To detect if a fitness cost was associated with the resistant allele, a 3-year field comparative growth analysis was conducted in north-eastern Italy in 2012 (experiment 1), 2013 (experiment 2), and 2014 (experiment 3) on the farm where the original resistant population 05-31 had been collected. In each experiment E. crus-galli plants growth were followed from the beginning of May to mid/end of September (i.e., to maize harvest). R and S plants were grown in virtually non-competitive (hereafter called no-competition) and intraspecific competitive situations: well-spaced (plants were 2 m apart, no-competition) as well as at 3, 7, and 20 plants m^-2 (competition). In competitive situations, each target plant was surrounded by 12 equally spaced plants of the other phenotype (i.e., the S target plant was surrounded by R plants and vice-versa) (Beruto et al., 1996). The experimental design was a completely randomized block with three replicates (see Supplementary Figure S1).

At the beginning of May, seeds of the two sub-populations S and R were chemically scarified in concentrated sulfuric acid (96%) for 20 min, carefully rinsed and conserved in a solution of KNO₃ at 2% (wt/V) until sowing. Sowing was done in the field by putting a few seeds in equidistant pockets (dibbling method). To make sure that the germinated plants were of the correct phenotype, some sterilized soil was put in a buried plastic tube (10 cm × 10 cm) and 15 seeds were sown inside and watered daily. Then seedlings were thinned to one plant and plastic tubes were removed to allow the plants to grow freely.

Phenology of the two phenotypes was monitored throughout the growing cycle: beginning of flowering was recorded for the target plants in all treatments; panicle numbers were counted weekly and covered with glassine bags (Glassine Bag Co. Global Polythene, Corringham, Essex, United Kingdom) just prior to the onset of seed shattering.

Temperature sum was calculated based on Growing-Degree Day equation (GDD):

where T_max and T_min are the daily maximum and minimum temperatures, respectively, and T_b is the base temperature, estimated to be 10.5°C for E. crus-galli (Sartorato and Pignata, 2008). When T_min was ≤T_b, the value of T_b was considered. Air temperature and rainfall data were recorded by the nearest ARPAV (Agenzia Regionale per la Prevenzione e Protezione Ambientale del Veneto¹) weather station (about 8 km away from the experimental field).

In 2012 and 2013 other three well-spaced plants for each phenotype were grown and panicles of these plants were covered with glassine bags in order to evaluate the total number of seeds produced (both retained and shattered). Seed loss (shattered seeds) was estimated as the average difference between the panicles’ weight of the covered plants and those uncovered (i.e., the plants in no-competition situation described above). A correlation between panicles’ weight and the number of panicles was also calculated. This correlation allowed the estimation of the panicles weight in 2014 through counting only the number of panicles.

A destructive sampling was done concurrently with maize harvested in a nearby field: target plants were harvested, different plant parts (stem, leaves, and panicles) were separated and oven dried at 105°C for 36 h, and dry weights were then recorded. In the first year, seeds were separated from rachides and mean dry weight of 100 seeds (in three replicates) was also evaluated to (1) estimate the number of seeds produced by each plant in the different conditions, (2) calculate a correlation between panicle weight and seed weight in order to use the former as indication of reproduction effort (RE) of each phenotype in the different growth conditions (i.e., seed production was estimated using panicle dry weight). The intra-specific competitive ability was evaluated as discussed by Ross and Harper (1972). The space available for plants in no-competitive condition was calculated considering a cube of the radius circle of 1 m (as spaced plants were 2 m apart).

The dose–response data and ALS enzyme assay data were analyzed using a non-linear regression analysis based on the log-logistic equation (Seefeldt et al., 1995):

where Y is plant survival or fresh weight, C and D are the lower and upper asymptotes at high and zero doses, respectively (note: in dose–response data, for biological reasons, and to improve the estimates of the parameters, the upper and lower asymptotes were forced to 100 and 0, respectively), I₅₀ is the dose giving the 50% response, b is the slope and x the herbicide rate. For dose–response experiments, doses giving the 50% response, i.e., LD₅₀ (based on survival data) and GR₅₀ (based on fresh weight data), and relative standard errors, were calculated using the macro BIOASSAY97 developed by Onofri (2005) and running in Windows Excel environment. The resistance index (RI) for the different herbicides was calculated as the ratio between the LD₅₀ (GR₅₀) of sub-population R and LD₅₀ (GR₅₀) of sub-population S.

Growth analyses data were analyzed using the software STATISTICA 7 (Hilbe, 2007): linear regression was used to calculate the relation between panicle and seed dry weight and between panicle number and panicle dry weight in experiment 1 in order to estimate the panicle dry weight in experiment 3. To test differences in slopes among regression lines the ANCOVA was performed.

The results from the outdoor dose–response pot experiment confirmed the preliminary test conducted by GIRE on a seed sample collected in 2005 from plants that had survived a herbicide treatment with nicosulfuron on the same farm. Plant survival and fresh weight gave similar results (Table 2). It was confirmed that S phenotype was adequately controlled by all ALS inhibitors tested, whereas R phenotype was highly cross-resistant to all ALS inhibitors with high RI calculated considering both plant survival and fresh weight (Table 2). Apart from nicosulfuron, which controlled more than 90% of R plants when applied at 320 g a.i. ha^-1 (i.e., 8x), for all other herbicides plant survival was always higher than 50% of the untreated control. Therefore it was impossible to fit the data with the log-logistic equation and LD₅₀ values were considered as being higher than the maximum herbicide dose used, e.g., for penoxsulam it was >653 g a.i. ha^-1 (Table 2). For fresh weight, it was not possible to calculate GR₅₀ for imazamox: plants were not damaged even at the highest dose applied (640 g a.i. ha^-1, i.e., 16x). Instead, GR₅₀ were calculated for bispyribac-Na and penoxsulam, indicating that plants were damaged, though alive, starting from dose 1x and 4x, respectively. These two herbicides were also more effective on the S phenotype and plants were already damaged at very low doses, therefore the RI resulted as very high (Table 2).

Molecular analyses proved that a target-site resistance mechanism is involved. The PCR amplification with ECH-ALS-F and ECH-3R led to obtaining a 1,870 bp sequence including all the known conserved domains identified previously carrying mutations endowing resistance. In all the resistant plants a double nucleotide substitution GC-AA, giving an Ala-Asn change at amino acid position 122 of the ALS gene, was detected. All plants were heterozygous for the mutation at the ALS locus. In the susceptible plants not one known mutation was identified.

The in vitro measurements of ALS activity indicated a sigmoidal response to increasing doses of imazamox and nicosulfuron (Figure 1) and the data were well fitted by the log-logistic equation (Table 3). The results of the in vitro bioassay support the presence of a target-site mediated as main resistance mechanism. The R/S ratios calculated were 7.4 for imazamox and 413 for nicosulfuron. The large difference in I₅₀ between R (I₅₀ = 7980) and S (I₅₀ = 19) sub-populations led to the exclusion of non-target-site mediated resistance mechanisms for the selecting agent of the resistant phenotype, i.e., nicosulfuron. This is in keeping with the results obtained from the molecular analyses.

All E. crus-galli seedlings used as target plants were analyzed to verify the presence or not of the mutant ALS allele. All resistant plants had the double substitution GC-AA, giving an Ala-122-Asn amino-acid change, whereas none of the susceptible plants had any nucleotide substitution.

The onset of flowering was delayed of about 1 month in 2013 in comparison to 2012 and 2014 (Table 4). In all years a later flowering in R plants with respect to S ones was recorded in both competitive and non-competitive situations (Table 4). Considering the cumulative number of panicles under no-competition in relation to the temperature sum (Figure 2), the same delay of about 1 month was recorded in 2013. It is worth mentioning that the temperature sum at the end of August 2013 was around 1,000°C d, whereas this value was reached at the beginning of August in 2012 and 2014. The production of panicles for R and S plants in 2013 was similar, whereas at harvest R plants produced 55% and 70% less panicles than S plants in 2012 and 2014, respectively (Figure 2). Considering the intraspecific competition, the larger differences between R and S plants were recorded in 2013 and 2014 at the lowest density (3 plants m^-2) and progressively disappeared at higher plant densities (Figure 3). No significant differences were detected in 2012, likely due to the higher variability in panicles produced by S plants.

Evaluation of the dry biomass, in particular the panicle dry weight produced by S and R plants, consolidated the data from the phenological analyses. The highly significant correlation (R² = 0.99) between seed dry weight (separated from rachides) and panicle dry weight in the first year of experiment allowed this latter variable to be used to estimate the seed production of S and R plants. In 2012 and 2013, panicle number and panicle dry weight were recorded for R and S plants and a linear regression analysis between these two parameters was obtained. In 2012, the R² was 0.98 (linear equation: y = 0.675x), whereas in 2013 the R² was slightly lower (0.93). Bartlett’s test for homogeneity of variance revealed that it was not possible to analyze all data together. Therefore, because of the higher R² only 2012 data were used for the estimation of panicle dry weight in the last experiment.

The comparison between different vegetative plant parts (stems and leaves) did not highlight any difference between R and S plants (data not shown). Considering the total dry weight (Figure 4, on the right): a difference was detected in only two cases (i.e., 2012 – 7 plants m^-2 and 2013 – 3 plants m^-2). Instead, considering panicle dry weight, S plants allocated more biomass to the reproductive organs than R ones especially in no-competition and at lower plant densities (Figure 4, on the left). At the highest plant density (i.e., 20 plants m^-2) the competition among plants was high and panicle production very low, giving no differences in reproductive effort between S and R plants. Data from 2012 and 2014 were consistent, whereas in 2013 a slight difference (not significant) in panicle dry weight between S and R plants was detected only in the 3 plants m^-2 situation (Figure 4). Other differences between S and R phenotypes were detected analyzing seed weight in 2012. Seed dry weight of S plants was about 30% higher than in R plants in no-competition situation (Figure 5A). Also in this case, the difference decreased enhancing the competition level. However, the mean weights of 100 seeds evaluated for the different situations have a different trend: mean seed weight increased with competition level for S plants, whereas it decreased for R ones (Figure 5B). In no-competition situation, the mean seed weight of S plants was lower than that of R ones, but S plants produced much more biomass. The resulting estimated number of seeds produced by R and S plants was significantly different (Figure 5C), i.e., the lower weight of a single seed in the S plant increases the difference in the number of seeds between S and R plants. The S plant produced 43% more seeds than the R one (Figure 5C). This difference disappeared at the highest plant density.

In order to compare the intra-specific competitive ability (i.e., between R and S plants), the relation between total dry weight (or panicle dry weight) and the three-dimensional space that plants have available, expressed as cube of radius of the circle, was calculated (Figure 6) (Ross and Harper, 1972). Considering the total dry weight, the slope of the regression lines between S and R plants was not significantly different in both 2012 and 2014 (t-value = 0.719, P = 0.512 in 2012; t-value = 1.02, P = 0.366 in 2014) (Figure 6A). Different behavior is observed for panicle dry weight (Figure 6B). The slopes of the regression lines between S and R plants were significantly different in both 2012 and 2014 (t-value 7.410, P = 0.002 in 2012; t-value 3.209, P = 0.03 in 2014). The more space available for plant growth the bigger the difference is in the biomass allocated to panicles (and therefore to seeds) between R and S plants. This implies that R plants have a lower potential for seed production, in other words they suffer from a fitness penalty in comparison with S plants.

This study reports for the first time the Ala-122-Asn amino-acid change in the ALS gene that confers a high level of cross-resistance to ALS inhibitors worldwide. Resistance-endowing mutations at Ala-122 have so far been reported in seven weed species (Tranel et al., 2017). The Ala-122-Thr endowed resistance only to imidazolinones but no resistance to sulfonylureas and pyrimidinylthiobenzoates in Solanum ptycanthum, Amaranthus retroflexus, A. powellii, and Apera spica-venti (Tranel et al., 2017). In E. crus-galli from Arkansas, this amino acid change conferred resistance not only to imidazolinones but also to triazolopyrimidines (Riar et al., 2013). The Ala-122-Val did not confer a broad resistance to the ALS inhibitors either (Riar et al., 2013; Tranel et al., 2017). Instead, the double point mutation GCT-TAT that determines an Ala-Tyr amino-acid change, detected only in R. raphanistrum (Han et al., 2012) conferred high-level and broad-spectrum resistance to ALS-inhibiting herbicides (Han et al., 2012; Tranel et al., 2017). Similarly, in E. crus-galli populations analyzed herein, the Ala-122-Asn amino-acid change in the ALS gene confers a broad spectrum resistance. Overall, resistance-endowing mutations at Ala-122 of the ALS gene can determine a different resistance pattern according to weed species and amino acid change. This reinforces the idea that each resistance case needs a specific study.

The resistance costs associated to specific ALS resistance-endowing mutations that have evolved in field weed populations have been examined in a limited number of cases. Here, resistant and susceptible plants of E. crus-galli with similar genetic background were selected and the detection of a target-site mediated resistance mechanism involved has allowed fitness costs associated with resistance to ALS herbicides to be estimated and interpreted.

Results of the 3-year growth analysis indicate that a fitness cost is present in plants bearing the Ala-122-Asn substitution. Differences were observed in growth and development: a slower development of R phenotype in relation to the S one was recorded (Table 3). Even if during the second year experiment, the date of beginning of flowering was delayed by 1 month with respect to 2012 and 2014, this 1 month delay was recorded throughout the life cycle, from appearance of the first panicle until harvest time (Table 3). As previously argued, the results of field experiments can hardly be treated quantitatively because many indeterminate factors, including the weather, may be of great importance (De Wit, 1960). However, our results highlight that the differences between R and S plants were consistent in different years and climatic conditions. Furthermore, in no-competition condition, R plants produced about 40% fewer panicles than S plants (Figure 2) and the high panicle dry weights recorded for S plants indicate that they allocated more biomass to reproductive organs than the R ones. It can be inferred that the lack of differences between R and S in the number of panicles produced in 2013 can be attributed to the different climatic condition. During 2013, there was a rather rainy spring with respect to 2012 and 2014 that delayed seedling emergence and therefore the subsequent phenological growth phases.

Considering the competitiveness of R and S phenotypes, the results indicated that the higher the plant density the less the difference was between R and S phenotypes in terms of number of panicles produced. No differences were recorded at the highest plant density. Hence, at low plant densities the susceptible plants exhibited a potential fitness advantage. The only study of fitness costs on mutation at Ala-122 was conducted on R. Raphanistrum having the Ala-122-Tyr mutation and no evidence of pleiotropic effects on plant growth were recorded between R and S plants (Li et al., 2013).

Overall, it seems that fitness costs related to ALS resistance vary according to the species, the mutated allele conferring resistance, the genetic background, experimental conditions (glass-house, field) and environmental conditions involved. This study reveals that fitness costs associated to Ala-122-Asn mutant E. crus-galli plants are evident under good growth conditions, but progressively disappear in highly competitive conditions.

Experiments conducted under field conditions allow fitness to be compared in a real situation where resistance evolves and has to be managed. This study proves that the evolved resistance to ALS inhibitors in E. crus-galli due to the Ala-122-Asn amino-acid change is associated with a fitness cost. However, the impact of this penalty in the absence of ALS inhibitors may be influenced by many factors, including environmental conditions (Gassmann, 2005). Different climatic conditions during 2013 influenced the relative plant growth, reducing the disadvantages of R plants toward the S plants.

The differential population dynamics observed between R and S E. crus-galli phenotypes can be a good starting point for devising a resistance management strategy. Differences in life history traits can be manipulated in fields to minimize or reverse the resistance evolution (Vila-Aiub et al., 2011). In our case the lack of a fitness cost at high plant densities suggests that keeping the infestation density as low as possible can increase the reproductive success of the S phenotype and therefore contribute to lowering the ratio between R and S alleles. In a situation where ALS-resistance is well established, a longer-term management based on crop rotation, lack of selection pressure from ALS-inhibiting herbicides and the use of any other available control tool should help to induce some change in the R/S plant ratio.

The different weather conditions encountered in 2013 may suggest other management options. The frequent and relatively abundant rainfall events in spring delayed maize sowing and weed control operations for about 1 month in relation to the normal schedule. As a consequence, the emergence of E. crus-galli, usually characterized by two main fluxes of emergences regulated by rainfall and temperature, was instead only one large and longer flux. In this case, the delay in sowing helped the control of weeds.

Another experiment was conducted in the same maize field where the conventional weed control strategy (with ALS inhibitors) adopted by the farm was compared with another one that did not include these herbicides for 4 years. The results showed that the simple measure of ceasing to use the selecting agent does not significantly modify the ratio between R and S plants when the barnyardgrass seed bank initially had more than 95% of “R seeds” (Scarabel et al., unpublished data). It is clear that the long-lived seed bank of E. crus-galli (Holm et al., 1977) imposes long-term resistance management strategies when the seed bank contains predominantly “R seeds.”

It is widely accepted that herbicides are not silver bullets (e.g., Mortensen et al., 2000) for the management of resistance, but rather diversified control tools should be used as “little hammers” (Barzman et al., 2015). If adequately embedded in a medium or long-term integrated weed management strategy, the presence of R plants with a fitness penalty provides an opportunity to minimize or reverse herbicide resistance evolution.