and 5.5% for galactosylation and fucosylation levels respectively) and RTX (mean absolute deviations from target values of 3.9% and 7.1% for galactosylation and fucosylation levels respectively). With a reasonable degree of precision, the graphs of Figure 5A,B can be used to infer the galactosylation and/or fucosylation of a given TZM or RTX sample that is captured on the sensor surface via an interaction with covalently grafted protein A. Hence, we demonstrated that the AUC can be used to infer important aspects of the glycosylation profile of IgGs, whether they are captured from a purified sample or not. Indeed, these results serve to validate the linear relationships found in Figure 5A,B. This conclusion validates the use of the sequential assay to monitor the glycosylation of cell culture samples, regardless of their purification state. TZM and RTX being two distinct IgG1s, this suggests a possible generalization of the assay to all IgG1s.

3.6. Independence of the AUC From the IgG and FcγR Concentrations

We calculated the AUC of the normalized dissociation phases obtained by injecting multiple concentrations (5, 10, 15, 20, 25 and 50 nM) of TZM for capture before the injection of FcγRIIA/B at a fixed concentration (100 nM). The FcγRIIA/B concentration was selected so as to be sufficient to record SPR sensorgrams with an acceptable signal‐to‐noise ratio (in our case 10 RU of FcγRIIA/B were found to be sufficient), while low enough to avoid saturation of the captured IgGs. With normalization of the signal, small variations in the TZM capture level from one cycle to another were canceled. Of interest, our results tend to indicate the absence of variation of the AUC when the concentration of TZM injected for capture was varied between 5 and 20 nM, as shown in Figure 6 (overlapping confidence intervals for these TZM concentrations). Sensorgrams obtained by injecting 25 nM of TZM or more (which translated to a SPR signal of 500 RU or more) showed a slower dissociation kinetic, which translated into a higher AUC, presumably because the IgG‐FcγRII interaction became affected by mass transport in those conditions, owing to the large amount of captured ligand (TZM) accelerating the IgG‐FcγR binding rates, and enhancing rebinding. Overall, these observations suggested that cell culture samples could be directly injected into the SPR instrument for both IgG quantification and glycosylation characterization if their concentration is included between 5 and 20 nM (towards the beginning of production) or if they have been previously diluted to be in this range.

Figure 6.

AUC of the dissociation phase of normalized SPR sensorgrams and corresponding 95% confidence intervals calculated by injecting 6 concentrations (5, 10, 15, 20, 25 and 50 nM) of purified TZM‐AF and 100 nM of FcγRIIA (A) or FcγRIIB (B) in triplicate. Double stars (**) indicate significant differences (2‐way student t test at a 95% confidence threshold) of the AUC average values.

We also proved that the normalization procedure enabled an accurate comparison of dissociation phases and their respective AUCs, regardless of the injected FcγR concentration. Indeed, as shown in Figure 7, when the sensorgrams of Figure 3 (corresponding to the injection of 30, 60, 100 or 300 nM of FcγRIIA/B) are normalized, all dissociation curves are overlaid, no matter the FcγRIIA/B concentration used for the interaction with TZM. Sensorgrams obtained with 10 nM of FcγRIIA/B were excluded from this figure to help visualization (they were noisier due to their lower signal‐to‐noise ratio), but they were also overlaid with all other concentrations during the dissociation phase. This result was expected since our previous study, where IgGs were injected over captured FcγRIIA/B, showed that the dissociation signal of FcγRIIA/B–IgG complexes (when normalized) is independent of the analyte concentration if equilibrium is reached at the end of the injection phase (Gaudreault et al. 2024). Therefore, our integrated and sequential assay is a strong and robust tool to both quantify IgGs and derive crucial information about their glycosylation, regardless of the concentration of the FcγRIIA/B solutions at hand.

Figure 7.

Overlay of normalized SPR sensorgrams recorded at 5°C by injecting 20 nM of TZM‐AF or TZM‐GT in a purified solution (left) or in a crude sample (right) followed by 4 concentrations (30, 60, 100 and 300 nM) of FcγRIIA (top) or FcγRIIB (bottom).

4. Discussion

In this manuscript, we describe an integrated sequential SPR assay, involving a biosensor surface with covalently bound protein A and soluble FcγRIIA/B injections for the determination of IgG concentration and glycoform composition (Figure 1). In our setup, antibody concentration is derived from the mass transport limited protein A‐IgG interaction, while IgG glycosylation state is derived from the protein A‐captured IgG interactions with FcγRIIA/B. Several key features of the assay make it a powerful tool for IgG quality monitoring throughout production processes.

First, information about titer and glycosylation state of IgGs contained in crude samples are effectively obtained without any prior purification. This is worth noting since SPR biosensing traditionally implies at least one purification step before biosensing: impurities within cell culture supernatant are more likely to bind to the biosensor reference surface than on the sensing surface, in turn hampering any subsequent analysis. Our integrated sequential assay efficiently resolves this issue, by allowing in‐situ purification of crude samples, leading to exploitable positive double‐referenced sensorgrams for the IgG‐FcγRIIA/B interactions (Figure 2, Figure 3 and Table 3). Said purification is based on the same principle as protein A affinity chromatography: the protein A‐decorated biosensor surface retained IgGs thanks to a highly selective interaction with the CH₂‐CH₃ interface of their Fc region (Rispens and Vidarsson 2014). One may have thought that a possible overlap with the binding site of FcγRs (which is located at the lower hinge region of the IgGs), would impact the IgG‐FcγR interaction. Despite the proximity between the two binding sites on IgGs, our results did not show evidence of steric hindrance. Indeed, no major impact on the IgG‐FcγR interaction was observed in our sequential assay, when compared to an assay where receptors are tethered to the biosensor surface and IgGs are injected. Hence, our computed K_D values (Figure 3 and Table 3) are very similar to those previously reported (Gaudreault et al. 2024) with the opposite orientation. This contrasts with observations in the literature, that is, in their SPR assay, Kwon et al. reported that capturing IgGs via the interaction of their Fab region targeting oligomeric amyloid‐β (OAβ) sterically hindered subsequent interaction with protein A, thus skewing the measured IgG‐protein A interaction (Kwon et al. 2015). Moreover, our assay is not affected by the potential binding of the TZM VH3 region on the protein A as suggested by some (Ghose et al. 2005), since we obtained similar results with RTX, an IgG that does not contain this region.

Second, our monitoring approach for glycosylation characterization does not rely on parameters identified by fitting ill‐suited models. We previously observed poor fits to a 1:1 kinetic model for the IgG‐FcγR interactions, particularly towards the beginning of the association and dissociation phases, even with optimized SPR assay conditions (Cambay et al. 2019). Multiple factors could explain these poor fits, including heterogeneities in glycoform populations within an IgG lot. Therefore, we here propose a model‐free analysis focusing on the shape of the SPR signal. This is achieved via the calculation of the area under the curve (AUC) during the dissociation phase. As curves are normalized, a strength of this metric is that it is not dependent on analyte concentration (i.e., FcγRIIA/B) (Figure 7). Hence, only one analyte injection is enough to obtain a quantifiable metric for glycoform characterization. This is once again remarkably different from methods aiming to characterize glycoform populations from binding affinities, i.e., from equilibrium responses recorded from multiple injections of analyte.

Third, while the vast majority of reported SPR experiments are performed at room temperature, working at a temperature of 5°C slows down kinetic interactions, in turn unravelling differences in the kinetic behavior of IgG glycoforms. In the specific case of IgG‐FcγRIIA/B complexes, a difference in the dissociation curves of galactosylated and core fucosylated glycoforms was observed (Figure 4) and prompted us toward the calculation of the AUC as a quantitative metric to infer IgG terminal galactosylation and core fucosylation (Figure 5).

Fourth, our assay relies on the use of FcγRII, a quite uncommon feature in SPR assays aiming to characterize IgG glycoform composition. In the literature, FcγRIII is indeed more frequently reported for IgG characterization by SPR (Forest‐Nault et al. 2021). However, the effect of core fucosylation on the IgG‐FcγRIII interaction is orders of magnitude larger than the effect of other glycans, obfuscating their contribution (Cambay et al. 2020; Subedi and Barb 2016). In contrast, FcγRIIA and FcγRIIB have rapid interaction kinetics with IgGs, enabling a shorter injection time while discriminating both galactosylated (FcγRIIA) and fucosylated (FcγRIIB) glycoforms when injected at a low temperature. FcγRIIA/B are thus advantageous to get quality attribute measurements rapidly and efficiently, an asset for process analytical technology.

Fifth, the glycans that our assay quantifies have a major impact on the therapeutic efficacy of IgGs and thus are defined as two of their CQAs. Indeed, the presence of core fucose is known to reduce ADCC (Falconer et al. 2018; Karampatzakis et al. 2021; Yamane‐Ohnuki et al. 2004). Galactose promotes the interaction of IgGs with C1q complexes, thus enhancing the CDC (Ghaderi et al. 2010; Goetze et al. 2011). The impacts of terminal galactosylation and core fucosylation on the effector functions and therapeutic efficacy of IgGs justify the need to monitor them. In contrast, for high mannose glycoforms, we found opposite trends for TZM and RTX samples, with weak correlation between their level of expression and the AUC (coefficient of determination R² < 0.2 with both receptors for RTX samples, data not shown). This observation once again supports the specificity of our assay for terminal galactosylation and core fucosylation. However, although we weigh mono‐galactosylated and di‐galactosylated glycoforms differently in our calculation of the galactosylation (see Supporting Information S1: Tables S1–S3), the resolution of our assay does not allow mono‐galactosylated and di‐galactosylated glycoforms differentiation. This heterogeneity can impact the FcγR and C1q binding of mAbs, and could be clinically relevant (Aoyama et al. 2019).

Our experiments were conducted with 6 lots (Table 2) of two different type 1 IgGs (TZM and RTX), hence supporting its broad application to the characterization of any IgG1. Furthermore, the robustness of the assay to variations in the IgG density on the protein A‐decorated surface (Figure 6), the IgG purity (Figure 5) and the FcγRII concentration (Figure 7) opens the door to at‐line bioprocess monitoring. In that regard, crucial information about IgG concentration and glycosylation state is obtained in less than 12 min, with very little biological material consumption (approximately 0.87 µg of IgG) and no need for prior purification. Moreover, for this study, we were able to record over 1000 sensorgrams with the same protein A surface without any significant calibration drift, indicating that our sequential assay is stable over time. SPR biosensors have already been proposed as ideal tools for PAT (Mandenius and Gustavsson 2015; Wang et al. 2024) and our work demonstrates the feasibility and advantages of this approach for one of the most important therapeutic proteins, IgG.

5. Conclusion

The assay proposed in this study demonstrated its capacity to rapidly measure IgG titer, terminal galactosylation and core fucosylation, whether IgGs are purified or not. Our assay could thus form the basis of an SPR‐based PAT tool to monitor monoclonal antibody production. Indeed, the implementation of QbD approaches, that require rapid and precise determination of CQAs during production to be effective, would be greatly eased by integrated methods such as the one proposed in this study. We firmly believe that such an integration will benefit the biomanufacturing sector.

Author Contributions

Ilona Metayer performed SPR experiments, data analysis and writing of the manuscript. Catherine Forest‐Nault produced TZM and FcγRII lots. Julie Guimond generated stable CHO cell lines for RTX production. Simon Joubert supervised cell culture and protein purification. Olivier Henry, Yves Durocher, Gregory De Crescenzo and Jimmy Gaudreault provided mentorship, supervision and revised the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

20250416 Supplementary data.

Data Availability Statement

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