Introduction
Horseweed [Erigeron canadensis L.; syn.: Conyza canadensis (L.) Cronq.] is an agriculturally important weed worldwide (Holm et al. Reference Holm, Doll, Holm, Pancho and Herberger1997). In the United States, it is found in annual and perennial crops, such as soybean [Glycine max (L.) Merr.], corn (Zea mays L.), cotton (Gossypium hirsutum L.), orchards, and vineyards (Moretti et al. Reference Moretti, Sosnoskie, Shrestha, Wright, Hembree, Jasieniuk and Hanson2016; Steckel and Gwathmey Reference Steckel and Gwathmey2009). Erigeron canadensis competition can reduce yield by 32% to 69% in corn (Ford et al. Reference Ford, Soltani, Robinson, Nurse, McFadden and Sikkema2014; Soltani et al. Reference Soltani, Shropshire and Sikkema2021) and 40% to 90% in soybean (Agostinetto et al. Reference Agostinetto, Silva and Vargas2018; Trezzi et al. Reference Trezzi, Balbinot, Benin, Debastiani, Patel and Miotto2013; Weaver Reference Weaver2001). Furthermore, E. canadensis is difficult to control due to numerous factors. First, E. canadensis is the most widespread glyphosate-resistant weed in the world and it is also resistant to other common sites of action (SOAs) such as acetolactate synthase (ALS) inhibitors (Heap Reference Heap2014). Second, E. canadensis has an extremely aggressive root system, and plants can grow up to 2.30-m tall (Bhowmik and Bekech Reference Bhowmik and Bekech1993; Weaver Reference Weaver2001). Third, E. canadensis is a highly prolific seed producer; a single plant can produce up to 200,000 windblown seeds per year (Bhowmik and Bekech Reference Bhowmik and Bekech1993; Weaver Reference Weaver2001). Wind-mediated seeds can travel long distances and infest a wide variety of different crops, even if they are relatively distant (Shields et al. Reference Shields, Dauer, VanGessel and Neumann2006), which helps to explain the cosmopolitan nature of this species (Heap Reference Heap2014; Holm et al. Reference Holm, Doll, Holm, Pancho and Herberger1997). Fourth, E. canadensis can emerge in either the fall or spring, making it both a summer and winter annual, depending on environmental conditions (Schramski et al. Reference Schramski, Sprague and Patterson2021). Taken together, these traits make E. canadensis one of the most troublesome weed species (Van Wychen Reference Van Wychen2024).
Glyphosate has been widely used to control E. canadensis in both the fall and spring as part of burndown herbicide programs as well as in postemergence in-season applications to glyphosate-resistant crops (i.e., Roundup Ready®). Glyphosate-resistant E. canadensis was reported in the early 2000s, shortly after the widespread adoption of glyphosate-resistant crops, and is among the first glyphosate-resistant species identified (VanGessel Reference VanGessel2001). Due to the extreme selection pressure imposed by heavy glyphosate usage, glyphosate-resistant E. canadensis became frequent and has rapidly spread across the United States, limiting the effective use of this chemistry and therefore glyphosate-resistant crop technology (Davis et al. Reference Davis, Gibson and Johnson2008; Flessner et al. Reference Flessner, McElroy, McCurdy, Toombs, Wehtje, Burmester, Price and Ducar2015; Heap Reference Heap2023; Koger et al. Reference Koger, Poston, Hayes and Montgomery2004). In addition to glyphosate resistance, resistance to ALS, photosystem I (PSI), and photosystem II (PSII) inhibitors has been identified in E. canadensis accessions in the United States (Heap Reference Heap2023; Moretti et al. Reference Moretti, Bobadilla and Hanson2021). Several of these accessions are resistant to two SOAs, termed multiple resistant. To date, multiple resistant E. canadensis accessions include resistance to: glyphosate and ALS inhibitors, glyphosate and PSI inhibitors, and triazine and ALS inhibitors (Byker et al. Reference Byker, Soltani, Robinson, Tardif, Lawton and Sikkema2013; Davis et al. Reference Davis, Kruger, Stachler, Loux and Johnson2009; Matzrafi et al. Reference Matzrafi, Lazar, Sibony and Rubin2015; Moretti et al. Reference Moretti, Hanson, Hembree and Shrestha2013). Specifically in Michigan, glyphosate-resistant E. canadensis accessions were first reported in 2007 in Mason County and have since become widespread, commonly in combination with ALS inhibitors (Hill Reference Hill2024).
New technologies to control resistant accessions of E. canadensis are currently being introduced in the form of multiple herbicide resistance traits into glyphosate-resistant crops. Specifically, in soybean, new varieties include: Roundup Ready 2 Xtend® (glyphosate and dicamba resistant, Bayer Crop Science, St. Louis, Missouri, USA), Roundup Ready 2 XtendFlex® (glyphosate, dicamba, and glufosinate resistant, Bayer Crop Science, St. Louis, Missouri, USA), Enlist E3™ (2,4-D choline, glyphosate, and glufosinate resistant, Corteva Agriscience, Indianapolis, IN, USA) and LibertyLink GT27® (glyphosate, glufosinate, and isoxaflutole resistant, BASF, Ludwigshafen, Germany). Roundup Ready 2 Xtend® was first commercially available in 2017, while Enlist E3™ was first available in 2020 (Dodson Reference Dodson2022). The introduction of these multiple herbicide resistant soybean varieties primarily enables farmers to apply auxin-mimicking herbicides (Group 4) after soybean emergence. These new soybean varieties will likely increase the use of 2,4-D or dicamba for in-season weed control in soybean. Auxinic herbicide use is not unique in Michigan cropping systems, as these herbicides are commonly used in corn and wheat (Triticum aestivum L.) (Supplementary Table S1 in Supplementary File 1). Therefore, E. canadensis accessions within the state have already been exposed to auxinic herbicides, and additional recurrent use of these herbicides will increase the selection pressure for auxin resistance in E. canadensis.
Due to the insurgence of herbicide-resistant E. canadensis in the United States, especially in a relatively short time frame (Holm et al. Reference Holm, Doll, Holm, Pancho and Herberger1997; VanGessel Reference VanGessel2001), and considering new crop-resistant technologies, we are in critical need of information pertaining to the main contributing factors that select for resistance and how growers can alter use practices to delay resistance evolution. One approach is to use epidemiology theory to proactively—not reactively—predict and manage herbicide-resistance evolution (Comont et al. Reference Comont, Hicks, Crook, Hull, Cocciantelli, Hadfield, Childs, Freckleton and Neve2019; Comont and Neve Reference Comont and Neve2021). A epidemiological understanding of the main drivers of herbicide-resistance evolution would permit us to educate growers and the agricultural industry as early as possible on how to avoid herbicide-resistance evolution and preserve the use of Group 4–resistant crops (Evans et al. Reference Evans, Tranel, Hager, Schutte, Wu, Chatham and Davis2016). Therefore, the objectives of this study are: (1) conduct dose–response assays to assess the current resistance spectrum of E. canadensis accessions to glyphosate, dicamba, and 2,4-D collected in Michigan; and (2) predict and determine the main factors in row-crop production that contribute to herbicide-resistance evolution in these accessions.
Materials and Methods
Dose–Response Assay
Greenhouse dose–response experiments were conducted from August 2021 to September 2023 at Michigan State University in East Lansing, MI, USA. The experimental design consisted of a randomized complete block design with four replications repeated twice. We utilized 20 E. canadensis accessions in this study collected from eight counties in Michigan (Table 1). Within an accession, seeds from multiple plants were hand threshed, cleaned, and pooled for the remaining analysis. Once cleaned, the accessions were planted in flats containing potting medium and placed in the greenhouse (16-h light at 26 C; 8-h dark at 18 C). When two true leaves were present, each plant was transplanted into a 12 by 12 cm pot (Shuttle Pot®, East Jordan Plastics, East Jordan, Michigan, USA) containing the same medium. When transplants reached approximately 12 cm in diameter, herbicide treatments were made. Plants were watered daily and fertilizer (NPK, 15-07-25, ICL Specialty Fertilizers, Summerville, South Carolina, USA) was applied weekly.
Table 1. Mean herbicide dose required for 50% biomass reduction (ED50) in 20 Erigeron canadensis accessions collected from eight counties in Michigan.
a ED50 calculated using dry plant biomass harvested 21 d after treatment using the drc package in R. Values in parentheses are ±SE. Field-recommended labeled rates are 1.26, 0.56, and 1.07 kg ae ha−1 for glyphosate, dicamba, and 2,4-D, respectively.
The dose–response experiment consisted of nine rates of the following herbicides: glyphosate (Roundup PowerMAX® 3, 575 g ae L−1, Bayer Crop Science, St. Louis, Missouri, USA), dicamba (XtendiMax®, 350 g ae L−1, Bayer Crop Science, St. Louis, Missouri, USA), and 2,4-D choline (Enlist One™, 455 g ae L−1, Corteva Agriscience Indianapolis, IN, USA) (Table 1). For glyphosate, the dose treatments were: 0, 0.25, 0.5, 1, 2, 4, 8, 16, and 32 times the recommended labeled rate of 1.26 kg ae ha−1. For dicamba and 2,4-D, the dose treatments were: 0, 0.016, 0.031, 0.062, 0.125, 0.25, 0.33, 0.5, and 1 times the recommended labeled rates of 0.56 kg ae ha−1 and 1.07 kg ae ha−1, respectively. Additionally, 2% v/v dry ammonium sulfate was included with glyphosate and 2,4-D treatments. Herbicide treatments were applied using a single-track sprayer (Generation 4, DeVries Manufacturing, Hollandale, MN) equipped with an 8001E TeeJet® flat-fan nozzle (TeeJet® Technologies, Wheaton, IL) calibrated to deliver 187 L ha−1 at 193 kPa of pressure. Visual injury ratings were performed at 7, 14, and 21 d after treatment (DAT). Aboveground dry biomass was obtained at 21 DAT. Plants were cut at the soil surface and dried at 66 C for 7 d before dry biomass was recorded.
Statistical Analysis
Three- and four-parameter log-logistic models as well as a three-parameter Weibull model were fit to the data to determine the pattern of biomass reduction per herbicide and accession (Equations 1–3), where the independent variable was herbicide rate, and the dependent variable was dry biomass using the drc package in R (Ritz et al. Reference Ritz, Baty, Streibig and Gerhard2015). Model fit was assessed using the drc modelFit function in R, following the methods outlined in Knezevic et al. (Reference Knezevic, Streibig and Ritz2007). Models that have P-values >0.05 were chosen for the analysis with a few exceptions in which models that yielded the smallest SE values were chosen for analysis (Table 2).
$$f\left( x \right) = {{d} \over{{1 + \left[ {{x} \over{{\left( {{\rm{ED}}50} \right)b}}} \right]}}}$$
$$f\left( x \right) = d + {{{\left( {a - d} \right)}} \over {{1 + \left[ {{x} \over{{\left( {{\rm{ED}}50} \right)b\;}}} \right]}}}$$
$$f\left( {x\ |\ c,{\rm{\lambda }},k} \right) = 1 - {e^{\left[ {{{{{\left( {x - c} \right)}^k}}} \over{\lambda }} \right]}}$$
Table 2. List of statistical models used to generate mean herbicide dose required for 50% biomass reduction (ED50 in kg ae ha−1 ± SE) in 20 Erigeron canadensis accessions collected from eight counties in Michigan. a
a Models were chosen using the modelFit function in R (Knezevic et al. Reference Knezevic, Streibig and Ritz2007). LL.3, log logistic three-parameter model; LL.3u, log logistic three-parameter model with upper limit of 1; LL2.3, log logistic type 2 three-parameter model; LL2.3u, log logistic type 2 three-parameter model with upper limit of 1; LL.4, log logistic four-parameter model; LL2.4, log logistic type 2 four-parameter model; W1.3, Weibull three-parameter model; W1.4, Weibull four-parameter model; W2.3u, Weibull type 2 three-parameter model with upper limit of 1; W2.4, Weibull type 2 four-parameter model.
For all the equations, f(x) represents the effect of the herbicide at a given dose x, a is the response level when the dose x is the highest, d is the response level when the dose x is the lowest, ED50 is the dose causing 50% biomass reduction, and b is Hill’s slope, which is how steep the dose–response curve is (Muse et al. Reference Muse, Mwalili and Ngesa2021). For the Weibull function, f (x/c, λ, k) is the probability that the ED50 is either less than or equal to a given dose x, c is the location parameter, λ is the scale parameter, and k is the shape parameter (Hallinan Reference Hallinan1993). Accessions with ED50 values greater than the glyphosate recommended field use rate were considered resistant, and those with values lower than the field use rate were considered susceptible for subsequent analysis (Supplemental Fig. S1 in Supplementary File 1). Accessions with ED50 values greater than dicamba and 2,4-D recommended field use rates were considered to have reduced sensitivity, and those with values lower than field use rate were considered susceptible for subsequent analysis.
To investigate the association among ED50 values from different herbicides, correlation analysis was performed using Pearson’s correlation coefficient (Equation 4) using the Hmisc package in R (Harrell 2023).
$$r={{{\left( {n\;\sum x\;y} \right) - \left( {\sum x} \right)*\left( {\sum y} \right)}} \over{{\sqrt {[n\;\sum \;{x^2}\; - \;{{(\sum \;x)}^2}\;\left] {\; \cdot\ } \right[(n\;\sum \;{y^2}\;\; - \left( {\sum \;y{)^2}} \right]} }}}$$
For this equation, r represents the correlation coefficient, x and y represent the individual data points for ED50 values to be correlated, and x 2 and y 2 represent the average of the respective sets of ED50 values. Pearson’s correlation coefficient (r) ranges from −1 to 1 (negative and positive correlation, respectively), where a value of: 0 indicates no; 0 < r < ±0.20, weak; ±0.20 < r < ±0.40, moderate; ±0.40 < r < ±0.80, strong; and ±0.80 < r < ±1.00, very strong correlation (Pearson Reference Pearson1895).
Dose–response data were further analyzed using logistic regression (odds ratios) to access the influence previous management history had on the occurrence of resistant accessions using the glm function in R with bionomial family and logic link function (R Development Core Team 2021). Odds ratio analysis was performed to verify the strength and direction of association between two variables following the methods of Hailpern and Visintainer (Reference Hailpern and Visintainer2003). Further, logistic regression using odds ratios is robust for this type of data, as it makes no assumptions of variable distribution and provides valid model estimates irrespective of study design (Harrell 2015). For this study, odds ratio analyses were performed using the herbicide ED50 values (glyphosate, dicamba, and 2,4-D) generated in the dose–response assays in objective 1, location the accession was collected, previous herbicide-resistance screening, and 8-yr crop rotation history in the location where the accessions were collected (Table 3). Michigan State University Plant and Pest Diagnostics previously evaluated these accessions for herbicide response (screened with one and four times the recommend field use rate) to glyphosate, 2,4-D amine, and cloransulam (Hill Reference Hill2024). Crop rotation data from 2015 to 2022 were collected from the USDA-NASS Cropland Data Layer using the CroplandCROS web application (USDA-NASS 2022). Once crop rotation information was extracted, the frequency of a particular crop grown in that rotation was calculated for each rotation and categorized as: high (greater than 50% of the years contained that crop), medium (20% to 50% of the years contained that crop), and low (less than 20% of the years contained that crop).
Table 3. Categorical variables associated with Erigeron canadensis accessions used in odds ratio analysis.
a County accession collection within Michigan.
b Phenotyped through dose–response assays in objective 1 of this study; G, glyphosate; D, dicamba; E, 2,4-D; R, resistant; RS, reduced sensitivity; S, susceptible.
c Plant and Pest Diagnostics at Michigan State University previously evaluated these accessions for herbicide response (screened with 1× and 4× the recommended field use rate) to glyphosate, 2,4-D, dicamba, and cloransulam (Hill Reference Hill2024).
d Potato (Solanum tuberosum L.).
e Crop rotation data were collected from USDA-NASS Cropland Data Layer using the CroplandCROS web application (USDA 2022). Once crop rotation information was extracted, the frequency of a particular crop grown in that rotation was calculated for each rotation and categorized as: high (greater than 50% of the years contained that crop), medium (20–50% of the years contained that crop), and low (less than 20% of the years contained that crop).
Results and Discussion
Dose Response
Glyphosate
Out of the 20 E. canadensis accessions evaluated in this study, 12 (60%) were resistant to glyphosate (i.e., survived the recommended field use rate of 1.26 kg ae ha−1) (Table 1). Susceptible accession ED50 values varied substantially, ranging between <0.32 kg ae ha−1 (the lowest dose applied) to 1.15 kg ae ha−1 (∼0.9 times the field use rate). Half of the susceptible accessions were collected from separate specific locations but within Ingham County, while the remaining accessions were distributed across Montcalm, Isabella, and Delta counties. Resistant accession ED50 values also varied widely, ranging between 1.85 kg ae ha−1 (1.5 times the field use rate) to >40.32 kg ae ha−1 (the highest rate applied). The accessions with ED50 values >40.32 kg ae ha−1 were collected from Macomb and Montcalm counties. Interestingly, out of the seven accessions collected in Montcalm County, six were resistant to glyphosate, with ED50 values ranging from 2.79 to >40.32 kg ae ha−1.
Adopting Roundup Ready® (i.e., glyphosate-resistant) soybean simplified weed management, was economical, and saved growers multiple passes through the field for weed management. The adoption rate of Roundup Ready® soybean increased dramatically since introduction in 1996, reaching nearly 90% of acreage in the United States by 2008 (Dodson Reference Dodson2022). Michigan has followed suit, with 93% of soybean and 92% of corn hectarage being Roundup Ready® (USDA-NASS 2023). Annual glyphosate usage in Michigan is approximately 8.06 and 6.43 million L in soybean and corn, respectively. In the short term, this single-SOA weed control tactic allowed for higher yields with greatly reduced cost (Duke and Powles Reference Duke and Powles2009); however, it also led to successive selection events that in turn selected for widespread resistance in many troublesome weed species, including E. canadensis (Green Reference Green2007). Perhaps unsurprisingly then, we observed in our initial dose–response results that Michigan counties that are the largest producers of corn and soybean (Table 3) also have the highest level of glyphosate resistance (Table 1).
2,4-D and Dicamba
Out of the 20 accessions screened in our study, 7 (35%) had reduced sensitivity to 2,4-D (i.e., survived the recommended field use rate of 1.07 kg ae ha−1) and 4 (20%) had reduced sensitivity to dicamba (i.e., survived the recommended field use rate of 0.56 kg ae ha−1) (Table 1). These populations were considered as having “reduced sensitivity” to 2,4-D and dicamba, as they survived the highest dose applied; however, as we did not apply a dose greater than the field use rate, it is unclear if they would be resistant in an agronomic setting. Overall, ED50 values for 2,4-D and dicamba varied widely across all 20 accessions, ranging between less than the lowest dose applied (<0.02 kg ae ha−1 and <0.01 kg ae ha−1 for 2,4-D and dicamba, respectively) to the recommended field use rate (>1.07 kg ae ha−1 and >0.56 kg ae ha−1 for 2,4-D and dicamba, respectively). Susceptible accessions for 2,4-D were found in all counties screened, except for Macomb and Cass, and in all counties for dicamba. Accessions with reduced sensitivity to 2,4-D were collected from Cass (one accession), Ingham (one accession), Isabella (one accession), Macomb (one accession), and Montcalm (three accessions) counties, while accessions with reduced sensitivity to dicamba were collected from Ingham (one accession), Isabella (one accession), and Montcalm (two accessions) counties.
To combat the rise in glyphosate resistance, agricultural companies have recently developed and released soybean varieties, Xtend/XtendFlex® and Enlist E3™, that are resistant to the auxinic herbicides dicamba and 2,4-D, respectively (Skelton et al. Reference Skelton, Simpson, Peterson and Riechers2017). We screened our accessions of E. canadensis with 2,4-D and dicamba in an attempt to understand baseline sensitivity in this species in Michigan before these technologies are widely adopted. Our results show that in counties with high frequency of glyphosate-resistant E. canadensis, there is a corresponding decrease in auxinic herbicide sensitivity. Perhaps unsurprisingly, we see this pattern the most in Michigan counties that are the largest producers of corn and soybean (USDA-NASS 2024). For instance, 45% of all accessions with reduced sensitivity to auxin herbicides were collected from Montcalm County (Table 1), which is the 22nd producer of corn and 26th in soybean out of 83 counties in Michigan (USDA-NASS 2024). Auxin-mimicking herbicides have been extensively used in corn for a long time due to its natural tolerance; therefore, it is possible usage in corn started the selection process for auxinic resistance in E. canadensis that has the potential to be exacerbated by the future increased usage of auxinic herbicides in soybean.
Correlation
Overall, 30% of collected accessions are resistant or have reduced sensitivity to two or more of the herbicides tested. Among these multiple resistant populations, five are resistant to glyphosate and 2,4-D (Cass, Macomb, and Montcalm counties), two are resistant to glyphosate and dicamba (Montcalm county), and three have reduced sensitivity to both 2,4-D and dicamba (Ingham and Montcalm counties) (Table 1). Furthermore, two of collected accessions are resistant to all three herbicides assayed (Montcalm county) while five are susceptible to all three herbicides tested (Delta, Ingham, and Montcalm counties).
When ED50 values were compared among accessions treated with the auxinic herbicides 2,4-D and dicamba, Pearson’s correlation coefficient was moderately positive, r = 0.35, indicating that increases in the ED50 for 2,4-D corresponds to increases in the ED50 value for dicamba (Figure 1). Furthermore, when ED50 values were compared among accessions treated with 2,4-D and glyphosate, Pearson’s correlation coefficient was also moderately positive, r = 0.21, indicating glyphosate and 2,4-D resistance are positively correlated as well (Figure 1). Finally, when ED50 values were compared among accessions treated with dicamba and glyphosate, Pearson’s correlation coefficient was weakly positive, r = 0.15 (Figure 1).
Figure 1. Pearson’s correlation analysis between the mean herbicide dose (ED50 in kg ae ha−1) required for 50% biomass reduction in 20 Erigeron canadensis accessions collected from eight counties in Michigan. ED50 values were calculated using dry plant biomass harvested 21 d after treatment using the drc package in R. Field-recommended labeled rates are 1.26, 0.56, and 1.07 kg ae ha−1 for glyphosate, dicamba, and 2,4-D, respectively.
The moderately positive Pearson’s correlation coefficient between dicamba and 2,4-D is interesting, as these herbicides are both Group 4 herbicides, but are from different chemical families and have been shown to have different protein binding partners, IAA16 and IAA2, respectively (LeClere et al. Reference LeClere, Wu, Westra and Sammons2018; Todd et al. Reference Todd, Figueiredo, Morran, Soni, Preston, Kubeš, Napier and Gaines2020). Even though these herbicides are from different chemical families, a non–target site resistance (NTSR) mechanism(s) could be responsible for the moderate, but positive, correlation between these herbicides. Although not reported yet in E. canadensis, these NTSR mechanisms have been implied in cross-resistance for Group 4 herbicides in wild radish (Raphanus raphanistrum L.) (Goggin et al. Reference Goggin, Cawthray and Powles2016).
The moderate and weak correlation between glyphosate and 2,4-D and glyphosate and dicamba and glyphosate is notable. Regardless of strength, it is interesting to find a correlation between glyphosate and auxin-mimicking herbicides, as these herbicides are from different SOA and chemical families. Furthermore E. canadensis is a strongly self-pollinated species and thus lacks the ability to rapidly stack multiple resistance traits in a single individual or population, which is predicted in outcrossing species such as Palmer amaranth (Amaranthus palmeri S. Watson) and waterhemp [Amaranthus tuberculatus (Moq.) Sauer] (Gaines et al. Reference Gaines, Duke, Morran, Rigon, Tranel, Küpper and Dayan2020; Powles Reference Powles2008). These positive correlations are particularly concerning from a management perspective where growers will likely turn to Xtend/XtendFlex® and Enlist E3® technologies in an attempt to control glyphosate-resistant weeds, yet these populations seem primed to develop resistance to Group 4 herbicides already via previous exposure to Group 4 herbicides used in grass crops.
Odds Ratio
Previous Herbicide Resistance
Dose–response data were further analyzed using logistic regression to access the influence previous management history had on the occurrence of resistant accessions. Odds ratio analyses were performed using the ED50 values generated from the dose–response assays performed in objective 1 (Table 1), the previous herbicide-resistance screening performed at Plant and Pest Diagnostics at Michigan State University (Hill Reference Hill2024), the county the accession was collected, and the crop rotation information from the location where the accession was collected using the USDA-NASS Cropland Data Layer CroplandCROS web application (USDA-NASS 2022) (Table 3). In total, 174 pairwise comparisons were performed for the analysis (Supplementary Table S2 in Supplementary File 2).
Among all combinations, nine were statistically meaningful (P-value < 0.20; Table 4) (Andrade Reference Andrade2019). Out of the significant pairwise comparisons, 33% involved comparisons of herbicide-resistance phenotypes identified in objective 1 to different SOAs previously screened (Hill Reference Hill2024). First, if an E. canadensis accession has reduced sensitivity to 2,4-D, the odds the accession is glyphosate resistant increases by 20% (odds ratio [OR] = 0.20, P = 0.19; Table 4). Second, if an E. canadensis accession is resistant to cloransulam, the odds the accession is glyphosate-resistant increases by 15% (OR = 0.15, P = 0.08; Table 4). Finally, if an E. canadensis accession has reduced sensitivity to 2,4-D, the odds the accession has reduced sensitivity to dicamba increases by 8% (OR = 0.08, P = 0.03; Table 4).
Table 4. Results of odds ratio analysis of significant pairwise comparisons among categorical variables
a Plant and Pest Diagnostics at Michigan State University and dose–response assays in objective 1 of this study evaluated these accessions for herbicide response to glyphosate, 2,4-D, dicamba, and cloransulam (Hill Reference Hill2024); R, resistant; RS, reduced sensitivity.
b County accession collection within Michigan.
c Crop rotation data were collected from USDA-NASS Cropland Data Layer using the CroplandCROS web application (USDA 2022). Once crop rotation information was extracted, the frequency of a particular crop grown in that rotation was calculated for each rotation and categorized as: high (H = greater than 50% of the years contained that crop), medium (M = 20–50% of the years contained that crop), and low (L = less than 20% of the years contained that crop).
Overall, our results suggest that resistance to one herbicide is associated with resistance to another in E. canadensis, a finding that is supported by previously published literature. In general, when one herbicide fails to control a weed, the primary alternative is to use a different herbicide chemical family or SOA to control those individuals, which can often lead to multiple herbicide resistance in which accessions are resistant to herbicides with different herbicide chemical families or SOAs (Beckie and Tardif Reference Beckie and Tardif2012). Specifically, the co-occurrence of ALS and glyphosate resistance in E. canadensis has been reported by Byker et al. (Reference Byker, Soltani, Robinson, Tardif, Lawton and Sikkema2013). Furthermore, cross-resistance has occurred in wild mustard (Sinapis arvensis L.) to three auxin-mimicking herbicides from different chemical families (2,4-D, dicamba, and/or quinclorac) (Heap and Morrison Reference Heap and Morrison2002). Another case reported multiple resistance in Sumatran fleabane (Erigeron. sumatrensis Retz.) to paraquat, glyphosate, and chlorimuron (Albrecht et al. Reference Albrecht, Pereira, Souza, Zobiole, Albrecht and Adegas2020). The results found in this study highlight the propensity of E. canadensis to evolve multiple resistance, especially in accessions that were already resistant to one SOA.
Location
The second most influential variable was location, for which combinations between the counties where the accessions were collected were analyzed to determine resistance. Specifically, glyphosate-resistant accessions were 2,400% (OR = 24.00, P = 0.04; Table 4) more likely to occur in Ingham County and 1,200% (OR = 12.00, P = 0.13; Table 4) more likely to occur in Isabella County when glyphosate resistance was already present in the neighboring county, Montcalm (Table 4; Figure 2). This is not surprising, as these counties have similar crop rotations—corn, soybean, and winter wheat—and thus share herbicide SOAs used in those rotations (USDA-NASS 2022). Interestingly, Isabella and Montcalm counties share a border, and because E. canadensis seed dispersal is wind mediated, the spread of resistance is potentially favored by the short distance between these counties.
Figure 2. Location and resistance phenotype of Erigeron canadensis accessions that were collected throughout Michigan. Crop rotation data were collected from USDA-NASS Cropland Data Layer using the CroplandCROS web application from an 8-yr period (2015–2022; USDA 2022): (A) soybean, (B) corn, (C) winter wheat, (D) potatoes, and (E) and pasture.
Crop Rotation Frequency
The third most impactful variable was crop rotation frequency. When corn is planted between 20% to 50% of the years evaluated (medium frequency), the odds of E. canadensis having reduced dicamba sensitivity increases by 600% (OR = 6.00, P = 0.15; Table 4). When soybean is planted in <20% of the years evaluated (low frequency), the odds of E. canadensis having reduced 2,4-D sensitivity increases by 560% (OR = 5.60, P = 0.17; Table 4). When soybean is planted in >50% of the years (high frequency), the odds of E. canadensis having reduced 2,4-D sensitivity increases by 18% (OR = 0.18, P = 0.17; Table 4). Finally, when winter wheat is planted in <20% of years, the odds of E. canadensis having reduced dicamba sensitivity increases by 15% (OR = 0.15 P = 0.17; Table 4).
Crop rotation dictates what herbicide SOAs are used within a particular crop and across the rotation. With the introduction of auxin-resistant soybean varieties, there is potential that this SOA will be used more frequently or even continuously in the crop rotation, especially for soybean–corn rotations (Figure 2A and 2B). Furthermore, using the same herbicide SOA increases the selection pressure and speed at which E. canadensis evolves resistance to those SOAs (Evans et al. Reference Evans, Tranel, Hager, Schutte, Wu, Chatham and Davis2016). Specifically, the propensity of E. canadensis to evolve dicamba resistance increased in our study when corn was present at medium frequency in the rotation. Furthermore, our finding that having soybean represented at low frequency in the crop rotation leads to an increase in 2,4-D resistance is supported by the high frequency of grass crops (i.e., corn and pasture) as the main components of these low-soybean rotations (Table 3; Figure 2). Overall, these results suggest that having a medium to high frequency of grass crops, which are naturally tolerant to auxinic herbicides, in the rotation will predispose E. canadensis to have reduced sensitivity to auxinic herbicides when they are used more frequently in the future as adoption of auxin-resistant soybean increases. This is potentially already occurring, as our data indicated that there was a marginal increase in 2,4-D resistance in rotations with high frequency of soybean (Table 4).
The epidemiological approach utilized in this study begins to elucidate the main agronomic predictors of herbicide resistance in E. canadensis in Michigan. These predictors reiterate the importance of diverse management strategies and crop rotations to prevent future resistance evolution. The introduction of Group 4–resistant soybean does not enhance cropping system diversity, as these herbicides have already been used long term in corn and wheat phases of crop rotations. Alarmingly, we are already able to find E. canadensis accessions that have reduced sensitivity to 2,4-D and dicamba; therefore, it is possible that the introduction of Enlist™ and Xtend® technologies may only offer short-term relief and control of glyphosate-resistant weeds (Evans et al. Reference Evans, Tranel, Hager, Schutte, Wu, Chatham and Davis2016; LeClere et al. Reference LeClere, Wu, Westra and Sammons2018; Todd et al. Reference Todd, Figueiredo, Morran, Soni, Preston, Kubeš, Napier and Gaines2020) if used without other nonchemical management strategies. Not coincidentally, counties in the same agroclimatic zone are prone to have similar food production systems and therefore comparable overall management strategies and resistance issues (Evans et al. Reference Evans, Tranel, Hager, Schutte, Wu, Chatham and Davis2016).
Perhaps expectedly, similarity in food production and management strategies favors the evolution and spread of herbicide resistance in E. canadensis, highlighting that lack of diversity is a major factor in resistance evolution. Furthermore, in a wind-dispersed species such as E. canadensis, once herbicide resistance is established in a certain county, the proximity of counties potentially leads to the spread of herbicide-resistant accessions. The epidemiological approach performed in this study focused on evaluating a medium-sized dataset of categorical variables that are controllable by farmers. However, other drivers may be influencing the shift toward herbicide resistance of E. canadensis in Michigan, which will require more research to explore and integrate to form robust predictions. Overall, growers have the ability to proactively manage herbicide-resistance evolution progression of E. canadensis in Michigan by adopting integrated weed management (IWM) techniques to slow successive selection events that occur in low-diversity management systems. Specifically, IWM principles of crop rotation diversity and thus herbicide SOA diversity will enable farmers to diminish the threat of herbicide-resistance evolution in E. canadensis on their properties and in their regions.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/wsc.2025.6
Acknowledgments
A special thank you to Plant and Pest Diagnostics at Michigan State University for providing accessions for this study.
Funding statement
This work was supported by the Michigan Soybean Committee and Michigan State University Project GREEEN (Generating Research and Extension to meet Economic and Environmental Needs).
Competing interests
The authors declare no conflicts of interest.
Open access
