Anal. Chem. 2007, 79, 1599-1603
Impact of Differential Recovery in Bioanalysis: The Example of Bortezomib in Whole Blood Adam H. Brockman,* Panos Hatsis, Martin Paton, and Jing-Tao Wu
Department of Drug Metabolism and Pharmacokinetics, Millennium Pharmaceuticals, Inc., 45 Sidney Street, Cambridge, Massachusetts 02139
A key assumption in pharmaceutical bioanalysis is that spiked standards mimic incurred samples in every analytical aspect. Although deviations from this assumption have been reported in terms of the difference in ion suppression or metabolite interference, the difference of extraction recovery and its impact has been rarely reported and is often characterized as unlikely. In this work, we demonstrated the presence and significance of differential recovery using a real-world example: the assay of bortezomib in whole blood. Recovery differences of up to 10-fold were observed between the spiked standards and the incurred samples when different extraction methods were used. Because of its high impact, it is important that the potential of differential recovery between standards and incurred samples be evaluated during method validation. A simple time course incubation experiment was proposed to screen compounds for potential differential recovery during method validation in heterogeneous matrixes, such as whole blood and tissue. The use of this approach and the interpretation of the results from this experiment were demonstrated using bortezomib in whole blood as an example. The differential recovery of bortezomib is likely to be driven by slow binding to the proteosome present in red blood cells. Spiked samples, however, do not have sufficient time for binding to occur. The bioanalytical subdiscipline of drug metabolism and pharmacokinetics is relied upon to provide ultratrace measurements of small-molecule drugs in biological matrixes. These measurements are heavily scrutinized and regulated as they form the basis for the assessment of safety margin and pharmacokinetic/ pharmacodynamic (PK/PD) relationship. The industrial standard for the basis of measurement is described in the FDA’s guidance document, Bioanalytical Method Validation.1 The common approach to bioanalysis is an assay at the nanogram per milliliter range (ppb) based on liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). This assessment is always conducted with some risk based on the major underlying assump* To whom correspondence should be addressed. Current address: Merck Research Laboratories-Boston, 33 Ave. Louis Pasteur, Boston, MA 02215. E-mail:
[email protected]. Tel: 617.930.0375. Fax: 617-820-5127. (1) Bioanalytical Method Validation; Center for Drug Evaluation and Research, U.S. Food and Drug Administration, U.S. Department of Health and Human Services: Rockville, MD, 2001. 10.1021/ac061680c CCC: $37.00 Published on Web 12/29/2006
© 2007 American Chemical Society
tion of the approach,2 namely, that the response of the spiked samples used for standard curve and quality control sample preparation mirrors the response of in vivo samples. Needless to say, the failure of this assumption would have a broad impact on the assessment of PK/PD relationship and safety margin for pharmaceuticals. Unfortunately, this underlying assumption of bioanalytical chemistry is most often taken for granted. It is illustrative, therefore, to explore certain scenarios where this assumption would fail. For instance, if the in vivo-derived, i.e., incurred, sample presents vastly different suppression/enhancement ionization effects than the spiked matrix and the internal standard does not effectively compensate for this effect, then accepted analytical results could be misleading. We could refer to this effect as a “differential matrix effect”, and it has been widely observed in the literature.3-6 Accordingly, the assessment and elimination of this effect is a standard part of bioanalytical method development. In addition, if metabolism were reversed by the method then measured parent species would be elevated,7-9 resulting in a “differential metabolism effect”. Finally, if the recovery of the analyte is markedly different from incurred samples than from spiked samples, then we can likewise refer to this phenomenon as “differential recovery”. To the best of our knowledge, there are limited or very few examples of this specific effect in the literature,10 particularly in whole blood. All three of these categories of observed deviations can result in misleading bioanalytical results, and they all have common attributes. First, these deviations cannot be disproved easily because doing so requires proof of a negative, i.e., that there are no differential effects in the bioanalytical method. Second, all of (2) James, C. A.; Breda, M.; Frigerio, E. J. Pharm. Biomed. Anal. 2004, 35, 887-893. (3) Bonfiglio, R.; King, R. C.; Olah, T. V.; Merkle, K. Rapid Commun. Mass Sepctrom. 1999, 13, 1175-1185. (4) Buhrman, D.; Price, P.; Rudewicz, P. J. Am. Soc. Mass Spectrom. 1996, 7, 1099-1105. (5) Matuszewski, B. K.; Constanzer, M. L.; Chavez-Eng, C. M. Anal. Chem. 2003, 75, 3019-3030. (6) King, R. C.; Bonfiglio, R.; Fernandez-Metzler, C.; Miller-Stein, C.; Olah, T. V. J. Am. Soc. Mass Spectrom. 2000, 11, 942-950. (7) Constanzer, M. L.; Chavez, C. M.; Matuszewski, B. K.; Carlin, J.; Graham, D. J. Chromatogr., B 1997, 693, 117-129. (8) Matuszewski, B. K.; Chavez, C. M.; Constanzer, M. L. J. Chromatogr., B 1998, 716, 195-208. (9) Jemal, M.; Ouyang, Z.; Powell, M. L. Rapid Commun. Mass Sepctrom. 2002, 16, 1538-1547. (10) James, C. A.; Breda, M.; Baratte, S.; Casati, M.; Grassi, S.; Pellagatta, B.; Sarati, S.; Frigerio, E. Chromatographia 2004, 59, S149-S156.
Analytical Chemistry, Vol. 79, No. 4, February 15, 2007 1599
Table 1. Extraction Methods and Conditions Used for Bortezomib Analysis
Figure 1. Structure of bortezomib.
these deviations require incurred samples in order to test the negative hypothesis. In view of the fact that we are trained in our field to prepare a bioanalytical method well in advance of in vivo sample generation, these differential effects present a formidable challenge. A common practice that has been adopted is to validate our methods as best we can using previous experience and theory and then to refine the methods after incurred samples have arrived. In this article, we present the determination of bortezomib in whole blood as a real-world example of differential recovery. We also advocate a more systematic approach to evaluate the potential of differential recovery during method validation in the absence of incurred samples. It should be noted that bortezomib is designed to bind strongly to the proteosome, which is present in millimolar quantities within red blood cells. Therefore, the differential recovery of bortezomib is likely to be caused by slow binding to the proteosome in in vivo-derived red blood cells. Spiked standards or quality control samples are generally not incubated for sufficient time to effect this binding. Furthermore, differential recovery has not been widely observed in matrixes such as plasma and serum, and this phenomenon may be more likely in heterogeneous matrixes such as whole blood and tissue. EXPERIMENTAL SECTION Chemicals and Reagents. Bortezomib is an FDA-approved drug for multiple myeloma from Millennium Pharmaceuticals. The structure of bortezomib is shown in Figure 1 and is unlike most marketed drugs in that the boronic acid functionality is present in the structure. Boronic acids are weakly acidic, and they are known to form “dative bonds” (weak covalent bonds) with diols. In addition, bortezomib is known to be chemically unstable in many matrixes, so it was handled with additional precautions as noted below. HPLC grade acetonitrile, methyl tert-butyl ether (MTBE) and water were purchased from EMD Chemicals (Gibbstown, NJ). Mobile phases for liquid chromatography were mixtures of acetonitrile and water acidified with 0.1% formic acid (EMD Chemicals). Some of the extractions developed in this study employed hydrochloric acid, which was obtained from Mallinckrodt Baker, Inc. (Phillipsburg, NJ). The in vivo rat and mouse samples used in this study were from Millennium’s animal facilities, and blank mouse, rat, or monkey whole blood (K3EDTA anticoagulant) was from Lampire Biological Laboratories (Pipersville, PA). Whole blood was treated according to the simulated conditions under study. Incubation experiments attempting to capture the behavior of the compound in vivo used fresh-drawn whole blood with intact red blood cells. The blank whole blood used for in vitro comparisons was stored at 4 °C. Incurred samples were generally frozen for no longer than 1 or 2 weeks 1600 Analytical Chemistry, Vol. 79, No. 4, February 15, 2007
1 2 3 4
protein precipitation extraction (PPE)
liquid-liquid extraction (LLE)
unbuffered 0.1% formic acid 1% formic acid 0.1 M HCl
unbuffered 0.1% formic acid 1% formic acid 0.1 M HCl
Instrumentation. All experiments presented in this paper were performed on an Applied Biosystems/MDS Sciex API4000 triple quadrupole mass spectrometer (Concord, ON, Canada) using electrospray ionization in the positive ion mode. The mass spectrometer was operated at an electrospray voltage of 5000 V and a declustering potential of 80 V. The Turbo-V heat injectors were operated at 600 °C. The analysis of bortezomib was performed in multiple reaction monitoring mode (MRM) using the precursor/fragment ion transition m/z 367.2/226.2 with a collision offset voltage of 26 V and a CAD gas setting of 10 (arbitrary units). A separate MRM channel was used to monitor the transition m/z 376.1/234.2 for C13-labeled bortezomib, which was used as the internal standard (IS). Both MRM transitions used a 100-ms dwell time, a 5-ms pause, unit resolution for Q1, and unit resolution for Q3. Liquid chromatography was performed with an Agilent 1100 series binary pump and mobile-phase degasser (Agilent Technologies, Palo Alto, CA). A total of 20 µL of sample was injected onto a Betasil C18 column (30 × 2.1 mm i.d., Thermo Electron Corp., Waltham, MA) using a CTC PAL autosampler (Leap Technologies, Carrboro, NC). Elution of bortezomib was achieved using gradient elution with water/acetonitrile mobile phases. Sample Preparation. A wide range of extraction methods were developed and tested for the analysis of bortezomib in whole blood. All methods, however, addressed the limited stability of bortezomib in the same way. Whole blood was always collected onto K3EDTA anticoagulant, which was found to improve stability compared to heparin. Furthermore, all extractions were performed on wet ice to minimize instability. Samples were subjected to at least one freeze/thaw cycle in order to lyse blood cells. The extraction methods presented below are summarized in Table 1. Protein precipitation extraction (PPE) was performed using a 4:1 ratio of acetonitrile (containing IS) to sample. The effect of pH on protein precipitation was explored by using the acetonitrile as is, i.e., unbuffered, acidified with 0.1 or 1% formic acid or acidified with 0.1 M hydrochloric acid. Samples were then vortexed for 5 min and centrifuged at 3000g for 10 min. A total of 100 µL of supernatant was evaporated and reconstituted in 100 µL of mobile phase. (Refer to Table 1.) Liquid-liquid extraction (LLE) was performed using 50 µL of sample, 50 µL of IS in acetonitrile, and 2000 µL of MTBE. As was the case with protein precipitation, the pH of the mixture was adjusted using 500 µL of pure water, i.e., unbuffered conditions, water acidified with 0.1 or 1% formic acid, or water acidified with 0.1 M hydrochloric acid. The mixture was then vortexed for 5 min and centrifuged at 3000g for 10 min. Approximately 1500 µL of supernatant was taken, evaporated, and reconstituted to 100 µL with mobile phase. (Refer to Table 1.)
Figure 2. Comparison of recovery and ion suppression for spiked and incurred blood samples. Recovery was calculated as outlined in the Experimental Section.
Determination of Recovery. For the purposes of this paper, traditional recovery refers to the characterization of recovery using spiked reagents, as is commonly reported in the literature.5 Both integrated peak area response and analyte-to-internal standard ratio were utilized for the calculation of recovery. Recovery was calculated by taking a ratio of the response of an extracted spiked sample to a postextraction spiked blank (the extraction residue was reconstituted in injection solvent containing the IS). Ionization suppression/enhancement was assessed by comparing a sample prepared in mobile phase and the postextraction spiked blank. Incurred sample recovery was measured by pooling a series of incurred whole blood samples from a cyno monkey study to create a larger volume of incurred sample. The pooled sample was used as an example to characterize the differences between the various extraction procedures outlined above and in Table 1. Incurred sample recovery was calculated on a relative basis by determining the ratio of the concentration of bortezomib in whole blood using each extraction method, to the average concentration of bortezomib in the pooled sample. The calculation of the concentration of the pooled sample is based on the assumption that LLE with 0.1 M HCl would give 100% recovery. Whether or not this is the case is not important, but rather the trend in
recovery in going from one set of extraction conditions to the next is what matters. It is worth mentioning that the list of extraction procedures explored in this work is not complete, but nonetheless represents a list that is exhaustive within the approaches known to be reliable for bortezomib analysis. Time Course Incubations. Bortezomib (100 nM) was incubated with fresh rat whole blood in order to simulate physiological conditions and study multiple extraction conditions. Aliquots of the incubation mixture were drawn at predetermined time points and extracted using each extraction procedure according to Table 1. Plots of area ratio (signal obtained for bortezomib to that obtained for C13-labeled bortezomib) versus time were constructed following LC/MS analysis to determine whether differential recovery had taken place over the course of the incubation. Intact, fresh red blood cells were utilized in the whole blood incubations, and physiological conditions were approximated by incubating at 37 °C. Whole Blood Pellet Washing. A whole blood pellet washing experiment was undertaken in order to further illustrate the differential recovery effect observed for bortezomib. If bortezomib is truly susceptible to differential recovery from blood, then unbuffered PPE should show high recovery in spiked samples and low recovery in incurred samples relative to LLE with 0.1 M Analytical Chemistry, Vol. 79, No. 4, February 15, 2007
1601
Table 2. Ten Monkey Whole Blood Samples Were Created by Pooling Multiple Study Time Pointsa Velcade peak area ratio
% theoretical (method C control)
method method method sample A 1001A 0.1222 1001 B 0.1289 1002A 0.1706 1002B 0.1416 1003A 0.1490 1003B 0.1267 2003A 0.0703 2003B 0.0791 2004A 0.0479 2004B 0.0534 10 ng/mL 0.1223 QC average
B 1.4213 1.5418 1.6632 1.5345 1.6509 1.7000 0.8891 1.2121 0.5192 0.7958 0.1646
C 1.2491 1.3248 1.3213 1.1869 1.0499 1.2408 0.7412 0.9954 0.4436 0.6335 0.1313
method method sample 1001A 1001 B 1002A 1002B 1003A 1003B 2003A 2003B 2004A 2004B 10 ng/mL QC
A -90.2 -90.3 -87.1 -88.1 -85.8 -89.8 -90.5 -92.1 -89.2 -91.6 -6.9
B 13.8 16.4 25.9 29.3 57.2 37.0 20.0 21.8 17.0 25.6 25.3
-89.5
26.4
Figure 3. Effect of a differential recovery effect on calibration curves.
a
Method A is PPE with 0.1% formic acid, method B is LLE with 0.1% formic acid, and method C is LLE with 0.1 M HCl. The results were normalized to method C.
HCl. Spiked and incurred rat whole blood samples were subjected to an initial unbuffered PPE using the procedure described above. The resulting pellet was washed with water to remove excess bortezomib and IS. The pellet was then resuspended in water using sonication and was subjected to another two cycles of PPE. The final extraction was performed using LLE with 0.1 M HCl, and the supernatants from each extraction were analyzed by LC/ MS. Data analysis was performed by taking the ratio of the signal obtained for bortezomib to that obtained for C13-labeled bortezomib. Presentation of the data was facilitated by normalizing to the maximum ratio obtained for spiked and incurred samples. RESULTS AND DISCUSSION Differential Recovery of Bortezomib in Whole Blood. Differential recovery effects pose a unique challenge during method development/validation since they generally cannot be determined using a traditional recovery experiment. The reason for this is that differential recovery effects are only observed in incurred samples, whereas traditional recovery employs spiked samples and postextraction spiked blanks. We have formulated a simple validation experiment to determine whether or not differential recovery is observed for a given method with reasonable certainty. Samples from different subjects could be used individually or pooled together, and a series of extractions could be performed to assess the possibility of a differential recovery effect. This is what was done in this work, and the results are shown in Figure 2. The incurred sample results presented in Figure 2 are normalized to the value obtained using LLE with 0.1 M HCL since incurred sample recovery does not have an absolute reference point. Examination of Figure 2 reveals that the spiked sample recovery increases from a low of 47% with unbuffered PPE, to 71% with LLE and 0.1 M HCl. This does not entirely explain the typical increase observed with actual study samples, which ranges from 5- to 10-fold using different methods (data not shown). This discrepancy, however, is accounted for when the incurred sample recovery for unbuffered PPE is compared to that for LLE with 0.1 M HCl. In this case, a greater than 5-fold increase in recovery 1602 Analytical Chemistry, Vol. 79, No. 4, February 15, 2007
Figure 4. Whole blood pellet washing experiment with spiked and incurred samples to illustrate the differential recovery effect for bortezomib in rat whole blood.
is observed, which is more in line with observed results. Clearly, these observations show the need for incurred samples to more accurately measure the true recovery of a bioanalytical method. LLE with 0.1 M HCl was chosen as the preferred method for the analysis of bortezomib in blood as a result of the high recovery and low ion suppression. Unbuffered PPE was compared to LLE with 0.1% formic acid for more than seven in vivo mouse, rat, and monkey whole blood pharmacokinetic studies. We summarize the monkey results for three conditions here. The results are presented in Table 2 and show a factor of ∼10 difference, but with a high degree of variability. The reason for the large discrepancy between the two methods is a result of the difference in recovery between spiked calibration/quality control samples and incurred samples. This is further illustrated in Figure 3, which shows the effect of spiked standards on the analytical standard curve, for the case when spiked sample recovery does not accurately reflect incurred sample recovery. If the spiked standards have a higher recovery than incurred samples, then the slope of the standard curve is increased above what it should be, which results in underestimation of the levels of bortezomib in blood. The discrepancy between spiked sample and incurred sample recovery is further demonstrated with the whole blood pellet washing experiment. Figure 4 shows the amount of bortezomib recovered after successive unbuffered PPE steps for spiked and incurred samples. Unbuffered PPE extracted a large amount of bortezomib from spiked samples after the first extraction, and each subsequent extraction including LLE with 0.1 M HCl only
Figure 5. Effect of incubation time and extraction conditions on the extraction of bortezomib from rat whole blood.
extracted less than 5% of the amount from the first extraction. In contrast, the amount of bortezomib extracted from incurred samples was relatively similar among three consecutive extractions with unbuffered PPE, suggesting a slow extraction rate. It is worth mentioning that incurred samples were monitored for known metabolites to assess the possibility of metabolite conversion; however, no evidence was found to support this possibility (data not shown). The fact that bortezomib can be extracted after multiple pellet washing and subsequent re-extraction cycles with unbuffered PPE is consistent with the finding that no metabolite conversion is involved. The use of LLE with 0.1 M HCl successfully released the majority amount of bortezomib from incurred samples. Systematic Approach To Detect Differential Recovery: Time Course Incubation Experiments. The above experiments clearly demonstrated the impact of differential recovery and the analytical approach to investigate and correct any differential recovery. However, this process is very labor intensive and requires the use of incurred samples; therefore, it is not practical to adopt this approach for every compound during the method validation stage. In this work, we propose a simple incubation experiment that can be used to screen for differential recovery during the method validation stage without the use of incurred samples. In this approach, a spiked sample in the matrix is prepared and incubated at 37 °C for a period of time, e.g., 0.5 h, and the response of this sample is compared with that of a freshly spiked sample. The purpose of the incubation is to allow any binding or cell penetration process to occur as if they would in vivo. If the response is similar between the two samples, the risk of any differential recovery can be considered as minimal and no further investigation needs to be carried out. If the responses of the
samples are significantly different, this should trigger an investigation into the stability or differential recovery of the compound. It is worth mentioning that we cannot differentiate between degradation and differential recovery in this time course incubation experiment alone. Additional experiments need to be conducted to evaluate whether this difference in response can be mitigated by varying the extraction conditions, which is a characteristic sign of differential recovery. An example of this investigation is demonstrated with the time course incubation experiment of bortezomib in whole blood. Figure 5 shows the disappearance as a function of time for bortezomib incubated in whole blood at 37 °C. The effect of various extraction procedures on the observed disappearance of bortezomib is also shown. Clearly, the type of extraction procedure used to extract bortezomib from blood has a profound impact on the observed recovery. For instance, unbuffered PPE is unable to extract bortezomib from blood, and almost all of the bortezomib is lost after ∼10 h. In contrast, PPE with 0.1 M HCl gives almost 100% recovery of bortezomib compared to the start of the incubation experiment. This demonstrates the disappearance of bortezomib was due to differential recovery when an unbuffered PPE method was used. CONCLUSIONS The case of a high-impact differential recovery was demonstrated with the assay for bortezomib in whole blood. An exhaustive search of extraction conditions was used to investigate and correct the differential recovery of bortezomib in whole blood. Recovery differences of up to 10-fold between spiked and incurred samples were observed using different extraction methods. To mitigate the risk of differential recovery, a simple time course incubation experiment was proposed to screen compounds during the method validation stage without the use of incurred samples. The use of this approach was demonstrated in the bortezomib example. ACKNOWLEDGMENT The authors thank Li Yu, Gerald Miwa, Raj Nagaraja, and J. Scott Daniels in DMPK at Millennium for their support and helpful suggestions for this work. We also thank Tom Veraeghe, Alex Hemeryck, and their colleagues at J&J Beers, Belgium, for providing the original strong acid LLE method for bortezomib and valuable comments on the manuscript. Finally, thanks are also due to Ed Brewer and Jean Pineault of Tandem Labs for analyzing the samples used to prepare Table 2.
Received for review November 16, 2006.
September
6,
2006.
Accepted
AC061680C
Analytical Chemistry, Vol. 79, No. 4, February 15, 2007
1603
