A High-Resolution, Accurate-Mass Approach for Ultra-High Throughput Screening Plasma Protein Binding

A High-Resolution, Accurate-Mass Approach for Ultra-High Throughput Screening Plasma Protein Binding Keeley Murphy, 1 Patrick Bennett, 1 Francois Espourteille, 2 Maciej Bromirski 3 1 Thermo Fisher Scientific, San Jose,CA; 2 Thermo Fisher Scientific, Franklin, MA; 3 Thermo Fisher Scientific, Bremen, Germany

Overview Purpose: To evaluate high throughput liquid chromatography (LC) multiplexing and high-resolution, accurate-mass analysis using a single generic method and rapid sample cycle times for protein plasma binding (PPB) assay measurements. Methods: A panel of commercially available drug compounds was incubated in an in vitro serum protein binding assay and were analyzed using a Thermo Scientific TM Exactive TM Plus mass spectrometer and Thermo Scientific TM Transcend TM LC multiplexing system. Sample results from LC multiplexing analysis were compared to results collected with single channel analysis as well as to results previously collected using LCMS/MS analysis with a triple quadrupole mass spectrometer. Results: Results for the incubated compound sets in the PPB assay meet the LOQ requirement of 5 nm for 85% of the compounds analyzed and a linear dynamic range of 3 orders of magnitude. The coefficient of variance between sample replicates across multiple channels is demonstrated to be less than 15% for all compounds and 5% or less for most. The percentage of unbound compound was calculated for all drugs with results correlating well to results previously collected by LCMS/MS for both single channel and multi-channel analysis. An overall cycle time of 19 seconds per sample was achieved indicating that fast LC gradients and multi-channel analysis did not decrease assay performance. Introduction In-vitro screening assays commonly used in the adsorption, distribution, metabolism and excretion (ADME) development phase of drug discovery rely on analytical methods that are straightforward and robust while maintaining a high level of performance for data quality. Currently the majority of these assays are analyzed using triple quadrupole mass spectrometers with LC gradients that are less than 2 minutes in duration. While effective, this approach requires the development of MS/MS conditions and instrument methods that contain idle mass spectrometry time during the desalting and column re-equilibration phases of the LC. The implementation of high-resolution, full scan mass spectrometry coupled with high throughput LC multiplexing eliminates the need for MS/MS method development while more efficiently utilizing instrument analysis time. Methods Sample Preparation A panel of twenty-two commercially available drug compounds was incubated in an in vitro serum protein binding assay, at an incubation concentration of 10 µm. Samples were incubated for 6.5 hours in a dialysis block followed by protein precipitation. Protein precipitation was performed by first adding 150 ml of acetonitrile containing internal standard compound (Alprenolol) to a 96-well 340-mL V-bottomed storage plate followed by addition of 50 ml of each of the assay samples. Calibration curves were also generated for each compound. A working stock solution of 50 mm in DMSO was first made for each assayed compound. A five-point standard curve at concentrations of 5, 50, 500, 1000 and 2000 nm was prepared for each compound by serial dilution from the working stock solution into blank mixed matrix using an eight channel pipette 1. Three groups of assay samples were prepared for each compound analyzed: Buffer and Serum samples from the dialysis block for % Free calculations as well as a Time Zero sample. Liquid Chromatography Gradient elution was accomplished using water (A) + 0.1% Formic Acid (v/v) and Acetonitrile (B) + 0.1% Formic Acid (v/v). The gradient was held at 98% aqueous for 0.25 minutes, ramped to 98% B over 0.35 minutes, and held at 98% B for 0.2 minutes before returning to the starting conditions at 2% B for a 0.4 minute equilibration time. LC multiplexing parameters for the Transcend system were optimized to achieve a 19 second sample collection cycle time. The optimized LCMS method was used for all sample analysis. Assay samples were analyzed using a four channel LC multiplexing system with each individual sample injected onto each LC channel in replicate injections to measure single channel and inter-channel variability across all analytical columns. Chromatographic separation was performed using a C18, 2.1 x 30 mm, 3µm column with 5 µl injections made for each sample. All sample injections were completed using a Thermo Scientific Transcend LX-4 system with a dual injector arm with DLW (Dynamic Load and Wash) at a flow rate of 900 µl/min 2 A High-Resolution, Accurate-Mass Approach for Ultra-High Throughput Screening Plasma Protein

N L: 1.15E 8 m /z= 267.1842-267.1868 M S PPB_HTS_Sin gle_set2_357 N L: 2.16E 7 m /z= 250.1788-250.1814 M S PPB_HTS_Sin gle_set2_357 Mass Spectrometry Samples were analyzed using an Exactive Plus mass spectrometer in Full Scan mode (m/z 220 900) with a resolution setting of 35,000 (FWHM) at m/z 200 and a spectral speed of 7 Hz. Generic ion source conditions were used for all sample collection including vaporizer temperature 550 C, capillary temperature 350 C, sheath gas of 55 arbitrary units, and an auxiliary gas of 20 arbitrary units. The instrument was calibrated in positive ion mode before sample acquisition using Pierce LTQ Velos ESI Positive Ion Calibration Solution. Data Analysis Data was acquired using Thermo Scientific Xcalibur 2.2 and Exactive Tune 2.1 software. Chromatographic data review and calibration curve generation was performed and reported using QuickCalc TM software (powered by Gubbs Inc., GMSU Gubbs Mass Spec Utilities, Atlanta, GA). Peak area measurements in the buffer chamber of the dialysis plate were compared to the peak area measurement in the serum chamber of the dialysis plate to calculate the percent of unbound compound (% Free) at assay equilibrium 1. The % Free values for each compound replicate were compared across LC channels and also compared to values previously obtained using a triple quadrupole mass spectrometer. The coefficient of variation of the area ratio for the Buffer, Serum, and Time Zero samples we also calculated and compared across LC channels to measure variability across LC columns. Results Data File Review The collection of multiple data points stored in a single data file allows for rapid sample injections and short sample to sample cycle times, but also requires software tools that simplify data review and automatically correlate individual chromatographic peaks to their respective injections (Figure 1). QuickCalc TM software was used to automatically correlate all sample injections contained in each data file with their respective chromatographic peaks areas as well as sample types and texts contained in the original sequence list. Additionally, identification of the LC channel to which each sample was injected was included with the sample information and was used to sort and calculate signal response and variation across each channel (Figure 2). FIGURE 1. Data file review of 64 sample injections in a single data file using Qual Browser. Analyte Desipramine, Internal Standard Alprenolol. Spectra display for RT at 11.59 min, T 0 replicate number 3. Desipramine _Buffer-1 RT: 0.00-21.00 20.81 100 80 20.15 19.80 60 11.59 8.96 10.27 18.83 19.49 40 2.36 6.29 8.30 1.70 3.01 3.68 4.98 5.62 7.59 7.96 0.73 1.04 4.32 17.51 18.19 20 12.58 13.91 14.57 15.55 16.20 16.87 0 15.54 100 16.86 18.18 20.80 80 14.88 16.19 19.48 8.95 11.58 10.26 1.03 2.35 3.67 14.56 60 0.72 4.97 1.69 3.00 4.31 6.28 8.29 5.61 6.93 7.59 40 20 12.58 13.90 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Time (min) Relative Abundance PPB_HTS_Single_Set2_357 #4970 RT: 11.58 AV: 1 NL: 1.60E7 T: FTMS + c ESI Full ms [220.00-900.00] 100 267.1849 Buffer Rep 1 Rep 2 Rep 3 Acebutolol 5.1% 4.1% 2.2% Buspirone 1.0% 1.8% 3.2% Carbamazepine 4.6% 8.4% Chlorpheniramine 5.0% 2.8% 6.2% Clomipramin 6.9% 3.2% 2.9% Clozapine 3.0% 0.9% 3.4% Desipramine 2.6% 4.2% 2.9% Diltiazem 1.2% 1.9% 2.2% Erythromycin 3.7% 2.6% 3.6% Fluphenazine 4.4% 6.8% Haloperidol 8.3% 5.9% 8.1% Imipramine 4.7% 1.3% 3.2% Metoprolol 3.8% 5.2% 4.5% Nefazadone 3.4% 1 11.4% Phentolamine 12.2% 15.4% 9.4% Propranolol 3.4% 2.9% 1.4% Retonavir 14.2% 13.4% n/a Thioridazine 6.3% 2.2% 5.5% Ticlopidine 4.9% 10.6% 7.1% Timolol 3.5% 8.2% Verapamil 5.1% 4.5% 90 80 Relative Abundance 70 60 50 40 30 20 10 0 250.1797 268.1884 255.0625 251.1833 261.1328 246.0067 247.0849 248.1669 253.0864 256.1460 257.0603 260.1854 262.1348 263.2366 265.1505 269.1927 246 248 250 252 254 256 258 260 262 264 266 268 270 m/z FIGURE 3. %CV for sample replicat representative compounds. FIGURE 2. Data file review of 64 sample injections in a single data file using QuickCalc. Analyte Desipramine, Internal Standard Alprenolol. Display for RT at 11.59 min, T 0 replicate number 3 Clomipramine %CV Imipramine %CV Ery B T B Thermo Scientifi c Poster Note PN ASMS13_Th228_KMurphy_E 07/13S 3

Inter-Channel Variability The overall sample acquisition rate from injection to injection can be greatly improved when variability between LC columns is minimized and a given analyte can be injected across all four LC channels without a major variation in signal response due to column variability. To demonstrate the overall variability between the four LC channels used in the experiment, each sample was sequentially injected directly onto each of the four LC channels and the coefficient of variation calculated using the resulting area ratio values.(table 1.) The calculated %CV for the area ratio response was less than 15% for all samples, the %CV was no greater than 10% for 92% of the samples analyzed and the %CV was no greater than 5% for 73% of the samples analyzed. The variation in response between LC channels is illustrated for six representative compounds below.(figure 3.) Although twenty-two compounds were incubated in PPB assay, one compound did not provide sufficient signal response to accurately measured and has been excluded from all tables and figures. TABLE 1. %CV for sample replicate across each LC channel. Each value in the table represents the variation in response for a single sample injected to each of the four LC channels. All samples are 15% CV or less, 92% of samples 10% CV or less, and 73% of samples 5% CV or less. Rep 1 Rep 2 Rep 3 Rep 1 Rep 2 Rep 3 Rep 1 Rep 2 Rep 3 Acebutolol 5.1% 4.1% 2.2% 0.7% 3.4% 5.1% 2.5% 5.3% Buspirone 1.0% 1.8% 3.2% 1.0% 1.3% 2.3% 6.5% 3.2% 3.6% Carbamazepine 4.6% 8.4% 10.9% 6.4% 7.8% 5.1% 9.5% 5.1% Chlorpheniramine 5.0% 2.8% 6.2% 3.2% n/a 2.6% 3.1% 9.4% 4.5% Clomipramin 6.9% 3.2% 2.9% 4.6% 3.3% 4.9% 5.4% 4.2% 2.3% Clozapine 3.0% 0.9% 3.4% 2.9% 1.8% 2.3% 2.9% 3.7% 2.5% Desipramine 2.6% 4.2% 2.9% 6.2% 2.4% 2.9% 2.6% 3.3% 1.8% Diltiazem 1.2% 1.9% 2.2% 0.9% 2.7% 3.5% 3.6% 2.8% 3.0% Erythromycin 3.7% 2.6% 3.6% 3.6% 2.7% 1.3% 2.9% 1.7% 1.5% Fluphenazine 4.4% 6.8% 2.9% 6.9% 3.0% 6.4% 2.4% 3.4% Haloperidol 8.3% 5.9% 8.1% 4.3% 3.8% 3.4% 3.8% 4.7% 6.4% Imipramine 4.7% 1.3% 3.2% 3.5% 4.1% 1.7% 3.7% 3.0% 1.5% Metoprolol 3.8% 5.2% 4.5% 8.8% 0.9% 6.1% 2.9% 2.8% 4.7% Nefazadone 3.4% 1 11.4% 6.1% 6.7% 9.1% 4.4% 4.9% 5.1% Phentolamine 12.2% 15.4% 9.4% 1.2% 1.1% 2.7% 1.2% n/a Propranolol 3.4% 2.9% 1.4% 9.3% n/a 1.7% 0.6% 2.3% Retonavir 14.2% 13.4% n/a 15.2% 7.7% 12.8% 1 7.6% 13.0% Thioridazine 6.3% 2.2% 5.5% 4.1% 6.6% 4.9% 7.6% 8.6% Ticlopidine 4.9% 10.6% 7.1% 3.3% 4.8% 3.2% 1.4% 2.1% 2.8% Timolol 3.5% 8.2% 6.6% 5.3% 7.8% 3.7% 3.2% 3.8% Verapamil 5.1% 4.5% 3.0% 1.3% 4.5% 2.6% 2.1% 4.7% % Free 9 8 7 6 5 4 3 2 Ave A FIGURE 3. %CV for sample replicates injected across each LC channel for six representative compounds. Clomipramine %CV Erythromycin %CV Haloperidol %CV Imipramine %CV Timolol %CV Verapamil %CV Buffer Serum T0 4 A High-Resolution, Accurate-Mass Approach for Ultra-High Throughput Screening Plasma Protein

% Free Calculation Percent free or the unbound amount of compound in the protein binding assay was calculated by determining the ratio of signal response between the buffer and serum samples 1. The % Free was calculated using both signal response from a single channel as well as signal response from across all channels to determine the viability of the multi-channel analysis approach. The % Free for the single channel analysis was calculated using the buffer/serum signal response from the same LC channel for all injection replicates and the average %Free was reported for each compound. The % Free for the multi-channel analysis was calculated by mixing the signal response for buffer and serum within a sample replicate between each of the four LC channels. For example, buffer from channel 1 was compared to serum from channel 2 etc. The calculated %Free values for replicates were averaged and reported. The average % Free values for each compound were plotted in a bar chart to illustrate differences in the % Free values between single channel analysis, multi-channel analysis and results previously collected by LC-MS/MS (Figure 4). The average %Free values for both the single channel and multi-channel correlate well to each other as well as to the LC- MS/MS. Two compounds in the multi-channel analysis were observed with a difference of more than 25% from the LC-MS/MS analysis, Nefazadone and Clomipramine. FIGURE 5. Calibration curve of concentration point analyzed n channels. Y = (0.0031608 * X) + (-7.9885E- 04) 7.00000 Area Ratio 6.00000 5.00000 4.00000 3.00000 R^2 = 0.99876 2.00000 RSD, 1. 1.00000 RSD, 3.71 0.00000 RSD, 11.51-1.00000 0 500 FIGURE 4. Average % Free for individual compounds across each analysis type. Nineteen of twenty-one compounds analyzed using high resolution mass spectrometry demonstrate a % difference of less than 25% when compare to results collected using LCMS/MS % Free 9 8 7 6 5 4 3 2 Average % Free Variation AVG %Free Single Channel AVG %Free Multi Channel AVG %Free MS/MS Each compound analyzed in the PPB assay was evaluated in a concentration curve to evaluate overall sensitivity and linear dynamic range. All compounds were serially diluted using PPB matrix blank solution with concentrations ranging from 5 nm to 2000 nm. Each concentration point was injected across each LC channel to determine inter-channel variability. The calibration curves for all compounds were generated using a linear regression and 1/x 2 weighting. Individual calibration points exceeding a percent difference of more than 20% of the regression line fit were excluded from all calculations. Three compounds did not provide adequate signal response for accurate quantitation of the 5 nm concentration point. Although three compounds could not be measured at the 5 nm concentration point only one compound did not provide enough signal response to be measure in the binding assay. The majority of the compounds provided adequate sensitivity for all concentration points in the calibration curve, with all compounds providing a linear response across the concentration range. Additionally, all compounds and their replicate injections at each concentration point across each LC channel demonstrated RSD values of less than 15%. An example calibration curve for the multi-channel analysis is displayed for Desipramine (Figure 5). Thermo Scientifi c Poster Note PN ASMS13_Th228_KMurphy_E 07/13S 5

FIGURE 5. Calibration curve of Desipramine from 5nm to 2000 nm. Each concentration point analyzed n=4 with replicates injected across all four LC channels. Y = (0.0031608 * X) + (-7.9885E- 04) 7.00000 Area Ratio 6.00000 5.00000 4.00000 R^2 = 0.99876 Desipramine RSD, 1.72 3.00000 RSD, 2.57 2.00000 RSD, 1.83 1.00000 RSD, 3.71 0.00000 RSD, 11.51-1.00000 0 500 1000 1500 2000 2500 Concentration (ng/ml) Conclusion 85% of the compounds analyzed met the assay calibration curve LOQ of 5 nm and 95% provided adequate signal for assay reporting. 92% of the sample replicates injected across multi-channel system have a %CV of 10% or less. Negligible variation was observed for % Free values in the multi-channel analysis. Sample cycle time of 19 seconds per sample. High resolution full scan mass spectrometry coupled LC multiplexing using a single generic method provides suitable alternative to traditional LCMS/MS analysis References 1. Zhang J, Shou WZ, Vath M, Kieltyka K, Maloney J, Elvebak L, Stewart J, Herbst J, Weller HN., Rapid Commun Mass Spectrom. 2010 Dec 30;24(24):3593 Acknowledgements We would like to thank Dr. Jun Zhang and Jennifer Maloney of Bristol-Myers Squibb, Wallingford, CT for sample preparation and assistance with data processing. 6 A High-Resolution, Accurate-Mass Approach for Ultra-High Throughput Screening Plasma Protein

www.thermoscientific.com 2013 Thermo Fisher Scientific Inc. All rights reserved. ISO is a trademark of the International Standards Organization. GMSU Gubbs Mass Spec Utilities is a trademark of Gubbs Inc. Microsoft and Excel are registered trademarks of Microsoft Corporation. All other trademarks are the property of Thermo Fisher Scientific, Inc. and its subsidiaries. Specifications, terms and pricing are subject to change. Not all products are available in all countries. Please consult your local sales representative for details. Thermo Fisher Scientific, San Jose, CA USA is ISO 9001:2008 Certified. Africa-Other +27 11 570 1840 Australia +61 3 9757 4300 Austria +43 1 333 50 34 0 Belgium +32 53 73 42 41 Canada +1 800 530 8447 China +86 10 8419 3588 Denmark +45 70 23 62 60 Europe-Other +43 1 333 50 34 0 Finland/Norway/Sweden +46 8 556 468 00 France +33 1 60 92 48 00 Germany +49 6103 408 1014 India +91 22 6742 9434 Italy +39 02 950 591 Japan +81 45 453 9100 Latin America +1 561 688 8700 Middle East +43 1 333 50 34 0 Netherlands +31 76 579 55 55 New Zealand +64 9 980 6700 Russia/CIS +43 1 333 50 34 0 South Africa +27 11 570 1840 Spain +34 914 845 965 Switzerland +41 61 716 77 00 UK +44 1442 233555 USA +1 800 532 4752 ASMS13_Th228_KMurphy_E 07/13S