The analytes evaluated for this method were the following: EMPA (ethyl methylphosphonic acid, CAS 1832-53-7), the metabolite of VX (O-ethyl S-(2-diisopropylaminoethyl) methylphosphonothioate, CAS 50782-69-9); IMPA (isopropyl methylphosphonic acid, CAS 1832-54-8), the metabolite of GB (isopropyl methylphosphonofluoridate CAS 107-44-8); PMPA (pinacolyl methylphosphonic acid, CAS 616-52-4), the metabolite of GD (pinacolyl methylphosphonofluoridate, CAS 96-64-0); CMPA (cyclohexyl methylphosphonic acid, CAS 1932-60-1), the metabolite of GF (cyclohexyl methylphosphonofluoridate, CAS 329-99-7); MMPA (2-(methyl)propyl methylphosphonic acid, CAS 1604-38-2), the metabolite of VR (O-2-(methyl)propyl S-2-(diethylaminoethyl) methylphosphonothioate. CAS 159939-87-4).

The internal standards were isotopically labeled as follows: EMPA, ethyl-D₅; IMPA, isopropyl-¹³C₃; PMPA, trimethylpropyl-¹³C₆; CMPA, cyclohexyl-¹³C₆; IMPA, methylphosphonyl-¹³C. The structures of the organophosphorus metabolites and internal standards are shown in Figure 1.

The analytical calibrators were prepared at concentrations of 0.5, 1.0, 2.0, 5.0, 10, 25, 50, 100 ng/mL of each analyte in serum; the quality control samples were prepared at 4.0 and 40 ng/mL of each analyte in serum. A primary stock (20 μg/mL of each analyte in methanol) purchased from Cerilliant Corporation (Round Rock, TX) was used to make stock solution I (4.0 μg/mL) and stock solution II (0.2 μg/mL). Stock solution I was prepared by aliquoting 400 μL 20 μg/mL stock to a glass vial containing 1600 μL of serum. The second stock solution was prepared by pipetting 100 μL of stock solution I to a glass vial containing 1900 μL of serum. Calibrators 0.5, 1.0, 2.0, and 5.0 ng/mL were made by aliquoting 25, 50, 100, 250 μL, respectively, of stock solution II to 10.0 mL volumetric flasks. Calibrators 10, 25, 50, and 100 ng/mL were prepared by delivering 25, 63, 125, and 250 μL, respectively, of stock solution I to their appropriate 10.0 mL volumetric flask. Quality control low (4.0 ng/mL) and quality control high (40 ng/mL) were prepared by adding 200 μL of stock solution II and 100 μL of stock solution I, respectively, to 10.0 mL volumetric flasks. An additional solution was prepared at 0.25 ng/mL in serum to evaluate the limits of detection for the assay, by delivering 17 μL of stock solution II to a 10.0 mL volumetric flask. Spiked samples, 0.75 and 3.0 ng/mL, were prepared by aliquoting 37.5, and 150 μL of stock solution II to their respective 10.0 mL volumetric flask. Spiked samples, 15, 70, and 90 ng/mL were prepared by delivering 37.5, 175, and 225 μL of stock solution I into their respective 10.0 mL volumetric flasks. All volumetric flasks were filled to 10.0 mL using serum purchased from Tennessee Blood Services Corporation (Memphis, TN). The serum was pooled from five donors, shipped at 4°C, received two days later, and subsequently stored at −20°C until use. Once prepared, all fortified samples, calibrators, and quality control samples were maintained at −20°C.

Internal standard was prepared by dilution of a 500 ng/mL aqueous mixture (Cerilliant Corporation, Round Rock, TX) to a concentration of 23.8 ng/mL in DI water.

Organic-free 18.2 MΩ Type I water from a purifier purchased from Aqua Solutions, Inc. (Jasper, GA) was used in these studies. Pelletized 98% ammonium fluoride was purchased from Alfa Aesar (Ward Hill, MA) and molecular biology grade DEPC-treated 5M ammonium acetate from CalBiochem (La Jolla, CA). HPLC-grade acetonitrile and methanol were purchased from Tedia (Fairfield, Ohio). All human serum was purchased from Tennessee Blood Services Corporation (Memphis, TN).

To prepare samples for extraction, 25 μL of 23.8 ng/mL internal standard was aliquoted into each well of a 2 ml 96-well NUNC plate. Fifty microliters of well-mixed serum was added to the internal standard, followed by 1000 μl of acetonitrile. The plate was sealed using a sheet of Thermo Scientific Adhesive PCR Sealing Foil (Hudson, NH) and vortexed for 5 minutes on a ThermoLab Systems Wellmix (Hudson, NH). The NUNC plate was then centrifuged for 5 minutes at 3000 rpm.

The solid phase extraction was automated by using the Life Science Zephyr (Hopkinton, MA). The Phenomenex Strata Si-1 SPE 96-well plate (55 μM, 70 Å, 100 mg) was pretreated with 1000 μl of 25% water in acetonitrile, followed by 1000 μl of acetonitrile. The entire sample mixture with precipitate was then loaded to the SPE 96-well plate, and rinsed by a two-step process of 1000 μl of acetonitrile, followed by 1000 μl of 7% water in acetonitrile. The solutions were pulled through the SPE 96-well plate using vacuum. The analytes were then eluted with 1000 μl of 28% water in acetonitrile and collected in a clean 96-well NUNC plate.

The eluent was heated at 70 °C under 25 psi of nitrogen in a TurboVap for 30 minutes and then evaporated to dryness after increasing flow rate to 70 psi. The lower initial flow rate eliminated solvent splashing. The sample was then reconstituted in 100 μl of 5% water in acetonitrile, mixed by pipetting twice, and transferred to a 150 μl 96-well PCR plate and sealed.

The HPLC separation was performed using an Agilent 1200 HPLC with a well-plate autosampler (Santa Clara, CA). A Waters Atlantis® HILIC 2.1-mm × 50-mm with 3-μm particles (70% porosity)HPLC column was used (Milford, MA). This hydrophilic interaction chromatography (HILIC) column consists of high purity, non-bonded silica particles. The injector was programmed to draw 20 μL of sample and wash the injector needle for 10 s in the wash port using 50% water in acetonitrile. The sample was then injected onto the column using an isocratic mobile phase consisting of 92% acetonitrile and 8% 1.0 mM ammonium fluoride with an initial flow rate of 500 μL/min. Following elution of the analytes, the flow rate was increased to 1000 μL/min at 3.01 min to remove any late eluting impurities. The flow rate returned to 500 μL/min at 5.01 min to ensure stable pressure for the following injection. This method allows a 5-min injection-to-injection cycle time.

The mass spectrometric analysis was performed using an API 4000 triple quadrupole QTrap mass spectrometer from Applied Biosystems (Foster City, CA) controlled by Analyst software. The mass spectrometer was operated in multiple-reaction-monitoring (MRM) mode using negative electrospray ionization with assisted heating. The specific operating conditions are listed in Table 1 with the proposed fragment ions.

The specific settings used were curtain gas (CUR), 10 psi; nebulizer gas (GS1), 40 psi; turbo gas (GS2), 40 psi; GS2 temperature (TEM), 550 °C; collision gas, nitrogen; collision gas (CAD), Medium, producing a gas pressure reading of 3.3 × 10⁻⁵ Torr; ionspray potential (IS), −4500 V; entrance potential (EP), −10; interface heater (IHE), on.

The data were analyzed using Analyst 1.5.2, which was provided with the instrument. This software is used to review the chromatograms for retention times, baselines and possible interferences. Quantitative analysis of the data by automated and manual integrations, linear regression, and calculation of accuracies and correlation coefficients was also performed with this software package. The chromatographic data were smoothed three times prior to integration with a bunching factor between 1 and 3, and fitted by linear regression using 1/x weighting.

The techniques and materials in this method do not pose any special hazards. General considerations include exercising universal precautions, such as wearing appropriate personal protective equipment, when handling chemicals and serum samples. The high voltage employed in electrospray ionization should also be considered a hazard, and the safety interlocks provided by the instrument manufacturer should not be altered. For instrument-specific safety concerns, please consult the manufacturer.

Initial evaluation of a protein precipitation using acetonitrile or acetone for the sample preparation of these compounds from serum resulted in lower sensitivities than desired. As exposures may be small and/or a sample may be collected many hours to days following exposure, it is important to have the highest sensitivity possible. Additionally, the presence of interfering matrix components was detected in select transitions, which would negatively impact the specificity of the analysis. To accomplish the goals of sensitivity and specificity, additional sample preparation was necessary.

Solid phase extraction was selected for sample preparation due to automation and high throughput potentials. The initial solvents and volumes used for the extraction were selected from Swaim, et al [10] from urine sample extraction. The sorbent was conditioned with 1000 μL of 25% water in acetonitrile, followed by 1000 μL of acetonitrile. After the sample mixture, comprised of 100 μL of serum, 25 μL of internal standard solution and 1000 μL of acetonitrile) was loaded, the sorbent was rinsed with 1000 μL of acetonitrile, followed by 1000 μL of 10% water in acetonitrile. Finally, the compounds were eluted using 1000 μL of 25% water in acetonitrile and collected in a 2 mL 96-well NUNC plate. Using a serum matrix fortified at 10 ng/mL, each step within the solid phase extraction protocol was then evaluated in triplicate for optimal recovery. To evaluate the loading step, 50 μL of serum was diluted with acetonitrile, with additional deionized water as needed, to yield an aqueous solution of 7, 12 and 17%. The seven percent aqueous loading solution resulted in the highest responses with no detected breakthrough. Next, the second wash step was assessed at acetonitrile content ranging from 93 – 83%. The wash step of 93% acetonitrile yielded the highest recoveries with minimal losses detected. No adjustments were made to the elution composition as complete elution was achieved in one step with the 25% water/75% acetonitrile mixture. Final extraction recoveries, determined at 10.0 ng/mL and calculated by the ratio of the measured concentration to the spiked concentration, were as follows: EMPA, 88% (standard deviation of 17), IMPA, 76%(13) MMPA , 92% (9.6)), PMPA, 94% (10), and CMPA , 95% (7.7). With the optimized solid phase extraction parameters, no analyte was detected within the two captured wash steps. Following the elution, a second elution step was evaluated to remove the analytes still remaining on the SPE. This second elution showed minimal (<1%) amounts of PMPA, CMPA and MMPA, approximately 5% of EMPA and no detectable IMPA remained on the SPE following the initial elution.

Previous studies have reported successful separation of these compounds with HILIC chromatography using an isocratic ammonium acetate buffer coupled with acetonitrile (14% 20 mM ammonium acetate: 86% acetonitrile) as the mobile phase [10, 11]. This mobile phase composition was used as a starting point for the separation of these compounds and is shown in Figure 2b. Since the mass spectrometric intensity of these compounds has been augmented with the addition of various solvents [16] post column, alternative buffers were investigated to determine if increases in response could be achieved. Ammonium fluoride has been reported to result in the ionization of additional compounds and may in fact enhance ionization when using negative electrospray [17, 18]. With this information the HILIC chromatographic separation was evaluated replacing the 20 mM ammonium acetate buffer with 1 mM ammonium fluoride buffer. This change resulted in an increased signal for all compounds (Table 2), with an average 7.7-fold improvement in peak area response, with astandard deviation of the peak area response between the compounds was found to be 0.77. Using the 1 mM ammonium fluoride increased retention; therefore, to minimize runtime, the isocratic mobile phase composition was adjusted to 8% 1 mM ammonium fluoride and 92% acetonitrile to achieve retention factors of 4.6–6.1 (Figure 2a). This change also resulted in further 2-fold signal increase from the 14% ammonium fluoride mobile phase, which is equivalent to at least a 10-fold signal increase (Table 2), indicating the impact of higher acetonitrile content on the ionization efficiency of these compounds.

Twenty sets of calibrators with QC samples were extracted and analyzed by LC-MS/MS to determine the precision and bias of this method over time. Multiple analysts prepared no more than two standard sets with QC samples per day over a period of 36 days. Quality control characterization data are presented in Table 3. The inter-day relative standard deviations for both quality control samples for all five compounds were below 8% and the accuracy was between 101–105% for all compounds. These results confirm this method provides acceptable precision and accuracy for these compounds according to the FDA Bioanalytical Method Development Guidelines [19].

The limits of detection (LOD) for the five compounds were estimated using low level spiked serum with concentrations from 0.25 – 1.0 ng/mL. Each low level spiked sample was analyzed 20 times within a two-week period, with no more than five replicates within one day. The absolute values of the standard deviations were then plotted versus concentration with the intercept of the least-squares fit of this line equal to S₀ with 3S₀ being the LOD [20]. The LODs for this method were determined and reported in Table 3. These estimates, ranging from 0.25–0.50 ng/mL, exceed other methods developed for serum which reported limits of detection from 3.9 – 30 ng/mL for EMPA[5, 6] and 16 ng/mL for PMPA and CMPA [7].

One hundred individual serum samples were extracted and analyzed to evaluate the presence of nerve agent metabolites in the serum of persons with no known exposure. A representative sample is shown in Figure 3. No peaks were observed above the lowest calibration point; therefore, results above the detection limit would indicate that an exposure to these compounds has occurred

To evaluate accuracy and precision across the calibration range, serum samples were prepared at the following concentrations: 0.75, 3, 15, 70, and 90 ng/mL. These samples were extracted and analyzed five times to generate the statistical data presented in Table 4. The relative standard deviations for all five levels were less than 10%, with the exception of the 0.75 ng/mL spike of IMPA which resulted in a variance of 18%. Similarly, the four highest levels resulted in accuracy between 96% and 112%. However, for the lowest spiked sample at a concentration of 0.75 ng/mL the accuracies were between 78–93%. This drop in accuracy and increase in variability at the low concentration may be due to the closeness of the value to the limit of detection.

To widen application of this method to samples with a metabolite concentration greater than 100 ng/mL, a 4.0 μg/mL fortified sample was prepared in serum. This sample was diluted to 4.0 and 40.0 ng/mL using blank serum; the diluted samples were then extracted and analyzed in triplicate. The average quantitated values ranged from 3.91–4.27 ng/mL and 39.5–41.8 ng/mL for the two levels. When the quantitated values were calculated with the dilution factor, the accuracies ranged from 97.8–106%.

The stability of these compounds in serum stored for 6 months at −20°C was evaluated at 4.0 ng/mL and 40.0 ng/mL. The stored samples were analyzed in triplicate and quantitated relative to freshly prepared calibrators. The concentrations of the two levels were within the 10.0% of the original spiked concentration for all five analytes. Given the variability of the method, these results indicated that the metabolites were stable for at least 6 months stored frozen.

Due to a high background response resulting from the ammonium fluoride buffer, the initial transition selected for CMPA (177>79 m/z) did not result in a discernible peak within the reportable range. An alternative transition of 177>63 m/z was evaluated; however, this new transition was only detectable to about 1 ng/mL due to limited signal. Similarly, the confirmatory ion for PMPA was not consistently identified below 2 ng/mL due to a high background response. Previous publications have indicated that this background may be attributed to the impurities within the ammonium fluoride used for the buffer and potentially the interaction of this buffer with the stationary phase [18]. It should be noted that these issues were only detected in the confirmatory ions and do not affect the reported limits of detection for these compounds