Glyphosate accumulation, translocation, and

Europ. J. Agronomy 74 (2016) 133–143

Contents lists available at ScienceDirect

European Journal of Agronomy journal homepage: www.elsevier.com/locate/eja

Glyphosate accumulation, translocation, and biological effects in Coffea arabica after single and multiple exposures Lars C. Schrübbers a,∗ , Bernal E. Valverde a,b , Bjarne W. Strobel a , Nina Cedergreen a a b

Department of Plant and Environmental Sciences, University of Copenhagen, Frederiksberg C, Denmark Investigación y Desarrollo en Agricultura Tropical S.A. (IDEA Tropical), Alajuela, Costa Rica

a r t i c l e

i n f o

Article history: Received 15 July 2015 Received in revised form 22 November 2015 Accepted 24 November 2015 Available online 31 December 2015 Keywords: Glyphosate Coffea Arabica Coffee Spray drift UHPLC-MS/MS LC–MS Shikimic acid accumulation Multiple applications

a b s t r a c t In perennial crops like coffee, glyphosate drift exposure can occur multiple times during its commercial life span. Due to limited glyphosate degradation in higher plants, a potential accumulation of glyphosate could lead to increased biological effects with increased exposure frequency. In this study, we investigated glyphosate translocation over time, and its concentration and biological effects after single and multiple simulated spray-drift exposures. Additionally, shikimic acid/glyphosate ratios were used as biomarkers for glyphosate binding to its target enzyme. Four weeks after the exposure, glyphosate was continuously translocated. Shikimic acid levels were linear correlated with glyphosate levels. After two months, however, glyphosate appeared to have reduced activity. In the greenhouse, multiple applications resulted in higher internal glyphosate concentrations. The time of application, however, was more important regarding biological effects than the number of applications both in the greenhouse and in the field. In the field, berry yield, the most important biological response variable, was reduced 26% by the first out of four sequential applications of glyphosate at 64 g a.e. ha−1 each. The three subsequent applications did not reduce yield any further. © 2015 Published by Elsevier B.V.

1. Introduction The non-selective, post-emergence herbicide glyphosate [N(phosphonomethyl) glycine], was first commercialized approximately 40 years ago. Today, it is the most used plant protection chemical worldwide (Steinmann et al., 2012). Once taken up, glyphosate is mobile in both xylem and phloem. Because of fast reloading into the phloem (Preston and Wakelin, 2008), glyphosate translocation is, however, predominant in the phloem from source to sink (Shaner, 2009), following the path of sucrose and other photosynthates. Its target site is the enzyme 5enolpyruvylshikimate-3-phosphate synthase (EPSPS) and upon its inhibition by glyphosate shikimic acid accumulates. The mode of action and the resulting accumulation of shikimic acid are well documented (Amrhein et al., 1980; De Maria et al., 2006; Schonbrunn et al., 2001). An increased level of shikimic acid can be used as biomarker for a recent exposure to glyphosate. An accumulation of shikimic acid can additionally be used as an effect measure on a molecular level indicating the degree of EPSPS inhibition.

∗ Corresponding author. Fax: +45 35332398. E-mail address: [email protected] (L.C. Schrübbers). http://dx.doi.org/10.1016/j.eja.2015.11.023 1161-0301/© 2015 Published by Elsevier B.V.

Because of its non-selective mode of action, glyphosate possesses the risk of damaging the crop through unintended spray drift. Various studies have been carried out to evaluate glyphosate crop toxicity via spray drift simulation experiments. Annual crops such as corn (Buehring et al., 2007), rice (Koger et al., 2005), soybean (Ding et al., 2011), peanut (Lassiter et al., 2007), and wheat (Rolder et al., 2007) were investigated as well as perennial plants grown as annuals including potato (Felix et al., 2011) and tomato (Gilreath et al., 2001), and woody perennial plants such as grape vine (Al-Khatib et al., 1993), cherry trees (Al-Khatib et al., 1992) and coffee (Franca et al., 2013; Schrübbers et al., 2014). The latter group, in contrast to annual plants and perennials grown as annuals, is potentially exposed many times during their commercial lifespan. Coffee is grown in regions without cold winters; making weed control necessary all year round; especially in the rainy season with favorable conditions for weed growth (Staver et al., 2001). Multiple exposures make an evaluation of glyphosate accumulation important for perennial crops, as a potential accumulation could lead to increased effects with time. Glyphosate already present in the plant could add to an otherwise non-damaging dose to reach an internal concentration high enough to cause a toxic effect. Damage will only take place if glyphosate is not sufficiently degraded, excreted, and/or stored in a place or form that avoids its interaction with

134

L.C. Schrübbers et al. / Europ. J. Agronomy 74 (2016) 133–143

the target enzyme within the interval of the multiple applications (Dayan et al., 2010; Duke, 2011; Pfleeger et al., 2014). Despite a reported slow or absent glyphosate degradation in higher plants (Duke, 2011), the literature is sparse on studies evaluating glyphosate concentration and effects in plants several weeks after single and, especially, multiple exposures (Pfleeger et al., 2014). An example are glyphosate resistant (GR) soybean leaves and stems, in which 3–4 mg glyphosate kg−1 were detected several weeks after single or multiple applications (Arregui et al., 2004). Generally, higher glyphosate concentrations were found in plants exposed multiple times to the herbicide. Information on the risk of accumulating internal concentrations by multiple exposures could potentially also be obtained from degradation curves after a single application, because an accumulation of glyphosate requires the chemical to be still present from the previous application. However, a decrease in glyphosate concentration might mainly be caused by a dilution effect due to plant growth, as shown in corn (Bernal et al., 2012). This dilution effect can be difficult to estimate for longer periods when plant growth varies with plant age and environmental conditions. Additionally, the biological effect of glyphosate will depend on the mobility of the herbicide within the plant both between organs and cell compartments, as the target site EPSPS is mainly located in the cytoplasm of active meristems (Shaner, 2009). Studies with GR horseweed (Ge et al., 2010, 2011) have shown that translocation of glyphosate to vacuoles can decrease the biological effect and thereby act as a resistance mechanism. As aforesaid, the accumulation of shikimic acid is an important identification criterion for glyphosate reaching its target enzyme. The correlation of glyphosate and shikimic acid concentrations over time has to our knowledge not been evaluated. The few studies we are aware of analyzing both analytes in the same sample either used GR soybean plants (Duke et al., 2003; Reddy et al., 2004) that do not accumulate shikimic acid after glyphosate exposure; or investigated only short intervals after application, i.e. 72 h in cotton (Pline et al., 2002) or 7 days using several plant species (Reddy et al., 2008). The limited glyphosate degradation in higher plants (Duke, 2011), however, makes the evaluation of the shikimic acid/glyphosate ratio over longer intervals interesting. When glyphosate is degraded, its main metabolite is aminomethylphosphonic acid (AMPA). AMPA is considered phytotoxic in itself; but less potent than glyphosate and has a different mode of action, as no shikimic acid accumulation is observed in plants exposed to AMPA only (Reddy et al., 2004). To investigate the biological effects of glyphosate in a plant over time, knowledge of AMPA concentrations are therefore necessary. The aim of this study was to investigate glyphosate accumulation, translocation and biological effects of one and multiple sub lethal doses in coffee plants over extended periods. Coffee was chosen as an example of a perennial crop typically experiencing several glyphosate applications per year. It was also selected for its high economic importance and because glyphosate related damage symptoms are often observed in coffee plantations despite careful glyphosate application (Arizaleta et al., 2008; Franca et al., 2010; Rodrigues et al., 2003). Coffee leaves were analyzed for glyphosate and AMPA after a single glyphosate application in different plant parts (compartments) at different intervals to obtain information on glyphosate translocation in the plant and thereby its ability to reach the target site after several weeks. Additionally, samples previously analyzed in our laboratory (Schrübbers et al., 2014) for shikimic acid were analyzed for glyphosate and correlated to evaluate in situ target site binding. Multiple exposures were applied in a greenhouse and field trial to estimate possible glyphosate accumulation and the effect on growth, plant health and coffee yield. Because of the slow degradation in higher plants, we hypothesize higher glyphosate concentration and increased phytotoxicity would be observed after several glyphosate applications.

2. Materials and methods 2.1. Chemicals and solutions Glyphosate (purity ≥97.0%) and AMPA (99%) were purchased from Sigma–Aldrich (Steinheim, Germany). The isotope labeled internal standards 1,2-13 C2 15 N glyphosate (98%) and 13 C 15 N AMPA (99% for 13 C, 34% for 15 N) were purchased from Dr. Ehrenstorfer (Augsburg, Germany). The organic solvents methanol and acetonitrile were HPLC grade, obtained from Rathburn (Walkerburn, Scotland). For the mobile phase, LC–MS grade (Chromasolv® ) acetonitrile from Sigma–Aldrich (Steinheim, Germany) was used. Dichloromethane, potassium hydroxide (p.a.) and boric acid (all EMSURE® p.a.) were purchased from Merck (Darmstadt, Germany). Hydrochloric acid (37%, AnalaR NORMAPUR) and sodium hydroxide were obtained from VWR (Fontenay-sous-Bois, France), and J.T. Baker (Deventer, Netherlands), respectively. Ammonium formate (MS grade ≥99.0%), ammonium hydroxide (LC–MS grade), formic acid (98%), and 9-fluorenylmethylchloroformate (FMOC chloride; ≥99%) were all from Sigma–Aldrich (Steinheim, Germany). All FMOC solutions were prepared the same day in acetonitrile. Borate buffer solutions consisted of 500 mM boric acid adjusted to pH 9 with 1 M NaOH. Stock solutions were prepared by dissolving 25.0 mg powder of the analyte in 25.0 mL MilliQ water. To dissolve the material, the solutions were sonicated with a Branson 1510E-DTH (Branson Ultrasonic Corporation, USA) at 40 ◦ C; 2 × 8 min for glyphosate and 1 × 8 min for AMPA. The internal standards were obtained as aqueous solutions from the manufacturer. Dilutions to prepare working solution were made in MilliQ. Standard curves were prepared by mixing different volumes of working solutions and water to give 1.2 mL with 1.2 mL borate buffer in 15 mL falcon tubes. Subsequently, 600 ␮L 10 mM FMOC solution were added, mixed and let stand for one hour. After the derivatization was completed, 2 mL dichloromethane were added, vortex mixed, centrifuged at 4000 rpm and the upper water phase was filtered (0.2 ␮m PTFE membrane, 13 mm syringe filer, VWR International, USA) into a LC-vial prior to injection. 2.2. Sample preparation Glyphosate was determined according to a method previously developed for coffee leaves (Schrübbers et al., 2015). The method was carried out with minor alterations and an instrumental change; the high performance liquid chromatography (HPLC) single quadrupole of the original method was substituted by an ultra performance liquid chromatography (UPLC) triple quadrupole (QqQ) tandem mass spectrometer system. Briefly coffee leaves (Coffea arabica L.) were stored for at least one night at −80 ◦ C and subsequently freeze-dried (Hetosicc, Heto, Birkerød Denmark). To homogenize and to reduce the particle size the sample was placed in a 250 mL PEHD bottle together with three stainless steel balls (15 mm in diameter) and ground to a fine powder by shaking using a conventional paint mixer (SO-40a, Fluid Management Europe, IDEX Corporation, Sassenheim, Netherlands). Samples for the correlation of shikimic acid and glyphosate concentration were prepared as described in Schrübbers et al. (2014) and 0.5 g was used for the analysis of glyphosate. For all other samples 1 g or all available material was used. The powder was added to a 50 mL falcon tube and spiked with internal standard (0.8 ␮g and 0.38 ␮g 1,2-13 C2 15 N glyphosate and 13 C 15 N AMPA, respectively). The extraction was carried out with water (18 mL) and 1 M HCl (2 mL). The sample was shaken, sonicated (Branson 1200, Branson Ultrasonic Corporation, USA) for 5 min and thereafter placed on a horizontal shaker (build in-house) for 30 min at 1.1 Hz. After the extraction, the sample was centrifuged for 5 min at 4000 rpm at 21 ◦ C. An aliquot of the supernatant (10 mL) was loaded to a 6 mL, 500 mg Strata-X solid


135

Fig. 1. Proposed structure of the product ions of glyphosate and AMPA. The internal standard product ions can be derived from the analyte product ions. The positions of the stable isotopes are marked with an asterisk. Quantifiers are on the left, qualifier ions are on the right side.

phase extraction (SPE) cartridge (Phenomenex, Torrance, CA, USA) conditioned with methanol (10 mL) and 0.1% formic acid (10 mL). The analytes were not retained on the cartridges and the eluate was collected in a 15 mL falcon tube, followed by 0.1% formic acid (3 mL) into the same falcon tube. The eluate was transferred to a 50 mL falcon tube containing 4 M KOH (0.35 mL) and 500 mM borate buffer, pH 9 (4 mL). The derivatization reaction was started by adding 35 mM FMOC solution (4 mL). After shaking, the reaction was carried out for one hour at room temperature and the sample was subsequently centrifuged for 5 min at 4000. An aliquot of 3 mL supernatant was transferred to a new 15 mL falcon tube containing dichloromethane (2 mL). The sample was vortexed, centrifuged at 4000 rpm and the upper water phase was filtered (0.2 ␮m PTFE membrane, 13 mm syringe filer, VWR International, USA) into a LC-vial for analysis. 2.3. Liquid chromatography tandem mass spectrometry For the chemical analysis a UPLC system equipped with degasser, autosampler and binary pump (Acquity I-class, Waters, Milford, MA, USA) was coupled to a triple quadrupole tandem mass spectrometer (Xevo TQD, Waters, Milford, MA, USA) with electrospray ionization (Zspray). As analytical column an Acquity UPLC® BEH C18, 1.7 ␮m, 150 mm × 2.1 mm i.d. (Waters, Milford, MA, USA), connected to a Acquity UPLC® BEH C18 VanGuard Pre-column, 300 Å, 2.5 ␮m, 5 mm × 2.1 mm i.d. (Waters, Milford, MA, USA) was used. The binary mobile phase consisted of 5 mM ammonium formate, pH 9 with 10% acetonitrile (solvent A) and acetonitrile (solvent B) delivered with a flow rate of 0.45 mL min−1 . The injection volume was 2.5 ␮L and the column temperature was kept at 40 ◦ C. The gradient was set to 1% solvent B from 0.0 to 0.5 min, increasing linear to 30% B from 0.5 min to 2.0 min. From 2.0 min to 2.1 min solvent B was increased to 90% and kept for 1.5 min until 3.6 min. From 3.6 to 3.7 solvent B returned to the starting condition of 1% and kept for 1.3 min until 5 min. The ion source temperature was 150 ◦ C with a capillary voltage of 1.1 kV. The desolvation temperature was 500 ◦ C with nitrogen gas flow of 1000 L h−1 . The cone gas flow was 2 L h−1 . The LM and HM resolution 1 and 2 were set to 10.0 for all four resolution settings. The tandem mass settings for the analytes and internal standards are given in Table 1. For the first

Fig. 2. Total amount glyphosate over time given in ␮g per plant, detected after a single glyphosate application at two different rates in three leaf compartments, divided according to the exposure level. The oldest leaves of the plants (“slightly exposed”) were partially covered by the younger leaves (“highly exposed”) at the day of spray application. The youngest leaves, grown after the spray application (“not directly exposed”) made up the third compartment. Error bars are given as standard deviation (n = 3). Negative error bars are not shown.

minute after the injection the flow was directed to waste to reduce ion source contamination with salts from the samples. Proposed structures of the fragments for the analytes and internal standards (ISTD) are given in Fig. 1. All glyphosate concentrations are given as dry weight. 2.4. Method validation The method was validated by determining its accuracy and precision, as suggested by the SANCO guide (European Commission Health and Consumer Protection Directorate General, 2013) using two blank samples and two spiking levels. The lower spike level was selected after preliminary tests, to be close to the expected limit of quantification (LOQ) and the higher level 10 times the expected LOQ. The limit of detection (LOD) was calculated as the mean blank value plus 3 times the standard deviation of the lower spike level.

136


Table 1 Tandem mass spectrometry parameters settings including precursor [M+H]+ and product ions serving as quantifier and qualifier for the derivatized analytes glyphosate and AMPA and the corresponding internal standard. Compound

Precursor ion (m/z)a

Product ion (m/z)a

Cone voltage (V)

Collision energy (eV)

Glyphosate-FMOC

392 392 334 334 395 395 336 336

88 179 156 112 91 179 158 114

32 32 28 28 32 32 28 28

22 24 8 16 22 24 8 16

AMPA-FMOC 1,2-13 C2

15

N Glyphosate-FMOC

13

C

a

Quantifiers are in bold.

15

N AMPA-FMOC

Fig. 3. Shikimic acid concentration of greenhouse samples in comparison to the corresponding glyphosate concentration. Shikimic acid concentration was obtained from Schrübbers et al.16 and is shown next to the corresponding glyphosate concentrations (n = 3) in the same sample (A). Shikimic acid concentration significant different from the controls are marked with an asterisk. The same dataset presented as shikimic acid concentration per glyphosate concentration (B). Error bars are given as standard deviation.

The LOQ was calculated in the same way but using 10 times the standard deviation. The spike levels were 0.1 and 1 mg kg−1 plant powder respectively. 2.5. Overview of the experiments conducted In total three greenhouse experiments and one field experiment were carried out. While the plants in first two greenhouse experiments were exposed to glyphosate only once, multiple applications were carried out for the third experiment. The first greenhouse experiment (glyphosate accumulation and allocation over time; single application) served to evaluate glyphosate accumulation and allocation over time. The second greenhouse experiment (correlation of shikimic acid and glyphosate concentrations) allowed correlating shikimic acid levels to glyphosate levels within the same sample. The third greenhouse experiment (multiple applications; greenhouse experiment) as well as the field experiment (multiple applications; field experiment) was used to evaluate glyphosate concentrations and biological effects after multiple exposures. AMPA was not detected in any sample in concentrations higher than the LOQ. 2.6. Glyphosate accumulation and allocation over time; single application Greenhouse experiments were conducted in Denmark at Højbakkegård (Taastrup, position: 55◦ 40 12 N, 12◦ 18 34 E). Coffee plants cv. Caturra were obtained from a coffee farm in Costa Rica and grown under greenhouse conditions with supplementary heating and artificial light (day/night; 26/20 ◦ C, 12 h/12 h) Each plant was grown in an individual pot, with soil (Pindstrup Færdigblanding 1, pH 5.5–6, low fertilization state) from a commercial source

(Pindstrup Mosebrug, Denmark). Plants were kept well-watered throughout the entire experiment. Water was provided by flooding the base of the pots. Plants were fertilized using a commercial mixture (Pioner NPK Macro 14-3-23 + Mg plus Pioner Micro M + Fe, Azelis, Lyngby, Denmark) added to the irrigation water in quantities of 20 kg and 2.5 L in 244 L water, respectively. The fertilizer solution was adjusted with nitric acid to pH 5.5. The light intensity was measured (LI-250 light meter, LI-COR Biosciences, Lincoln, USA) and varied between 1300 ␮mol m−2 s−1 and 200 ␮mol m−2 s−1 during the experiment. At the start of the experiment plants were 29.8 cm tall (16% relative standard deviation, RSD) with 30.5 leaves per plant (24% RSD). Glyphosate was applied using a spraying cabinet equipped with Hardi LD-02-110 hydraulic nozzles delivering 150 L ha−1 at 4 bar. Plants were sprayed from above; spray solutions contained Tween 20 (0.1% v/v). Tween 20 was used as adjuvant to reduce the water surface tension and increase glyphosate uptake; simulating a commercial glyphosate formulation. Glyphosate was of technical grade (Cheminova, Denmark) containing 570 g a.e. (acid equivalent) L−1 as the isopropylamine salt (770 g L−1 ). A full field rate was considered to be, 4 L Roundup® ha−1 (Carvalho et al., 2013; Franca et al., 2013) with 360 g a.e. L−1 glyphosate content, corresponding to 1440 g a.e. ha−1 . The rates applied were 0, 27, and 107 g a.e. ha−1 . Plants were marked with red ribbons to distinguish leaves with different exposure levels. Three types of exposure levels were defined; leaves that received a full dose (“highly exposed”), leaves that were only slightly exposed because they were shielded by the younger leaves that received the full dose (“sightly exposed”) and new leaves grown after the spray application and therefore not exposed directly (“not directly exposed”). The experiment was started with nine plants per treatment and the complete leaf material of three replicates were sampled at 3, 28 and 84 days after the treatment (DAT) and ana-


lyzed for glyphosate and AMPA; only two replicates were analyzed for the controls. For the first sampling event (3 DAT) no leaf material was available for the “not directly exposed” compartment due to the limited growth of the coffee plants. In order to calculate the amount of glyphosate per compartment, the concentration found per gram was multiplied with the total amount of sample material present. During the 84 day duration of the experiment, plant height increased approximately 50% (to 147% of the starting height) and fresh weight of the canopy approximately doubled (weight increased to 202% of the starting weight). 2.7. Correlation of shikimic acid and glyphosate concentrations Shikimic acid concentrations were taken from a previous publication of our laboratory (Schrübbers et al., 2014). Greenhouse conditions and glyphosate rates applied were as described above for the experiment: glyphosate accumulation and allocation over time; single application. The rates applied were 0, 86, and 432 g a.e. ha−1 . Seven-month old, 16–36 cm tall plants (n = 12), bearing 12–20 leaves were treated with glyphosate and the four youngest, fully developed (i.e. horizontal stage) leaves were sampled 7, 14, 28 and 56 DAT. For each treatment three replicates were analyzed for glyphosate. The shikimic acid and glyphosate contents were obtained from the extraction of the same homogenized sample. 2.8. Multiple applications; greenhouse experiment Greenhouse conditions and glyphosate application were as described above for the experiment: glyphosate accumulation and allocation over time; single application. Plants were 28.5 cm tall (13% relative standard deviation, RSD) with 30.5 leaves per plant (24% RSD). Four spraying events, with rates of 0, 27, 54, and 107 g a.e. ha−1 were carried out on the 22nd November 2011, 20th January 2012, 21st March 2012 and 21st May 2012, imitating a 2 month interval between spraying in the field (application scheme and results in Fig. 4). In May 2012, 168 days after the first glyphosate application, plants were sampled by taking every leaf growing directly attached to the main stem. Leaves were washed with water to remove glyphosate adhered to the surface. Plant height approximately doubled during the 168 day experiment (height increased to 217% of the start height) and fresh weight of the canopy increase about five times (weight increased to 515% of start weight). 2.9. Multiple applications; field experiment Field studies were conducted at a commercial coffee farm in Tambor, Alajuela, Costa Rica (10◦ 02 34 N, 84◦ 14 20 W). Coffee plants cv. Caturra, eighteen-month old with a mean height of 132 cm (7% relative standard deviation, RSD, n = 52) were selected and tagged with metal labels in November 2011. Coffee plots were hand-weeded during the course of the experiment. In total one to four glyphosate applications were made, on the 23rd November 2011, 25th May 2012, 19th August 2012 and 22nd November 2012 (application scheme and results in Fig. 5). The herbicide was sprayed on the foliage using a CO2 -operated manual sprayer modified to facilitate the application of 30 mL of spray solution containing 0, 360 or 1800 mg glyphosate a.e. L−1 , equivalent to 0, 11 and 54 mg a.e. per plant (n = 6). The rate applied was calculated based on the mean plant diameter of 147 (8.8% RSD). Field rates were equivalent to 64 g a.e. ha−1 for the 11 mg a.e. per plant and 318 g a.e. ha−1 for the 54 mg per plant. The high rate was applied only the during the last spraying event to ensure detectable glyphosate levels after several months and to be able to compare the effect on fruit yield of a high rate with multiple applications. Spray volume was delivered by a single Tee Jet 8001VS flat fan nozzle at 2.07 bar. The volume selected (30 mL per plant) was sufficient to wet the foliage without

137

runoff. Glyphosate solutions were prepared from a Roundup® 35.6 SL formulation (Monsanto) containing 356 g a.e. L−1 as isopropylamine salt. The pH was adjusted to 5.0 using a commercial pH corrector (Indicate 79% L, Germinare, Mexico). No additional surfactant was added. Plywood boards were used to shield neighboring plants avoiding involuntary spraying and drift. Samples for glyphosate quantification were collected on the 5th July 2013 by randomly picking leaves from different parts of the plants. Additionally the branch leaf coverage was estimated twice, shortly after the last spraying event in November 2012 at 1 and 6 days after the last treatment, because it was noted that several plants had shed many leaves during the course of the experiment. The two estimates were pooled. The leaf coverage was assessed by visual inspecting the plants with the evaluator not knowing the treatment. If all branches were bearing full number of leaves the branch leaf coverage was set to 100%. If branches were partly bald, the amount of coverage was estimated for each individual tree, e.g. if half of the branches were completely bald, the leaf coverage was estimated to be 50%. Yield was assessed at three occasions, December 2012, December 2013, and December 2014 by collecting the ripe red beans and measuring the total fruit weight per plant. 2.10. Statistical treatment For statistical comparison of mean values Analysis of Variance (ANOVA, Holm Sidak method, SigmaPlot 12.3, Systat Software Inc, USA) or a t-test was used. 3. Results 3.1. Glyphosate and AMPA analysis The recovery of the modified method was determined and LOQ and LOD values were calculated. Glyphosate and AMPA were not detected in the blank samples. The mean recovery (n = 5) for the lower spiking level (0.1 mg kg−1 ) was 84.6% (8.4% relative standard deviation, RSD) and 106.3% (19.6% RSD) for glyphosate and AMPA, respectively. The mean recovery (n = 5) for the high spiking level (1 mg kg−1 ) was 111.2% (13.7% RSD) and 102.9% (6.9% RSD) for glyphosate and AMPA, respectively. The calculated LOD and LOQ values for glyphosate were 21 and 71 ␮g kg−1 of freeze-dried leaf powder, respectively. For AMPA the LOD and LOQ were 63 and 209 ␮g kg−1 , respectively. 3.2. Glyphosate accumulation and allocation over time; single application In order to evaluate mobility of glyphosate over extended periods, the total amount present in the leaves was determined in three leaf compartments. The glyphosate amount (in ␮g) was calculated by multiplying the measured concentration by the weight of the total compartment sampled (Fig. 2). Not directly exposed leaves were not yet developed three DAT. The total amount i.e. the addition of glyphosate found in all three compartments decreased over time for the low rate (27 g a.e. ha−1 ) from 9.4 to 5.5 ␮g plant−1 , while it increased from 13 to 38 ␮g plant−1 for the high rate (107 g a.e. ha−1 ). However, none of the differences in total glyphosate amount found within one rate was statistically significant (ANOVA, P > 0.05), despite more than doubling for the high rate from 3 to 84 DAT. For the highly exposed compartment at 84 DAT the glyphosate amount in ␮g was lower (ANOVA, P > 0.05) with 1.2 and 8.1 ␮g for the low and the high rate, respectively compared to the two earlier sampling times (low rate: 8.6 and 6.9 ␮g; high rate: 12.0 and 15.1 ␮g), indicating allocation away from the exposed leaves. The

138


Fig. 4. Coffee plants displaying symptoms as defined as percentage of leaves with glyphosate symptoms (A), above soil biomass (B), and internal glyphosate concentrations (C), after multiple glyphosate applications of different rates. Controls are displayed as the first bar from the right. Mean glyphosate concentrations were measured in all leaves attached directly to the stem after 1 up to 4 applications, sampled on 25th May 2012, 168 days after the first spraying event. For the evaluation of leaves displaying symptoms and above soil biomass an additional rate was taken into account (54 g a.e. ha−1 ). The last glyphosate application (21st May 2012; marked with the hash key) was not considered for the percentage of leaves with glyphosate symptoms and above soil biomass, as glyphosate is slow acting and this application would not affect the biological response within three days. Treatments that were different from the controls are marked with an asterisk. Error bars are given as standard deviations.


139

Fig. 5. Coffee bean yield in kg plant−1 of two sequential years (grey bars) and coffee plant branch leaf coverage (blue diamonds) in % after a single or multiple glyphosate applications of 64 g a.e. ha−1 . The last application rate in November 2012 was either 61 or 306 g a.e. ha−1 . Beans were harvested December 2012 and December 2013. Coffee plant branch leaf coverage was estimated November 2012 only. Leaves were sampled for glyphosate analysis in July 2013. Error bars are given as 95% confidence interval (n = 6, controls n = 10); for the yield the error bars are given for the combined yield from both harvests. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

not directly exposed new leaf compartment seemed to be the sink for this glyphosate, since there was a trend (ANOVA, P > 0.05) for both rates towards an increase in glyphosate from 28 DAT to 84 DAT; for the low rate glyphosate increased from 1.4 to 3.9 ␮g, while the high rate increased from 4.2 to 26.4 ␮g.

3.3. Correlation of shikimic acid and glyphosate concentrations To compare shikimic acid accumulation with internal glyphosate concentrations, leaf samples previously analyzed for shikimic acid (Schrübbers et al., 2014) were evaluated for glyphosate (Fig. 3). Glyphosate was absent or at low concentration in the controls. The highest concentration (0.23 mg kg−1 ) in the controls was detected 7 DAT. Plants treated with the two different rates had high glyphosate concentrations at 7 and 14 DAT for both rates. The concentration decreased to 1.47 mg kg−1 at 28 DAT for the low rate and remained low (0.68 mg kg−1 ). The high rate (432 g a.e. ha−1 ) resulted in high glyphosate concentrations up to 28 DAT (7.82 mg kg−1 ) decreasing to 4.0 mg kg−1 , 56 DAT. In general, a higher shikimic acid level was detected when a higher glyphosate level was present (Pearson Product Moment Correlation, P < 0.05). However, seven days were not sufficient to accumulate a significant amount of shikimic acid in coffee plants (Schrübbers et al., 2014). Separating the samples by sampling date, the correlation pattern (i.e. slopes in Fig. 3B) 14 DAT and 28 DAT were similar and significant (Pearson Product Moment Correlation, P < 0.05). For the first and the last sampling dates, 7 DAT and 56 DAT, shikimic acid and glyphosate levels were not significantly correlated (Pearson Product Moment Correlation, P > 0.05). Neither was the shikimic

acid accumulation significant compared to the controls at these two sampling dates for any of the doses (Schrübbers et al., 2014).

3.4. Multiple applications; greenhouse experiment To investigate the toxic effect of internal glyphosate concentrations on coffee plants, single and multiple glyphosate applications were carried out. The last (fourth) application (21st May 2012) was carried out three days before sampling, well before toxicity symptoms could develop. Therefore this last application was not considered for the assessed biological response variables; only for internal concentrations (Fig. 4A, B and S2 of the Supporting information). The plants sprayed on this last application were treated as not sprayed for the biological response. Glyphosate symptoms visible as deformed leaves, definitions are described elsewhere (Schrübbers et al., 2014), developed at all three rates; predominantly for the two higher rate (Fig. 4A). However, no trend was observed with respect to number of applications. The first two applications resulted in clear glyphosate symptoms, more severe the higher the applied rate. The third application alone did not produce toxicity symptoms. No difference was observed (ANOVA, P > 0.05) for the above soil biomass (Fig. 4B) or height (S2 of the Supporting information) 168 DAT. A trend, however, towards a reduced above soil biomass was observed for the first application. The glyphosate levels detected demonstrated that multiple applications led to a higher accumulation of the herbicide (Fig. 4C). The two single glyphosate applications (22. November 2011 and 21. May 2012 respectively), did not result in a significant internal buildup of glyphosate compared to the controls. Thus, a threshold for a significant glyphosate accumulation was two applications for the high rate and 3 applications for the low rate. The high-

140


est concentration, 1.1 and 5.3 mg kg−1 for the low and the high rate respectively, was detected after 4 applications for both rates analyzed. The minor glyphosate amount detected in the control plant was most likely caused by leaf-to-leaf contact with sprayed plants, leading to cross-contamination. Another, less likely, possible glyphosate contamination source would be the water from flooding the greenhouse tables. 3.5. Multiple applications; field experiment Similar to the greenhouse experiment single and multiple glyphosate applications were made in the field trial. Besides glyphosate measurements, branch leaf coverage was estimated and coffee berries were collected in three subsequent years. The treatments and results for the yield of the first two years and leaf coverage are shown in Fig. 5. The yield of the third year, December 2014 is only presented in Fig. S3 of the Supporting information since no trend with respect to the glyphosate application was observed for the last fruit harvest. In general the yield was lower in the second year (December 2013) compared to the first harvest (December 2012). The lowest combined yield of the two first years was obtained when the plants were exposed to the first two applications (2.3 kg fruit plant−1 ) followed by plants exposed solely to the first application (2.5 kg plant−1 ). The highest combined yield was obtained for plants exposed only to the second and third application (3.8 kg plant−1 ). None of the treatments significantly affected yield compared to the controls (ANOVA, P > 0.05). Plant height was similar across the treatments (data not shown). The leaf coverage was estimated shortly after the last spraying in November 2012. The two estimates did not differ significantly (paired t-test, P > 0.05, n = 52) and the mean value of both measurements was used for the evaluation (Fig. 5). The highest leaf coverage was observed in the control plants, and in plants that received only the last application, with 76 and 79% coverage, respectively. The lowest leaf coverage was observed for the plants exposed the first three applications with 49%; the result for this treatment was significantly different from the controls (ANOVA, P < 0.05). As for yield, leaf coverage was reduced when plants were exposed at an early stage rather than multiple times. When separating the plants according to exposed and not directly exposed at an individual spraying event (first, second, third or fourth) a significant difference (t-test, P < 0.05) was found for the yield for the first spraying event and for the first and second spraying events for the leaf coverage. Leaf coverage and yield were correlated (Pearson Product Moment Correlation, P < 0.01); suggesting that the yield reduction is, directly or indirectly, caused by the leaf shedding. Foliage for glyphosate analysis was sampled in July 2013 (approximately 8 months after the last application). No glyphosate above the LOQ was detected. Traces (concentration > LOD, but