Journal of Immunological Methods 416 (2015) 94–104

Contents lists available at ScienceDirect

Journal of Immunological Methods journal homepage: www.elsevier.com/locate/jim

Development and characterization of a pre-treatment procedure to eliminate human monoclonal antibody therapeutic drug and matrix interference in cell-based functional neutralizing antibody assays Weifeng Xu ⁎,1, Hao Jiang 1, Craig Titsch, Jonathan R. Haulenbeek, Renuka C. Pillutla, Anne-Françoise Aubry, Binodh S. DeSilva, Mark E. Arnold, Jianing Zeng, Robert W. Dodge ⁎⁎ Analytical and Bioanalytical Development, Bristol-Myers Squibb, Princeton, NJ 08540, United States

a r t i c l e

i n f o

Article history: Received 12 September 2014 Received in revised form 25 October 2014 Accepted 3 November 2014 Available online 8 November 2014 Keywords: Neutralizing antibody Bioassay Sample pre-treatment Acid dissociation Monoclonal antibody therapeutics LC-MS/MS

a b s t r a c t Biological therapeutics can induce an undesirable immune response resulting in the formation of anti-drug antibodies (ADA), including neutralizing antibodies (NAbs). Functional (usually cellbased) NAb assays are preferred to determine NAb presence in patient serum, but are often subject to interferences from numerous serum factors, such as growth factors and disease-related cytokines. Many functional cell-based NAb assays are essentially drug concentration assays that imply the presence of NAbs by the detection of small changes in functional drug concentration. Any drug contained in the test sample will increase the total amount of drug in the assay, thus reducing the sensitivity of NAb detection. Biotin-drug Extraction with Acid Dissociation (BEAD) has been successfully applied to extract ADA, thereby removing drug and other interfering factors from human serum samples. However, to date there has been no report to estimate the residual drug level after BEAD treatment when the drug itself is a human monoclonal antibody; mainly due to the limitation of traditional ligand-binding assays. Here we describe a universal BEAD optimization procedure for human monoclonal antibody (mAb) drugs by using a LC–MS/MS method to simultaneously measure drug (a mutant human IgG4), NAb positive control (a mouse IgG), and endogenous human IgGs as an indicator of nonspecific carry-over in the BEAD eluate. This is the first report demonstrating that residual human mAb drug level in clinical sample can be measured after BEAD pre-treatment, which is critical for further BEAD procedure optimization and downstream immunogenicity testing. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Numerous biological therapeutics, including recombinant proteins, peptides, nucleic acids and carbohydrates, have been

⁎ Corresponding author. Tel.: +1 609 252 3120. ⁎⁎ Corresponding author. Tel.: +1 609 252 5204. E-mail addresses: [email protected] (W. Xu), [email protected] (R.W. Dodge). 1 Equal contributors.

http://dx.doi.org/10.1016/j.jim.2014.11.005 0022-1759/© 2014 Elsevier B.V. All rights reserved.

approved for clinical use (Food and Drug Administration). However, administration of biological therapeutics has the potential to induce undesirable immunogenicity, resulting in the development of anti-drug antibodies (ADA), including neutralizing antibodies (NAbs) (Dodge et al., 2009; Civoli et al., 2012; Food and Drug Administration, 2013). NAbs diminish therapeutic efficacy by either preventing the drug from binding to its target or inhibiting downstream signaling upon binding due to steric hindrance. In some cases, NAbs can cross-react and neutralize the biological activity of an endogenous counterpart, resulting in the impairment of an essential normal physiological

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

function and life-threatening adverse effects (Casadevall et al., 2002). Monoclonal antibody (mAb) therapeutics have been established as a class of favorable therapeutic agents as demonstrated by the approval of dozens of such products into the market (Beck et al., 2008; Beck et al., 2010). These therapeutic molecules provide high specificity toward their intended targets and often exhibit less off-target toxicities when compared to small molecule drugs. However, even fully humanized mAb therapeutics still have the potential to induce ADA, including NAb (McCutcheon et al., 2010) (Hu et al., 2009). It is currently recommended by health authorities that a functional cell-based NAb assay is preferred whenever possible for the characterization of the neutralizing potential of ADA (Food and Drug Administration, 2013). In a conventional cellbased functional NAb assay, known amounts of drug (system drug) are pre-incubated with samples before adding to the functional bioassay. Any statistically significant signal change (level of functional drug), when compared to drug only control, implies the presence of NAb in the test samples. If NAb positive samples also contain drug from treatment, NAb will be preoccupied with drug in the sample and will no longer be available to bind to the system drug in the bioassay; this would lead to reduced NAb detection and/or no NAb detection, depending on the molar ratio of NAb to drug in the test samples. Since mAb therapeutics usually are dosed at high concentrations and have long half-lives, they typically are present at high circulating levels in serum throughout sampling. These high levels of mAb therapeutics in test samples are particularly problematic for functional cell-based NAb bioassays (Hu et al., 2009). A cell-based assay to detect drug concentration should represent the in vivo mechanism of action of the drug as closely as possible (Food and Drug Administration, 2013). As many mAb drugs interact with cell surface receptors, a functional bioassay will typically involve human cells. A clinical serum sample to be assessed for the presence of NAbs will contain many components (growth factors, cytokines, etc.) that may affect cells in the bioassay. Thus, the matrix factors in the sample may affect the functional assay readout regardless of the presence of NAbs. While small perturbations from matrix interfering factors on functional bioassay readout may be tolerable, large subject-to-subject and time point-to-time point variation may make it impossible to accurately characterize a sample for the presence of NAbs. To circumvent both drug interference and matrix interfering factors, we developed a procedure to extract ADA with modified BEAD prior to characterizing their neutralizing ability with a cell-based functional assay (Lofgren et al., 2006). Several important questions need to be answered when an extensive pre-treatment sample extraction procedure is put in place prior to a functional cell based assay. First, the recovery of the NAb should be determined. This recovery should be assessed in the presence of varying amounts of drug in a sample to determine relative efficiency of the pre-treatment extraction method in samples having varying levels of drug. A second question about efficiency and consistency of a pretreatment procedure is what is the level of drug remaining in a sample after the extraction. While it is straightforward to measure drug concentration in a sample using a ligand binding assay, it may not be trivial to quantitate levels of mAb drug in

95

the presence of varying concentrations of a NAb. This is because NAb will interfere with anti-idiotypic capture and detection reagents in a ligand binding assay. Also, generic ligand binding assays using non-specific Fc-based detection reagents will not work in the presence of high concentrations of human IgG. Therefore, to answer key questions about the efficiency and consistency of our optimized pre-treatment extraction procedure, we developed an LC–MS/MS assay to simultaneously measure drug (a mutant human IgG4) and NAb PC (a mouse IgG). Also, as a measurement of non-specific serum proteins carried over into a sample after extraction procedure, we measured endogenous human IgG concentration in the BEAD eluate. This is the first report that a fully human mAb drug can be accurately detected after clinical samples are pre-treated with BEAD procedure. 2. Material and methods 2.1. Reagents RPMI-1640, DMEM, heat-inactivated fetal bovine serum (FBS), 2-mercaptoethanol, hygromycin B and G418 were all purchased from Gibco/Life Technology (Grand Island, NY). Hen Egg Lysozyme (HEL) peptide (48-62, DGSTDYGILQINSRW) was ordered from Genscript (Piscataway, NJ). AphaLISA mouse IL-2 Kit was purchased from PerkinElmer (Waltham, MA) and SeraMag Magnetic Streptavidin Blocked Microparticles were from Thermo (Rockford, IL). Pooled human serum, individual normal human serum and individual diseased human serum were purchased from Bioreclamation (Hicksville, NY). The 3A9 mouse T cell and LK35.2 mouse B cell lines were obtained from the BMS Cell Bank. The neutralizing positive control antibody was a monoclonal mouse affinity-purified antibody to the drug product produced in house. The drug product (designated S229P mutant hIgG4) was manufactured in house at BMS and the biotinylation of the drug product was done in house or by Radix BioSolutions (Georgetown, TX). 2.2. Cell culture The 3A9 and LK35.2 cells overexpressing our drug target and its ligand were expanded in vented cap cell culture flasks (BD Falcon, Franklin Lakes, NJ) at 37 °C, 5% CO2 and 95% relative humidity (RH). The growth medium was RPMI 1640 with 10% heat-inactivated FBS, 55 μM 2-mercaptoethonal and 2 mM L-glutamine supplemented with 800 μg/mL Hygromycin B for 3A9 cells and 500 μg/mL G418 for LK35.2 cells, respectively. Both cell lines grow rapidly and may be split 1:10 at approximately 2 day intervals. Cells were frozen with pure FBS supplemented with 10% DMSO. Upon thawing, cells were washed once and resuspended in DMEM medium supplemented with 10% FBS and 20 μM HEL peptide and used directly in the bioassay. 2.3. Preparation of test samples for biotin-drug extraction and acid dissociation (BEAD) procedure Different concentrations of NAb PC were spiked in pooled or individual healthy human serum, with or without 100 μg/ml drug product, and incubated at room temperature on a rotator for 4 h to allow antigen–antibody binding and for the formation

96

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

of immune complexes. Different concentrations of drug product were also spiked into human serum and prepared in the same way. Samples were then aliquoted and frozen at −70 °C until use. 2.4. Bead extraction and acid dissociation (BEAD) procedure We adapted a previously published solid phase or bead extraction with acid dissociation procedure to extract ADA with minor modifications (Lofgren et al., 2006). Briefly, 100 μL of human serum samples and controls were first mixed with an equal volume of 400 mM glycine-HCl, at pH 2.0, and incubated at room temperature (RT) for 60 min on a shaker (Labnet Orbit P4, Woodbridge, NJ) at 1200 rpm. To each sample, 27.5 μL 1.8 M Trizma Base (pH 8.8) containing 182 μg/mL biotinylated-drug was added for 90 min on the same shaker at 1200 rpm. ADAs, dissociated from drug product, would bind to biotin-drug and were then immobilized on 25 μL streptavidin-coated magnetic beads added at 6 mg/mL. Bead-complexes were then captured using a microplate magnet (96 F Magnet LifeSep Biomagnetic Separators, BioTek, Winooski, VT) and manually washed three times with PBST. Bound ADAs were released from the beadcomplex by a second acid treatment with 90 μL of 300 mM glycine at pH 3.0, for 10 min shaking at 1200 rpm at RT. Eighty microliters of this ADAs-containing acid solution was then transferred to a new plate and neutralized with 25 μL of 400 mM HEPES and complete DMEM medium mixed at 1:1.7 ratio. 2.5. Bioassay for detecting neutralizing activity The IL-2-releasing bioassay was used to assess the absence, presence, or amount of the anti-therapeutic protein neutralizing antibodies in the samples. Briefly, 40 μL of BEAD eluate from 2.4 was incubated with 40 μL of 500 ng/ml of system drug in a NUNC edge plate for a minimum of 20 min at RT. Fifty thousand freshly thawed T cells, washed and resuspended in 40 μL of complete DMEM medium containing 20 μM of a stimulating peptide, were then added and incubated at RT for another 20 min. At the last step, 25,000 freshly thawed B cells, washed and resuspended in 40 μL of complete DMEM medium containing 20 μM of stimulation peptide, were added, mixed and the whole plate was transferred to an incubator set at 37 °C, 5% CO2, and 95% humidity for 20 to 24 h. The contents in the wells were mixed by gently tapping on the plate side, centrifuged at 1200 rpm, and 20 μL of supernatant was removed from each well. The levels of IL-2 expressed were analyzed using a mouse IL-2 alphaLISA kit from PerkinElmer following the manufacturer's instructions. The optical density of the contents of each well was determined using an EnSpire (PerkinElmer, Waltham, MA) plate reader with default 96-well alphaLISA protocol. 2.6. LC–MS/MS assay Detailed assay development and qualification of LC–MS/MS method can be found here (Jiang et al., 2014). Briefly, the BEAD eluate samples (25 μL) were mixed with a stable isotopically labeled (SIL) form of the drug as an internal standard, followed by reduction with dithiothreitol at 60 °C for 30 min and alkylation with iodoacetamide at 30 °C for 30 min. The treated

samples were then digested with trypsin at 50 °C for 90 min. The digestion reaction was quenched by adding 25 μL of 10% formic acid in acetonitrile. The tryptic digests were filtered prior to injection into a high performance liquid chromatography system (HPLC, model LC-30 AD, Shimadzu Scientific Instruments, Inc.) fitted with an Acquity UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm, Waters Co.). The HPLC mobile phases of 0.1% formic acid (A) and 0.1% acetonitrile (B) were delivered under a gradient program, 5%B to 42%B for 9 min. The flow-rate was 0.15 mL/min. The HPLC eluent was introduced into an AB Sciex 5500 triple-quadrupole mass spectrometer, followed by ionization in positive electrospray mode (curtain gas, 30 units; CAD gas, 8 units; gas 1, 40 units; gas 2, 50 units; ion spray voltage, +4000 V; temperature, 500 °C), and analyzed by the triple quadrupole analyzers in multiple reaction monitoring (MRM) mode. Specific and sensitive tryptic surrogate peptides for the drug, human IgGs, and mouse NAb were monitored simultaneously and acquired chromatographic peaks were integrated by the Analyst® software (version 1.5.1, AB SCIEX), followed by exporting peak areas to Watson LIMSTM (version 7.3, Thermo Fisher Scientific Inc.) for calibration curve regression and back-calculation of the concentrations of quality control samples and BEAD eluate. 2.7. Size exclusion chromatography Samples of biotin-conjugated drug reagent were analyzed using size exclusion chromatography (SEC) on an Agilent 1100 HPLC instrument (Agilent Technologies). The separation was performed using a Yarra 3000 column (4.6 × 300 mm, 3 μm, Phenomenex Inc.). Samples were injected neat (10 μL) and eluted at a flow rate of 0.5 mL/min over 15 min with 100 mM sodium phosphate and 0.025% sodium azide. Spectra were collected using a diode array detector set at 280 nm (reference at 360 nm). The resultant chromatograms were analyzed using Openlab Chemstation Edition (Revision C.01.04). 2.8. High resolution mass spectrometry (HRMS) Samples of biotin-conjugated drug were mixed with an equal volume of 10 mM tris(2-chloroethyl)phosphate (TCEP). The resultant mixture was injected (5 μL) into an Acquity H-class Bio ultra high pressure liquid chromatography system (UPLC, Waters Co.) fitted with a Kinetex C8 column (2.1 × 50 mm, 1.7 μm, Phenomenex Inc.). The mobile phases of 0.1% formic acid aqueous (A) and 0.1% formic acid in acetonitrile (B) were delivered under a gradient program, 20% B to 80% B over 7.0 min (curve factor 4), 80% B to 90% B over 0.5 min (curve factor 4), followed by re-equilibration. The flow rate was set to 0.3 mL/min and the column was held at 80 °C. The column eluent was introduced into a Bruker Daltonik MaXis 4G q-TOF mass spectrometer. The ionization source was set to positive polarity mode with a capillary voltage at 4.5 kV, nebulizer pressure at 1.6 bar, dry gas flow at 9.0 L/min, and temperature at 220 °C. The mass analyzer was calibrated between 300 and 2900 m/z and spectra were collected at 1.0 Hz. Profile mass spectra were summed from 3.4 to 3.7 min, smoothed, baseline subtracted and deconvoluted using a maximum entropy equation in Compass DataAnalysis 4.2 (Bruker Daltonim GmbH).

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

3. Results 3.1. Therapeutic-induced IL-2 production Treatment of target-overexpressing 3A9 and LK35.2 cells in the presence of stimulatory peptide with our therapeutic candidate induced a robust up-regulation of interleukin-2 (IL-2) in a concentration-dependent manner (Fig. 1A). We further chose a drug concentration around EC70 (500 ng/ml) and included a mouse monoclonal positive control NAb in the system, and the effect of the drug was blocked and the IL-2 production was inhibited, with an EC50 around 1 μg/ml (Fig. 1B). Although both drug and NAb neutralization curves had very good signal to noise ratios, 23 and 20, respectively, the assay had a flattened curve when 5% normal human serum was present (Data not shown). Even at this low serum concentration, individual serum sample showed high variation (Data not shown). Since the expected drug concentration in the test samples is 50–125 μg/ml, we decided to pre-treat samples with Biotin-drug Extraction with Acid Dissociation. 3.2. Residual drug estimation by alphaLISA Due to the bivalent nature of the antibody drug and the high concentration of total drug in the test samples (50–125 μg/mL),

97

there is a high probability for the formation of heterogeneous immune complexes containing one free drug and one biotindrug binding to each arm of NAb. This heterogeneous immune complex would ultimately lead to the carry-over of drug in the BEAD eluate (Fig. 7, green circle). Although the hIgG4 therapeutic contains a point mutation (S229P) to prevent the antigen binding fragment (Fab) arm exchange (Angal et al., 1993), currently there is no S229P-specific antibody to distinguish the therapeutic from endogenous hIgG4. We first attempted to estimate the amount of residual hIgG4 drug in the BEAD eluate via a ligand binding assay using an alphaLISA kit from PerkinElmer. When normal pooled human sera were spiked with increasing amounts of the therapeutic in the presence of a constant amount of NAb PCs or vice versa, an increasing amount of human IgG4 was detected in the BEAD eluate (Fig. 2A, B); this was expected and consistent with our initial concern with the ligand binding assay format. However, when different concentrations of serum were treated with BEAD procedure without any NAb PCs or therapeutic, similar amounts of hIgG4 in BEAD eluate were detected, suggesting that endogenous hIgG4 accounted for a significant portion of the detected hIgG4 in the BEAD eluate (Fig. 2C) and alphaLISA ligand binding assay was not capable of distinguishing the therapeutic from endogenous human IgG. NAb PC recovery after the BEAD procedure was also measured by alphaLISA kit for mouse IgG detection. The amount of NAb PCs recovered was similar to that of hIgG4 when NAbs were spiked at 123 ng/mL. When spiked NAbs reached 370 ng/mL, recovered NAb PCs were 5-fold higher than that of hIgG4 (Fig. 2B & D). This means that all of the hIgG4 therapeutic was complexed with NAb PC when NAb PC was spiked at concentrations above 123 ng/mL. Together with the fact that currently there is no good antibody available to distinguish S229P containing hIgG4 from endogenous hIgG4 in a ligand binding assay, the specificity of the LC–MS/MS technology was evaluated to distinguish the hIgG4 therapeutic from the endogenous hIgG4 (Furlong et al., 2014). 3.3. Specificity of LC–MS/MS to detect S229P mutant hIgG4, endogenous hIgG and mouse IgG Two peptides, one from the antigen binding region and one containing the S229P mutation in the hinge region, were found to be very specific for the detection of modified hIgG4 therapeutics by LC–MS/MS, with no cross-interference from either endogenous hIgG (containing hIgG4) or mouse IgG. We also developed an assay to specifically detect total hIgG and mouse IgG with good sensitivity and precision without crossinterference with each other (Tables 1A and 1B). 3.4. Residual drug due to addition of biotin labeled-drug, NAb PC and drug contained in the testing samples

Fig. 1. Drug induced IL-2 production. T cell line expressing target and B cell expressing the ligand were incubated with drug or drug plus NAb overnight prior to measuring IL-2 production by alphaLISA. (A) Increasing amount of drug resulted in increasing production of IL-2; (B) increasing amount of NAb PCs (pre-incubated with 1 μg/mL of drug before adding to bioassay) led to lower IL-2 production in a dose-dependent manner.

LC–MS/MS results showed that NAb alone without drug in pooled human serum samples resulted in a residual drug level ranging from 100 ng/mL to about 250 ng/mL (Fig. 3A). Since there was no therapeutic in the test sample, the residual drug in the final BEAD eluate must come from the added biotinconjugated drug during the BEAD process. Indeed, a mock BEAD procedure without adding biotin-drug showed no residual drug

98

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

Fig. 2. AlphaLISA to measure mouse IgG and human IgG4 in the BEAD eluate. (A) Increasing human IgG4 in BEAD eluate in the presence of increasing concentration of IgG4 drug with constant amount of NAb PCs in a sample; (B) increasing human IgG4 in BEAD eluate in the presence of constant concentration of IgG4 drug and increasing amount of NAb PC. (C) Increasing human IgG4 in BEAD eluate with increasing amount of human serum only. (D) Increasing amount of mIgG (NAb PCs) with increasing amount of spiked NAb with constant amount of hIgG4 drug. Data are from one of three experiments with similar results. Unit on y axis: ng/mL.

compared to samples subjected to the normal BEAD process (Fig. 4A vs. 4A″, column 1). The fact that mock BEAD group had about 1000 ng/mL total IgG but undetectable residual drug level confirmed the specificity of hIgG detection by LC–MS/MS, and excluded the possibility that the residual drug detected by LC–MS/MS was due to the interference of high amount of hIgG in the final BEAD elute (Fig. 4A″ and B″, column 1, also see Table 1B). The presence of 100 μg/mL drug in the test samples without any NAb led to only about 10 ng/mL of additional residual drug compared to no NAb and no drug group (Fig. 3B vs. 3A, Table 2). This extremely low drug carry-over rate (~0.01%) indicated

that our BEAD procedure removed drug from test samples with high efficiency. Although there was a tendency for increased drug carry-over rate from about 0.01% to 0.07% in the presence of increasing amount of NAb PC in the samples, the overall carry-over rate and absolute amount of drug carry-over due to 100 μg/mL of drug in the testing samples was still low, relative to 500 ng/mL system drug in the downstream functional cell assay. Drug alone in DMEM culture medium (containing 10% FBS) had a similar level of residual drug compared to NAb alone or NAb and drug in human serum (Fig. 3C vs. 3A and 3B). The fact that higher amounts of drug in the absence of NAb also led to

Table 1A Accuracy and precision of the LC–MS/MS peptide recovery. Surrogate peptide ID⁎

LLOQ (ng/mL)

Accuracy (%Dev)

Between-run precision (%CV)

Within-run precision (%CV)

ASGI (drug, VH) YGPP (drug, CH2) TVAA (hIgGs, CL) VVSV (hIgGs, CH2) VNSA (mNAb, CH2)

50 50 50 50 50

b±2.6 b±3.6 b±3.0 b±8.6 b±4.2

≤2.6 ≤8.0 ≤1.7 ≤9.7 ≤3.9

≤13.6 ≤6.6 ≤6.1 ≤7.7 ≤6.2

⁎ Surrogate peptides were named according to their first four amino acid codes in the sequences.

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104 Table 1B Specificity of LC–MS/MS for the drug, mouse NAb PC and human IgG.

Drug Drug Mouse NAb PC Mouse NAb PC Human IgG

Conc, μg/mL

NAb PC conca, μg/mL

Drug conca, μg/mL

Human IgG conca, μg/mL

1 20 1 20 1

N.D. N.D. 1 21 N.D.

1 21 N.D. N.D. N.D.

N.D. N.D. N.D. N.D. 0.62b

N.D., not detectable. The reference material of the intravenous immunoglobulin (IVIG), an extracted IgG product from human plasma, was spiked into the blank BEAD matrix at a concentration of 1 μg/mL. The result suggests that ∼62% of the total human IgG contain the peptide VVSV. a Mean concentration from three replicates. b The value represents the concentration of human IgG containing the quantitation peptide VVSV, because this peptide is only contained in human IgG1, IgG3, and IgG4.

higher residual drug level suggested that drug could nonspecifically bind to the magnetic beads in a concentrationdependent manner (Fig. 3C). In conclusion, biotin-drug added during sample pretreatment was the main contributor to the residual drug while drug alone at even 100 μg/mL had minimal contribution. However, the presence of NAb facilitated carry-over of the drug in a dose-dependent manner; most likely due to the formation of more biotin-drug/NAb/free drug complex (Fig. 7). 3.5. Residual hIgG level as an indication of non-specific serum factor carry-over We used LC–MS/MS to measure total hIgG, including S229P mutant IgG4 drug, in BEAD eluate. Human serum alone without spiked NAb or drug resulted in about 1000 ng/mL of hIgG in BEAD eluate (Fig. 3D, NAb at 0 μg/mL). However, increased amounts of NAb led to a 3-fold increase of IgG carry-over, from 1000 ng/mL to over 3000 ng/mL. Since there was no crossinterference between human and mouse IgG detection by the LC–MS/MS method (Table 1B), this significant increase of IgG carry-over may be due to natural antibodies against mouse immunoglobulin presented in about 10% of the human population (Klee, 2000). The more mouse NAb PC captured on the beads, the more anti-mouse natural hIgG Abs will also be presented on the beads through NAb PC (Fig. 5). Interestingly, total hIgG detected in the drug-spiked DMEM culture medium group was essentially the same as the hIgG4 drug detected from the same samples (Fig. 3C & F), demonstrating the accuracy of the LC–MS/MS method. 3.6. Pre-blocked magnetic beads have a lower total residual IgG During the assay development, a new version of preblocked streptavidin-coated magnetic beads from the same vendor became available which is “blocked using a proprietary method to help prevent nonspecific binding of proteins”. Since nonspecific carry-over of hIgG could reach 3000 ng/mL with the non-blocked version of the beads, indicating that other serum factors, such as growth factors, and cytokines, may also have the chance to be nonspecifically carried over and affect the cells in our NAb assay. To determine whether the preblocked version of the streptavidin-coated magnetic beads has reduced non-specific carry-over, we compared two versions of

99

the beads side by side. The pre-blocked beads indeed had significantly lower total hIgG carry-over with an average reduction of about 30% (Fig. 3D vs. 3D″ and 3E vs. 3E″). While the pre-blocked beads tended to have more drug carry-over when NAb was lower, at higher NAb concentrations, the drug carry-over actually was lower with the non-blocked beads. This essentially reduced the sample variations, and therefore, we used the pre-blocked beads for all the future studies. 3.7. Selection of different sources of biotin-conjugated drug reagent Since biotin-conjugated drug was one of the major factors that contributed to residual drug, we tested whether LC–MS/ MS could also be helpful in characterizing different sources of biotin-drug. We compared two lots of biotinylated-drugs, and measured residual drug and total hIgG level as well as IL-2 production by the bioassay. While the bioassay showed similar IL-2 production for both biotin-drugs (Fig. 5C), LC–MS/MS clearly showed less drug carry-over with Lot#1 of biotin-drug (Fig. 5A). We also characterized both biotin-drug lots using size exclusion chromatography (SEC) and High Resolution Mass Spectrometry (HRMS). Although both preparations had similar heavy and light chain biotin conjugation efficiency, Lot#2 contained more free biotin reagent, as well as high-molecular weight species, likely drug aggregates (Fig. 6. Peaks eluting prior to the drug peak total 13.4%, as compared to Lot#1 with 1.7%). These impurities essentially reduced the amount of effective biotin-drug thus reducing the ratio of biotin-drug to the free drug in the samples, which in turn will potentially lead to more residual drug. This slight difference may not be obvious with a few bioassays but may be statistically significant when multiple runs were analyzed for cut point and sensitivity during assay validation. 3.8. NAb recovery Although recovery of mouse NAb PC could be measured with alphaLISA as in Fig. 2D, LC–MS/MS has the capacity to test multiple signature peptides simultaneously without additional assay time and cost. To measure NAb PC recovery by LC–MS/MS, we spiked known amounts of mouse NAb PCs into either pooled normal human sera or BEAD buffer. Human sera samples were then extracted with BEAD and the NAb PCs were measured with LC–MS/MS in both groups after an appropriate dilution. The ratio of the NAb PC concentration in the BEAD-treated sample to that of the nontreated sample was calculated as the NAb PC recovery. NAb PCs spiked at 0.625 up to 20 μg/mL had a recovery of over 40% (Table 3). However, when NAb PC was spiked at 40 μg/mL, at which point the molar ratio of NAb PC and added biotin-drug was close to 1:1, the recovery dropped to 34.7% (Table 3). This reduced recovery suggested that more biotin-drug or a longer incubation time may be needed to achieve equilibrium for maximal recovery of NAb when NAb is abundant in the test samples. 4. Discussion Overcoming matrix and drug interference is a critical method development task for immunogenicity testing, so that a robust and rugged method can be validated to support drug

100

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

Fig. 3. LC–MS/MS to measure IgG4 drug and total human IgG in BEAD eluate. Pooled healthy human serum (A, B, D, E and A″, B″, D″ E″) or DMEM culture medium containing 10% FBS (C, F and C″, F″) were spiked with NAb and/or drug as indicated and subjected to BEAD treatment. Unit: μg/mL. Straight line indicates absence of the material and 100 μg/mL means that 100 μg/ml of drug was present in all samples within the group. (A–C): hIgG4 drug measurement with nonblocked beads; (A″–C″): hIgG4 drug measurement with pre-blocked beads; (D–E): hIgG measurement with non-blocked beads; (D″–E″): hIgG measurement with pre-blocked beads. Data are from one of three experiments with similar results.

development (DeSilva and Garofolo, 2014). Although the BEAD procedure has been successfully applied to remove interfering matrix factors or excess amount of drug in test samples prior to ADA or cell-based NAb assays (Lofgren et al., 2006), there has been no report of monitoring the residual drug and absolute NAb recovery in the final BEAD eluate, especially when the drug itself is a human mAb. The ability to verify residual drug level in samples for a NAb assay, which is essentially a functional drug quantitation assay, is the key to having confidence in sensitivity of the assay during sample analysis. We report here for the first time the detailed characterization of the BEAD procedure

applied as a pre-treatment step in a cell-based functional NAb assay. One of our original concerns was the efficiency of our BEAD procedure for the removal of drug contained in test samples, which could be as high as 125 μg/mL. If residual drug level is high due to high drug concentration and low efficiency of drug removal, cut point determined with traditional way of using either normal human serum or diseased in-study serum without treatment may not be relevant. The LC–MS/MS data clearly showed that drug was removed with high efficiency: only 0.01% of 100 μg/mL drug (~10 ng/mL) was still present

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

101

Fig. 4. Biotinylated-drug contributes to residual drug in BEAD eluate. (A, B) BEAD procedures were carried out with either normal human serum, 10%FBS or PBS; (A″, B″) “Mock” BEAD procedures were carried out without adding biotinylateddrug during BEAD treatment. Data are from one of three experiments with similar results.

after extraction. This essentially eliminated our concern about cut point inconsistency when samples with or without high amount of drug were used for the cut point determination. However, when samples also contained 20 μg/mL NAb PC, the remaining drug was higher at about 0.07%. This is expected, since more NAbs in test samples contribute to the formation of more free-drug/NAb/biotin-drug complex when both added biotin-drug and drug in the sample have a fixed concentration,

Fig. 5. Comparison of different preparations of biotinylated drugs. Lot #1 (mesh) vs. Lot# 2 (black and white squares) were used in the BEAD procedure to treat the same batch of samples, followed by the bioassay. (A, B) Residual drug and total IgG in the BEAD elute were measured by LC/MS–MS. (C) IL-2 production was measured with alphaLISA. Data are from one of three experiments with similar results.

which ultimately leads to more residual drug (Figs. 3 and 7). This higher rate of residual drug level, however, will not affect NAb detection, as LC–MS/MS data showed that NAb recovery at

Table 2 Comparison of residual drug level in BEAD eluate with or without 100 μg/mL drug from Fig. 3A & B. NAb, μg/mL

+0 μg/mL druga

+100 μg/mL drugb

Extra residual drug in the presence of 100 μg/mL drugc

Drug carry-over rated, %

0 4 5 6 20

112 101 117 118 238

123 150 164 190 286

11 49 47 72 48

0.011 0.049 0.047 0.072 0.048

a b c d

Residual drug concentration measured in BEAD elute from NAb +0 μg/mL drug samples. Residual drug concentration measured in BEAD elute from NAb +100 μg/mL drug samples. (b–a). c/1000. Unit: ng/mL.

102

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

Fig. 6. Physiochemical characterization of biotin-drug using SEC and HRMS. The reduced drug was analyzed by HRMS to assess the number of biotins distributed on both the light (A) and heavy (B) chains. Similar distributions for both lots (1 and 2) can be seen. The relatively small peak for unlabeled heavy chain (B) in both lots of biotindrug suggests that little drug has not been labeled in each lot. Further characterization to assess high molecular weight (HMW) species and free biotin reagent was carried out using SEC (C). Lot #2 has a greater percentage of HMW species and free biotin reagent by relative area percent as compared to Lot #1.

any point was much higher than that of residual drug (Tables 2 & 3). This data provides direct evidence that our BEAD pretreatment removed most of the drug and had good NAb recovery. Table 3 Recovery of NAb PC after BEAD treatment. NAb PC spiked, ng/mL

NAb PC measured after BEAD extraction, ng/mL

Recovery, %

0 625 1250 2500 5000 10,000 20,000 40,000

0 286.4 616.1 1184.1 2193.0 4344.0 8439.1 13,865.6

N/A 45.8 49.3 47.4 43.9 43.4 42.2 34.7

Known amount of mouse NAb PCs were spiked into either pooled normal human sera or BEAD buffer. Human serum samples were then extracted with the BEAD and the NAb PCs were measured with LC–MS/MS in both groups after appropriate dilution. The ratio of the NAb-PC concentration in the BEAD-treated sample to that of the nontreated sample was calculated as the NAb PC recovery. Samples were analyzed in triplicate and the mean value at each concentration point was used for calculation.

It was surprising to observe that biotin-drug alone contributed more than 100 ng/mL of residual drug in the final BEAD eluate. This is acceptable for the current cell-based NAb assay since the system drug in the assay is at 500 ng/mL. However, this residual drug would be a problem in assays where the system drug is close to 100 ng/mL, since the residual drug 100 ng/mL would effectively double the amount of drug in an actual sample. Originally we thought that the presence of residual drug might be due to incomplete biotinylation of the drug product, so that a significant portion of the drug had no biotin. Detailed characterization showed that this was not the case (Fig. 6). The other possibility then is that streptavidinbiotin-drug was washed off from the magnetic beads due to low pH during the second acid dissociation. Beads with better resistance to low-pH or beads covalently conjugated with drug, instead of using biotin-drug and streptavidin-coated beads, could be helpful. Alternatively, one NAb can bind to two biotindrugs, but only one of the biotin-drug binds to streptavidin on the beads, and the free biotin-drug could be dissociated and carried over to the final BEAD eluate. A specific surrogate peptide, which is contained in all S229P mutant hIgG4 drugs (Furlong et al., 2014), was used in

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

103

Fig. 7. Outline of the biotin-drug extraction and acid dissociation (BEAD) procedure. AD, acid dissociation. Depending on the relative ratio of the free drug and the biotindrug, multiple species could be formed: if free drug is predominant, then two free drugs complexed with one NAb will be the dominant species, which will be eventually washed away (red circle). If one free drug and one biotin-drug are complexed with NAb (green circle), that free drug will be carried over to the final elute upon 2nd acid dissociation. In addition, free drug and endogenous IgG may be carried over through nonspecific interaction with the beads. Or, due to the instability of the streptavidincoated beads at low pH, streptavidin-biotin-drug may be present in the final eluate as well upon 2nd AD.

the LC–MS/MS assay to measure the residual drug. Similarly, any modified human mAb drugs other than IgG4 subtype could also use the mutant peptide for LC–MS/MS analysis. In addition, unique peptides in the antigen binding region of any subtype of human mAb drug can be used, so that our LC–MS/MS method might be applied to all human mAb drugs. Characterizing the recovery of a NAb PC by LC–MS/MS is straightforward, as often the positive control in a NAb assay is from a non-human species or is a monoclonal antibody having a unique sequence. While characterizing the recovery of a positive control does not ensure recovery of all possible human anti-drug antibodies in real clinical samples, it determines the order of magnitude of the recovery, thus providing a guideline for further optimization. For example, a poor NAb PC recovery could be due to low affinity or instability of NAb PC at low pH, or it could be due to insufficient amount of biotin-drug or streptavidin-coated beads and sub-optimal incubation time. In conclusion, we have developed and optimized an efficient and consistent pre-treatment method for testing clinical samples containing high therapeutic concentrations. This example clearly shows the complexity of developing sensitive and selective Nab assays, which involves multiple analytical technologies, disciplines and scientific judgment. The critical information enabling this optimization is the accurate measurement of residual drug levels using a LC–MS/MS assay. In addition, the LC–MS/MS method developed in our laboratory simultaneously verified recovery of the positive control, as well as endogenous hIgG as an indicator of nonspecific serum factor carry-over. This information could provide critical guidance as whether the low sensitivity of cell-based NAb assays is due to high amount of residual drug, poor NAb recovery, or residual

serum factors. Our well-characterized pretreatment procedure has been successfully applied to the assay validation and sample analysis for multiple human mAb therapeutics. Acknowledgments The authors would like to thank Mr. Bruce Stouffer, Drs. Jennifer Cummings and Marina Ichetovkin for review and comments. References Angal, S., King, D.J., Bodmer, M.W., Turner, A., Lawson, A.D., Roberts, G., Pedley, B., Adair, J.R., 1993. A single amino acid substitution abolishes the heterogeneity of chimeric mouse/human (IgG4) antibody. Mol. Immunol. 30, 105. Beck, A., Wagner-Rousset, E., Bussat, M.C., Lokteff, M., Klinguer-Hamour, C., Haeuw, J.F., Goetsch, L., Wurch, T., Van Dorsselaer, A., Corvaia, N., 2008. Trends in glycosylation, glycoanalysis and glycoengineering of therapeutic antibodies and Fc-fusion proteins. Curr. Pharm. Biotechnol. 9, 482. Beck, A., Wurch, T., Bailly, C., Corvaia, N., 2010. Strategies and challenges for the next generation of therapeutic antibodies. Nat. Rev. Immunol. 10, 345. Casadevall, N., Nataf, J., Viron, B., Kolta, A., Kiladjian, J.-J., Martin-Dupont, P., Michaud, P., Papo, T., Ugo, V., Teyssandier, I., Varet, B., Mayeux, P., 2002. Pure red-cell aplasia and antierythropoietin antibodies in patients treated with recombinant erythropoietin. N. Engl. J. Med. 346, 469. Civoli, F., Kroenke, M.A., Reynhardt, K., Zhuang, Y., Kaliyaperumal, A., Gupta, S., 2012. Development and optimization of neutralizing antibody assays to monitor clinical immunogenicity. Bioanalysis 4, 2725. DeSilva, B., Garofolo, F., 2014. Matrix interference in ligand-binding assays: challenge or solution? Bioanalysis 6, 1029. Dodge, R., Daus, C., Yaskanin, D., 2009. Challenges in developing antidrug antibody screening assays. Bioanalysis 1, 699. Food and Drug Administration, 2013. DRAFT — Guidance for Industry Immunogenicity Assessment for Therapeutic Protein Products. U.S. Department of Health and Human Services.

104

W. Xu et al. / Journal of Immunological Methods 416 (2015) 94–104

Food and Drug Administration, n.a. F.A.D.P. http://www.accessdata.fda.gov/ scripts/cder/drugsatfda/index.cfm. Furlong, M.T., Titsch, C., Xu, W., Jiang, H., Jemal, M., Zeng, J., 2014. An exploratory universal LC–MS/MS assay for bioanalysis of hinge region-stabilized human IgG4 mAbs in clinical studies. Bioanalysis 6 (13), 1747–1758. Hu, J., Gupta, S., Swanson, S.J., Zhuang, Y., 2009. A bioactive drug quantitation based approach for the detection of anti-drug neutralizing antibodies in human serum. J. Immunol. Methods 345, 70. Jiang, H., Xu, W., Titsch, C.A., Furlong, M.T., Dodge, R., Voronin, K., Allentoff, A., Zeng, J., Aubry, A.F., DeSilva, B.S., Arnold, M.E., 2014. Innovative use of LC–MS/MS for simultaneous quantitation of neutralizing antibody,

residual drug, and human immunoglobulin G in immunogenicity assay development. Anal. Chem. 86, 2673. Klee, G.G., 2000. Human anti-mouse antibodies. Arch. Pathol. Lab. Med. 124, 921. Lofgren, J.A., Wala, I., Koren, E., Swanson, S.J., Jing, S., 2006. Detection of neutralizing anti-therapeutic protein antibodies in serum or plasma samples containing high levels of the therapeutic protein. J. Immunol. Methods 308, 101. McCutcheon, K.M., Quarmby, V., Song, A., 2010. Development and optimization of a cell-based neutralizing antibody assay using a sample pre-treatment step to eliminate serum interference. J. Immunol. Methods 358, 35.