Montmorillonites Can Tightly Bind Glyphosate and Paraquat Reducing Toxin Exposures and Toxicity.

Among the numerous contaminants of soil, glyphosate and paraquat are two of the most widely used herbicides that are commonly detected in the environment. Soil and sediment contaminated with glyphosate, paraquat, and other environmental toxins can be mobilized and redistributed to lawns, vegetable gardens, parks, and water supplies in vulnerable communities at the site of disasters such as hurricanes and flooding. Glyphosate and paraquat bind strongly to soils containing clays, making their bioavailability (bioaccessibility) from these types of soil very low. Because of their affinity for clay-based soils, it is possible that montmorillonite clays could be administered as a therapeutic agent in the diet of animals and humans to decrease short-term exposure and toxicity. In this study, we investigated the sorption mechanisms of glyphosate and paraquat onto active surfaces of calcium montmorillonite (CM) and sodium montmorillonite (SM) clays and derived binding parameters, including capacity, affinity, and enthalpy. Additionally, we used these parameters to predict the reduction in bioavailability under different pH and temperature conditions and to estimate the theoretical dose of clay that could protect against severe paraquat toxicity and lethality. Computational modeling and simulation studies depicted toxin sorption mechanisms at different pH values. Additionally, a toxin-sensitive living organism (Hydra vulgaris) was used to confirm the safety of the clay and its ability to protect against toxicity from glyphosate and paraquat. The high efficacy of CM and SM shown in this study supports the natural binding activity of glyphosate and paraquat to clay-based soils. Following disasters and medical emergencies, montmorillonite clays could be administered by capsules and tablets, or added to food and flavored water, to reduce toxin bioavailability and human and animal exposures.

Introduction

The use of pesticides is of major importance to agriculture, but their widespread occurrence and persistence in the environment can be hazardous to living organisms. Moreover, the potential health hazards of pesticides (and other environmental contaminants) can be enhanced by hurricanes and flooding, which can mobilize and redistribute these chemicals in contaminated soil and sediment, resulting in increased contamination of food and water at the site of disasters.

Glyphosate is one of the most commonly used organophosphorus herbicides to control weeds. An important factor contributing to the dominant use of glyphosate is the introduction of transgenic, glyphosate-resistant crops in 1996. Almost 90% of all transgenic crops grown worldwide are glyphosate-resistant, and the adoption of these crops is increasing at a steady pace. Its mode of action is by inhibiting enzymes involved in the synthesis of three amino acids: tyrosine, tryptophan, and phenylalanine.[1] Importantly, glyphosate has been shown to bind tightly to soils, especially those containing montmorillonite clay,[2] making its bioavailability (bioaccessibility) low for uptake by roots and translocation to nearby plants. In soils with little or no clay in their composition, glyphosate can be bioavailable to plants.[3] The potential risk of glyphosate exposure is a contemporary and controversial issue. In 2015, glyphosate was classified as “probably carcinogenic in humans” by International Agency for Research on Cancer (IARC).[4] Case–control studies of occupational exposure in the United States, Canada, and Sweden found increased risks for non-Hodgkin lymphoma.[5,6] However, other agencies, such as the European Food Safety Authority, have concluded that the evidence for carcinogenicity is insufficient.[7,8] The risk of occupational exposure to glyphosate is increased in agricultural workers, and the toxin can be detected in blood and urine samples from those who are highly exposed to glyphosate.[9]

Paraquat is one of the most widely used herbicides due to its rapid, contact-dependent killing of weeds and plants and its inactivation upon reaching the soil. Like glyphosate, its sorption by clay minerals is believed to be a major mechanism for its inactivation in soil.[10] Paraquat has been associated with an increased risk of Parkinson’s disease.[11] It is also the leading cause of death from pesticide self-poisoning and is classified as “restricted use” by the US Environmental Protection Agency. The very high case-fatality rate of paraquat is due to both its inherent toxicity and the lack of any effective treatment for exposed individuals.[12]

Calcium montmorillonite (CM) clay has been shown to be safe for short-term consumption in multiple animal models and clinical trials in the United States and Africa.[13−41] Both CM and sodium montmorillonite (SM) are routinely used in animal feeds as therapy for the mitigation of aflatoxin, worldwide. In this study, we have characterized the sorption of glyphosate and paraquat to CM and SM clays and investigated chemical and structural mechanisms of sorption. Equilibrium isothermal analyses and dosimetry studies were conducted to derive binding parameters and gain insight into (1) surface capacities and affinities, (2) potential mechanisms of sorption, (3) thermodynamics of toxin/surface interactions, and (4) predicted reduction in bioavailability as a function of the estimated dose of the sorbent. Additionally, we performed molecular dynamics (MD) simulations and structural analyses of SM in the presence of glyphosate and paraquat at pH 2 and 7 to study the surface chemistry of the sorption reaction and the effect of pH. Finally, we used a toxin-sensitive living organism (Hydra vulgaris) to confirm the safety of CM and SM and their efficacy in reducing the toxicity of glyphosate and paraquat.

Results

Glyphosate

The molecular representations of toxins and different forms of glyphosate as a function of pH are shown in Figure . Isothermal analysis of glyphosate sorption onto CM and SM surfaces at 37 °C (T1) in an aqueous solution at pH 2 and 7 are shown in Figure A,B. The plots fit the Langmuir model as indicated by (1) a good correlation coefficient (r2 ≥ 0.9), (2) a residual standard error (rse) ranging from 0.022 to 0.11 indicating a variability in Qmax data less than 11%, and (3) a curved shape indicating saturation. Glyphosate sorption onto clay sorbents showed higher binding capacity at pH 2 (Qmax = 0.71 and 0.96 for CM and SM, respectively) than at pH 7 (Qmax = 0.26 and 0.41 for CM and SM, respectively). Isotherms run at 26 °C (T2) (Figure C,D) fit the Langmuir model with r2 ≥ 0.9 and rse ranging from 0.022 to 0.095, indicating a variability in Qmax data less than 7%. Binding capacities at pH 2 (Qmax = 1.27 and 0.9 for CM and SM, respectively) were also higher than at pH 7 (Qmax = 0.32 and 0.3 for CM and SM, respectively). The calculated enthalpies for CM were equal to −75 and −33 kJ/mol at pH 2 and 7, and the enthalpies for SM were equal to −33 and −41 kJ/mol at pH 2 and 7, respectively. Within the simulations, glyphosate was primarily bound to clay through its positive amide group (74%) at pH 7 (Figure E, encircled in green). At pH 2, glyphosate was primarily bound to clay through its positive amide group (51%) or through its protonated phosphate (43%) (Figure E, encircled in green and orange, respectively). Other binding conformations were observed but with lower probability. The computationally predicted average residence time per binding event of glyphosate bound to clay at pH 2 (0.59 ± 1.2 ns with a maximum duration of 20.08 ns) was longer than the residence time at pH 7 (0.10 ± 0.09 ns with a maximum duration of 2.42 ns) (Figure ).

Figure 1

Molecular contact distance criteria and sets of atoms used to categorize the binding configurations of glyphosate and paraquat (A). Different forms of glyphosate as a function of pH (B).

Figure 2

Isotherm data (triangles and squares) and Langmuir plots (curves) of glyphosate binding onto CM (A, C) and SM (B, D). Top panels (A, B) showing observed and predicted Qmax at 37 °C (T1), and middle panels (C, D) showing observed and predicted Qmax at 26 °C (T2), all in an aqueous solution at pH 2 (orange) and 7 (blue). Enthalpy (ΔH) was calculated from isotherms run at 37 and 26 °C using the van’t Hoff equation. All measurements were independently triplicated. Representative MD simulation snapshots of montmorillonite in the presence of glyphosate at pH 2 and 7 (E). SM is shown in van der Waals representation. Glyphosate molecules are shown in licorice representation with hydrogens omitted. The most prominent binding conformations of glyphosate and their corresponding propensities are encircled in green dotted lines adjacent to the image of the simulation system and green doubled lines within the image of the simulation system. The second most prominent binding conformation of glyphosate at pH 2 is encircled in orange dotted lines adjacent to the image of the simulation system and orange doubled lines within the image of the simulation system.

Figure 3

Residence time per binding event distributions for glyphosate and paraquat at pH 2 and 7 according to MD simulations. The (%) number of instances per residence time bin is calculated as the number of binding instances for a specific duration over the total number of binding instances. Error bars represent the standard deviation calculated over multiple simulation runs.

Using the rearranged Langmuir equation described in Materials and Methods, the toxin bioavailability can be predicted for combinations of nominal toxin concentration Cnom and clay concentration Cclay. Figure A shows the predicted fraction bioavailable for 30 ppm glyphosate. For glyphosate, both CM and SM at concentrations of 0.1–0.2% result in the bioavailability of no more than a few percent. Exposure to glyphosate resulted in a time- and concentration-dependent decrease in hydra morphology scores (see Supporting Information Figure S1), with complete loss of viability at t ≥ 44 h for C ≥ 40 ppm glyphosate. The next lowest concentration of 30 ppm glyphosate was selected for directly testing the effects of clay. Based on the predicted reduction in bioavailability shown in Figure A, with 0.1% CM or SM, the hydra morphology scores at 30 ppm glyphosate would be ≥9. Figure B shows the experimental results of direct comparisons of the toxin with and without clay. For glyphosate, complete protection was observed for 0.1% CM, resulting in morphological scores equal to the hydra media control group. Additionally, only partial protection from glyphosate toxicity was observed by the heat-collapsed CM, where the interlayers were dehydroxylated and collapsed.

Figure 4

Predicted free fraction (%) of glyphosate at a nominal concentration of 30 ppm for different clays under different pH and temperature as a function of clay concentration (%) (A). Predictions were made by rearrangement of the Langmuir equation. Protection of hydra by CM and collapsed CM at a 0.1% inclusion rate against 30 ppm glyphosate (B). CM showed complete (100%) protection against glyphosate, whereas 33% protection was shown from the collapsed CM treatment group at the end point. Bands indicate 95% confidence intervals on the mean response.

Paraquat

As shown in Figure A–D, the isotherm plots for paraquat fit the Langmuir model with r2 > 0.9 and rse ranging from 0.026 to 0.055 (indicating a variability in the Qmax data less than 18.9%), high binding capacity (Qmax = 0.43 and 0.44 for CM and SM, respectively), and high affinity (Kd = 1.5 × 106 and 2.0 × 106 for CM and SM, respectively). The thermodynamics at 37 °C (T1) and 26 °C (T2) showed binding enthalpies for paraquat sorption onto the surfaces of CM and SM equal to −44 and −32 kJ/mol, respectively. The sorption parameters for glyphosate and paraquat on CM and SM are summarized in Table . Within the simulations, paraquat at both pH 2 and 7 was bound to the clay through both of its bipyridinium rings being approximately parallel to the clay surface (90 and 93%, respectively) (left and right panels of Figure E, respectively, encircled in green). The computationally predicted average residence time per binding event of paraquat bound to clay was similar at pH 2 and 7 (2.84 ± 6.33 ns with a maximum duration of 39.98 ns, and 4.06 ± 7.45 ns with a maximum duration of 40.02 ns, respectively).

Figure 5

Isotherm data (triangles and squares) and Langmuir plots (curves) of paraquat binding onto surfaces of CM (A, C) and SM (B, D). Top panels (A, B) showing observed and predicted Qmax at 37 °C (T1), and middle panels (C, D) showing observed and predicted Qmax at 26 °C (T2), all in an aqueous solution at pH 7. Enthalpy (ΔH) was calculated from isotherms run at 37 and 26 °C using the van’t Hoff equation. All measurements were independently triplicated. Representative MD simulation snapshots of montmorillonite in the presence of paraquat at pH 2 and 7 (E). SM is shown in van der Waals representation. Paraquat molecules are shown in licorice representation with hydrogens omitted. The most prominent binding conformations of paraquat and their corresponding propensities are encircled in green dotted lines adjacent to the image of the simulation system and green doubled lines within the image of the simulation system.

Table 1

Summary Table of Sorption Parameters for CM and SM Sorbentsa

	glyphosate				paraquat
sorbents	pH 2, T1	pH 7, T1	pH 2, T2	pH 7, T2	pH 7, T1	pH 7, T2
CM	Qmax = 0.71	Qmax = 0.26	Qmax = 1.27	Qmax = 0.32	Qmax = 0.43	Qmax = 0.29
	Kd = 1.1 × 105	Kd = 1.5 × 105	Kd = 3.2 × 105	Kd = 2.4 × 105	Kd = 1.5 × 106	Kd = 2.8 × 106
	rse = 0.037	rse = 0.022	rse = 0.095	rse = 0.056	rse = 0.033	rse = 0.055
SM	Qmax = 0.96	Qmax = 0.41	Qmax = 0.9	Qmax = 0.3	Qmax = 0.44	Qmax = 0.44
	Kd = 1.2 × 105	Kd = 1.4 × 105	Kd = 1.9 × 105	Kd = 2.5 × 105	Kd = 2.0 × 106	Kd = 3.4 × 106
	rse = 0.11	rse = 0.038	rse = 0.062	rse = 0.022	rse = 0.026	rse = 0.046

CM, calcium montmorillonite; SM, sodium montmorillonite; T1 = 37 °C; T2 = 26 °C; Qmax, binding capacity; Kd, binding affinity; rse, residual standard error.

In Figure A, clay concentrations as low as 0.1–0.2% resulted in a predicted bioavailability of less than 2% of 20 ppm paraquat. At the same clay concentration, temperature, and pH, paraquat was predicted to be less bioavailable than glyphosate. Exposure to paraquat resulted in a time- and concentration-dependent decrease in hydra morphology scores (see Supporting Information Figure S2), with complete loss of viability at t ≥ 44 h for C ≥ 30 ppm paraquat. The next lowest concentration of 20 ppm paraquat was selected for directly testing the effects of clay. Based on the predicted reduction in bioavailability shown in Figure A, with 0.1% CM or SM, complete protection was predicted for paraquat at 20 ppm. Figure B shows the experimental results of direct comparisons of paraquat without clay and the complete protection observed for both CM and SM at either 0.1 or 0.2%.

Figure 6

Predicted free fraction (%) of paraquat at a nominal concentration of 20 ppm for different clays under different pH and temperature as a function of clay concentration (%) (A). Predictions were made by rearrangement of the Langmuir equation. Complete (100%) protection by CM and SM at 0.1 and 0.2% inclusion rates, respectively, was shown against 20 ppm paraquat (B). Bands indicate 95% confidence intervals on the mean response.

Discussion

Our previous results have indicated that CM clay is stable and safe for short-term animal and human consumption.[42,43] Moreover, the inclusion of CM in the diet of humans had no negative impact on palatability (texture, taste, and odor) or acceptability of the diet from our clinical trials in Ghana,[37] Texas,[39] and Kenya.[40]

In the current study, sorption isotherms were conducted at 37 °C (T1) in an aqueous solution at pH 2 and 7 to simulate body temperature and stomach and intestinal pH conditions. The net negative charge on both CM and SM clays is the primary factor in clay dispersion, and changes in pH affect clay dispersion by altering the net charge on clay particles. However, the inclusion rate of clay in isotherms run at pH 2 and 7 was only 0.002%. We believe that this very low level excludes the possibility of aggregation effects. The fact that glyphosate binding capacity at pH 2 (Qmax = 0.71 and 0.96 for CM and SM, respectively) was higher than at pH 7 (Qmax = 0.26 and 0.41 for CM and SM, respectively) (Figure A,B) may be due to the zwitterionic property of glyphosate and the resulting structural conformers at different pH values. At pH 2, glyphosate is expected to be protonated resulting in a net positive charge.[44] Based on our isothermal data, we hypothesized that the binding of glyphosate at pH 2 is probably the result of electrostatic interactions of the positively charged amide group with the negatively charged interlamellar surfaces of CM and SM and a hydrophobic effect at the protonated carboxyl group on the toxin. Glyphosate (at pH 7) possesses negative charges resulting from the deprotonation at the carboxyl and the phosphate groups, resulting in a less favorable electrostatic interaction. Similarly, the simulations suggest that glyphosate is primarily bound to clay interlayers through its positive amide group at pH 2 and 7 (Figure E, encircled in green). Due to protonation at pH 2, the simulations suggest that glyphosate can also bind to the clay interlayer through hydrogen bond interactions between its phosphate group and the oxygens of the clay siloxane surface. This is consistent with previous literatures indicating that phosphate in the soil can compete with glyphosate for sorption sites and can reduce the sorption of both compounds, especially in acidic soil.[45] Besides electrostatic interactions, it is possible that interlamellar cations (such as Ca2+) could interact with oxygens from phosphonic acid and glycine moieties associated with glyphosate and contribute to the mechanism. More work is warranted in this area.

To determine the available clay surface area for binding at saturation (plateau surface density), we calculated the surface area of active binding sites on CM and compared it to the maximum binding area of glyphosate derived from isothermal analysis. The surface area of CM interlamellar binding sites = 850–70 = 780 m2/g = 7.8 × 1022 Å2/g. Based on the maximum binding capacity derived from the isotherms, the number of glyphosate molecules bound at pH 2 = 0.71 mol/kg × 6.02 × 1023 molecules/mol = 4.3 × 1020 molecules/g (or plateau surface density). Since the topological polar surface area of one glyphosate molecule is 113 Å2, the total binding area for the glyphosate = 4.3 × 1020 × 113 = 4.9 × 1022 Å2/g, which is smaller than the available clay surface area (7.8 × 1022 Å2/g). The plateau surface density and the high binding capacity suggest that montmorillonite contains more than adequate binding sites for monolayer sorption, with space left over. The monolayer sorption (instead of a multilayer phenomenon) is consistent with the curved shape and homogeneous site requirement of the Langmuir model.

Thermodynamic analysis of the interaction can determine if the sorption is physical or chemical, spontaneous or nonspontaneous, and also exothermic or endothermic. To calculate binding enthalpy, isotherms were conducted at both 37 °C (T1) and 26 °C (T2), and then the derived Kd values were applied in the van’t Hoff equation. Figure C,D shows isothermal sorption at 26 °C (T2) onto CM and SM. The high enthalpy values indicate that surface interaction involves tight binding at active sites (i.e., chemisorption), instead of weak interactions or physisorption (less than 20 kJ/mol). Our thermodynamic data is consistent with the high stability of the clay/toxin complex and low toxin desorption. These results suggest that glyphosate should bind strongly in the stomach, with sustained binding in the intestine. Both binding sites for glyphosate have high capacity and high affinity and involve chemisorption interactions. The isothermal and thermodynamic analyses were similar for glyphosate binding to CM and SM.

Additionally, we rearranged the Langmuir equation to predict the bioavailability of glyphosate as a function of nominal toxin concentration and clay concentration. Based on the predicted reduction in glyphosate bioavailability shown with 0.1% CM or SM (Figure A), the glyphosate hydra morphology scores at 30 ppm would be predicted to be ≥9. The hydra data in Figure B showed that CM at the inclusion rate of only 0.1% completely protected hydra against a 30 ppm glyphosate toxicity, resulting in morphological scores no different from the hydra media control group. This protection is in alignment with the high binding capacity, affinity, and enthalpy for glyphosate sorption onto the surfaces of CM. Interestingly, only partial protection was observed by the heat-collapsed CM, where the interlayers were dehydroxylated and collapsed. This is indirect evidence that the main site of sorption for glyphosate is found within the CM interlayer on negatively charged surfaces.

Paraquat is another herbicide naturally bound by soils containing clays. It is widely used in agriculture and also used as an agent for suicide due to its potent toxicity.[11,12] Because paraquat has permanent positive charges on the quaternary nitrogens that are not pH-dependent, isotherms were conducted at pH 7 in water. As shown in Figure A,B, the high binding capacity of paraquat at 37 C (T1) onto both CM and SM indicates that the maximum concentration of paraquat bound (Qmax) is 0.44 mol/kg × 6.02 × 1023 molecules/mol = 2.6 × 1020 molecules/g. Since the topological polar surface area of one paraquat molecule is equal to 7.8 Å2, the coverage area (plateau surface density) for the maximum amount of paraquat bound is equal to 2.6 × 1020 × 7.8 = 2 × 1021 Å2/g, which is an order of magnitude smaller than the surface area of clay available for binding (7.8 × 1022 Å2/g). This data and the high binding capacity of paraquat (like glyphosate) suggest that montmorillonite clay contains more than adequate binding sites for monolayer coverage, with space left over. This sorption of paraquat was also tight (i.e., chemisorption), as suggested by the thermodynamics of the reaction at 37 °C (T1) and 26 °C (T2) (Figure C,D). The predominant binding conformations for paraquat indicate that toxin binding to the clay interlayers at pH 2 and 7 was nearly identical within the simulation. This suggests that the sorption of paraquat was unaffected by pH. Simulations indicate that paraquat primarily binds to clay with both bipyridinium rings being approximately parallel to the surface (Figure E, encircled in green).

As with glyphosate, we rearranged the Langmuir equation to predict the bioavailability of paraquat as a function of nominal toxin concentration and clay concentration (Figure A). The hydra bioassay (Figure B) confirms that CM and SM at an inclusion rate of only 0.1% were able to prevent the toxicity of 20 ppm paraquat, which resulted in 100% lethality in hydra without clays. At the same clay concentration, the predicted bioavailability of paraquat is even lower than that of glyphosate. In line with this result, the computationally derived distributions of residence time per binding events suggest that paraquat has a higher probability to retain a bound conformation compared to glyphosate at pH 2 and 7 (Figure ). Thus, bound paraquat molecules are less likely to desorb from the clay, thereby presumably decreasing toxin bioavailability in the presence of clay. This may be attributed to the two positively charged nitrogens in the paraquat bipyridinium rings, as well as its planar structure, allowing both nitrogens to simultaneously form favorable electrostatic interactions with the negatively charged clay surfaces. It is worth noting that the computationally predicted residence time per binding event should be considered an underestimate of the total dwell time of a compound within a clay as the former considers a single binding event, whereas the latter considers an aggregate time of multiple binding events. Due to the potent toxicity of paraquat and the incidence of paraquat poisoning, we predicted (from the Langmuir equation) the theoretical amount of clay that would be needed to reduce the bioavailability of a lethal level of paraquat (35 ppm)[46] by 99.9%. The predicted concentration of CM to protect against 35 ppm paraquat was equal to 0.2% w/w or 2 g clay/kg. Assuming the ingestion of 35 mg of paraquat during an emergency, 2 g of CM could be administered using gavage, capsules, tablets, or flavored water to reduce human or animal exposures. Our results suggest that CM binds to paraquat very effectively, and this clay (or similar clays) could be effective as toxin enterosorbents during emergencies. In further work, we are investigating the sorption of other substances and using MD simulations and computational methods to design effective clay-based sorbents for the mitigation of hazardous Superfund chemicals.

Based on a combination of in vitro, in vivo, and in silico studies, CM and SM clays were found to be effective sorbents with high binding capacities, affinities, and enthalpies for glyphosate and paraquat. These studies suggest that effective levels of montmorillonite clays can be used to reduce human and animal exposures to toxins from contaminated food and water during disasters and emergencies. These clays may also facilitate the treatment of lethal doses of paraquat in humans and animals.

Materials and Methods

Materials and Reagents

Ca2+/Na+-montmorillonite with 20% (or more) sodium behaves qualitatively as a sodium montmorillonite clay (SM), whereas montmorillonite with 90% (or more) calcium in the interlayer behaves similarly to calcium montmorillonite (CM).[47] CM clay was purchased from Engelhard Chemical Corporation (and is now available from BASF in Lampertheim, Germany) with a total surface area as high as 850 m2/g, an external surface area of approximately 70 m2/g, and a cation exchange capacity equal to 89.2 cmol/kg.[48] SM clay was obtained from the University of Missouri-Columbia with a cation exchange capacity equal to 90.5 cmol/kg. The generic formula for these clays is (Na,Ca)0.3(Al,Mg)2Si4O10(OH)2·nH2O. The molecular models of CM and SM have been constructed in silico and published by our laboratory, along with important physicochemical properties of CM and SM clays based on X-ray fluorescence analysis.[49] From X-ray diffraction (XRD) of CM, trace levels of feldspars and mica were identified; XRD of SM showed 2.1% mica, 0.8% iron oxides, and trace levels of other mineral impurities.[50,51] To investigate the importance of intact interlamellar spaces, montmorillonite clay was collapsed by heating samples at 200 °C for 30 min and 800 °C for 1 h.[52] Reagents for high-pressure liquid chromatography (HPLC) including acetonitrile, ammonium acetate, and pH buffers (4.0, 7.0, and 10.0) were purchased from VWR (Atlanta, GA). Glyphosate and paraquat were purchased from Sigma-Aldrich (Saint Louis, MO).

Analytical Chemistry

Glyphosate and paraquat were analyzed using a Waters Acquity ultraperformance liquid chromatography/mass spectrometry/ mass spectrometry (LC/MS/MS) equipped with triple quadrupole. For glyphosate, an Acquity BEH C18 column (2.1 × 50 mm, 5 μm) was used for separation and kept at 20 °C. A gradient elution using water with 0.1% formic acid (eluent A) and acetonitrile with 0.1% formic acid (eluent B) was carried out at a flow rate of 0.3 mL/min. The gradient program for elution was 5% eluent B (initial) and 5–100% eluent B (from 0 to 10 min). Formic acid in the mobile phase was used to promote protonation of the amino group. The injection volume was 10 μL for each analysis. The mass spectrometer was used with electrospray ionization (ESI) interface and operated in a negative ion mode. The spray voltage was maintained at 4.5 kV. The source temperature was kept at 225 °C. The monitored precursor and product ions were m/z 168 to 63 and 81.[53] Paraquat was detected using a HILIC column (2.1 × 100 mm, 3 μm) at 30 °C. Separation using a mobile phase containing 10 mM ammonium acetate with 0.1% formic acid (eluent A) and acetonitrile (eluent B) was carried out at a flow rate of 0.2 mL/min and an injection volume of 10 μL. The gradient program was used for elution: 60% eluent B (initial), 60–20% eluent B (from 0 to 7 min), and 20% eluent B (7–8 min). The triple quadrupole mass spectrometer was operated in a positive mode with a capillary voltage at 4.5 kV. The source temperature was kept at 350 °C. Positively charged molecular ions of paraquat were monitored for precursor and products at m/z 186 to 171 and 155.[54] For both methods, the mass spectrometer was operated under multiple reaction monitoring (MRM) mode. Nitrogen gas was used as the collision and curtain gas, and argon gas was used as the nebulizer and heater gas. Empower analyst software was used to control the LC/MS/MS system and acquire the data.

The limits of detection (LOD) for glyphosate and paraquat were 0.5 ppb and 10 ppb, respectively, with excellent reproducibility and sensitivity of the detection methods. Standard toxin solutions were spiked before and after 2 h of agitating, and the relative standard deviations (RSD) were <4%, showing a high recovery percentage and limited nonspecific binding. The detection methods were validated using standard calibration curves. Standard solutions of glyphosate and paraquat were prepared in distilled water at concentration gradients between 20 and 0.5 ppm to plot the standard curves. The standard curves for glyphosate and paraquat were linear (r2 > 0.99) between signal intensity (y axis) and toxin concentration (x axis) and were described by the following equations: y = 32 185x + 41 034 and y = 106x–720 669, respectively.

Sorption Isotherm Experiments

The toxin stock solutions were individually prepared by dissolving pure crystals into distilled water to yield 12 ppm (μg/mL) glyphosate and 5 ppm paraquat solutions. The stock concentrations were set based on chemical solubility in water to prevent the occurrence of precipitation and the optimal ratio of toxin/clay to reach saturation (equilibrium) on isotherm plots. For isothermal analysis, 0.002% w/w of adsorbents were added to toxin solutions with an increasing gradient. The concentration gradients of toxin solutions were achieved by adding a calculated amount of toxin stock solution along with a complementary volume of distilled water at pH 2 and 7 for glyphosate and pH 7 for paraquat. Additionally, there were three controls consisting of distilled water, toxin solution without an adsorbent, and 0.002% adsorbent in distilled water. The control and test groups were agitated at 1000 rpm on an IKA electric shaker (VIBRAX VXR basic, Werke, Germany) for 2 h at body temperature (37 °C) and ambient temperature (26 °C) for thermodynamic experiments. All samples were then centrifuged at 2000g for 20 min to separate the clay/toxin complex from solution.

From equilibrium isotherms, the toxin concentration in solution was detected by LC/MS/MS and was calculated from the peak area at the toxin retention time. The amount sorbed for each data point was calculated from the concentration difference between test and control groups. These data were then plotted using Table-Curve two-dimensional (2D) and a computer program that was developed using Microsoft Excel to derive values for the variable parameters.[48,55] The best fit for the data was a Langmuir model, which was used to plot equilibrium isotherms from triplicate analysis.[55] The isotherm equation was entered as user-defined functionsq = toxin adsorbed (mol/kg), Qmax = maximum capacity (mol/kg), Kd = distribution constant, Cw = toxin equilibrium concentration

In Table-Curve 2D, estimates for the Qmax and Kd were taken from the double-logarithmic plot of the isotherm. The plot displays a break in the curve. The value on the x axis where the curve breaks is an estimate of Kd–1. The value on the y axis where the curve breaks is an estimate of Qmax.[48,56][48,56] The Qmax was taken from the fit of the Langmuir model to the sorption data. The definition of Kd was given byThe enthalpy (ΔH) was a thermodynamic parameter indicating the total heat released or absorbed during the sorption reaction. It was calculated from the van’t Hoff equation by comparing individual Kd values at different temperatures (37 and 26 °C)R (ideal gas constant) = 8.314 J mol–1 K–1, T (absolute temperature) = 273 + t (°C).

Molecular Modeling of Glyphosate and Paraquat Sorptions onto Clay Surfaces

The montmorillonite clay structure used in this study with the stoichiometry (Si4)IV(Al1.67Mg0.33)VIO10(OH)2 was modeled based on atomic coordinates from the Interface MD model database.[57,58] At pH 7, the structure of a single 2.5 nm layer (5 × 3 × 1) was extracted from the database and periodically replicated to build a single 5 nm layer (10 × 6 × 1). At pH 2, the structure of a single 5 nm (10 × 6 × 1) layer with neutral SiOH and aluminate edges was extracted from the database.[57,58] Subsequently, for the clay model at both pH 2 and 7, the single 5 nm layers were independently replicated to form a second layer with a d001 spacing of 21 Å.[52,59] The layered montmorillonite then was solvated in a 90 Å3 water box with 48 individual glyphosate or paraquat molecules distributed in random configurations and orientations. The initial molecular structures of glyphosate at pH 7 and paraquat at pH 2 and 7 were extracted from the ZINC database.[60] The protonated glyphosate at pH 2 was generated by adding hydrogens to the molecular structure at pH 7 using UCSF Chimera.[61] Random configurations of glyphosate and paraquat originated from short 1 ns simulations of the single molecules at infinite dilution using the generalized Born with a simple switching implicit solvent model.[62]

Multi-ns MD simulations were performed in CHARMM[63] for each modeled system (SM in the presence of glyphosate at pH 7, SM in the presence of glyphosate at pH 2, SM in the presence of paraquat at pH 7, and SM in the presence of paraquat at pH 2). Prior to MD simulations, each system was first energetically minimized and equilibrated to adjust the models to the applied molecular mechanics force field. The energetic minimizations were implemented through 500 steps of steepest descent minimization, 500 steps of adopted Newton–Rapson minimization, and 500 steps of steepest descent minimization. The modeled systems were equilibrated for 1 ns. During the energy minimization and equilibration stages, the montmorillonite layers and either paraquat or glyphosate were constrained using 1 kcal mol–1 Å–1 harmonic constraints on all heavy atoms. Then, all constraints were released, and the modeled systems were simulated for 50 ns under a temperature of 300 K and a pressure of 1 atm. The SHAKE algorithm[64] was used to constrain hydrogen bond lengths. MD simulation snapshots were extracted at 20 ps intervals for subsequent analysis. Five replicate 50 ns MD simulations were performed for each modeled system, each with different initial velocities, for an aggregate of 200 ns per system. All energy minimizations and MD simulations were performed using CHARMM, v41b1.

Parameters and topologies within the simulations were extracted from the INTERFACE force field.[57] The INTERFACE force field can operate as an extension of several commonly used harmonic force fields, including CHARMM,[61] thus enabling simulations of toxin binding on organic/biomolecular and inorganic interfaces,[65−69] including montmorillonite.[58] The topologies of glyphosate and paraquat were generated through CGENFF[70] with low penalties using the molecular structures of the two molecules at pH 2 and 7. The CHARMM36 force field[71] was used to model the water and counter ions.

Structural Analysis and Classification of Glyphosate and Paraquat Binding Conformations

In-house FORTRAN programs and modules were developed to perform a structural analysis on the MD simulation snapshots. The analysis was aimed to explore the binding properties and extract the most prominent binding modes of glyphosate and paraquat to SM at pH 2 and 7. After ∼5 ns, the systems reached equilibrium, based on the convergence of the number of instances at which a toxin was in contact with the clay interlayer. Thus, in the analysis, the first 5 ns of the simulation was considered as an additional equilibration stage, and the following 45 ns of each of the five replicate simulations were analyzed, with snapshots extracted every 20 ps.

As the clay layers were stable and did not undulate throughout the simulations, two planes were defined to represent the two clay layers based on the aluminum atoms at the center of each clay layer (Figure , solid black line). A binding event between glyphosate or paraquat molecules with the clay interlayer was recorded when at least two of their nonhydrogen atoms were within 6.5 Å of either plane. The 6.5 Å cutoff accounts for the distance between the plane and the surface (3 Å) and a 3.5 Å distance cutoff. In addition, the residence time between any of the two molecules and clay interlayer was recorded when glyphosate or paraquat was bound to the clay for a minimum of 40 ps with an analysis resolution of 20 ps. While recording “residence time per binding event”, a grace period of 20 ps was allowed, which accounts for any instances where the compound may spontaneously lose contact to SM and then regain contact in the same binding mode.

To identify the most prominent binding modes of glyphosate and paraquat to SM within the simulations, a statistical analysis was performed on the interactions between the clay interlayer and toxins. The binding conformations of glyphosate or paraquat molecules to the SM interlayer were classified into binding modes based on which set of atoms were in contact with the interlayer surface. Glyphosate was divided into three sets of atoms: the phosphate group, the central carbons and amide group, and the carboxyl group. Paraquat was divided into four sets of atoms: the methyl group, the nitrogens and carbons at the para positions of the aromatic rings, the carbons at the ortho and meta positions of one side of the aromatic rings, and the carbons at the ortho and meta positions of the opposite side of the aromatic rings (Figure ). Upon iterative analysis, certain geometric criteria were defined to cluster the binding conformations into different binding modes. Thus, structurally similar binding conformations were categorized into the same binding mode, while different conformations were categorized in distinct binding modes. The analysis was performed independently for glyphosate and paraquat at pH 2 and 7.

Predicted Bioavailability

The Langmuir model can be rearranged to predict the freely available fraction of toxin as follows. Noting that for a nominal (total) concentration Cnom and clay concentration Cclay, the value of q = (Cnom – Cw)/Cclay, the Langmuir model can be rearranged so that free concentration Cw is given by the solution of the quadratic equationThus, Cw can be predicted using the quadratic formula (taking the positive root). The predicted percent bioavailability is then given by dividing by the nominal concentration: 100 × Cw/Cnom.

Hydra Bioassay

H. vulgaris was obtained from Environment Canada (Montreal) and maintained at 18 °C. The hydra classification method[72] was used with modification[73] to rate the morphology of the adult hydra as an indicator of toxicity and involves a scoring system from 0 to 10, with 10 being “unaffected”. Mature and nonbudding hydra of similar size were chosen for testing to minimize differences between samples. All test solution tubes were capped and prepared by shaking at 1000 rpm for 2 h and centrifugation at 2000g for 20 min prior to exposure of hydra to the toxin in pyrex dishes. All experiments were conducted with three hydra per treatment or control group.

To establish the dose and time dependences of response to toxins alone, experiments were first conducted with glyphosate (0, 10, 30, 40, and 60 ppm) and paraquat (0, 5, 10, 20, and 30 ppm), with hydra morphological response scored and recorded at 0, 4, 20, 28, 44, 68, and 92 h.[74,75] The concentration (C)–time (t)–response (y) relationship was fit to both a power-exponential and a Hill modelThe responses at C = 0 and t = 0 time were constrained to be 10, because all untreated hydra were uniformly unaffected. Model fitting was conducted using nonlinear least squares (nls function in the R statistical software, version 3.4.2); model fits were compared using the AIC, and confidence intervals were calculated based on Monte Carlo estimates from the R package propagate. Additionally, to predict the effect of sorbents, the response was calculated based on the predicted bioavailability (described above) at different nominal concentrations Cnom and different clay concentrations Cclay.

Based on the response to toxins alone and bioavailability predictions, subsequent testing was performed at concentrations of 30 ppm glyphosate and 20 ppm paraquat, with and without cotreatment with sorbent at 0.1 or 0.2%. Responses were scored and recorded at the same time points as mentioned above from 0 to 92 h. A comparison between different treatment and control groups was conducted by comparing confidence intervals from fitting a linear model in time.