Widespread Occurrence of Glyphosate-Resistant Hairy Fleabane (Erigeron bonariensis L.) in Colombia and Weed Control Alternatives

Abstract

Glyphosate, the most applied herbicide globally, offers effective non-selective and post-emergent weed control. Evolution of herbicide-resistant weeds is commonly associated with the recurrent application of herbicides with the same mode of action. Native to South America, hairy fleabane (Erigeron bonariensis L.) is the most problematic weed in this sub-continent and has previously been confirmed glyphosate resistant. This research aimed at characterizing glyphosate-resistant populations, thus estimating the frequency of resistance, resistance levels and identifying effective herbicide alternatives to control glyphosate-resistant populations. Glyphosate resistance characterization was initially conducted on ten suspected populations collected in plantain, banana, cassava, passionfruit, papaya, and drybean crops. Two resistant populations were selected and further characterized through dose-response tests; in addition, response to alternative herbicides (paraquat, glufosinate, 2,4-D, pyraflufen-ethyl, and mesotrione) was determined. All surveyed hairy fleabane populations survived (≥80% of individuals) the recommended glyphosate rate of 1080 g ae ha⁻¹; conversely, five populations collected from non-cropping areas were effectively controlled at this same rate. Dose-response tests estimated resistance factors of 3.15- to 22.3-fold versus the most susceptible population. Interestingly, resistance profile and dose-response tests detected hormesis responses at field-recommended rates. The most effective herbicide alternatives to control glyphosate-resistant hairy fleabane were pyraflufen-ethyl and mesotrione.

1. Introduction

Chemical weed control revolutionized agriculture worldwide in the 20th century. Advancements in cost-effective agricultural production systems, assisted by agrochemicals, have, in part, sustained the demand for food in an increasing human population in recent decades [1,2]. Glyphosate has become the world’s most successful herbicide since its introduction in 1974; this molecule offers non-selective, systemic, and effective post-emergence weed control [3,4]. In Colombia, glyphosate is also the most commercialized and commonly used agrochemical [5]. Thus, growers frequently spray glyphosate prior to crop sowing as a broadcast application or directed to weeds in a postemergence spray within several annual and perennial crops. Other common uses include over-the-top applications on glyphosate-resistant crops, such as maize and soybean, directed spray on waysides, and non-crop industrial applications.

Glyphosate acts by disrupting the shikimate pathway in the chloroplast, ultimately resulting in the arrest of aromatic amino acid biosynthesis via inhibition of the enzyme 3-phosphoshikimate 1-carboxyvinyltransferase (EPSP synthase; EC 2.5.1.19) [3,6]. The shikimate pathway is responsible for processing circa 20% of the carbon fixed by plants, biosynthesizing essential precursors and metabolic compounds such as vitamins, lignins, alkaloids, and flavonoids. Thus, glyphosate inhibition of EPSP synthase impacts several fundamental processes in the plant, chiefly photosynthesis and carbohydrate metabolism [3,7].

The Asteraceae species, commonly known as hairy fleabane (Erigeron bonariensis L.) (synonym: Conyza bonariensis [L.] Cronquist), represents a cosmopolitan native from South America [8]. It was initially described in Argentina and, subsequently, has further naturalized to warm areas worldwide [9]. This species possesses a weedy, aggressive, and invasive behavior, characterized by noxiousness as well as resilience to environmental stress [10]. Competition by Erigeron species in soybean may result in >50% yield loss, and it is currently considered to be the most problematic glyphosate-resistant broadleaf weed in South America [11]. Under tropical Colombian weather conditions, hairy fleabane is well-adapted to a wide range of temperature, precipitation, and altitude conditions ranging from sea level to above 3900 m above sea level (MASL) [8,10]. Moreover, in Colombia, hairy fleabane has been reported as an important weed species in annual and perennial crops, such as banana, plantain, cassava, vegetables, coffee, rice, corn, and diverse fruits; it is also commonly found in roadsides, irrigation channels, and waste areas [12,13,14].

The world’s first glyphosate-resistant hairy fleabane was reported in grape orchards in South Africa [3,15]. In Colombia, the first report was in 2006 in coffee bean groves [12,15] and more recently, a research report characterized glyphosate resistance in a hairy fleabane population from banana plantations in Magdalena province [16].

This research aimed at estimating the frequency and level of resistance in hairy fleabane populations collected from perennial and annual crops in Colombia (i.e., banana, plantain, papaya, passionfruit, cassava, and red beans) and to evaluate alternative herbicide control options for the effective management of glyphosate-resistant hairy fleabane.

2. Materials and Methods

2.1. Plant Material

A total of 10 locations were surveyed for the presence of hairy fleabane, including annual and perennial crops, and a roadside in Colombia (

Table 1. Hairy fleabane populations tested.

ID	Crop	Department (Province)	Glyphosate Use Pattern Dose/Sprays per Year/Years	Regional Context	Coordinates Latitude N/Longitude W	Altitude MASL 1
P2	Asparagus S	Cundinamarca	No use	H/CA	4°40′28.3″/74°13′1.2″	2516
P3	Rainforest S	Magdalena a	No use	U	11°6′24.9″/75°4′13″	2019
P4	Cassava	Meta	1424 g ae ha−1/4 spray/8	CA	3° °30′28.1″/73°43′21.8″	371
P5	Railway b	Cordoba, Spain	ND	Non-crop land	37°47′4.48″/5°09′18.2″	ND
P6	Railway b	Cordoba, Spain	ND	Non-crop land	38°01′40.2″/4°04′41.1″	ND
P7	Rainforest S	Cundinamarca	No use	U/near to CA	4°17′20.7″/73°59′42″	2855
P8	Rainforest S	Cundinamarca	No use	U/near to CA	4°21′38.5″/73°59′6.14″	2778
P9	Plantain	Meta	1424 g ae ha−1/8 sprays/10	CA	4°1′26.0″/73°13′51.5″	221
P10	Plantain	Antioquia	1068 g ae ha−1/4 sprays/>10	CA	7°53′40.6″/76°37′17.5″	34
P12	Red Beans	Cundinamarca	1068 g ae ha−1/1–2 sprays/18	CA	4°13′42.6″/73°57′29.3″	2056
P13	Urban area	Antioquia	ND	Non crop/near to CA	7°51′59.6″/76°36′14.4″	51
P14	Plantain	Meta	1424 g ae ha−1/4 sprays/4	CA	3°30′23.9″/73°44′49.8″	321
P15	Papaya	Meta	1424 g ae ha−1/10 sprays/5	CA	3°11′16.2″/73°35′54.3″	342
P16	Banana	Magdalena a	1068 g ae ha−1/6 sprays/>20	CA	10°45′3.6″/74°7′10.5″	53
P17	Shrubland S	Cundinamarca	No use	Sh	4°13′38.4″/74°57′13.3″	2026
P18	Plantain	Antioquia	1068 g ae ha−1/6 sprays/>20	CA	7°51′59.6″/76°36′14.4″	56
P19	Red Beans	Cundinamarca	1068 g ae ha−1/1–2 sprays/18	CA	4°14′58.6″/73°58′41.2″	2316
P20	Passionfruit	Santander	1424 g ae ha−1/6 sprays/8	CA	7°4′35″/73°12′46.7″	1256

¹ MASL = meters above sea level; ^S Putative glyphosate-susceptible (S); ^a Reference populations characterized by Quintero-Pertuz [16]; ^b Characterized by Amaro-Blanco [17]; ND = No records available; H = Horticulture; CA = Conventional agriculture; Sh = Shrubland; U = Understory in the rainforest.

). Locations were selected based on the anecdotal reports of the weed by growers, citing putative glyphosate resistance. As a resistance reference, two previously confirmed glyphosate-resistant populations [17] collected on railway sides in Cordoba (Spain) were used; these accessions were kindly conveyed by Dr. Rafael de Prado. A third glyphosate-resistant and one glyphosate-susceptible population collected in banana plantations, previously characterized by Dr. Irma Quintero, were also used in the study [16]. Three putative glyphosate-susceptible populations were collected from the understory in rainforests, and a fourth from a farm where glyphosate had not been used for over ten years. The coordinates and altitude of locations where collections occurred were recorded using Garmin^® 12 (Garmin International, Olathe, KS, United States) (Table 1). The taxonomical identification of flowered individuals originating from the collected populations, was performed by the botany experts at the Instituto Nacional de Ciencias Naturales of the Universidad Nacional de Colombia (UNAL).

Hairy fleabane achenes were planted in plastic pots filled with peat moss, watered, and covered with plastic film. Pots were then placed in a germination chamber at 28–32 °C, a 16 h photoperiod, and fertilized bi-weekly with 0.1 g per pot of the water-soluble fertilizer Soluplant^® Inicio (Agafert, Bari, Italy). After 45 days of planting, 2 cm height seedlings were transplanted by placing one plant per pot. The substrate was peat and organic soil, in a 1:1 proportion mixture. Plants were grown under glasshouse conditions at the Universidad Nacional de Colombia (4°38′10.8″ N; 74°5′19.4″ W, altitude 2610 m). Conditions were day/night temperature of 36/28 °C, relative humidity of 47%, and photoperiod of 12 h. Plants were fertilized bi-weekly using the same fertilizer solution as used in the germination chamber. Glyphosate was sprayed at the rosette stage (BBCH 19), with diameter plants ranging from 9 to 12 cm. The application equipment was a spray chamber equipped with a flat-fan nozzle Tee-Jet^® 80-02 (Spray Systems Co., Aurora, IL, USA) at 275 KPa pressure and 200 L ha⁻¹ application volume.

2.2. Resistance Profile Test

The experimental design was a completely randomized, factorial arrangement with three replicates. The first factor was population (accession) with 18 levels; the second factor was glyphosate rate with 3 levels: zero, X, and 2X, where X was equal to the recommended glyphosate rate of 1080 g acid equivalents (ae) ha⁻¹ [18,19]. Round Up^® Activo SL (Monsanto, St. Louis, MO, USA) was the commercial product used in the study, containing 363 g ae L⁻¹ as glyphosate monopotassium salt.

Assessments were performed 21 days after treatment (DAT). The evaluations included percent visual control (PVC), percent survival (PS), and fresh biomass in grams (FW); dry weight (DW) determinations were conducted after drying at 40 °C for 48 h. Relative biomass for FW and DW were calculated as percent in comparison vs. untreated for each population. Plant survival assessment was based on visual appearance, leaf coloration, and meristem turgidity, which are indicators purporting the plant’s potential to recover, reach maturity and produce achenes. Percent visual control was based on a scale of 0 to 100%, where zero was equivalent to no-observable damage in plants, and 100% was equivalent to dry or necrotic, brown, and non-phototactically active plants.

2.3. Dose-Response Tests

Based on results from the resistant profile test, two confirmed resistant (R) populations with different resistance levels were chosen for further characterization via dose-response tests. In addition, one R population from Spain (P₅) and two confirmed glyphosate-susceptible (S) populations (P₃ and P₇) were used as reference. Glyphosate application timing was identical, as indicated in the resistance profile test. The experiment was conducted under a completely randomized design with a factorial structure possessing four biological replicates (plants). The first factor was population with five levels (three R and two S) and the second factor was glyphosate rate with seven levels: 0, 102, 276, 750, 2039, 5542, and 15,064 g ae ha⁻¹. PVC and PS were assessed at 21 DAT based on the aforementioned criteria. Dose-response experiments were repeated in time to better estimate, characterize and corroborate the unexpected hormesis observed response.

2.4. Alternative Chemical Control Options for Glyphosate-Resistant Populations

The efficacy of five herbicide alternatives of different mode of action to glyphosate, according to the Herbicide Resistance Action Committee (HRAC), were assessed over three R populations (P₁₀ from plantain; P₁₅ from papaya; P₂₀ from passionfruit). These alternative herbicides were sprayed using the rates and adjuvant systems recommended in the respective commercial product labels as described in

Table 2. Alternative herbicides tested, mode of action, and doses [15].

Herbicide	HRAC MoA	Dose (g ae-ai ha−1)	Concentration (g ae-ai L−1)	Formulation	Brand	Manufacturer (Country)
2,4-D dimethylamine salt	4	720	720	SL	Amina	Invesa (Colombia)
Glufosinate ammonium	10	225	150	SC	Finale®	BASF (Germany)
Mesotrione	27	100	480	SC	Callisto®	Syngenta (Switzerland)
Paraquat dichloride	22	600	200	SL	Gramoxone®	Syngenta (Switzerland)
Pyraflufen-ethyl	14	8	26.5	SL	Et-Herb®	Nihon Nohyaku (Japan)

. The application timing was rosette (BBCH: 19), as indicated in the Plant Material Section, and plants were sprayed with a 200 L ha⁻¹ spray volume. Populations P₁₀ and P₁₅ were previously characterized using dose-response tests, whilst the third population, P₂₀, was classified as R₂ in the resistance profile test. The experiment was comprised of a complete randomized design in a factorial arrangement with four biological replicates (plants). The first factor was population with three levels (P₁₀, P₁₅, and P₂₀) and the second factor was herbicide with six levels (paraquat, 2,4-D, mesotrione, glufosinate, and pyraflufen-ethyl). Assessment variables were DW at 28 DAT and PVC using the same aforementioned criteria at 14 and 28 DAT.

2.5. Statistical Analyses

Statistical analyses were performed using R-Studio version 1.1.463 (R-Studio, Inc., Boston, MA, USA). For the resistance profile test, data were analyzed through the Effect Model with Restricted Maximum Likelihood (REML) using the nlme package [20,21]. This tool permitted the selection of the most parsimonious model considering the log-likelihood ratio, the lowest Akaike information criterion (AIC), and the Bayesian information criterion (BIC) [22,23]. For each of the variables, a comparison of the fixed and random effects was conducted using the function in the ghlt package [24]. Further, the response to glyphosate was scored independently for each variable, segregating into susceptible (S) and resistant (R) populations using the most repeated value or mode (see criteria in

Table 3. Criteria for susceptibility and resistance level in the resistance profile test.

Category	Sub- Category	Survival a (%)	Visual Control c (%)	Relative Fresh Biomass (%)	Relative Dry Biomass (%)
Susceptible	S	<80% at 1X b	>80% at 1X	Less than 20% at 1X	Less than 50% at 1X
Resistant	R1	>80% at 1X	<80% at 1X	>20% at 1X but <20% at 2X	>50% at 1X but <50% at 2X
	R2	>80% at 2X	<80% at 2X	>20% at 2X	>50% at 2X

^a Adapted from Beres et al. [26]; ^b X = 1080 g ae ha⁻¹; ^c Adapted from Zabala et al. and Panozzo et al. [22,25].

) [22,25,26].

For the dose-response test, analyses were performed using the drc package in R [27,28,29]. The model with the best fit was selected using the mselect function, which considers the higher log-likelihood and lack-of-fit, and the lower AIC and residual variance [28,29].

Regression was performed using the four-parameter log-logistic model [29,30], which is a curve that is symmetric at the inflection point “e”, or ED₅₀, in the formula:

$$y=\mathrm{c}+\left(\mathrm{d}-\mathrm{c}\right)/\left(1+e^\left(b\times \left(\log x-\log e\right)\right)\right)$$

(1)

where “y” is the response (PVC, PS, FW, or DW); “c” and “d” are lower and upper limits, respectively; “e” is Euler’s number; “b” is the slope at the parameter “e”; “x” the herbicide rate. Parameter “e” is also called ED₅₀ or GR₅₀ corresponding to LD₅₀ in the case of mortality [27,28,29,30]. Resistance factors (RFs) were calculated by the division of the R/S ratio, considering the GR₅₀ estimate for each resistant (R) and the susceptible (S) population, respectively. Estimation of R/S ratios is instrumental in determining the magnitude of resistance of different populations to a particular herbicide [31,32].

Hormesis was evaluated by fit to the Brain–Cousens model with the formula:

$$y=\mathrm{c}+\left(\mathrm{d}-\mathrm{c}+fx\right)/\left(1+e^\left(b\times \left(\log x-\log e\right)\right)\right)$$

(2)

where “c” and “d” are the lower and upper asymptotes; “b” and “e” do not have interpretation; and “f” is the size of the hormesis effect which must be different from zero for the model to have meaning [28,29,33].

Statistical analyses for the study of alternative herbicides were performed through variance analyses and the Tukey test, utilizing the function HSD test in the agricolae package of R [22,25,34].

3. Results

3.1. Resistance Profile Test

Consistently, PVC, PS, FW, and DW data parameters had better fit to the mixed model using dose as a fixed effect and population as a random effect, as described in

Table 4. Explanatory power and random effects importance in the mixed model for percent visual control (PVC), percent survival (PS), fresh weight (FW), and dry weight (DW) in 18 hairy fleabane populations.

Variable	Fix Effect	Pop.	Pop. x Rep	Log-Likelihood	AIC a	BIC b	Lik. c Ratio	Df d	p-Value
PS	Dose	x	-	−38.80	85.61	97.91	30.90	4	<0.0001
PVC	Dose	x	-	−74. 85	1495.70	1507.99	68.43	4	<0.0001
FW	Dose	x	-	−338.85	685.71	698.02	42.71	4	<0.0001
DW	Dose	x	-	−100.73	209.45	221.75	52.82	4	<0.0001

^a AIC = Akaike information criterion; ^b BIC = Bayesian information criterion; ^c Lik = Likelihood; ^d Df = Degrees of freedom; See Supplementary Table S1 for detailed analyses.

. The interaction between populations and replicates was not significant and the mixed model had higher explanatory power than using a linear model. The dose effect was significant (p-value ≤ 0.05) for all variables assessed (HSD test) and the mixed model was effective in removing the variation caused by the population factor. Importantly, this allowed separation of the response to glyphosate, for each population, utilizing the criteria described in Table 3 and segregating groups into: glyphosate-susceptible (S) and two resistant population levels (R₁ and R₂).

3.1.1. Reference and Putative Susceptible Populations

The reference-susceptible population (P₃) collected from the understory in a rainforest was confirmed glyphosate-susceptible according to the characterization by Quintero-Pertuz et al. [16]. In addition, four putative susceptible populations were classified as S and these included two rainforest accessions (P₇, P₈), a shrubland (P₁₇), and one from a farm where glyphosate had not been sprayed for over ten years (P₂). Conversely, three R accessions (P₅, P₆, and P₁₆) were classified as glyphosate-resistant (R₂) in concordance to their respective previous reports [16,17]; the sub-categories for the reference populations are summarized in

Table 5. Hairy fleabane response to glyphosate in reference resistant (R) and susceptible (S) populations evaluated with the resistance profile test.

Expected Behavior	Crop/Landscape	Code	PS	PVC	Rel-FW a	Rel-DW b	Overall Performance (Mode) c
Resistant	Railway [17] (Spain)	P5	R2	R2	R2	R2	R2
		P6	R2	R2	R1	R2	R2
	Banana [16]	P16	R2	R2	R2	R2	R2
Susceptible	Rainforest [16]	P3	S	S	S	S	S
Putative Susceptible	Asparagus	P2	S	S	S	S	S
	Understory	P7	S	S	S	S	S
		P8	R1	S	S	S	S
	Shrubland	P17	S	S	S	S	S

^a Rel-FW = relative fresh biomass 21 DAT, ^b Rel-DW = Relative dry biomass 21 DAT; ^c mode = Most repeated value; S = Susceptible; R₁ = Low resistance level; R₂ = High resistance level. PS = Percent survival; PVC = Percent visual control

.

3.1.2. Surveyed Populations

All sampled populations were confirmed resistant to glyphosate, including nine from agricultural systems where glyphosate was routinely sprayed and one population collected on the roadside near a banana plantation; see Table 1.

Based on the four efficacy parameters evaluated, PVC, PS, FW, and DW, populations had differential responses to the commercial glyphosate rate of 1080 g ae ha⁻¹, when comparing both sub-categories of glyphosate-resistance (R₁ and R₂), as well as in comparison with the susceptible populations as described in Figure 1. Consistently, all five S populations were effectively controlled with the 1X glyphosate rate (1080 g ae ha⁻¹), resulting in an average visual control of 89.2% (SE ± 5.92) (Figure 2a). In these five S populations the 2X rate resulted in an average visual control of 96% (SE ± 2.45). Conversely, the R₁ populations (low resistance) had a PVC value of 67.5% (SE ± 5.81) at the 1X rate and 80% (SE ± 3.87) at the 2X field rate. A low PVC value of 16.82% (SE ± 6.96) was recorded at the 1X rate for the R₂ group, and only 43.03% PVC (SE ± 9.54) materialized at the 2X rate (Figure 2a). Please refer to the pictures in Figure A1 and further detail provided in the Supplementary Table S2.

When glyphosate was sprayed at a 1X rate, 60% of plants within susceptible populations survived, and only 20% survived to the 2X glyphosate rate. Conversely, >90% of plants comprised in the R₁ and R₂ categories survived the 1X or 2X glyphosate rates as plotted in Figure 2b.

The calculated relative fresh biomass (Rel-FW) and the relative dry weight (Rel-DW) suggested significant differences between 1X and 2X rates (Figure 2c,d). Furthermore, in S populations, the Rel-FW was only an average of 10.49% (±3.51) at the dose the 1X and was reduced to only 4.68% (±1.62) at 2X (Figure 2c).

The Rel-FW average for the R₁ group treated with the recommended glyphosate rate was less affected compared to the S populations, where the mean was 60.63% (±1.62). When the 2X glyphosate rate was sprayed, the FW was also higher in R₁ compared to S with an estimated value of 17.36% (±5.04).

The R₂ populations demonstrated an Rel-FW average of 109.16% (±13.99) at a 1X rate, which implied that these R₂ populations had more biomass when sprayed with the recommended glyphosate rate, thus suggesting a hormesis response versus the untreated plants. Interestingly, several populations classified as R₂ plants had greater plant height than the untreated plants and tended to flower earlier when compared with untreated plants. Hormesis occurs when a dose-response relationship has a uncharacteristic effect at low doses compared to high doses, with stimulatory responses at a subtoxic level of toxin and, thus, enhanced growth parameters [35,36]. A geographical description of resistance and susceptibility for each population and surveyed locations is described in Figure 3.

3.2. Dose-Response Test for Glyphosate to Control Hairy Fleabane

3.2.1. Models Fitted and Resistance Factors

PS and PVS data were fitted to the four-parameter log-logistic model using p-values > 0.05 in a lack-of-fit test for all five populations tested [29] (see summary in

Table 6. Models fitted and parameters in dose-response test on hairy fleabane populations.

Model Fitted a	Lack-of Fit p-Value b	Pop.	b	c	d	e c	f (p-Value) e	RF7 d	RF3
	Percent survival (PS)
LL4	0.995966	P7	1.9	0	1	326.6	-
		P3	7.8	0	1	651.3	-
		P5	0.8	0	1	45,454.90	-	139.2	69.8
		P10	0.9	0	1	6857.90	-	20.3	10.2
		P15	1.8	0	1	6638	-	21	10.5
	Visual control (PVS)
LL4	0.506233	P7	−3.5	0	100	129.5	-
		P3	−2.2	0	100	152	-
		P5	−1	0	100	1278.10	-	9.9	8.4
		P10	−9.9	0	100	1581.60	-	12.2	10.4
		P15	−6.3	0	100	4552.30	-	35.1	29.9
	Dry biomass (DW)
LL4	0.961805	P7	3.3	0.2	1.8	109.0	-
		P3	2.5	0.2	2.1	165.7	-
		P5	1	0	1.8	1935.50	-	17.8	11.7
BC5	0.081436	P7	3.2	0.2	1.8	92.1	0.0052 (0.9411)
		P10	13	0.3	0.9	763.1	0.0050 (7.51 × 10−7)	8.7
		P15	7.0	0.4	1.3	4490.40	0.0007 (5.05 × 10−8)	73.7
	Relative biomass (Rel-DW)
LL4	0.992759	P7	3.3	9.4	100	108.9	-
		P3	2.5	8.5	100	166	-
		P5	1	−0.8	100	1950.30	-	17.9	11.7
BC4	0.143583	P7	3.4	9.2	100	67.6	1.5803 (0.9364)
		P10	13	36,0	100	753	0.7549 (2.2 × 10−16)	3.15
		P15	8.7	33.7	100	4652.50	0.0641 (2.2 × 10−10)	22.3

^a LL4 = 4 parameter log-logistic model; BC4 and BC5 = Brain–Cousens models with 4 or 5 parameters; ^b p-value for the lack-of-fit test should be >0.05 for the model to describe properly the data; ^c For log-logistic models parameters e is the ED₅₀ in g ae ha⁻¹; ^d RF₇ = Resistance factor vs. P₇; RF₃ = Resistance factor vs. P₃; ^e Parameter f must be ≠ 0 and p-value < 0.05 for significative hormesis [27,28,29]. See detailed analyses in Supplementary Table S3.

and detailed analyses in Supplementary Table S3). Two were S populations collected from the understory in Colombia (P₃–P₇) whilst three were classified as R: the first was collected in a plantain grove (P₁₀), the second P₁₅ from a papaya orchard, and the third P_5, from a railway in Spain and used as R-reference.

Susceptibility to glyphosate was higher in P₇, compared to P₃, in all the variables evaluated; accordingly, calculated R/S ratios or resistance factors (RF) were higher in the P₇ population.

In R populations, the four-parameter log-logistic model for the PS parameter predicted the highest lethal dose (LD₅₀) in the railroad population from Spain (P₅; LD₅₀ = 45,454.9 g ae ha⁻¹), which was appreciably higher than populations collected in plantain (P₁₀) or papaya (P₁₅). As anticipated, the lowest attained values for PS were found in S populations; LD₅₀ estimates were only 326.6 and 651.3 g ae ha⁻¹, respectively, for P₇ and P₃.

The estimated ED₅₀ for PVC indicated that 50% visual control was attained in S populations when sprayed with 129.5~152 g ae glyphosate ha⁻¹, while, in the reference resistant population (P₅), 1278.1 g ae ha⁻¹ was required to achieve the same level of control. In the population collected from a plantain area (P₁₀), a higher rate of 1581.6 g ae glyphosate ha⁻¹ was required, and an even higher dose was required to attain 50% visual control in the population collected from a papaya crop (P₁₅), where the calculated ED₅₀ was 4490 g ae ha⁻¹.

3.2.2. Models for Biomass

Dry biomass (DW and Rel-DW) in the R population P₅, as well as the S populations (P₃ and P₇), had the best fit to the log-logistic model with four parameters described in Table 6. Considering population P₇, the ED₅₀ for dry biomass (DW) was 109 g ae ha⁻¹, whereas that estimate for P₃ was higher and quantified at 165.7 g ae ha⁻¹. The reference R population (P₅) required a rate of 1950.3 g ae ha⁻¹ to reduce dry biomass population by 50% and this resulted in a resistance factor of 17.9 (vs. P₇).

In contrast, the dry biomass evaluations in R populations P₁₀ and P₁₅ had best fit to the five parameters Brain–Cousens model because of the hormesis response curve, calculating an f parameter that was significatively different from zero (p-value < 0.05 in the t-test).

Compared to the most susceptible population P₇, the resistance factor (R/S ratio) calculated for the R population collected from a papaya orchard (P₁₅) was >10 (22.3 for Rel-DW). In comparison, the population collected from a plantain grove (P₁₀) had an intermediate resistance factor (3.15 for Rel-DW) [32].

3.2.3. Graphical Output

Figure 4 shows the graphical outputs for Rel-DW data modeled for P₃, P₅, P₇, P₁₀, and P₁₅. For additional analyzed variables, please refer to Figure A2. Considering the S populations (P₃ and P₇), curves possessed steeper slopes and faster responses to glyphosate rates compared to R accessions. The Rel-DW variable fitted to the four-parameter log-logistic model described the typical “s-shape” curves (Figure 4a; however, curves for biomass in R populations had a similar pattern but less pronounced and skewed to the right (Figure 4a,b).

Two populations demonstrating hormesis (P₁₀ and P₁₅) and fitted to the Brain–Cousens model for biomass. The modeled data reveals an increase in biomass within lower rates, followed by a peak in increased response and then a decrease in plant biomass (Figure 4b). The hormesis response found in P₁₀ vs. P₁₅ was a similar to a non-monotonic response; however, the asymptote dose in P₁₀ occurred at a lower rate compared to P₁₅ (i.e., ~750 vs. ~2039 g ae ha⁻¹). In addition, the second trial repeated in time demonstrated that the hormesis effect was consistent across trials for the population collected from a papaya area (P₁₅); see Figure 5 and Figure A3.

3.3. Alternative Herbicides Evaluation

The study used populations P₁₀ (low resistance level), P₁₅, and P₂₀ (high resistance). The factorial ANOVA analyses for Rel-DW at 28 DAT demonstrated that population, herbicide, and their interaction were highly significant (p-value < 0.001); see Supplementary Table S4. Therefore, the performance and response to each herbicide active ingredient differ depending on the target population evaluated.

Biomass (Rel-DW) assessment in the population from a plantain area (P₁₀) indicated that the best alternatives for effective weed management were 2,4-D and pyraflufen-ethyl (Figure 6a), followed by mesotrione and glufosinate, and, lastly, paraquat. Conversely, P₁₅ collected from a papaya farmland, the most effective herbicide alternatives were 2,4-D, glufosinate, mesotrione, and pyraflufen-ethyl (Figure 6b). Population P₂₀ from a passionfruit orchard had the highest dry biomass values. Importantly for P₂₀, application of 2,4-D, glufosinate, and paraquat had the lowest biomass reductions compared to pyraflufen-ethyl and mesotrione (Figure 6c).

For the PVC evaluation, ANOVA estimated significant effects for all factors (herbicide, population, and time) and concomitant interactions at the 14 and 28 DAT evaluation times. The effect on evaluation over time is explained by the speed and mode of action for each herbicide, as summarized in

Table 7. Visual control of glyphosate-resistant hairy fleabane populations with different herbicides at 14 and 28 days after treatment (DAT).

	Dose	P10				P15				P20
Herbicide a	g ai-ea ha−1	14 DAT b		28 DAT		14 DAT		28 DAT		14 DAT		28 DAT
2,4-D	720	85.0	a	88.75	a	83.75	a	85.0	a	82.5	a	25.0	b
Glufosinate	150	78.75	ab	47.5	b	81.25	a	56.25	b	56.65	b	25.0	b
Mesotrione	100	66.25	ab	40.0	b	80.0	a	92.5	a	83.75	a	86.25	a
Paraquat	600	52.5	a	20.0	b	60.0	a	50.0	b	20.0	c	30.0	b
Pyraflufen-ethyl	8	80.0	ab	82.5	a	72.5	a	77.5	ab	80.0	a	90.0	a
Pr (>F)	-	0.0176		7.27 × 10−5		0.0712		0.00113		1.77 × 10−7		1.01 × 10−8

^a 2,4-D; glufosinate; mesotrione; paraquat; pyraflufen-ethyl; ^b Letters = Tukey groups (HSD); See detailed analyses in Supplementary Table S4.

. The pyridinium herbicide paraquat was the least effective of the evaluated treatments, it had the highest biomass across R populations and the lowest visual control effect.

Pyraflufen-ethyl and mesotrione demonstrated the best overall performance for PVC across all three populations and across assessment times, whereas 2,4-D only effectively controlled the P₁₀ and P₁₅ populations. Interestingly, the auxin herbicide 2,4-D had a satisfactory control (82.5%), of the population from the passionfruit field (P₂₀) at the 14 DAT assessment; however, at 28 DAT, the control was reduced significantly and only delivered 25% visual control. Glufosinate offered acceptable control levels (>80%) at the 14 DAT in only one population (P₁₅), but the control declined over time, and plants recovered from the application at the end of the trial (28 DAT).

4. Discussion

4.1. Resistance to Glyphosate in Hairy Fleabane Is Widespread

Consistent with expected results, the resistance profile test demonstrated that the three S populations (non-exposed to glyphosate) were effectively controlled with the field-recommended glyphosate rate of 1080 g ae ha⁻¹. Furthermore, resistance in low and high levels, was present in 100% of the locations where collections were made in areas of prior glyphosate use. A high resistance level (R₂) was detected in 80% of the surveyed fields. A survey performed in Erigeron sp. populations in Brazil between 2014 and 2018 detected glyphosate-resistance in 71.2% out of 1184 samples collected [37]. Previously, in the United States, 100% of the E. canadensis from agricultural areas had high levels of resistance in Ohio, whereas, in Iowa, more than 90% were sites with varying levels of R at the plant level.

These results purport the seriousness of the glyphosate resistance problem in Colombia, suggesting that resistance has achieved similar hairy fleabane prevalence levels as those found in other countries within agricultural landscapes.

Resistance was also present in the P₁₃ population, collected in a street near a banana plantation, which also scored a high level of resistance (R₂) in the resistance profile test. Similarly, in Spain, hairy fleabane populations were R in 41.7% of populations from railways [17]. In our study, the confirmed R resistance populations P₅ and P₆ were also included. Furthermore, the estimated ED₅₀ on relative biomass (Rel-DW) in P₅, which was 1950.3 g ae ha⁻¹, was very close to that estimated in the population reported by Amaro-Blanco [17], who found that this value was 1972.4 g ae ha⁻¹.

In Iowa and Ohio, US, researchers found that 45% of the non-agricultural sampled places were R and that these accounted for high survival (>80%), even at 40 times the recommended field rate (i.e., 33,600 g ae ha⁻¹) [26]. These results confirmed that non-agricultural areas also serve as a refuge for R genes and enhance the chance of resistance spreading to other areas.

4.2. Levels of Susceptibility and Resistance to Glyphosate in Hairy Fleabane in Colombia

The dose-response test allowed estimation of the susceptibility levels found in two S populations, which resulted in intermediate and similar values when comparing ED₅₀ with the values reported by other studies in the same species [16,17,18,19,38,39,40,41,42,43,44]. In summary, these studies report ED₅₀ values based on dry biomass in S biotypes from 34.8 to 335 g ae ha⁻¹.

Our results estimated values of 165.7 and 109 g ae ha⁻¹ for ED₅₀ in P₃ and P₇, respectively (over biomass). These values are comparable with the aforementioned studies referenced in this publication. For detailed benchmark values, please refer to Supplementary Table S5.

For all assessments (survival, control, and biomass), we found that ED₅₀ values contrasted between the R and S populations. Population P₁₅ (from the passionfruit grove) had the highest ED₅₀ values considering control and biomass and were the most resistant according to tests for biomass and visual control. Population P₅ (the reference from the railway) had the highest ED₅₀ for survival rate. P₁₀ (from plantain crop) had the lower ED₅₀ values. Dry biomass ED₅₀ values are also comparable with the values reported previously; thus, for R populations, the ED₅₀ values may vary from 1129.9 to 14,261 g ae ha⁻¹ [16,17,18,19,38,39,40,41,42,43,44].

For the variate survival in the R-reference P₅ from the railway in Spain, the model estimated a very high LD₅₀ (45,454.9 g ae ha⁻¹). This was similar to one R biotype described in the United States by Moretti et al., 2016 [41], for which LD₅₀ was 40,862 g ae ha⁻¹; that particular biotype had an estimated ED₅₀ for DW of 1279 g ae ha⁻¹, which was similar to the one estimated for P₅ in our study (1935.5 g ae ha⁻¹).

These extremely different estimates (survival vs. biomass) suggest clearly that the rate to kill 50% of these R plants can be much higher than the rate to decrease 50% of the biomass. Those high field rates to control R biotypes can be unaffordable for growers using only glyphosate. In the case of both hairy fleabane populations from Colombia, the estimated LD₅₀ values were much lower: 6857.9 and 6638 g ae ha⁻¹ for P₁₀ and P₁₅, respectively. This means, on average, six times the field rate for glyphosate, which also results in a soaring extra cost.

The resistance factors (R/S ratios) found in the populations from papaya (P₁₅) and the reference from the railway (P₅) were 22.3-fold and 17.8-fold, and are considered high resistance, whereas the population from plantain RF was 3.15-fold, which is considered low resistance [32,43]. These RFs values are, in addition, similar to the results of other authors [16,17,18,19,38,39,40,41,42,43,44].

4.3. Hormesis Triggered by Glyphosate-Resistant Hairy Fleabane

Biomass data in two populations from Colombia fitted to the Brain–Cousens model for hormesis [27,28,29,33] and the response were consistent across trials. Hormetic response to glyphosate has been largely documented to occur in more than 20 species at doses from 1.9 to 730 g ae ha⁻¹ [37,45], and has been also previously documented in hairy fleabane.

In South Africa, hairy fleabane accessions classified as R showed hormesis when the trials were performed under 27 °C conditions but not at 15 °C [43]. In five out of nine E. canadensis R populations from Indiana (USA), dose-response trials revealed hormesis when assessing FW [46].

A study in Brazil found that a glyphosate dose of 90 to 360 g ae ha⁻¹ in E. sumatrensis increased plant height and had earlier flowering, while there was no effect on the number of capitula per plant after treatment [39]. Our results found a hormesis response with an increase in biomass at the recommended rate in the dose profile test (Figure 2c,d), as well as in the dose-response test (Figure 4b and Figure A2). In addition, we found earlier flowering and an increase in plant height, indicating that some Colombian biotypes may have a larger tolerance to glyphosate applications than those reported by the aforementioned studies in South Africa, Brazil, or United States.

Other cases reporting a hormesis effect using the recommended doses of the herbicide have been published in ACCase-target-site-resistant (Alopercus myosuroides Huds.) biotypes; this resulted in a 147% increase in shoot fresh weight after exposure to rates similar to those recommended for the field [47]. Such an increase in biomass at recommended doses results in a higher ability for weeds to compete with the crop and an enhancement of the reproductive potential for resistance biotypes [36].

4.4. Herbicidal Alternatives for Effective Control of Glyphosate-Resistant Hairy Fleabane

Based on biomass reduction and visual control, the best alternatives tested to control R hairy fleabane were pyraflufen-ethyl and mesotrione. In contrast, paraquat had the highest biomass and the lowest PVC, and was the least effective herbicide tested on all three populations. The opposite results were reported with the same mode of action (HRAC: 22), and diquat achieved 80% control on R populations from Spain [19].

Our results suggest that multiple resistance mechanisms, including vacuolar sequestration by ABC-transporters, may have conferred resistance to both paraquat and glyphosate in these populations, as reported in hairy fleabane populations from the United States [42]. Further research is needed to confirm multiple resistance in these populations from Colombia. So, paraquat (and subsequently diquat) can be discarded as an alternative and effective control method for the cases reported in Colombia.

Glufosinate applied at the rate recommended on the label in Colombia (150 g ai ha⁻¹) offered acceptable control levels (>80%) at 14 DAT in only one population, but the control declined over time and, at 28 DAA, plants recovered from the application and the reduction in biomass was similar to that of paraquat. In populations from Spain, glufosinate offered effective control (>80%) when using a much higher dose of 750 g ai ha⁻¹ [19], so glufosinate could be evaluated with higher doses to set a control alternative with it.

Interestingly, in the case of the auxin herbicide 2,4-D, the population from the passionfruit field (P₂₀) had a satisfactory control (82.5%) in the 14 DAT assessment; however, the control reduced significantly to only 25% at 28 DAT. Multiple resistance in hairy fleabane including 2,4-D and glyphosate has been recently discovered [48,49], so it is necessary to investigate if P₂₀ is 2,4-D resistant or if 2,4-D may also require a higher dose to prevent plant recovery.

Pyraflufen, mesotrione, or 2,4-D can also be tank-mixed with glyphosate to improve the synergistic effect on either resistant or susceptible populations, as found in Spain [19]. In addition, mixtures can be sprayed in pre-plant-burndown or in post-emergence to crop on those crops where selectivity can be achieved and when crop growth and competition may complete effective control of weeds [50].

Tank-mixing glyphosate with pyraflufen-ethyl, glufosinate, or mesotrione also widens the control spectrum. However, tank mixes would increase production cost, a common consequence of resistance [3].

5. Conclusions

A high prevalence of glyphosate resistance was found in hairy fleabane biotypes from Colombia. Populations collected in non-cropping areas, not exposed to glyphosate, remained susceptible to the herbicide. Resistance factor estimates varied from low (3.5-fold) resistance levels in one population collected from plantain (P₁₀) areas to high (22.3-fold) resistance in a population collected from a papaya (P₁₅) farmland, which incidentally had higher levels of resistance compared to the reference population from Spain (17.8-fold). Furthermore, a hormesis response was detected at the recommended glyphosate rates, purporting the ability of hairy fleabane to adapt to adverse environments, including chemical weed control. Two alternative modes of action, pyraflufen-ethyl (8 g ai ha⁻¹) and mesotrione (100 g ai ha⁻¹), delivered circa 80% control of three glyphosate-resistant hairy fleabane biotypes, representing resistance management options in solo applications or in a mixture with glyphosate. Unexpectedly, paraquat (600 g ai ha⁻¹), 2,4-D (720 g ae ha⁻¹) and glufosinate (100 g ai ha⁻¹) offered only inconsistent and partial control of glyphosate resistant hairy fleabane populations. Evolved herbicide resistance is certainly a modern concern in agriculture, calls for attention from the crop protection industry, researchers, and growers to continue pursuing solutions to prevent crop and food losses.