MS Assay for the Simultaneous Quantitation of

Sep 11, 2013 - Drug Safety Evaluation, Bristol-Myers Squibb, New Brunswick, New Jersey, United .... (Woburn, MA), and AnaSpec (Fremont, CA), Table S1...
8 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/ac

Fully Validated LC-MS/MS Assay for the Simultaneous Quantitation of Coadministered Therapeutic Antibodies in Cynomolgus Monkey Serum Hao Jiang,*,† Jianing Zeng,*,† Craig Titsch,† Kimberly Voronin,‡ Billy Akinsanya,† Linlin Luo,† Hongwu Shen,§ Dharmesh D. Desai,† Alban Allentoff,‡ Anne-Françoise Aubry,† Binodh S. DeSilva,† and Mark E. Arnold† †

Analytical and Bioanalytical Development and ‡Discovery Chemistry Synthesis, Bristol-Myers Squibb, Princeton, New Jersey, United States § Drug Safety Evaluation, Bristol-Myers Squibb, New Brunswick, New Jersey, United States S Supporting Information *

ABSTRACT: An LC-MS/MS assay was developed and fully validated for the simultaneous quantitation of two coadministered human monoclonal antibodies (mAbs), mAb-A and mAb-B of IgG4 subclass, in monkey serum. The total serum proteins were digested with trypsin at 50 °C for 30 min after methanol denaturation and precipitation, dithiothreitol reduction, and iodoacetamide alkylation. The tryptic peptides were chromatographically separated with a C18 column (2.1 × 100 mm, 1.7 μm) with mobile phases of 0.1% formic acid in water and acetonitrile. Four peptides, a unique peptide for each mAb and two confirmatory peptides from different antibody domains, were simultaneously quantified by LC-MS/ MS in the multiple reaction-monitoring mode. Stable isotopically labeled peptides with flanking amino acids on C- and N-terminals were used as internal standards to minimize the variability during sample processing and detection. The LC-MS/MS assay showed lower limit of quantitation (LLOQ) at 5 μg/mL for mAb-A and 25 μg/mL for mAb-B. The intra- and interassay precision (%CV) was within 10.0% and 8.1%, respectively, and the accuracy (%Dev) was within ±5.4% for all the peptides. Other validation parameters, including sensitivity, selectivity, dilution linearity, processing recovery and matrix effect, autosampler carryover, run size, stability, and data reproducibility, were all evaluated. The confirmatory peptides played a critical role in confirming quantitation accuracy and the integrity of the drugs in the study samples. The robustness of the LC-MS/MS assay and the data agreement with the ligand binding data demonstrated that LC-MS/MS is a reliable and complementary approach for the quantitation of coadministered antibody drugs.

P

graphic run time often necessary to achieve good selectivity.14−19 Direct LC-MS/MS analysis after trypsin digestion of total plasma/serum protein pellets has been recently reported for absolute quantitation of mAb drugs.20−22 The total proteins in plasma/serum were denatured by water-miscible organic solvents (methanol, acetonitrile, etc.) followed by a centrifugation to form a pellet, which removed some endogenous substances (phospholipids, peptides, etc.) prior to trypsin digestion. In those papers, the reduction and alkylation steps were not applied during sample processing to save time and effort, considering that no cysteine residues were contained in the target peptides. However, in our experience from this study, the steps of reduction and alkylation are essential to good data accuracy and precision. In the present study, the total serum proteins were denatured, reduced, alkylated, and trypsindigested prior to direct analysis with LC-MS/MS for the quantitation of the peptides for two coadministrated human

harmaceutical companies have a growing interest to complement their small molecule portfolios with biotherapeutics/biologics, because biologics have higher specificities and are less prone to toxicological and drug−drug interaction liabilities than small molecule drugs. In particular, the use of monoclonal antibody (mAb) drugs has achieved considerable success in oncology in recent years.1,2 Bioanalytical methods that can be developed quickly to measure concentrations of therapeutic mAbs in biological fluids are needed to support drug development. Liquid chromatography−triple quadrupole mass spectrometry (LC-MS/MS), a selective and sensitive bioanalytical tool for quantifying small molecule drugs, has been applied to quantitation of therapeutic peptides and proteins in complex biological matrixes (plasma, serum, etc.).3−19 Its advantages over ligand binding assays (LBAs) are capability to distinguish drugs from their degradation products,4 ability to detect and quantify peptides in specific regions of the proteins for specificity, feasibility of multiplex analysis, and wider dynamic range. However, extensive sample cleanup and analyte enrichment is usually needed to achieve the required sensitivity, and the throughput may be limited by the long chromato© 2013 American Chemical Society

Received: July 31, 2013 Accepted: September 11, 2013 Published: September 11, 2013 9859

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

Each protein pellet was obtained by vigorously vortexing for 2 min and centrifuging at 200g (rcf) for 2 min. The supernatant was manually dumped into a waste container, and the residual supernatant was removed by upside-down touching the opening of microtubes with clean and dry paper towels for 2−3 s. The retained protein pellet was resuspended with 50 μL of digestion buffer (100 mM ammonium bicarbonate) by vigorously vortexing for 2 min until a uniform white protein suspension was obtained. The denatured protein solution was further reduced with 10 μL of 100 mM DTT at 60 °C for 60 min and alkylated with 25 μL of 100 mM IAA at 30 °C for 30 min, in a preheated thermomixer at 1000 rpm. Then 25 μL of the SIL-f-IS working solution (containing 2 μg/mL of SIL-fGLEW and SIL-f-VVSV, and 10 μg/mL of SIL-f-TVAA and SIL-f-ASGI in 10 mM PBS) was added to each sample, except for blank samples to which 25 μL of PBS was added. The digestion was initiated by adding 25 μL of 8 mg/mL Sigma trypsin (prepared in 0.1% formic acid/water solution) and incubating at 50 °C in a preheated thermomixer (1000 rpm) for 30 min. The digestion reaction was then quenched by adding 25 μL of 10% formic acid/water solution. The final tryptic digest was centrifuged at 5000g (rcf) for 5 min, and the supernatant (∼120 μL) was transferred using a JANUS Mini liquid handler (Perkin-Elmer, Downers Grove, IL) to a clean polypropylene 96-well collection plate for LC-MS/MS analysis. Brief centrifugation and vortexing after each addition of reagents were performed to ensure well-to-well consistency during sample processing. LC-MS/MS Analysis. The total serum protein digest (2 μL) was injected into a high performance liquid chromatography system (model LC-30AD, Shimadzu Scientific Instruments, Inc., Columbia, MD) fitted with an Acquity UPLC BEH C18 column (2.1 × 100 mm, 1.7 μm, Waters Co., Milford, MA). The mobile phases of 0.1% formic acid (A) and 0.1% acetonitrile (B) were delivered under a gradient program, 5% B to 35%B for 9.5 min. The flow-rate was 0.6 mL/min. The eluent was introduced to an AB Sciex 5500 triple-quadrupole mass spectrometer (Foster City, CA) during the period of 5− 9.5 min, controlled by the instrument built-in switching valve. Then the eluent was ionized 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, 600 °C) and analyzed by the triple quadrupole analyzers in multiple reaction monitoring (MRM) mode. The mass-dependent ion transitions and quadrupole parameters are listed in Table S1 (Supporting Information). Data acquisition was divided into two periods, 0−7.2 min for GLEW/SIL-f-GLEW and ASGI/SIL-f-ASGI, and 7.2−12 min for TVAA/SIL-f-TVAA and VVSV/SIL-f-VVSV. The acquired chromatographic peaks were integrated by the Analyst software (version 1.5.1, AB SCIEX), followed by exporting to Watson LIMS (version 7.3, Thermo Fisher Scientific Inc.) for calibration curve regression and backcalculation of the concentrations of QCs and study samples. Statistical data were calculated with Watson LIMS and plotted with Microsoft Excel (version 2007, Microsoft Co.). Assay Validation and Sample Analysis. The LC-MS/MS assay validation and sample analysis were conducted according to current regulatory guidelines and BMS internal SOPs for small molecules. The acceptance criteria of ±15% for calibration standards and QCs (±20% at the LLOQ) were applied. Toxicokinetic Study in Monkeys. According to the protocol approved by the Institutional Animal Care and Use Committee of BMS, vehicle control (group 1), two doses of

mAbs in cynomolgus monkey serum. After all the tryptic peptides were screened, one unique peptide for each drug and two confirmatory peptides from different antibody domains were chosen for LC-MS/MS quantitation. This multiplex LCMS/MS assay has been developed and fully validated for the simultaneous determination of two coadministered human mAb drugs in monkey serum. The assay validation and the sample analysis were conducted according to regulatory guidelines23,24 and Bristol-Myers Squibb (BMS) internal standard operating procedures (SOPs) for LC-MS assays, as, at present, there is no guidance on bioanalysis of large molecules by LC-MS/MS. Samples from a 4-week cynomolgus monkey toxicology study were tested by the LC-MS/MS and two LBAs. Good agreement between the results demonstrated the feasibility and reliability of LC-MS/MS for the absolute quantitation of codosed antibody drugs in serum for drug development.

■

EXPERIMENTAL SECTION Reagents and Chemicals. Human mAb-A and mAb-B are both IgG4 (molecular weight ∼150 kDa) obtained from BMS Research and Development. Four pairs of peptides were chemically synthesized by BMS Research and Development (Princeton, NJ), GenScript (Piscataway, NJ), NeoBioLab (Woburn, MA), and AnaSpec (Fremont, CA), Table S1 (Supporting Information). The unlabeled synthetic peptides were used for optimizing mass spectrometer parameters and for evaluating recovery and matrix effect but not for quantitation. Each peptide was named with N-terminal first four amino acid codes in the sequence, i.e., GLEW (mAb-A peptide), ASGI (mAb-B peptide), VVSV (human IgG4 peptide 1), and TVAA (human IgG4 peptide 2). The internal standards (IS) were the synthetic peptides containing stable isotopically labeled (SIL) amino acids ([13C6,15N]leucine or [13C5,15N]valine) and flanking amino acids on the N-terminal and/or C-terminal and are designated as SIL-f-GLEW, SIL-f-ASGI, SIL-f-VVSV, and SIL-f-TVAA. The stock solutions of these synthetic peptides were all prepared in 20% acetonitrile/water solution. Methanol (HPLC grade), acetonitrile (HPLC grade), 2-propanol (HPLC grade), ammonium bicarbonate, dithiothreitol (DTT), iodoacetamide (IAA), phosphate-buffered saline (PBS) tablet, and trypsin from bovine pancreas (catalogue No. T1426, TCPK treated) were all purchased from Sigma-Aldrich (St. Louis, MO). Sequencing grade trypsin (catalogue No. V5111) was purchased from Promega (Madison, WI). Formic acid was obtained from EMD (Gibbstown, NJ). Control monkey serum was purchased from Bioreclamation Inc. (Westbury, NY). Preparation of Calibration Standards and Quality Control Samples (QCs). Calibration standards (5, 10, 25, 50, 100, 200, 400, and 500 μg/mL) and QCs (5, 15, 25, 60, 125, 250, 400, 5000 μg/mL) were prepared from the stock solutions of mAb-A (10.0 mg/mL) and mAb-B (19.8 mg/mL) by serial dilutions with two different lots of control monkey serum. The identical nominal concentrations were assigned to the unique peptide GLEW representing the concentration of mAb-A, the unique peptide ASGI representing the concentration of mAb-B, and the confirmatory peptides VVSV or TVAA representing the concentration of mAb-A and mAb-B (VVSV and TVAA were contained in both mAb-A and mAb-B). Sample Processing Procedures. Aliquots of 25 μL of the serum samples (blanks, standards, QCs, or study samples) were pipetted into the wells of a 96-well polypropylene microtube plate (containing twelve eight-tube strips), followed by a precipitation of total serum proteins with 100 μL of methanol. 9860

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

mAb-A alone (groups 2 and 3), one dose of mAb-B alone (group 4), and one combination dose of mAb-A and mAb-B (group 5) were administered via intravenous infusion once weekly for 4 weeks. Blood samples were collected prior to dosing on days 1, 8, 15, and 22, and on days 1 and 22 at approximately 0.5, 24, 48, 72, 168 h postdose. Each processed serum sample was split into three portions, stored at −70 °C, and shipped to three laboratories for the LBA of mAb-A, the LBA of mAb-B, and the LC-MS/MS combo assay, respectively. The toxicokinetic parameters were calculated using Phoenix WinNonlin (version 6.2, Pharsight Co.), and correlation analyses were performed by GraphPad Prism (version 5.0, GraphPad Software Inc.).

collision energy, and collision cell exit potential) were individually optimized using step values for each parameter. Monkey serum samples with and without the spiked drugs (final concentration of 1 mg/mL as positive control for either mAb-A or mAb-B) were processed and then analyzed using the optimized LC-MS/MS method. The chromatograms were then compared to determine the selectivity in the blank monkey serum sample and the relative sensitivity of every tryptic peptide. On the basis of the screening results, the usable peptide candidates were further narrowed down to those peptides that showed prominent MRM responses (sensitivity) and lack of interference peaks (selectivity). There were no unique peptides on the light chain Fab domain and heavy chain Fc domain that showed acceptable selectivity; therefore, the common peptides TVAA and VVSV were selected as the confirmatory peptides representing the light chain Fab and heavy chain Fc domains, respectively. One of the unique peptides from the heavy chain Fab was selected for each drug (GLEW for mAb-A and ASGI for mAb-B), which were CDR peptides with a high signal-to-noise ratio, Table S2 (Supporting Information). Denaturation and Precipitation. In this method, methanol denatured and precipitated the serum proteins and separated total proteins from endogenous substances such as lipids and soluble peptides, which remained in the supernatant. The strength of vortexing (gentle vs vigorous) and the ratio of methanol to serum (4:1 vs 6:1, v/v) did not significantly alter the MRM responses of the peptides (only GLEW and VVSV were evaluated as the representatives of Fab peptides and Fc peptides in the drugs, data not shown). As shown in Figure S1 (Supporting Information), heat denaturation (95 °C) significantly increased the response of GLEW but decreased the response of VVSV. After an investigation, it was found that the increased response of GLEW was caused by ionization enhancement by matrix components, and the decreased response of VVSV was caused by the ionization suppression by matrix components (see the discussion in Processing Recovery and Matrix Effect). Different types of modifiers (6 M guanidine chloride, RapiGest SF, or 1% Triton X-100/1% acetonitrile) in the digestion buffer (100 mM ammonium bicarbonate) did not improve the MRM responses of GLEW and VVSV, Figure S2 (Supporting Information). Thus, sample denaturation with vigorous vortexing and methanol-to-serum ratio of 4:1 without 95 °C heat denaturation and modifiers were implemented in the final sample processing. Reduction and Alkylation. Reduction (with DTT) and alkylation (with IAA) did not affect the MRM response of GLEW but doubled the response of VVSV, Figure S2 (Supporting Information). This phenomenon might be related to the different locations of the peptides in the protein structure, even though neither peptide contains cysteine residues and the reduction and alkylation may not be needed for trypsin digestion. GLEW is located in the CDR and the heavy-chain variable region (VH), which is on the surface of the three-dimensional structure and easily accessible to trypsin for digestion. VVSV is located in the second heavy-chain constant region (CH2) which is very close to the intrachain disulfide bond. In theory, methanol denaturation causes interior hydrophobic side chains to twist to the surface where they interact favorably with high concentrations of methanol. Thus, the three-dimensional structure of the protein is completely altered,25,26 and some regional peptides (such as VVSV) might be partially buried inside of the structure. In this situation, the

■

RESULTS Method Development. A practical approach was applied to optimize every critical step during method development, which was essential for the success of developing a robust LC-MS/MS assay for drug development. This experimental protocol has been successfully applied for absolute quantitation of other antibodies, and thus subsequent efforts on method development were significantly reduced. In Silico Digestion and the Basic Local Alignment Search Tool (BLAST) Search. In silico trypsin digestion (PeptideMass, http://web.expasy.org/peptide_mass/) was used to predict complementarity determining region (CDR) peptides and confirmatory peptides, based on the known amino acid sequences and functional structures of the mAb drugs. Then the online protein BLAST (http://www.ncbi.nlm.nih.gov/ BLAST/Blast.cgi?PAGE=Proteins) was used to predict the selectivity of each tryptic peptide in monkey serum by querying the protein databases (taxonomy ID: 9541; database: nonredundant protein sequences). The querying results without significant hits of the identical sequences suggest potential lack of interfering peptides from monkey serum. The predicted tryptic peptides with 30 amino acids or more were not considered in this study due to the difficulties in LC-MS/MS detection and chemical synthesis of the SIL-f-IS. In the results, ∼60% of the tryptic peptides were absent from monkey protein sequences; however, ∼70% of which within mAb-A and mAb-B were common to both drugs. As a consequence, there were only two unique peptides from Fab domain and none from Fc domain Table S2 (Supporting Information). Tryptic Peptide Screening. The neat drug solution of mAbA or mAb-B (1 mg/mL in PBS) were digested with the sequencing grade trypsin and analyzed with LC-MS/MS using a 30-min HPLC gradient program. Precursor and product ions were monitored in the information-dependent acquisition (IDA) mode. The tryptic peptides with significant MS responses were identified, and the MRM ion transitions were selected based on the m/z values of the precursor ion and the product ion. The results showed that ∼65% in silico tryptic peptides were highly abundant and identifiable Table S2 (Supporting Information). LC-MS/MS parameters in MRM mode were then optimized by injecting 2-μL of the tryptic digests for each testing below. Waters ACQUITY UPLC C18 columns of different particle technology (BEH vs HSS) and different lengths (2.1 × 50 mm vs 2.1 × 100 mm) were evaluated. The BEH 100-mm column gave the best peak resolutions and the lowest MRM baseline. The MRM parameters (such as gas flow rates, ionization spray voltage, temperature, ion transitions, declustering potential, 9861

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

Figure 1. Representative chromatograms of the unique peptides at the LLOQ levels. Significant analyte peaks were observed in the LLOQ samples: 5 μg/mL for GLEW (a1), and 25 μg/mL for ASGI (b1). No interference peaks were observed in the corresponding blank samples, a0 and b0. For the internal standards, no interference peaks were observed in the blank samples (a0′ and b0′) in comparison with the samples spiked with the SIL-fISs (a1′ and b1′).

reduction/alkylation will help unfold all peptide chains and makes all the peptides accessible to trypsin digestion, which not

only increases digestion efficiency but also ensures digestion reproducibility. Therefore, reduction and alkylation are 9862

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

Figure 2. Procedure for evaluation of processing recovery and matrix effect. For the processing recovery samples, the neat proteins or the peptides were spiked into the samples before or after each process. For the matrix effect samples, the neat peptides were spiked into 20% acetonitrile/water with 0.1% formic acid to obtain sample e.

Method Validation and Sample Analysis. Calibration Curves and Regression Models. Calibration curve ranges were 5−500 μg/mL for GLEW and VVSV, and 25−500 μg/mL for ASGI and TVAA, representing the concentrations of the mAb drugs in serum samples. The latter two peptides had a higher LLOQ due to their lower MS responses, but the sensitivity adequately covered the minimal drug concentration in the study samples. Two regression models (linear vs quadratic) were evaluated by analyzing the calibration standard data from all available validation runs with the BMS internal regression analysis software. Plots of weighted residuals were used to determine the appropriate weighting (x, 1/x, or 1/x2, data not shown). Based on a scientific assessment of the results from statistical analysis, a linear (1/x2 weighting) regression model was selected. Accuracy and Precision. The accuracy, expressed as the percent deviation (%Dev) from the nominal concentration, for the calibration standards and QCs in the assay validation and sample analysis were calculated. All five validation runs and five sample analysis runs passed the acceptance criteria. The accuracy (%Dev) and precision (%CV) information, based on the maximum value from all concentration levels of the QCs, was obtained using one-way ANOVA in Watson, Tables S3−S6 (Supporting Information). The intra- and interassay precisions (%CV) and the accuracy (%Dev) were within 3.5%, 3.5%, and ±4.1% for GLEW, 9.0%, 3.4%, and ±4.9% for ASGI, 10.0%, 8.1%, and ±5.4% for VVSV, and 3.0%, 3.2%, and 3.5% for TVAA, respectively.

essential to an efficient trypsin digestion for good assay accuracy and precision. Internal Standardization. SIL-intact mAb is the best choice for an internal standard in absolute protein quantitation, but it is not always easy to obtain. Using chemically synthesized SIL peptides is an alternative and straightforward approach. In this study, we used [13C6,15N]leucine- or [13C5,15N]valine-labeled peptides with two to five flanking amino acids on the Nterminal and/or C-terminal end to track the sample-to-sample variability during trypsin digestion and LC-MS/MS analysis. The SIL provided a sufficient mass shift to distinguish each pair of the labeled and the unlabeled peptides. The number (two to five amino acids) of flanking amino acids (Table S1, Supporting Information) did not make a difference on assay performance for any of the SIL-f-IS, suggesting that two flanking amino acids sufficed for tracking the variability during trypsin digestion. Trypsin Digestion. The amount of trypsin (200 vs 400 μg in 25 μL of the serum sample) and the incubation time (30 vs 60 min) did not make significant changes to the MRM responses of both GLEW and VVSV (data not shown). The temperatures of 37 °C, 50 °C, and 60 °C provided similar responses for GLEW (CDR, Fab); however, the temperature of 50 °C doubled the response for VVSV (Fc) compared to 37 °C and 60 °C. A higher temperature of 60 °C compared to 50 °C unexpectedly decreased the VVSV response, which may be related to the instability of VVSV.27 The digestion condition of 200 μg of trypsin for 30 min at 50 °C was used in this study, Figure S1 (Supporting Information). 9863

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

for trypsin digestion were ∼100% except for VVSV (∼80%), suggesting low digestion efficiency for VVSV. The recoveries for postdigestion treatment were ∼80% for all the peptides, due to a ∼20% sample loss during transfer of the supernatant. The matrix effect on MRM detection was determined at the low and high QC concentrations for each peptide. The response of the peptide in the spiked control serum digest was compared with that in the spiked solution of 20% acetonitrile/water with 0.1% formic acid. The mean of matrix factors (MF) at the concentrations of 60 and 400 μg/mL in six lots of monkey serum were 1.9 and 1.4 for GLEW, 3.7 and 3.0 for ASGI, 0.3 and 0.4 for VVSV, and 1.4 and 1.2 for TVAA. The results indicated significant MS response enhancement or suppression by coeluting interfering components, which was peptidedependent. However, the observed variability and difference in the recovery and the matrix effect did not affect data accuracy and precision, because they were well compensated by the SILf-ISs during sample processing and LC-MS/MS analysis. Autosampler Carryover and Run Size. The carryover of each target peptide was tested by analyzing, in six replicates, a blank serum digest injected directly after the high QC (400 μg/mL). The carryover was calculated as the percent response in the blank sample compared to the response in the high QC sample. The carryover of each SIL-f-IS was determined similarly. No carryover effect was observed for any of the peptides and their SIL-f-ISs, except for VVSV (0.11%). However, the impact to the LLOQ sample was less than 20% and acceptable. It has been demonstrated that the LC-MS/MS method can be performed acceptably for a run size up to 192 samples, including calibration standards, QCs, and study samples. In Vitro Sample Stability in Serum. The following in vitro stability for the target peptides in monkey serum was demonstrated by evaluating the deviations of the mean predicted QC concentrations from their nominal concentrations: room temperature for 49 h, five freeze−thaw cycles, and −70 °C frozen storage for 33 days. The reinjection integrity for 178 h at 4 °C autosampler was demonstrated by reinjecting all standards and QCs from an analyzed batch sample. The processed sample stability for 102 h at 4 °C was confirmed by testing the processed and the stored QCs against freshly prepared standards. Sample Analysis and Incurred Sample Reanalysis (ISR). Serum samples from the monkey study were processed and analyzed for the four peptides simultaneously, and more than 10% of the samples (49 ISR samples) were reanalyzed to evaluate the data reproducibility. The results met the acceptance criteria (more than 2/3 of reanalysis results were within 10% deviation from the mean of the original and reanalysis results) for all the peptides, Figure S5 (Supporting Information). The ISR samples from groups 2 and 3 (dosed with only mAb-A) generated concentration values for GLEW, TVAA, and VVSV; the samples from group 4 (dosed with only mAb-B) generated concentration values for ASGI, TVAA, and VVSV; the samples from group 5 (dosed with both mAb-A and mAb-B) generated concentration values for GLEW, ASGI, TVAA, and VVSV. The good correlation between the original and the repeat values demonstrated the data reproducibility, Figure S6 (Supporting Information). Confirmatory Peptides. The peptides VVSV and TVAA were selected as the confirmatory peptides to ensure the data accuracy of the unique peptides GLEW and ASGI and the integrity of the drugs in vivo and in vitro. Both mAb drugs have similar antibody Y-shape structure and homology on constant regions of both heavy chains and light chains. VVSV is located

Lower Limit of Quantitation (LLOQ). The LLOQ for each analyte was assessed using serum samples at 5 μg/mL for GLEW and VVSV, and at 25 μg/mL for ASGI and TVAA. Ten different lots of control serum were spiked to obtain ten LLOQ samples. All ten LLOQ samples were within ±7.6%, ± 17.3% (nine out of ten lots serum), ±12.6%, and ±8.1% of the corresponding nominal values of GLEW, ASGI, VVSV, and TVAA, respectively, suggesting that measurements at the LLOQ levels were accurate. Selectivity. For the processed control monkey serum samples, the absence of interfering components in the chromatograms for each target peptide and its SIL-f-IS demonstrated assay selectivity in the matrix, Figure 1 and Figure S3 (Supporting Information). Interference from the coadministered drug (mAb-B) with the detection of GLEW was evaluated by spiking 2000 μg/mL of mAb-B to the QCs at low and high concentrations; interference of the coadministered drug (mAb-A) with the detection of ASGI was evaluated by spiking 4000 μg/mL mAb-A to the QCs at low and high concentrations. The %Dev was within ±0.2% for GLEW and ±4.9% for ASGI, suggesting no interference from the coadministered drugs. To evaluate the interference of antidrug antibodies (ADA) with the detection of GLEW and VVSV, 13 mouse monoclonal anti-mAb-A antibodies were individually spiked into the LLOQ samples to reach a final concentration of 10 μg/mL. The results demonstrated no ADA interference with the LLOQ measurements, Table S7 (Supporting Information). The peptides ASGI and TVAA were not measured in these samples, because the mAb-A concentration of 5 μg/mL was below the LLOQ for ASGI and TVAA (25 μg/mL). The ADA for mAbB was not evaluated, because the ADAs for mAb-B were not available in the lab when the experiment was conducted. In the step of sample pretreatment, the protein−protein interaction between mAbs and ADAs must be disrupted by 4 volumes of methanol, which was confirmed by using mAb-A as a representative antigen. Dilution Linearity. Dilution linearity was evaluated to support expected study sample concentrations above the upper limit of quantitation (ULOQ). Dilution QC (5000 μg/ mL) was diluted with monkey serum at different dilution factors. Triplicate aliquots were analyzed. Based on the %Dev of the mean concentration against the nominal concentration and the %CV, a good dilution linearity was demonstrated within the curve range of 5−500 μg/mL for GLEW and VVSV, and 25−500 μg/mL for ASGI and TVAA, Figure S4 (Supporting Information). Low recoveries at 5000 μg/mL for GLEW (74.7%) and VVSV (20.1%) but not ASGI (107.9%) and TVAA (102.6%) suggested either insufficient trypsin digestion or MS saturation. Based on the MRM chromatograms, VVSV reached the MRM detection upper limit (∼107 counts per minute) at 1000 μg/mL and showed split chromatographic peaks at 5000 μg/mL, indicating that MS detection saturation caused the low recoveries. Processing Recovery and Matrix Effect. The processing recovery of each target peptide from monkey serum at different processing steps (denaturation, trypsin digestion, and postdigestion treatment) was determined at the low and high concentrations (60 and 400 μg/mL), Figure 2. The response ratio (analyte to SIL-f-IS) in monkey serum samples spiked with the analyte mixture (either the mixture of mAb-A and mAb-B, or the mixture of GLEW, ASGI, VVSV, and TVAA) prior to each processing step was compared with that spiked after the processing step. The recoveries for denaturation and precipitation were ∼100% for all the peptides. The recoveries 9864

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

Figure 3. Ratios of confirmatory peptides to unique peptides. GLEW and ASGI were the unique peptides for mAb-A and mAb-B, respectively. VVSV and TVAA were the confirmatory peptides, the common peptides for both drugs.

Figure 4. Agreement of LC-MS/MS and LBA data. Mean concentration data (group 5, coadministered with the both drugs) from LC-MS/MS and LBA were comparable.

The finding of the high ratios of VVSV/ASGI (or TVAA/ASGI) in group 4 samples underscores the utility of the confirmatory peptides in the LC-MS/MS analysis. It also highlighted the power of the LC-MS/MS technique for simultaneously monitoring multiple peptides from various regions of the proteins by opening up additional ion transition channels. Comparison of LC-MS/MS vs Ligand-Binding Assay. Crossvalidation between the LC-MS/MS assay and the LBA for mAb-A were conducted by analyzing a set of QCs with both assays and comparing the data. The QC results (≤7.8% difference at different concentration levels) demonstrated that the LC-MS/ MS data were comparable to the LBA data. For study sample analysis, three aliquots of each study samples were analyzed by three laboratories in parallel. The concentration data of the study samples from group 5 (the coadministration group) obtained by LBA and LC-MS/MS were compared, Figure 4a,b, and found to agree well. Parson correlation of toxicokinetic data indicated that the LC-MS/MS data were highly correlated with the LBA data, P < 0.0001, Table S8 (Supporting Information). This good agreement between LC-MS/MS data and LBA data may be due to the fact that there was no soluble ligand or significant ADA levels in the samples. ADAs in 24/40 (60%) monkey serum samples were observed, but the amounts of ADAs were not significant (signal-to-noise ratio ∼2). On the basis of the LBA results, no pharmacokinetic profile change was observed when comparing monkeys with and without ADAs.

on the heavy chain constant region 2 (CH2) of both mAb drugs (IgG4). TVAA is located on the light chain constant region (CL) of both mAb drugs (IgG4). Equal amounts of mAb-A and mAb-B were spiked in the calibration standards, and the nominal concentration was defined for a single molecule. However, VVSV and TVAA were derived from both molecules. In this circumstance, when only mAb-A drug was present in the study samples, the ratio of confirmatory peptide to unique peptide equals (VVSV*2)/GLEW or (TVAA*2)/GLEW. When both mAb drugs were present in the study samples, the ratio of confirmatory peptide to unique peptide equals (VVSV*2)/(GLEW + ASGI) or (TVAA*2)/(GLEW + ASGI) as shown in Figure 3. The concentration ratio of VVSV (or TVAA) to GLEW was plotted for the samples from groups 2 and 3 monkeys (dosed with mAb-A), the concentration ratio of VVSV (or TVAA) to ASGI was plotted for the samples from group 4 monkeys (dosed with mAb-B), and the concentration ratio of VVSV (or TVAA) to the sum of GLEW and ASGI was plotted for the samples from group 5 monkeys (codosed with mAb-A and mAb-B). Most of the ratios were distributed in the range of 0.85−1.15 (i.e., 85−115%) for both confirmatory peptides. The wider distribution of the ratios for VVSV was attributed to the relatively high variability of VVSV in the assay. The ratios of VVSV/ASGI and TVAA/ASGI in group 4 samples were both higher than other ratios in the plots, indicating that the concentrations of ASGI were ∼10% lower than those of the confirmatory peptides VVSV or TVAA. This might be related to the in vivo modification (oxidation) of the methionine residue 28−30 in ASGI, because methionine oxidation as a result of sample processing ex vivo was not observed (data not shown). In addition, the tight QC data of ASGI (−6.4% to 11.1% deviation from nominal concentrations) in the run ruled out the possibility of data bias.

■

DISCUSSION It is critical to choose selective and sensitive peptides as the surrogates for the quantitative determination of mAb concentrations in biological matrixes. In this assay, a unique peptide was selected for each drug to distinguish them from 9865

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

precipitation) to remove protein−protein interactions between the mAb drugs and their ADAs or targets. The linear curve range was wide by the nature of the dynamic range of MS detection. The LLOQ was improved by changing the sample aliquot volumes, LC injection volume, and MS parameters. Importantly, ease of multiplexing was a significant advantage of LC-MS/MS. Capability of monitoring multiple MRM ion transitions during data acquisition allowed the detection of multiple target peptides for the simultaneous quantitation of two codosed mAbs.

each other and endogenous peptides. Due to the high sequence homology between these two IgG4 drugs, the peptides in the CDR were the only ones that were selective enough. The screening results (Table S2, Supporting Information) for sensitivity and selectivity indicated that GLEW was the only unique peptide for mAb-A, and ASGI for mAb-B. All other CDR peptides were either not MS sensitive or not selective. In contrast, it seems that there were more usable peptide candidates for confirmatory peptides, but only a few peptides candidates showed good MS sensitivity and selectivity in the screening experiment. In addition, only those peptides located at different antibody domains from the unique peptide could be considered as confirmatory peptides. Eventually only a limited number of peptides could be used as confirmatory peptides. Among the peptides examined for sensitivity, VVSV (Fc) and TVAA (light chain, Fab) were the best ones because they are present in the drugs but absent from monkey IgGs. This assay could be adapted for use in other animal species. The sample processing method is not matrix- and species-dependent. The peptides GLEW and ASGI are drug specific and can be used for different matrixes and species, including human. The peptides TVAA and VVSV are human IgG specific and only applicable to nonhuman matrixes. In this study, the conditions of predigestion treatments (protein denaturation, reduction, and alkylation) and trypsin digestion were optimized. The steps of reduction and alkylation were essential to good data accuracy and precision, especially for the target peptides, which are located in the Fc region. The results of VVSV from the samples without reduction and alkylation showed very poor data accuracy and precision, and high intrarun variability. As for the trypsin digestion, the enzymatic reaction was completed within 30 min due to the high trypsin-to-protein ratio (1:10, w/w) and the high temperature (50 °C). The doubled amount of trypsin did not increase responses of the peptides. Unless very selective cleanup of the protein is performed prior to the digestion, the complexity of digested samples causes matrix effects and interference from coeluting tryptic peptides and endogenous substances and thus make low-level quantitation difficult. In this study, selective sample cleanup (such as immunocapture) was not applied because the required LLOQ (∼50 μg/mL in monkey serum) was easily achieved. The matrix effects on LC-MS/MS detection were observed for the target peptides, either signal enhancement or suppression. SIL internal standards should correct matrix effects, because they have the same LC-MS/MS performance as the corresponding peptides. It must be noted that the stability data from the validation only represented the stability of the quantified peptide fragments from the mAb drugs during bioanalytical processing and storage (so-called in vitro stability). The in vivo and in vitro changes on some of the functional sites and structure integrity of the mAbs could not be determined in this assay, unless they occurred on the target peptides as previously discussed for the oxidation of ASGI. This assay measured total drugs in serum samples because total serum proteins were digested and the representative tryptic peptides were analyzed. In this study, the advantage of the LC-MS/MS assay over the LBA was its powers in selectivity, flexibility, and multiplexing. The LC-MS/MS assay was able to distinguish the target peptides from other interfering peptides by the high resolutions of chromatographic separation and mass analyzers. It also allowed harsh conditions for sample pretreatments (organic solvents for protein

■

CONCLUSIONS This manuscript reports a fully validated LC-MS/MS method for the simultaneous quantitation of two codosed human mAbs drugs in cynomolgus serum samples. One unique peptide for each drug was quantified to represent the intact protein concentration. Two confirmatory peptides from different mAb domains were also monitored to confirm the quantitation accuracy and the integrity of the drugs in the study samples. The method development strategy assured the good performance of the method as demonstrated by the validation. The method development strategy, sample processing protocol, and the LC-MS/MS conditions are general and may be applicable for quantitation of other monoclonal antibodies in biological matrixes.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. Fax: 609-252-3845. *E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank Dr. Jonathan R. Haulenbeek and Dr. Qin C. Ji for the discussion, review, and comments.

■

REFERENCES

(1) Scott, A. M.; Wolchok, J. D.; Old, L. J. Nat. Rev. Cancer 2012, 12, 278−287. (2) Elvin, J. G.; Couston, R. G.; van der Walle, C. F. Int. J. Pharm. 2013, 440, 83−98. (3) van den Broek, I.; Niessen, W. M. A.; van Dongen, W. D. J. Chromatogr., B. 2013, 929, 161−79. (4) Tamvakopoulos, C. Mass Spectrom. Rev. 2007, 26, 389−402. (5) van den Broek, I.; Sparidans, R. W.; Schellens, J. H. M.; Beijnen, J. H. J. Chromatogr., B 2008, 872, 1−22. (6) Liu, H.; Manuilov, A. V.; Chumsae, C.; Babineau, M. L.; Tarcsa, E. Anal. Biochem. 2011, 414, 147−153. (7) Hagman, C.; Ricke, D.; Ewert, S.; Bek, S.; Falchetto, R.; Bitsch, F. Anal. Chem. 2008, 80, 1290−1296. (8) Ewles, M.; Goodwin, L. Bioanalysis 2011, 3, 1379−1397. (9) Kirsch, S.; Widart, J.; Louette, J.; Focant, J. F.; De Pauw, E. J. Chromatogr., A. 2007, 1153, 300−306. (10) Becher, F.; Pruvost, A.; Clement, G.; Tabet, J. C.; Ezan, E. Anal. Chem. 2006, 8, 2306−2313. (11) Buscher, B. A. P.; Gerritsen, H.; van Schöll, I.; Cnubben, N. H. P.; Brüll, L. P. J. Chromatogr., B 2007, 852, 631−634. (12) van Platerink, C. J.; Janssen, H. G.; Horsten, R.; Haverkamp, J. J. Chromatogr., B 2006, 830, 151−157. 9866

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867

Analytical Chemistry

Article

(13) Wu, S. T.; Ouyang, Z.; Olah, T. V.; Jemal, M. Rapid Commun. Mass Spectrom. 2011, 25, 281−290. (14) Berna, M. J.; Zhen, Y.; Watson, D. E.; Hale, J. E.; Ackermann, B. L. Anal. Chem. 2007, 79, 4199−4205. (15) Wolf, R.; Hoffmann, T.; Rosche, F.; Demuth, H. U. J. Chromatogr., B 2004, 803, 91−99. (16) Li, H.; Ortiz, R.; Tran, L.; Hall, M.; Spahr, C.; Walker, K.; Laudemann, J.; Miller, S.; Salimi-Moosavi, H.; Lee, J. W. Anal. Chem. 2012, 84, 1267−1273. (17) Neubert, H.; Muirhead, D.; Kabir, M.; Grace, C.; Cleton, A.; Arends, R. Anal. Chem. 2013, 85, 1719−1726. (18) Heinig, K.; Wirz, T. Anal. Chem. 2009, 81, 3705−3713. (19) Kushnir, M. M.; Rockwood, A. L.; Roberts, W. L.; Abraham, D.; Hoofnagle, A. N.; Meikle, A. W. Clin. Chem. 2013, 59, 982−990. (20) Ouyang, Z.; Furlong, M. T.; Wu, S.; Sleczka, B.; Tamura, J.; Wang, H.; Suchard, S.; Suri, A.; Olah, T.; Tymiak, A.; Jemal, M. Bioanalysis 2012, 4, 17−28. (21) Liu, G.; Ji, Q. C.; Dodge, R.; Sun, H.; Shuster, D.; Zhao, Q.; Arnold, M. J. Chromatogr., A 2013, 1284, 155−162. (22) Yuan, L.; Arnold, M. E.; Aubry, A. F.; Ji, Q. C. Bioanalysis 2012, 4, 2887−2896. (23) Guidance for Industry: Bioanalytical Method Validation. Food and Drug Administration, 2001. http://www.fda.gov/downloads/ Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ ucm070107.pdf (accessed Mar 21, 2013). (24) Guideline on bioanalytical method validation. European medicines agency, 2011. http://www.ema.europa.eu/docs/en_GB/ document_library/Scientific_guideline/2011/08/WC500109686.pdf (accessed Mar 21, 2013). (25) Babu, K. R.; Douglas, D. J. Biochemistry 2000, 39, 14702−14710. (26) Babu, K. R.; Moradian, A.; Douglas, D. J. J. Am. Soc. Mass Spectrom. 2001, 12, 317−328. (27) Furlong, M. T.; Ouyang, Z.; Wu, S.; Tamura, J.; Olah, T.; Tymiak, A.; Jemal, M. Biomed. Chromatogr. 2012, 26, 1024−1032. (28) Vogt, W. Free Radic. Biol. Med. 1995, 18, 93−105. (29) Schöneich, C. Biochim. Biophys. Acta 2005, 1703, 111−119. (30) Stadtman, E. R.; Van Remmen, H.; Richardson, A.; Wehr, N. B.; Levine, R. L. Biochim. Biophys. Acta 2005, 1703, 135−140.

9867

dx.doi.org/10.1021/ac402420v | Anal. Chem. 2013, 85, 9859−9867