The Toxicity of Herbicides to Non-Target Aquatic Plants and Algae

Pest Management Science Pest Manag Sci 61:1152–1160 (2005) DOI: 10.1002/ps.1117

The toxicity of herbicides to non-target aquatic plants and algae: assessment of predictive factors and hazard Nina Cedergreen∗ and Jens C Streibig Højbakkegard˚ Alle´ 13, 2630 Tastrup,˚ Denmark

Abstract: Widely used herbicides sometimes inadvertently contaminate surface waters. In this study we evaluate the toxicity of herbicides to aquatic plants and algae and relate it to environmental herbicide concentrations and exposure scenarios, herbicide formulation and mode of action. This was done experimentally for ten herbicides, using the aquatic macrophyte Lemna minor L. and the green alga Pseudokirchneriella subcapitata (Korshikov) Hindak, supplemented with a database study comprising algae toxicity data for 146 herbicides. The laboratory study showed that herbicide formulations in general did not enhance herbicide efﬁcacy in the aquatic environment. The Roundup formulation of glyphosate proved to be the only exception, decreasing the EC50 of the technical product for both L. minor and P. subcapitata approximately fourfold. Comparison of the sensitivity of L. minor and P. subcapitata revealed up to 1000-fold higher sensitivity of L. minor for the herbicides categorized as weak acids (pKa < 5), emphasizing the importance of higher plants in hazard assessment. Database analyses showed that no herbicide group, categorized by site of action, was signiﬁcantly more toxic than another. Synthetic auxins were the exception as they are virtually non-toxic to unicellular algae. There was no strong correlation between toxicity to algae and Kow of the herbicides, not even within groups having the same site of action. Evaluating all data, few herbicides were toxic at concentrations below 1 µglitre−1,whichis the 99.9th percentile of the herbicide concentrations measured in the Danish surveillance programme. Joint action of several herbicides cannot however be excluded.  2005 Society of Chemical Industry

Keywords: Lemna minor; algae; herbicide formulations; database study; hazard assessment

1 INTRODUCTION the function of aquatic ecosystems. Aquatic plants aid The aquatic toxicity of insecticides has been a sub- in stabilizing the sediment both in lakes and running ject of intense investigation and compounds have waters, and their presence affects sedimentation rates, often been shown to affect aquatic fauna even at flow velocity, nutrient uptake and recirculation.8 In very low concentrations.1,2 Quantitatively, however, addition, they provide refuges for insects, crustaceans insecticides form a minor part of the pesticides used and fish, and act as substrates for surface-living micro- in agriculture in Europe, accounting for only 11% of organisms, snails and other epiphyte grazers.9 Micro- the pesticide volume used in 2002, and for the Nordic algae provide the basis for a range of food-webs in the countries, Denmark and Finland, accounting for less aquatic environment and are therefore fundamental to than 4% of the total volume used.3 Herbicides, in the functioning of aquatic ecosystems.9 particular, account for more than 35% of the pesti- In the terrestrial environment, studies of the effect cide volume used in Europe and for more than 74% of herbicides on non-target plants have demonstrated of the use in the Nordic countries, and are therefore that even doses as low as 1% of field dose can frequently found in the aquatic environment.4–6 Their affect plant growth and reproduction.10,11 Knowing effect on the aquatic environment is nonetheless not that spray drift events are estimated to lie within the as well investigated as that of insecticides in the peer- range 1–10% of field dose within 10 m of the sprayed reviewed literature. Toxicity tests on aquatic species area, and in orchards up to 35%,10 and considering are required for regulatory purposes,7 but data are not the intense agricultural practice on fertile lands close always readily available and reviews rarely made. to natural water bodies, it is inevitable that aquatic Herbicides are designed to kill unwanted plants. plants will also be exposed to spray-drift events known Consequently the most sensitive group of aquatic non- to harm terrestrial species. However, while terrestrial target organisms is expected to be aquatic plants and plants are only exposed partly and at one defined algae. Aquatic plants and algae play a pivotal role for time, the whole aquatic plant can be exposed to the

∗ Correspondence to: Nina Cedergreen, Højbakkegard˚ Alle´ 13, 2630 Tastrup,˚ Denmark E-mail: [email protected] Contract/grant sponsor: Danish Environmental Protection Agency; contract/grant number: M 7041-0312 (Received 18 January 2005; revised version received 9 June 2005; accepted 10 June 2005) Published online 30 September 2005  2005 Society of Chemical Industry. Pest Manag Sci 1526–498X/2005/$30.00 1152 Prediction of toxicity of herbicides to non-target aquatic plants and algae herbicide over a long period, especially for plants in 2 MATERIALS AND METHODS stagnant waters. This increased exposure area and 2.1 Herbicides time might make aquatic plants more sensitive than The aquatic toxicity of ten different herbicides, terrestrial plants. As the herbicides are diluted when representing seven sites of action, was tested on the entering the aquatic environment, their aquatic tox- floating macrophyte L. minor and the micro-alga P. icity could also be lower than their toxicity in the subcapitata. Both technical and formulated products terrestrial environment. were tested. Information on the herbicides, their mode When discussing the toxicity of herbicides, the focus of action, technical purity and formulations is given in is mostly on the active compound. However, herbi- Table 1. cides are formulated to increase their efficacy against target plants. The question is whether the type of 2.2 Lemna tests formulation also increase herbicide efficacy against 2.2.1 Plants aquatic plants. The aquatic toxicity of various com- Lemna minor was collected from a local pond pounds with no specific biochemical target site can to in Copenhagen, Denmark in 1999. The plants a large degree be explained by their lipid solubility, were surface sterilized according to Landolt and 12,13 16 given as Kow. Hence, lipophilic compounds added Kandeler and the clone has since then been to the formulations could potentially exacerbate the kept in Ehrlenmeyer flasks in ‘K’-medium,17 pH 5, toxicity of the active compounds, either by physical or at 24 ◦C and a continuous photon flux density biochemical interactions, as seen in terrestrial plants, of 85–120 µmol m−2 s−1 (photosynthetically active or simply by adding to the herbicide toxicity.14,15 For range; PAR). The flasks and medium are sealed with compounds with specific sites of action, such as her- cotton wool and autoclaved before the weekly transfer bicides, aquatic toxicity is not expected to depend on of plants to new media. The sensitivity of the Lemna Kow but primarily on the specific site of action of the clone towards standard test compounds is equivalent herbicide. to that of other Lemna clones.18 The aims of this study were: (1) to evaluate the toxicity of herbicides against aquatic plants and algae 2.2.2 Experimental method compared with that against terrestrial plants, (2) to The experiments were performed in six-well TC-test assess the efficacy of formulated herbicides when plates (CM Lab. Aps, Vordingborg, Denmark) and applied to the aquatic environment, (3) to deter- were initiated by transferring one L. minor frond mine whether specific groups of herbicides are more to 10 ml of medium containing a selected concen- toxic to aquatic flora than others and (4) to evaluate tration of herbicide. Eight herbicide dilutions were the implications of the results for the hazard assess- used for each herbicide. The concentrations increased ment of herbicides in the aquatic environment. This by a factor two between each step. Every dilution was done experimentally, using the aquatic macro- series consisted of three replicates per concentration phyte Lemna minor L. and the green micro-alga and six controls. The concentration range was deter- Pseudokirchneriella subcapitata (Korshikov) Hindak mined from previous experiments, and the highest (= Selenastrum capricornutum Printz), supplemented concentration set to approximately 10 times the esti- with a database study comprising a larger number of mated EC50. When placed in the well, the plants herbicides. were photographed with a digital camera alongside a

Table 1. Properties of the herbicides tested in the Lemna and alga test systems. The log Kow values are given at pH 5 and pH 7 for the weak acids. Where a value was not available, this is denoted by ‘na’. The name of the formulated product is followed by the amount of active ingredient (AI) in the product. The information is obtained from the suppliers or The e-Pesticide Manual.22 The following abbreviations are used: EPSP (5-enoylpyruvyl shikimic acid 3-phosphate), ALS (acetolactate synthase), PSII (photosystem II), PSI (photosystem I) and HPPD (p-hydroxyphenylpyruvate dioxygenase)

Log Kow at pKa Technical Formulated Herbicide Site of action pH 5/7 (acids) purity product/AI Company

MCPA Auxin agonist 0.59/0.71 3.07 93% Hormotec 750 g litre−1 AH Marks Mecoprop Auxin agonist na/0.10 3.78 89% Optica 600 g litre−1 AH Marks Glyphosate EPSP synthase inhibitor −3.2/na 2.18, 5.77 min. 95% Roundup 360 g litre−1b Monsanto − Triasulfuron ALS inhibitor 1.1/−0.59 4.64 97% Logran 200 g kg 1 Syngenta − Metsulfuron-methyl ALS inhibitor na/−1.75 3.30 98.5% Ally 200 g kg 1 Du Pont Terbuthylazine PSII inhibitor 3.21 — 99% Gardoprim 500 g kg−1 Syngenta − Bentazone PSII inhibitor 0.77/−0.46 3.30 >90% Basagran 480 g litre 1 BASF − − Diquat PSI electron diversion −4.60 — 374 g L 1a Reglone 374 g litre 1 Syngenta Mesotrione 4 HPPD inhibitor 0.11c 3.12 79% Callisto 100 g kg−1 Syngenta Pendimethalin Microtubule assembly inhibitor 5.18 — 98% Stomp 400 g litre−1 BASF a Diquat bromide. b Danish registration No. 48-1. c http://pmep.cce.cornell.edu/proﬁles/herb-growthreg/fatty-alcohol-monuron/mesotrione/mesotrione let 602.html [15 December 2004]

Pest Manag Sci 61:1152–1160 (2005) 1153 N Cedergreen, JC Streibig

1 × 1 cm white plastic square, and surface area was Table 2. Classification of herbicides in groups with similar mode of determined by pixel counts using the computer pro- action according to The Herbicide Resistance Action Committee gram Photoshop 7.0. The plants were then placed in (HRAC) classification.21 The group number is given by the authors a growth cabinet at 24 ◦C and a continuous photon and is used in Fig. 2 µ −2 −1 flux density of 85–120 mol PAR photons m s . Mode of action Group After 7 days, the plants were photographed and surface areas determined. Relative growth rates were Inhibition of cellulose synthesis 1 calculated from changes in frond areas according to: PS I inhibition 2 Inhibition of cell division 3 (ln A − ln A )/T,whereA is the area at time T and T 0 T DHP inhibition (synthesis of folic acids)48 4 A0 is the area at the start of the experiment. Inhibition of pigment synthesis 5 Inhibition of microtubule assembly 6 2.3 Alga tests Inhibition of protoporphyrinogen oxidase 7 2.3.1 Algae PS II inhibition 8 The alga culture originated from Institut for Van- Inhibition of auxin transport 9 nforskning (NIVA) in Norway where it is called Inhibition of ALS (synthesis of branched-chain 10 NIVA-CHL 1, which is a strain of the alga P. sub- amino acids) capitata. The stock culture was kept in the dark at Inhibition of lipid synthesis (AACase) 11 5 ◦C. Inhibition of lipid synthesis (not AACase) 12 Inhibition of glutamine synthesis 13 Synthetic auxins 14 2.3.2 Experimental method Inhibition of EPSP-synthase (synthesis of 15 19 The algae test is described by Arensberg et al. aromatic amino acids) and Mayer et al.20 and is coherent with the ISO standards.21 Before the start of the experiments, 100 µl where toxicity data for algae were given, including of algae from the stock culture were transferred to a data on herbicide site of action, Kow,pKa,and 20-ml glass vial with 5 ml of fresh nutrient media21 (pH minimum and maximum recommended field dose. −1 8) giving a density of approximately 10 000 cells ml . The herbicides were classified in groups with similar The vial was closed with a lid with a 2-mm hole, to site or mode of action according to The Herbicide allow gas exchange. It was then placed in a holder on Resistance Action Committee (HRAC) classification ◦ a shaking table at 22 C and a continuous irradiance (Table 2).23 µ −2 −1 of 80 PAR photons mol m s . The culture was In the cases where solubility or Kow were given at diluted every 3 days two or three times, before it different pH values, the value closest to pH 7 was was ready for experimentation. For each experiment, chosen, as most algae tests are performed around eight herbicide dilutions (pH 8, buffered with sodium that pH. In the cases where toxicity for more than hydroxide or hydrochloric acid when necessary) with one species was given, data from Selenastrum sp were two replicates each and six controls were prepared and used, otherwise from Scenedesmus sp., as most tests incubated with the growing algae culture as described giving the species were performed on Selenastrum above. After 48 h the experiment was terminated and (as were the tests in this study). The toxicity was the pH of the algae culture measured (changes of normally given as EC50, but in some cases (<5%) >1 pH unit were not accepted). the ‘no observable effect concentration’ (NOEC) was Aliquots of 400 µl of algae culture were removed the toxicity parameter provided. NOEC values were from the vials at the start of the experiment and converted to estimated EC50 values by multiplying 24 and 48 h after experiment initiation. The aliquots by 10. This is a conservative estimate for EC50,as were mixed with 1.6 ml of acetone saturated with the relationship between NOEC defined as EC − 10 1 24,25 magnesium sulfate (12 g litre ) in test tubes, which or EC25 and EC50 varies between 2- and 9- were firmly closed and stored in darkness. Chlorophyll fold for logistic concentration–response curves with fluorescence of the samples was measured 24 h after slopes varying between 1 and 2. Slopes in that range experiment termination using a Turner Fluorometer covered more than 85% of 106 substances tested on (Quantech, USA) at an excitation wavelength of algae.26 The endpoints for the algae tests were not 420 nm and an emission wavelength of 670 nm. The given, but are most likely based on either cell counts relative growth rate of the algae population was or chlorophyll fluorescence, which are the endpoints determined as the slope of the regression of the ln- recommended by the International Organisation for transformed chlorophyll fluorescence as a function of Standardization.21 time, given in days. In order to relate the aquatic toxicity of the herbicides to realistic exposure levels, the toxicity 2.4 Database study exposure ratio (TER) was calculated. This was Data were retrieved from The e-Pesticide Manual,22 done by first calculating the predicted environmental as it also contains information on the recommended concentration (PEC), which is the concentration field dose. This information makes it possible to obtained by spraying the recommended field dose estimate worst-case exposure scenarios for the aquatic on a 30-cm water column, and then relating the EC50 environment. The database contained all herbicides of the herbicide to this concentration. Hence, TER

1154 Pest Manag Sci 61:1152–1160 (2005) Prediction of toxicity of herbicides to non-target aquatic plants and algae

24 can be described as EC50/PEC. A TER value of 1 herbicides tested, except glyphosate, the formulation means that the recommended field dose sprayed on of the compounds did not affect the toxicity by more 30-cm deep water will reduce algae growth by 50%. At than a factor of approximately two when assessed TER >1, a field dose sprayed on the water will reduce at the EC50 response level. For bentazone toxicity the growth of algae by less than 50% and at TER to algae, the difference was a little higher, with the <1, growth will be reduced by more than 50%. In a formulated compound being the least toxic. For the regulatory context, TER values smaller than 100 for Roundup formulation of glyphosate, the formulated acute animal tests or less than 10 for chronic tests and compound was approximately four times more toxic plant tests will require further tests and/or restrictions to L. minor and algae than the technical compound, 7 of use. In this study we used the numerical mean when based on EC50. This difference was repeated in recommended field dose to calculate PEC. It should two or more trials. In general the variance in the data of nonetheless be noted that the TER values will vary the test system was relatively small, resulting in EC50 depending on the local use, and can be several factors values determined with a coefficient of variation of lower than the mean. the fitted parameter (asymptotic standard error/mean) of 11 (±7)% (n = 77 curves) for the Lemna test ± = 2.5 Statistics and 15 ( 10)% (n 35 curves) for the algae test. 2.5.1 Concentration–response curves The variation in EC50 between experiments for the The effect of concentration on growth was tested same herbicide ranged from an EC50 CV of 22–42% = by an analysis of variance. If a significant dose (n 8–12 curves for each of five herbicides) in the = effect was found, data were described by a logistic Lemna test and of 29–39% (n 5–10 curves for each concentration–response model: of four herbicides) in the algae test. The EC10 and EC50 values in Table 3 are derived from representative d curves for each of the herbicides. Hence, the standard U = , 1 + exp[b(log(dose) − log(EC50))] deviations of these parameters represent the within- test variation. where U denotes plant response; d denotes the upper There was a large difference between the toxicities limits of the plant response at zero concentration of the of some of the herbicides to the two organisms herbicide; EC50 denotes the concentration required to tested. This difference was particularly obvious for reduce the area specific growth rate by half and b is the synthetic auxins, which had no effect on algae 27 proportional to the slope of the curve around EC50. even at concentrations approaching those of the The validity of the concentration–response model was spray-tank. Moreover, for the herbicides characterized compared to the analysis of variance by the lack-of-fit as being weak acids, toxicity to L. minor was up F-test.28 to 1000-fold larger than toxicity to P. subcapitata (Table 3). This difference was not directly coupled 2.5.2 Database study to the difference in pH between the two test systems. The relationship between the herbicide toxicity to This was demonstrated by determining the toxicity algae and Kow was tested using regression analysis, of metsulfuron-methyl to L. minor grown in algae test − while the relationship between toxicity to algae and medium (pH 8), which was 1.68 (±0.39) µg litre 1 and − the mode of action of the herbicide was tested by 1.21 (±0.12) µg litre 1 in two consecutive trials, and analysis of variance between the HRAC-classified therefore not substantially different from the toxicity groups, followed by a Tukey test. All analyses were measured at pH 5 (Table 2). performed using SAS. 3.2 Database study There were 146 herbicides for which the e-Pesticide 3RESULTS Manual provided toxicity data on algae. Of these, 55 3.1 Lemna and algae tests were tested on Selenastrum sp., 16 herbicides were The results of the Lemna and algae tests are given tested on Scenedesmus sp., six were tested on Chlorella in Table 3. None of the herbicide concentrations sp., two on Anabaena sp and one on Navicula sp measured in Danish surface waters exceed the EC10 For the remaining 66 herbicides, algae species were values, which is the effect level used as a substitute for not given. TER values could be calculated for 119 the traditional NOEC value in Denmark.21 Relating herbicides. the EC10 values to a worst case PEC, however, shows The frequency distributions of EC50 and TER both that an accidental spray on the water surface would appear log-normal (Fig. 1). The EC50 values range cause harm to the aquatic flora for at least seven from 0.0035 µg litre−1 to 3.92 g litre−1 with a median of out of the ten herbicides tested, if the sensitivity 160 µg litre−1, while the TER values range from 0.0002 of L. minor and P. subcapitata is representative of to 10 000 with a median of 2.18. Approximately the flora as a whole. A spray-drift even of 10% of 20% of the herbicides had EC50 values lower than field dose, which is a more realistic event,29 would 10 µg litre−1, which is around the maximal measured 4 give concentrations that exceed EC10 for triasulfuron, concentrations in Denmark. Of these, only aclonifen −1 terbuthylazine, diquat and pendimethalin. For all with an EC50 of 6.9 µg litre is registered for use in

Pest Manag Sci 61:1152–1160 (2005) 1155 N Cedergreen, JC Streibig

Denmark (Table 4). More than 50% of the herbicides had a TER >1, meaning that a full ﬁeld dose on 30 cm of water would reduce algal growth by more than 50%. An exposure scenario of 10% spray drift reduced the number of herbicides exceeding the algal EC50 to 32 herbicides (Table 4). Of these, terbuthylazine (TER:

umn. All concentrations 0.03), diquat bromide (TER: 0.09) and prosulfocarb

Measured in (TER:0.09)areregisteredinDenmark. There was a signiﬁcant but very small correlation between the aquatic toxicity of the herbicides, ) 0.14 550

− measured as both log algal EC50 and log TER, and 2 2 log Kow (r = 0.04, P = 0.02, n = 135; r = 0.07, 50 P = 0.003, n = 114). However, although signiﬁcant, EC . Also given are the maximum concentrations of regression coefﬁcients ranging from 4 to 7% suggest that the independent variables poorly predicted the sprayed on a 30-cm water col

21 variability in the EC50. Dividing the herbicides into 15 groups depending on mode of action (Table 2), did not improve the correlation between K and 10 ow 100 000 11 400 100 000 7 500

EC toxicity, when assessed within each group. Comparing > > the toxicity among groups showed that glyphosate, the only representative of EPSP-synthase inhibitors, and the synthetic auxins were particularly non-toxic to algae (Fig. 2). The only signiﬁcant difference between

50 groups of similar mode of action was between synthetic

EC auxins (Group 14) and inhibitors of cell division

13 600 (2293) 888 (920) 42 400 (13 400)(Group 10 3), microtubule 320 assembly inhibitors (Group 6), PS II inhibitors (Group 8) and ALS-inhibitors ed from a mean recommended ﬁeld dose for a range of technical and formulated herbicides (Group 10) (ANOVA and Tukey test). There was no signiﬁcant difference between groups of similar mode

10 of action when based on TER. 100 000 100 000 > > EC

4 DISCUSSION 4.1 The aquatic toxicity of herbicides The laboratory results showed that no herbicide 50 concentration measured in Danish surface waters Pseudokirchneriella subcapitata EC ncentration (PEC). PEC is calculat would be liable to affect either algae or Lemna minor growth by more than 10%, if they behave in nature

dard error of the parameter value in parentheses as they do in the laboratory. Assuming worst-case

.54) 1.57 (0.97) 292 (91)scenarios 677 (74) of a 106 (54) 100% spray-drift 597 (114) 0.014event (PEC, 1 Table 3), 10 more than half of the tested herbicides could cause a EC and the micro-alga >10% effect on the measured endpoints, and for the more realistic 10% spray-drift event four herbicides are still able to cause an adverse effect. For these very Lemna Alga potent herbicides, a 10- to 20-m spray-free buffer zone Lemna minor 50 is mandatory in Denmark.24 and the predicted environmental co EC 5

, Herbicides, however, rarely occur alone. In the 4 spraying season, combinations of spray drift, run-off and leaching from different ﬁelds could end up in Technical Formulated Technicalthe aquatic environment. Formulated Consequently, Denmark (reference) PEC for surface

10 water samples where pesticides have been measured, more than seven and up to 18 different herbicides were detected in more than 50% of the positive for the ﬂoating macrophyte 4 10 , and the EC values are given with the asymptotic stan samples in Denmark and similar results have been 1 − obtained by the National Water Quality Assessment 30 and EC

g litre Program (NAWQA) in the USA. Knowing that µ 50 several seemingly non-toxic concentrations can add EC up and exert toxic effects when enough compounds are present,26,31,32 it therefore cannot be excluded that MechlorpropGlyphosateTriasulfuronMetsulfuron-methyl 1490Terbuthylazine (422) 0.37 3780 (0.41)Bentazone (5360)Diquat 0.11 46 (0.04) 900 (31 918) 5480Mesotrione (632) 0.79Pendimethalin (0.44) 3560 (649) 35 (32) 966 1140 (387) 0.45 0.24 (2203) (0 (0.04) 11 200 10 (791) 200 (1280) 0.13 (0.08) 2560 92 (1453) 500 (18 6.8 400) (3.8) 105 90 (43) (21) — 0.15 (0.04) 270 000 (22 800) — 42 13 (40) 600 (4412) 634 (65) 20 (6) 64 700 (8740) 35 — (7) 115 (51) — 4.7 41 (4.8) (32) 15 405 4.6 (33) (9.2) 280 (87) 40 (17) 25 1770 (6) (1108) 134 (21) 15 — 420 (14) 977 (469) 55 (4) 6786 534 (1152) (35) — 1980 — (440) 12 (5) — 7810 (713) — 59 (10) 21 (5) 1 — 1.02 38 (12) 368 ( 42 380 80 (16) — 130 Table 3. HerbicideMCPA EC 248 (111) 4240 (700) 350 (159) 5790 (900) the herbicides found in Danish surface waters are given as herbicides do have an impact on the growth of aquatic

1156 Pest Manag Sci 61:1152–1160 (2005) Prediction of toxicity of herbicides to non-target aquatic plants and algae

35 25

30 20 25 15 20

15 10 10 5 Number of herbicides 5 0 0 1 1 10 10 0.1 100 0.1 100 0.01 1000 0.01 1000 0.001 10000 0.001 10000 100000 0.0001 1000000 µ -1 (a)EC50 Algae ( g l ) (b) TER

Figure 1. Frequency distribution of (a) the EC50 values of herbicides towards algae (n = 141) and (b) the toxicity exposure ratio (TER) (n = 119). A TER value of 1 denotes that an average ﬁeld dose of a herbicide sprayed on a 30-cm water column will yield a herbicide concentration equal to the EC50 of algae.

Table 4. Herbicides with a toxicity exposure ratio (TER) smaller than 0.1, with TER being calculated on the basis of alga EC50 and mean recommended dose22

−1 Name EC50 (µg litre ) Use in DK Name TER Use in DK

Cloransulam-methyl 0.0035 — Cloransulam-methyl 0.00022 — Paraquat dichloride 0.19 — Paraquat dichloride 0.00081 — Cyclosulfamuron 0.44 — Dichlobenil 0.0015 — Pyraflufen-ethyl 0.65 Trial Ametryn 0.0036 — Azafenidin 0.94 — Oxadiargyl 0.0046 Trial Benzobicyclon 1 — Pretilachlor 0.0053 — Cyhalofop-butyl 1 — Propisochlor 0.0058 — Oxadiargyl 1.2 Trial Terbumeton 0.006 — Diclosulam 1.6 — Pethoxamid 0.0088 — Dichlobenil 2 Not used Propachlor 0.0089 Not used Pretilachlor 2 — Flufenacet 0.0093 — Butafenacil 2.5 — Thiobencarb 0.011 — Propisochlor 2.8 — Alachlor 0.012 — Flufenacet 3.1 — Benzobicyclon 0.012 — Ametryn 3.6 — Cyhalofop-butyl 0.016 — TI-35 4.2 — Prodiamine 0.018 — Beflubutamid 4.45 — Norflurazon 0.023 — Pethoxamid 5 — Ethalfluralin 0.024 — Fentrazamide 6 — Terbutryn 0.026 — Flurochloridone 6.4 — Flurochloridone 0.031 — Trifloxysulfuron 6.5 — Chlorotoluron 0.032 — Aclonifen 6.9 Used Terbuthylazine 0.033 Used Prodiamine 7 — Cyclosulfamuron 0.035 — Ethalfluralin 9 — Tebuthiuron 0.041 — Florasulam 9 — Prometryn 0.042 — Terbumeton 9 — Oryzalin 0.065 — Isoproturon 0.072 Not used Tebutam 0.077 — Diquat bromide 0.090 Used Propham 0.091 — Prosulfocarb 0.094 Used Dimethachlor 0.098 Trial plants and algae. The impact of exposure to multiple including low doses of pesticides, are emerging, but pesticide residues should be evaluated and quantified currently only on invertebrates.36 together with other factors such as weed cutting and eutrophication. The latter factors are known to 4.2 Lemna minor versus Pseudokirchneriella have a major impact on the abundance and species subcapitata composition of aquatic macrophytes.33– 35 Studies that For some of the herbicides, there were large differences try to quantify the relative effect of different stressors, in toxicity between L. minor and the tested alga

Pest Manag Sci 61:1152–1160 (2005) 1157 N Cedergreen, JC Streibig

10000000 in terrestrial plants is known to be acidic (pH 5.8),40 1000000 might play a role for the charge of the herbicide 100000 molecule that has to cross the plasmalemma. This )

-1 hypothesis, however, has to be tested. 10000 g l µ 1000 100 4.3 The effect of formulation algae ( The herbicide formulations designed to help solubilize 50 10 sufﬁcient active ingredient for practical commercial EC 1 use and in other ways enhance efﬁcacy in the terrestrial 0.1 environment did not have the same effect when applied 0.01 to the aquatic environment (Table 3). The reason (a) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 could be that most of the oils and surfactants used 10000 in formulations either (1) help to make the active ingredient stick to the plant surface, a property that 1000 does not apply in the aquatic environment or (2) has 100 to occur in high concentrations to facilitate uptake, 10 a property that disappears when the compound is diluted in water. The only compound where the

TER 1 formulation proved to enhance the effect was the 0.1 Roundup formulation of glyphosate. The reason for this is probably that the blank formulation is in itself 0.01 phytotoxic. The blank formulation has proven to be 0.001 more toxic than glyphosate itself in tests with fish.41 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Consequently, the Roundup formulation should be (b) Category for mode of action (Table 3) treated as a mixture of two phytotoxic compounds Figure 2. The toxicity of 15 groups of herbicides towards algae given rather than as a mixture of a phytotoxic substance as (a) EC50 (n = 135) and (b) toxicity exposure ratio (TER) (n = 120). and an inert substance enhancing the phytotoxicity The mode of action of the different groups of herbicides is given by a of the active ingredient in the terrestrial environment, number explained in Table 2. All herbicides are denoted by a small as is the case with the other herbicides tested. It filled circle while the median toxicity within each group is given by a larger grey circle. should be noted, however, that there are other less toxic formulations on the market, and that Roundup is still among the least phytotoxic compounds tested species, P. subcapitata. For the synthetic auxins, the on Lemna minor and algae. insensitivity of the algae probably stems from the fact that they are unicellular organisms whose development is not governed by plant hormones such as auxins.37 4.4 Does site of action matter? The large differences found for the weak acids are For compounds that interact with cell membranes, nonetheless more interesting. A quick data-base search toxicity is closely related to the lipophilicity of the on herbicides, where toxicities for both Lemna sp and compound.12 This correlation does not apply when it an alga species were given, confirmed the tendency. comes to compounds with a specific site of action, such Weak acids were all more toxic to Lemna sp than as pesticides, as was shown in this study. For herbicides to algae, with a median of a 122-fold increase in we expected compounds with the same site of action to toxicity, ranging from 2 to 625 000-fold (n = 17). For have comparable toxicities. Inhibitors of acetolactate the neutral compounds the mean difference in toxicity synthase (ALS), for example, were expected to be (Lemna sp EC50/alga EC50) was 0.48, ranging from very potent, as they are in the terrestrial environment, 0.02 to 8.33 (n = 10).22 Neutral molecules are more compared with, for example, cell division inhibitors easily taken up than negatively charged ions, and it or synthetic auxins which are normally used in much has been shown that the toxicity of weak acids to larger doses.22 It was therefore surprising to find such algae is related to pH.38 The obvious explanation for large differences in aquatic toxicity within groups with the difference would therefore be that the experiments the same site of action, with both EC50 for algae were carried out at pH 5 (L. minor) and pH 8 (P. and TER varying by up to five orders of magnitude subcapitata) and the difference in toxicity was therefore (Fig. 2). In comparison, the field doses used to control due to more neutral herbicide molecules being present target weeds rarely exceed a factor of 100 within at low pH than at high pH. Because the high toxicity herbicide groups of the same site of action.22 The to L. minor was maintained also at high pH, this large difference within groups was not explained by explanation does not hold. The sensitivity of L. minor differences in Kow, which largely influence uptake is representative of other aquatic macrophytes, at least and partitioning of the active compound between cell for the weak acid metsulfuron-methyl39 and bentazone organelles.42 The large variance might therefore stem (unpublished results). It is therefore hypothesized that from differences in species-specific sensitivity between the pH of the apoplast of multi-cellular plants, which the algae species tested, which has been shown to vary

1158 Pest Manag Sci 61:1152–1160 (2005) Prediction of toxicity of herbicides to non-target aquatic plants and algae more than 100-fold,43,44 together with the variability ACKNOWLEDGEMENTS induced by different experimental set-ups. The authors wish to thank the chemical companies Nonetheless, there were two herbicide groups that AH Marks, BASF, Monsanto and Syngenta for their stood out as being virtually non-toxic to algae. These kind provision of technical and formulated herbicides. were the synthetic auxins and glyphosate, the latter The authors are also grateful to Axel Jørgensen for being the only representative of enolpyruvylshikimate providing information on glyphosate formulations and phosphate (EPSP) synthase inhibitors (Table 3). The to Richard Garnett and two anonymous reviewers reason for auxins being relatively non-toxic has already for reading through an earlier version of the been discussed. The low aquatic toxicity of glyphosate, manuscript. This work was supported by the Danish both towards L. minor and algae, was believed to Environmental Protection Agency grant nr. M 7041- be due to the compound being a zwitter-ion,22 0312. which makes diffusion through lipophilic membranes 45 difficult. Another herbicide with two pKa values below neutral pH was imazethapyr, which also has a REFERENCES −1 22 low algal toxicity of >100 mg litre . 1 Nørum U and Bjerregaard P, Ferskvandsinvertebraters bevægelsesadfærd som biomarkør for pesticid-exponering og 4.5 Implications for hazard assessment -effekt. 75. Danish Environmental Protection Agency. Bekæmpelsesmiddelforskning, Copenhagen, 66 pp. (2003). Evaluating the potential hazard of herbicides to the 2 de Vlaming V, DiGiorgio C, Fing S, Deanovic LA, de la aquatic environment using the TER and a safety Paz Carpio-Obeso M, Miller JL, Miller MJ and Richard NJ, factor7 of 10 showed, not surprisingly, that a majority Irrigation runoff insecticide pollution of rivers in the Imperial (62%) of the herbicides would be classified as Valley, California (USA). Environ Pollut 132:213–229 (2004). potentially toxic to the aquatic environment. The large 3 European Crop Protection Association, ECPA Statistical Review, D/03/EJ/12560. European Crop Protection Asso- majority of the most toxic ones are, however, not ciation, Brussels (2002). registered for use in Denmark (Table 4) and probably 4 National Environmental Research Institute (NERI), Database not in a number of other EU countries. For the for pesticides. National Surveillance of Water and Atmosphere remaining herbicides, use in the vicinity of the aquatic (NOVA) (2003). environment is regulated, at least in Denmark.24 The 5 Løkke H, Bichel-udvalget, in Rapport fra underudvalget om miljø og sundhed. Danish Environmental Protection Agency, large difference in toxicity between L. minor and Copenhagen, 83 pp. (2000). algae for some herbicides suggests that a safety factor 6 Solomon KR, Baker DB, Richards RP, Dixon DR, Klaine SJ, of 10, as proposed by the European Commission, LaPoint TW, Kendall RJ, Weisskopf CP, Giddings JM, is also insufficient to protect macrophytes, if the Giesy JP, Hall LW and Williams WM, Ecological risk assess- hazard assessment is made from algae tests alone. ment of atrazine in North American surface waters. Environ Toxicol Chem 15:31–74 (1996). The importance of including aquatic macrophytes 7 European Commission, Guidance document of aquatic eco- in hazard assessment, as is also required under the toxicology in the context of the Directive 91/414/EEC, European directive 91/414/EEC,7 is therefore once Sanco/3268/2001 rev. 4 (final), pp. 1–62 (2002). Available: again emphasised.25,46 http://europa.eu.int/comm/food/plant/protection/resources/ Evaluating hazard in terms of actual concentrations publications en.htm [15 December 2004]. 8 Sand-Jensen K and Krause-Jensen D, Broad-scale comparison measured showed that only a very few herbicides of photosynthesis in terrestrial and aquatic plant communities. −1 were toxic at concentration below 1 µg litre ,which OIKOS 80:203–208 (1997). is the concentration range of 99.9% of all measured 9WetzelRG,Limnology. Saunders College Publishing, Orlando, herbicides (n > 1100) in the National Surveillance Florida, USA, 763 pp. (1983). of Water and Atmosphere Programme.4 As a 10 Boutin C, Lee HB, Peart ET, Batchelor PS and Maguire RJ, Effects of the sulfonylurea herbicide metsulfuron-methyl on consequence, compared with terrestrial non-target growth and reproduction of five wetland and terrestrial plant plants, which are known to receive, and be affected species. Environ Toxicol Chem 19:2532–2541 (2000). by, spray drift events of 1–30% field dose,10,11,29 11 Bhatti MA, Al-Khatib K, Felsot AS, Parker R and Kadir S, aquatic plants seem to be better protected by the Effects of simulated chlorsulfuron drift on fruit yield and larger distances to the sprayed fields required for quality of sweet cherries (Prunus avium L.). Environ Toxicol Chem 14:537–544 (1995). potentially toxic herbicides, by adsorption of some 12 Wright DA and Welbourn P, Factors affecting toxicity, in 29 of the drift by bank vegetation and probably also Environmental toxicology, ed. by Wright DA and Welbourn P. by dilution of the herbicides in water. The problem University Press, Cambridge, UK, pp. 218–248 (2002). for the aquatic non-target plants, in contrast to 13 Russom CL, Breton RL, Walker JD and Bradbury SP, An terrestrial non-target plants, is the larger probability for overview of the use of quantitative structure–activity relationships for ranking and prioritizing large chemical multiple and prolonged herbicide exposure. There are inventories for environmental risk assessments. Environ examples where measured concentrations of several Toxicol Chem 22:1810–1821 (2003). simultaneously occurring herbicides have approached 14 Chow PNP and Grant CA, Research and development of levels of potential adverse effects.47 Assessing the effect agro-adjuvants: a bibliographic survey, in Adjuvants for of multiple chemical exposures and their interactions agrichemicals, ed by Foy CL. CRC Press, Boca Raton, FL, USA, pp. 3–16 (1992). with other stress factors form some of the future 15 Altenburger R, Nenza M and Schuürmann¨ G, Mixture toxicity challenges to understanding the aquatic toxicity of and its modelling by quantitative structure-activity relation- herbicides. ships. Environ Toxicol Chem 22:1900–1915 (2003).

Pest Manag Sci 61:1152–1160 (2005) 1159 N Cedergreen, JC Streibig

16 Landolt E and Kandeler R, Physiological characteristics, in The 33 Baattrup-Pedersen A, Larsen SE and Riis T, Composition family of Lemnaceae—a monographic study. Stiftung Rubel,¨ and richness of macrophyte communities in small Danish Zurich,¨ pp. 54–113 (1987). streams—Influence of environmental factors and weed 17 Maeng J and Khudairi AK, Studies on the flowering mechanism cutting. Hydrobiologia 495:171–179 (2003). of Lemna I. Amino acid changes during flower induction. 34 Baattrup-Pedersen A, Larsen SE and Riis T, Long-term effects Physiol Plant 28:264–270 (1973). of stream management on plant communities in two Danish 18 ISO Ringtest. Performed in Lemnatec (in press). lowland streams. Hydrobiologia 481:33–45 (2002). 19 Arensberg P, Hemmingsen VH and Nyholm N, A miniscale 35 Riis T and Sand-Jensen K, Historical changes in species algal toxicity test. Chemosphere 30:2103–2115 (1995). composition and richness accompanying pertubation and 20 Mayer P, Cuhel R and Nyholm N, A simple in vitro fluorescence eutrophication of Danish lowland streams over 100 years. method for biomass measurements in algal growth inhibition Freshwater Biol 46:269–280 (2001). tests. Wat Res 31:2525–2561 (1997). 36 Berenzen N, Kumke T, Schulz HK and Schulz R, Macroinver- 21 International Organization for Standardization. Water qual- tebrate community structure in agricultural streams: impact ity—Fresh water algal growth inhibition test with Scenedesmus of run-off related pesticide contamination. Ecotoxicol Environ subspicatus and Selenastrum capricornutum. International Orga- Saft 60:37–46 (2005). nization for Standardization, Geneva (1989). 37 Salisbury FB and Ross CW, Plant physiology.Wadsworth 22 Tomlin CDS (ed.), The e-Pesticide Manual. Version 2.2, British Publishing Company, Belmont, CA, 540 pp. (1985). Crop Protection Council (2002). Available on CD from: 38 Fahl GM, Kreft L, Altenburger R, Faust M, Boedeker W and www.bcpcbookshop.co.uk. Grimme LH, pH-dependent sorption, bioconcentration and 23 Boutsalis P, The world of herbicides according to HRAC classifica- algal toxicity of sulfonylurea herbicides. Aquat Toxicol tion on mode of action 2000. 1. AD + AM Design and Market- 31:175–187 (1995). ing, England (2000). Available: http://www.plantprotection. 39 Cedergreen N, Streibig JC and Spliid NH, Sensitivity of aquatic org/hrac/moa posters.html [15 December 2004]. plants to the herbicide metsulfuron-methyl. Ecotoxicol Environ 24 Danish Environmental Protection Agency. Rammer for vurder- Saft 57:153–161 (2004). ing af plantebeskyttelsesmidler, 7016–0002. Danish Environ- 40 Husted S and Schjoerring JK, Apoplastic pH and ammonium mental Protection Agency, Copenhagen, 53 pp. (1999). concentration in leaves of Brassica napus L. Plant Physiol 25 Davy M, Petrie R, Smrchek J, Kuchnicki T and Francois D, 109:1453–1460 (1995). Proposal to update non-target plant toxicity testing under 41 Folmar LC, Sanders HO and Julin AM, Toxicity of the NAFTA. USEPA and PMRA, USA, 163 pp. (2001). Avail- herbicide glyphosate and several of its formulations to fish and able: http://www.epa.gov/scipoly/sap/2001/june/sap14.pdf [15 aquatic invertebrates. Arch Environ Contam Toxicol 8:269–278 December 2004]. (1979). 26 Faust M and Scholze M, Competing concepts for the prediction 42 Trapp S, Model for uptake of xenobiotics in plants, in Plant of mixture toxicity: does the difference matter for regulatory contamination, Modelling and simulation of organic chemical purposes? EVKI-CT1999-00012, pp. 2–62 (2004). processes, ed by Trapp S and Mc Farlane JG. Lewis Publishers, 27 Streibig JC, Rudemo M and Jensen JE, Dose-response curves London, pp. 108–148 (1995). and statistical models, in Herbicide bioassay,edbyStreibigJC 43 Peterson HG, Boutin C, Freemark KE and Martin PA, Toxicity and Kudsk P. CRC Press, Boca Raton, FL, pp. 29–55 of hexazinone and diquat to green algae, diatoms, cyanobac- (1993). teria and duckweed. Aquat Toxicol 39:111–134 (1997). 28 Seefeldt SS, Jensen JE and Fuerst E, Log–logistic analysis 44 Fairchild JF, Ruessler DS and Carlson AR, Comparative sensi- of herbicide dose–response relationships. Weed Technol tivity of five species of macrophytes and six species of algae 9:218–227 (1995). to atrazine, metribuzin, alachlor, and metolachlor. Environ 29 Asman W, Jørgensen A and Jensen PK, Dry deposition and spray Toxicol Chem 17:1830–1834 (1998). drift of pesticides to nearby water bodies, 66. Danish Environmen- 45 Schonherr¨ J, A mechanistic analysis of penetration of glyphosate tal Protection Agency, Copenhagen, 171 pp, (2003). Avail- salts across astomatous cuticular membranes. Pest Manag Sci able: http://www.mst.dk/homepage/default.asp?Sub=http:// 58:343–351 (2002). www.mst.dk/udgiv/publications/2003/87-7972-945-2/html/ 46 Roshon RD and Stephenson GR, Comparison of two reference kap08 eng.htm [15 December 2004]. toxicants and their effect upon the growth and development 30 Martin JD, Crawford CG and Larson SJ. Pesticides in streams. of Myriophyllum sibiricum in axenic culture, in Environmental National Water Assessment Program (NAWQA), USA, 5 pp. toxicology and risk assessment, ed by Little EE, DeLonay AJ and (2003). Greenberg BM. American Society for Testing and Materials, 31 Backhaus T, Faust M, Scholze M, Gramatica P, Vighi M and West Conshohosken, PA, pp. 200–214 (1998). Grimme LH, Joint algal toxicity of phenylurea herbicides is 47 Cedergreen N, Streibig JC, Kudsk P, Møhlenberg F and Kam- equally predictable by concentration addition and indepen- per A, Er det synergi, nar˚ 0 + 0 = 1? Vand og Jord 10:88–91 dent action. Environ Toxicol Chem 23:258–264 (2004). (2003). 32 Faust M, Altenburger R, Backhaus T, Blanck H, Boedeker W, 48 Fedtke C, Microtubules, in Biochemistry and physiology Gramatica P, Hamer V, Scholze M, Vighi M and Grimme LH, of herbicide action. Springer-Verlag, Berlin, pp. 121–138 Predicting the joint algal toxicity of multi-component s- (1982). triazine mixtures at low-effect concentrations of individual toxicants. Aquat Toxicol 56:13–32 (2001).

1160 Pest Manag Sci 61:1152–1160 (2005)