Development and validation of a liquid chromatography tandem mass spectrometry assay for the quantitation of a protein therapeutic in cynomolgus monkey serum

Journal of Chromatography B, 988 (2015) 81–87

Contents lists available at ScienceDirect

Journal of Chromatography B journal homepage: www.elsevier.com/locate/chromb

Development and validation of a liquid chromatography tandem mass spectrometry assay for the quantitation of a protein therapeutic in cynomolgus monkey serum Yue Zhao a , Guowen Liu a,∗ , Aida Angeles a , Lora L. Hamuro a , Kevin J. Trouba b , Bonnie Wang c , Renuka C. Pillutla a , Binodh S. DeSilva a , Mark E. Arnold a , Jim X. Shen a a

Bioanalytical Sciences, Research & Development, Bristol-Myers Squibb Co., Route 206 and Province Line Road, Princeton, NJ 08543, USA Drug Safety Evaluation, Bristol-Myers Squibb Co., 4601 Highway 62 East, Mt Vernon, IN 47620, USA c Drug Safety Evaluation, Bristol-Myers Squibb Co., 1 Squibb Drive, New Brunswick, NJ 08903, USA b

a r t i c l e

i n f o

Article history: Received 24 October 2014 Received in revised form 6 February 2015 Accepted 8 February 2015 Available online 16 February 2015 Keywords: LC-MS/MS Protein therapeutics Bioanalysis Albumin removal Validation

a b s t r a c t We have developed and fully validated a fast and simple LC-MS/MS assay to quantitate a therapeutic protein BMS-A in cynomolgus monkey serum. Prior to trypsin digestion, a recently reported sample pretreatment method was applied to remove more than 95% of the total serum albumin and denature the proteins in the serum sample. The pretreatment procedure simplified the biological sample prior to digestion, improved digestion efficiency and reproducibility, and did not require reduction and alkylation. The denatured proteins were then digested with trypsin at 60 ◦ C for 30 min and the tryptic peptides were chromatographically separated on an Acquity CSH column (2.1 mm × 50 mm, 1.7 ␮m) using gradient elution. One surrogate peptide was used for quantitation and another surrogate peptide was selected for confirmation. Two corresponding stable isotope labeled peptides were used to compensate variations during LC-MS detection. The linear analytical range of the assay was 0.50–500 ␮g/mL. The accuracy (%Dev) was within ±5.4% and the total assay variation (%CV) was less than 12.0% for sample analysis. The validated method demonstrated good accuracy and precision and the application of the innovative albumin removal sample pretreatment method improved both assay sensitivity and robustness. The assay has been applied to a cynomolgus monkey toxicology study and the serum sample concentration data were in good agreement with data generated using a quantitative ligand-binding assay (LBA). The use of a confirmatory peptide, in addition to the quantitation peptide, ensured the integrity of the drug concentrations measured by the method. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Protein therapeutics is one of the fastest growing classes of drugs in the pharmaceutical industry because of their high specificities and low toxicity. Currently, more than 130 proteins or peptides have been approved for clinical use by the US Food and Drug Administration (FDA) [1]. The rapid growth of protein therapeutics places high demands on fast, accurate and rugged bioanalytical methods to support their development. In recent years, LC-MS/MS has been demonstrated as a promising

∗ Corresponding author at: Bristol-Myers Squibb Co, Bioanalytical Sciences, Research & Development, Route 206 and Province Line Road, Princeton, NJ 08543, United States. Tel.: +1 609 252 5343. E-mail address: [email protected] (G. Liu). http://dx.doi.org/10.1016/j.jchromb.2015.02.007 1570-0232/© 2015 Elsevier B.V. All rights reserved.

alternative platform for quantitation and characterization of protein therapeutics [1–3]. Compared to the ligand binding assay (LBA), which has been the gold standard in measuring the protein drug concentrations in biological fluids, the LC-MS/MS technique provides advantages including a wider dynamic range, faster method development, and improved specificity [4,5]. In addition, sample pretreatment/cleanup steps typically associated with LC-MS/MS often disrupt the interference from anti-drug antibodies (ADA) which can interfere with LBA methodologies [6,7]. For the advantages mentioned above and to keep up with the fast increasing demand for bioanalytical support on biologics at BristolMyers Squibb (BMS), we are developing LC-MS/MS based assays for protein therapeutics to serve as a complementary tool to LBA. Although it is possible to analyze intact proteins using high resolution MS techniques, the sensitivity and specificity are still

82

Y. Zhao et al. / J. Chromatogr. B 988 (2015) 81–87

suboptimal due to mass resolution limitations [8]. The most frequently applied technique for the quantitation of proteins in biological specimens involves performing enzymatic digestion followed by the measurement of one or more quantitation peptides (a peptide used to represent the entire protein) from the digested samples. In general, a quantitation peptide has the following desirable attributes: a unique sequence without interference from endogenous background; contains no unstable amino acids (e.g., methionine, cysteine), and produces good MS sensitivity with suitable chromatographic retention [8]. Understandably, sample digestion, the process to produce such surrogate peptide(s) from the target protein, is the most critical step for protein quantitation using LC-MS/MS. Variations from the digestion step will impact precision and accuracy of the bioanalytical assay. A number of approaches [9,10] have been proposed to compensate the variations during digestion. Of the most simple to implement, is the use of a stable-isotope-labeled (SIL)-protein as the internal standard (IS) at the beginning of sample preparation process. This is desirable since the SIL-IS will standardize both digestion efficiency, as well as the subsequent LC-MS/MS analysis. However, an SIL-protein is not always available due to the significant cost associated with its production, as well as the time required for generation. Additionally, the paucity of SIL-proteins is especially acute during the early stages of drug development. Therefore, how to minimize variation during the digestion step is a constant challenge for scientists applying LC-MS/MS to protein bioanalysis. As such, alternative bioanalytical methods to improve efficiency and consistency have been conceptualized by scientists (e.g. the use of SIL-peptides to compensate variations during the LC-MS/MS analysis). The complexity of biological samples (e.g., high levels of endogenous peptides) following digestion is another challenge affecting both sensitivity and assay ruggedness for LC-MS/MS quantitation of proteins. A number of different sample preparation procedures have been used to simplify the plasma/serum sample environment, these include pellet digestion [11] (remove small molecules), albumin depletion [12] (remove the most abundant endogenous protein) and immunocapture [13–15] (specific extraction of target protein). Sensitivity and ruggedness requirements of the assay often dictate sophistication and complexity in the cleanup. Typically, higher assay sensitivity and ruggedness requirements demand a thorough sample cleanup method at a higher cost (i.e. immunocapture). Given this, a cost effective and easy to use method, which maintains or improves assay ruggedness and sensitivity whilst keeping the cost low, is most desirable. We recently reported our experience of a novel albumin removal approach which is cost effective, easy to use, and high throughput [16]. This approach can remove more than 95% total serum albumin from the serum or plasma sample whilst improving assay ruggedness and sensitivity. There may be some concerns on adding extra variation during the albumin removal step since no internal standard is used to track this process. However, unlike immunoaffinity extraction, this step is simple and easy to implement. No significant variation was observed. In the current work, we are reporting an LC-MS/MS method for a protein drug candidate (BMS-A), focusing on the operational details and practical challenges for method development and validation. In general, the same practices for small molecule bioanalysis using LCMS/MS can be followed for protein therapeutics (large molecules). The single noticeable difference is the extra digestion step, which is unique to protein bioanalysis using the surrogate peptide approach. Our focus will be on how to practically evaluate “surrogate peptide generation efficiency” (the efficiency of generating the surrogate peptide from the target protein through the overall sample preparation process, including sample pretreatment and digestion. This is in constrast to analyte recovery for small molecule bioanalysis) and the factors that impact this evaluation. The assay uses the

albumin removal approach [16] as a pre-digestion cleanup followed by tryptic digestion for BMS-A in cynomolgus monkey serum. One quantitation peptide (ITYG, note, we used the first four amino acids from each tryptic peptide to name our quantitation peptides for BMS proprietary peptide sequence) and one confirmatory peptide VVSVLTVLHQDWLNGK (VVSV) were chosen for quantitation. Stable-isotope labeled versions of those two peptides were used as the internal standards and were added after the digestion step. This method was fully validated according to regulatory guidelines [17,18] and Bristol-Myers Squibb (BMS) internal standard operating procedures. Moreover, the newly developed albumin removal method was compared to the more frequently used pellet digestion method [11]. Serum samples obtained from a monkey toxicological study were analyzed by both LC-MS/MS and LBA, and data from both platforms were compared. 2. Experimental details 2.1. Chemicals, reagents, materials and apparatus HPLC grade methanol and isopropanol (IPA) were purchased from J.T. Baker (Phillipsburg, NJ, USA). Formic Acid (FA) (SupraPur grade) was purchased from EMD Chemicals (Gibbstown, NJ, USA). LC grade acetonitrile, trypsin from bovine pancreas (TPCK treated, salt free, lyophilized powder, ≥10,000 BAEE units/mg protein), Human Serum Albumin (HSA), ammonium bicarbonate, dithiothreitol (DTT), iodoacetamide (IAA) and trichloroacetic acid (TCA) were purchased from Sigma–Aldrich (St. Louis, MO, USA). Deionized water was generated in house using a NANOpure Diamond ultrapure water system from Barnstead International (Dubuque, IA, USA). Cynomolgus monkey serum was obtained from Bioreclamation, Inc. (Westbury, NY, USA). Protein BMS-A and the stable isotope-labeled surrogate peptide internal standards SIL-ITYG and SIL-VVSV were produced internally at BMS. 2.2. Equipment and apparatus All sample analyses were performed on a triple quadrupole 5500 mass spectrometer (AB Sciex, Foster City, CA), which was controlled by Analyst® 1.5.1 software. The mass spectrometer was coupled with a Shimadzu (Columbia, MD, USA) Nexera UHPLC system. The Shimadzu UHPLC system consists of two LC-30AD pumps, two DGU-20A5 degassers, one SIL-30ACMP autosampler and one CTO30AS column heater. Chromatographic separation was achieved on a Waters (Milford, MA, USA) Acquity UPLC CSH C18 column ˚ 1.7 ␮m, 2.1 mm × 50 mm). A thermomixer R model 5355 (130 A, from Eppendorf (Hamburg, Germany) and MTP Microblock was used for trypsin digestion. An automated liquid handler, JANUS® Mini from PerkinElmer (Waltham, MA, USA) was used for adding and transferring liquid. 2.3. Preparation of calibration standards and quality control samples for BMS-A Protein BMS-A stock solution (52.5 mg/mL) was diluted to prepare the calibration standard curves in cynomolgus serum with concentrations of 0.50, 1.25, 5.00, 20.0, 50.0, 200, 375 and 500 ␮g/mL. Quality control (QC) samples including lower limit of quantitation (LLOQ), low QC, geometric mean (GM) QC, middle QC, high QC and dilution QC were prepared similarly at concentrations of 0.50, 1.50, 20.0, 250, 425 and 5,000 ␮g/mL, respectively. After preparation, the QC samples were stored at −70 ◦ C before analysis. Two stock solutions of the stable isotope-labeled internal standard (SIL-ITYG and SIL-VVSV) were prepared in amber glass vial at a concentration of 1.00 mg/mL in 0.1% FA in 50:50 (v:v) water:methanol. The stock solutions were further diluted into a combined solution

Y. Zhao et al. / J. Chromatogr. B 988 (2015) 81–87

with a final concentration of 1000 ng/mL for both peptides with 10% formic acid (FA) in water and used as the internal standard working solution. The internal standard working solution was stored at 2–8 ◦ C and used within 3 months of preparation. 2.4. Sample preparation and trypsin digestion A 20 ␮L of serum sample (calibration standards, QCs or study samples) was mixed with 200 ␮L of IPA with 1.0% (by weight) TCA solution to generate protein pellets. After vortexing vigorously for 2 min, the samples were centrifuged at 1500 × g in a refrigerated centrifuge (5 ◦ C) for 5 min. The supernatant was removed using a JANUS Mini liquid handler and 200 ␮L of methanol was added to all samples to wash the pellets. The pellets were re-suspended in methanol, centrifuged at 2000 rpm for 2 min and the supernatant was removed using the JANUS Mini liquid handler. The washed pellets were re-suspended in 200 ␮L of 100 mM ammonium bicarbonate aqueous buffer and 25 ␮L of trypsin reagent (20 ␮g/␮L prepared in 100 mM ammonium bicarbonate) was added. The mixture was incubated at 60 ◦ C for 30 min with a shaker set at 750 rpm. Following incubation, the digestion was quenched by adding 50 ␮L IS working solution (containing 10% FA) to the samples. The final samples were vortex mixed and transferred to a 96-well plate for LC-MS analysis. For the reduction and alkylation experiments, after the protein pellets were resuspended in 100 mM ammonium bicarbonate buffer, 10 ␮L of 100 mM DTT was added to the samples and incubated at 60 ◦ C for 60 min to reduce the denatured protein. The samples were further alkylated with 10 ␮L of 100 mM IAA at 30 ◦ C for 30 min in the dark. The trypsin digestion step was then carried out as described above. 2.5. HPLC-MS/MS conditions Mobile phase A contained 0.1% FA in water and mobile phase B contained 0.1% FA in methanol. Chromatographic separation was achieved using gradient elution with a flow rate of 0.8 mL/min on a Waters CSH C18 UPLC column. The column temperature was set to maintain at 80 ◦ C and the following gradient was applied: mobile phase B was maintained at 15% B till 0.50 min; then a linear gradient increased over 5.5 min from 15% to 48%. The gradient was changed to 100% B in 0.01 min, held for 1 min; and decreased back to 15% in 0.01 min. The total run time is 8.00 min. A diverting valve was used with solvent diverted to waste between 0 and 3.2 min; switched back to MS from 3.2 to 3.8 min; diverted to waste from 3.8 to 5.0 min; switched back to MS for 0.5 min, and finally diverted to waste. The tryptic digested peptides, ITYG and VVSV, were monitored using positive ion electrospray ionization (ESI) with selective reaction monitoring (SRM). The MS conditions were optimized as listed: curtain gas and collision gas were set as 35 and 8; the turbo spray voltage was set at 4000 V and ion source gas 1 and gas 2 were both set at 65 psi. Entrance Potential (EP) was maintained at 10 V. The probe temperature was set at 650 ◦ C. For the SRM detection, doubly charged molecular ions for peptide ITYG and triply charged molecular ions for peptide VVSV were selected at Q1. The SRM transitions monitored for ITYG and SIL-ITYG were 1055.2/329.2 and 1060.1/329.2 with declustering potential (DP) set at 100 V, collision energy (CE) at 45 eV. For VVSV and SIL-VVSV, the SRM transitions were 603.5/805.7 and 605.8/809.2; and DP and CE were set at 50 and 20 eV, respectively. 2.6. Nonspecific adsorption transfer experiment The adsorption test was performed in two different solvents. A combined solution of SIL-ITYG and SIL-VVSV was prepared at

83

concentrations of 84.0 and 72.0 ng/mL by diluting their stock solutions using 0.1% FA in 50:50 (v:v) water:methanol and 0.1% FA in water, respectively. The solution was prepared and stored in amber glass vials. Transfer experiments were performed to confirm adsorption issue using a procedure widely used in the industry [19]. Briefly, one aliquot of the prepared solution was taken out to a collection plate and used as the control sample. The remaining solution was transferred into a clean amber glass vial, vortex mix and placed at room temperature for approximately 10 min. After that, an aliquot was transferred to the collection plate and the process was then repeated for 4 times.

2.7. Surrogate peptide generation efficiency and matrix effect evaluation Following the industry widely accepted practices for “recovery” and matrix effect evaluation for small molecule bioanlaysis [20], we have designed our experiment as follows: Sample A: serum sample spiked with BMS-A at different concentrations subject to digestion; Sample B: digested blank serum (blank serum that went through the digest process) samples spiked with synthetic SIL surrogate peptides at concentrations equivalent to theoretical values based on BMS-A in sample A; Sample C: neat solution spiked with SIL surrogate peptides at concentrations equivalent to theoretical values based on BMS-A in sample A. Please note that in our case the so called “recovery” is actually the surrogate peptide generation efficiency, which is a combination effect of recovery of the target protein during the albumin removal step and the digestion efficiency of the target protein after pre-treatment to produce the surrogate peptides. Surrogate peptide generation efficiency (or recovery) was determined by comparing the responses (peak area) of surrogate peptides (ITYG or VVSV) from Sample A to the peak area of the corresponding SIL peptides (SIL-ITYG or SIL-VVSV) from Sample B. Matrix effect was determined in 6 individual lots by dividing the responses of the SIL peptides in the presence of matrix (Sample B) to the absence of matrix (Sample C). The digestion efficiency and matrix effect were assessed at the low and high concentrations (1.50 and 425 ␮g/mL of BMS-A) in cynomolgus monkey serum. Two stock solutions of SIL-ITYG and SIL-VVSV at 1.00 mg/mL were prepared freshly in 1% HSA in water solution for this test. These two stock solutions were further diluted using digested serum blank samples to final concentrations of equivalent to 1.50 and 425 ␮g/mL of BMS-A on the day of the experiment, and were used as postdigestion spiking solutions.

3. Results and discussion 3.1. Method development 3.1.1. Digestion condition optimization To improve digestion efficiency and consistency, the conditions were carefully optimized. Specifically, incubation temperature from 40 to 90 ◦ C, incubation time from 15 to 60 min, and the amount of trypsin added between 5 and 40 mg/mL (25 ␮L) were tested. The results of optimization are as follows: ITYG had similar MS responses (normalized to IS response) when digestion occurred at a temperature higher than 50 ◦ C, while the response for VVSV decreased dramatically at the higher temperatures, which may be due in part to the instability of VVSV [21]. A digestion time of 30 min at 60 ◦ C yielded the best results for both ITYG and VVSV, and a 20 mg/mL trypsin solution under our conditions was found to be the optimal concentration. In the final protocol, 25 ␮L of 20 mg/mL trypsin was used at 60 ◦ C for 30 min to digest all samples. The final trypsin-to-protein ratio was approximately between 1:1 and 1:2.

Y. Zhao et al. / J. Chromatogr. B 988 (2015) 81–87

3.1.3. Benefits from sample pretreatment with albumin removal As noted, we used a previously reported albumin removal approach for the samples prior to trypsin digestion [16]. This approach was designed to remove albumin, the most abundant protein in serum or plasma, from the biological samples to simplify the matrix and increase digestion efficiency and consistency. Briefly, a solution containing 1% TCA in IPA was used to remove albumin from the serum sample before digestion. This 1% TCA denatured and precipitated therapeutic proteins while allowing albumin and other endogenous substances, such as lipids and soluble peptides, to remain in the supernatant. After removal of the supernatant,

12000 Peak Area

3.1.2. Nonspecific adsorption When working with neat solution, loss of proteins and peptides due to nonspecific binding to solid surfaces such as containers, extraction plates, pipette tips and HPLC tubing could happen if not handled properly [22]. Positively charged peptides can electro-statically interact with surfaces carrying negative charges (non-treated glass surfaces which contain residual silanol groups), while nonpolar amino acids can easily interact hydrophobically with the hydrophobic surface of polypropylene containers [23]. During our initial experiments, surrogate peptide generation efficiency was measured as more than 200% for both peptides. It was suspected that nonspecific adsorption loss for the spiking peptide solutions in neat solvent was the cause for the inaccurate measurements. As described previously, surrogate peptide generation efficiency was evaluated by comparing the instrument responses of peptides ITYG/VVSV from digested serum samples containing BMS-A with the instrument responses from digested blank serum samples post-spiked with SIL-ITYG/SIL-VVSV (the amount of SIL-ITYG/SIL-VVSV spiked was based on the theoretical values calculated from BMS-A). If the spiking peptide solution had a concentration lower than its theoretical value due to nonspecific adsorption loss during the preparation process, this would result in higher measured surrogate peptide generation efficiency. A nonspecific adsorption test was therefore performed by serially transferring the peptide neat solution to confirm this. A loss of approximate 35% to 60% (data not shown) was observed after 5 transfers compared to no transfer for SIL-ITYG and SIL-VVSV in the solution prepared in 0.1% FA in 50:50 (v:v) water: methanol (which was the condition for the stock solution preparation). For solution prepared in pure aqueous environment (0.1% FA in water), a more severe adsorption effect was observed. The author also wants to emphasize that the percentage decrease reported here was compared to the control samples which did not go through any further transfer step following preparation; however, a loss may had already occurred in the control sample. Fortunately, a carrier protein (HSA: human serum albumin) at a concentration of 1% (by weight) or digested blank serum samples successfully prevented the nonspecific binding loss of both peptides. This was demonstrated by the following experiment: firstly prepared the SIL-peptide stock solutions in 1% HSA in water; then diluted the stock solutions to a low concentration (84.0 ng/mL for SIL-ITYG and 72.0 ng/mL for SIL-VVSV) in 3 different solvents: extracted blank, 0.1% FA in water and 1% HSA in water. Responses for both peptides were compared from the three different preparations. As shown in Fig. 1, solutions prepared in digested blank serum and 1% HSA showed similar but much greater responses compared to the sample prepared in 0.1% FA in water for both peptides. In addition, no nonspecific adsorption loss was observed for both peptides prepared in 1% HSA aqueous solution according to the transfer experiment (data not shown). Therefore, when performing the surrogate peptide generation efficiency and matrix effect test, the stock solutions of SIL-ITYG and SIL-VVSV was prepared in 1% HSA in water and post-digestion spiking solution in the digested blank to prevent the adsorption issue.

(A)

8000 4000 0 digested blank serum

0.1% FA in water

1% HSA in water

160000 Peak Area

84

(B)

120000 80000 40000 0 digested blank serum

0.1% FA in water

1% HSA in water

Fig. 1. Instrument response for SIL-ITYG (A) and SIL-VVSV (B) prepared in digested blank serum, 0.1% FA in water and 1% HSA in water at concentrations of 84.0 and 72.0 ng/mL, respectively.

the precipitated proteins were washed and made ready for digestion. Due to the complex three dimensional structures of the proteins and the existence of multiple disulfide bonds, sample pretreatment was a critical step to produce a consistent and/or efficient digestion. Heat denaturation, reduction and/or alkylation before the digestion are often necessary to encourage protein denaturation/unfolding and accessibility of peptides to trypsin digestion [24–26]. A previously reported pellet digestion methodology has been shown to successfully eliminated the tedious sample pretreatment steps in some cases [9,11,27], but not all [25]. In our new albumin removal method, we tested if the application of the acidified solvent can further disrupt the protein structures and improve digestion efficiency. One of the quantitation peptide of BMS-A, VVSV, is a universal surrogate peptide that exists in all human immunoglobulin G1, G3 and G4 regions, and is in close proximity to the intra-chain disulfide bond [21]. A careful selection of the pretreatment method became essential in ensuring digestion reproducibility and efficiency. We compared the following sample pretreatment methods (1) traditional pellet digestion method using methanol, (2) pellet digestion followed by reduction using DTT and alkylation using IAA, (3) the newly developed albumin removal method using 1% TCA in IPA, and (4) the 1% TCA in IPA method followed by reduction and alkylation. The experiment was performed using two different serum lots with 4 replicates. The results showed (Fig. 2) that without reduction and alkylation, samples extracted with only methanol had lower MS responses compared to the other groups. Moreover, there was an approximate 30–50% difference in MS responses between the two serum lots tested for both ITYG and VVSV. In contrast, the MS responses from the other three groups (IPA, IPA or methanol with reduction and alkylation) produced similar results with minimal differences between the two individual serum lots. Therefore, we concluded there was no additional benefit with the reduction and alkylation steps when 1% TCA in IPA was used for sample pretreatment. Considering the reduction and alkylation step would increase the complexity of sample preparation and consume additional time and reagent, the 1% TCA in IPA was considered to be a better choice.

R ela t ive R esp on se (A/I S)

Y. Zhao et al. / J. Chromatogr. B 988 (2015) 81–87

tested. The assay demonstrated excellent accuracy and precision as reported previously [16]. Specificity was evaluated in six lots of individual cynomolgus monkey serum. Representative chromatograms of a blank sample, QC0 (blank sample containing IS only) and LLOQ are presented in Fig. 3 for ITYG and VVSV. There was no endogenous interference from blank matrix to ITYG, SIL-ITYG, VVSV and SIL-VVSV at the corresponding retention time (3.5 min for ITYG and 5.3 min for VVSV). The stable-isotope labeled internal standards (SIL-ITYG and SIL-VVSV) had no contributions to the analytes channel as well. In addition, the six serum lots were also evaluated at the LLOQ level (0.50 ␮g/mL) with all six lots generating predictable concentrations within the ±20% of the nominal concentrations for ITYG and five out of six met the ±20% acceptance criteria for VVSV.

8 6

ITYG

4

lot 1

2

lot 2

R ela t ive R esp on se (A/I S)

0

15 12 9

VVSV

6 3

lot 1

0

lot 2

3.2.2. Stability evaluation Freeze–thaw, room temperature and long-term storage stability of ITYG and VVSV were evaluated based on three QC levels (low, high and dilution) analyzed in triplicate. As shown in Table 1, BMSA was stable in cynomolgus monkey serum for at least 24 h at room temperature, at least 25 days when stored at −70 ◦ C and after 4 freeze–thaw cycles. Processed sample stability was established for at least 144 h at 2–8 ◦ C by testing the processed QCs against a freshly prepared curve. The extracted samples were stable for at least 180 h in the autosampler at 2–8 ◦ C.

Fig. 2. Comparison between different sample pretreatments with and without reduction and alkylation in samples prepared at 250 ␮g/mL in two different monkey serum lots. Each data point (relative response of analyte (A) to its internal standard (IS)) presented was an average of four measurements.

3.2.3. Surrogate peptide generation efficiency (recovery) and matrix effect Due to the unavailability of the reference materials of the ITYG and VVSV peptides, surrogate peptide generation efficiency test was performed by comparing the ITYG and VVSV responses from the digested sample to the SIL-ITYG and SIL-VVSV responses from the post-spiked sample. The surrogate peptide generation efficiency (recovery) was ∼83–98% for ITYG and ∼86–105% for VVSV. These numbers may not be the absolute values since many assumptions have been made during the evaluation. However, it did help to guide us through the method development and optimization process. The matrix effect (MS ion enhancement/suppression) was determined in six individual lots by calculating the matrix factor (MF)

3.2. Method validation 3.2.1. Accuracy and precision and specificity The analytical range was established over from 0.50 to 500 ␮g/mL with excellent linearity (see Supplementary Data). Different regression models (linear vs. quadratic with x, 1/x, or 1/x2 ) were evaluated by analyzing all accuracy and precision runs in the validation using BMS internal regression analysis software. A linear, 1/x2 -weighted regression model was chosen for both peptides. The regression coefficient was not less than 0.992 for any of the runs

M S R es p o n s e ( cp s )

900 A1

M S R es p o n s e ( cp s )

600

600

B1

400

300

0

85

200

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

0 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5

8.0

3500 B2

6000 A2 S/N=23

3000

S/N=22

4000 2000 2000

0

1000

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 Time (min)

4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 Time (min)

8.0

Fig. 3. Representative LC-MS/MS chromatograms of the two surrogate peptides: (1) ITYG (A1) and VVSV (B1) of a blank cynomolgus serum sample; (2) ITYG (A2) and VVSV (B2) of an LLOQ sample (0.50 ␮g/mL) in cynomolgus serum.

86

Y. Zhao et al. / J. Chromatogr. B 988 (2015) 81–87

Table 1 Stability information for BMS-A in cynomolgus monkey serum based on two surrogate peptides: (A) ITYG and (B) VVSV. (A) QC type (nominal conc.)

Low QC 1.50 (␮g/mL)

High QC 425.00 (␮g/mL)

Dilution QC 5000.00 (␮g/mL)

Sample condition

Mean conc.

%Dev

Mean conc.

%Dev

Mean conc.

%Dev

24 h at RT in serum After 4 freeze–thaw Cycle in serum 25 day at −70 ◦ C in serum 144 h at 2–8 ◦ C processed samples 168 h at 2–8 ◦ C Re-injection integrity

1.55 1.68 1.44 1.28 1.45

3.5 12.1 −3.8 −14.6 −3.2

437.17 463.61 427.44 428.54 429.64

2.9 9.1 0.6 0.8 1.1

4471.89 5148.83 4924.26 4765.86 4816.46

−10.6 3.0 −1.5 −4.7 −3.7

(B) QC type (nominal conc.)

Low QC 1.50 (␮g/mL)

Sample condition

Mean conc.

%Dev

Mean conc.

High QC 425.00 (␮g/mL) %Dev

Mean conc.

%Dev

24 h at RT in serum After 4 freeze–thaw cycle in serum 25 day at −70 ◦ C in serum 144 h at 2–8 ◦ C processed samples 168 h at 2–8 ◦ C Re-injection integrity

1.40 1.49 1.40 1.28 1.34

−6.5 −0.9 −6.8 −14.6 −10.9

456.95 379.24 441.39 406.62 450.35

7.5 −10.8 3.9 −4.3 6.0

4896.57 5182.71 5132.33 4750.14 5162.31

−2.1 3.7 2.6 −5.0 3.2

for each lot and comparing the responses of the SIL-peptides in the spiked digested blank to those of the spiked neat solvent (0.1% FA in water). The mean of MF were 1.14 (CV% 17.8%) and 0.97 (CV% 11.8%) for ITYG and VVSV, respectively. The results indicated very little or no ion enhancement for ITYG and VVSV. In addition, a postcolumn infusion experiment [28] was performed by continuously delivering neat solution of the SIL-peptides prepared at 1000 ng/mL in 0.1% FA in water directly to the MS ion source. A sample only contained neat solution (0.1% FA in water) was injected followed by an extracted blank sample. The results showed there was no observed difference in the MS signal for the two peptides when injecting neat solution vs. extracted blank. This result is in agreement with the result from the matrix effect experiment described above. 3.3. Method performance 3.3.1. Sample analysis After method validation, monkey serum samples from a toxicology study were analyzed. ITYG, a unique peptide to BMS-A, was selected as the primary quantitation peptide whose concentrations were reported for unknown samples; VVSV was used as the confirmatory peptide to ensure accurate quantitation results. For each run, the concentrations between VVSV and ITYG were compared. For 90% of all samples tested, the measured concentrations between the two surrogate peptides were within 10% of their mean concentration. In addition, the assay showed great reproducibility demonstrated by an incurred sample analysis (ISR) test which was reported previously [16]. As shown in Table 2, the assay accuracy

Dilution QC 5000.00 (␮g/mL)

was within ±4.0% for ITYG and ±5.4% for VVSV based on Low, GM, Mid and High QCs across seven runs. The total assay variation (%CV) was less than 12.0% for ITYG and 11.4% for VVSV.

3.3.2. Comparison of LC-MS/MS vs. ligand-binding assay Study sample results obtained by the validated LC-MS/MS assay were compared to the data generated using a validated LBA assay. In general, excellent correlation was obtained between two sets of data. Fig. 4 shows a correlation plot between LBA and LC-MS/MS concentrations in the X and Y axes, respectively with data generated at early time points (up to day 8). The two sets of data with slope of 1.1846 and R2 equals to 0.9713, indicating the LC-MS/MS data was slightly and consistently higher than the LBA data. In addition, some monkeys demonstrating antidrug antibody (ADA) formation at later phases of dosing, LC-MS/MS data showed much higher concentrations than those measured using LBA. Such data have been reported elsewhere recently [16]. In general, LC-MS/MS and LBA results were in good agreement from day 1 to day 8 in all animals; however, in the later time points (day 15 to day 25) higher concentrations were reported from the LC-MS/MS for animals developed significant immunogenicity. These results suggested that the LBA method was likely affected by ADA interferences which resulted in an underestimation of the drug concentrations while the LCMS/MS assay was not affected by the presence of ADA. Therefore, as an alternative technology to LBA in biotherapeutics bioanalysis, LC-MS/MS could also provide useful complimentary data in cases where ADA interference had a detrimental effect on the LBA.

Table 2 Accuracy and precision results during sample analysis for BMS-A in monkey serum based on two surrogate peptides: (A) ITYG and (B) VVSV. (A) Nominal conc.

Low (1.50 ␮g/mL)

GM (20.00 ␮g/mL)

Mid (250.00 ␮g/mL)

High (425.00 ␮g/mL)

Mean observed conc. %Dev Total variation (%CV) N Number of runs

1.56 4.0 10.4 20 7

20.13 0.6 12.0 20 7

257.64 3.1 10.4 20 7

434.15 2.2 10.8 20 7

Nominal Conc.

Low (1.50 ␮g/mL)

GM (20.00 ␮g/mL)

Mid (250.00 ␮g/mL)

High (425.00 ␮g/mL)

Mean observed conc. %Dev Total variation (%CV) N Number of runs

1.47 −2.0 11.4 20 7

19.42 −2.9 10.6 20 7

256.31 2.5 8.3 20 7

447.96 5.4 7.5 20 7

(B)

Y. Zhao et al. / J. Chromatogr. B 988 (2015) 81–87

References

1600 LC-MS/MS (µg/mL)

87

1400 y = 1.1846x + 6.7445 R² = 0.9713

1200 1000 800 600 400 200 0 0

200

400

600

800

1000

1200

1400

LBA (µg/mL) Fig. 4. Comparison of the concentration data generated from the LC-MS/MS assay and LBA. The linear regression equation and the plot describe a good correlation between the two sets of data.

4. Conclusions To conclude, a LC-MS/MS method was validated over a range of 0.50 to 500 ␮g/mL, which was sufficient for the repeat dose cynomolgus monkey toxicology study supported. Two surrogate peptides were measured with one as the quantitation peptide to represent the concentrations of the intact protein and the other one as the confirmatory peptide to ensure the quantitation accuracy. The digestion conditions were carefully optimized to improve digestion efficiency. Peptide adsorption issue was discovered and resolved during the surrogate peptide generation efficiency and matrix effect test. The validated method demonstrated good accuracy and precision. This LC-MS/MS method is fast, simple, cost effective and high throughput. The enclosed procedure is suitable for high throughput LC-MS/MS bioanalytical laboratories where protein therapeutic drugs concentrations are necessary or optimal to support development of protein therapeutics. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.jchromb.2015.02.007.

[1] I. van den Broek, W.M. Niessen, W.D. van Dongen, J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 929 (2013) 161–179. [2] S.T. Wu, Z. Ouyang, T.V. Olah, M. Jemal, Rapid Commun. Mass Spectrom.: RCM 25 (2011) 281–290. [3] M. Fernandez Ocana, I.T. James, M. Kabir, C. Grace, G. Yuan, S.W. Martin, H. Neubert, Anal. Chem. 84 (2012) 5959–5967. [4] E. Ezan, F. Bitsch, Bioanalysis 1 (2009) 1375–1388. [5] C. Tamvakopoulos, Mass Spectrom. Rev. 26 (2007) 389–402. [6] S.J. Wang, S.T. Wu, J. Gokemeijer, A. Fura, M. Krishna, P. Morin, G. Chen, K. Price, D. Wang-Iverson, T. Olah, R. Weiner, A. Tymiak, M. Jemal, Anal. Bioanal. Chem. 402 (2012) 1229–1239. [7] E. Ezan, Bioanalysis 3 (2011) 469–471. [8] J. Zheng, J. Mehl, Y. Zhu, B. Xin, T. Olah, Bioanalysis 6 (2014) 859–879. [9] G. Liu, Q.C. Ji, R. Dodge, H. Sun, D. Shuster, Q. Zhao, M. Arnold, J. Chromatogr, A 1284 (2013) 155–162. [10] K.J. Bronsema, R. Bischoff, N.C. van de Merbel, J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 893–894 (2012) 1–14. [11] Z. Ouyang, M.T. Furlong, S. Wu, B. Sleczka, J. Tamura, H. Wang, S. Suchard, A. Suri, T. Olah, A. Tymiak, M. Jemal, Bioanalysis 4 (2012) 17–28. [12] Q. Lu, X. Zheng, T. McIntosh, H. Davis, J.F. Nemeth, C. Pendley, S.L. Wu, W.S. Hancock, Anal. Chem. 81 (2009) 8715–8723. [13] H. Neubert, D. Muirhead, M. Kabir, C. Grace, A. Cleton, R. Arends, Anal. Chem. 85 (2013) 1719–1726. [14] J. Palandra, A. Finelli, M. Zhu, J. Masferrer, H. Neubert, Anal. Chem. 85 (2013) 5522–5529. [15] T. Liu, W.J. Qian, H.M. Mottaz, M.A. Gritsenko, A.D. Norbeck, R.J. Moore, S.O. Purvine, D.G. Camp, R.D. Smith 2nd, Mol. Cell Proteomics 5 (2006) 2167–2174. [16] G. Liu, Y. Zhao, A. Angeles, L.L. Hamuro, M.E. Arnold, J.X. Shen, Anal. Chem. 86 (2014) 8336–8343. [17] F.a.D.A. US Department of Health and Human Services, CDER, Rockville (2001). [18] European Medicines Agency (2012). [19] A.J. Ji, Z. Jiang, Y. Livson, J.A. Davis, J.X. Chu, N. Weng, Bioanalysis 2 (2010) 1573–1586. [20] B.K. Matuszewski, M.L. Constanzer, C.M. Chavez-Eng, Anal. Chem. 75 (2003) 3019–3030. [21] M.T. Furlong, Z. Ouyang, S. Wu, J. Tamura, T. Olah, A. Tymiak, M. Jemal, Biomed. Chromatogr. 26 (2012) 1024–1032. [22] A. Pezeshki, V. Vergote, S. Van Dorpe, B. Baert, C. Burvenich, A. Popkov, B. De Spiegeleer, J. Pharm. Biomed. Anal. 49 (2009) 607–612. [23] I. van den Broek, R.W. Sparidans, J.H. Schellens, J.H. Beijnen, J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 872 (2008) 1–22. [24] H. Li, R. Ortiz, L. Tran, M. Hall, C. Spahr, K. Walker, J. Laudemann, S. Miller, H. Salimi-Moosavi, J.W. Lee, Anal. Chem. 84 (2012) 1267–1273. [25] H. Jiang, J. Zeng, C. Titsch, K. Voronin, B. Akinsanya, L. Luo, H. Shen, D.D. Desai, A. Allentoff, A.F. Aubry, B.S. Desilva, M.E. Arnold, Anal. Chem. 85 (2013) 9859–9867. [26] O. Heudi, S. Barteau, D. Zimmer, J. Schmidt, K. Bill, N. Lehmann, C. Bauer, O. Kretz, Anal. Chem. 80 (2008) 4200–4207. [27] L. Yuan, M.E. Arnold, A.F. Aubry, Q.C. Ji, Bioanalysis 4 (2012) 2887–2896. [28] R. Bonfiglio, R.C. King, T.V. Olah, K. Merkle, Rapid Commun. Mass Spectrom. 13 (1999) 1175–1185.