Unique challenges required reassessment and alterations to critical reagents to rescue a neutralizing antibody assay

Abstract

Aim: To redevelop a neutralizing antibody (NAb) assay to be much more drug tolerant, have a large dynamic range and have high inhibition when using high levels of positive control (PC).

Materials & methods: Early assay data suggested that typical biotin labeling of the capture reagent (Drug 1, produced in a human cell line) was blocking it from binding with the PC or the detection target, and that the detection target was out competing the PC. Methodical biotin labeling experiments were performed at several challenge ratios and an Fc linker was added to the detection target.

Results & conclusion: A larger dynamic range, high inhibition and higher drug tolerance were achieved by adding an acid dissociation step to the assay, performing atypical biotin labeling of Drug 1 and switching to a detection target that contained an Fc linker to increase steric hinderance and decrease its binding affinity to Drug 1.

Plain Language Summary

Many of the drugs available today are produced by a living organism and these are called biologics. Biologics are larger than chemical drugs and the human body can detect them as foreign and create antibodies against them. This is called immunogenicity. When the antibodies created against the biologic blocks the drug's ability to work correctly, they are called neutralizing antibodies (NAbs). Testing for NAbs is one of the requirements of regulatory agencies for biologics. Here we describe challenges encountered developing an assay to test for NAbs against a biologic.

Tweetable Abstract

Development of a ligand binding neutralizing antibody (NAb) assay offered some unique challenges. The biotin conjugated capture drug and the ruthenylated detection target both required atypical labeling to achieve desired results.

Plain language summary

Article highlights.

• Unusual light chain biotin labeling of Drug 1 was interfering with its ability to bind to the positive control (PC), thereby lowering the PCs ability to inhibit the electrochemiluminescence (ECL) signal by blocking detection target binding.

• Labeling with high challenge ratios (i.e., 50:1 & 100:1) over labeled the light chain of the antibody (Ab), inhibiting Drug 1 from conjugating with the PC and the target, resulting is very low ECL signal and very little inhibition by the HQC.

• Given enough time, some of the detection target was able to colocalize with bound PC, so competition was not completely exclusive. By adding an Fc linker to the detection target, we were able to increase its size and steric hinderance (reducing its binding affinity), which reduced colocalization with bound PC.

• Having the ability to assess critical reagents with LC-MS and biolayer interferometry we were able to elucidate what was causing most of the problems we encountered and make appropriate adjustments to produce a better neutralizing antibody assay.

1. Introduction

Biotherapeutics risk being recognized by the immune system as foreign and eliciting an anti-drug antibody (ADA) immune response in dosed subjects. ADAs have the potential to impact efficacy and safety. ADA responses may be diverse and polyclonal among dosed patients potentially binding different epitopes on the therapeutic protein; some with little or no effect, but others that may neutralize the drug's pharmacologic activity. These types of ADAs are called neutralizing antibodies (NAbs) and may diminish the drugs pharmacodynamic effect, change its pharmacokinetic profile and/or compromise clinical efficacy [1–5]. Although non-NAb ADA do not inhibit the pharmacodynamic activity of the drug, and are often benign, in some cases can still be problematic. Some have been shown to increase the rate of drug clearance, resulting in a clinically similar outcome to that of NAbs [6]. When NAbs are present, in addition to compromising the drug efficacy by reducing the availability of biologically active drug, they can induce serious adverse effects if they cross-react and neutralize the activity of an endogenous homolog and impair its essential normal physiological function [7,8]. Immunogenicity analysis is applied according to a risk-based strategy in a tiered assay format [9]. Therefore, once ADAs are detected in the initial immunogenicity test (often a screening and confirmation immunoassay), at an appropriate time in clinical development it is important to further evaluate the ADA positive samples for neutralizing capability of the biological therapeutic.

NAb assay formats are commonly guided by the biotherapeutic method of action. Given the need of NAb assays to demonstrate drug neutralization in a biologic matrix, NAb assays are commonly very susceptible to drug and matrix interference [10]. Competitive ligand-binding assays (CLBAs) are a NAb assay format commonly applied to biotherapeutics with lower immunogenicity risk (i.e., antagonistic monoclonal antibody [mAb] binding to soluble targets), provide improved assay performance (have better drug tolerance, sensitivity and target tolerance) and are operationally simpler than cell-based NAb assays [11]. While the CLBA format may be simpler, one of the most common challenges of CLBA NAb assays is attaining sufficient assay drug tolerance. Drug tolerance is the ability to measure low levels of ADA in the presence of drug in the samples. Our objective for this study was to re-develop an insufficiently drug tolerant CLBA NAb assay through the addition of an acid dissociation step to improve the drug tolerance to greater than 1 μg/ml for assay use in an on-going clinical program.

Here we present the development and validation of a drug tolerant CLBA that is specific for the detection of NAbs generated against Drug 1 (a mAb generated against a 14 kDa human monomeric glycoprotein that functions as a cytokine). The assay design was originally comprised of biotinylated drug as the capture reagent, and a ruthenylated target (Ru)-target as the detection reagent. Ideally, NAbs should block the binding of the Ru-target and reduce the electrochemiluminescence (ECL) signal, with the negative control (NC) giving the highest ECL (0% inhibition) and the high positive/quality control (HQC) having the lowest ECL (e.g., >80% inhibition).

The critical reagents available for this assay created three unique challenges. First, this biologic drug was a mAb produced in a human cell line, which displayed distinctly different effects to biotin labeling than other mAbs we have previously worked with that were produced in Chinese hamster ovary (CHO) cells. Second, we did not have a suitable monoclonal or polyclonal Ab against the drug to use as a positive control (PC), that gave us the desired level of inhibition of the NC signal that we were hoping to achieve, and third, longer incubations with the detection target reduced PC inhibition of the ECL signal.

In addition to the challenges faced with the critical reagents, achieving drug tolerance was also a major concern. Under normal conditions, the original assay had less than 50 ng/ml drug tolerance at 500 ng/ml PC, so attempting an acid dissociation step seemed reasonable. When ADA in subject serum samples is already bound to drug that has not dissipated following dosing, this can inhibit the ADA from binding to the labeled drug. Often an acid dissociation is required. This consists of an acid pretreatment of samples to dissociate the drug-bound ADA, usually employing an acid at pH 2.5–3.0. This is followed by neutralization of the acidified sample with a basic buffer (i.e., [hydroxymethyl] aminomethane or TRIS). This has become a frequent practice for attaining higher assay drug and target tolerance levels [12,13].

Many surrogate polyclonal positive control Abs (PCs) used in ADA and NAb assays include a mix of binding Abs, some of which neutralize and others which do not. Additionally, in some cases, NAbs are only partly effective at blocking the active site of an antibody therapeutic and the labeled target can still bind to the biotinylated drug (especially with a smaller molecular weight target proteins), that will then reduce the ability of the PC to achieve a high percentage of inhibition. This can be caused when a PC binding affinity to the drug is substantially weaker than the drug binding affinity to its target. We generated and tested three different PCs and encountered different challenges with each but were finally able to achieve our goal of a robust NAb assay that is drug and target tolerant. The assay's sensitivity, precision, selectivity, drug tolerance, target interference, matrix interference, specificity, robustness and stability were evaluated during the validation.

2. Material & methods

2.1. Reagents

2.1.1. Drug

Drug 1: A mAb against a monomeric glycoprotein that functions as a cytokine, which was produced in Human PER.C6™ cell line (immortalized primary human embryonic retinal cells).

2.1.2. Positive controls

PC-1 was a polyclonal rabbit immunoglobulin against Drug 1 that was used in the original NAb assay, and it was produced, and affinity purified by LabCorp (PA, USA). PC-2 was an anti-Drug 1 Fab-Max-FH, human combinatorial antibody libraries (HuCAL) Fab bivalent antibody lacking the Fc region, which was produced by Bio-Rad Laboratories (CA, USA). PC-3 was a recombinant full length mAb (HuCAL) specific to antigen AbD35889ia-4GSK on Drug 1 containing disulfide bonds and was produced by Bio-Rad Laboratories. PC concentrations were set at 5000 ng/ml HQC, 1250 ng/ml MQC and 750 ng/ml LQC.

2.1.3. Negative control

Pooled Human Serum Cat# HMSRM (BioIVT, NY, USA).

2.1.4. Biotin labeling of abs

Drug 1 was buffer exchanged using PBS prewashed Zeba Spin columns Cat# 87764 (Thermo Fisher Scientific, MA, USA). The Drug 1 was then biotin labeled with EZ-Link NHS-LC-Biotin (Cat# 21336, Thermo Fisher Scientific) or EZ-link-NHS-PEG12-Biotin (Cat# A35389, Thermo Fisher Scientific) as per manufacturer's instructions. Specified volumes for each challenge ratio of the respective reconstituted biotin solutions were added to Drug 1. The Drug 1-biotin conjugation reactions were incubated at two different temperatures, ambient and 2–8°C, for 30–120 min before they were washed using PBS equilibrated Zeba Spin columns to remove any free biotin.

2.1.5. Liquid chromatography tandem mass spectrometry

Water solution containing 0.1% formic acid Catalog #900687 (Sigma-Aldrich, MO, USA), acetonitrile Catalog # 34851 (Sigma-Aldrich), acetonitrile solution containing 0.1% formic acid Catalog # 900686 (Sigma-Aldrich), formic acid Catalog # 533002 (Sigma-Aldrich).

2.1.6. Mass spectrometry

Biotin–drug (Bt-drug) samples were reduced by diluting the sample to 5 μg/ml containing 20 mM dithiothreitol (DTT) Cat# D1532 (Thermo Fisher Scientific), and incubating the sample at 60°C for 15 min. In preparation for LC-MS analysis, reduced Bt-Drug 1 was diluted in 10% glacial acetic acid. Bt-Drug 1 was analyzed on a Waters Acquity M-Class UPLC coupled with a Waters iKey 150 mm × 50 mm Protein BEH C4 300 A 1.7 μm column, Waters Synapt G2-Si mass spectrometer and Acquity Quadrupole Dalton-Based detector, all from Waters (MA, USA) to characterize the degree of labeling (DOL) and label efficiencies. Mobile phases consisted of water with 1% formic acid (Phase A) and acetonitrile with 1% formic acid (Phase B). The initial condition was 80:20 ratio of A:B phases was held until 3.25 min, and then a gradient method of flow was used in which the mobile phase conditions shifted to 30:70 A:B until 6 min, and then back to 80:20 A:B until the end of the run at the total run time of 7 min. The flow rate during the run was 5 μl/min. The samples were analyzed using reduced MS conditions previously described [14].

2.1.6. Biolayer interferometry

The drugs' dissociation constants (KD's) for the conjugated target and the PC Abs (i.e., PC-1, PC-2 and PC-3) were determined using the Octet 384 (ForteBio, CA, USA). Bt-Drug 1 was diluted to 1.5 μg/ml in 1× PBS +0.05% Tween-20 (running buffer) and immobilized onto streptavidin biosensor tips (Sartorious, CA, USA) for 5 min. Samples were serially titrated threefold to 100, 33.3, 11.1 and 3.7 nM concentrations in running buffer. KD's were calculated using the ForteBio Data Analysis 12.2 software.

2.1.7. Optimization of biotin labeling

To determine the DOL and label efficiency for the Bt-Drug 1 preparations, we followed the previously described method [14,15]. Optimization was performed in three batches: the original batch 1 (small-scale 5:1), batch 2 (a second small-scale 5:1) and batch 3 (large-scale 5 mg at 5:1). Scouting the best biotin labeling parameters for these batches of Bt-Drug 1 was performed using a challenge ratio of 5:1 biotin:Drug 1 with reaction incubation at 2–8°C. Every 10 min during the biotin labeling incubation, a small volume was taken from the reaction and analyzed by reduced LC-MS to track DOL and label efficiency values.

2.1.8. Preparation of Ru-target & Ru-target-Fc

Commercial target (Miltenyi Biotech, CA, USA) and target-Fc (Sino Biological, PA, USA) proteins were labeled with ruthenium by resuspending 150 nM vials of Sulfo-tag (MSD, MD, USA) in 50 μl of Milli-Q-Water. Specific Sulfo-Tag amounts corresponding to a molar ratio of 12:1 were then added to each protein, the reactions shielded from light and rotated end over end at ambient temperature for 2 h. After incubation, the ruthenium labeled reagents were added to PBS equilibrated Zeba Spin Desalting Columns to remove free Sulfo-tag.

Neutralization solution: 45% TRIS-HCL 1 M pH 9.5 (Teknova, CA, USA), into 55% StartingBlock™-TBS Blocking Buffer Cat# 37542 (Thermo Fisher Scientific).

2.1.9. General reagents

MSD Streptavidin Gold Plate Catalog #L15AS (Meso Scale Diagnostics, MD, USA), MSD Read Buffer T (4x) is a TRIS-based buffer containing tripropylamine (TPA) as a co-reactant for light generation in electrochemiluminescence immunoassays (Meso Scale Diagnostics). Dulbecco's phosphate-buffered saline (DPBS) 10X, no calcium, no magnesium Cat# 14200075 (Thermo Fisher Scientific).

2.2. Statistical methods

Cut point assessment: to determine whether or not a sample is positive or negative for NAb, a cut point must be established. A sample is considered positive for NAb if the percent inhibition generated by that sample is greater than or equal to the cut point value. The percentage of inhibition is determined by subtracting the sample ECL value from the NC ECL value (the reduction in signal), dividing this number by the NC ECL value, and multiplying this number times 100 or (NC-Sample)/NC) *100. To establish the cut point, 30 individual normal human serum samples were analyzed to provide a statistically valid assessment of both biological and analytical variability by two analysts on six separate occasions over 3 days, for a total of 180 observations. The analysis of the screening cut point proceeded according to the workflows outlined in [11,16,17] for the screening cut point analysis of ADA and NAb validations. After identification and exclusion of outliers (2 analytical and 5 biological) in screening sample data, the screening cut point of 32.11% inhibition was statistically calculated for a 1% false positive rate. The statistical analysis was performed using SAS 9.4 TS Level 1M5 software.

2.3. NAb assay method

Samples were diluted 4:1 in 2.25 M HCl and then incubated for 30 min at room temperature with shaking. A neutralization & capture solution was prepared containing 50 ng/ml biotinylated drug in 0.45 M TRIS. The pH of this neutralization & capture solution is very high (pH 9.0), so the Bt-drug is added just prior to quenching the acidified samples. Having the conjugated drug in basic conditions for a short period of time did not impact the NC assay response (data not shown). The neutralized samples (pH 7.4) contain equal volumes of the acidified samples (75 μl) and the neutralization & capture solution (75 μl). These samples were then incubated for 60 min shaking at room temperature. At the same time, the MSD standard SA plates were blocked with 150 μl/well starting block buffer for 1 h at room temperature. The blocking buffer was removed from the MSD plate, the plates tapped dry on an absorbent pad, and 60 μl of neutralized samples per well were then added to the MSD Streptavidin plate for one hour at room temperature with shaking. After washing the plate three-times with 1XPBST, MSD Read Buffer T (2X) was added and the plate was read on an MSD Sector Imager S600. To determine the percent inhibition, the formula ((NC-Sample)/NC) *100 was used.

3. Results

Developing a competitive ligand binding method to identify NAbs among ADAs and native IgG is challenging. The challenge may be compounded due to the idiosyncrasies of the assay critical reagents, including labeled drug and target, required to create an assay capable of demonstrating specific signal inhibition in a complex biologic matrix. For the method described, detecting NAbs to a therapeutic mAb (Drug 1), the generation of a biotin labelled drug was particularly challenging. Initial labeling attempts using internally standard challenge ratios of 8:1 and 20:1 (label: mAb) were not optimal. Subsequent labeling efforts using 5:1, 12:1 (PEG), 50:1, 100:1 resulted in successively higher challenge ratios producing inversely lower signals in the assay (Table 1). LC-MS Analysis showed that the majority of the biotin label was attaching to the light chain of the therapeutic mAb prior to attaching to the heavy chain (Figures 1–2) and the biotin labeling was possibly blocking the target binding site. There was also the possibility that heavy biotin labeling of the light chain was causing the labeled drug to attach to the streptavidin plate in an orientation that inhibited PC binding. To test this hypothesis, we labeled the mAb with biotin attached to a polyethylene glycol (PEG) linker which should have allowed more flexible movement of the drug and reduced steric hindrance. However, no improvement over a succinimidyl-6-(biotinamido) hexanoate (LC-biotin) labeled batch produced using the same challenge ratio was observed, supporting the conclusion that over labeling was reducing PC binding, and not sub-optimal streptavidin binding orientation (Table 1). While most therapeutic mAbs labeled for use in completive ligand binding NAb assays are not as sensitive to moderate levels of biotinylation, we hypothesized that very low levels of biotinylation may be necessary to optimize assay performance for this specific therapeutic mAb, as the biotin was preferentially binding to the light chain, and a high number of biotins on the light chain diminished performance.

Table 1.

List of the biotin-drug challenge ratios assessed, and the average percent inhibition measured with the HQC (5000 ng/ml positive control).

No Fc linker on ruthenium-target	Biotins per IgG	Percent incorporated	Heavy chain labeling	% Inhibition with HQC
Bt-drug (CR5:1)	1.20	47	No	59.9
Bt-drug (CR8:1)	0.74	37	No	66.7
PEG-Bt-drug (CR12:1)	1.10	100	No	39.8
Bt-drug (CR20:1)	1.86	82	No	54.9

With Fc linker on ruthenium-target
Bt-drug (CR5:1)	1.20	47	No	94.4
Bt-drug (CR8:1)	0.74	37	No	89.6
Bt-drug (CR20:1)	1.86	82	No	86.6
Bt-drug (CR50:1)	2.94	91	Yes	28.8†
Bt-drug (CR100:1)	4.39	100	Yes	0.0†

^†Electrochemiluminescence signal diminished 90+%.

Bt: Biotin.

Figure 1.

LC/MS analysis of LC-Bt-Drug-1 comparing 5:1 and 50:1 Bt-Drug challange ratios. (A1) Reduced LC-MS of the light chain of the final reagent preparation of the large-scale Bt-Drug 1 reagent at a challenge ratio of 5:1, showing an average of 1.11 biotins on the light chain. (A2) Reduced LC-MS of the heavy chain of the same 5:1 Bt-Drug 1 reagent showing no biotin labels on the heavy chain. (B1) Reduced LC-MS of the light chain of a Bt-Drug 1 reagent at a challenge ratio 50:1 showing an average of 2.94 biotins on the light chain. (B2) Reduced LC-MS of the heavy chain of the same 50:1 Bt-Drug 1 reagent showing an average of 3.54 biotins on the heavy chain.

Figure 2.

LC-MS comparison of the drug biotinylated at two different challenge ratios to unlabeled drug.

The optimized biotin labeling incubation time for Bt-Drug -1 batch 2, to produce a suitably biotin labeled reagent that did not interference with the binding activity of the drug, was determined empirically by percent inhibition in the assay. The biotinylated capture (Drug 1) was atypically sensitive to over labeling in this assay. The DOL with the final selected small scale labeling parameters was 1.2 biotin: IgG (0.63 biotin:IgG per each single light chain), 53% label efficiency (light chain only) and the heavy chain containing no labels. Typically, when an antibody is labeled with biotin, the heavy chain is labeled more than the light chain. However, in this specific case, the assay results indicate that if the DOL is greater than 0.63, the light chain was overlabeled. This can cause the labeled drug to lose its binding affinity to both the PC and the target. Following the biotin labeling parameters selected from the scouting experiments, a scaled-up batch of the 5:1 Bt-drug -1 batch 3 was developed and supported the entire assay validation. The DOL for the scaled-up batch (5 mg) was 1.1 biotin:IgG or 0.55 biotin:IgG per single light chain (See Table 1), which was calculated from the LC-MS analysis for this batch shown in Figure 1A. The best assay performance was obtained by labeling with a low, 5:1 (biotin: mAb) challenge ratio on ice for 80 min (Figure 1A). The ice was important to slow down the labeling reaction, since even a 3:1 (biotin: mAb) challenge ratio at RT for only 20 min resulted in slightly over labeled light chain of the drug (data not shown).

With the Bt-Drug 1 capture problem solved, a second challenge presented itself. The second PC tested, PC-2 (anti-Drug 1 Fab-Max-FH) worked better than PC-1 as a NAb with our Bt-drug, but the method had no drug tolerance. We found that the ADAs bound tightly to the drug and required a high acid concentration (450 mM HCl final) to show significantly improved drug tolerance. PC-2, an antibody-like molecule was only held together by ionic bonds and was denatured by the acid dissociation. Therefore, PC-2 would only be sufficient as a control if analyzing samples following a drug wash-out, where no acid dissociation was required. We obtained a third PC, PC-3, a Bio-Rad humanized recombinant full-sized Ab with disulfide (covalent) bonds at the hinge, which was able to survive the acid dissociation intact, but could only achieve a maximum of 40% inhibition over NC, and it was also rejected as an assay control. PC-1, a polyclonal rabbit Ab, was able to achieve 60–70% inhibition with the HQC after acid dissociation. Because PC-1 looked the most promising of the three, it was selected as the sole positive control antibody. However, we soon found that the percent inhibition of PC-1 decreased proportionally to ruthenium tagged targets' incubation time, suggesting that the ruthenium tagged target was either co-localizing with the PC on the drug or displacing the PC (data not shown).

To examine the time dependency of the inhibition, we tested the binding affinity between the antibody and its antigen and found that the target antigen bound the drug with an affinity 14-fold stronger than the interaction between PC and drug (Supplementary Figure S1). Since further increasing the PC's affinity for drug would be challenging, we decided to explore reducing the binding affinity of the target by increasing its size as a fusion protein and likely introducing limited steric hinderance. Hence, we acquired, and ruthenium labeled target linked with a recombinant Fc (crystallizable fragment region of human immunoglobulin). This reduced the binding affinity almost fourfold (Supplementary Figure S1) and improved the inhibition. With the HQC using the polyclonal PC-1, we were able to achieve over 90% inhibition of the NC signal.

With biotin labeling and PC optimization completed, the method required a few minor modifications to reach the optimum format. It was observed that assay performance improved with lower dilution factors. While maintaining the acid treatment and neutralization step, a 1:4 minimal required dilution (MRD) produced the best assay performance. Although slightly lower dilution factors were possible by using more concentrated acid and neutralization buffer, we found that strong acid, 2.25 M HCl (450 mM HCl final), does not work well with automation instruments, and stronger acid would be more likely to permanently damage the Abs, decreasing ECL signal overall.

Finally, with the correct dilution factor, the optimal biotin labeled drug ratio, using the sulfo-tag labeled target with the Fc linker, and selecting the best PC, we were able to achieve a consistently high percentage of inhibition with our HQC and acceptable drug tolerance (500 ng/ml PC still positive with 1 µg/ml drug, and 1 μg/ml PC still positive with 15 µg/ml drug), which is greater than the expected trough drug concentration of <1 µg/ml (Figure 3).

Figure 3.

Drug tolerance tested at 500, 600, 750 & 1000 ng/ml positive control and up to 15 μg/ml drug.

The final method performance results for the NAb assay validation are summarized in Table 2. After the removal of outliers (2 analytical and 5 biological resulting in 173 observations), a cut point of 32.11% inhibition was calculated, and when applied to the data, a false positive rate of 3.3% was observed for the population screened. In the validated method, a sample with a % inhibition greater than or equal to 32.11% inhibition is positive for NAbs against the drug. The sensitivity curves are shown in Figure 4, and the target tolerance is shown in Figure 5.

Table 2.

NAb validation summary.

Cut point	32.11 % inhibition
Sensitivity	459 ng/ml
Precision	Positive control	Intra-assay	Inter-assay
	HQC	0.4–11.6%	12.80%
	MQC	0.0–8.8%	6.70%
	LQC	0.7–13.0%	6.30%
Selectivity	Spiked NC	100% met criteria
Drug tolerance/interference	Positive control ng/ml	Drug μg/ml
	1000	15
	750	3–6
	500	1.0
Target tolerance/interference	Positive control ng/ml	Target pg/ml
	500	1000
	750	1000
	NC	1000
Matrix interference	Hemolysis	No interference observed
	Lipemia	No interference observed
	Biotin	No interference observed below 100 ng/ml
Specificity	2nd monoclonal antibody	No interference observed
Robustness	All controls met acceptance criteria when varying instruments, plate lots and maximum incubation times
Stability	Condition	Maximum Acceptable
	Freeze–thaw (-70°C to ambient)	3 cycles
	Refrigerated 5°C	7 days
	Ambient	24 h

Figure 4.

Sensitivity curves. (A) Sensitivity curve showing ECL vs positive control concentration, (B) A logarithmic curve of the sensitivity tested up to 10 μg/ml shows that the validated assay has over 90% inhibition at the HQC (5000 ng/ml).

ECL: Electrochemiluminescence.

Figure 5.

Target tolerance data shows that the assay has target tolerance at the LQC (750 ng/ml) up to 1000 pg/ml.

4. Discussion

NAb assays are required by regulatory agencies to characterize ADA and are implemented during clinical development according to mode of action and perceived immunogenic risk. The challenges in developing a functional NAb assay, whether cell based or CLBA, are multi-faceted. In vitro detection of a NAb preventing a drug from binding its target may be challenging due to interfering factors, including matrix effects and target and drug interference. The assay development described here encountered an additional hurdle that we had not faced in prior NAb assay development programs. Unlike our previous experience with CHO produced therapeutic mAbs, this human cell line derived mAb was not being conjugated in the heavy chain region. The biotin conjugation almost exclusively occurred in the light chain/variable region (Figure 1) and showed diminished performance in the assay (lower % inhibition of PCs and lower ECL for the NC) as DOL increased (Table 1). As a result, a modified approach to biotin labeling process was required to obtain acceptable assay performance. This included evaluating various challenge ratios, linker types, reaction temperatures and reaction times. The labeling incubation was performed on ice instead of room temperature to reduce the thermodynamic motion of the side chains and their accessibility to the conjugation process.

Lysines are the most commonly used amino acid residues for chemically linking conjugates to Abs because they are usually exposed on the surface of the antibody and the primary amines react efficiently with NHS-activated labels [18]. IgGs contain, on average, 80 lysine residues [19]; hence, labeling targets should be plentiful. In this case, the amine groups from lysine residues on the heavy chains were hindered or blocked by glycosylation, which was something we typically did not encounter with mAbs produced in CHO cells. The cell line used to produce IgG determines its glycosylation profile. Recombinant glycoproteins produced in mammalian cells, including mAbs, have shown different biological properties based on their glycan profiles [20]. The host cell line has a strong influence on glycosylation [21], with mAbs produced by PER.C6™ cells often characterized by abnormal glycosylation, being desialylated and fucosylated [22]. Therefore, this biotinylation issue could be unique to mAbs produced in this cell line PER.C6™. Different cell culture systems have various sugar populations and glycosylation enzymes that contribute to the final IgG-Fc glycosylation profile [23]. Studies have shown that the glycosylation of Abs produced in the human embryonic kidney 293-F cell line and Abs produced in the Chinese hamster ovary–K1 cell line vary greatly [20]. Drug 1 has N-linked glycosylation on each heavy chain and none on the light chain, which would lend credence to the glycosylation interference hypothesis. While reduced heavy chain conjugation might be associated with amino acid sequence of Drug 1 (i.e., fewer lysines in the Fc region), higher challenge ratios of biotin did show some conjugation to the heavy chain, but only after over conjugation to the light chain that reduced binding to both the PC (reduced inhibition) and the target-tag (reduced ECL for the NC) (Table 1). Alternatively, we did not have the capability to express Drug 1 in a different cell line (e.g., CHO) or try to deglycosylated Drug 1, to confirm glycosylation interference.

The Ru-labeled-target also created a challenge since it had a 14-fold higher affinity for the drug than the PC (Supplementary Figure S1). Because this NAb assay is stepwise, the PC is bound to the Bt-Drug 1 before the Ru-target is introduced (Figure 6). The strong acid required to dissociate the drug-PC complex in serum to achieve excellent drug tolerance suggests tight binding between the drug and PC, so it is unlikely that the target-tag was directly displacing the PC. Therefore, the target-tag was probably replacing the PC after its dissociation from drug, or able to co-localize with bound PC on the drug when given enough time. By conjugating an Fc- target fusion protein in place of the monomeric target, the Fc-target-tag decreased its binding affinity to the drug from 14-fold greater affinity than the PC for the target-tag to 3.6-fold with the Fc-target-tag. This also created a larger molecule (from 14 kDa to 40 kDa), with greater potential for steric hinderance, likely reducing the ability of the Fc-target fusion to co-localize with the drug bound PC. In spite of this, our target tolerance was over 1 ng/ml, well above the assay requirement of 15 pg/ml and the drug tolerance was over 15 μg/ml with 1 μg/ml PC. The conditions for the final assay are polyclonal PC-1, Bt-drug with a challenge ratio of 5:1 on ice for 80 min, 450 mM HCL acid dissociation and Ru-target with Fc (Figure 6).

Figure 6.

Final configuration of the neutralizing antibodies assay.

NAb: Neutralizing antibody; PC: Positive control.

5. Conclusion

We encountered a challenging competitive ligand binding method development, in which the therapeutic mAb had a stronger affinity for a soluble target than the PC and using our usual drug biotinylation approach was inadequate. Although each reagent and step provided challenges, we were able to develop and validate a robust NAb assay that was both drug and target tolerant.

CLBAs can be quite different from ADA assays, and one should not assume that a reagent that performs well in an ADA assay will also perform well in a NAb assay. With careful characterization of the biotinylate drug, and altering the conjugation reaction conditions, the best reagent was selected to develop this NAb assay. Having a critical reagents group to readily analyze and alter the reagents as needed to improve performance was a great resource to have available. With the knowledge gained from this validation and especially for those who do not have the ability to analyze their reagents with LC-MS, we would suggest adding a low biotin challenge ratio (i.e., 1:5) to those you typically run at the beginning of development, when labeling the drug.

The learnings from these experiments will be a major help in developing future NAb assays and allow us to recognize and address these issues sooner if they are encountered in the future.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/17576180.2024.2360363

Author contributions

All authors were involved in the overall design and execution of the study and authoring of the manuscript. In addition, B Rowe and K Carle conducted the experiments described, A Concannon was involved in reagent preparation, B Rowe, K Carle, Y Jiang, K Reese and E Meyer were involved in the analysis of data. E Meyer was responsible for data review, and A Gehman was responsible for statistical analyses. K Chen and K McCarthy were involved in assay design and validation.

Financial disclosure

All authors are employees of GSK PLC. Some authors are owners of GSK PLC shares. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, stock ownership or options and expert testimony.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

The authors state that they have obtained appropriate institutional review board approval for all experimental investigations. In addition, for investigations involving human subjects, informed consent has been obtained from the participants involved.