A Strategy of Plasma Protein Quantitation by Selective Reaction

Anal. Chem. 2003, 75, 7008-7014

A Strategy of Plasma Protein Quantitation by Selective Reaction Monitoring of an Intact Protein Qin C. Ji,* Ramona Rodila, Eric M. Gage, and Tawakol A. El-Shourbagy

Department of Drug Analysis, Abbott Laboratories, Abbott Park, Illinois 60064

Immunoassays are used extensively in the quantitative analysis of proteins in plasma, urine, and other biological matrixes to support preclinical and clinical studies. Although immunoassays are both sensitive and rapid, difficulties during development of these assays are compounded by the need to have a specific antibody or antigen to the protein of interest. Furthermore, calibration curves of immunoassays are inherently nonlinear, and the technique often detects many structurally related components in addition to the analyte of interest. We have developed a novel strategy of analyzing protein concentrations in plasma by utilizing 96-well solid-phase extraction and LC-MS/MS detection of the intact protein. This strategy has been successfully applied in method development and assay validation of quantitatively analyzing protein rK5 concentrations in monkey plasma samples. Additional techniques such as precolumn regeneration and column heating were also incorporated into the assay. Total run time for each sample was ∼15 min. An LLOQ of 99.2 ng/ mL from a sample volume of 50 µL, corresponding to only 380 fmol (3.97 ng) of the rK5 analyte being injected onto the analytical column (assuming 100% extraction recovery), was obtained. The validated linear dynamic range was between 99.2 and 52 920.0 ng/mL, with a correlation coefficient (r2) ranging from 0.9972 and 0.9994. The intraassay CV for this assay was between 0.6 and 3.8%, and the interassay CV was between 1.7 and 3.2%. Interassay mean accuracies were between 101.5 and 104.7%. The assay has proven rugged and specific and has been employed to generate data in support of preclinical studies. This strategy for rK5 assay could be used for the development of bioanalytical assays to provide preclinical and clinical support for other protein drug candidates and, furthermore, for the validation of biomarkers discovered from proteomic research. Bioanalytical assays for measuring concentrations of therapeutic compounds in biological matrixes, such as plasma and serum, are essential throughout drug development, preclinical, and clinical studies. Significant advances of biotechnology in drug discovery and development in recent years has led to an increasing number of therapeutic compounds that are proteins having molecular weights in excess of 10 000 amu. Traditionally, immunoassays are used as the primary bioanalytical method to ac* Corresponding author: (phone) (847) 937 5786; (fax) (847) 938 7789; (e-mail) [email protected].

7008 Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

complish these measurements. Although immunoassays are both sensitive and rapid, difficulties during development of these assays are compounded by the need to have a specific antibody or antigen to the protein of interest. Furthermore, calibration curves of immunoassays are inherently nonlinear, and the technique often detects many structurally related components in addition to the analyte of interest. In the field of proteomics, emphasis was originally placed on the relative quantification of proteins.1 Technologies such as twodimensional gel electrophoresis and matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) have been used for direct monitoring of intact proteins to obtain relative quantitation.2-4 MALDI-MS/MS has also been successful in quantitating cell line proteins by detecting peptide segments digested from the protein analyte and purified by gel electrophoresis.5 When LC and electrospray ionization techniques are involved, relative protein quantitation can be achieved by detection of a representative peptide fragment that is labeled with stable isotope tags either before or after protein digestion.3,6,7 Furthermore, absolute concentrations of proteins were obtained by adding stable isotope labeled proteins5 or peptides8 as internal standard before the digestion. Difficulties in the analysis of plasma samples arise from the fact that samples contain large quantities of various proteins and other compounds inherent to the matrix. Sample extraction techniques such as protein precipitation, liquid-liquid extraction, and solid-phase extraction (SPE) are often used to isolate small organic molecules from these samples for analysis. LC-MS and LC-MS/MS technology have found widespread use in bioanalytical assays for the concentration determination of these molecules in plasma.9,10 However, extracting specific proteins to a degree of efficiency conducive to LC-MS/MS remains a challenging task. This is perhaps the primary limitation in applying LC-MS/MS to the quantitative analysis of proteins in plasma samples. (1) Steen, H.; Pandey, A. Trends Biotechnol. 2002, 20, 361-364. (2) Ardekani, A. M.; Herman, E. H.; Sistare, F. D.; Liotta, L. A.; Petricoin, E. F., III. Curr. Ther. Res. 2001, 62, 803-819. (3) Patterson, S. D. Curr. Opin. Biotechnol. 2000, 11, 413-418. (4) Hamdan, M.; Righetti, P. G. Mass Spectrom. Rev. 2002, 21, 287-302. (5) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6591-6596. (6) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Nat. Biotechnol. 1999, 17, 994-999. (7) Stewart, I. I.; Thomson, T.; Figeys, D. Rapid Commun. Mass Spectrom. 2001, 15, 2456-2465. (8) Gerber, S. A.; Rush, J.; Stemman, O.; Kirschner, M. W.; Gygi S. P. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 6940-6945. (9) Jemal, M. Biomed. Chromatogr. 2000, 14, 422-429. (10) Lee, M. S.; Kerns, E. H. Mass Spectrom. Rev. 1999, 18, 187-279. 10.1021/ac034930n CCC: $25.00

© 2003 American Chemical Society Published on Web 11/15/2003

This assay has been validated following the FDA’s Guidance for Industry-Bioanalytical Method Validation16 and was applied to support toxicokinetic studies.

Figure 1. Structure of rK5.

In addition to the challenges encountered during sample extraction, mass spectrometry utilizing electrospray ionization of proteins has a tendency to form multiply charged ions, which complicates monitoring of the analyte. There are several approaches in monitoring ions of a protein for quantitative analysis. One is to employ selective ion monitoring, or full mass range monitoring, in which peak intensities of multiply charged ions are used directly for quantitative analysis.11,12 Another is to perform deconvolution of the multiply charged ion peaks to obtain a molecular weight peak for the compound and then use the intensity of this molecular weight peak for quantitative analysis.13 As shown in Figure 1, rK5 is a protein drug candidate with molecular weight of 10 464 amu. It is a specific and potent angiogenesis inhibitor that may be effective in the treatment of human brain glioma and other human tumors.14 In a previous paper,15 we presented our initial experiments to determine rK5 concentrations in human plasma utilizing a novel strategy for protein quantitation in support of clinical studies. A semiautomated SPE procedure in 96-well format was developed to effectively extract protein analytes from plasma samples, with subsequent analysis by LC-MS/MS. Selective reaction monitoring (SRM) was employed to monitor the ions of the intact protein analyte for quantitation. Although SRM and 96-well SPE have been widely used for the quantitation of small organic molecules and peptides, no report has been found describing the use of either SRM or SPE for the quantitative analysis of proteins with molecular weight of more than 10 000 amu to support preclinical and clinical studies. In this paper, we present the results from a novel strategy utilizing 96-well solid-phase extraction and LC-MS/MS detection of the intact protein to determine protein concentrations in cyno (monkey) plasma to support preclinical studies. A 15N isotopically labeled version of the rK5 protein was used as internal standard. (11) Garcia, M. C.; Hogenboom, A. C.; Zappey, H.; Irth, H. J Chromatogr., A 2002, 957, 187-199. (12) Mao, Y.; Moore, R. J.; Wagnon, K. B.; Pierce, J. T.; Debban, K. H.; Smith, C. S.; Dill, J. A.; Fuciarelli, A. F. Chem. Res. Toxicol. 1998, 11, 953-961. (13) Bergen, H. R.; Lacey, J. M.; O’Brien, J. F.; Naylor, S. Anal. Biochem. 2001, 296, 122-129. (14) Davidson, D. J.; Ennis, B.; Kherzai, A.; Egan, D. A.; Walter, K.; Gubbins, E. F.; Nuss, M.; Luo, Y.; Haskell, C.; Wang, Y.; Lesniewski, R.; Henkin, J. J. Clin. Invest., submitted. (15) Ji, Q. C.; Gage, E. M.; Rodila, R. C.; Chang, M. S.; El-Shourbagy, T. A. Rapid Commun. Mass Spectrom. 2003, 17, 794-799.

EXPERIMENTAL SECTION Chemicals and Reagents. High-purity methanol, acetonitrile, and water, all Omnisolv grade, were purchased from EMD, formerly EM Science (Gibbstown, NJ), along with HPLC grade hexanes. Glacial acetic acid was purchased from Aldrich (St. Louis, MO). Trifluoroacetic acid was purchased from EM Science. The stock solutions of rK5 and internal standard were produced at Abbott Laboratories (Abbott Park, IL). Normal cyno plasma with potassium EDTA as anticoagulant (NCP-KEDTA) was purchased from Lampire Biological Laboratories (Pipersville, PA). Instrumentation. Plasma solutions were transferred using a single-channel positive displacement hand-held pipet from Gilson (Middleton, WI). Multichannel hand-held electronic pipets were from BioHit (Helsinki, Finland), and solid-phase extraction plates were provided by Waters Corp. (Milford, MA). A Hamilton (Reno, NV) MicroLab AT 2 Plus automated liquid handler was used for the addition and mixing of internal standard, and a BeckmanCoulter (Fullerton, CA) square well plate collar and vacuum manifold base were used in the solid-phase extraction process. The Shimadzu (Kyoto, Japan) HPLC system included a Shimadzu LC-10 AD HPLC pump, a Shimadzu SIL-10A XL autosampler, and a Shimadzu SCL-10A system controller. A Hot Pocket column heater from Keystone Scientific (Bellefonte, PA) was also used. The valves used to control LC flow between the mass spectrometer inlet and waste line were from Valco Instruments (Houston, TX). An 1100 series HPLC pump and degasser system from Agilent Technologies (Palo Alto, CA) was used to deliver backwash solvent for the precolumn regeneration. An API-3000 mass spectrometer and computer control system were from PE Sciex (Toronto, ON, Canada). A 2.1 × 150 mm Symmetry 300 C18, 5-µm column from Waters was used as the analytical column, along with a precolumn consisting of a 3.9 × 20 mm Symmetry 300 C18, 5-µm cartridge, also from Waters, and an in-line filter with an A-110 2 µm titanium frit from Upchurch Scientific Inc. (Oak Harbor, WA). MassChrom version 1.1.1 was used as the data acquisition software. Preparation of Standard and QC Samples. Stock solutions from the same source and of the same concentration were used in preparation of all standard and QC samples. Separate working solutions for the standards and QCs were prepared by diluting the stock solution of the analyte with water. Standard levels 1-10 at concentrations of 99.23, 198.45, 661.50, 1653.75, 3307.50, 9922.50, 19 845.00, 33 075.00, 42 336.00, and 52 920.00 ng/mL were prepared by adding the appropriate volume of working solution into a class A volumetric flask and diluting to mark with pooled NCP-KEDTA. QC samples were also prepared in the same manner at concentrations of 235.20, 1176.00, 5880.00, 17 640.00, and 41 160.00 ng/mL. Standards and QCs were then aliquoted into polypropylene tubes and stored in freezers maintained at ∼-70 °C. Additional QCs were stored at ∼-20 °C for the purpose of stability evaluation. Sample Extraction. Samples were thawed at room temperature, followed by mixing to ensure homogeneity. Fifty microliters (16) Food and Drug Administration of the United States. Guidance for IndustryBioanalytical Method Validation; U.S. Department of Health and Human Services, Center for Drug Evaluation and Research (CDER), Center for Veterinary Medicine (CVM), May 200; http://www.fda.gov/cder/guidance/ index.htm.

Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

7009

of each plasma sample was loaded into the appropriate wells of a 96-well plate using a hand-held single-channel pipet. The Hamilton automated liquid handler was then used for the addition of 200 µL of an ∼5 µg/mL 15N rK5 solution as internal standard, followed by aspirating and dispensing six times to ensure mixing of solutions. A Waters Oasis HLB 60-mg solid-phase extraction plate was then conditioned by adding 1 mL of methanol to each well and drawing through with vacuum, followed by equilibration by adding 1 mL of water to each well and drawing through with vacuum. The contents of the 96-well plate were then transferred to the corresponding wells of the solid-phase extraction plate using a multichannel pipet, followed by application of vacuum. Each well was then washed by adding 1 mL of water with 0.2% trifluoroacetic acid and drawing through with vacuum. An additional wash was then performed by adding 1 mL of hexane to each well and drawing through with vacuum. Wells were then eluted by adding 0.8 mL of acetonitrile with 0.2% trifluoroacetic acid and drawing through by vacuum into a clean 96-well plate. The wells were then dried down over room-temperature nitrogen and reconstituted with 100 µL of water. Samples were then injected onto the LCMS/MS system for analysis. Chromatographic Conditions. A gradient HPLC method was employed for separation. The gradient consisted of two mobile phases, A and B. Mobile phase A consisted of 0.1% acetic acid and 0.02% trifluoroacetic acid in water, and mobile phase B consisted of 0.1% acetic acid and 0.02% trifluoroacetic acid in 80/ 20 (v/v) acetonitrile/water. The timing program was designed such that mobile phase B was maintained at 6.3% for the first 2.00 min. Mobile phase B was then ramped up to 62.5% between 2.00 and 5.00 min and was maintained at 62.5% from 5.00 to 12.00 min, followed by a reduction from 62.5 to 6.3% between 12.00 and 12.10 min. Mobile phase B was maintained at 6.3% from 12.10 to 15.00 min for the purpose of column reequilibration. The flow rate for this program was set to 0.2 mL/min, except during column reequilibration, when it was increased to 0.3 mL/min between 12.20 and 14.50 min. The analytical column was maintained at a temperature of 40 °C, and the injection volume was 40 µL. Mobile phase A was used as the injector wash solution, and the injector was rinsed with 1 mL of wash following each injection. The precolumn backwash mobile phase was composed of 50/50 (v/ v) acetonitrile/water with 0.1% trifluoroacetic acid. The use of this precolumn backwash mobile phase is discussed in LC-MS/MS Detection. MS/MS Detection. LC-MS/MS detection was performed using a PE Sciex API 3000 triple quadrupole mass spectrometer with a Turbo Ionspray ionization source operated in the positive ion mode. The computer control system was MassChrom version 1.1.1. The spray voltage was 4000 V. The source temperature was 350 °C. The auxiliary gas flow rate was 7.5 L/min. The nebulizer gas setting was 10, and curtain gas setting was 10. Other parameters were optimized by infusing the analyte with a mixture of 50:50 mobile phase A/B at a flow rate of 200 µL/min. The following were some other parameters of the mass spectrometer state file, which is a set of instrument parameters that allow optimal ion transmission for mass spectrometric detection. OR (orifice) was 46 V. RNG (focusing ring) was 260 V. Q0 (quad 0 rod offset) was -10 V. R01 (quad 1 rod offset) was -11 V. R02 (quad 2 rod offset) was -64 V. R03 (quad 3 rod offset) was -66 V. The SRM detection channel for rK5 was m/z 1495.5 f 1462.9. The SRM detection channel for the internal standard was m/z 1513.5 f 1480.8. 7010

Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

Quantitiation. The peak areas of rK5 and the internal standard were determined using Sciex MacQuan software version 1.6. For each analytical batch, a calibration curve was derived from the peak area ratios (analyte/internal standard) using weighted linear least-squares regression of the area ratio versus the concentration of the standards. A weighting of 1/x2 (where x is the concentration of a given standard) was used for curve fit. The regression equation for the calibration curve was used to back-calculate the measured concentration for each standard and QC, and the results were compared to the theoretical concentration to obtain the accuracy, expressed as a percentage of the theoretical value, for each standard and QC measured. RESULT AND DISCUSSION Sample Extraction. Due to the inherent nature of plasma matrixes, a given plasma sample may contain a considerable number of proteins, the concentrations of which may be significantly higher than the analyte of interest. In our initial method development work,15 it was found that a 60-mg Waters Oasis HLB extraction plate provides excellent retention for the analyte, which allows optimization of the wash step to remove as many interfering compounds as possible. After the initial aqueous wash, a strong organic solvent, hexane, was incorporated into the second wash step. The elution solvent (70:30 v/v acetonitrile/water with 0.1% acetic acid and 0.02% trifluoroacetic acid) that was used in the extraction procedure described in our initial report15 was replaced by acetonitrile with 0.2% trifluoroacetic acid (TFA), as the additional TFA greatly improves the extraction efficiency. Elimination of water from the elution solvent also greatly accelerates the dry-down process. LC-MS/MS Detection. Figure 2A shows the ESI MS spectrum of rK5 infused with a 1:1 mixture of mobile phases A and B at a flow rate of 200 µL/min. The ESI spectrum contained primarily the multiply charged ions with m/z equal to 1308.7, 1495.5, 1744.4, and 2093.2. The MS/MS spectrum of the precursor ion m/z 1496 is shown in Figure 2B. The ESI spectrum of internal standard is shown in Figure 2C. The ESI spectrum contained primarily the multiply charged ions with m/z equal to 1324.0, 1514.0, 1766.0, and 2118.0. The MS/MS spectrum of the precursor ion m/z 1514 is shown in Figure 2D. The collision energy for both rK5 and internal standard was 378 eV (the number of the precursor ion charges multiplied by the difference between Q0 and R02 voltages). Although the signal from SRM detection is generated from only one of the several charged states in the ion envelope of the protein, SRM detection has the benefit of significantly reducing background interference and therefore greatly improves the signal-to-noise ratio. A gradient HPLC method was employed during elution into the mass spectrometer. This aids in separating interfering compounds from the analyte. Although separation of background peaks cannot be observed using SRM monitoring, better chromatographic separation generally reduces the adverse effects of matrix in electrospray ionization. In our initial method, a Prevail C18, 5-µm HPLC column from Alltech (Deerfield, IL) was used. This column required conditioning with several injections of extracted plasma blanks before acceptable chromatographic peak shape was obtained.15 Utilization of the Symmetry300 C18, 5-µm HPLC column not only eliminates the need for column conditioning but also greatly reduces column carryover, which is one of the predominant challenges in LC analysis of proteins. This feature is one of the major factors that allow us to achieve the lower limit

Figure 2. Mass spectrum of rK5 (A), MS/MS spectrum of rK5 (B). mass spectrum of the internal standard (C), and MS/MS spectrum of internal standard (D).

Figure 3. Effect of temperature for column separation: ion chromatograph of rK5 at room temperature (A) and at 40 °C (B).

of quantitation (LLOQ) of the current assay, as discussed in the Assay Validation section. Dilution, the LLOQ of 99.23 ng/mL of this assay corresponds to less than 380 fmol (3.97 ng) of rK5 protein being injected into the analytical column (assuming 100% extraction recovery). Heating of the analytical column was also found to be indispensable in attaining good peak shape. As shown in Figure 3A, at room-temperature chromatograms of rK5 show overlapping peaks. When the column is heated to 40 °C, a vastly improved peak is obtained as shown in Figure 3B. Heating of the column is somewhat unconventional for the analysis of proteins due to the occurrence of protein precipitation at elevated column temperature, which can result in column deterioration and poor LC performance. However, in this case, it is believed that heating forces the protein into a uniform conformational arrangement, and no deterioration of the column was observed for the duration of the validation and sample analysis process. Another key factor in the success of this assay is the use of the precolumn and its subsequent regeneration after each sample.

Although the Oasis solid-phase extraction plate affords excellent recovery for the protein of interest, it also has a high retention of other background materials that are not removed by the additional wash steps. A significant amount of these are eluted with the protein analyte, and this can be visually observed in the extract. By employing precolumn regeneration, many of these can be removed before elution into the analytical column and mass spectrometer, which greatly improves the ruggedness and consistency of the assay. Prior to injection, the precolumn was in line with the LC flow as shown in Figure 4A. It was found that the retention of the analyte and internal standard in the precolumn is very similar to the retention in the analytical column. The 10-port valve was switched to the position shown in Figure 4B only after the analyte and internal standard eluted from the analytical column to the mass spectrometer. The backwash mobile phase, 50/50 (v/v) acetonitrile/water with 0.1% trifluoroacetic acid, was pumped by the backwash pump through the precolumn at a flow rate of 1.0 mL/min for 2.1 min. At this time the flow from the backwash pump moves through the precolumn in the direction opposite from the initial LC flow. This backward flow coupled with the organic content of the backwash mobile phase ensures that any material not eluted into the analytical column is removed from the precolumn prior to injection of the following sample. This helps to reduce carryover and prevents contamination from reaching the analytical column. After this initial backwash period, the solvent selection valve on the backwash pump was switched to select the equilibration mobile phase, which was prepared in such a way as to reproduce the initial chromatographic conditions, and consisted of 5/95 (v/v) acetonitrile/water with 0.1% acetic acid and 0.02% trifluoroacetic acid. The equilibration mobile phase was then pumped through the precolumn at a flow rate of 1.0 mL/ min for 1.0 min, followed by another 2.1 min at a flow rate of 0.2 mL/min, identical to that of the initial LC flow. This reconditions the precolumn and ensures that only a mobile phase equivalent to that used at the beginning of the gradient is present before Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

7011

Figure 4. HPLC backwash valve system: (A) valve position during sample analysis (B) and during precolumn regeneration process. Table 1. Statistical Calculation of Monkey Plasma Calibration Standards for Assay Linearity Validation standard level 1

2

3

4

5

6

7

8

9

10

theor concn (ng/mL) calcd concn (ng/mL)

99.2 100.7 102.6 100.3

198.5 192.9 186.6 193.6

661.5 667.0 671.7 668.5

1653.8 1620.7 1483.5 1669.7

3307.5 3187.0 3231.5 3256.0

9922.5 10161.7 10166.6 10154.7

19845.0 20380.5 20808.5 20749.7

33075.0 33610.7 34415.1 30739.6

42336.0 42813.2 43477.5 42162.4

52920.0 52005.7 52710.2 53730.5

mean % CV % theoretical

101.2 1.2 102.0

191.0 2.0 96.3

669.1 0.4 101.1

1591.3 6.1 96.2

3224.8 1.1 97.5

10161.0 0.1 102.4

20646.3 1.1 104.0

32921.8 5.9 99.5

42817.7 1.5 101.1

52815.5 1.6 99.8

the next sample is injected. At the end of the equilibration period, the 10-port valve was switched back to the starting position shown in Figure 4A in preparation for the next injection. This configuration allows washing of the precolumn without disrupting the flow to the mass spectrometer. The precolumn is washed off-line with the backwash mobile phase and reconditioned with the equilibration mobile phase as part of each injection. Assay Validation. The validation experiments were designed with reference to Guidance for Industry-Bioanalytical Method Validation recommended by the Food and Drug Administration (FDA) of the United States.16 The experimental design and results of some most important criteria of method validation are presented in following sections. 1. Linearity, LLOQ and ULOQ, Dilution. The evaluation of the linearity of the calibration curve was obtained from three consecutively prepared batches. The linear dynamic range evaluated was between 99.23 and 52 920.00 ng/mL. Across this assay range, the correlation coefficient (r2) was between 0.9972 and 0.9994. The mean back-calculated concentrations of the standards were between 96.2 and 104.0% of the theoretical concentrations (Table 1). A representative calibration curve is presented as Figure 5. Eighteen replicates of LLOQ samples from three separate runs were used to evaluate the precision and accuracy at the low end of the assay range. The CV was 4.8% and the accuracy, expressed as percent theoretical, was 105.3%. The precision and accuracy at the upper limit of quantitation (ULOQ) was determined in the same manner, with a CV equal to 1.7% and a mean accuracy of 100.3%. Representative LC-MS/MS chromatograms of ULOQ and LLOQ samples are shown in Figure 6A and B. The mass spectrometric data collection was started at 7.5 min after injection, 7012 Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

Figure 5. Example of standard calibration curves of rK5.

and the absolute retention time of both rK5 and internal standards is ∼8.28 min. 2. Accuracy and Precision. Eighteen replicates of QC samples from three consecutive runs were used to evaluate the precision and accuracy at each concentration level. The intraassay CV was between 0.6 and 3.8%, and the interassay CV was between 1.7 and 3.2%. The interassay mean accuracies, expressed as percent theoretical, were between 101.5 and 104.7% (Table 2). 3. Matrix Effect. The effect of matrix variation on the concentration determination of rK5 was investigated by preparing QC samples in NCP-KEDTA with two different unpooled lots of matrix, male and female. The QCs were evaluated using a calibration curve generated from the same standards used for the determination of linearity, accuracy, and precision. When quantitated on this curve, the mean accuracies of the QCs were between 92.0 and 104.4%, as shown in Table 3, demonstrating that the measured concentration of rK5 is independent of the sample matrix.

Table 3. Summary of Matrix Effect QC Sample Evaluation theor concn (ng/mL) matrix 1

2

Figure 6. Ion chromatographs of (A) high standard (52 920.00 ng/ mL rK5), (B) low standard (99.23 ng/mL rK5), and (C) blank plasma with internal standard. Table 2. Statistical Calculation of Monkey Plasma QC Samples for Assay Accuracy and Precision Validation quality control level 1

2

3

4

5

theor concn (ng/mL)

235.2

1176.0

5880.0

17640.0

41160.0

mean % CV % theoretical n

239.2 3.2 101.7 17

1208.7 2.1 102.8 18

6154.4 2.4 104.7 18

18047.4 1.7 102.3 18

41781.7 1.9 101.5 18

4. Selectivity. Selectivity was evaluated by extracting blank plasma samples from six different lots of matrix and comparing the response at the retention time of rK5 to the response at the LLOQ. No significant peaks were observed in any of the blank plasma samples. As shown in Figure 6B, LC-MS/MS response of the LLOQ sample was ∼1500 counts/s while noise peaks were almost not observed. This low noise level was primarily due to the high m/z values of the precursor and product ions, as the chance for chemical noise in this mass range was very small. As shown in Figure 6C, LC-MS/MS response for the blank plasma samples was ∼150 counts/s, which was significantly below the LC-MS/MS response of the LLOQ sample. This could possibly have resulted from carryover of the LC system rather than interference from the blank matrixes. 5. Extraction Recovery. To determine extraction recovery, recovery control solutions were prepared in the reconstitution

mean % CV % theoretical n mean % CV % theoretical n

235.20

5880.00

41160.00

244.80 2.6 104.1 3 216.35 1.3 92.0 3

6139.75 3.8 104.4 3 6108.65 0.8 103.9 3

41942.16 0.2 101.9 3 41940.89 0.0 101.9 3

solvent at known concentrations. Fifty microliters of recovery control solution was added to extracted NCP-KEDTA with internal standard prior to the drying step. After drying, samples were reconstituted as normal. The area ratio (analyte/internal standard) for the recovery controls was then determined at each level and compared to the area ratio obtained from extracted QC samples of the corresponding level. Extraction recovery was calculated by dividing the area ratios of individual QCs by the mean area ratio of the recovery control solutions. Overall mean extraction recovery evaluated at rK5 concentration levels of 5880.0 and 41 160.0 ng/ mL were calculated to be 85 and 72%. Extraction recovery is more than sufficient to achieve accurate, precise, and reproducible results at the LLOQ. 6. Stability. An integral part of method validation is to demonstrate that accurate measurement of the analyte concentration will not be compromised by the stability of the analyte under various stages during the sample analysis process. Stability at a given stage of preparation is specific to storage conditions, matrix, and container systems. The stability of samples subjected to multiple freeze/thaw cycles with a corresponding storage period at room temperature was evaluated by subjecting stability QC samples to conditions that would occur during repeat sample analysis. Freeze/thaw stability QCs were then assayed along with standards and control QCs that had undergone only one freeze/thaw cycle. Freeze/ thaw stability was investigated for samples stored at both -20 and -70 °C. Stability QC samples stored at -20 °C went through four additional freeze/thaw cycles relative to the control QC samples and were exposed to room temperature for 18.25 h. Comparisons of the mean concentrations at each level showed differences between -2.5 and 0.1%. Stability samples stored at -70 °C went through five additional freeze/thaw cycles relative to the control QC samples and were held at room temperature for 30.83 h. Comparisons of the mean concentrations at each level showed differences between -4.8 and 0.6%. The frozen storage stability of NCP-KEDTA samples was evaluated as follows. Multiple sets of stability QC samples were prepared and stored at both -20 and -70 °C. For the initial testing, one set of QC samples was assayed and quantitated using a set of calibration standards prepared on the same date as the QCs. After a documented period of time in frozen storage, the stability QC samples were retested using newly prepared calibration standards. The mean concentrations of each QC level were then compared to the mean concentrations determined from the initial testing. The time difference between the initial testing of the stability QCs and the preparation date of the calibration standards used for retesting was the established frozen storage stability period. For samples stored at -20 °C, stability has been Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

7013

established for at least 147 days, with mean percent differences between -7.1 and 8.3%. Stability for samples stored at -70 °C has also been established for 147 days, with mean percent differences between -1.7 and 8.7%. Frozen storage stability can be extended, if necessary, by preparing new standards and retesting the original stability QC samples. Stability of the analyte in reconstituted solution was also investigated in order to demonstrate stability both before and during injection of an analytical run. A batch consisting of standards and QCs was injected onto the LC-MS/MS. After a period of storage in the autosampler, the batch was reinjected, and the reinjected QCs quantitated using the originally injected standard curve. The time difference between the first injection of the final QC and the reinjection of the final QC was the demonstrated stability period. The mean accuracies of reinjected QCs, expressed as percent theoretical, were between 98.7 and 101.6%, establishing reconstituted stability for at least 72.5 h when stored at ∼10 °C in the autosampler. Stability of the analyte in aqueous solution was also established at both ambient and refrigerated temperatures as part of the validation. Reference standards of rK5 used in the preparation of standard and QC samples were received in a frozen, buffered solution and were not reused after the initial thaw. For the purposes of validation, stock solution stability was assumed to be independent of concentration. However, to account for varying concentrations of buffer, stability was tested for both undiluted stock solutions and stock solutions that were diluted to a concentration approximately equal to that of the least concentrated working solution. Results for the determination of stock solution stability were calculated by comparing mean response ratios (area of response per unit of concentration) of stability solutions to mean response ratios of freshly prepared control solutions, with the acceptance criteria of (5%. Room-temperature stability of rK5 solutions has been established for at least 23 h in both diluted and undiluted states. Refrigerated stability of rK5 has been established for at least 24 h for undiluted solutions. Refrigerated stability for diluted solutions was not investigated, as working solutions are to be freshly prepared for each standard or QC preparation. Room-temperature stability for the internal standard at both diluted and undiluted concentrations was also tested, with stability being established for at least 71 and 48 h, respectively. Assay Performance for Toxicokinetic Studies. This validated assay has been applied to the concentration determination of monkey plasma samples in support of toxicokinetic studies for the development of rK5 as a therapeutic agent. An example rK5 concentration profile after intravenous dosing is shown in Figure 7. The linear dynamic range of the assay provided appropriate coverage of the study sample concentrations without sample dilution. All 11 batches prepared during sample analysis of a toxicity study were within the assay acceptance criteria. Accuracy and precision of QC samples during analysis are shown in Table 4. The mean accuracies of these QC samples ranged from 97.5 to 104.3%, with CV between 1.9 and 5.3%.

7014 Analytical Chemistry, Vol. 75, No. 24, December 15, 2003

Figure 7. Example of a toxicokinetic curve from a toxicology study. Table 4. Statistical Calculation of QC Performance for a Toxicology Study Monkey Sample Analysis quality control level

theor concn (ng/mL) mean % CV % theoretical n

1

2

3

4

5

235.2 229.3 5.3 97.5 27

1176.0 1191.0 3.3 101.3 27

5880.0 6129.9 2.2 104.3 28

17640.0 18146.0 2.8 102.9 28

41160.0 42588.8 1.9 103.5 28

CONCLUSIONS We have described a novel strategy of protein quantitation for the analysis of rK5 in monkey plasma. Rather than follow a traditional immunoassay approach, a 96-well solid-phase extraction technique was employed with subsequent LC-MS/MS detection of the intact protein. Precolumn regeneration and column heating were also integrated into the assay. The assay has been validated and utilized in preclinical studies, with an LLOQ of 99.2 ng/mL obtained from a sample volume of 50 µL, and a corresponding on-column concentration of only 380 fmol (3.97 ng). The assay is rugged and specific and provides accurate and precise data in support of clinical and preclinical trials. Although quantitative measurement of a specific peptide from protein digestion may also meet the needs of protein quantitation in plasma and other biological matrixes, direct monitoring of the intact protein provides a facile and expedient solution. This approach can also be used as a guide in the development of bioanalytical assays for other proteins and, furthermore, for the quantitation of biomarkers discovered in proteomic research. ACKNOWLEDGMENT The authors thank Dr. Don J. Davidson and Dr. Evelyn M. McKeegan for helpful discussions and providing the internal standard.

Received for review August 8, 2003. Accepted October 8, 2003. AC034930N