In modern agriculture, habitats interspersed among croplands are impacted by multiple stressors related to farming intensification. Habitats intimately associated with croplands, including wetlands, riparian strips and associated river systems, as well as woodlots and hedgerows, are known to contribute positively to biodiversity. This is particularly evident in areas of intensive agriculture such as the St. Lawrence—Great Lakes region of eastern Canada (http://www.statcan.gc.ca/). It is therefore critical to determine the impact of pesticides and other stressors on the biodiversity of native plants (and wild plant species in general) within these habitats. Wild species include native and non-native plants (species that have been introduced since the arrival of Europeans in North America) that are not used as agricultural crops but that constitute the flora of current agroecosystems.
In regulatory testing, there is a large gap between the actual phytotoxicity tests performed with crops (although mostly nonnative species, they are not considered wild species or as part of the natural flora of agroecosystems) under artificial conditions and the subsequent risk assessment conducted with the aim of protecting wild plants and wildlife habitats. Phytotoxicity tests designed to evaluate the impact of pesticides on terrestrial and aquatic emergent plants are generally conducted under controlled greenhouse conditions using crop plants grown individually in pots; effects are assessed at the juvenile stage, usually before reproduction occurs 1, 2. These tests are used for the risk assessment of wild plants growing within communities exposed to variable outdoor conditions at various phenological stages.
Several experiments designed to address some of the uncertainties identified in guidelines currently used by pesticide registrants have demonstrated clear shortcomings 3. Results indicated that plant sensitivity is both herbicide and species dependent and that no obvious pattern emerged with 18 wild and crop plant species tested with five different herbicides. For some herbicides, results based on crop sensitivity may result in the under protection of wild species whereas, for others, testing with crop species may provide overprotection.
Many wild plant species are circumpolar or have spread to different regions through human introductions. Some species that have been introduced in Canada and that are often considered weeds are in need of protection in their native ranges. For example, Epipactis helleborine is an introduced orchid in Canada that is protected in Denmark 4. Cerastium arvense, although a native species in eastern Canada, is often considered a weed in Canada, whereas it is listed as an endangered species in Germany 5. Conversely, Solidago canadensis, which is an introduced weed in Europe, is a valuable native species in North America. It is likely that future tests conducted by registrants for phytotoxicity estimates will include some of these ubiquitous species.
In recent years, many wild plant species have been successfully used in phytotoxicity studies 3, 6–12. Historically, registrants have been reluctant to use wild species (as per Organization for Economic Co-operation and Development Guidelines 2), based on allegations that a large variability may exist among different ecotypes. This has not been substantiated, although it has been shown that different crop varieties (or cultivars) elicited very different IC50 (inhibition concentration causing a 50% effect) values 3, demonstrating that, at the very least, an uncertainty factor may be required for risk assessments using crop plants.
Another source of uncertainty in phytotoxicity testing is the variability in the growing conditions often found in greenhouses as well as in the outdoors. For instance, variations in temperature and water availability modified the sensitivity of Abutilon theophrasti to glyphosate 13. It is well known that greenhouse conditions fluctuate because they are subjected to a certain extent to prevailing outdoor conditions such as temperature and sunlight. This may be unavoidable and could possibly be considered appropriate given that outdoor natural conditions also fluctuate. The magnitude of effect that variable conditions have on species sensitivity to herbicides requires further scrutiny in the context of regulatory risk assessment.
The purpose of this research was threefold: to measure the variability in phytotoxicity response to two herbicides among ecotypes (plant populations from different geographical areas) of wild species originating from different parts of the world, to measure reproducibility by assessing the temporal variability in herbicide response of plants tested at different times of the year when growing conditions vary, and to investigate the biotic (germination and growth patterns) and abiotic (temperature and light intensity) factors that may contribute to differences among ecotypes and lead to uncertainties and inconsistencies in phytotoxicity testing. Uncertainty factors for risk assessments were suggested based on the results of previous experimental work using crop cultivars 3 and those of the present study with wild ecotypes and temporal/seasonal fluctuation experiments.
MATERIALS AND METHODS
All experimental work was conducted between January 2007 and January 2009. Plants were grown in a greenhouse or in a growth chamber with 16 h daylight at temperatures varying from 16 to 43°C. The photosynthetically active radiation (PAR) ranged from 106 to 1,959 µmol m−2 s−1.
Two commercial herbicide formulations, Aatrex Liquid 480® (Syngenta Crop Protection Canada) containing 480 g/L atrazine [6-chloro-N-ethyl-N′-(1-methylethyl)-1,3,5-triazine-2,4-diamine] and Round-Up Original® or Vision® (Monsanto Canada), both formulations containing 356 g/L glyphosate [N-(phosphonomethyl) glycine], were included in the present study. Label rates selected were 1,296 g active ingredient (a.i.) ha−1 for atrazine and 2,136 g a.i. ha−1 for glyphosate. A nonionic surfactant, Agral 90® (Norac Concepts), containing nonylphenoxy polyethyoxyethanol was added to glyphosate, as recommended on the label to improve the coverage of herbicide spray mixtures. No surfactant was used with atrazine. Herbicides will henceforth be referred to using their active ingredients (atrazine and glyphosate).
All seeds were obtained from commercial seed suppliers, donated, or collected by the researchers from wild populations. Plants were grown from seed and transplanted into 10-cm-diameter by 9-cm-height plastic pots containing a 3:1 Promix BX with Mycorise® Pro (Premier Horticulture) soil:sand mixture within 14 d of germination. Tests with atrazine and glyphosate were performed separately. A one-time application of herbicide was performed when plants within a given group reached the three- to five-leaf stage. Herbicides were applied using a track spray booth (de Vries Manufacturing) equipped with a TeeJet 8002E flat-fan nozzle (Spraying Systems) delivering 6.75 ml/m2 at 206.84 kPa. Depending on the species, six to 10 herbicide doses (which followed a geometric progression) plus controls with five or six replicates per dose were used for each ecotype and for every species. One plant per pot was used in all cases, and the whole experimental work consisted of a total of over 4,500 pots over the length of the experiment. In each experiment, plants were randomly assigned a numerical identification tag to prevent bias during measurements, and the layout was set as a completely randomized design. At 28 d after herbicide exposure, all above-ground green plant material was harvested and placed in a forced-air dryer for a minimum of 72 h at approximately 70°C for dry biomass determination.
Ecotype variability experiment
Eight different herbaceous broad-leaf species from four families and with different life spans were included in the ecotype variability experiment (Table 1). These were obtained from various countries in Europe as well as different regions of North America. Nine or ten doses of a given herbicide (atrazine doses from 13 to 1,380 g a.i. ha−1 and glyphosate doses from 21 to 2,277 g a.i. ha−1, both following a geometric progression of 1.7), each with six replicates, were tested on all ecotypes. Extra plants of each species were grown so that prior to spraying five plants per ecotype per species could be harvested to examine prespray variability in above-ground biomass. All ecotypes from a given species were tested during the same period to avoid possible seasonal variability.
Table 1. Overview of the plant species and ecotypes selected for testing
To account for potential size differences between the ecotypes of a given species, the total percentage of control based on above-ground biomass (calculated as weight of each plant over the average weight of controls) was determined for each ecotype and herbicide. A two-way analysis of variance (ANOVA) comparing the total percentage of control was performed for each species to determine whether there were significant differences in herbicide sensitivity among the ecotypes tested, doses, and their interaction. In several instances, the assumptions of the two-way ANOVA could not be met, in which case the nonparametric Friedman's test was performed using population means.
An inhibition concentration (IC25), defined as the dosage that results in a 25% reduction in biomass compared with controls, was calculated for every population and herbicide using nonlinear regression methods based on the relationships between herbicide dose and biomass 14. If a suitable model could not be generated or assumptions of normality and homogeneity of variance could not be met even after transformation of the data, the linear interpolation method for sublethal toxicity, also known as the “inhibition concentration approach,” was used 15.
The calculation of the hazardous dosage (HD) was performed with the ETX program developed by Aldenberg and Slob 16. This method is based on the species sensitivity distribution for the calculation of the hazardous dose that will protect 95% of the species with 50% [median HD5 or HD5(50)] and 95% [HD5(95)] confidence levels. The median, or 50% confidence, recommended for risk assessment in some countries 17, should be favored because it is considered a better estimate of the hazardous dose, but the more conservative 95% confidence level is considered a safer value.
Seed weights were recorded for every ecotype of each species included in the ecotype variability experiment. Three batches of 50 seeds for each ecotype of each species were prepared, and the average seed weight was determined. The same seeds were not weighed in each replicate. Whether differences in seed size existed between ecotypes of the same species was determined by using a one-way ANOVA or a Student's t test. Kruskal-Wallis test statistics were used when assumptions of normality and homogeneity of variance were not met.
For the germination tests, two pretreatments involving a 0- or 1-month stratification period coupled with two different environments (a greenhouse and a growth chamber, which exhibited differences in light intensity and temperature fluctuations) were used. Every species' ecotype was subjected to each combination of stratification period and environment for a total of four experimental treatments. For each experimental treatment, three replicates of 50 seeds each were included. Seeds were sown in 90-mm Petri dishes filled with a 3:1 Promix to sand mixture, following procedures described by Baskin and Baskin 18 and White et al. 19 All seeds were surface-sown except for Centaurea cyanus and Rumex crispus, which were evenly covered with approximately 15 ml of the aforementioned soil mixture to bury the seeds adequately. Dishes were top-watered as required to ensure neither the seeds nor the soil dried out, and monitored for a 28 to 31 d test period. For replicates that underwent stratification, seeds were sown as described above and plastic covers secured with Parafilm® laboratory wax (Pechiney Plastic Packaging), were placed on each dish to prevent moisture loss during the stratification period. Stratification occurred in a dark refrigerator at 2 to 4°C, and, once complete, the Petri dishes were placed in the appropriate environments and monitored the same as the nonstratified replicates. Germination was recorded every 2 to 3 d and at the end of the test period, the average total percentage germination for each of the species in each of the experimental treatments was determined.
Upon completion of testing, the germination data were arcsin transformed, and differences in germination between experimental treatments were evaluated for each species using a three-way ANOVA with ecotype, environment, and stratification as covariates. Interactions were tested among the three variables. In addition, the data were summarized to indicate the minimum number of days required to reach both 50 and 70% germination (if applicable).
Temporal variability experiment
Seven different herbaceous species from five families, including four wild species and three crops, were used in the temporal variability experiment (Table 1). Six to nine doses were tested (atrazine doses from 32 to 1,387 g a.i. ha−1 and glyphosate doses from 53 to 2,285 g a.i. ha−1, following a geometric progression ranging from 1.3 to 1.6), on six replicates for each herbicide. In addition, a comparison between plants growing under greenhouse conditions versus growth chamber conditions was performed for one crop and one wild species. In all experiments, temperatures and PAR were carefully recorded. As in the ecotype variability experiment, the IC25 and HD values were calculated for every herbicide and species where applicable.
Statistical analyses were conducted in Systat version 11 20 or Statistica version 7.1 21 unless otherwise specified.
In the ecotype variability experiment, the dose–response curves of ecotypes for a given species were remarkably similar, with the exception of Prunella vulgaris tested with atrazine (Figs. 1 and 2). Nevertheless, statistical analysis indicated that, for several species, significant differences in herbicide sensitivity existed among the ecotypes. Herbicide sensitivity (IC25 g a.i. ha−1) among ecotypes of the same species varied from a factor of less than two in eight cases (of 16) to more than one order of magnitude in the case of atrazine-tested P. vulgaris (Table 2). Furthermore, 30% of cases (14/46) exhibited sensitivities of less then 10% label rate, which is estimated to be the level that can be found drifting into field margins and other habitats adjacent to crop fields 22. If the most sensitive ecotypes from each species had been selected when tested with the herbicide atrazine, the sensitivity would have ranged from 2.5% (Bellis perennis, ecotype western North America) to 30% (Inula helenium, eastern North America) of the label rate. This is on average three times less than if the most tolerant ecotypes had been chosen for each species, with a range of sensitivity between 4.0% (R. crispus, eastern North America) and 100% (P. vulgaris, United Kingdom; and P. vulgaris, Germany) label rate. With glyphosate, the discrepancy was less pronounced, ranging from 2.8% (B. perennis, western North America) and 39.4% (R. hirta, Germany) for the most sensitive ecotypes to 4.6% (B. perennis, Germany) and 66.2% (R. hirta, midwestern North America) for the least sensitive.
Table 2. Summary of IC25 (defined as the dosage that resulted in a 25% reduction in biomass) with atrazine and glyphosate for species used in the ecotype sensitivity experiment. Label rates selected for atrazine were 1296 g a. i. ha−1 and for glyphosate, 2136 g a. i. ha−1
The variability in biomass of the different ecotypes prior to being sprayed with atrazine was measured. Differences already existed at the three- to five-leaf stage between most ecotypes at the time of spraying (Fig. 3). Interestingly, the smaller ecotypes of some species were not the most sensitive within the species (Table 2). For example, ecotypes of B. perennis (western North America), C. cyanus (United Kingdom), I. helenium (eastern North America), and R. hirta (midwestern North America) were the largest at the time of spraying but exhibited the lowest IC25s for atrazine (Table 2). No consistency was found for the other species.
Based on the average weight of 50 seeds, significant differences in seed size did exist among ecotypes of C. cyanus, P. vulgaris, and R. hirta (Table 3). There was no obvious trend supporting the idea that seeds from a certain area (North America or Europe) were either larger or smaller than the other. Significant differences in germination between ecotypes were observed for all species (Table 4). However, despite the differences in germination, the time required to reach 50 and 70% germination for each ecotype was usually similar. For five of the eight species included in the present study, stratification significantly affected germination. The location where seeds were germinated also significantly affected germination, with five species showing a preference for a specific growing environment. A few significant interactions were found among some of the variables (not shown). In general, seeds from ecotypes in North America germinated better than those from Europe.
Table 3. The average weight of 50 seeds for each ecotype tested
One-way analysis of variance results are also shown which indicate whether significant differences in seed size existed between ecotypes of the same species. A Student's test was performed with I. helenium because only two ecotypes were tested. F statistic = variance of the group means/mean of the within-group variances.
See Table 1 for complete botanical names of species.
Results of analyses of variance are also shown which indicate whether significant differences in germination existed between ecotypes of the same species, environment (greenhouse or growth chamber) and stratification (0 or 1 month). F statistic = variance of the group means/mean of the within-group variances.
See Table 1 for complete botanical names of species.
The results of the temporal variability experiment demonstrated that plant species exhibited variable levels of herbicide sensitivity when grown in a greenhouse at different times of the year or when grown under greenhouse or growth chamber conditions (Table 5). In numerous cases, more than one order of magnitude difference occurred in the IC25 values among seasons for both crops (Solanum lycopersicon, atrazine; Lactuca sativa, atrazine and glyphosate) and noncrops (Geum canadense, glyphosate; Chrysanthemum leucanthemum and R. hirta, atrazine and glyphosate). In many cases, the 10% threshold (corresponding to the accepted drift level) was reached, depending on the time of the year. For example, G. canadense was more sensitive when tested with both atrazine and glyphosate in the summer (6.8–2.0% label rate) than when tested in the spring (16.2–21.1% label rate). In contrast, C. leucanthemum was more sensitive in the fall (0.9–5.3%) than in other seasons (21.1–59.0%).
Table 5. Summary of IC25 (defined as the dosage that resulted in a 25% reduction in biomass) with atrazine and glyphosate for species used in the temporal variability experiment in greenhouses and growth chambers. The 95% confidence interval and percent of label rate (in italics) are presented below each IC25 value
Growth Chamber Winter
See Table 1 for complete botanical names of species.
Minimum and maximum temperatures did not vary greatly during the course of the year in the greenhouse (16–19.6°C and 31.1–43°C, respectively). In spring and summer, minimal artificial light was used because there was sufficient natural sunlight and also to prevent overheating (PAR = 1,959 µmol m−2 s−1 recorded on a sunny day at noon), although in the fall and winter the artificial lights were continually on for 16-h photoperiods to supplement natural sunlight (PAR: cloudy day = 106–156 µmol m−2 s−1, sunny day = 266– 299 µmol m−2 s−1 recorded at noon). Not surprisingly, plants received more total hours of sunshine and at a greater intensity in the summer than during other seasons (Table 6). In the growth chambers, artificial light provided between 364 and 462 µmol m−2 s−1 for 16-h photoperiods, and temperatures ranged between 14.8°C and 31.5°C.
Table 6. Total hours of sunshine for each post spray period (28 d) for each species
Amount of sunlight (hours)
See Table 1 for complete botanical names of species.
The HD calculated using species sensitivity distributions with the ecotype experiment data revealed that a factor of two generally separated the least sensitive and the most sensitive ecotypes (Table 7). In the case of crop cultivars 3, the ratio of the hazardous dose between the least and the most sensitive cultivars was also low but more variable, likely as a result of the small sample size. In marked contrast, seasonal variability elicited a pronounced discrepancy in response between plants tested at different times of the year; the ratio between the highest and lowest HD5 ranged from 34 to 305 (Table 7). The large ratio between the HD with 50% and 95% confidence intervals, especially with regard to crop cultivars, demonstrates some uncertainties and may mean that more species should be tested for a more accurate prediction of the HD value 16.
Table 7. HD5s (Hazard doses protecting 95% of species) calculated for the ecotype sentitivity experiment, the temporal (seasonal) variability experiment and the crop cultivar experiment in White and Boutin 3
HD5s were calculated with 50% [HD5(50)] and 95% [HD5(95)] confidence levels; data followed a logistic distribution in all cases. For each herbicide in each experiment, the lowest or the highest IC25s (defined as the dosage that resulted in a 25% reduction in biomass) for ecotypes, seasons or cultivars were used to calculate the hazardous doses. MCPA = (4-chloro-2-methylphenoxy)acetic acid.
The objective of any risk assessment prior to pesticide registration, especially for herbicides, is to ensure that nontarget plants will not be unduly harmed by off-target movement, including overspray, runoff, drifting, or leaching outside the intended spray zones. Appropriate species selection and test conditions are recognized as contentious issues in the phytotoxicity tests used in herbicide risk assessments (3, 22–24; see also http://www.epa.gov/scipoly/SAP/meetings/2001/june/sap14.pdf). Realistically, only a few species can be used to represent the hundreds of species that have to be protected. The present study revealed that some variation in sensitivity to herbicides existed among ecotypes of different plant species and that conclusions regarding the phytotoxicity of any given herbicide may differ depending on the ecotypes chosen for inclusion in risk assessment. A previous study conducted with several crop species showed that the range of herbicide sensitivity among different cultivars could be quite extensive 3. In the present study, several ecotypes differed in their seed size, percentage germination, and germination requirements, as well as in their growth patterns; however, these differences were not found to be related to any pattern of sensitivity observed among the species tested.
These findings agree with previous research, in which inequality in herbicide susceptibility among ecotypes has been demonstrated for several weed species 25–27. However, many of the studies included species with previous exposure to herbicides and were often focused on understanding the development of herbicide resistance 28. Zhou et al.13 showed that plants of Abutilon theophrasti sprayed with sublethal doses of glyphosate were less susceptible when treated with a full rate 4 d later. Other studies have shown that populations of several plant species differed in their response to herbicides depending on whether they originated from areas historically exposed to herbicides 29, 30. Plant populations collected from cropland areas where herbicide pressure has existed for some time appeared to have developed lower polymorphism than plant populations emanating from untreated areas, as a result of more resistant genotypes or phenotypes being selected 29, 30. Thus, it may be that the differences noted among ecotypes in the present study stem from the past exposure of their populations to glyphosate or atrazine, two widely used herbicides.
Other factors that were not measured in the present study may explain some of the discrepancy in the observed herbicide sensitivity among ecotypes and, to some extent, among species. Effects observed on plants in phytotoxicity tests tend to deviate from effects observed in other organisms, such as birds and mammals, in that the dosage does not indicate the concentration in the organism but merely how much of the chemical was applied per unit ground area. Baril et al. 31 showed that plant traits such as height, leaf shape, leaf length, branching pattern, and overall morphology (e.g., grass-like, erect, climbing herbaceous or woody form) explained some of the variance in pesticide residues per unit dose detected in leaves and shoot samples. Differences in leaf surface characteristics (stomata frequency, cuticle thickness, wax type, trichome density, and trichome types) examined in many species did not seem to explain the susceptibility differences between ecotypes 25, 26, 32. Conversely, Wilkinson 33 concluded that a portion of the observed ecotype resistance to a postemergence herbicide of a woody species, Tamarix pentandra, may be due to apparent differences in epicuticular wax quantity and quality. Zhou et al. 13 noted that leaves of drought-stressed plants of A. theophrasti angled downward and produced thicker cuticles, which reduced overall glyphosate efficacy. The ecotypes used in the present study presented visible dissimilarities, including significant differences in biomass at the time of herbicide spraying, but further research is required to evaluate the importance of particular morphological characteristics.
The actual herbicide uptake in an individual plant may depend not only on species specific traits but also on timing and environmental factors. In the present study, it was shown that test conditions induced a large variability in a given species' response to herbicides. Both crops and wild plant species responded quite variably when they were tested in different seasons as well as when tested in a greenhouse or in growth chambers. This was likely due to the different prevailing conditions; however, other factors might have played a role. Spray conditions are unlikely factors, insofar as the same equipment was used throughout the experimental work, and the spray booth was duly calibrated prior to each spray event. Slight differences in doses may explain some of the variability, but this cannot be assessed in the present study because only nominal concentrations are known. However, in the past, measured concentrations were found to be close to the intended concentrations 3, 10; R. L. Dalton, MSc thesis, Carleton University, Ottawa, ON, Canada]. Zhou et al. 13 showed that stressed plants (through drought or flooding) were more tolerant to glyphosate; however, sensitivity increased with cooler temperatures. Contradictory results emerged from other studies on the effects of abiotic factors, such as temperature and light, on herbicide efficacy 34, 35. All these factors have to be taken into account in phytotoxicity testing even in greenhouses, where conditions are considered relatively homogeneous compared with natural environments.
The hazardous doses calculated showed that seasonal fluctuations are a greater source of discrepancy than variability among ecotypes of the tested species. From the results obtained in the present study, given the limited amount of sunlight in the fall and winter, it may be appropriate to propose that the test duration after spray (14–28 d, depending on guidelines) be based on environmental resources such as hours of sunlight and/or hours of temperature above a certain threshold rather than number of days.
When considering the HD5(50), which is the best estimate of the hazardous dose 16, the present study showed that more than one order of magnitude should be considered as an acceptable uncertainty factor to account for the intrinsic variability in plant sensitivity to herbicides together with the variability caused by extrinsic factors. To improve risk assessments, further research is necessary to understand the significance of this variability and to better predict potentially detrimental effects.
The strict and limited species selection and protocol of current guidelines for phytotoxicity tests (e.g., U.S. EPA 1) raises the following question: how can results be extrapolated to known variable conditions under which plant species naturally live? In a very elaborate experiment, described in her Master's thesis, R.L. Dalton showed that plants grown in microcosm and placed outdoors were much more variable in their response to both atrazine and glyphosate than when similar microcosms were grown under more uniform greenhouse conditions. Other studies that included single-species tests in greenhouses and the outdoors yielded conflicting results because of multiple confounding and unexplained environmental factors 36, 37. The obvious conclusion from the present study and others is that strict adherence to current guidelines (e.g., U.S. EPA 1 and Organization for Economic Co-operation and Development 2) is likely to generate inconsistency in risk assessment and that the protection of wild plants and habitats within agroecosystems is unresolved.
This research was made possible through funding provided by Environment Canada (Pesticide Science Fund). We are very grateful to Stephan Reuter for providing seeds from Germany.