Anal. Chem. 2007, 79, 8010-8015
Flexible Automated Approach for Quantitative Liquid Handling of Complex Biological Samples Joe Palandra,† David Weller, Gary Hudson, Jeff Li, Sarah Osgood,‡ Emily Hudson,‡ Min Zhong,‡ Lisa Buchholz,‡ and Lucinda H. Cohen*
Bioanalytical Research, Department of Pharmacokinetics, Dynamics & Metabolism, Pfizer Global Research and Development, 2800 Plymouth Road, Ann Arbor, Michigan 48105
A fully automated protein precipitation technique for biological sample preparation has been developed for the quantitation of drugs in various biological matrixes. All liquid handling during sample preparation was automated using a Hamilton MicroLab Star Robotic workstation, which included the preparation of standards and controls from a Watson laboratory information management system generated work list, shaking of 96-well plates, and vacuum application. Processing time is less than 30 s per sample or ∼45 min per 96-well plate, which is then immediately ready for injection onto an LC-MS/MS system. An overview of the process workflow is discussed, including the software development. Validation data are also provided, including specific liquid class data as well as comparative data of automated vs manual preparation using both quality controls and actual sample data. The efficiencies gained from this automated approach are described. The advancement of combinatorial chemistry along with highthroughput screening have helped accelerate the discovery process for pharmaceutical compounds. This in turn has increased the demand for higher throughput bioanalysis. To that end, bioanalytical laboratories are continually looking for new strategies to improve sample preparation and analytical detection. Due to its rapid speed, specificity, and sensitivity, high-performance liquid chromatography coupled with tandem mass spectrometry (LCMS/MS) has become the bedrock of any modern day bioanalytical laboratory. With LC-MS/MS systems fully leveraged, the focus has now turned to the bioanalytical procedure required before samples are introduced into the mass spectrometer, i.e., sample preparation. Sample preparation has been and continues to be the critical step for fast, accurate, and reliable LC-MS/MS assays.1-5 Conventional biological matrix extraction methods such as solid* To whom correspondence should be addressed. Merck Research Laboratories, Mail Stop RY800-B201, P.O. Box 2000, Rahway, NJ 07065. E-mail:
[email protected]. † Current address: Pfizer Research and Development, Chesterfield, MO 63017. ‡ Permanent address: Pfizer Research and Development, Eastern Point Rd., Groton, CT 06340. (1) Peng, S. X.; Branch, T. M.; King, S. L. Anal. Chem. 2001, 73, 708-714. (2) Woolf, E. J.; Fu, I.; Matuszewski, B. K. J. Pharm. Biomed. Anal. 1998, 18, 347-357. (3) Henion, J.; Brewer, E.; Rule, G. Anal. Chem. 1998, 70, 650A-656A. (4) Hsieh, Y.; Bryant, M. S.; Brisson, J. M, Ng, K.; Korfmacher, WA. J. Chromatogr., B 2002, 767, 353-362.
8010 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007
phase extraction, liquid-liquid extraction, and protein precipitation (PPT) are both time-consuming and labor-intensive. Due to its simplicity and universality, protein precipitation continues to be the main methodology employed for biological sample preparation, especially in the drug discovery arena.6-9 However, all three major sample preparation methods remain tedious processes and, consequently, bottlenecks limiting the throughput of current fast LC-MS/MS analysis. Current trends in sample preparation have focused on the use of 96-well technology as well as automated sample preparation via 96-well parallel processing.10-16 While the introduction of 96-well liquid handlers, such as the Tomtec Quadra, has helped overcome some of these bottlenecks, a completely automated sample preparation solution remains elusive.17-20 Current 96-well automation is composed solely of the liquid-transfer steps of sample preparation, but not the preparation of standards and quality control samples or sample dilutions. In addition, current automation is incapable of interfacing with either laboratory information management systems (LIMS) such as Watson or mass spectrometric data acquisition software. This report demonstrates the utility of a commercially available (5) Beaudry, F; Le Blanc, Y. J. C.; Coutu, M; Brown, N. K. Rapid Commun. Mass Spectrom. 1998, 12, 1216-1222. (6) Yang, L.; Wu, N.; Rudewicz, P. J. J. Chromatogr., A 2001, 926, 43-55. (7) Watt, P.; Morrison, D.; Locker, K. L.; Evans, D. C. Anal. Chem. 2000, 72, 979. (8) Ramos, L.; Bakhtiar, R.; Tse, F. L. S. Rapid Commun. Mass Spectrom. 2000, 14, 740. (9) Peng, S. X.; King, S. L.; Bornes, D. M.; Foltz, D. J.; Baker, T. R.; Natchus, M. G. Anal. Chem. 2000, 72, 1913. (10) Grant, R. P.; Cameron, C.; Mackenzie-McMurter, ?? Rapid Commun. Mass Spectrom. 2002, 16, 1785-1792. (11) Kaye, B.; Herron, W. J.; Macrae, P. V.; Robinson, S. R.; Stopher, D. A.; Venn, R.; Wild, W. Anal. Chem. 1996, 68, 1658. (12) Allanson, J. P.; Briddlecombe, R. A.; Jones, A. E.; Pleasance, S. Rapid Commun. Mass Spectrom. 1996, 10, 811. (13) Zweigenbaum, J.; Henion, J.; Steinborner, S.; Wachs, T. Anal. Chem. 1999, 71, 2294. (14) Simpson, H; Berthemy, A; Buhrman, D; Burton, R; Newton, J; Kealy, M.; Wells, D.; Wu, D. Rapid Commun. Mass Spectrom. 1998, 12, 75-82. (15) Harrison, A. C.; Walker, D. K. J. Pharm. Biomed. Anal. 1998, 16, 777783. (16) Hempenius, J.; Wieling, J.; Brakenhoff, J. P. G.; Maris, F. A; Jonkman, J. H. G. J. Chromatogr., B 1998, 714, 361-368. (17) Xu, N.; Kim, G. E.; Hope, G.; Azza, W.; Brendan, S.; A.; Min, S. C.; Tawakol, A. J. Pharm. Biomed. Anal. 2004, 16, 189-195. (18) Ross, R. T. LC-GC 1999, S4-S6. (19) Weng, N.; Bu, H.; Chen, Y.; Shou, W. Z.; Jiang, X.; Halls, T. D. J. J. Pharm. Biomed. Anal. 2002, 28, 1115-1126. (20) Song, Q.; Heiko, J.; Tang, Y.; Li, A. C; Addison, T.; McCort-Tipton, M.; Beato, B.; Naidong, W. J. Chromatogr., B 2005, 814, 105-114. 10.1021/ac070618s CCC: $37.00
© 2007 American Chemical Society Published on Web 09/25/2007
Figure 1. Hamilton MicrolabStar Deck layout.
Figure 2. Hamilton interface program (HIP).
advanced robotic workstation, the Hamilton MicroLab Star, and the resultant fully automated 96-well PPT methodology for biological sample preparation. In addition, we have developed a Visual Basic software tool, which allows the scientists to design the liquid handling process with a range of flexible options. This fully automated method avoids some of the manual pitfalls in semiautomated approaches such as preparation of standards or controls, as well as the vortexing and centrifugation of samples. Furthermore, we will describe the implementation of this process and resultant routine utilization. EXPERIMENTAL MATERIAL AND METHODS Automated PPT Procedure. The Hamilton MicroLab Star workstation is equipped with a 96-core pipetting head as well as a 8-channel pipetting head using 300-µL tips (Figure 1). Using a
Watson LIMS-generated worklist along with the Hamilton Interface Program (HIP), a Visual Basic interface shown in Figure 2, the user determines the number of standards, quality controls (QC), concentration of each standard and QC, sample and protein precipitation solution volumes, dilution factors for each sample, and sample positions on the final completed plate. The totally automated program features aliquot volumes ranging from 5 to 300 µL, up to 60-fold sample dilutions, an automated prompt to the scientist for the volume of blank biological matrix required for the assay, and sample preparation times of less than 30 s per sample. Total capacity per assay is 4 96-well plates, or 392 samples, which includes standards and quality controls. In Figure 3 , we describe the process workflow using the Hamilton. Study Submission. The in-house custom-designed HIP utilizes a Visual Basic interface and represents the first step toward Analytical Chemistry, Vol. 79, No. 21, November 1, 2007
8011
Figure 3. Protein precipitation workflow using the Hamilton.
processing samples on the Hamilton. A screen capture of the HIP is shown in Figure 2. The contents of the submission include various analytical and study parameters such as sample and precipitation solution volume, Watson LIMS study identification number, Excel file location created from Watson LIMS, and any special comments. Once submitted, an accession number is automatically generated containing all pertinent information required for the Hamilton. The accession number is used for tracking purposes during the sample processing and metrics generation. The current program allows the laboratory technician operating the robotics a unique view for study tracking. The Analyst view is intended to use with the scientists submitting samples for processing. The Source File Database options allow the scientists to define what LIMS system generated the original sequence file, i.e., Watson for in vivo studies and Galileo for in vitro studies. Processing options include simply aliquotting samples and dilutions, preparing samples and dilutions with the addition of internal standard, and preparation of samples through the entire process to LC/MS/MS supernatant. If multiple studies are run for the same analyte, the scientist may recall previous studies from the archive and submit without extensive data entry. The HIP also permits scientists to recall studies once submitted, make revisions as necessary, and then resubmit the revised study to the processing queue. File Handling on Hamilton: Sequence Creation. Once the study has been submitted, that information is brought into the Hamilton by entering the accession number generated by the HIP. The program begins by extracting necessary information from a Watson-created Excel file and reorganizing the information into a text delimited file utilized by the Hamilton. It then creates sequence files from the text file. The sequence files determine a list of chronological pipetting steps performed on the Hamilton. Tip Sparing Optimization for CORE-96. The PPT solution is added to the target plate through the CORE-96 head. This component can contain up to 96 tips simultaneously and completes the pipetting process for an entire plate in a matter of seconds. Because frequently not all of the 96 wells on the target plate are used, an algorithm was designed to remove extra tips based on the layout of the target plate. The extra tips are stored in an empty rack for later use. On average, more than 48 tips can be saved per experiment. Tip optimization is conducted in no more that a few minutes. Preparation of Standards and Quality Controls. The highest standard is first prepared on the Hamilton by aliquoting an analyte stock solution into blank matrix, with the program ensuring adherence to a ratio of at least 95% matrix content. The 8012
Analytical Chemistry, Vol. 79, No. 21, November 1, 2007
remaining standards are then created via serial dilution. This same procedure is used for the creation of QCs as well. Aliquoting of Samples to a 96-Well Plate. Using the sequence files created earlier, samples are aliquoted to an intermediate plate based on the position layout on the final target plate(s). Any sample dilutions required are also performed during this step. Sample dilutions are particularly important for pharmacokinetic studies spanning a 3-5 order of magnitude concentration range. Transfer of PPT Solution and Biological Samples to PPT Plate. Using the equipped CORE-96 head, the Hamilton next transfers the PPT solution containing internal standard (IS) to the Sirocco PPT filtration plate. The same tips are then used to aspirate and dispense biological samples from the intermediate holding plate to the PPT plate. Shaking and Elution from PPT Plate to Clean 96-Well Plate. Utilizing the shaker unit on the Hamilton deck, the PPT plate is first sealed with a Beckman Coulter aluminum foil lid and then shaken for 60 s. Once the shaking is complete, the PPT plate is then transferred to the automated vacuum system (AVS) on the Hamilton using the i-swap arm. Once correctly positioned, the vacuum system begins the 4-min vacuum/elution process into an empty 96-well plate. After vacuum application, the plate can be analyzed by LC-MS/MS. Reagents. Proprietary test compounds were obtained from Pfizer Global Research and Development (Ann Arbor, MI). Ammonium formate, formic acid, and DMSO were obtained from Sigma (St. Louis, MO). HPLC grade acetonitrile, 2-propranol, and water were obtained from EM Science (Gibbstown, NJ). Blank Sprague-Dawley rat (with EDTA as an anticoagulant) plasma was obtained from Pfizer Global Research and Development (Ann Arbor, MI). Sirocco plates were obtained from Waters Corp. (Millford, MA). LC-MS/MS Methodology. LC-MS/MS analysis was carried out using a high-performance liquid chromatography system consisting of a Shimadzu binary pump with CTC PAL autosampler interfaced to an API 4000 Sciex triple-quadrupole tandem mass spectrometer (Applied Biosystems, Foster City, CA). The analyte was separated on a Hypersil Gold column (2.1 × 50 mm, Thermo Electron, Waltham, MA). The mobile phase consisted of solvent A (5 mM ammonium formate in water) and solvent B (acetonitrile). The gradient was as follows: solvent B was held at 10% for 0.3 min, linearly ramped from 10 to 75% in 1.7 min, ramped from 75 to 95% in 0.7 min, and then immediately returned to 10% for reequilibration. Total run time was 3 min with a flow rate of 0.30 mL/min. The mass spectrometer was operated in positive ion mode with electrospray ionization using multiple reaction
Table 1. Validation of Hamilton Liquid Classes (a) rat plasma weight (g) replicate
5 µL
100 µL
(b) DMSO weight (g) 300 µL
replicate
5 µL
100 µL
300 µL
1 2 3 4 5 6 7 8
0.0048 0.0049 0.0050 0.0050 0.0052 0.0051 0.0052 0.0050
0.1006 0.1001 0.1000 0.1003 0.0996 0.1000 0.1003 0.0998
0.3004 0.3003 0.3006 0.3008 0.2997 0.3002 0.3004 0.3009
1 2 3 4 5 6 7 8
0.0050 0.0049 0.0049 0.0050 0.0051 0.0051 0.0051 0.0049
0.0997 0.0996 0.0995 0.0997 0.0995 0.0996 0.1000 0.0999
0.3015 0.3014 0.3015 0.3009 0.3011 0.3010 0.3015 0.3002
mean Std Dev precision (%) accuracy (%)
0.0050 1.39 × 10-4 2.76 0.50
0.1001 3.14 × 10-4 0.31 0.09
0.3004 3.76 × 10-4 0.13 0.14
mean Std Dev precision (%) accuracy (%)
0.0050 9.26 × 10-5 1.85 0.00
0.0997 1.81 × 10-4 0.18 -0.31
0.3011 4.50 × 10-4 0.15 0.38
(c) 2-propanol/H20 weight (g) replicate
5 µL
100 µL
(d) acetonitrile weight (g)
300 µL
replicate
5 µL
100 µL
300 µL
1 2 3 4 5 6 7 8
0.0050 0.0053 0.0048 0.0056 0.0052 0.0049 0.0051 0.0052
0.1005 0.0988 0.1001 0.1000 0.1001 0.0999 0.1000 0.0992
0.2991 0.2999 0.3000 0.3003 0.2993 0.2997 0.3008 0.3006
1 2 3 4 5 6 7 8
0.0048 0.0050 0.0053 0.0054 0.0051 0.0049 0.0053 0.0054
0.0999 0.1008 0.1003 0.1005 0.1005 0.1000 0.1009 0.1005
0.3005 0.3005 0.3001 0.3009 0.2994 0.3003 0.3004 0.3003
mean Std Dev precision (%) accuracy (%)
0.0051 2.50 × 10-4 4.87 2.75
0.0998 5.50 × 10-4 0.55 -0.18
0.3000 5.95 × 10-4 0.20 -0.01
mean Std Dev precision (%) accuracy (%)
0.0052 2.33 × 10-4 4.52 3.00
0.1004 3.49 × 10-4 0.35 0.42
0.3003 4.31 × 10-4 0.14 0.10
Table 2. Quality Control Comparison theoretical concentration 2 ng/mL replicate number 1 2 3 4 5 6 mean Std Dev precision (%) accuracy (%)
Hamilton
20 ng/mL manual
1.81 1.82 1.88 1.82 1.77 1.76
1.73 1.75 1.92 1.90 1.74 1.89
1.81 0.04 2.37 -9.50
1.82 0.09 4.95 -8.92
200 ng/mL
Hamilton
manual
18.6 18.2 19.3 20.3 19.9 20.6
20.5 21.7 19.9 20.3 21.2 20.1
175 174 175 180 180 187
187 199 195 193 188 187
1750 1760 1800 1810 1810 1840
1830 1890 1800 1930 1880 1910
19.5 0.95 4.90 -2.58
20.6 0.69 3.37 3.08
179 4.93 2.76 -10.75
192 4.97 2.60 -4.25
1795 33.91 1.89 -10.25
1873 49.26 2.63 -6.33
monitoring. All raw data were processed using Analyst Software, version 1.4 (Applied Biosystems/MDS Sciex Inc., Ontario, Canada). RESULTS AND DISCUSSION Validation of Liquid Classes. The highly flexible Hamilton robotics system is capable of handling multiple liquid class types within a given method. This feature allows accurate pipetting of a variety of solvents or matrixes by assigning a specific liquid class to an individual solvent throughout the method. However, in order to ensure accurate pipetting, each liquid class must be independently calibrated, typically on a quarterly basis. This is accomplished using the eight-channel head and a Mettler balance, which is incorporated into the deck layout of the Hamilton, shown
Hamilton
manual
2000 ng/mL Hamilton
manual
in Figure 4. Several liquid classes required calibration and validation, including a plasma liquid class for analysis of biological matrix, an organic liquid class for transfer of stock solutions, and finally a tissue liquid class, a mixture of tissue suspended in 1:1 2-propanol/water. The validation process for each liquid class was conducted by measuring the weight of eight replicates, one replicate per channel on the unit across a pipetting range of 5300 µL. Volumes were calculated using the density of each liquid class. Results of the liquid class validation are shown in Table 1. For volumes of 5, 100, and 300 µL, the volume measurements shown were calculated from the weight values divided by the density of Analytical Chemistry, Vol. 79, No. 21, November 1, 2007
8013
Figure 4. Hamilton MicrolabStar Deck layout with integrated balance.
Figure 5. Plasma sample data comparison. Manual vs Hamilton. Slope, 1.03; intercept, -0.05.
the given liquid class. All four liquid classes including plasma, DMSO, acetonitrile, and the 2-propanol/water mixture showed excellent precision and accuracy within 5% at the 5-µL volume and within 0.5% at 100- and 300-µL volumes. Comparison for Quality Controls and Study Samples. Once the liquid classes were verified and tested, the next step of the validation process involved direct comparison of quality control samples prepared manually versus the Hamilton automation. QCs were prepared at four concentration levels: 2, 20, 200, and 2000 ng/mL in Sprague-Dawley rat EDTA plasma used with a proprietary test compound. Manual and Hamilton prepared quality controls were all quantitated against a Hamilton prepared standard curve. Several other compounds’ study samples were prepared and compared using manual and Hamilton preparation with similar correlations (results not shown). The quantitative results for quality controls are shown in Table 2. Again, excellent precision (within 5%) and accuracy (within 11%) were observed at all four concentration levels. Figure 5 displays the correlation between Hamilton and manual preparation for plasma study samples. In addition, incurred brain tissue samples, which were prepared manually as well as on the Hamilton, were quantitated against a standard curve prepared on the Hamilton only, with the results shown in Figure 6. Both comparisons of 8014 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007
Figure 6. Tissue sample data comparison. Manual vs Hamilton. Slope, 0.95; intercept, -0.005.
manual and automated preparation of plasma and tissue study samples over 3 orders of magnitude concentration range for plasma and 2.5-order concentration range for tissue exhibited a correlation coefficient (r2) of 0.999. CONCLUSIONS Through use of in-house custom-designed software (HIP) and the Hamilton MicroLab Star liquid handling workstation, protein precipitation sample preparation has been successfully automated and implemented for several biological matrix samples. This in turn has greatly reduced the intensive labor as well as the possibilities of systematic error associated with the manual volumetric transfers. Significant advantages in terms of efficiency and throughput have now been realized by this automated process while maintaining the integrity of the precision and accuracy. Over 1200 studies and 61 000 samples have been prepared in our laboratory since implementation. The original goal to automate greater than 80% of the sample workload using this approach was not only met but exceeded. Furthermore, the open configurability of both the Hamilton hardware and software allows for the continual refinement, enhancements, or additions to our program, which in turn enables us to keep pace with the ever-
changing landscape and external demands that drive and shape our future.
discussion. The contributions of scientists working in the Discovery Bioanalytical Research group are also gratefully acknowledged.
ACKNOWLEDGMENT
Received for review March 28, 2007. Accepted July 25, 2007.
The authors thank Steve Michael, Cathy Knupp, Chris Holliman, Doug Fast, and James Yergey for helpful review and
AC070618S
Analytical Chemistry, Vol. 79, No. 21, November 1, 2007
8015
