Herbicide resistance is an large global problem observed in 250 weed species for 23 of the 26 known herbicide sites of action (SoA) (Heap, 2016). The overuse over many generations of herbicides targeting a few specific SoA in weeds has led to the evolution of weed populations surviving increasing doses of herbicides. Evolution through natural selection in plants is a continuous process which allows weed populations to eventually overcome most selection pressures that humans apply as eradication strategies. Therefore, the use of herbicides with novel SoA might control herbicide resistant weeds for a while, but inevitably some populations will evolve resistance. The herbicide resistance mechanisms are an important factor in the strength and speed at which herbicide resistance can evolve de novo or spread from neighboring populations. Non-target site resistance (NTSR) implies to repurpose pre-existent stress and defense enzyme pathways to defend the plant against herbicides. NTSR pre-dates herbicide use and implicate a wide diversity of pathways and is inherited in complex manners due to polygenic identity (Délye et al., 2011; Délye, 2013). NTSR normally confers multiple resistance to other herbicide SoA (Burnet et al., 1994; Petit et al., 2009; Cummins et al., 2013; Délye, 2013). Some weeds have shown to be resistant to herbicides that they had never been exposed to Espeby et al. (2011). Therefore, management of herbicide resistance cannot rely solely on chemical solutions and must incorporate evolutionary biology knowledge when dealing with NTSR resistant weeds (Neve et al., 2009, 2014).

Traditionally, fitness is defined as the number of viable and fertile offspring contributing to the next generation. In plants, this implies that fitness could only be measured by evaluating the quantity and quality of seeds (Vila-Aiub et al., 2009). However, the number of seeds produced can depend on the health and growth of the individual. This is based on the allocation of resource theory stating that individual’s metabolism has a limiting amount of energy to allocate to vegetative growth versus reproduction and that any extra energy allocated to health (for example if plants live in a sub-optimum environment) will have a negative effect on its reproductive ability (Maxwell et al., 1990; Herms and Mattson, 1992; Park et al., 2004; Vila-Aiub et al., 2015). Therefore, fitness can also be evaluated by using proxies such as growth and health measurements. Specifically regarding agricultural weeds, fitness has been defined as survival and reproductive success in field conditions (Menchari et al., 2007) which is directly linked to competitive ability. Therefore, evaluating traits linked to competitive ability such as germination rate, vegetative growth characteristics and time needed to reach specific growth stages can indicate relative fitness differences.

The direction of natural selection depends on the selective pressure (e.g., herbicide) and also on the fitness of individuals facing this new pressure. In weeds, it was expected, and has also been observed, that herbicide resistant individuals had a lower fitness than their susceptible counterparts in the absence of the herbicide in terms of biomass and/or seed production (Bergelson and Purrington, 1996; Purrington, 2000; Ashigh and Tardif, 2009). The presence of a fitness cost could dictate different management strategies in order to reduce the resistant population. A resistant weed population with a strong fitness penalty could, in theory, be reverted to a susceptible status if the herbicide is not used.

Different resistance mechanisms [target-site (TSR) vs. NTSR] result in different fitness (Roux et al., 2005; Vila-Aiub et al., 2009; Wang et al., 2010). NTSR mechanisms are hypothesized to have a fitness cost because of negative consequences due to the changes in pathway dynamics which could even alter ecological interactions. Also negative effects on energetic resources being diverted from growth and reproduction to defense mechanism have been suggested (Jasieniuk et al., 1996; Vila-Aiub et al., 2009). Increased detoxification due to cytochrome P450 monooxygenase (P450s) in an acetyl-CoA carboxylase inhibitor (ACCase) NTSR resistant Lolium rigidum was correlated with a reduced seed production, biomass and growth rate (Ashigh and Tardif, 2009). Similarly, NTSR resistant Bromus tectorum had a reduced biomass, seed number and leaf area (Park and Mallory-Smith, 2005). However, the assumption of negative fitness in resistant weeds has been challenged as more observations of neutral and even positive fitness have arised. Neutral fitness has been observed in glyphosate NTSR resistant Amaranthus powellii (Giacomini et al., 2014). Positive fitness was observed in ACCase resistant Setaria viridis and triazine resistant Phalaris paradoxa (Schönfeld et al., 1987). NTSR mechanisms are diverse and a multitude of large gene families are involved such as P450s, glycosyltransferase (GT), glutathione S-transferase (GST), ABC-transporters, esterase, etc. (de Carvalho et al., 2009; Délye, 2013). These gene families all act in different ways (e.g., conjugation, compartmentalisation, transport) to allow the plant to survive. This multitude of factors is why the fitness of resistant weeds must be evaluated on a case by case basis (Lehnhoff et al., 2013a; Vila-Aiub et al., 2015).

Acetolactate synthase (ALS) inhibitor herbicides inhibit the synthesis of branched-chain amino acids valine, leucine and isoleucine (Duggleby et al., 2008). ALS herbicides have a high number of resistant weed species than any other SoA (Heap, 2016). Loose silky bentgrass [Apera spica-venti (L.) Beauv.] is one of the most serious grass weed in Northern, Eastern and Central Europe (Hamouzová et al., 2011; Massa and Gerhards, 2011; Schulz et al., 2011). At a density of 200 plants m^-2, A. spica-venti reduces the yield of winter cereals up to 30%, and it can cause greater yield losses than Alopecurus myosuroides (Melander, 1995; Melander et al., 2008). The highest number of resistance cases in A. spica-venti has been reported for ALS herbicides (10 cases) followed by photosystem II (seven cases) and ACCase (three cases) (Heap, 2016). Two studies have investigated fitness differences between ALS resistant and susceptible A. spica-venti biotypes, both evaluating germination efficiency (Soukup et al., 2006; Gerhards and Massa, 2011). These studies revealed contradictory results where Soukup et al. (2006) found no fitness differences in a population with unknown resistance mechanism, while Gerhards and Massa (2011) found a threefold increase in germination rates in the TSR resistant biotypes. However, accurate estimation of fitness of herbicide resistant weeds is a difficult methodological task. Because of the importance of both environmental conditions and genetic background, experimental conditions and genotypes to be tested must be thoroughly controlled otherwise the fitness assessment cannot be solely attributed to the resistance allele(s) (Jasieniuk et al., 1996; Ashigh and Tardif, 2009; Vila-Aiub et al., 2009).

The lack of information of fitness in NTSR resistant A. spica-venti limits the implementation of evolution-knowledge-based resistance management strategies. This study aimed to evaluate the fitness and growth characteristics throughout the life cycle of ALS NTSR resistant A. spica-venti biotype with a randomized genetic background. The genetic backgrounds of resistant and susceptible populations were randomized over two successive generations. ALS resistance level and mechanisms were assessed. Fitness was estimated as seed germination at three different temperatures. A target-neighborhood experiment was conducted with three crop density where time to multiple key growth stages were recorded, as well as biomass and seed yield and compared to non-competitive conditions. We hypothesized that either no or higher rate of seed germination would be observed based on previous studies (Soukup et al., 2006; Massa and Gerhards, 2011) and that no differences in growth characteristics would be found based on previous field observations of ALS resistant and susceptible A. spica-venti populations in Denmark (Babineau, personal observation).

The fitness and competitive ability of one ALS NTSR resistant population of A. spica-venti was tested against a susceptible biotype throughout its life cycle. No differences were found in terms of tiller number, biomass, time to stem elongation, time to first visible inflorescence, time to 50% mature seed, nor in final yield. The absence of biomass differences in competitive conditions was also observed in NTSR resistant Avena fatua (Lehnhoff et al., 2013a). A significant correlation was observed for the resistant population in terms of seed germination, time to emerged inflorescence (INFEM), and time to 100% flowering (FLO).

The resistant population germinated 9 to 20 °Cd earlier than the susceptible population which in Denmark would represent 1–2 days difference at the time of sowing of winter cereals. This result is consistent with the results reported for TSR ALS resistant B. tectorum and Kochia scoparia biotypes which showed significantly higher germination rate at low temperatures (5°C) compared to the corresponding susceptible biotypes (Dyer et al., 1993; Park et al., 2004). Differences in germination rate were observed only at the lowest temperature unlike a previous study of ALS resistant A. spica-venti where a threefold increase in germination for the TSR (Pro197, Trp574, and Arg377) resistant biotypes was observed (Gerhards and Massa, 2011; Massa et al., 2011) at 20°C/15°C with 12 h photoperiod. This difference could be due to resistance mechanisms as TSR and NTSR have very different genetic implications and different genes are modified to allow survival (Délye, 2013). Single nucleotide mutations are hypothesized to have negative pleiotropic effect on gene function whereas NTSR mechanisms are thought to have negative effect on growth as a limited amount of resources have to be allocated to either defense or growth (Vila-Aiub et al., 2015). Therefore, the modified ALS gene, which is expressed in seeds and is essential in the synthesis of amino acids leucine, valine and isoleucine, in the TSR individuals could explain the increased germination rate found by Gerhards and Massa (2011). A reduction in total germination for the resistant population was observed in this study compared to the susceptible, in contrast to Soukup et al. (2006) who found no differences between four ALS resistant and four susceptible field populations of A. spica-venti tested in Petri dishes (20°C) and pot experiment. The difference in total germination could be due to differences in dormancy proportions between resistant and susceptible, as A. spica-venti has previously shown variation in seed primary dormancy (Andersson and Åkerblom Espeby, 2009).

Weeds that germinate earlier can better compete with crop and other weeds which gives them a considerable advantage (Kleemann and Gill, 2013; Owen et al., 2015). More than 30 different P450s are expressed in A. spica-venti seeds at different germination stages and NTSR, via an increased cytochrome P450s activity, can be hypothesized to have a pleiotropic effect on germination (Babineau et al., 2017). The results of this study also show differences in the magnitude with germination at 16°C showing the largest difference between resistant and susceptible population compared to 10 and 22°C. Fitness differences between germination temperatures were also observed in resistant B. tectorum showing germination differences between ALS resistant and susceptible populations only at low (5°C vs. 15°C and 25°C) temperature (Park et al., 2004). This could be explained by the known association of germination efficiency and temperature and might reflect regional adaption. Alternatively, there could be linkage between genes associated with germination timing and P450s under selection for ALS resistance. This is supported by the fact that P450s genes in Arabidopsis thaliana have been shown to evolved from ancient whole genome and successive gene duplication of P450s tandem arrays resulting in P40s genes being tightly linked and distributed on multiple chromosomes (Bak et al., 2011). Several P450s (CYP707A1-A4) have been found to be directly involved in decreasing the levels of abscisic acid (ABA) during seed imbibition in A. thaliana which allows seeds to germinate by lowering dormancy (Schopfer et al., 1979; Kushiro et al., 2004). The same relationship between P450s and seed germination has not yet been established in A. spica-venti, but the increased germination in individuals showing a high variation in P450s due to herbicide selection this is possible.

At the medium density of 96 wheat plants m^-2 the ALS NTSR resistant population took less time to reach INFEM and FLO by 383 and 196 °Cd respectively, which corresponds to around 25 and 13 days, respectively, in field condition in Denmark. The importance of timing of trait has been underlined before (Paris et al., 2008). Similarly, NTSR resistant A. fatua reached anthesis earlier than susceptible biotypes (Lehnhoff et al., 2013b). Flowering time appears to be a plastic trait as flowering time in Raphanus raphanistrum was halved within five generations using directional selection (Ashworth et al., 2016). The ability to reach inflorescence emergence and flowering earlier is an advantage to weed populations allowing them to escape potential eradication by late season management strategies or harvesting. Herbicide NTSR resistance mechanisms can have different pleiotropic effects that could manifest themselves only at specific growth stages. In the field, difference in flowering time could imply selective interbreeding only between members of a resistant population which could result in sub-population structuring and differentiation over consecutive generations. Therefore, different ALS resistance mechanisms present in neighboring populations would not mix due to gene flow and populations could instead become increasingly genetically isolated.

The effect of crop competition on fitness and growth characteristics of ALS resistant biotypes shows an interesting pattern in this study. There were no significant differences at low densities (D0 and D2) neither at the high density (D8) in any of the five growth stages. Growth differences only manifested at 96 winter wheat plants m^-2 (D4). This result indicates that high crop density reduces the fitness of resistant biotypes to an equal level to the fitness of susceptible populations as previous studies have been shown (Park et al., 2004; Paris et al., 2008). Recovery of growth timing differences at intermediate crop density and not at the lower densities has also been shown previously at D4 in A. fatua (Lehnhoff et al., 2013a,b). This could indicate differences in the regulation of NTSR constitutively expressed genes. Competition is known to influence the traits that exhibit fitness differences. Evolution in competitive conditions will favor fitness cost on early growth traits while evolution in non-competitive conditions will favor fitness cost on later growth traits (Paris et al., 2008). However, a fitness advantage of the resistant population in both early (germination) and later (inflorescence emergence and flowering) growth stages might imply very different and variable competitive conditions when ALS resistance evolved. The crop density in which we observe significant growth differences is much lower than the realistic wheat density (200–300 m^-2) in farm conditions. We did not observed any differences in the growth of NTSR A. spica-venti at the lower end of the farm wheat density spectrum (D8; 192 m^-2). Therefore, the timing of growth stages differences observed in this study most likely will not be observed in field conditions at normal wheat densities. However, in crops with lower farm densities or different competitive ability compared to wheat, the differences observed here in the timing of growth stages for this weed could be observable and would be worth investigating.

This study did not demonstrated a fitness cost of resistance, as often hypothesized, but found a correlation with earlier germination and growth which could be a fitness benefit in some field conditions. Earlier germination can translate to a better access to nitrogen and nutrients and early vigor. However, this benefit could turn into a cost in case of false seedbed or pre-emergence application of foliar-active herbicides.

Fitness benefits have been attributed to two possible processes: compensatory and replacement hypothesis (McKenzie et al., 1982; Andersson, 2003; Paris et al., 2008; Darmency et al., 2015). Compensatory mechanism implies specific pleiotropic consequences, where the resistance allele is linked to other alleles (modifier genes) that help compensate for the fitness cost of resistance (Andersson, 2003; Darmency et al., 2015). These modifier genes can be linked to resistance alleles from the beginning or can evolve rapidly (McKenzie et al., 1982). The replacement hypothesis argues that resistance alleles carrying a fitness penalty can be replaced over time with similar resistance alleles that do not carry a fitness penalty (Paris et al., 2008). In the case of the resistant population examined in this study, both hypotheses could apply since it is unknown how long ago the resistance evolved and therefore replacement or modifier genes could have had enough time to evolve and be fixed.

In general, fitness is defined as “to survive and produce a number of fertile and viable offspring that will contribute to the next generation” (Barker, 2009). In strict evolutionary terms, there were no fitness cost or benefit observed in this study as the final seed number (viability was not measured) was not different between resistant and susceptible A. spica-venti. However, in an agricultural landscape, plants (crops and weeds) are managed throughout their life cycle and their competitive ability has a direct effect on their survival which often cannot be measured solely based on seed production. Small differences in competitive ability or in management practices can have large effect on weed population level which will obviously be observed in seed yield at the end of the season. But this difference in seed yield will not identify the precise life stage and precise differences that led to the higher total fitness which is key in weed management. Therefore fitness evaluation and measurement in agricultural conditions has to be enlarged to encompass growth and competitive differences (Vila-Aiub et al., 2009, 2015) much more prominently than viable offspring evaluation seen in natural environmental studies (Leimu et al., 2006). For these reasons, we surmise that fitness cost or benefits in weeds should consistently aim to assess plants throughout their life cycle. The strict definition of fitness based on viable offspring number is not appropriate for evaluating plants in intensively managed anthropogenic environments.

Another theoretical aspect emerging from this study is the use of randomized genetic background in studying NTSR mechanisms. This randomization of the genetic background method used to compare individual fitness in herbicide resistant weeds was originally designed for one allele difference, e.g., TSR and was used as such (Roux et al., 2004; Menchari et al., 2007; Delye et al., 2013). The handful of studies investigating fitness cost of NTSR mechanisms have used resistant and susceptible individuals from the same population (P450s; Vila-Aiub et al., 2005), reduced translocation (Wakelin and Preston, 2006; Pedersen et al., 2007), or the genetic background randomization (EPSPS gene amplification, Giacomini et al., 2014). However, just as in our study, the exact NTSR allele(s) was not identified (e.g., which P450s isoenzyme) and fitness evaluation was based on bulk assessment of involvement of one or multiple large gene families. The problem with evaluating fitness cost from NTSR is that, as mentioned before, these mechanisms are assumed to be polygenic and generally involves multiple alleles, from multiple loci across a varied functional background. All of these have not only their own evolutionary history regarding herbicide resistance, genetic linkage and inheritance, but most likely have their own fitness cost associated. This bulk assessment of NTSR fitness cost comprises a mixture of genes that could have additive, multiplicative, agonist, antagonist, and compensatory effects on fitness phenotype depending on the combination that has evolved in a particular population in a particular weed species.

The aim of the genetic randomization method is to have similar genetic background except for the alleles conferring resistance by equally mixing the whole genetic background of both herbicide resistant and susceptible individuals by performing crosses for a few generations. In this context, it is uncertain how the various NTSR alleles will be distributed after a few generation of genetic randomization background; they could be randomly distributed but then no “true” susceptible will be found, they could be lost, or some alleles could co-segregate due to linkage or recombination. The latter would result in individuals that are resistant to herbicide due to NTSR but not the same allele combinations than found in the field which could not guide weed management strategies. The assessment of field populations displaying NTSR will be difficult because of the reasons explained above and also because NTSR mechanisms pre-exist at a certain degree in every individuals no matter their phenotype. The study of the bulk effect of NTSR mechanism on fitness is useful as it represents more closely the situation in the field and also gives a response to the question of the presence or absence of a fitness cost. In the absence of a fitness cost, the resistant individuals are not likely to decrease without the herbicide and other management strategies can be sought. If a fitness cost (or benefit) is detected from the bulk analysis, then the identification of the specific alleles and their individual effect on fitness cost becomes necessary.

An appropriate method for estimating fitness cost from individual NTSR alleles and their respective roles in fitness would be to create different lines with very similar genetic backgrounds (using near isogenic lines or model plant A. thaliana) but differing by only one allele. Increasing gradually the number of NTSR alleles in some lines could allow to effectively estimate the epistatic effect (cumulative, antagonist, etc.) of the different genes involved in NTSR. This method requires time, large spatial resources and genetic transformation ability (if used on a weed species) but most importantly requires to identify each NTSR alleles thought to be involved in herbicide resistance in a particular individuals showing a fitness cost which would necessitate a differential gene expression pre-analysis. This method could also allow testing the hypothesis of modifier genes. To our knowledge, no study has performed such an analysis yet but with the increasing number of herbicide NTSR fitness cost studies (Park and Mallory-Smith, 2005; Vila-Aiub et al., 2005; Pedersen et al., 2007; Preston and Wakelin, 2008; Giacomini et al., 2014) and transcriptomic studies identifying herbicide resistant genes (Hu et al., 2009; Cummins et al., 2013; Gaines et al., 2014; Duhoux et al., 2015) this method has the basic knowledge available to be used. Once used across a range of NTSR alleles, the comparison with different NTSR weed species or populations might inform about the universality of fitness cost associated to certain alleles which could help weed management. In our study for example, the inconsistent fitness difference observed (germination and flowering stage only) could be explained by the role of different NTSR genes having pleiotropic effects at different life stages.

The presence of fitness differences between herbicide resistant and susceptible biotypes has been used previously to develop management strategies that exploit those differences (Neve et al., 2009, 2014; Preston et al., 2009), however, most have focused on a fitness cost assumption. The finding of potential beneficial differences in ALS NTSR resistant A. spica-venti, as in this study and by Gerhards and Massa (Gerhards and Massa, 2011), calls for the development of management strategies that take into account different direction and magnitudes of fitness consequences (Colbach et al., 2006). The early germination observed for the resistant biotype in this study could imply a selective eradication by pre-emergence application of foliar-active herbicides or stale seedbed methods. The earlier flowering and seed production stages will be difficult to manage. However, flowering differences were recovered at a crop density lower than in Danish fields, which means that at the current sowing densities of winter wheat these differences are most likely non-existent and would not cause new management problems. The management of such growth differences in NTSR biotypes cannot be chemically managed because of the high risk of multiple resistance, which emphasizes the importance of herbicide resistance prevention strategies.

Lastly, the results observed here come from only one NTSR population. Therefore, the investigation of more A. spica-venti NTSR populations in terms of germination and growth characteristics might confirm, or nuance, the results observed here, as sample number is an important factor is the estimation of herbicide resistance fitness cost in weeds (Cousens et al., 1997).

This study observed a correlation between germination and growth differences and NTSR resistance in an A. spica-venti population. The resistant population germinated 9–20 GDD earlier at 10, 16, and 22°C, reached inflorescence 383 GDD earlier and flowered 196 GDD earlier at a density of 96 wheat m^-2 compared to the susceptible population. No differences were observed at other wheat densities or regarding tiller number, biomass, time to stem elongation, time to first visible inflorescence, time to 50% mature seed, nor in final yield. The differences in growth characteristics identified could be used to better manage ALS NTSR loose silky bentgrass populations in the future.

MB designed, conducted, analyzed data and drafted the manuscript. SM, MK, and PK designed the experiment, advised with data interpretation and reviewed the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.