Quantification of Gemcitabine Incorporation into Human DNA by LC

Jul 13, 2010 - Eli Lilly and Company, Indianapolis, Indiana 46285, and Advion BioServices Inc., Ithaca, New York 14850. Anal. Chem. , 2010, 82 (15), ...
0 downloads 0 Views 638KB Size
Download PDF
Anal. Chem. 2010, 82, 6576–6583

Quantification of Gemcitabine Incorporation into Human DNA by LC/MS/MS as a Surrogate Measure for Target Engagement Enaksha R. Wickremsinhe,*,† Barry S. Lutzke,† Barry R. Jones,‡ Gary A. Schultz,‡ Angela B. Freeman,† Susan E. Pratt,† Angela M. Bones,† and Bradley L. Ackermann† Eli Lilly and Company, Indianapolis, Indiana 46285, and Advion BioServices Inc., Ithaca, New York 14850 In this study, we report a method for direct determination of gemcitabine incorporation into human DNA. Gemcitabine (dFdC), a structural analog of the nucleoside deoxycytidine (dC), derives its primary antitumor activity through interruption of DNA synthesis. Unlike other surrogate measures, DNA incorporation provides a mechanistic end point useful for dose optimization. DNA samples (ca. 25 µg) were hydrolyzed using a two-step enzymatic procedure to release dFdC which was subsequently quantified by LC-ESI-MS/MS using stable isotope labeled internal standards and selected reaction monitoring (SRM). dFdC was quantitated and reported relative to deoxyguanosine (dG) since dG is the complementary base for both dFdC and dC. The SRM channel for dG was detuned using collision energy as the attenuating parameter in order to accommodate the difference in relative abundance for these two analytes (>104) and enable simultaneous quantification from the same injection. The assay was shown to be independent of the amount of DNA analyzed. The method was validated for clinical use using a 3 day procedure assessing precision, accuracy, stability, selectivity, and robustness. The validated ranges for dFdC and dG were 5-7500 pg/mL and 0.1-150 µg/mL, respectively. Results are presented which confirm that the ratio of dFdC to dG in DNA isolated from tumor cells incubated with dFdC increases with increased exposure to the drug and that dFdC can also be quantified from DNA extracted from blood. Gemcitabine (Gemzar, difluorodeoxycytidine, 2′,2′-difluorodeoxycytidine, dFdC) is a pyrimidine antimetabolite with a broad spectrum of antitumor activity against several human malignancies including pancreatic,1,2 ovarian,3,4 lung,5,6 breast,7,8 and bladder.9,10 It is phosphorylated intracellularly by deoxycytidine kinase to * To whom correspondence should be addressed. E-mail: [email protected]. Fax: 317-433-6432. † Eli Lilly and Company. ‡ Advion BioServices Inc. (1) Burris, H. A., III; Moore, M. J.; Andersen, J.; Green, M. R.; Rothenberg, M. L.; Modiano, M. R.; Cripps, M. C.; Portenoy, R. K.; Storniolo, A. M.; Tarassoff, P.; Nelson, R.; Dorr, F. A.; Stephens, C. D.; Von Hoff, D. D. J. Clin. Oncol. 1997, 15, 2403–2413. (2) Hilbig, A.; Oettle, H. Expert Rev. Anticancer Ther. 2008, 8, 511–523. (3) Lorusso, D.; Di Stefano, A.; Fanfani, F.; Scambia, G. Ann. Oncol. 2006, 17, 188–194.

6576

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

dFdC triphosphate (dFdCTP) and gets incorporated into DNA leading to apoptosis.11 dFdC is a close structural analog of the nucleoside deoxycytidine (dC). Therefore, both dFdC and dC can be incorporated into DNA by forming hydrogen bonds with dG, the complementary base, during DNA synthesis. dFdC is also rapidly and extensively metabolized by cytidine deaminase to its primary metabolite, 2′,2′-difluoro-deoxyuridine (dFdU).12 In order to optimize gemcitabine therapy, a predictable correlation between drug exposure and efficacy is required. Unfortunately, plasma concentrations of gemcitabine do not allow for sufficient dose optimization because it does not represent the exposure at the target (tumor) level. Alternatively, investigators have resorted to monitor intracellular levels of its active form, dFdCTP, in peripheral blood mononuclear cells (PBMC).13-16 However, this still does not provide a direct measure of the intracellular concentration at the target tissue. Moreover, these methods require tedious processing at the clinical sites that involve multiple centrifugation steps to isolate PBMCs and the use of perchloric acid to stabilize dFdCTP. Another issue with dFdCTP measurement is the need to normalize the results to account for differences in sample size. When analyzing dFdCTP from PBMCs, the amount of dFdCTP is typically reported normalized to the total number of cells taken for analysis or the total protein in the PBMC sample. The quantification of total protein has a greater tendency to introduce imprecision into the overall measurement due to frequent con(4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15)

(16)

Thigpen, T. Semin. Oncol. 2006, 33, S26–S32. Mornex, F.; Girard, N. Ann. Oncol. 2006, 17, 1743–1747. Favaretto, A. G. Ann. Oncol. 2006, 17, 82–85. Silvestris, N.; Cinieri, S.; La Torre, I.; Pezzella, G.; Numico, G.; Orlando, L.; Lorusso, V. Breast 2008, 17, 220–226. Dent, S.; Messersmith, H.; Trudeau, M. Breast Cancer Res. Treat. 2008, 108, 319–331. Bellmunt, J.; Albiol, S.; de Olano, A. R.; Pujadas, J.; Maroto, P. Ann. Oncol. 2006, 17, 113–117. El Karak, F.; Flechon, A. Expert Opin. Pharmacother. 2007, 8, 3251–3256. Plunkett, W.; Peng, H.; Yi-Zhend, X.; Heinemann, V.; Grunewald, R.; Varsha, G. Semin. Oncol. 1995, 22, 3–10. Sparidans, R. W.; Crul, M.; Schellens, J. H. M.; Beijnen, J. H. J. Chromatogr., B 2002, 780, 423–430. Losa, R.; Sierra, M. I.; Blay, P.; Blanco, D.; Buesa, J. M. Chromatographia 2004, 59, 493–496. Nishi, R.; Yamauchi, T.; Ueda, T. Cancer Sci. 2006, 97, 1274–1278. Veltkamp, S. A.; Hillebrand, M. J. X.; Rosing, H.; Jansen, R. S.; Wickremsinhe, E. R.; Perkins, E. J.; Schellens, J. H. M.; Beijnen, J. H. J. Mass Spectrom. 2006, 4, 1633–1642. Abbruzzese, J. L.; Grunewald, R.; Weeks, E. A.; Gravel, D.; Adams, T.; Nowak, B.; Mineishi, S.; Tarassoff, P.; Satterlee, W.; Raber, M. N.; Plunkett, W. J. Clin. Oncol. 1991, 9, 491–498. 10.1021/ac100984h  2010 American Chemical Society Published on Web 07/13/2010

Figure 1. Schematic showing the implementation of this procedure for clinical trials.

tamination of the PBMC sample with plasma proteins. Although the concentration of dFdCTP (in PBMCs) has been quantified in many clinical studies, the data have failed to show a clinical correlation.17-20 Therefore, the most direct technique would be to be able to directly measure dFdC incorporated into tumor DNA. In addition to providing a direct method to assess target engagement, this approach would be highly versatile since DNA isolated from any source could be analyzed (e.g., blood, tissue, tumor biopsy, etc.). This communication reports methodology developed for quantitating the amount of dFdC that is incorporated into DNA (target engagement) along with the validation of the technique to support clinical application. Key attributes of this methodology include: (1) the use of standard methods to extract and hydrolyze DNA, (2) the use of LC/MS/MS analysis to achieve the sensitivity, selectivity, and precision needed for clinical application, (3) normalizing the amount of dFdC released relative to the native nucleoside dG, and (4) simultaneous quantification of both analytes by LC/MS/MS by extensive detuning of the signal for dG. On the basis of the results presented, this method is believed to represent an improved surrogate for quantitating dFdC target engagement. A schematic of the processes involved in the implementation of this procedure to support clinical trials is illustrated in Figure 1. EXPERIMENTAL SECTION Materials and Reagents. Methanol and HPLC-grade bottled water were from Burdick and Jackson (Morristown, NJ). Trifluoroacetic acid (TFA) was from J. T. Baker (Phillipsburg, NJ). Ammonium formate was from Sigma-Aldrich (St. Louis, MO). Magnesium chloride 6-hydrate (MgCl2), Teknova Tris-HCl (pH 6.8), and Teknova Tris-HCl (pH 8.3) were from Fisher Scientific (Pittsburgh, PA). Diethylpyrocarbonate (DEPC) treated water was from Invitrogen (Carlsbad, CA). Gemcitabine (17) Grunewald, R.; Kantarjian, H.; Du, M.; Faucher, K.; Tarassoff, P.; Plunkett, W. J. Clin. Oncol. 1992, 10, 406–413. (18) Touroutoglou, N.; Gravel, D.; Raber, M. N.; Plunkett, W.; Abbruzzese, J. L. Ann. Oncol. 1998, 9, 1003–1008. (19) Patel, S. R.; Gandhi, V.; Jenkins, J.; Papadopolous, N.; Burgess, M. A.; Plager, C.; Plunkett, W.; Benjamin, R. S. J. Clin. Oncol. 2001, 19, 3483–3489. (20) Licea-Perez, H.; Wang, S.; Szapacs, M. E. Steroids 2008, 7, 601–610.

(dFdC) and the stable-isotope labeled internal standard of dFdC (dFdC-IS) were synthesized at Eli Lilly and Co (Indianapolis, IN). Deoxyguanosine (dG), deoxycytidine (dC), deoxyadenosine (dA), and thymidine (T) were from Sigma-Aldrich (St. Louis, MO). The stable-isotope labeled internal standard of deoxyguanosine (dG-IS) was from Medical Isotopes, Inc. (Pelham, NH). 1.5 mL nonstick flip-cap microfuge tubes were from Phenix Brand (Candler, NC). The structures of the key analytes are shown in Figure 2. Preparation of dFdC, dG, dFdC-IS, and dG-IS Stock and Working Solutions. Stock solutions of dFdC and the dFdC-IS were prepared in HPLC grade water at 0.1 mg/mL concentration. Stock solutions of dG and dG-IS were prepared in HPLC grade water at 1.0 mg/mL concentration. Stocks of dFdC and dG were prepared from separate weighings for the standard curve and validation/QC samples, respectively. Stock solutions of dC, dA, and T were prepared in HPLC grade water at 2.0 mg/mL concentration. All stock solutions were stored in 1.5 mL nonstick microfuge tubes and stored at approximately -20 °C. A working solution, combining dA, dC, dG, T, and dFdC, was prepared using the stock solutions to yield 150 µg/mL dA, dC, dG, and T and 7.5 ng/mL dFdC. Hydration buffer (mixture of 3 mL of Tris-EDTA buffer (pH 8) and 20 mL of DEPC water) was used as the diluent. This 1:1:1:1 mixture of dG, dA, dC, and T served as the proxy matrix for this assay. An intermediate stock solution of dFdC-IS was prepared by diluting the 1 mg/mL stock with HPLC grade water at 2.5 µg/mL. A combined IS working solution was made using the 1 mg/mL dG-IS and the 2.5 µg/mL dFdC-IS resulting in and 5 µg/mL dG-IS and 50 ng/mL dFdC-IS. Preparation of Standard Curve and QC/Validation Samples. The standard curve samples were prepared in duplicate for each analytical run. The dilution scheme for the preparation of the standard curve samples and the validation/QC samples is summarized in Table 1. The amount of dC, dA, and T in the standard curve, validation, and QC samples were similar to the amount of dG in order to mimick the approximately 1:1:1:1 ratio of the four bases in DNA. The hydration buffer was used as the diluent. Selectivity zero samples were prepared by combining dC, dA, and T at a concentration of 150 µg/mL along with dG-IS and dFdCIS. Selectivity double blank samples were prepared by combining Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6577

Figure 2. Structures of gemcitabine (dFdC), deoxyguanosine (dG), stable isotope labeled gemcitabine (dFdC-IS), and deoxyguanosine (dGIS) and the respective fragmentation yielding the product ion used for quantification. Table 1. Dilution Scheme for the Preparation of Standard Curve (STD) and Quality Control (QC) Samples STD or QC ID

concentration of dGa, dAb, dCc, and Td (ng/mL)

concentration of dFdCe (pg/mL)

Std 9 Std 8 Std 7 Std 6 Std 5 Std 4 Std 3 Std 2 Std 1 QC ULOQf QC 3 QC 2 intermediate QC QC 1 QC LLOQg

150 000 100 000 50 000 25 000 10 000 2500 1000 200 100 150 000 120 000 75 000 5000 400 100

7500 5000 2500 1250 500 125 50 10 5 7500 6000 3750 250 20 5

a Deoxyguanosine. b Deoxyadenosine. c Deoxycytidine. quantification.

d

Thymidine. e Difluorodeoxy cytidine. f Upper limit of quantification. g Lower limit of

dC, dA, and T at a concentration of 150 µg/mL. These samples were analyzed in duplicate for each run to determine the presence of any interference that would affect the quantitation of dFdC and dG. Preparation of Hydrolysis Reagent Mixture. Hydrolysis mix I was prepared by combining 11 µL of 10 mM MgCl2, 15 µL of 10 mM Tris-HCl (pH 6.8), and 4 µL of deoxyribonuclease I 6578

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

(4000 units/mL, Sigma-Aldrich, St. Louis, MO) per DNA sample to be hydrolyzed. Hydrolysis mix 2 was prepared by combining 12.5 µL of 100 mM Tris-HCl (pH 8.3), 3.75 µL of snake venom phosphodiesterase (124 units/mL, Worthington Biochemical Corporation, Lakewood, NJ), and 3.75 µL of bacterial alkaline phosphatase (200 units/mL, Sigma-Aldrich, St. Louis, MO) per DNA sample to be hydrolyzed. Multiple

Figure 3. Fragment ion profile for dG showing the most abundant ions (top) and the relationship between collision energy and detector response for 152 m/z ion (bottom).

volumes were prepared at a time, as needed. DNA Hydrolysis. Aliquots of DNA (5-50 µg) were placed in 1.4 mL polypropylene tubes or a polypropylene 96 well block (Matrix Technologies Inc., Hudson, New Hampshire) and diluted with DEPC water to 115 µL. The DNA was denatured by heating at 95 °C for 5 min followed by chilling for 3 min. Twenty µL of the internal standard working solution and 30 µL of the hydrolysis mix 1 was added and incubated at 37 °C for 3 h. Then, 20 µL of the hydrolysis mix 2 was added and incubated overnight (minimum of 16 h) at 37 °C. Heating, chilling, and incubations were performed using a GeneAmp PCR System 9700 thermocycler (Applied BioSystems, Foster City, CA). The samples were centrifuged for 3 min at approximately 3200, and the supernatant was used for LC/MS/MS analysis. Aliquots of the DNA sample prior to hydrolysis as well as following hydrolysis were visualized on a 2% agarose E-gel using a low range quantitative DNA ladder (from Invitrogen, Carlsbad, CA) and quantified using a Nanodrop ND-1000 microvolume spectrophotometer (NanoDrop, Wilmington, DE) as a check to ensure proper sample handling and processing. Sample Preparation (Non-DNA Samples, Proxy Matrix). Aliquots (115 µL) of each non-DNA sample (standard, QC, blank, etc.) were placed into a polypropylene 96 well block. Thirty µL of the hydrolysis mix 1, 20 µL of the hydrolysis mix 2, and 20 µL of the internal standard working solution were added; the block was capped and vortexed vigorously. The resulting cocktail was used for LC/MS/MS analysis.

In Vitro Experiments. HEK 293 cells (ATCC, Manassas, VA) were grown in DMEM media supplemented with 1 mM pyruvate and 10% FBS in T150 flasks (all cell culture products including PBS were from Invitrogen, Carlsbad, CA). When cells were approximately 50-60% confluent, they were exposed to dFdC concentrations ranging from 0.025 to 10 µM, for approximately 24 h. The cells were harvested by removing the media followed by washing 3 times with PBS. The control (untreated) cell pellet was spiked with 2 µL of 1 mM dFdC to evaluate the efficiency of the DNA extraction procedure to remove nonincorporated dFdC from the final DNA sample. DNA was extracted from these cells using the Qiagen AllPrep DNA/RNA Kit, following the manufacturer’s procedure (Qiagen Inc., Valencia, CA). Additionally, cell lines A549, HepG2, Calu6, HCT116, and HEK293 (ATCC, Manassas, VA) were also cultured as described above. When cells were approximately 50-60% confluent, they were exposed to 0.1, 1.0, and 10 µM dFdC for approximately 24 h in the presence of tetrahydrouridine (THU, from Calbiochem, LaJolla, CA) to inhibit the deactivation of dFdC by cytidine deaminase. The cells were harvested; the control (untreated) cell pellets were spiked with 2 µL of 1 mM dFdC, and DNA was extracted as describe previously. In Vivo Experiments. Three male beagle dogs were given a single 25 mg/kg intravenous bolus dose (formulated in 0.9% sodium chloride injection, USP) of gemcitabine via the jugular vein. The study was conducted in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council, 1996) and all applicable Eli Lilly and Company policies and guidelines. Six mL aliquots of blood were collected into EDTA containing vacutainer tubes from each dog at predose and at approximately 72, 96, 120, 144, 168, and 240 h post dose. The time points were selected on the basis of a previous pilot experiment. DNA was isolated from blood with the Qiagen/Gentra Puregene Blood Kit Plus protocol using the red blood cell lysis solution, cell lysis solution, RNase solution, and the protein precipitation solution (all from Qiagen Inc., Valencia, CA). DNA Hydrolysis Experiments. Pooled aliquots of DNA obtained from the dog experiment were transferred to microfuge tubes, in triplicate, to represent 10, 15, and 25 µg DNA. These were diluted with DEPC water to 115 µL and subjected to hydrolysis. DNA Stability Experiments. A sample of DNA that contained incorporated dFdC was required to conduct stability studies. This was accomplished by obtaining DNA from a patient undergoing treatment with Gemzar (gemcitabine) under an open IRB protocol for Receipt of Human Biospecimens for Research established with the Methodist Hospital (Indianapolis, IN). DNA was isolated from blood with the Qiagen/Gentra Puregene Blood Kit Plus protocol described above. The DNA was reconstituted in the hydration buffer at 0.6 µg/ mL and pipetted as 25 µg aliquots (41.7 µL) into microfuge tubes. Triplicate samples were immediately stored at -20 °C to evaluate frozen storage stability at approximately 2, 6, and 12 months; three were stored at room temperature for 24 h benchtop stability, and three were frozen and thawed three times to demonstrate freeze/ thaw stability. Reinjection reproducibility for hydrolyzed dFdC and dG samples was evaluated by storing processed samples at room temperature for 52 h. Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6579

For all stability samples, the ratios of gemcitabine (in pg/mL) to deoxyguanosine (in µg/mL) were compared and expressed as relative error (RE %) to the pooled human DNA sample at time zero. All stability samples were assayed in replicates of three. LC System and Chromatographic Conditions. A Shimadzu HPLC system consisting of LC-10AD vp pumps, SCL-10AVP pump controller, DGU-20A5 online degassing system, and FCV-12AH 6-port diverter valve (Shimadzu Co, Columbia, MD) were used in combination with a LEAP autosampler (LEAP Scientific Co. Carrboro, NC). A Phenomenex Synergi Hydro-RP (2 mm × 50 mm, 4 µm) HPLC column (Phenomenex, Torrance, CA) was used. Mobile phase solvent A consisted of HPLC-grade bottled water/1 M ammonium formate/TFA (1000/10/1 v/v), and solvent B consisted of methanol/1 M ammonium formate/TFA (1000/10/1 v/v). The flow rate was 0.6 mL/min; the column temperature was held at ambient, and the injection volume was 25 µL. Analytes were separated using a linear gradient starting from 2% solvent B to 15% solvent B in 3 min, following which the effluent was diverted to waste and washed at 80% solvent B for 1.5 min and re-equilibrated at 2% solvent B for 2 min. Mass Spectrometric Conditions. Mass spectrometric data were generated using an API 4000 triple quadrupole mass spectrometer and acquired using Analyst Software, v 1.4 (Applied BioSystems, Foster City, CA). Full scan and selected reaction monitoring (SRM) acquisitions were performed at unit resolution using positive ion atmospheric pressure ionization at a source temperature of approximately 500-600 °C and an IonSpray voltage of approximately 3500 V. UHP nitrogen was used as the nebulizer gas, curtain gas, collision gas, and turboionspray gas. Full scan and product ion spectra were acquired for each analyte and its internal standard via direct infusion. The instrument conditions were set up to enable the quantification of both dFdC and dG during a single chromatographic run. In each case, the transitions monitored corresponded to the favorable loss of the ribose moiety to yield the protonated nucleotide base (Figure 2). The SRM transitions (±0.2 m/z) and the corresponding collision energy (eV) are as follows: dFdC 264.1f112.1 (19 eV), dFdC-IS 269.1f117.1 (19 eV), dG 268.1f152.1 (60 eV), and dG-IS 283.1f162.1 (55 eV). The exact masses to be monitored are optimized after calibration and tuning. RESULTS AND DISCUSSION We have developed a methodology that enables quantification of gemcitabine (dFdC) following its incorporation into DNA that serves as a direct measure of its target engagement. The importance of quantifying dFdC along with dG was critical since dG served as a surrogate internal standard to track any discrepancies or inaccuracies during sample collection, processing, and hydrolysis. This also eliminated the need to accurately measure the quantity of DNA used for the assay, as long as the quantity was within the specified acceptable range. The amount of dFdC in each sample was normalized to the amount of dG and was expressed as picograms dFdC per microgram dG. dG was selected since it is the complementary base for dFdC as well as dC. This assay can easily be adopted for use at multiple clinical sites, and the quantification can be performed on DNA isolated from any source (blood, tissues, tumor biopsies, etc.). Establishing LC/MS/MS Conditions. Preliminary experiments had shown that the amount of dFdC incorporated into DNA 6580

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

Figure 4. Chromatograms representing the lower limit of quantification for dFdC at 5 pg/mL (top) and dG at 0.1 µg/mL (middle) and the stable isotope labeled internal standards for dFdC and dG (bottom). All chromatograms are plotted as time (minutes) against mass spectrometric response. The nominal transitions monitored (m/z): dFdC 264f112, dG 268f152, dFdC-IS 269f117, dG-IS 283f162. X-axis is time (minutes) and Y-axis is intensity (cps).

could range between 1 and 50 molecules per million molecules of dG. This presented a challenge of quantitating dFdC and dG in a single assay due to the saturation of the mass spectrometric response by the overwhelming abundance of dG. This was addressed by detuning the response for dG using collision energy (CE). Detuning CE enabled the deattenuation of the dG SRM transition by as much as 70-fold (Figure 3) and resulted in achieving a linear response over the entire calibration range for dG (0.1-150 µg/mL). It is noted that the CE for dG-IS was detuned to a slightly lesser degree in order to conserve the amount of dG-IS used during analysis. This approach was significantly more efficient compared to the use of a lesser abundant fragment ion (Figure 3) or varying the declustering potential (DP). Varying the DP between 0 and 400 V resulted in only a 2-fold change in response. Analogous strategies, including the monitoring of a lesser abundant SRM transition, quantification based on the natural isotope etc. have been successfully used for quantitating biomarkers and other endogenous molecules present in high concentrations.21,22 The dynamic range of the assay was established on the basis of preliminary experiments (data not shown) using DNA from in vitro grown cells treated with dFdC. The chromatographic profile (21) Midttun, O.; Hustad, S.; Ueland, P. M. Rapid Commun. Mass Spectrom. 2009, 23, 1371–1379. (22) Bioanalytical Method Validation, Guidance for Industry, Food and Drug Administration, 2001; http://www.fda.gov/cder/guidance/index.htm.

Table 2. Stability of dFdC and dG in the Presence of Proxy Matrix and Hydrolysis Reagents dFdC (pg/mL) 24 h room temperature 24 h at 37 °C 52 h reinjection reproducibility

a

dG (µg/mL)

nominal conc

mean

% REa

nominal conc

mean

% REa

20 6000 20 6000 5 3750 7500

19.47 5517 19.78 5371 4.84 3631 7299

-2.6 -8.0 -1.1 -10.5 -3.2 -3.2 -2.7

0.4 120 0.4 120 0.1 75 150

0.38 110.1 0.39 108.1 0.104 72.65 142.1

-4.3 -8.3 -1.3 -9.9 4.0 -3.1 -5.3

% RE ) (mean - nominal conc)/mean × 100.

Table 3. Stability of dFdC and dG Incorporated into DNA dFdC (pg/mL) replicates day zero 24 h room temperature 3 cycles freeze/thaw 65 days frozen 202 days frozen 375 days frozen a

6 3 3 3 3 3

mean

%REa

336.19 na 335.14 -0.31 334.42 -0.53 334.33 -3.59 339.47 0.97 296.41 -12.58

dG (µg/mL) mean

%REa

36.67 36.11 36.97 36.48 33.30 33.23

na -1.52 0.84 -0.52 -9.63 -9.85

% RE ) (mean - mean day zero)/mean day zero × 100.

representing the lowest standard (5 pg/mL dFdC and 0.1 µg/mL dG) and the corresponding stable-labeled internal standards are shown as Figure 4. The standard at the LLOQ (standard 1) was used as the system suitability injection before each analysis to assess sensitivity of the instrumentation before sample acquisition. There were no interferences observed in the blank samples that corresponded with the analyte retention times that could have interfered with the quantification. The amount of dG in each sample is relative to the amount of DNA in the original sample. Therefore, in order to accurately mimic this condition, the amounts of dC, dA, and T in the standard curve, validation, and QC samples were prepared to be equal to the concentration of dG. This unique feature adds to the accuracy and robustness of the overall assay, compared to the traditional use of a constant concentration of the proxy matrix. Validation. The interassay precision (% relative error, RE) and accuracy (% relative standard deviation, RSD) for dFdC and dG was established at the lower limit of quantification (LLOQ), the upper limit of quantification (ULOQ), and a midpoint concentration. Overall, the interassay accuracy for both analytes was between 1 and 9% RSD, and the precision was between +6 and -11% RE, falling well within the limits set forth by the FDA.23 Additionally, there was no difference between the data generated using proxy matrix samples processed with and without the hydrolysis procedure, indicating no impact on the individual nucleosides or its analogs as a result of the hydrolysis procedure. Stability. The stability of dFdC and dG was evaluated in the proxy matrix and in the presence of hydrolysis mixes 1 and 2. The results show that both analytes were stable at both room temperature and 37 °C over a period of approximately 24 h (Table 2). The final extracts used for LC/MS/MS were also stable at room temperature for approximately 52 h, thus enabling the reinjection of the samples for reanalysis (if needed). (23) Cohen, R. M.; Wolfenden, R. J. Biol. Chem. 1971, 246, 7561–7565.

Figure 5. Relationship between the amount of DNA used for analysis and the resulting concentrations of dFdC (Y-axis) and dG (Z-axis).

The stability of dFdC and dG incorporated into DNA was evaluated using a pooled DNA sample that contained incorporated dFdC. The results show that both analytes were stable for 24 h at room temperature, stable through 3 freeze/thaw cycles, and stable over 12 months of frozen storage at -20 °C (Table 3). The criteria for acceptable stability are defined as falling within 20% of the nominal or time zero concentration. Hydrolysis Efficiency. The efficiency of the hydrolysis and release of individual nucleosides was evaluated by hydrolyzing different amounts of a pooled DNA sample containing incorporated dFdC. The results showed a linear increase in the amount of dFdC and dG with increasing amounts of DNA as depicted in Figure 5, and the ratio of dFdC to dG, calculated as pg of dFdC per µg of dG, was constant across the range. This demonstrates that the amount of DNA needed for the analysis does not need to be accurate as long as it is between 10 and 25 µg per sample. In a previous experiment using genomic human liver DNA (purchased from BioChain, Hayward, CA), we demonstrated a linear increase in the amount of dG released from 5 µg up to 50 µg of DNA (data not shown). This information is extremely important from a perspective of collecting and extracting DNA at multiple clinical sites or laboratories, since accurate quantification of each DNA sample is not critical, as long as an approximate concentration is known to proceed with the hydrolysis. The presence of DNA in the sample prior to hydrolysis as well as successful digestion of the DNA following treatment with the hydrolysis enzyme mix was visualized on an agarose gel prior to LC/MS/MS analysis, to ensure proper sample preparation. A low DNA mass ladder corresponding to a 100-2000 base pair range was used as a calibrant. The gels showed the presence of DNA Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6581

Table 4. Incorporation of dFdC into HCT116 Cell DNA Following 24 h Incubation with Varying Concentrations of dFdCa

a

treatment (µM dFdC)

dFdC conc (pg/mL)

dG conc (µg/mL)

pg of dFdC per µg of dG

control control 0.025 0.025 0.1 0.1 0.25 0.25 1 1 2.5 2.5 10 10

no peak no peak 55.4 44.5 189 176 524 349 717 1110 1330 1190 579 883

47.3 27.3 51.7 44.5 46 42.2 43.1 40.5 26.4 39.7 39.2 42.2 25.4 39.8

1.1 1.0 4.1 4.2 12.2 8.6 27.2 28.0 33.9 28.2 22.8 22.2

Experiment was conducted in duplicate.

Figure 6. Incorporation of dFdC into HEK293, HepG2, Calu6, HCT116, and A549 cell DNA following 24 h incubation with 0.1, 1, and 10 µM concentrations of dFdC.

in the sample prehydrolysis and the absence of DNA posthydrolysis, demonstrating successful digestion of the DNA. In Vitro Studies. Incorporation of dFdC into the DNA of HEK293 cells exposed to dFdC for approximately 24 h showed a linear dose response relationship from 0.025 µM up to approximately a 1 µM dFdC concentration (Table 4). The range of dFdC and dG concentrations from all samples fell within the assay range (5-7500 pg/mL for dFdC and 0.1-150 µg/mL for dG). Data from five human tumor cell lines cultured in vitro and exposed to different concentrations of dFdC are summarized in Figure 6. All the cell lines showed an increase in the amount of dFdC that was incorporated into the DNA when incubated in the presence of increasing concentrations of dFdC (from 0.1 to 10 µM). The difference in the amounts of dFdC incorporated across the cell lines seem to correlate with the differences in the doubling times for each cell line (HCT116 and A549 have relatively fast doubling times while HepG2 and HEK293 have longer doubling times). The amount of dFdC incorporated into the HEK293 cells in this experiment was higher compared to the previous experiment. This is most likely due to the fact that these five cell lines were exposed to dFdC in the presence of THU. THU is a potent inhibitor of cytidine deaminase, the enzyme responsible for rapid 6582

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

Figure 7. Incorporation of dFdC into DNA from dog blood following a single 25 mg/kg intravenous administration of gemcitabine (n ) 3 per time point).

inactivation of dFdC12,24. Therefore, incubating dFdC in the presence of THU increases the availability of dFdC. In both of these experiments, the control samples were spiked with dFdC prior to DNA extraction. Analysis of these samples did not show any detectable quantities of dFdC. This data shows the complete removal of all nonincorporated dFdC from the final DNA sample that is subjected to hydrolysis, thereby ensuring that the dFdC quantified by LC/MS/MS is what was released from the DNA following the treatment with the hydrolysis enzyme mix. Additionally, this also enables the isolation of DNA from whole blood without having to be concerned with contamination of the DNA sample with free dFdC that could be in the systemic circulation at the time of the blood collection. In Vivo Studies. The exposure profile for dFdC incorporated into the DNA isolated from dog blood is depicted in Figure 7. This experiment demonstrated the feasibility of the use of blood to extract DNA and quantify incorporated dFdC and also validated the sensitivity of the assay to detect and quantify the pharmacodynamic range. The highest concentration (Cmax) of incorporated dFdC in dog blood was around 5 to 6 days post dose, which is immediately following the nadir for the absolute neutrophil count (data not shown).

In summary, we have demonstrated a novel approach to quantitate the engagement of dFdC with its cellular target. The key aspects of this technique include (1) the quantification of dG as a surrogate and the normalization of the dFdC concentrations per microgram of dG, (2) the ability to quantify both analytes (dFdC and dG) within the same analytical run, despite the greater than 4-orders of magnitude difference in the relative concentrations, (3) the successful use of a proxy matrix consisting of dC, dA, and T, (4) the demonstration that any nonincorporated dFdC is efficiently removed from the final DNA sample, (5) the demonstration that it is not critical to use a fixed amount of DNA for the (hydrolysis) analysis; therefore, there is no need to quantify the DNA prior to hydrolysis, and (6) there are no stability issues which enables the shipping of DNA samples collected from multiple sources and sites to a central laboratory for analysis.

The data generated using this technique is being used to develop a pharmacokinetic/pharmacodynamic (PK/PD) model that will help understand the efficacy/toxicity and guide the best dosing regimen for clinical development of a gemcitabine (dFdC) prodrug that can be administered orally. ACKNOWLEDGMENT The author wishes to thank Drs. Thomas Daly, Ignacio GarciaRibas, and Anne Dantzig for their scientific inputs and stimulating discussion during the development of this technique.

Received for review April 14, 2010. Accepted June 9, 2010. AC100984H

Analytical Chemistry, Vol. 82, No. 15, August 1, 2010

6583