A Method for the Analysis of Glyphosate, Aminomethylphosphonic Acid, and Glufosinate in Human Urine Using Liquid Chromatography-Tandem Mass Spectrometry

The extensive use of herbicides, such as glyphosate and glufosinate, in crop production during recent decades has raised concerns about human exposure. Nevertheless, analysis of trace levels of these herbicides in human biospecimens has been challenging. Here, we describe a method for the determination of urinary glyphosate, its degradation product aminomethylphosphonic acid (AMPA), and glufosinate using liquid chromatography-tandem mass spectrometry (LC–MS/MS). The method was optimized using isotopically labelled internal standards (13C2, 15N-glyphosate, 13C, 15N, D2-AMPA, and D3-glufosinate) and solid-phase extraction (SPE) with cation-exchange and anion-exchange cartridges. The method provides excellent chromatographic retention, resolution and peak shape of target analytes without the need for strong acidic mobile phases and derivatization steps. The instrument linearity was in the range of 0.1–100 ng/mL, with R > 0.99 in the matrix for all analytes. The method detection limits (MDLs) and the method quantification limits (MQLs) were in the ranges of 0.12 (AMPA and glufosinate)–0.14 (glyphosate) ng/mL and 0.40 (AMPA)–0.48 (glyphosate) ng/mL, respectively. The recoveries of analytes spiked into urine matrix ranged from 79.1% to 119%, with coefficients of variation (CVs) of 4–10%. Repeated analysis of samples for over 2 weeks showed intra-day and inter-day analytical variations of 3.13–10.8% and 5.93–12.9%, respectively. The matrix effects for glyphosate, AMPA, and glufosinate spiked into urine matrix averaged −14.4%, 13.2%, and 22.2%, respectively. The method was further validated through the analysis of external quality assurance proficiency test (PT) urine samples. The method offers optimal sensitivity, accuracy, and precision for the urine-based assessment of human exposure to glyphosate, AMPA, and glufosinate.


Introduction
Glyphosate (N-(phosphonomethyl)glycine) and glufosinate (2-amino-4-(hydroxy(methyl) phosphoryl)butanoic acid) are non-selective, broad-spectrum herbicides used in both agricultural and non-agricultural sectors. Their use in agriculture has greatly increased since the development of crop strains genetically modified to tolerate them. The current annual use of glyphosate, the most widely used herbicide, is estimated at 600,000-750,000 tons of active ingredients and is expected to increase to 740,000-920,000 tons by 2025 [1]. The United States accounts for 19% of the global glyphosate usage and >100,000 tons of glyphosate have been applied annually in agriculture since 2010 [2]. Glufosinate is mainly used to The molecular structures of the target analytes are shown in Figure 1. Glyphosate (10 µg/mL in water), 13 C 2 , 15 N-glyphosate (100 µg/mL in water), AMPA (100 µg/mL in water), and 13 C, 15 N,D 2 -AMPA (100 µg/mL in water) with purities of 95-98% were purchased from Cambridge Isotope Laboratories (Andover, MA, USA). Glufosinate and D 3 -glufosinate (purity ≥ 95%) were from Toronto Research Chemicals (Toronto, ON, Canada). Primary stock solutions of glufosinate and D 3 -glufosinate (1 mg/mL) were prepared in water. Working standard solutions were diluted from stock solutions using water:acetonitrile (ACN) (95:5, v/v) containing 0.1% formic acid. Formic acid (88%) and ammonium hydroxide (NH 4 OH; 28-30%) of analytical grade were obtained from Sigma-Aldrich (St. Louis, MO, USA). LC passivation solution containing 10 M medronic acid was from Restek Corp (Bellefonte, PA, USA). Water, methanol (MeOH), and ACN were purchased from Fisher Scientific (Waltham, MA, USA). Oasis ® MAX cartridges (60 mg/3 mL) and Oasis ® MCX cartridges (60 mg/3 mL) were obtained from Waters Corp. (Milford, MA, USA).
fortified human urine samples and analyzing external quality assurance proficiency test (PT) urine samples.
A small number of archived human urine samples previously collected for other studies were analyzed [33]. Institutional Review Board approvals were obtained from New York State Department of Health for the analysis of de-identified urine samples (under exempt category) to demonstrate application of the method developed in this study.

Sample Preparation
A 250 µ L aliquot of each urine sample was transferred into a 15 mL polypropylene (PP) tube. Urine samples were fortified with the target compounds and internal standards at 0.5, 1, and 5 ng/mL concentrations (in water: ACN [95:5 v/v] containing 0.1% formic acid) for method optimization and validation. The sample was vortexed vigorously and kept at room temperature for 30 min. The mixture was loaded onto an Oasis MCX cartridge that had been preconditioned with 2 mL MeOH and 2 mL water. The eluate was collected immediately, as the target analytes were not absorbed by the cation-exchange cartridges (this step was for purification and removal of cationic interferences). The cartridge was then washed with 2 mL water, and the eluate was collected and combined. Thereafter, 2.5 mL of 3% NH4OH (v/v) aqueous solution was added and vortexed vigorously. The mixture (~5 mL in total) was then loaded onto an Oasis MAX cartridge preconditioned with 2 mL MeOH, 2 mL water, and 1 mL of 3% NH4OH. The cartridge was washed with 2 mL of 3% NH4OH and 2 mL MeOH, and moisture was removed using a vacuum pump for 3 min. The analytes were then eluted into a 15 mL PP tube with 3 mL of 3% formic acid in MeOH (v/v), and the eluate was evaporated to dryness under N2 at 40 °C. The residue was reconstituted in 250 µ L of water: ACN (95:5, v/v) containing 0.1% A small number of archived human urine samples previously collected for other studies were analyzed [33]. Institutional Review Board approvals were obtained from New York State Department of Health for the analysis of de-identified urine samples (under exempt category) to demonstrate application of the method developed in this study.

Sample Preparation
A 250 µL aliquot of each urine sample was transferred into a 15 mL polypropylene (PP) tube. Urine samples were fortified with the target compounds and internal standards at 0.5, 1, and 5 ng/mL concentrations (in water: ACN [95:5 v/v] containing 0.1% formic acid) for method optimization and validation. The sample was vortexed vigorously and kept at room temperature for 30 min. The mixture was loaded onto an Oasis MCX cartridge that had been preconditioned with 2 mL MeOH and 2 mL water. The eluate was collected immediately, as the target analytes were not absorbed by the cation-exchange cartridges (this step was for purification and removal of cationic interferences). The cartridge was then washed with 2 mL water, and the eluate was collected and combined. Thereafter, 2.5 mL of 3% NH 4 OH (v/v) aqueous solution was added and vortexed vigorously. The mixture (~5 mL in total) was then loaded onto an Oasis MAX cartridge preconditioned with 2 mL MeOH, 2 mL water, and 1 mL of 3% NH 4 OH. The cartridge was washed with 2 mL of 3% NH 4 OH and 2 mL MeOH, and moisture was removed using a vacuum pump for 3 min. The analytes were then eluted into a 15 mL PP tube with 3 mL of 3% formic acid in MeOH (v/v), and the eluate was evaporated to dryness under N 2 at 40 • C. The residue was reconstituted in 250 µL of water: ACN (95:5, v/v) containing 0.1% formic acid, vortexed vigorously, and transferred into a glass vial. Finally, 20 µL of the sample was injected into the LC-MS/MS instrument.

LC-MS/MS
Identification and detection of the target analytes were performed using an AB Sciex 5500 Q-trap mass spectrometer (Framingham, MA, USA) coupled with a Shimadzu LC-30 AD ultra-high-performance liquid chromatograph (Shimadzu Corp., Kyoto, Japan). Analytes were separated on a Gemini ® C6-Phenyl column (150 × 4.6 mm, 5 µm; Phenomenex, Torrance, CA, USA) connected to a Betasil C18 guard column (20 × 2.1 mm, 5 µm; Thermo Fisher Scientific, Waltham, MA, USA). The mobile phases were water (A) and ACN (B) each containing 0.1% formic acid (v/v). The following mobile-phase gradient program was used: hold at 5% B for 2 min, linear ramp to 95% B over 8 min, hold at 95% B for 1 min, then return to initial conditions in over 1 min, and equilibrate at initial conditions for additional 2 min prior to the next injection. The column temperature was maintained at 40 • C; the autosampler temperature was 15 • C; and the mobile phase flow rate was 0.8 mL/min.
The target analytes were determined using negative-ion electrospray ionization (ESI) in the multiple reaction monitoring (MRM) mode. The MRM parameters, including declustering potential (DP), collision energy (CE), and collision cell exit potential (CXP), are shown in Table S1. The IonSpray voltage was −5.5 kV; the ionization source temperature was 500 • C; and the curtain gas flow rate was 20 psi. Data were acquired and processed using the Analyst software, version 1. formic acid, vortexed vigorously, and transferred into a glass vial. Finally, 20 µ L of the sample was injected into the LC-MS/MS instrument.

LC-MS/MS
Identification and detection of the target analytes were performed using an AB Sciex 5500 Q-trap mass spectrometer (Framingham, MA, USA) coupled with a Shimadzu LC-30 AD ultra-high-performance liquid chromatograph (Shimadzu Corp., Kyoto, Japan). Analytes were separated on a Gemini ® C6-Phenyl column (150 × 4.6 mm, 5 µ m; Phenomenex, Torrance, CA, USA) connected to a Betasil C18 guard column (20 × 2.1 mm, 5 µ m; Thermo Fisher Scientific, Waltham, MA, USA). The mobile phases were water (A) and ACN (B) each containing 0.1% formic acid (v/v). The following mobile-phase gradient program was used: hold at 5% B for 2 min, linear ramp to 95% B over 8 min, hold at 95% B for 1 min, then return to initial conditions in over 1 min, and equilibrate at initial conditions for additional 2 min prior to the next injection. The column temperature was maintained at 40 °C; the autosampler temperature was 15 °C; and the mobile phase flow rate was 0.8 mL/min.
The target analytes were determined using negative-ion electrospray ionization (ESI) in the multiple reaction monitoring (MRM) mode. The MRM parameters, including declustering potential (DP), collision energy (CE), and collision cell exit potential (CXP), are shown in Table S1. The IonSpray voltage was −5.5 kV; the ionization source temperature was 500 °C; and the curtain gas flow rate was 20 psi. Data were acquired and processed using the Analyst software, version 1.7.2 (AB Sciex, Framingham, MA, USA). Typical MS/MS chromatograms of the target compounds in standard solution are shown in Figure  2.

Method Validation
The method was validated by following a protocol of the New York State Department of Health (Wadsworth Center, Laboratory of Organic Analytical Chemistry; available at: https://www.wadsworth.org/sites/default/files/WebDoc/NYS%20DOH%20MML-301-06SOP.pdf (accessed on 10 March 2022)). Calibration curves were constructed for standards prepared both in neat solution and in fortified urine matrix. Calibration standards ranged in concentrations from 0.05 to 100 ng/mL, with 10 ng/mL of labelled internal standards, diluted from stock solutions with HPLC-grade water: ACN (95:5, v/v) containing 0.1% formic acid. Matrix-matched calibration curves were prepared by spiking various concentrations of the target analytes (0.1, 0.2, 0.5, 1, 2, 5, 10, 20, 50, and 100 ng/mL) into pooled urine.
Matrix effect was calculated as the percentage of signal enhancement or suppression, as shown in Equation (1): where A and B are the slopes of analytes from the matrix-matched calibration curve and calibration curve prepared in neat solution, respectively. The instrument detection limit (IDL) and instrument quantification limit (IQL) were defined as the concentrations of analytes in solvent that produced a peak with a signal-tonoise ratio (S/N) of 3 and 10, respectively. To estimate the method detection limit (MDL) and method quantification limit (MQL), six pooled urine samples were fortified with each target analyte individually at 0.5 ng/mL, a concentration that yielded peaks with S/N values of 11.3, 7.3, and 12.8 for glyphosate, AMPA, and glufosinate, respectively. MDL and MQL were calculated as 3 and 10 times the standard deviation (SD) measured in matrix, spiked at 0.5 ng/mL, respectively.
The accuracy of the method was determined as the recoveries of analytes spiked at three different concentrations (0.5, 1 and 5 ng/mL) in pooled urine. Procedural blank samples (water in place of urine) were included to monitor for background levels contamination. The precision of the method was assessed by intra-day and inter-day variations, which were calculated as the percentage of the coefficient of variation (%CV) of the measured concentrations in six pooled urine samples spiked at 0.5, 1, and 5 ng/mL, respectively. The inter-day CV was measured by repeated injection of fortified samples over a period of 2 weeks.

Chromatography and Mass Spectrometry
Reported LC-MS/MS methods for the determination of glyphosate and AMPA in urine are summarized in Table 1. Due to the highly polar and hydrophilic nature of the target analytes, chromatographic retention and separation using conventional reversedphase columns (e.g., C18 column) is arduous, resulting in their co-elution with other matrix components. Retention of such analytes can be improved by reversed-phase ion-pair chromatography [27], which is based on the addition of ion pair reagents in the mobile phase to promote the formation of ion pairs. The increase in the hydrophobic character of the electrically neutral ion pair results in a greater affinity for the reverse stationary phase. Because of their strong hydrophobic interactions, the ion pair reagents cannot be completely flushed out of the LC column even through extensive washing, and thus require the use of a dedicated column for a particular application. Hydrophilic interaction liquid chromatography (HILIC) columns enable the retention and separation of hydrophilic compounds, but they often lead to poor peak shape due to interactions with metals in the stationary phase or the chromatographic hardware [34]. Alternatively, considering the low pKa values of the analytes (0.8 for the first phosphonate of glyphosate, 0.9 for the first phosphonate of AMPA, and 0.8 for the phosphonate of glufosinate [26,30]), an anionexchange column was expected to offer efficient retention. However, a high concentration of an acid (e.g., 1% formic acid) was needed in the mobile phases to maintain an optimal peak shape [26]. Other studies have employed a cationic (−H + ) guard column. Although this resulted in separation, glyphosate was eluted within a short retention time, while the peak shape of AMPA was poor [35,36]. In this study, we compared the performance of different chromatographic columns, including reversed-phase (C18-, C8-, C6-Phenyl), HILIC, and anion-exchange (polymer-based NH 2 , hydroxide-selective anion-exchange) columns (data not shown), and found that the C6-Phenyl column exhibited the best chromatographic performance. All analytes were well separated, and the peak shape of AMPA and glufosinate was symmetrical (Figures 2 and 3). However, peak tailing was observed for glyphosate (Figure 3), probably due to the chelation of glyphosate with metal ions in the LC system [37]. Hsiao et al. recommended the addition of 5 µm medronic acid in the mobile phase (passivation solution) to eliminate chelation by metal ions and improve peak shape for metal-sensitive compounds [37]. Nevertheless, we observed a reduced intensity (by~2-fold) for all analytes when medronic acid was added in mobile phases, indicative of ionization suppression. As an alternative, we passivated the LC system by injecting 10 mM medronic acid before analyzing real samples, i.e., we injected 20 µL of 10 mM medronic acid at the beginning of the analytical run (with the mobile phases directed to waste instead of the mass spectrometer). After this passivation, no ionization suppression was found and all analytes including glyphosate exhibited sharp and symmetrical peaks ( Figure 3) for at least 300 subsequent injections.
Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 6 low pKa values of the analytes (0.8 for the first phosphonate of glyphosate, 0.9 for the phosphonate of AMPA, and 0.8 for the phosphonate of glufosinate [26,30]), an anion change column was expected to offer efficient retention. However, a high concentra of an acid (e.g., 1% formic acid) was needed in the mobile phases to maintain an opt peak shape [26]. Other studies have employed a cationic (−H + ) guard column. Altho this resulted in separation, glyphosate was eluted within a short retention time, while peak shape of AMPA was poor [35,36]. In this study, we compared the performanc different chromatographic columns, including reversed-phase (C18-, C8-, C6-Phen HILIC, and anion-exchange (polymer-based NH2, hydroxide-selective anion-excha columns (data not shown), and found that the C6-Phenyl column exhibited the best c matographic performance. All analytes were well separated, and the peak shape of AM and glufosinate was symmetrical (Figures 2 and 3). However, peak tailing was obser for glyphosate (Figure 3), probably due to the chelation of glyphosate with metal ion the LC system [37]. Hsiao et al. recommended the addition of 5 µ M medronic acid in mobile phase (passivation solution) to eliminate chelation by metal ions and improve p shape for metal-sensitive compounds [37]. Nevertheless, we observed a reduced inten (by ~2-fold) for all analytes when medronic acid was added in mobile phases, indica of ionization suppression. As an alternative, we passivated the LC system by injectin mM medronic acid before analyzing real samples, i.e., we injected 20 µ L of 10 medronic acid at the beginning of the analytical run (with the mobile phases directe waste instead of the mass spectrometer). After this passivation, no ionization suppres was found and all analytes including glyphosate exhibited sharp and symmetrical p ( Figure 3) for at least 300 subsequent injections.

Optimization of Sample Cleanup
Because of their low pKa values, we expected anion-exchange cartridges, which positively charged and can bind negatively charged target analytes, to be effective for application [26,38]. We first optimized a mixed-mode anion-exchange cartridge (Oa MAX cartridge), which contains sorbents having both hydrophobic and anion-excha functionalities. Indeed, MAX cartridges provided excellent recoveries for all target lytes after optimization of elution solvents. However, matrix components were not c pletely removed, as we observed strong ionization suppression of glyphosate glufosinate, which resulted in poor sensitivity. For example, the S/N values of glypho

Optimization of Sample Cleanup
Because of their low pKa values, we expected anion-exchange cartridges, which are positively charged and can bind negatively charged target analytes, to be effective for this application [26,38]. We first optimized a mixed-mode anion-exchange cartridge (Oasis ® MAX cartridge), which contains sorbents having both hydrophobic and anion-exchange functionalities. Indeed, MAX cartridges provided excellent recoveries for all target analytes after optimization of elution solvents. However, matrix components were not completely removed, as we observed strong ionization suppression of glyphosate and glufosinate, which resulted in poor sensitivity. For example, the S/N values of glyphosate and glufosinate in pooled urine spiked at 0.5 ng/mL were <3 and 3.8, respectively ( Figure S1). Therefore, we introduced an additional purification step to reduce matrix effects. We compared several cartridges for cleanup, including reversed-phase cartridges (hydrophilic lipophilic balanced (HLB) solid-phase extraction (SPE) cartridges, C18, and graphitized non-porous carbon) and mixed-mode strong cation-exchange cartridges (Oasis ® MCX) (data not shown). We found that a pre-cleanup step in which samples were passed through MCX cartridges (as described above) significantly reduced matrix effects, and thus increased the method sensitivity. The responses of all analytes, especially that of glyphosate, increased considerably after MCX pre-cleanup ( Figure S1). The peak area of glyphosate was >10-fold higher in urine sample passed through MCX and MAX cartridges than in those that passed only through MAX (Table S2). These results highlighted the efficacy of MCX SPE as a pre-cleanup step for the improvement of method sensitivity. We believe that this is due to the efficient removal of cationic components from the matrix. An earlier study reported the successful use of MCX cartridges for pre-cleanup in the analysis of glyphosate and AMPA in foodstuffs [38]. LC-MS/MS chromatograms obtained following a combination of cation-exchange and anion-exchange SPE cartridges in the preparation of urine samples showed well-resolved peaks in samples fortified at 0.5 ng/mL (Figure 4). and glufosinate in pooled urine spiked at 0.5 ng/mL were <3 and 3.8, respectively ( Figure  S1). Therefore, we introduced an additional purification step to reduce matrix effects. We compared several cartridges for cleanup, including reversed-phase cartridges (hydrophilic lipophilic balanced (HLB) solid-phase extraction (SPE) cartridges, C18, and graphitized non-porous carbon) and mixed-mode strong cation-exchange cartridges (Oasis ® MCX) (data not shown). We found that a pre-cleanup step in which samples were passed through MCX cartridges (as described above) significantly reduced matrix effects, and thus increased the method sensitivity. The responses of all analytes, especially that of glyphosate, increased considerably after MCX pre-cleanup ( Figure S1). The peak area of glyphosate was >10-fold higher in urine sample passed through MCX and MAX cartridges than in those that passed only through MAX (Table S2). These results highlighted the efficacy of MCX SPE as a pre-cleanup step for the improvement of method sensitivity. We believe that this is due to the efficient removal of cationic components from the matrix. An earlier study reported the successful use of MCX cartridges for pre-cleanup in the analysis of glyphosate and AMPA in foodstuffs [38]. LC-MS/MS chromatograms obtained following a combination of cation-exchange and anion-exchange SPE cartridges in the preparation of urine samples showed well-resolved peaks in samples fortified at 0.5 ng/mL (Figure 4).

Method Validation
We assessed the linearity of the instrument by injecting analytical standards prepared both in solvent (0.05-100 ng/mL) and urine matrix (0.1-100 ng/mL). An excellent linearity was found for all analytes with R values >0.99 (Table 2). We assessed the accuracy
We determined the sensitivity of the method as IDLs/IQLs as well as MDLs/MQLs through the injection of standards and fortified urine samples. The respective IDLs and IQLs were 0.01 and 0.05 ng/mL for all target analytes. The MDLs/MQLs were 0.14/0.48, 0.12/0.40, and 0.12/0.41 ng/mL for glyphosate, AMPA, and glufosinate, respectively. The sensitivity of our method is comparable to those found in several previous studies [11,[39][40][41][42], and slightly higher than those of others [27,43] (Table 1). We expect that further improvements in MDLs/MQLs could be accomplished through inclusion of additional sample volumes available for extraction. The matrix effect is a common phenomenon in LC-MS analysis, especially in the ESI mode, that involves enhancement or suppression of analyte responses by matrix components [48]. We observed an ionization suppression for glyphosate (matrix effect: −14.4%) and ionization enhancements for AMPA (13.2%) and glufosinate (22.2%) ( Table 2). The ionization suppression may explain the lower recoveries of glyphosate in fortified samples, which were in the range of 79.1-84.4%. However, the addition of labelled internal standards for quantification enabled correction for matrix effects.
We also validated our method by analyzing external quality assurance proficiency test (PT) urine samples, offered by the German External Quality Assessment Scheme (G-EQUAS) and the Quebec External Quality Assessment Scheme for Organic Substances in Urine (OSEQAS). Our results were within the acceptable ranges of assigned values, indicating high accuracy of our method (Table 3). Table 2. Optimized analytical parameters for the analysis of glyphosate, AMPA, and glufosinate in human urine. AMPA, aminomethylphosphonic acid; IDL, instrument detection limit; IQL, instrument quantification limit; MDL, method detection limit; MQL, method quantification limit.  Next, we applied the validated method to the determination of concentrations in twenty human urine samples randomly collected from the populations of the US states of Iowa (n = 10) and New York (n = 10). In samples from Iowa, we detected glyphosate in six out of ten samples (mean: 1.18 ng/mL) and AMPA in five out of ten samples (mean: 0.88 ng/mL). In samples from New York, we found glyphosate in only one sample (0.53 ng/mL) and did not detect AMPA in any samples (Table 4). Samples from Iowa were collected from adult males living in a rural farming region, whereas those from New York were from a population of office workers including adult males and females. Glufosinate was not found in any of the samples, probably due to its rapid metabolism, and low usage (~200-fold lower than glyphosate) [1,2,17]. However, with the exponential increase in glufosinate usage, its concentration in human urine may increase in the future. Overall, our results suggest the feasibility of measuring glyphosate, AMPA, and glufosinate in biomonitoring studies using the current method.

Glyphosate
In comparison to previous studies (Table 1), our method has several advantages for application in human biomonitoring studies: (1) Excellent chromatographic retention, resolution, and peak shape, which were achieved through the use of less corrosive (i.e., less acidic) mobile phases. Previous studies used very high concentrations of acids in mobile phases (i.e., 1% formic acid or acetic acid) [26,40], used ion-pairing reagents (i.e., heptafluorobutyric acid) [27,31], or applied derivatization steps [32] to enhance sensitivity and selectivity. Such techniques are tedious, time-consuming, or corrosive; (2) Our method provides excellent sensitivity and uses a smaller sample volume (250 µL) compared with other methods that used 0.5-2.5 mL urine [25,36]. Although one method reported relatively higher sensitivity [36], that method used a dilute-and-shoot method, which can affect selectivity and sensitivity due to matrix interferences. Furthermore, it was not clear if the reported detection limit for that method was that of the method or the IDL. (3) Our method has been validated through various QC parameters and successful participation in external assurance schemes while previously reported method did not report such external validation protocols (Tables 3 and 4). Table 4. Glyphosate, AMPA, and glufosinate concentrations measured in twenty human urine samples randomly collected from the general populations in Iowa (n = 10) and New York (n = 10), USA. Calculated concentrations are provided for those between MDL and MQL.

Conclusions
We have developed and validated an LC-MS/MS method for the determination of glyphosate, AMPA, and glufosinate in human urine. Passage of samples in sequence through a combination of cation-and anion-exchange solid-phase extraction cartridges for purification reduced matrix effects and increased sensitivity. Chromatographic and