A high-throughput liquid chromatography/tandem mass spectrometry method for simultaneous quantification of a hydrophobic drug candidate and its hydrophilic metabolite in human urine with a fully automated liquid/ liquid extraction
ABT-869 (A-741439) is an investigational new drug candidate under development by Abbott Laboratories. ABT-869 is hydrophobic, but is oxidized in the body to A-849529, a hydrophilic metabolite that includes both carboxyl and amino groups. Poor solubility of ABT-869 in aqueous matrix causes simultaneous analysis of both ABT-869 and its metabolite within the same extraction and injection to be extremely difficult in human urine. In this paper, a high-performance liquid chromatography/tandem mass spectrometry (HPLC/MS/MS) method has been developed and vali- dated for high-speed simultaneous quantitation of the hydrophobic ABT-869 and its hydrophilic metabolite, A-849529, in human urine. The deuterated internal standards, A-741439D4 and A-849529D4, were used in this method. The disparate properties of the two analytes were mediated by treating samples with acetonitrile, adjusting pH with an extraction buffer, and optimizing the extraction solvent and mobile phase composition. For a 100 mL urine sample volume, the lower limit of quantitation was approximately 1 ng/mL for both ABT-869 and A-849529. The calibration curve was linear from 1.09 to 595.13 ng/mL for ABT-869, and 1.10 to 600.48 ng/mL for A-849529 (r2 > 0.9975 for both ABT-869 and A-849529). Because the method employs simultaneous quantification, high throughput is achieved despite the presence of both a hydrophobic analyte and its hydrophilic metabolite in human urine.
ABT-869 has been demonstrated as a novel ATP-competitive inhibitor of the vascular endothelian growth factors (VEGF), platelet-derived growth factors (PDGF), and receptor tyrosine kinases (RTKs).1 Currently, it is being developed for cancer treatments by Abbott Laboratories.2 A methyl group of ABT-869, shown in Fig. 1(a), is oxidized to form the carboxylic acid metabolite, A-849529, as shown in Fig. 1(b). An assay for the simultaneous determination of both ABT- 869 and A-849529 in human plasma has been reported recently.3 This method has been successfully used to support clinical trials. A direct conversion of the plasma method into a urine method was first attempted, but was unsuccessful. As previously mentioned, ABT-869 is hydrophobic while its metabolite is hydrophilic, which explains why ABT-869 is almost insoluble in human urine but the metabolite is highly soluble. The high ionic strength of urine further intensifies this disparity. Meanwhile, ABT-869 has a tendency to adhere to the storage vessel even if spiked into human urine. Therefore, any transferred aliquot is no longer representative of the entire collected sample, even if aggressive mixing is involved. When the extraction procedure described in the human plasma method was initially used to extract ABT-869 and its metabolite during early development of the urine method, only the metabolite could be extracted successfully. Although addition of surfactants can help improve the solubility of hydrophobic analyte in aqueous matrix, the surfactants can be introduced into the mass spectrometer through sample preparation. Moreover, the use of surfac- tants can significantly suppress the ionization efficiency4 by reducing the surface tension of mobile phase, neutralizing the Coulomb’s repulsion of charges or even quenching the ionization.5 In order to retrieve ABT-869 efficiently, urine samples were treated with acetonitrile in the original storage
prior to analysis.
An accurate and sensitive high-performance liquid chromatography/tandem mass spectrometry (HPLC/MS/ MS) method for the simultaneous quantification of a hydrophobic analyte and its hydrophilic metabolite in human urine has been developed to reduce the compli- cations that arise from their dissimilar chemical properties. The extraction is high throughput, and fully automated in a 96-well format. Mass spectrometric detection has proved sensitive and selective, and the assay has been validated and utilized in support of multiple clinical trials.
EXPERIMENTAL
Materials and reagents
All solutions with an aqueous component, including the HPLC mobile phase, were prepared using purified de- ionized water from a Millipore Milli-Q (Billerica, MA, USA). Formic acid, glacial acetic acid and ammonium hydroxide were from EMD Chemicals, formerly EM Science (Gibbstown, NJ, USA). Ammonium acetate was from J.T. Baker (Phillipsburg, NJ, USA). All of these reagents are A.C.S. grade. Hexanes, methanol, acetonitrile, and ethyl acetate, all HPLC grade, were also supplied by EMD Chemicals. ABT- 869, A-849529, deuterated ABT-869D4 and A-849529D4
(shown in Figs. 1(c) and 1(d)) were synthesized by Abbott Laboratories (North Chicago, IL, USA). Blank human urine was from Biological Specialty Corporation (Colmar, PA, USA).
Instrumentation
The HPLC system consisted of an SIL-HTc autosampler and LC-10AD VP pump from Shimadzu Corporation (Kyoto, Japan). An API-3000 mass spectrometer was from MDS Sciex (Thornhill, Ontario). Data was acquired and processed by Analyst 1.3.2 software, also from MDS Sciex. Watson LIMS, from ThermoElectron Corporation (Waltham, MA, USA), was used for data storage and regression. Chromatographic separation was performed with a SymmetryShieldTM column from Waters (Milford, MA, USA) and a Zorbax guard column from Agilent (Palo Alto, CA, USA). A MicroLab AT Plus 2 automated liquid handler from Hamilton Company (Reno, NV, USA) was used to transfer samples, reagents, internal standards, and extraction solvent. A VX2500 multi- tube vortexer from VWR (West Chester, PA, USA) was used for mixing purposes. A multi-channel evaporator, modified in-house, was used to dry down the organic extract. A centrifuge from Jouan (West Chester, PA, USA) was used to ensure that reconstituted samples were collected in the bottom of each well after mixing.
LC/MS/MS detection
Separation of analytes and internal standards was accom- plished using a Waters SymmetryShieldTM RP8, 5 mm,2.1 × 150 mm analytical column with an Agilent Zorbax 300SB-C8, 5 mM, 2.1 × 20 mm guard column. The isocratic separation used a flow rate of 0.3 mL/min. The mobile phase consisted of 0.1% formic acid in 50:50 acetonitrile/water (v/v). An autosampler needle wash solution containing 50:50 methanol/water was also used. A backwash solution of 0.1% formic acid in 90:10 acetonitrile/water (v/v) was also employed for guard column regeneration after the analytes had benn eluted into the analytical column and the guard column had been taken offline. The backwash system significantly reduces extraction residues, thus protecting the analytical column.
The analytes and internal standards were monitored using an MDS Sciex API 3000 triple quadrupole mass spectrometer with a turboionspray (ESI) interface. The mass spectrometer was operated in a positive ion multiple reaction monitoring (MRM) mode. The source temperature setting was 3008C.
The pressures of the nebulizer and collision gases were set to 8 and 6, respectively. Curtain gas was set to 10. The following precursor-to-product ion reaction channels were monitored with dwell times of 200 ms: m/z 376.1 → 251.3 for ABT-869, m/z 406.1 → 251.3 for A-849529, m/z 380.2 → 255.3 for A-741439D4, and m/z 410.2 → 255.3 for A-849529D4.
High-purity nitrogen was used as the collision gas. Mass spectrometric data acquisition was initiated 2.0 min after the sample was injected, and lasted for 4.4 min. Total chromato- graphic run time was 6.5 min. Figure 2 shows example chromatograms of a reference solution containing approxi- mately 38 ng/mL of ABT-869, 100 ng/mL of A-849529, and 25 ng/mL each of A-741439D4 and A-849529D4.
Data processing and quantification
Peak areas of the analytes and internal standards were obtained using the SCIEX AnalystTM software. A calibration curve was derived from the peak area ratios versus the concentration of the standards. A weighting factor of 1/x2 (where x is the concentration of a given standard) was used for curve fitting. The regression equation for the calibration curve was then used to back-calculate the measured concentrations. For each standard and quality control (QC) sample, the results were compared to the theoretical concentrations to obtain the accuracy, expressed as a % bias from nominal concentration of each level measured. Results from the QC samples were used to verify accuracy and precision of the analytical results for the study samples.
Preparation of standards and QC samples
Stock solutions used for standard and QC preparations were made from independent weighings. Solutions were prepared from solid powders and dissolved with 1:1 (v/v) aceto- nitrile/water. Sonication may be required to completely dissolve ABT-869 into the mixture. Stock solutions were stored in a refrigerator at approximately 48C. In order to spike the pre-diluted mixture of ABT-869 and A-849529 into the drug-free human urine, working solutions of intermedi- ate concentrations were prepared by combining both ABT- 869 and A-849529 stock solutions and diluting with 1:1 (v/v) acetonitrile/water. Standards and QC samples were then prepared by diluting the working solutions in volumetric flasks with drug-free human urine. Ten standard levels ranged in concentration from 1.09 to 595.13 ng/mL for ABT- 869, and 1.10 to 600.48 ng/mL for A-849529. Five QC levels ranged in concentration from 2.09 to 478.13 ng/mL for ABT- 869, and from 2.11 to 482.80 ng/mL for A-849529. For storage, volumes of 0.5 mL of the standard and QC samples were
aliquoted into 2 mL polypropylene test tubes, and were frozen at approximately —708C until use.
Internal standard stock solutions were prepared in an identical fashion to their non-deuterated counterparts. A working internal standard solution of A-741439D4 and A-849529D4 at approximately 1 mg/mL each was then prepared from the stock solutions in 25:75 (v/v) aceto- nitrile/water and refrigerated at approximately 48C.
Liquid/liquid extraction (LLE) procedure
Samples were prepared using a 96-well LLE technique. All liquid transfers except for the initial addition of acetonitrile were performed by the Hamilton Microlab AT2 Plus automated liquid handler. The procedure is summarized as follows.Samples were thawed at room temperature, mixed thoroughly and centrifuged. To the original sample tubes, 125 mL of acetonitrile were added to 0.5 mL of sample, creating a 1:4 mixture of acetonitrile/urine. After thoroughly mixing and centrifuging, 100 mL of each sample were added to the appropriate wells of a 96-well polypropylene plate. Then, 50 mL of internal standard solution were added to each well except the well for the double blank. The samples and internal standard were mixed by aspirating and dispensing 100 mL six times. After addition of 100 mL buffer (100 mM ammonium acetate solution with 0.1% ammonium hydrox- ide) to each well, 1200 mL of extraction solvent (1:11, hexanes/ethyl acetate, v/v) were added. Each well was mixed by aspirating and dispensing 300 mL ten times. Then 900 mL of the organic layer were transferred into a clean 96- well plate after a settling time of approximately 45 s. The supernatant was then evaporated to dryness under a stream of nitrogen at room temperature. The extracts were reconstituted with 200 mL reconstitution solution (1:1, mobile phase/water, v/v). The plate was capped and shaken using a multi-tube vortexer for approximately 3 min for dissolution. After centrifuging at 3000 rpm for approximately 5 min to ensure that reconstituted samples were collected in the bottom of each well, a volume of 40 mL of each sample was injected into the LC/MS/MS system for data acquisition.
RESULTS AND DISCUSSION
Method development A fully automated LLE method in 96-well format was used to extract the analytes from human urine samples. An LC/MS/ MS method was used to generate the chromatographic peak areas of the analytes and internal standard for quantification. Because both ABT-869 and A-849529 have amino groups, they are easily ionized in the positive electrospray ionization (ESI) mode. Due to the chemical complexity of human urine as well as the challenges of absorption and distribution resulting from the disparate properties of the analytes, numerous difficulties were encountered during develop- ment of this method. They are summarized as follows.
Extraction efficiency
ABT-869, exhibiting a high degree of hydrophobicity, is very difficult to dissolve into urine but readily extracts into an organic phase. Conversely, A-849529 is hydrophilic in nature and easily dissolves into urine, but is difficult to extract into an organic phase. It is necessary to select an organic solvent or solvent mixture that simultaneously gives the required extraction efficiencies for both ABT-869 and A-849529. Typically, hexanes are more favorable to non-polar or hydrophobic analytes, while ethyl acetate is more favorable to polar or hydrophilic analytes. Much effort was put into finding an organic mixture with adequate extraction efficiencies for both analytes. The extraction solvent was then optimized by employing a mixture of hexanes and ethyl acetate in a ratio of 1:11 (v/v). The ratio should be strictly controlled to attain the optimal extraction recovery for both analytes.
Effects of storage A test was designed to find out if inconsistencies for ABT-869 were caused by adsorption to the polypropylene tubes used for storage. As an alternative, glass tubes were also tested for sample storage, along with sonication of both polypropylene and glass tubes in an attempt to liberate ABT-869 into the matrix. After up to 10 min of sonication, no improvement was observed for either glass or polypropylene tubes.
Effects of acetonitrile and normal human plasma
It was later found that the addition of either acetonitrile (ACN) or normal human plasma (NHP) into human urine samples significantly improveed the accuracy and precision of ABT-869. However, the QC samples treated with ACN showed slightly better accuracy or smaller %bias than the QC samples treated with NHP. Different ACN/urine ratios from 1:5 to 4:10 were tested and there was no significant %bias difference observed. Therefore, the ratio 1:4 of ACN to urine (v/v) was selected for the sample treatment in this method.
Heat-seal technique
A heat-seal method with vigorous mixing was initially tested in order to increase the extraction recovery. Unfortunately, the technique could not be used due to an emulsion being formed after shaking. This attributed to the high percentage of ethyl acetate in the extraction solvent. Additional residues were produced at the bottom of each well once the so-called supernatant was dried under nitrogen, and this approach also required much more time to dry down prior to reconstitution.
Effects of buffer solution in extraction
Buffer pH also plays a pivotal role in obtaining a clean chromatogram. At first, 100 mM ammonium acetate with 0.2% acetic acid was used as a buffer, identical to that used in the previously described plasma method. Several ‘ghost’ peaks were observed, as shown in Fig. 3(a). These peaks disappeared if a plate was injected after overnight storage in the autosampler (maintained at ~108C). Evidently, to eliminate these ghost peaks, an equilibration time (e.g. overnight) was required, which is unacceptable for regulated work due to the inconsistent nature of the chromatograms. The process would also have had a severe impact on assay throughput. However, after changing the buffer to 100 mM ammonium acetate with 0.1% ammonium hydroxide, the ghost peaks completely disappeared without time for equilibration, and a clean chromatogram could be obtained, as shown in Fig. 3(b).
Method validation results
Validation and samples analysis experiments were designed in accordance with ‘Guidance for Industry, Bioanalytical Method Validation’ issued by the US Food and Drug Administration (FDA) in May 2001.6 This section briefly describes validation results for the simultaneous determination of ABT-869 and its oxidized metabolite, A-849529, in human urine using deuterated analogs as internal standards.
Calibration curves
In order to demonstrate the accuracy and precision of the method, three consecutive batches were extracted and injected. All tested standards were within the acceptance criteria, having biases less than or equal to 15%. Tables 1(a) A-849529. The maximum coefficient of variation (CV) was 3.3% for ABT-869 and 1.5% for A-849529. The coefficient of determination (r2) was greater than or equal to 0.997516 for ABT-869 and 0.999473 for A-849529.
Quantitation limits
The standard with the lowest concentration is used as the lower limit of quantitation (LLOQ). The LLOQs were 1.09 ng/mL for ABT-869 and 1.10 ng/mL for A-849529. Three consecutive batches with six replicates each were used for LLOQ evaluation, although LLOQ samples were not used to construct the calibration curve. The CV at the LLOQ was 7.5% with a mean bias of —3.0% for ABT-869. The CV was 4.5% with a mean bias of —0.3% for A-849529. A statistical summary of LLOQ results is shown in Table 2.
The standard with the highest concentration is used as the upper limit of quantitation (ULOQ). The ULOQs were 595.13 ng/mL for ABT-869 and 600.48 ng/mL for A-849529. Three consecutive batches with six replicates each were used for ULOQ evaluation, and were not used to construct the calibration curve. The CV at the ULOQ was 3.0% with a mean bias of —2.3% for ABT-869. The CV was 2.0% with a mean bias of —1.4% for A-849529. A statistical summary of ULOQ values is shown in Table 3.
Accuracy and precision of QCs
Five levels of QC samples were prepared. Three consecutive batches were extracted and injected, with six replicates at each included. Statistical summaries of QC results are listed in Tables 4(a) and 4(b) for ABT-869 and A-849529, respectively. A total of 87 of 90 ABT-869 QC samples and 90 of 90 of A-849529 QC sample were within the acceptance criteria of 15%. The maximum CV was 5.4% for ABT-869 and 3.9% for A-849529. The mean bias was between —10.8% and 1.6% for ABT-869 and between —1.9% and 8.1% for A-849529.
Selectivity of the method
Six lots of matrix were screened for interference from endogenous matrix components. The blank matrix samples were extracted both with and without addition of internal standard. The area response of the LLOQ was compared to the responses of the blank matrices at the retention of analytes. For ABT-869, six of six lots with internal standard had a signal-to-noise ratio (S/N) greater than 5 or did not have integrateable peaks, and five of six lots without internal standard had S/N greater than 5 or did not have integrate- able peaks. For A-849529, six of six lots with internal standard had S/N greater than 5 or did not have integrate- able peaks, and six of six lots without internal standard had S/N greater than 5 or did not have integrateable peaks. Figure 4 presents the chromatogram for an LLOQ sample, while Fig. 5 shows the chromatogram of blank matrix without internal standard added. No significant interference was observed from blank human urine samples.
Stability
Due to the irreversibility of addition of ACN into the human urine samples, two types of stability QC samples were required: normal QC samples and ACN-treated QC samples. The normal QC samples were prepared in normal human urine, and stored as such during the stability testing period. ACN-treated QC samples were normal QC samples with ACN added in the ratio of 1:4 ACN/human urine, as described above, and were stored as such during stability
testing. The normal QC samples demonstrate sample stability prior to the first extraction, while ACN-treated QC samples demonstrate sample stability after ACN is added. It should also be noted that no additional ACN is added to ACN-treated QC samples in future extractions.
Freeze/thaw (F/T) and short-term room temperature evaluation
The normal F/T QC stability samples went through at least one additional F/T cycle before analysis to cover a possible power loss or other unintentional thaw. The ACN-treated F/T QC samples were prepared as follows: after the normal QC sample had gone through one F/T cycle, ACN was added into the normal QC samples in a ratio 1:4. The ACN-treated QC samples went through at least three more F/T cycles before analysis. A record was kept to determine the time the samples were exposed to room temperature. The concen- trations of F/T QC samples were calculated using a calibration curve and were compared to the concentrations of the control samples, which did not undergo any additional F/T cycles. The frozen temperature was approximately
—708C. The % differences were determined by comparing the means of the calculated F/T QC concentrations with the mean concentrations of the controls.
Tables 5(a) and 5(b) show the results for F/T stability QC samples. No significant change in concentration of the QC samples was found during the experiment. Stability has been established for both ABT-869 and A-849529 in human urine at room temperature for 11 h and four F/T cycles for ACN- treated F/T QC samples, and 6 h and two F/T cycles for normal QC samples.
Frozen storage evaluation
Normal frozen storage QC samples were placed in the freezer after preparation, and were stored at approximately —708C. The ACN-treated frozen storage QC samples were prepared by putting normal QC samples through one F/T cycle, and upon thawing ACN was added in a 1:4 ACN/ urine ratio as described above.
For stability controls, one set of QC samples was analyzed to generate day one concentrations for frozen storage stability. The same control values were used to evaluate both normal frozen storage QC samples and ACN-treated frozen storage QC samples. The date of initial testing was used for day one values of normal QC samples, while for ACN-treated QC samples, the date of ACN addition was designated as day one. After a documented period of time, the QC samples from the same preparation as the controls were tested for frozen storage stability using a set of freshly prepared calibration standards and QC samples.
The date of preparation for the standards used during stability testing was used as the endpoint for stability testing. It was found that both ABT-869 and A-849529 in human urine were stable for at least 33 days for normal QC samples, and 13 days for ACN-treated QC samples when stored at
approximately —708C.
Autosampler stability and batch storage stability were established for at least 100 h and 102 h for both ABT-869 and A-849529 at approximately 108C.
Stock solution stability at both room temperature and refrigerator temperature was established previously.3
Matrix effect
Matrix effect experiments were conducted to demonstrate that assay performance is independent of variations in sample matrix. Matrix effect QC samples were prepared in six different lots of matrix at a concentration approximately equal to the lowest QC sample. The matrix effect QC samples were treated as unknowns and analyzed. The measured concentrations of these matrix effect QC samples were compared with their theoretical values. The maximum CV was 5.2% for ABT-869 and 8.4% for A-849529. Mean biases were between —8.2% and —1.0% for ABT-869 and between 1.9% and 13.8% for A-849529. From this, it was determined that there is no significant matrix effect for either ABT-869 or A-849529 in human urine.
Extraction recovery
The extraction recovery of the analytes from urine was estimated by comparing the instrument response of extracted samples (regular run QC samples) with solutions of the same concentrations that did not undergo extraction. Mean extraction recoveries were 49% and 35% for ABT-869 and A-849529, respectively. The recovery was sufficient to obtain reproducible and accurate quantification at the desired LLOQ. These comparable extraction recovery values can be attributed to the use of an extraction solvent that was optimized for simultaneous extraction of the two analytes, as discussed above.
Application to clinical studies
This method has been successfully applied to the determi- nation of ABT-869 and A-849529 in human urine for two phase I studies, which evaluate the safety and pharmaco- kinetics of ABT-869. An additional objective of these studies was to obtain a preliminary assessment of anti-tumor activity in subjects with advanced non-hematologic malignancies.
CONCLUSIONS
A sensitive, reliable, and robust HPLC/MS/MS method for the simultaneous determination of ABT-869 and its oxidized metabolite, A-849529, in human urine has been successfully developed and validated while maintaining high assay throughput. The method simultaneously accommodates the hydrophobicity of ABT-869 and the hydrophilicity of its metabolite in the presence of a complex urine matrix. The extraction procedure was adapted into a fully automatic liquid/liquid extraction in 96-well format. The method has been successfully applied to the quantification of ABT-869 and A-849529 in clinical trials. This method can also be used as a model in the development of other assays to simultaneously quantitate analytes with significantly differ- ent behaviors Linifanib in a given matrix.