Plant material and treatments
Seven different Brassica types (hereafter, varieties) were used in this study. These included three crop varieties of Brassica napus, two wild varieties of Brassica rapa L., and one wild variety of Brassica nigra L. and Brassica juncea L each. Two of the B. napus varieties were derived from a cv. Westar genetic background representing a single homozygous transgenic trait in glyphosate resistant Canola® (B. napus RR), and a non-transgenic segregating variety (B. napus null) [20]. The third B. napus variety used was the non-transgenic B. napus cv. Sponsor, which was included to determine if plant responses to herbicide drift can be generalized to Canola® cultivars with different genetic heritage. A transgenic glufosinate resistant variety of B. napus was not available for these studies. The remaining varieties included plants grown from seeds of two populations of B. rapa collected from weedy populations in Oregon and Northern California, a single population of B. nigra collected from a weedy population in Oregon, and a single population of B. juncea (PI649101), obtained from the USDA-GRIN national germplasm repository. The cultivated and wild species used here represent a portion of a hybridization complex between diploid (B. rapa, B. nigra) and tetraploid (B. napus, B. juncea) species [12]. B. rapa and B. juncea are sexually compatible with B. napus but represent self-incompatible and self-compatible modes of fertilization respectively. B. nigra has not been shown to be easily hybridized with B. napus[12] but shares a genome with the crop species. Additionally, B. nigra is frequently found as a weed in the production regions of the US (pers obs).
Plants were seeded in 15.24 cm (6 inches) diameter pots in standard potting media (Seedling Mix No. 1, OBC Northwest, Canby, OR) and cultivated in greenhouses at 20–30°C temperature and 16/8 hr day/night light regime. Two temporal replicate experiments were planted 2 weeks apart (June 10, 2009 and June 24, 2009) with variety groups randomized and rotated in position on separate greenhouse benches. Replicates were examined for a total of 100 days from the day of seeding encompassing the termination of flowering for the majority of plants under greenhouse conditions. Replicates were examined in the same greenhouse facility and plants were rotated in position on the greenhouse benches to assure environmental uniformity. Within each temporal replicate, 8 individually potted biological replicates of each variety were examined for each treatment except for B. nigra and B. juncea varieties, which suffered from variable germination. In replicate one, 6 biological reps per treatment/control were used for B. juncea while 4 reps per treatment and 6 reps for control were used for B. nigra. In replicate two, 7 replicates were used per treatment and 9 for control for B. juncea, while B. nigra had 8 replicates for all treatments/control. As a result, temporal replicate one had a total of 262 plants, while temporal replicate two had 277.
Four herbicide stress treatments were used. Treatments involved two brand-name herbicides, Liberty® (glufosinate-ammonium) and Roundup® (glyphosate, isopropylamine salt) applied at a simulated drift level concentration of 5% (0.05) and 10% (0.10) of the field application rate (f.a.r.) expected near Canola® agriculture: (glufosinate f.a.r. = 2.48 L/Ha; 0.05 = 0.12 L/Ha, 0.10 = 0.25 L/Ha; glyphosate f.a.r. = 2.34 L/Ha; 0.05 = 0.177 L/Ha, 0.01 = 0.234 L/Ha). Glufosinate treatments included ammonium sulfate in the spray mixture (3 lbs/acre) and glyphosate treatments included the surfactant “Preference” (0.5% v/v) following suggested rates. Treatments were applied using a track sprayer (Model RC5000-100EP, Mandel Scientific Company, Ltd. Guelph, Ontario, Canada). After herbicide applications had dried, plants were placed in the greenhouse and arranged in a randomized design to minimize spatial effects. Control plants were left unsprayed. Herbicide treatments were designed to simulate the drift of herbicides onto escaped crop and weed populations in adjacent non-crop habitats. As development times are variable between the varieties, herbicide drift treatments were applied 4 weeks after seeding. At this time, the majority of the varieties were either at the pre-bolting or bolting stage but no varieties had initiated flowering. No pollinators were released within the greenhouses, preventing unintentional cross-pollination of varieties. Non-transgenic, self-fertile varieties (B. napus and B. juncea) were not restricted in the development of seed pods (siliques).
Data collection
Aboveground biomass (BIO), the total number of flowers (FA), the number of days to bolting (BOLT), days to first flower (DTF), and duration of flowering (DUR) were recorded for each individual plant. Days to first flower was recorded for all plants when the first flower-like structure with four petals was produced. Duration of flowering was recorded as the time from first flower to the termination of flowering (last fully formed flower) under greenhouse conditions. At the conclusion of flowering, plants were watered for 7 days before harvest to allow any developing siliques to elongate. At harvest, the number of flower attempts was counted by manually counting the siliques and pedicels on each raceme except for B. nigra due to the extremely large number of flowers on each plant of this species. Total aboveground biomass was collected and weighed after being dried in a 60°C drying oven (Blue M Model POM-326E, Thermal Product Solutions, New Columbia, PA) for 5 days.
Herbicide drift exposure could alter a plants ability to produce seeds either by impacting male function, female function, or both. For self-fertile species (B. napus, B. juncea), we evaluated the impact of herbicide treatments on reproduction by measuring the proportion of successful siliques vs. unsuccessful siliques. Measurements of successful self-fertility cannot distinguish reductions in reproductive fitness that arise either due to impacts on the stamen or on the pistil. Additionally, B. rapa and B. nigra varieties in this experiment are self-incompatible so additional measures of male and female function were conducted. Herbicide effects on male function were evaluated by digital photography and image analysis of anther morphology. Anthers were collected from the stamens of all varieties in all treatments from at least three flowers per plant, and three plants per treatment. Twenty-one days after herbicide applications, anthers were sampled from freshly opened flowers and placed in a 5% sucrose solution and MTT viability stain [21]. We attempted to assess pollen viability with the viability stain, however, complications with pollen extraction from the deformed anthers obtained from glyphosate treated plants precluded quantitative measures of pollen viability. Instead, we quantified morphological deformities by measuring the anther length (L), width (W), and the W/L ratio (R) from prepared slides. Image analysis was conducted using ImageJ Software [22].
To evaluate female function, manual pollinations were performed between B. napus cv. RR as a paternal parent and B. napus cv. Null, B. napus cv. Sponsor, and B. rapa OR as maternal parents. Crosses were not performed on B. rapa CA or B. juncea due to low sample sizes of recovered flowers, nor were crosses made to B. nigra due to high incompatibility with B. napus[12]. Pistils were hand pollinated at 10 days post treatment to assess the viability of pistils on plants in the early stages of recovery from herbicide drift. At 21 days post treatment, a second evaluation of pistil function on the same plants was conducted. The second evaluation corresponded to the time at which “recovered” flowers were observed. At least 3 individual flowers were pollinated on at least three plants in each treatment. Due to limited available pistils on B. napus plants at both pollination time points, it was necessary to pool the manual pollinations for cv. Null and cv. Sponsor varieties. The percent of successful manual pollinations was used to determine the viability of pistils at both the pre-recovery (10 day) and post-recovery (21 day) time points.
Data was initially analyzed as multivariate data with MANOVA but due to a lack of correlation between response variables (data not shown), data were further analyzed with ANOVA (PROC GLM) using SAS 9.2 (SAS/STAT). The two different herbicide types were examined using contrast statements for comparisons to control. Our experimental factors included Treatment (T), Variety (V), and Rep (R); all interaction effects were tested and included TxV, TxR, RxT, and TxVxR. When interactions were significant, examination of the simple treatment effects was performed [23]. Pistil viability measurements were analyzed using a nonparametric Mann–Whitney Wilcoxon Test in R [24].