Tolerance to Glyphosate in Broadleaf Buttonweed and White-Eye Biotypes

Infesting species of Rubiaceae family have great importance in soybean crops in Southwest of Paraná and North of Santa Catarina, especially due to the selection of tolerant populations caused by the pressure of selection exerted by intensive use of glyphosate. The study aimed to evaluate the levels of tolerance to glyphosate in biotypes of broadleaf buttonweed (Borreria latifolia) (BL) and whiteeye (Richardia brasiliensis) (RB) by means of dose-response curves. Two experiments were conducted in a greenhouse, in completely randomized design (CRD), with four replications. Fourteen BL biotypes and eleven RB biotypes were evaluated, collected in soybean fields of Paraná and Santa Catarina. The doses of glyphosate were 0, 74, 163, 360, 792 and 1,742 g ha-1, applied when the plants had 6 to 8 leaves. Visual control was evaluated at 14 and 28 days after application (DAA), as well as mass of green shoots and mass of dry shoots at 28 DAA. It was verified that there is response variability among the biotypes studied. BL biotypes 277, 283 and 300, and RB biotypes 283, 285 and Papanduva were not controlled with glyphosate at doses superior to those normally used in crops (720 g ha-1), indicating selection by the recurrent use of glyphosate. The tolerance factors ranged from 1.1 to 4.1 and 2.8 to 8.1 for RB and BL, respectively. The greatest difficulties of control were reported in areas where RR technology had been adopted for more than six years.


INTRODUCTION
Rubiaceae is one of the greatest angiosperm families, with around 13,000 species worldwide, and 1,500 are found in Brazil (Delprete, 1999;Souza and Lorenzi, 2005). Among the species found in Brazil with high economic value, we can highlight coffee tree (Coffea sp.), ipecacuanha (Psychotria ipecacuanha), and gardenia (Gardenia jasminoides), for instance. On the other hand, many others are considered weeds (Kissmann and Groth, 2000), such as genera Borreria (Spermacoce) and Richardia, which are of great relevance in Southwest of Paraná, especially in soybean fields. As they interfere negatively with the crop yield, there is need to control these infesting species.
Chemical control is one of the main methods employed for weed control. However, the intensive and improper use of herbicides can provoke negative effects on the environment, such as the selection of weed species that are tolerant and resistant to the different existent herbicide mechanisms of action (Ferreira et al., 2009). The pressure of selection exerted by a given herbicide on the weed community can cause increased proportion of a species over the others because of tolerance to the herbicide (Christoffoleti et al., 2008).
Tolerance, also defined as natural resistance (Duke, 2011), is a feature of the species even before the application of the herbicide in a given area (Vargas et al., 2009). On the other hand, resistance is the ability of a plant to survive and reproduce after exposition to a dose of herbicide normally fatal for its wild biotype (Christoffoleti et al., 2008). Nevertheless, existence of resistance presupposes that there is a population of a weed biotype previously susceptible to the herbicide (Heap and Lebaron, 2001;Gazziero et al., 2009), what does not occur with populations tolerant to herbicides.
Glyphosate is the most applied herbicide worldwide due to low cost, great efficiency and broad adoption, by farmers, of genetically modified crops (GMCs) with resistance to this herbicide. The highest risk related to the adoption of GMCs in Brazil and South America is due to the selection of weeds with resistance and tolerance to herbicides (Cerdeira et al., 2010). In areas with frequent use of glyphosate, as in genetically modified soybean (RR) fields, coffee trees and orchards areas, there has been selection of species tolerant to this product.
Modifications in the composition of weed populations and seed banks have been registered in several countries because of the intensive use of glyphosate (Cerdeira et al., 2007;Cerdeira et al., 2010;Webster and Nichols, 2012). In Brazil, the intensified use of glyphosate causes the interspecific selection of some weed species, as wandering-jew (Commelina spp.), morning glory (Ipomoea spp.), white-eye (R. brasiliensis Gomes) and broadleaf buttonweed (B. latifolia (Aubl.) K. Schum). The reason for differential tolerance was elucidated in some species, as Commelina benghalensis, Ipomoea grandifolia and Amaranthus hybridus (Monquero et al., 2004), Ipomoea nil, Tridax procumbens and Spermacoce latifolia (Galon et al., 2013), although the knowledge of the mechanism of tolerance has not been yet elucidated for others.
B. latifolia and R. brasiliensis have been documented in the literature as being tolerant to glyphosate (Sharma and Singh, 2001;Lacerda and Victoria Filho, 2004;Monquero et al., 2004;Cerdeira et al., 2010). Species or populations of the same weed, or even a single biotype, may present variability regarding tolerance to a given herbicide. In order to obtain control of around 90% of Spermacoce latifolia (sin. B. latifolia), the application of doses above 720 g a.e. (acid equivalent) ha -1 is necessary (Lacerda and Victoria Filho, 2004). Applications of glyphosate (720 g a.e. ha -1 ) to seedlings of B. latifolia with one-three and four-six leaves promoted control of 100% and 81%, respectively (Ramires et al., 2011).
Several studies conducted on R. brasiliensis have found that, regarding the species, there is also great variability of control by means of glyphosate. Use of glyphosate alone at a dose of 420 g ha -1 resulted in 90% control at 14 DAA (Monquero et al., 2004). Vitorino et al. (2012) reported level of 99.5% control of white-eye at 28 DAA with the employment of glyphosate at a dose of 720 g ha -1 , and it was superior to the other alternative herbicides tested. On the other hand, use of glyphosate alone at a dose of 770 g ha -1 presented low level of efficiency to control R. brasiliensis, only reaching 14% (Sharma and Singh, 2001).
In order to confirm the occurrence of resistance or differential tolerance, the performance of dose-response curves is recommended, and through them it is possible to know the dose needed 3 to control 50% population (C 50 ) as well as and the dose needed to reduce by 50% the population dry mass (GR 50 ) (Gazziero et al., 2009). These variables enable the calculation of either the resistance factor (RF) (Hall et al., 1998) or the tolerance factor (TF).
The hypothesis is that, due to selection for recurrent use of glyphosate, biotypes collected at different fields present differential tolerance to the herbicide. This study aimed at evaluating, through dose-response curves, the levels of tolerance to glyphosate of B. latifolia and R. brasiliensis biotypes.

MATERIAL AND METHODS
Two experiments were conducted between the months of January and April 2014. The species employed were B. latifolia and R. brasiliensis. The biotypes used in each experiment and their respective collection sites are listed in Table 1. The biotypes were collected in soybean fields with a history of glyphosate herbicide use and high infestation. B. latifolia and R. brasiliensis susceptible biotypes were obtained from the company Cosmos Agrícola Ltda., located in São Paulo/ SP. R. brasiliensis biotype 271 was also considered susceptible for being collected at an area without history of use of glyphosate.
Concomitant with the collection of seeds of the species studied, questionnaires were applied to the owners of the areas so that to obtain information about the history of the management adopted by the farmers. After analysis of control levels, the biotypes highlighted in bold in the table were selected for presentation of the results since showed responses in contrast to the glyphosate ( Table 1).
The experiments were conducted in a completely randomized design with four replicates in a 6 x 6 factorial, in which the first factor represented the number of biotypes of each species (B. latifolia and R. brasiliensis) (Tabel 1) and the second factor by six doses of glyphosate (0, 74, 163, 360, 792 and 1,742 g a.e. ha -1 ).  The seeds of B. latifolia were submitted to break of dormancy by heating at 60 o C for 30 min and subsequent immersion in 2% potassium nitrate for 3 h. The seeds of R. brasiliensis, in turn, do not need to overcome dormancy. The seeds were placed to germinate in gearbox-type boxes with double layer of germination paper moistened with distilled water, allocated in BOD-type germination chamber at 25 o C and photoperiod of 12 h. Around 15 days after germination, two seedlings of each species were transplanted into polyethylene pots with a capacity of 5 dm³, containing Distroferric Red Latosol collected at an area free of Rubiaceae species propagules. Thinning was performed 20 days after the transplantation, and only one plant per pot remained.
When the plants had six to eight completely expanded leaves, the treatments were applied with CO 2 -pressurized backpack sprayer at 43 lbf pol -2 , with 1 m wide bar, 110.02-type fan nozzle tips, spaced 0.5 m apart , totaling a volume of 200 L ha -1 . Meteorological conditions in the beginning and end of the applications were as follows: air temperature (AT o C): 24 and 21.8; air relative humidity (RH %): 84 and 88.6.
At 14 and 28 days after application (DAA) of the treatments, visual control evaluations were conducted, based on the scale proposed by Frans et al. (1986), on which 0 represents null effect of herbicide symptoms on the plants and 100% represents their death. After the last evaluation (28 DAA), collection and weighing of the shoots of the plants were performed so that to obtain the mass of green shoots (GM). Subsequently, the plants were placed in an incubator at 60 o C until they reach constant mass, and the mass of dry shoots (DM) was obtained.
The visual control data were transformed into 100 -control, and GM and DM were expressed as percentage in relation to the control. The data obtained were submitted to the analysis of results variance by F test at 5% error probability through WinStat software (Machado and Conceição, 2005). The relation between quantitative factor levels and response variables was adjusted by non-linear regression, by means of SigmaPlot Version 10.0 software. The values of C 50 (dose required to achieve 50% control) and GR 50 (dose required to reduce DM by 50%) were calculated with help of the adjusted model. Thus, through the adjusted equations, C 80 or GR 80 were also calculated, using an Excel® spreadsheet.
The tolerance factor (TF) was calculated through the quotients of C 50 or GR 50 of each biotype by C 50 or GR 50 of the biotype most susceptible to the herbicide. This factor represents a comparative index of the biotype with greatest tolerance in relation to the biotype with lowest tolerance (Hall et al., 1998). There was also analysis of Pearson correlation between control data and DM data by means of Genes software (Cruz, 2006).

RESULTS AND DISCUSSION
All variables analyzed for B. latifolia and R. brasiliensis presented significance of biotypes factor x doses at 5% probability.

Response of B. latifolia biotypes
There was high response variability of B. latifolia biotypes to glyphosate ( Figure 1A and 1B). At 28 days after application (DAA), there was a considerable increase in the control of B. latifolia biotypes ( Figure 1B), compared to 14 DAA ( Figure 1A).
According as the herbicide doses were increased, it was possible to distinguish biotypes regarding tolerance. Both the evaluation performed at 14 days after application (DAA) and the one performed at 28 DAA presented superior control efficacy on biotypes São Paulo and Papanduva, that is, they showed greater susceptibility to glyphosate, while for biotypes 277, 287 and 283, the control efficacy was lower and, therefore, there was greater tolerance to the herbicide ( Figure 1A and B). The dose of 360 g ha -1 of glyphosate was strongly selective, both at 14 and 28 DAA, as it caused very large control differences between biotypes. For this dose, the control level for biotype São Paulo was around 100%, while the control levels for biotypes 283, 277 and 300 were between 15% and 26%.
The aforementioned results show variability of tolerance to glyphosate among B. latifolia biotypes collected at different locations of Paraná and North of Santa Catarina.   Analysis of the literature indicates that the control of B. latifolia is variable, depending mainly on the dose of glyphosate employed and on the development stage of the plants (Monquero et al., 2004;Lacerda;Victoria Filho, 2004;Cerdeira et al., 2010;Galon et al., 2013). A study developed by Lacerda and Victoria Filho (2004) evidenced need of doses superior to 540 g ha -1 to obtain control near to 90% for B. latifolia, evidencing that this species shows intermediate susceptibility to glyphosate doses. The use of 720 g ha -1 of glyphosate in seedlings of B. latifolia at one-three and four-six leaves development stage resulted in control levels of 100% and 81% at 35 DAA, respectively (Ramires et al. 2011), demonstrating the importance of the development stage to achieve good levels of control.
Not touching to C 50 values adjusted by SigmaPlot at 14 DAA (Table 2), the highest values were reached by biotypes 277, 283, 287 and 300, totaling 560.87; 540.45; 702.31 and 527.80 g ha -1 , respectively, while the biotype considered susceptible presented only 208.79 g ha -1 . In relation to C 80 , the same biotypes stood out; however, biotype 287 presented C 80 value higher than the highest dose tested in this study, differing from the others. Biotypes Papanduva and São Paulo presented the lowest values of C 50 and C 80 , in this order, with values of 153 and 350 g ha -1 for Papanduva and 208 and 269 g ha -1 for São Paulo, respectively.
Biotypes 277, 283, 287 and 300 showed the highest C 50 values adjusted by SigmaPlot at 28 DAA (Table 3), they being 403.73; 405.16; 267.09 and 384.80 g ha -1 , respectively, while the susceptible biotype showed C 50 of only 99.41 g ha -1 , much lower than the others. As shown by C 50 , the great tolerance of biotype 287 stands out, requiring 709 g ha -1 to reach 80% control, with need of a dose usually used in the fields (720 g ha -1 ). However, biotypes São Paulo and Papanduva required approximately 160 g ha -1 to reach 80% control. The tolerance factors (TF) varied between  1.0, for biotype São Paulo, and 4.1, for biotypes 277 and 283 (Table 3), indicating that these biotypes required a dose four times greater than the susceptible biotype (São Paulo) to achieve 50% control.
The results of GM and DM reinforce the data presented for visual control, evidencing high variability of response to glyphosate among the B. latifolia biotypes tested. It is possible to verify there is a high reduction of GM ( Figure 2A) and DM ( Figure 2B) with increased doses of glyphosate. Biotypes 277, 283 and 300 stood out due to the smallest reductions of DM, of 79%, 79% and 81%, in this order, with use of the highest dose of glyphosate. Nevertheless, it is verified that biotype São Paulo showed approximately 90% DM reduction with only 163 g ha -1 , making its greater susceptibility to the herbicide evident.
GR 80 values, estimated by Excel Program, varied between 115 and 1,033 g ha -1 (Table 4). Biotypes São Paulo and 287 were the most susceptible to the herbicide, presenting the lowest values of GR 50 and GR 80 . Biotypes 277, 300, 283 and Papanduva required doses greater than 800 g ha -1 to reduce by 80% DM (GR 80 ), superior to that usually used in commercial fields It is worth mentioning that biotype Papanduva, although presenting a low GR 50 (99.26 g ha -1 ), required 7 954 g ha -1 of glyphosate to reach GR 80 , i.e., a 900% dose increase was required to achieve this reduction.
The tolerance factors (TF) varied between 1.0, for biotype São Paulo and 287, and 4, for biotypes 277, 283 and 300 (Table 4), indicating these biotypes required a dose four times greater than the susceptible biotype (São Paulo) for a 50% DM reduction.
No biotype had TF inferior to 1, that is, the biotype that was considered susceptible, acquired from a company that sells weed seeds, was effectively the most susceptible to glyphosate among those studied.
The correlation between the TFs calculated by visual control at 28 DAA and DM was 0.88 (Table 3). The TFs calculated by means of DM were slightly inferior to those calculated by the control variable, but the TFs obtained by both variables allowed the identification of the most tolerant biotypes.
The correlations between the visual control and DM for B. latifolia biotypes in the evaluation at 28 DAA were high and significant (Table 3). The values of the correlation coefficients for control x DM presented significance at 1% and 5%, ranging from 0.87 to 0.97, with a general correlation coefficient of 0.91.

Response of R. brasiliensis biotypes
R. brasiliensis biotypes presented contrasting responses of control by glyphosate in the evaluation performed at 14 and 28 DAA ( Figures 3A and 3B). A more pronounced control was observed at 28 DAA ( Figure 3A), compared to 14 DAA ( Figure 3B), making it possible to classify the biotypes regarding tolerance to glyphosate. Biotype 283 had unsatisfactory control levels at 14 and 28 DAA ( Figure 3A and 3B), but biotypes São Paulo, 271 and 277 were satisfactorily controlled according as the doses were increased, evidencing their higher susceptibility to glyphosate.
It is important to emphasize that biotypes 271 and São Paulo were collected at areas with no glyphosate application history and, therefore, the higher tolerance in the other biotypes is probably associated with the selection pressure exerted by the herbicide.
The literature shows a great result variability of control of R. brasiliensis through glyphosate, because while many studies indicate the difficulty of controlling this weed with this herbicide, some indicate satisfactory levels of control at low doses. The greater tolerance presented by R. brasiliensis biotype 283 in this study is similar to that demonstrated in other studies with the same species, such as the study developed by Sharma and Singh (2001), where they observed that use of glyphosate alone at a dose of 577 g ha -1 for control of R. brasiliensis presented low efficiency, reaching a maximum of 14% (Sharma and Singh, 2001).
However, 99.5% control of this weed at 28 DAA with 540 g ha -1 of glyphosate, superior to other alternative herbicides tested, was verified by Vitorino et al. (2012). When studying the species 8 Richardia scabra, belonging to the same genus as R. brasiliensis, Reddy and Singh (1992) obtained control levels higher than 94%, with 375 and 750 g ha -1 of glyphosate.
The contrasting results may be related to differences between biotypes, and also to the development stage of the plant at the time of application. The application of glyphosate alone at the dose of 420 g ha -1 resulted in 60% control of R. brasiliensis plants at initial development stage at 7 DAA, and 90% at 14 DAA, but the dose of 2160 g ha -1 of glyphosate did not efficiently control this weed in adult phase (Monquero et al., 2004).
C 50 values adjusted by SigmaPlot at 14 DAA for biotypes Papanduva and 283 were greater than 1000 g ha -1 (Table 5). However, the C 50 value of the susceptible biotype was of only 219.7 g ha -1 .
In relation to C 80 , the same biotypes stood out, presenting C 80 value higher than the highest dose tested in this study. Biotypes 283 and Papanduva are included in the most tolerant group (C 80 above 1000). The susceptible biotype showed C 80 of 415 g ha -1 .
In the evaluation performed at 28 DAA, the control levels of the biotypes were increased, which resulted in C 50 values adjusted by SigmaPlot lower than in the evaluation at 14 DAA. Biotypes 283, 285 and Papanduva showed the highest C 50 values, totaling 735.56; 468.71 and 429.80 g ha -1 , respectively, while the susceptible biotype showed C 50 of 90.56 g ha -1 (Table 6). Biotype 283 presented C 80 higher than the highest dose used in the experiment (1742 g ha -1 ), while biotype 271, considered susceptible, required only 153 g ha -1 to cause the same effect ( Table 6).  Vertical bars represent the standard error of the mean of each treatment.

9
There was an expressive reduction of DM with increased glyphosate doses (Figure 4), reinforcing the data presented for visual control, and it was also possible to demonstrate a high variability of glyphosate response among the R. brasiliensis biotypes evaluated.
Biotype Papanduva showed a GR 80 superior to the highest dose used, that is, it was not able to reduce 80% biotype DM. For comparison purposes, biotype 271, considered susceptible, presented GR 80 of only 87 g ha -1 .
In R. brasiliensis species, the correlation coefficients between the visual control and DM for the biotypes in the evaluation at 28 DAA were smaller, and only three were significant (Table 6), varying between 0.49 and 0.92. The general correlation, considering all biotypes, was 0.75.
The aforementioned results evidenced the high variability of response to glyphosate between B. latifolia and R. brasiliensis biotypes. This variability can be attributed to the glyphosate selection process to which weed populations are submitted. From the late 1990s, the intensification of glyphosate use in soybean fields in South of Brazil increased with the widespread and rapid Vertical bars represent the standard error of the mean of each treatment.  adoption by farmers of genetically modified crops (GMC), which are resistant to glyphosate. The selection of weeds with resistance and/or tolerance to herbicide is considered the process of highest risky in the adoption of GMC in Brazil and South America (Cerdeira et al., 2010).
Since the introduction of genetically modified cultivars resistant to glyphosate in soybean cultivation, there has been selection of species tolerant to this herbicide, as day flower (Commelina spp.), morning glory (Ipomoea spp.), white-eye (R. brasiliensis), and broadleaf buttonweed (B. latifolia), among others. This phenomenon has occurred in other parts of the world, in species as Ambrosia artemisiifolia and Sesbania exaltata, for instance. In Brazil, there is also a great increase in the area infested with glyphosate-resistant biotypes, mainly in species of Conyza and Digitaria insularis (Correia et al., 2010;Lamego et al., 2013).
Weeds are biological organisms evolving in response to environmental changes (disturbance and stress), which determine changes in species composition in a given area (Radosevich et al., 2007). In addition, the recurrent use of the same herbicide or herbicides with the same mechanism of action and spectrum of weed control for several years causes changes in the infesting flora, since there is intensification of the process of selection of resistant weeds, tolerant to the sorting agent (Monquero et al., 2004).
In the literature, the distinctive responses among different studies evaluating the response of B. latifolia and R. brasiliensis can be attributed to several factors. Firstly, the genetic variability inherent to the biotypes. Secondly, the development conditions of the plants before and after the application of the herbicide, such as solar radiation, temperature, water availability, and relative humidity. Thirdly, the physiological stage of weeds at the time of application, as younger plants present lower defense to the action of the herbicides (Chauhan and Abugho, 2012). Dias et al. (2013) verified that day flower plants become four times more tolerant to glyphosate every ten phenological development units on the Biologische Bundesanstalt, Bundessortenamt and Chemical industry (BBCH) scale, adapted from Hess et al., 1997. In other words, chemical control measures at early stages of day flower growth reduce the dose of glyphosate and, consequently, its cost of application, with decrease in the selection pressure on biotypes tolerant to this herbicide, and reduction of damages to the agroecosystem.
Application of glyphosate with a maximum of three pairs of leaves resulted in a control level of 87.5% with 810 g ha -1 in Borreria densiflora, of the same genus of B. latifolia, and preliminary observations of post-emergence applications showed that the weed becomes tolerant when it presents the phenology of 4-5 pairs of leaves, thus causing the failure of control by herbicides (Martins and Christoffoleti, 2014). Takano et al. (2013) observed that glyphosate alone resulted in 99.5% control of plants at the phenological stage above 10 leaves in B. latifolia, whereas R. brasiliensis presented only 82.5% control for the same conditions mentioned above.
There are at least six general mechanisms that may explain weed tolerance to herbicides: reduced absorption and translocation; increased metabolism of the herbicide for substances with less phytotoxic activity; compartmentalization of the herbicide molecule; lack of herbicide affinity with the specific site of action, and over expression of the target enzyme (Galon et al., With the results of this study, it was possible to prove that R. brasiliensis biotype 271, collected at an area with no glyphosate application history, was the most susceptible among the biotypes studied, contributing to the affirmation that weed cropping and plant management systems in fields have influence on the selection of more tolerant biotypes. Increasing the selection pressure of the herbicide on a population promotes survival of the most tolerant individuals and leads to the mortality of the most susceptible, which increases the frequency of tolerance genes in the surviving population, leading to the gradual increase in C 50 value, similarly to what occurs with populations in which the herbicide resistance character is dominated by multiple genes (Neve et al., 2003). Thus, it is believed that the populations with the highest C 50 values, detected in this experiment, are located at a more advanced stage of the glyphosate tolerance evolution process.
In recent times, the incidence of tolerant species and resistant populations in Brazil assumed quite large proportions, and levels of tolerance to worrisome herbicides have already been reached by several weeds throughout the country (Galon et al., 2013). Many pieces of research in the literature have studied the mechanisms of resistance to glyphosate in weed populations; however, little focus is given on the mechanisms of tolerance to this herbicide in weeds. It is very important to investigate the differences between biotypes of tolerant species collected in different situations of selection pressure, thus allowing the identification of changes in populations after increasing the selection pressure imposed by the herbicides.
The existence of populations with greater tolerance indicates need to conduct joint actions, whether by the farmers, through conscious use of herbicides, or by the companies, by providing a greater diversity of herbicides with differentiated mechanisms of action, as well as the interest of the academic community and research bodies in deepening the knowledge of biology and physiological mechanisms that confer this character to the weeds present in the cropping systems.
In general, variability of glyphosate herbicide response was observed among B. latifolia and R. brasiliensis biotypes collected at different locations in Paraná and Santa Catarina. Some biotypes of both species studied were not controlled with doses of glyphosate above those usually used in the crops, evidencing selection by recurrent use of glyphosate. The results showed a positive relation between the difficulty of controlling some biotypes of both species studied and the period of use of glyphosate in agricultural areas.