How Far Can Pesticides Spread? A Case Study on No-Tillage Farming Systems and Surrounding Forest Patches

With the growing global concern on pesticide management, the relationship between its environmental recalcitrance, food security and human health has never been more relevant. Pesticides residues are known to cause signicant environmental contamination. The present study focused on long-term no-tillage farming systems and subtropical forests in Brazil, where Glyphosate (GLY) has been applied for more than 35 years. GLY and the main breakdown product of glyphosate, aminomethylphosphonic acid (AMPA) were determined by high-performance liquid chromatography coupled with a uorescence detector. In addition, the presence of carbamates, organochlorines, organophosphates and triazines were also detected. GLY and AMPA were omnipresent in soil samples, reaching values higher than those described for soils previously in the literature. GLY and AMPA were observed respectively at peak concentrations of 66.38 and 26.03 mg/kg soil. GLY was strongly associated with forest soil properties, while AMPA associated with no-tillage soil properties. The texture was a signicant factor that contributed to the maximum discrimination of the results since clay and sand contents inuence GLY and AMPA retention in soils. This was the rst study to report DDT and metabolites in consolidated no-tillage soils in Brazil (a pesticide fully banned 2009). Based on human risk assessment and the potential risk of GLY to local soil communities, this study offers a baseline for future studies that will be crucial to understand the toxicity mechanisms on non-target soil biota. were found in all studied areas, and GLY concentrations are the highest ever reported in the world. It was not possible to identify differences due to sampling location along the catena (upper, middle or lower slope) in the different transects, and only one area (NT-B2) had signicant higher AMPA concentrations than the surrounding SF. Based on the high carcinogenic indices calculated in the present study, such concentrations may promote adverse effects on the metabolic homeostasis of humans, but also of soil quality-related edaphic organisms. The results of the present study indicate that land-use inuences the retention of such contaminants since acidic soils with high metal contents increase GLY adsorption. Soil texture discriminated the maximum variation between the study sites, highlighting the importance of clay and sand contents to GLY and AMPA availability. This study can provide bases to guide limits for pesticide applications based on different soil types in Brazil. As so, environmental legislation, such as the CONAMA/420 resolution, requires urgent revision as the values and guiding criteria for pesticides residues in soils report only to commercially prohibited products but not for the commercially approved pesticides (no guiding values exist for GLY and AMPA in soils). DDT and metabolite residues in NT soils conrmed the persistence of these compounds due to high environmental stability. Despite the impossibility to show real values, this is the rst study to report DDT and metabolites in consolidated no-tillage soils in Brazil. Faced with the potential risk to soil communities, future studies using ecotoxicological assays are needed to determine its bioavailability and to understand the mechanisms of toxicity to non-target organisms, essential for the maintenance of ecosystem services.

widespread use of glyphosate was also useful in the worldwide expansion of no-tillage agriculture, particularly in Brazil, where this soilconservation practice occupies over 32 million hectares, almost 60% of the annual crop area of the country 56 .
In the present study, was performed a pesticide screening in no-tillage farming systems (NT) where pesticides have been regularly applied (> 35 years) and in surrounding fragments of secondary Atlantic forest (SF) in intermediate to advanced regeneration state. The study also evaluated the relationships between GLY and AMPA concentrations and soil's physical and chemical characteristics and performed a human health risk assessment using an indirect probabilistic risk model. We hypothesised that location along the catena in uences the concentrations of pesticide residues (e.g. by runoff or wind dispersion) with higher residue levels at sites located at the bottom of the slope when compared to the top.

Study area
The study areas were selected due to the long-term history of pesticide use (> 35 years) under no-tillage farming system (NT) in the State of Parana, Brazil. Study area details are shown in Table 4, and rainfall conditions are showed in Table S4 (SD). The sampled areas (A, B and C) are located in the region with the largest number of pesticide-applied rural establishments (63%) and the second most sprayed per hectare per year in Brazil (9.81 mg pesticides/kg soil) 6 . The exact locations of the sample sites are not included due to privacy reasons but may be provided upon request to the authors and conditioned to acceptance by the owners. The coordinates provided correspond to the closest settlement. The herbicide GLY was applied on average three times a year, with 7.7 L/ha/year or 2.16 mg active ingredient/kg soil/year, and was at least sprayed one month before soil sampled in site NT-A and NT-B and the previous year in NT-C. In NT, GLY had a maximum application rate of three times per year. Table 4: Description of the main characteristics of the three agroecosystems. NT: no-tillage farming system. SF: secondary forest fragment.
Soil classi cation according to FAO (2014) 96 and climate classi cation, according to Köppen-Geiger (1936): CFa -humid subtropical climate, CFb -temperate oceanic climate. All these areas are found within the Atlantic forest biome.  101 . Calibration curves were prepared using GLY (97% purity, Sigma Aldrich), and AMPA (99% purity, Sigma Aldrich) standards. An uncontaminated matrix sample from 1876, con rmed by Remor et al. (2015) 102 , was also analysed in triplicate after forti cation procedures with standard solutions. Other details of the analytical procedure are shown in Supplementary Table S5.
Trace residues analyses of other semi-volatile organic pesticides were performed for the same soil samples. Due to technical problems, the initial analyses could only be repeated after a 480 days storage period at -20°C. The reanalyses were performed by Centro de Biologia Experimental Oceanus LTDA. Screening for the presence of semi-volatile organic pesticides (thirteen-one organochlorines, thirteen organophosphorus, six carbamates, two triazine, and one pyrethroid compounds along with their metabolites) were extracted by a sonication method, according to US EPA method 3550C 103 , followed by gas chromatography device coupled with mass spectrometry, according to US EPA method 8270D 104 .

Data analysis
The matrices of physicalchemical properties of soils were analysed by Kaiser-Meyer-Olkin (KMO) criteria (KMO > 0.5), which evaluates the degree of colinearity among variables 105 . The scores resulting from the principal component analysis (PCA) were evaluated for signi cance using the Kruskal-Wallis test, performing the post-hoc Dunn's test, since the data displayed non-normal distribution (Shapiro-Wilk test) 106 .
The GLY and AMPA values were analysed per transect (upper, middle and lower area) and compared among themselves and with the respective forest in each study area. GLY and AMPA concentrations were also evaluated among NT farms (NT-A, NT-B and NT-C), as well as among forests (SF-A, SF-B, SF-C). As GLY and AMPA concentrations displayed normal distribution (Shapiro-Wilk test) and homoscedasticity of variance (Bartlett test) 108 , they were submitted analysis of variance (ANOVA) test followed by Tukey's post-hoc test. To identify possible pesticide mobility in the catena, two-way ANOVA analysis was used to test for interactions between NT slope position at each NT site using GLY and AMPA values.
The percentage of AMPA in relation to the total modi ed glycines was measured according to equation (1) % AMPA = C AMPA (mg AMPA/kg soil) / C glyphosate (mg GLY/kg soil) + C AMPA (mg AMPA/kg soil) * 100 The AMPA:GLY ratio equation (2)  (2) AMPA:GLY ratio = C AMPA (mg AMPA/kg soil) / C glyphosate (mg GLY/kg soil) The total extracted GLY (TEG) from the soil samples was also calculated considering the AMPA concentration, expressed in GLY mass equivalent basis, added to the GLY concentration, according to the equation (3) TEG = C glyphosate + C AMPA x MW glyphosate / MW AMPA where C glyphosate and C AMPA represent GLY and AMPA concentrations, respectively, MW glyphosate represents GLY molar weight (169 g/mol), and MW AMPA represents AMPA molar weight (111 g/mol).
Soil properties without correlation with GLY, AMPA and AMPA:GLY values according to Spearman's test were not used in the canonical correlation analysis (CCA). The canonical variation represents the values of chemical and physical parameters related to the pesticide concentrations. The eigenvalues represent a correlation between each pair of canonical root axes. A canonical discriminant function was created to separate the groups by maximising the variation between them to the variation within each group. Multivariate analysis of variance (MANOVA) and Pillai test were used as a preliminary test of canonical discriminant function to identify whether variations in treatment levels have a higher in uence on data variance than error.
In all statistical tests, the level of signi cance was 0.05. All data analyses were performed using R (R Core Team) 110 and packages for statistical computing are provided in Table S6 (SD). The semi-volatile organic pesticides were not included in the statistical analysis due to the long-term storage of the samples (480 days at -20ºC).

Human risk assessment
A generic probabilistic risk model was developed to estimate the carcinogenic risk of GLY concentrations found in the present study. The calculation of incremental lifetime cancer risk (ILRC) was performed according to the methodology described by Exposure Factors Manual (2011) 85 and Generic Exposure Routes Assumptions and Data Source Document (2014) 111 and is fully described in the Supplementary data. Details of the parameters used to estimate the risk of exposure to human health are presented in Table S7 (SD). Physiological and behavioural data related to human lifestyle were used according to Qu et al. (2019) 112 .

Soil physical and chemical properties
The physical and chemical soil properties are represented in Table 1, and details of sampling transects in Supplementary Table S1 (  Glyphosate (GLY) and aminomethylphosphonic acid (AMPA) residues The highest GLY concentration in this study was found in a forest soil sample (SF-site A), which concomitantly also represents the highest GLY concentration found in any site studied worldwide (66.38 mg GLY/kg soil - Table 2). GLY reached concentrations up to 66.38 mg/kg soil, while AMPA was detected up to 26.03 mg/kg soil.
There were no signi cant differences observed between the SF areas for GLY, AMPA, AMPA:GLY ratio or TEG (    Screening pesticides occurrence Due to technical di culties, the initial pesticide screening of fty-four parental compounds and metabolites could only be rerun after 480 days of storage. The long-term storage led to the degradation of compounds present in soils, and only the most persistent ones were detected. Since the concentrations would not re ect real concentrations found in the environment, only the presence/absence of these metabolites is provided in Table S2 (SD). Please note that the absence of a particular metabolite does not necessarily mean that it was not present but that its concentration was below detection limits. Traces of p,p'-DDT and its metabolite p,p'-DDE were found above the limit of detection in all NT areas. The metabolite p,p'-DDD was also present in NT-A.

Relationships between soil properties, GLY and AMPA
According to the linear regression (Figure 2A), GLY concentrations were positively correlated with Al 3+ and clay content, and negatively correlated with S, pH and sand content. On the other hand, AMPA concentrations showed positive correlations with Ca 2+ , CEC and clay content, and negative correlations with sand content. AMPA:GLY ratio showed a positive correlation with Ca 2+ , Mg 2+ and CEC. TEG showed positive correlations with Al 3+ , clay content and negative correlations with S, pH and sand content (Table S3 -SD). In the canonical correlation analysis (Figure 2B), the principal axis represented 76% of the total canonical variation (F = 1.6, p = 0.04), with GLY and AMPA:GLY ratio correlating with SF soils while AMPA correlated with NT soils.
In the canonical discriminant analyses ( Figure 3A), the principal discriminant axis Can1 explained 66.7% of the maximisation between groups. This axis seemed to discriminate the samples according to soil properties, mainly between sandy and clayey soils (F = 36.71, p < 0.0001; Figure 3B). Additionally, it differentiated GLY and AMPA concentrations that were dependent on clay and sand content. This was con rmed by the TEG values that showed a positive correlation to clay and negative correlation to sand contents. The axis Can2 explained 21.1% of the maximisation between groups, suggesting a separation between land-use, with NT-farming systems appearing in the positive component and SF-systems in the negative (with the exception of SF-A).

Human risk assessment
The indirect human risk assessment was calculated based on the incremental lifetime cancer risk (ILCR) values presented in Table 3. The model that estimates the probabilistic risk from generic characteristics was built for adults aged between 18 and 70 years old. Even considering the uncertainties of the indirect exposure assessment model, the results showed that GLY and AMPA concentrations in the areas sampled may represent a carcinogenic risk to public health, since they are higher than the reference risk value of 1 x 10 -4 .  33 ). As for the AMPA concentrations in the present study, they are also higher than those found in Portugal (0.73 mg/kg soil) 35 , US (0.18 mg/kg soil) 58 or Argentina (4.20 mg /kg soil) 33 , but lower than maximum values reported for the latter country 33 . Linear correlation and regression analyses con rm the hypothesis of pseudo-persistence attributed to AMPA due to maximum soil half-life 33 , the continuous GLY application and the maximum half-life of the precursor molecule.
In terms of our hypothesis regarding sample location along the catena, we found no differences in either GLY of AMPA concentrations at upper, middle and lower transects, implying that if there was any particle transport with AMPA or drift-derived movement on-site, this did not re ect itself in higher pesticide residues at the transects located in the lowest part of the slope.
Furthermore, only at one site (B) were signi cant differences in AMPA levels detected between contents in an NT transect (2) and the forest, which may be related to higher pH, Ca 2+ and P levels in NT. The higher levels of P in NT-B2 to SF-B may in uence the mineralisation of pollutants in the soil and explain the high levels of AMPA in this transect. Previous studies using microcosms indicate that phosphate levels can cause negative regulation of carbon-phosphorus lyase (C-P lyase) 59,60 , an enzyme responsible for degrading GLY and AMPA. The soil microbial community (not measured in the present study) may also have a preference for the use of inorganic phosphorus as the source of energy, which may help increase the availability of soluble phosphates by solubilising inorganic phosphate 61  Previous literature showed that pesticide degradation by microorganisms is lower in soils with more acid pH 58 . Nonetheless, our results show that is not the most important factor that can explain GLY and AMPA levels in NT and SF. The availability of GLY molecules increases as pH increases as a result of lime application 63 and mineral fertilization 64 . However, these practices may become hazards to surrounding areas due to the sorption and desorption dynamics of modi ed glycine 63 . Liming in NT farms promotes an electrostatic repulsion of GLY due to excess negative electric charges in the soil, which reduces the formation of hydrogen bonds and releases GLY from the chelating reaction 63 . The result is reduced sorption and increased availability of the precursor molecule for degradation, which also may explain AMPA concentrations in NT sites. The AMPA:GLY ratio showed a negative correlation to soils with high pH, thus supporting a higher degradation of GLY observed in NT farms.  33 . To explain the slow degradability of GLY in forest soils, further studies are needed in order to determine whether the high concentrations of AMPA found in soils under NT practices may be also associated with the diversity and role of microbial communities present (e.g., Pseudomonas spp. and Agrobacterium spp.) encoding active ingredient enzymes (eg. oxidoreductases) 65 that use GLY as a source of carbon, nitrogen and phosphorus 66 .
GLY sorption is favoured in acidic environments 63,67 , due to the protonated soil solution that promotes the formation of different GLY speciation forms with a positive net charge, facilitating the formation of GLY functional group complexes (R-NH, R-COOH and R-PO(OH)) with the metals present in those soils 68,69 . Previous studies 9,70-72 consider that amorphous metal oxides and hydroxides play a major role in GLY sorption. The solubility and availability of Al 3+ of the cation exchange complex are pH-dependent, and the levels of Al 3+ and H + Al observed in the SFs of the present study showed a negative correlation with pH. This raises the hypothesis that GLY retention in forest soils increases the half-life of this active product, reducing availability for biological degradation. This can be the result of the forest environmental characteristics, with high metal ion contents, acidic pH, as well as reduced light and higher C sources. The mechanism of GLY retention in forests may represent a slow and continuous source of AMPA for environmental compartments, highlighting, therefore, the need to preserve the stability of forest systems due to their buffering effects on environmental contaminants.
In the present study, soil texture was the factor with the highest discriminant variation. According to the literature, GLY mobility is in uenced by soil texture 10,63 , with clayey soils being less favourable for mobility 73 , while sandy soils increase transport phenomena 74 and vertical mobility 51 to groundwater 75 . In addition to the reduced content of metal oxides (for complexation of the R-PO(OH) group with soil colloids), the speed of water ow in macropores of sandy soils in uences the transport of bound and unbound contaminants. Fast ow reduces the time for adsorption (non-instantaneous process) of GLY with the soil matrix to be in equilibrium 74 . This soil contamination poses a risk not only to food quality and human health (since these compounds have been identi ed as endocrine disruptors) but also to ecological processes for ecosystem services maintenance [76][77][78][79] . The presence of GLY and AMPA in soil may promote toxicity to key species for biodiversity conservation 80 , which are fundamental for maintaining the interactions that ensure environmental services.
In addition to the risks to soil organisms, GLY is also considered as a probable carcinogen for humans 81 . In this sense, the use of tools that can assess exposure risk to humans are of utmost importance. These tools can be used directly, indirectly and from reconstructed scenarios.
According to EFSA 82 , chronic exposure of pure GLY concentrations at the rate of 1.4 mg GLY/kg/day increased the incidence of malignant lymphomas in rats, while chronic exposure of parental rabbits (1 mg GLY/kg/day) in uenced offspring development due to delayed ossi cation and increased skeletal and cardiac malformations. In a recent review, Van Bruggen et al. (2018) 83 pointed to the positive correlations between GLY use and increased diseases such as kidney disease, attention de cit hyperactivity disorder (ADHD), Alzheimer's and Parkinson's disease, abortions, and dermatological diseases. In previous studies, continuous low exposure (70 mg GLY/kg/day) was suggested to promote neurotoxic effects by altering acetylcholinesterase activity 84 . GLY increased production of reactive oxygen species (> 42 mg GLY/L), leukocyte DNA damage (85 to 1,690 mg GLY/L) and reduced DNA methylation (42 mg GLY/L) have been reported in vitro 83 .
Even if the concentrations that presented risk in vitro are higher than those obtained in the present study, the contamination levels observed in the areas can be considered a problem of global concern. When calculating the human risk equations, oral intake was determined to be one of the most important constants, contributing to the risk. Its importance is even greater if we consider that the limits of GLY in food, as coffee (1 mg/kg BR vs 0.1 mg/kg EU), sugar cane (1 mg/kg BR vs 0.05 mg/kg EU), soybean (10 mg/kg BR vs 0.05 mg/kg EU) and drinking water (0.5 mg/L BR vs 0.0001 mg/L EU) permitted in Brazil are well above the limits accepted by the EU 6 . Despite the human health risk assessment having uncertainties and limitations due to generalisations 85 , the results obtained from the present model should be taken as a warning, and should be used to review GLY application regulations and policies. In countries such as Sri Lanka and El Salvador, for example, GLY has been banned, since it can trigger chronic kidney disease of unknown etiology in farmers 86 .
The occurrence of p,p'-DDT, p,p'-DDD and p,p'-DDE in soils at the sample sites also represents a risk to ecosystems, based on their lipophilic pro les (log Kow > 3  [90][91][92] . Despite the adoption of soil conservation practices like no-tillage that result in important environmental bene ts, the use of more sustainable and less-intensive pest management techniques need to be further explored and developed for widespread use on-farm, in order to reduce dependence on chemical pesticides. Current intensive agricultural practices require pesticide application in most cases, but the high residual levels found in the present study highlight the need for further studies on the reasons for the high persistence of both GLY and AMPA in soils 93 as well as studies on its toxicity to soil organisms 94 and for less persistent alternative herbicides and improved weed management practices. Finally, our results also re ect the need to update the Brazilian Resolution 420 of the National Environmental Council (CONAMA/420) 95 , in order to review guideline values for contaminants such as DDT. As for GLY and AMPA residues in Brazilian soils, at the moment, there is no legislation which limits the maximum concentration in soils or number of applications. This is particularly concerning considering the frequency of applications each year (3 times at the farms studied), and because every ve spraying events was estimated to result in the increase of 1 mg GLY/kg soil in farms in neighbouring Argentina 33 .

Conclusion
This study collected soil from three different farming areas and nearby forest patches in order to perform a pesticide screening. This is the rst study in Brazil that quanti ed GLY and AMPA in soils under NT farming practices and surrounding subtropical SF. GLY and its main metabolite AMPA were found in all studied areas, and GLY concentrations are the highest ever reported in the world. It was not possible to identify differences due to sampling location along the catena (upper, middle or lower slope) in the different transects, and only one area (NT-B2) had signi cant higher AMPA concentrations than the surrounding SF. Based on the high carcinogenic indices calculated in the present study, such concentrations may promote adverse effects on the metabolic homeostasis of humans, but also of soil quality-related edaphic organisms. The results of the present study indicate that land-use in uences the retention of such contaminants since acidic soils with high metal contents increase GLY adsorption. Soil texture discriminated the maximum variation between the study sites, highlighting the importance of clay and sand contents to GLY and AMPA availability. This study can provide bases to guide limits for pesticide applications based on different soil types in Brazil. As so, environmental legislation, such as the CONAMA/420 resolution, requires urgent revision as the values and guiding criteria for pesticides residues in soils report only to commercially prohibited products but not for the commercially approved pesticides (no guiding values exist for GLY and AMPA in soils). DDT and metabolite residues in NT soils con rmed the persistence of these compounds due to high environmental stability. Despite the impossibility to show real values, this is the rst study to report DDT and metabolites in consolidated no-tillage soils in Brazil. Faced with the potential risk to soil communities, future studies using ecotoxicological assays are needed to determine its bioavailability and to understand the mechanisms of toxicity to non-target organisms, essential for the maintenance of ecosystem services.