http://jla.sagepub.com/Journal of Laboratory Automation

http://jla.sagepub.com/content/17/3/211The online version of this article can be found at:

 DOI: 10.1177/2211068211429775

2012 17: 211 originally published online 24 January 2012Journal of Laboratory AutomationHao Jiang, Zheng Ouyang, Jianing Zeng, Long Yuan, Naiyu Zheng, Mohammed Jemal and Mark E. Arnold

and Dilution to Benefit the Regulated BioanalysisA User-Friendly Robotic Sample Preparation Program for Fully Automated Biological Sample Pipetting

  

Published by:

http://www.sagepublications.com

On behalf of: 

  Society for Laboratory Automation and Screening

can be found at:Journal of Laboratory AutomationAdditional services and information for    

  http://jla.sagepub.com/cgi/alertsEmail Alerts:

 

http://jla.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

What is This? 

- Jan 24, 2012OnlineFirst Version of Record  

- May 11, 2012Version of Record >>

by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from by guest on October 11, 2013jla.sagepub.comDownloaded from

Journal of Laboratory Automation17(3) 211 –221© 2012 Society for Laboratory Automation and ScreeningDOI: 10.1177/2211068211429775http://jala.sagepub.com

Introduction

Bioanalytical sample pipetting and dilution are a bottleneck in bioanalytical process due to the complexities in sample preparation. At least four types of samples need to be included in an analytical run, namely, blanks, standards, quality controls (QCs), and study samples.1 The sequence and replicates of these samples in an analytical run are dif-ferent run to run. When using 96-well plates for sample processing, the locations of blanks, standards, QCs, and study samples on the plates vary from run to run. Study samples and QCs are required to be interspersed in the 96-well plate and are bracketed by standards placed at the front and back of the run. In addition, when the concentra-tions of samples are beyond standard curve ranges, study samples need to be diluted up front before being transferred to 96-well plates for further sample clean up (liquid-liquid extraction, solid-phase extraction, or protein precipitation,

etc.). Sometimes different dilution factors (DF) should apply to different bioanalytical samples in a run (e.g., phar-macokinetic samples), which make the process even more complicated. If an internal standard is needed for an assay, the internal standard should be added to all standards, QCs, and study samples except for blanks to which a solvent is added. The overall process in preparing samples for an ana-lytical run (less than 96 samples × two plates) is very labor intensive and extremely time-consuming if the samples are

429775 JLAXXX10.1177/2211068211429775Jiang et al.Journal of Laboratory Automation

1Bioanalytical Sciences Department, Bristol-Myers Squibb Company, Princeton, New Jersey, USA2Bioanalytical & Discovery Analytical Sciences Department, Bristol-Myers Squibb Company, Princeton, New Jersey, USA

Received July 15, 2011.

Corresponding Author:Hao Jiang or Jianing Zeng, Bristol-Myers Squibb, Bioanalytical Sciences, Rt. 206 & Province Line Rd, Princeton, NJ 08543 Email: hao.jiang@bms.com or jianing.zeng@bms.com

A User-Friendly Robotic Sample Preparation Program for Fully Automated Biological Sample Pipetting and Dilution to Benefit the Regulated Bioanalysis

Hao Jiang1, Zheng Ouyang2, Jianing Zeng1, Long Yuan1, Naiyu Zheng1, Mohammed Jemal2, and Mark E. Arnold1

Abstract

Biological sample dilution is a rate-limiting step in bioanalytical sample preparation when the concentrations of samples are beyond standard curve ranges, especially when multiple dilution factors are needed in an analytical run. We have developed and validated a Microsoft Excel–based robotic sample preparation program (RSPP) that automatically transforms Watson worklist sample information (identification, sequence and dilution factor) to comma-separated value (CSV) files. The Freedom EVO liquid handler software imports and transforms the CSV files to executable worklists (.gwl files), allowing the robot to perform sample dilutions at variable dilution factors. The dynamic dilution range is 1- to 1000-fold and divided into three dilution steps: 1- to 10-, 11- to 100-, and 101- to 1000-fold. The whole process, including pipetting samples, diluting samples, and adding internal standard(s), is accomplished within 1 h for two racks of samples (96 samples/rack). This platform also supports online sample extraction (liquid-liquid extraction, solid-phase extraction, protein precipitation, etc.) using 96 multichannel arms. This fully automated and validated sample dilution and preparation process has been applied to several drug development programs. The results demonstrate that application of the RSPP for fully automated sample processing is efficient and rugged. The RSPP not only saved more than 50% of the time in sample pipetting and dilution but also reduced human errors. The generated bioanalytical data are accurate and precise; therefore, this application can be used in regulated bioanalysis.

Keywords

robotic sample preparation program, automated, sample dilution and preparation

212 Journal of Laboratory Automation 17(3)

pipetted and diluted manually. Generally, it takes about 2 to 4 h to complete this step of sample preparation depending on sample quantity, number of samples needed to be diluted, and steps of the dilution process. Besides, it is very hard to control human errors during the long and boring process in sample pipetting and dilution.

Current available liquid-handling robots such as Tecan Freedom EVO, Tecan Genesis RSP, Hamilton STAR line, and JANUS Automated Workstation have been widely used for liquid pipetting and sample preparation in laboratories. Some studies2–11 reported that automations had been applied for pipetting samples and solvents to facilitate sample prep-aration and extraction, but samples had to be diluted manu-ally before automatic sample extraction, or all the samples should have a consistent DF and preassigned locations on 96-well plates; otherwise, it is not possible to accommodate the process to different analytical runs with different sam-ple numbers, DF, and varied sample locations on 96-well plates. Although there is a worklist function available in some robots, it is very time-consuming and error prone to program a robotic script for each sample’s dilution in every analytical run. Some robots support worklist functions to allow bioanalyst’s entry of sample information (e.g., source position, destination position, volume, and liquid class) to a comma-separated values (CSV) worklist that is executable by robotic scripts, but it is still not practical or efficient to edit the CSV worklist for every analytical run. A Microsoft Visual C++ program12 has been developed recently that is capable of automatically generating run-specific robot scripts for fully automated sample pipetting, dilution, and downstream sample extraction. This program is the first successful program for automated sample dilution, which makes the sample preparation process fully automated. However, a unique script has to be generated for every individual run, and the complexity of the generated script with several hundred robot commands makes it impossi-ble to effectively troubleshoot during sample pipetting and dilution.

We have developed and validated a Microsoft Excel–based robotic sample preparation program (RSPP). The RSPP contains a dilution calculation spreadsheet and a Visual Basic for Applications (VBA) macro to automati-cally transform sample information from a Watson worklist to executable CSV worklists that contain each sample’s dilution scheme, source well positions, destination well posi-tions, and liquid classes. A preprogrammed robotic script that operates within the Freedom EVO software (EVOware) with several executable commands transforms the CSV files to executable worklists (.gwl files) and then executes the sample pipetting and dilution. It is simple and easy for troubleshooting errors in sample pipetting and dilution because EVOware records each step of liquid pipetting and is able to continue pipetting from an unexpected stop in a run. The RSPP program can be integrated within other robot

commands to facilitate a fully automated sample prepara-tion and extraction.

ExperimentalMaterials

Drug-free K2EDTA plasma specimens were purchased from

Bioreclamation Inc. (Hicksville, NY). Deionized water was prepared from an in-house Barnstead Nanopure Diamond system (Dubuque, IA). All solvents for sample extraction and LC-MS/MS were obtained from different vendors at the highest grade available. Reference materials of BMS-650032 and isotopically labeled BMS-650032 (internal standard) were obtained from Bristol-Myers Squibb Company (New Brunswick, NJ). Dilution plates (96-square well, 2.0 mL) were obtained from VWR Co. and 96-microtube plates (1.1 mL) from National Scientific Supply Co. (Claremont, CA).

InstrumentationThe liquid-handling robot used in this study was a Tecan Freedom EVO150 liquid-handling unit equipped with an eight-channel liquid handler (LiHa) arm and a 96-multichannel arm (96-MCA). The system was controlled by Freedom EVO software, which had been validated for use in regulated bioanalysis by the department staff and the inter-nal Informatics Quality Assurance Group. The Freedom EVO worktable was loaded with one disposable tip carrier (DiTi 200/1000, 2 Pos.), one solvent reservoir carrier (Trough 100 mL, 3 Pos.), three 96-well plate carriers (96-Well, 3 Pos.), two 96-sample racks (16 × 6), and one 96-MCA carrier (see Fig. 1). The following labware was used for sample dilution and liquid-liquid extraction: one tray of 200-µL DiTi tips, one tray of 1000-µL DiTi tips, three 100-mL troughs (containing the diluent, the internal standard, and the blank solvent, respectively), three 96-square well plates (dilution plates, labeled as A, B, and C), one 96-microtube plate (a final sample plate, labeled as D) for each rack of samples, and three boxes of 96-MCA tips. A balance (Mettler AG285) communicating with the workstation computer was used to weigh and record weights of the liquid pipetted by the Freedom EVO.

RSPPThe RSPP is a Microsoft Excel–based application that consists of a calculation spreadsheet and a VBA macro (Fig. 2A, B). The calculation spreadsheet contains an IF-THEN function formula to determine six dilution parameters (source rack label, source position, destination rack label, destination position, volume, and liquid class) for samples and the dilu-ent at each dilution step. The calculation is based on DF

Jiang et al. 213

from Watson worklist and preset parameters (final diluted sample volume, internal standard volume, and liquid classes for pipetting samples; internal standards; diluents; and diluted samples). The macro was programmed with the VBA, which has the following functions:

1. copy and paste sample identification and dilution factors (DF) from a Watson worklist (an Excel file that is exported from Watson) to the calculation spreadsheet,

2. clean and reformat the resulting spreadsheet with dilution parameters, and

3. generate eight CSV files for sample pipetting, diluting, and adding internal standard (IS)

The dilution range is 1- to 1000-fold, which is divided into three dilution steps: 1- to 10-, 11-to 100-, and 101- to 1000-fold. An example of the dilution scheme for a final volume of 50 µL samples is shown in Table 1. For samples with DF ≤2, each sample (blanks, standards, QCs, or study samples) is pipetted from a sample vial in a sample rack (sample rack 1 or sample rack 2; Fig. 1B) to a final 96-microtube plate D (1D or 2D; Fig. 1B), and then an appropriate volume of diluent (blank plasma) is added fol-lowed by automated pipette mixing. Similarly, for samples with 2 < DF ≤ 10, each sample is pipetted to a plate C for dilution; for samples with 10 < DF ≤ 100, each sample is pipetted to a plate A for step 1 dilution and after the first step of dilution to a plate C for step 2 dilution; for samples

A

DiTi 200

Wash Sta�on

Trough 100 mL

Plasma

Dilu�on 96-Well Plate

Plas

ma

Balance

DiTi 1000

Waste

Wat

er

66)

66)

DiTi 200

Wash Sta�on

Trough 100 mL

1A 2A 1D

Dilu�on 96-Well Plate

Final 96-Microtube Plate

asm

a

B

Sam

ple

Rack

1 (1

6

Sam

ple

Rack

2 (1

6DiTi 200

DiTi 1000

Waste

1A

1B

1C

2A

2B

2C

1D

2D

Pla

ISSo

lven

t

S S96-MCA Carrier

1D’ 1D

96-Well LLE Collec�on Plate

Final 96-Microtube PlateC

1D

2D’

Extrac�onSolvent

1D

2D

BufferRecon.Solu�on

Figure 1. Freedom EVO worktable configuration. (A) System gravimetric verification. (B) Sample diliution. (C) Liquid-liquid extraction.

214 Journal of Laboratory Automation 17(3)

Figure 2. The interface of the robotic sample preparation program. (A) Calculation spreadsheet. (B) Visual Basic for Applications macro.

with 100 < DF ≤ 1000, each sample is pipetted to a plate A for step 1 dilution, to a plate B for step 2 dilution, and to a plate C for step 3 dilution. After the last dilution step for

each sample, the final diluted sample is pipetted to the plate D. IS solution or a blank solvent is added to each sample according to the specified sample types. A minimum sample

Jiang et al. 215

Table 1. Representative Dilution Schemes for Diluted Samples with a Final Volume of 50 µL

Dilution Factor

Step 1 (DF 1-10) Step 2 (DF 11-100) Step 3 (DF 101-1000)Final

Pipetting IS Pipetting

Worklist 2 Worklist 3 Worklist 4 Worklist 5 Worklist 6 Worklist 7 Worklist 8 Worklist 1

Study Sample (µL)

Diluent (µL)

Diluted Sample (µL)

Diluent (µL)

Diluted Sample (µL)

Diluent (µL)

Diluted Sample (µL)

Diluent (µL)

1 50 0 0 0 0 0 0 502 25 25 0 0 0 0 0 505 25 100 50 0 0 0 0 5010 25 225 50 0 0 0 0 5020 25 225 50 50 50 0 0 5050 25 225 50 200 50 0 0 50100 25 225 50 450 50 0 0 50200 25 225 50 450 50 50 50 50500 25 225 50 450 50 200 50 501000 25 225 50 450 50 450 50 50

Figure 3. Representative comma-separated value (CSV) worklists generated by the robotic sample preparation program macro. When transferring CSV worklists to .executable GWL files, EVOware reads columns A to F (representing the dilution parameters of source rack label, source position, destination rack label, destination position, volume, and liquid class, respectively) of worklists 1, columns G to L of worklists 2, columns M to R of worklists 3, columns S to X of worklists 4, columns Y to AD of worklists 5, columns AE to AJ of worklists 6, columns AK to AP of worklists 7, and columns AQ to AV of worklists 8.

volume of 25 µL is required for step 1 dilution to ensure pipetting accuracy, and minimum 50 µL of samples is required for step 2 and 3 dilutions. Six dilution parameters for each sample at each step dilution are located in the spe-cific columns in the CSV worklists (Fig. 3).

The RSPP is user friendly with a simple interface. The VBA macro is executed by clicking the hyperlink of the RSPP macro (the instruction panel; Fig. 2A); a pop-up

window then guides users to select an exported Watson worklists. Lastly, a second pop-up window directs users to generate eight CSV files to a user-defined project folder. The program is closed automatically after successfully gen-erating CSV worklists. For multiple analytical runs, these eight CSV worklists will be updated for each analytical run, and a run-specific backup worklist (with a time stamp on the file name) will be generated.

216 Journal of Laboratory Automation 17(3)

Verification of Freedom EVO System Performance

Deionized water was used as the standard solvent for sys-tem verification. The manufacturer’s default “water” liq-uid class was used for water aspirating and dispensing. Six aliquots of 25 µL and 900 µL of water, respectively, were aspirated and dispensed onto the online balance by each channel pipettor (Fig. 1A). The weight of each aliquot was recorded by EVOware. The accuracy (%Dev) and preci-sion (%CV) for each pipettor at the volumes of 25 and 900 µL should be within 5%, based on the internal standard operation procedure.

Verification of Plasma Pipetting Accuracy and PrecisionThe manufacturer’s default “serum” liquid class was used as a template for creating a customized liquid class. A volume of 25, 50, 100, 200, 225, and 450 µL plasma (rat, dog, mon-key, rabbit, mouse, or human plasma) was pipetted by Freedom EVO onto the balance in replicates of six, respec-tively. The same volumes for these plasmas were manually pipetted using calibrated handheld pipettors onto the balance in replicates of six. The mean of six aliquots of rat plasma at each volume was used as the nominal value for calcula-tions. The mean %Dev of the weights from Freedom EVO pipetting against the nominal values and %CV at each vol-ume were calculated to determine the pipetting accuracy and precision.

The impact of various sample volumes in tubes on pipetting accuracy and precision was also evaluated for two types of sample tubes. The BMS-customized 5 mL tube has a built-in insert with a narrower inner diame-ter (˜5 mm). The Corning cryogenic tube has a bigger inner diameter (˜10 mm, catalog no. 430491). Six rep-licate 25 µL volumes of plasma were pipetted from each type of tube containing 50, 75, 100, and 200 µL of rat plasma, respectively, and weighed. The %Dev and %CV were calculated to evaluate the accuracy and precision.

BMS-650032 Standards and QC Sample PreparationStandards (5.00, 10.0, 20.0, 50.0, 100, 500, 1000, and 2000 ng/mL) and QCs (5.00, 15.0, 25, 125, 1000, 1600 and 50 000 ng/mL) were prepared using calibrated handheld pipettors in drug-free dog plasma by spiking BMS-650032 stock solutions (1.0 mg/mL in methanol from two separate weighings) and serial dilutions with dog plasma. All QCs were aliquotted into polypropyl-ene tubes and stored at approximately –20 °C. The standards were freshly prepared on the day of sample preparation.

Plasma Sample Collection

Four male dogs were orally dosed with 400 mg of BMS-650032 on days 1 and 8. Whole-blood (˜1 mL) samples were collected into tubes containing K

2EDTA at time

points 0, 1, 2, 3, 4, 8, and 24 h after dosing. The plasma was separated after centrifugation and then transferred to poly-propylene tubes for storage at –20 °C.

Plasma Sample Dilution Using the RSPP and Freedom EVOStudy samples and standards/QCs were placed in sample racks according to the sequence in the Watson worklist. Two sets of the standards bracketed all study samples and QCs in a run. QCs were interspersed among study samples. QCs at the concentration of 25, 1600, and 50 000 ng/mL (defined as RSPP QCs) were diluted at dilution factors of 1, 2, 5, 10, 20, and 200 in replicates of six to test dilution accu-racy and precision. For each rack of samples, three 96-well dilution plates (1A or 2A sample dilution plate, 1B or 2B sample dilution plate, and 1C or 2C sample dilution plate) and one 96-microtube plate (1D or 2D sample dilution plate) were placed on the plate carrier as shown in Figure 1B. Appropriate volumes of the diluent (drug-free dog plasma), the IS solution (100 ng/mL of D

9-BMS-650032 in 50%

acetonitrile/water), and blank solvent (50% acetonitrile/water) were poured into the 100 mL trough reservoirs. RSPP-generated CSV worklist were imported and trans-formed into .gwl files by “worklist import” command in Freedom EVO scripts. The .gwl files were then executed by the user selecting the “worklist execution” command to pipette and dilute samples. A minimum volume of 25 µL of a plasma sample was pipetted for the dilutions. The maxi-mum volume in a well was 500 µL. Five cycles of aspirat-ing and dispensing after pipetting the diluent (plasma) were automatically conducted. There was also an option to use manual mixing by modifying the liquid class by not activat-ing the “mix after dispense” function. Several pauses were inserted in between every dilution step to allow an optional manual mixing. Once the final volume of 50 µL of each blank, standard, QC, and study sample were pipetted into the final sample plate, 50 µL of the IS solution was auto-matically pipetted to each sample.

Liquid-Liquid Extraction and LC-MS/MS DetectionAfter completion of sample pipetting and dilution, a semi-automated liquid-liquid extraction was conducted by the 96-MCA that pipetted 100 µL of 1.0 M ammonium formate buffer (pH ˜3) and 600 µL of ethyl acetate/hexane (10:90, v/v) into each sample in the final sample plate, followed by an offline vortexing for 1 min (Fig. 1C). After centrifugation for 4 min at 2000×g, the upper organic layer was pipetted

Jiang et al. 217

to a collection 96-well plate (labeled as 1D′ or 2D′ in Fig. 1C) and evaporated under nitrogen for 30 min at 40 °C. The residue was reconstituted in 100 µL of 5 mM ammonium bicarbonate in acetonitrile/water (50:50, v/v) and then injected into the LC-MS/MS system for analyses.13 Chromatographic separation was achieved in 4 min on a Waters Atlantis dC18 analytical column (2.1 × 50 mm, 3 µm) with the mobile phases A (10 mM ammonium bicarbon-ate) and B (acetonitrile) under a gradient program. Detection was accomplished using a Sciex API4000 triple quadrupole mass spectrometer using positive ion electrospray and mul-tiple reaction monitoring (BMS-650032, m/z 748 > 535; isotopically labeled BMS-650032, m/z 757 > 536).

Results and DiscussionFreedom EVO Pipetting Accuracy and Precision

Water is commonly used as a standard solvent to evaluate pipetting performance because its specific gravity is known (˜0.998 g/mL at 25 °C) and it is easy to obtain in laboratories.

In this study, good pipetting accuracies (%Dev ≤ ±2.1%) and precisions (%CV ≤ 1.6%) were demonstrated at both volumes of 25 and 900 µL for eight channel pipettors (Table 2). The results showed that the Freedom EVO was accurate and precise in pipetting 25 to 900 µL liquids.

The relative gravity (density) of human plasma is about 2% higher than that of water.14 It is not practical to calibrate Freedom EVO based on the plasma relative gravity15 because it is labor intensive and has minimal benefit in improving pipetting accuracy. In addition, minor differences in the plasma from different species or individuals may exist. In this study, a relative comparison between manual pipetting (using handheld pipettors) and automated pipetting (by Freedom EVO) was conducted. The mean weights of six ali-quots of plasma from different sources were compared with the mean weights of rat plasma by manual pipetting. The maximum %Dev (2.5%) and %CV (4.0%) were observed at the lowest pipetted volume (25 µL); the %Dev and %CV at all the other volumes were within ±2.0% (Table 3).

Based on the mean weight of rat plasma at different volumes, the dilution bias (%Dev) at each dilution factor (2, 5, 10, 20, 50, 100, 200, 500, and 1000) was calculated

Table 2. System Verification for Pipetting Accuracy and Precision

Volume of Water (µL) 25 (n = 6) 900 (n = 6)

Tip Mean Weight (mg) Mean %Dev %CV Mean Weight (mg) Mean %Dev %CV

1 25.3 1.1 1.2 906.0 0.7 0.02 24.9 -0.3 1.0 905.9 0.6 0.13 25.1 0.5 1.1 905.5 0.6 0.14 24.7 -1.1 1.3 905.5 0.6 0.35 24.5 -2.1 1.6 904.1 0.5 0.06 24.7 -1.1 1.1 905.0 0.6 0.07 24.6 -1.5 0.6 905.4 0.6 0.08 25.3 1.1 0.6 906.0 0.7 0.0

Table 3. The Mean, %CV, and %Dev of Plasma Weights at Different Pipetting Volumes (Manual vs. Automated)

Nominal Volume (µL)

Mean Weight (mg) of

Rat Plasma by Manual Pipetting, (%CV)

Mean Weight (mg) of Plasma Pipetted by Freedom EVO in Six Replicates (%CV; %Deva)

Rat Mouse Rabbit Monkey Dog Human

25 24.9 (1.6) 25.4 (0.8; 2.0) 25.7 (0.5; 3.2) 25.5 (0.7; 2.4) 25.5 (2.5; 2.4) 25.8 (0.9; 3.6) 25.9 (0.5; 4.0)50 49.7 (1.8) 50.6 (0.5; 1.8) 50.5 (0.3; 1.6) 50.8 (0.4; 2.2) 50.7 (0.2; 2.0) 50.7 (0.4; 2.0) 50.7 (0.5; 2.0)100 101.8 (0.7) 101.3 (0.2; –0.5) 101.4 (0.2; –0.4) 101.7 (0.2; –0.1) 101.7 (0.2; –0.1) 101.6 (0.1; –0.2) 102.0 (1.0; 0.2)200 202.6 (2.1) 204.2 (0.1; 0.8) 204.3 (0.1; 0.8) 204.6 (0.1; 1.0) 204.8 (0.0; 1.1) 204.9 (0.1; 1.1) 204.9 (0.1; 1.1)225 226.9 (0.9) 230.0 (0.1; 1.4) 229.9 (0.1; 1.3) 230.1 (0.1; 1.4) 230.5 (0.1; 1.6) 230.7 (0.2; 1.7) 230.6 (0.0; 1.6)450 458.7 (0.5) 458.1 (0.1; –0.1) 458.6 (0.0; 0.0) 458.8 (0.1; 0.0) 459.3 (0.0; 0.1) 459.5 (0.1; 0.2) 459.9 (0.2; 0.3)

a. The weight from manual pipetting (by handheld pipettors) was used as the nominal value.

218 Journal of Laboratory Automation 17(3)

from the nominal dilution factors (DFn) and the actual dilu-tion factors (DFa), expressed as %Dev = (1/DFa – 1/DFn)/(1/DFn) × 100% (Table 4). The estimated dilution bias (%Dev) from the nominal value was within ±2.0% at the dilution factors 2 to 1000. The bias was slightly increased with the increase of the dilution steps. In contrast, manual dilution results had greater biases at all the dilution factors.

Plasma pipetting accuracy was further demonstrated by pipetting 25 µL of plasma from different volumes of rat plasma in two types of sample tubes (Table 5). The %Dev from the nominal value (defined as 25.4 mg by pipetting 25 µL of rat plasma from the 100 mL trough by the Freedom EVO; Table 3) was within ±1.9%, and %CV was within 1.9%. The results showed that the plasma pipetting accu-racy and precision were independent of the sample volume in the tubes; that is, even 50 µL of plasma in the tubes was sufficient for pipetting 25 µL volumes.

Accuracy and Precision of Sample Dilution

To evaluate sample dilution accuracy and precision, RSPP prepared QCs containing BMS-650032 were diluted at different dilution factors followed by LC-MS/MS analy-ses. The measured concentrations of these QCs were obtained by back-calculating from standard curves and multiplying by the corresponding dilution factors. The mean %Dev of the concentrations from corresponding nominal concentrations was within ±7.7% at different dilution factors and concentration levels (Table 6). These results confirm the estimation of the dilution bias from Table 4. The results of incurred study sample reanalysis show good reproducibility (Table 7). The %Dev of the repeat value from the mean of the initial value and the repeat value was within ±7.6% for samples from different animals and collection times.

Table 5. Impact of Sample Volume on the Accuracy and Precision of Pipetting from Tubes

Initial Volume in Tube (µL)

Weight (mg) of 25 µL Plasma from BMS-Customized Tube Weight (mg) of 25 µL Plasma from Corning Tube

50 75 100 200 50 75 100 200

24.9 24.7 25.2 25.2 25.9 25.0 25.7 25.6 25.4 25.2 25.0 25.3 25.2 25.0 25.3 25.2 25.8 25.0 25.7 25.5 25.2 25.2 25.2 25.5 25.8 24.9 25.4 24.3 25.9 24.5 25.4 25.2 25.0 25.1 25.5 24.5 25.8 25.8 25.9 25.1 25.3 24.4 24.9 24.7 25.4 25.1 25.1 25.0Mean 25.4 24.9 25.3 24.9 25.6 25.1 25.4 25.3%Deva -0.1 -2.0 -0.5 -1.9 0.7 -1.2 0.1 -0.5SD 0.4 0.3 0.3 0.5 0.3 0.4 0.3 0.2%CV 1.5 1.2 1.2 1.9 1.3 1.7 1.2 0.9

a. The nominal value was defined as the mean weight (25.4 mg) of 25 µL rat plasma pipetted by Freedom EVO (Table 3).

Table 4. Estimated Dilution Biases (%Dev) from Automated Dilution Compared with Manual Dilution

Automated Manual

Nominal DF (DFn) Dilution Scheme (µL) Actual DF (DFa) %Dev* Actual DF (DFa) %Dev*

2 (25 + 25) 2.0 0.0 2.0 0.05 (25 + 100) 5.0 0.2 5.1 -1.710 (25 + 225) 10.1 -0.5 10.1 -1.12 (50 + 50) 2.0 0.0 2.0 0.05 (50 + 200) 5.0 -0.7 5.1 -1.510 (50 + 450) 10.1 -0.5 10.2 -2.220 (25 + 225) ; (50 + 50) 20.1 -0.5 20.2 -1.150 (25 + 225); (50 + 200) 50.6 -1.3 51.3 -2.6100 (25 + 225); (50 + 450) 101.1 -1.1 103.4 -3.3200 (25 + 225); (50 + 450); (50 + 50) 202.2 -1.1 206.9 -3.3500 (25 + 225); (50 + 450); (50 + 200) 509.0 -1.8 525.1 -4.81000 (25 + 225); (50 + 450); (50 + 450) 1016.3 -1.6 1058.2 -5.5

*%Dev = (1/DFa - 1/DFn)/(1/DFn) ´ 100%, herein, DFn = Nominal DF, DFa = Actual DF

Jiang et al. 219

Freedom EVO System Configuration

Each carrier and labware on the worktable was well cali-brated and aligned on x-, y-, and z-axis positions. The z-max value was critical for accurately pipetting liquid from sam-ple tubes and the wells of the 96-well plates. The liquid detection mode was turned on in the liquid class for plasma pipetting to check of liquid volumes. The function of “mix after dispense” was activated to allow automatic mixing during the sample dilution process. Disposable tips (DiTi 200 µL and DiTi 1000 µL) were used to avoid carryover and dilution effects from fixed tips.16,17 The integrated RSPP

Freedom EVO system is not limiting and provides flexibil-ity for using different types and dimensions of carriers and labware; however, a set of standardized carriers and labware is recommended to ensure reproducibility and ruggedness in sample pipetting and dilution.

Advantages and Future ImprovementsThe RSPP is an interface program bridging Watson and EVOware. Thus, automated sample pipetting and dilution can be accomplished based on sample sequence and dilu-tion factors from a Watson worklist. The entire process for

Table 6. Accuracy and Precision of the Concentrations of BMS-650032 in Diluted Samples

Sample5 × LLOQ QC

(25 ng/mL) High QC (1600 ng/mL)Dilution QC

(50 000 ng/mL)

Run ID Dilution Factor 1 2 1 2 5 10 20 200

1 %CV (n = 6) 3.8 NAa 1.7 NA 2.1 2.9 0.9 1.0 %Dev -1.3 NA -0.6 NA -2.5 -0.5 -3.5 -7.72 %CV (n = 6) 3.9 7.6 3.5 2.2 1.2 2.8 6.2 2.3 %Dev -5.0 -6.7 3.2 -6.5 2.5 4.0 6.7 -2.33 %CV (n = 6) 6.2 6.4 2.1 1.9 2.1 3.6 1.7 1.6 %Dev -5.0 -3.3 1.3 -5.3 3.4 3.7 -1.8 -5.8

a. Not available.

Table 7. Reproducibility Demonstrated by the Incurred Sample Reanalysis Results

Number Sample IdentificationInitial Value

(ng/mL)Repeat Value

(ng/mL)Mean

(ng/mL)Absolute % Deviation of Values from Mean

Dilution Factor

1 Animal 1, day 1, 1 h 13 048.47 14 555.66 13 802.07 5.5 202 Animal 1, day 1, 2 h 33 599.94 33 831.26 33 715.60 0.3 503 Animal 1, day 1, 4 h 52 765.33 57 405.45 55 085.39 4.2 504 Animal 1, day 8, 4 h 28 205.64 29 013.13 28 609.39 1.4 505 Animal 1, day 8, 8 h 24 367.19 23 697.26 24 032.23 1.4 506 Animal 2, day 1, 2 h 55 760.54 55 604.93 55 682.74 0.1 507 Animal 2, day 1, 3 h 72 060.16 68 738.49 70 399.33 2.4 508 Animal 2, day 8, 1 h 14 083.52 16 075.20 15 079.36 6.6 209 Animal 2, day 8, 2 h 24 667.09 25 064.90 24 866.00 0.8 50

10 Animal 2, day 8, 3 h 47 847.39 48 040.40 47 943.90 0.2 5011 Animal 3, day 1, 3 h 15 920.19 17 271.13 16 595.66 4.1 5012 Animal 3, day 1, 4 h 9439.29 11 001.40 10 220.35 7.6 5013 Animal 3, day 1, 8 h 2681.97 2980.12 2831.05 5.3 5014 Animal 3, day 8, 1 h 13 262.83 14 438.23 13 850.53 4.2 2015 Animal 3, day 8, 4 h 5747.48 5381.79 5564.64 3.3 5016 Animal 4, day 1, 1 h 13 711.41 14 338.18 14 024.80 2.2 2017 Animal 4, day 1, 4 h 10 991.09 10 754.75 10 872.92 1.1 5018 Animal 4, day 8, 2 h 3569.42 3758.36 3663.89 2.6 5019 Animal 4, day 8, 4 h 1993.90 1881.94 1937.92 2.9 5020 Animal 4, day 8, 8 h 263.12 284.54 273.83 3.9 50

220 Journal of Laboratory Automation 17(3)

192 samples, two 96-well plates, can be completed within an hour, which saves more than 50% in sample preparation time compared with a manual process. In addition, this automated process provides significantly better data reproducibility. The improved functions of EVOware at tracking and recov-ering the dilution process (resuming dilution process from an unexpected stop) make the RSPP application more pow-erful and controllable because EVOware records each step of liquid pipetting and is able to continue pipetting from an unexpected stop in a run. Both the RSPP and EVOware have been validated in house according to regulatory guid-ance and are being applied to the regulated nonclinical and clinical bioanalytical sample analyses.

It should be noted that several factors may affect dilution accuracy and precision: (1) The quality of biological sam-ples is critical to the pipetting accuracy. For example, gel-like and clotted plasma samples, and samples that may be less homogeneous (e.g., tissue homogenates), may not be accurately pipetted; therefore, checking the sample quality and homogeneity before sample dilution is necessary. (2) Appropriate integer dilution factors are recommended to use, such as 2, 5, 10, 20, 50, 100, 200, 500, and 1000. Dilution factors such as 11, 12. . ., 21, 22, . . ., 51, 52, . . ., 101, 102, . . ., 201, 202,. . ., 501, 502, are not recommended because the initial sample pipetting volumes may be less than the lowest qualified pipetting volume (25 µL), which may cause dilution bias.

The following improvements on the RSPP will be con-sidered in the future: (1) a Watson worklist can be trans-formed directly to one .gwl file without generating any CSV worklist, because the .gwl file is the final executable file for pipetting; (2) standards and QCs can be pipetted from sample tubes in a rack separate from the study sample racks to avoid rearranging samples in a sample sequence consistent with Watson; and (3) standards or QCs can be multipipetted from same sample tubes for preparing repli-cate standards and QCs in a run. These improvements should make the process even more user-friendly, efficient, and convenient.

This Microsoft Excel–based RSPP makes it possible to automatically pipette and dilute samples based on the sequence and dilution factors from a Watson worklist. This fully automated sample dilution and preparation process greatly saves effort and time in biological sample analysis and provides higher quality bioanalytical data.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, author-ship, and/or publication of this article.

References

1. Food and Drug Adminstration. Guidance for Industry: Bioan-alytical Method Validation. http://www.fda.gov/downloads/ Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ ucm070107.pdf.

2. Ma, J.; Shi, J.; Le, H.; Cho, R.; Huang, J. C.; Miao, S.; Wong, B. K. A Fully Automated Plasma Protein Precipita-tion Sample Preparation Method for LC-MS/MS Bioanalysis. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 2008, 862, 219–226.

3. Deng, Y.; Wu, J. T.; Lloyd, T. L.; Chi, C. L.; Olah, T. V.; Unger, S. E. High-Speed Gradient Parallel Liquid Chromatography/Tandem Mass Spectrometry with Fully Automated Sample Preparation for Bioanalysis: 30 Seconds per Sample from Plasma. Rapid Commun. Mass Spectrom. 2002, 16, 1116–1123.

4. Zhang, N.; Hoffman, K. L.; Li, W.; Rossi, D. T. Semi-automated 96-Well Liquid-Liquid Extraction for Quantita-tion of Drugs in Biological Fluids. J. Pharm. Biomed. Anal. 2000, 22, 131–138.

5. Harrison, A. C.; Walker, D. K. Automated 96-Well Solid Phase Extraction for the Determination of Doramectin in Cat-tle Plasma. J. Pharm. Biomed. Anal. 1998, 16, 777–783.

6. Kronkvist, K.; Gustavsson, M.; Wendel, A. K.; Jaegfeldt, H. Automated Sample Preparation for the Determination of Budesonide in Plasma Samples by Liquid Chromatography and Tandem Mass Spectrometry. J. Chromatogr. A. 1998, 823, 401–409.

7. Polettini, A.; Groppi, A.; Vignali, C.; Montagna, M. Fully-Automated Systematic Toxicological Analysis of Drugs, Poi-sons, and Metabolites in Whole Blood, Urine, and Plasma by Gas Chromatography-Full Scan Mass Spectrometry. J. Chro-matogr. B Biomed. Sci. Appl. 1998, 713, 265–279.

8. Hsieh, J. Y.; Lin, C.; Matuszewski, B. K.; Dobrinska, M. R. Fully Automated Methods for the Determination of Hydro-chlorothiazide in Human Plasma and Urine. J. Pharm. Biomed. Anal. 1994, 12, 1555–1562.

9. Hempenius, J.; Hendriks, G.; Hingstman, J.; Mensink, C. K.; Jonkman, J. H.; Lin, C. C. An Automated Analytical Method for the Determination of Felbamate in Human Plasma by Robotic Sample Preparation and Reversed-Phase High Per-formance Liquid Chromatography. J. Pharm. Biomed. Anal. 1994, 12, 1443–1451.

10. Holman, J. W.; Felder, R. A. Robotic Automation of Cyclo-sporine Analysis in Whole Blood. Clin. Chem. 1992, 38, 1440–1443.

11. Hempenius, J.; Wieling, J.; Jonkman, J. H.; de Noord, O. E.; Coenegracht, P. M.; Doornbos, D. A. Development of a Laboratory Robotic System for Automated Bioanalytical Methods—I. The Determination of Theophylline in Human Plasma: A Comparison between the Robotized and Manual Method. J. Pharm. Biomed. Anal. 1990, 8, 313–320.

Jiang et al. 221

12. Gu, H.; Unger, S.; Deng, Y. Automated Tecan Programming for Bioanalytical Sample Preparation with EZTecan. Assay Drug Dev. Technol. 2006, 4, 721–733.

13. Jiang, H.; Zeng, J.; Peng, Q.; Zheng, N.; Kandossi, H.; Valentine, J.; Arnold, M. E. A Convenient Strategy for Quantitative Determination of Drug Concentrations in Tis-sue Homogenates Using a Liquid Chromatography/Tandem Mass Spectrometry Assay for Plasma Samples. Manuscript in preparation.

14. Trudnowski, R. J.; Rico, R. C. Specific Gravity of Blood and Plasma at 4 and 37 °C. Clin. Chem. 1974, 20, 615–616.

15. Xie, I. H.; Wang, M. H.; Carpenter, R.; Wu, H. Y.Automated Calibration of TECAN Genesis Liquid Handling Workstation Utilizing an Online Balance and Density Meter. Assay Drug Dev. Technol. 2004, 2, 71–80.

16. Gu, H.; Deng, Y. Dilution Effect in Multichannel Liquid- Handling System Equipped with Fixed Tips: Problems and Solutions for Bioanalytical Sample Preparation. J. Assoc. Lab. Automat. 2007, 12, 355–362.

17. Dong, H.; Ouyang, Z.; Liu, J.; Jemal, M. The Use of a Dual Dye Photometric Calibration Method to Identify Possible Sample Dilution from an Automated Multichannel Liquid-Handling System. J. Assoc. Lab. Automat. 2006, 11, 60–64.