**3. Results and discussion**

Resistance to neurotoxic insecticides varied generally among the different populations of *L. coffeella* in Brazil. RT50 varied from 1.02 to 522. Low resistance levels were observed for chlorantraniliprole insecticides (1.02-3.23 times), abamectin (1.19-4.80 times), and deltamethrin (1.05-5.35 times).

On the other hand, intermediate resistance was observed for thiamethoxam (1.11-10.61 times) and chlorpyrifos (4.53-18.63 times), while resistance was high for profenofos (65.3-522 times) (Table 1). Higher levels of organophosphate resistance were observed in Minas Gerais (Abaeté dos Mendes, Rio Paranaíba and Carmo do Paranaíba), Pernambuco (Guaranhuns), and São Paulo (Franca).

The RT50 values are supported by the LT50 values, which were variable among populations and insecticides. The population from Carmo do Paranaíba-MG was noteworthy as it took 89.93 h for 50% of the population to die after contact with the insecticide thiamethoxam. The organo‐ phosphate and pyrethroid insecticides had lower lethal times. Chlorantraniliprole showed lower LT50 of 8.59 h.

Two canonical axes were significant among the five canonical axes identified, showing linear associations between LT50 of the insecticides with the geographical regions of the population origins of *L. coffeella*, which showed that the four canonical axes were significant, with the first three axes explaining 90% of the total variance data (Table 1 and Figure 2). The highest absolute values of the canonical coefficients show which insecticides most contributed to the standard deviation of resistance among the different localities. For the first canonical axis of greater importance in the analysis, the insecticides chlorpyrifos, profenofos, and deltamethrin showed positive correlations and higher values of coefficients and thus higher contributions to the differences between the resistant populations (Table 2). Profenofos and deltamethrin, with a positive relationship, contributed to the pattern of divergence on the second axis.

The opposite relationship was observed for assistance with the chlorpyrifos insecticide on the third and fourth axes. On the fifth and sixth axes, a positive relationship was observed between the profenofos insecticide and the standard deviation. It is important to highlight that the new insecticide chlorantraniliprole did not contribute to the resistance of populations (Table 2). Graphs of this analysis done with the first two axes explained 92% of the total variance of the data to show the grouping between locations (Table 2 and Figure 2).

The weight of organophosphate (profenofos and chlorpyrifos) and pyrethroid (deltamethrin) insecticides on the first two axes enhanced the resistance process since they are among the main groups with examples of insect resistance (quotation). Two grouping patterns were observed, with one group for the populations of *L. coffeella* Rio Parnaíba-MG, Carmo do Paranaíba-MG, and Abaeté dos Mendes-MG and a second group for the populations of Santa Teresa-ES and Guaraciaba-MG, but these patterns did not occur in the other populations (Figure 2).


a *df*x = degrees of freedom (numerator/denominator).

b *R*2 *x*b = canonical correlation square.

separately into Petri dishes (9.0 × 1.5 cm). Ten larvae of *L. coffeella* reared in the lab were transferred to each Petri dish using a fine-tipped brush. The Petri dishes with the larvae were kept in the BOD incubator (model SP-500) at 25°C ± 1°C until the time of evaluation. The experiments were conducted in a completely randomized design with four replications.

Preliminary tests using only discs soaked in water were carried out to observe caterpillar mortality over a 48-h period. This was necessary to estimate the maximum evaluation time after bioassay assembly that causes 20% lower mortality in the control [18]. Thus, to have a mortality range from 0% to 100%, evaluations were made at 2, 6, 12, 16, 24, 32 and 48 h (treatments) after bioassay assembly. The time intervals were assessed in independent experimental units, to avoid pseudoreplicates. We considered insects dead when they did not

To determine the spatial dependence of *L. coffeella* insecticide resistance, the semivariance statistical model of LT50 values to *L. coffeella* populations for each insecticide and the distance between sampling locations of each population were used. The distance between the sampling sites of each insect population was determined using geographic coordinates with a global positioning system (GPS 12, Garmin International, Olathe, KS). The semivariograms were estimated from the semivariance data of the LTs50 of each population for each insecticide and used as dependent variables in regression analysis, with the distance between the sampling sites as an independent variable. The first inflection point of the semivariogram curve represents the maximum distance of interference between the populations of *L. coffeella* in

Resistance to neurotoxic insecticides varied generally among the different populations of *L. coffeella* in Brazil. RT50 varied from 1.02 to 522. Low resistance levels were observed for chlorantraniliprole insecticides (1.02-3.23 times), abamectin (1.19-4.80 times), and deltamethrin

On the other hand, intermediate resistance was observed for thiamethoxam (1.11-10.61 times) and chlorpyrifos (4.53-18.63 times), while resistance was high for profenofos (65.3-522 times) (Table 1). Higher levels of organophosphate resistance were observed in Minas Gerais (Abaeté dos Mendes, Rio Paranaíba and Carmo do Paranaíba), Pernambuco (Guaranhuns), and São

The RT50 values are supported by the LT50 values, which were variable among populations and insecticides. The population from Carmo do Paranaíba-MG was noteworthy as it took 89.93 h for 50% of the population to die after contact with the insecticide thiamethoxam. The organo‐ phosphate and pyrethroid insecticides had lower lethal times. Chlorantraniliprole showed

move after being touched with the fine-tipped brush.

**2.4. Spatial dependence of insecticide resistance**

relation to susceptibility to a given insecticide.

**3. Results and discussion**

(1.05-5.35 times).

8 Insecticides Resistance

Paulo (Franca).

lower LT50 of 8.59 h.

**Table 2.** Canonical axes and coefficients (grouped in the canonical structure) of mortalities of *Leucoptera coffeella* caused by six neurotoxic insecticides.

**Figure 2.** Ordination diagram showing discrimination of insecticide resistance between Brazilian populations of *Leu‐ coptera coffeella*. Spherical gray symbols are centroids of treatments and represent the average of canonical variable classes. Large circles indicates treatment groups with no significant difference between them (approximate *F* test, *P* < 0. 05), based on the Mahalanobis distance (D2) between averages.

The semivariogram models related to the LT50 values of *L. coffeella* with the distance between the sampling sites obtained for only two insecticides, the organophosphates chlorpyrifos and the pirimiphos. The first inflection points for the models were lengths of 169 and 1,956 km for the insecticides chlorpyrifos and pirimiphos (Figure 3). Therefore, these were the maximum distances between the interference resistance levels of the *L. coffeella* sampling sites.

Our study reported high variations in the resistance ratio (RT50) of the organophosphates profenofos (522 times) and chlorpyrifos (19 times) compared to the susceptible population of *L. coffeella*. This large variation represented by RT50 indicates that populations show differences in susceptibility and greater or lesser sensitivity to the enzyme acetylcholinesterase since variations were observed between populations that died the fastest and those that died more slowly.

This shows that this group of insecticides is extremely important in managing resistance because of its intense use, with this group being highly toxic and presenting higher neurotoxic action [19]. Many studies on resistance to the organophosphate insecticide group showed high variation in the mortality of the resistant population compared to other lepidopteran popula‐ tions [20,21]. Extensive insecticide use in coffee crops and high death speed are among the main factors of resistance [22]. Fragoso et al. [13] observed up to 22 applications of organo‐ phosphate insecticides, detecting high levels of resistance when larvae were kept exposed to the discriminating concentration. These concentrations were higher than those tested for profenofos and chlorpyrifos in our study.

**Figure 2.** Ordination diagram showing discrimination of insecticide resistance between Brazilian populations of *Leu‐ coptera coffeella*. Spherical gray symbols are centroids of treatments and represent the average of canonical variable classes. Large circles indicates treatment groups with no significant difference between them (approximate *F* test, *P* < 0.

The semivariogram models related to the LT50 values of *L. coffeella* with the distance between the sampling sites obtained for only two insecticides, the organophosphates chlorpyrifos and the pirimiphos. The first inflection points for the models were lengths of 169 and 1,956 km for the insecticides chlorpyrifos and pirimiphos (Figure 3). Therefore, these were the maximum

Our study reported high variations in the resistance ratio (RT50) of the organophosphates profenofos (522 times) and chlorpyrifos (19 times) compared to the susceptible population of *L. coffeella*. This large variation represented by RT50 indicates that populations show differences in susceptibility and greater or lesser sensitivity to the enzyme acetylcholinesterase since variations were observed between populations that died the fastest and those that died more

This shows that this group of insecticides is extremely important in managing resistance because of its intense use, with this group being highly toxic and presenting higher neurotoxic action [19]. Many studies on resistance to the organophosphate insecticide group showed high variation in the mortality of the resistant population compared to other lepidopteran popula‐ tions [20,21]. Extensive insecticide use in coffee crops and high death speed are among the main factors of resistance [22]. Fragoso et al. [13] observed up to 22 applications of organo‐ phosphate insecticides, detecting high levels of resistance when larvae were kept exposed to the discriminating concentration. These concentrations were higher than those tested for

distances between the interference resistance levels of the *L. coffeella* sampling sites.

05), based on the Mahalanobis distance (D2) between averages.

profenofos and chlorpyrifos in our study.

slowly.

10 Insecticides Resistance

**Figure 3.** Semivariogram of the LT50 of chlorpyrifos, profenofos, and deltamethrin according to the distance between sampled points from populations of *Leucoptera coffeella*. The first inflection point of the semivariogram curve, repre‐ sented by a down ward-pointing arrow, represents the maximum distance of interference of the resistance to the insec‐ ticides.

On the other hand, chlorantraniliprole, abamectin, and deltamethrin insecticides showed low levels of RT50 variation. The result with the chlorantraniliprole insecticide was as expected since this insecticide has only recently been commercialized [23-25] and has a highly efficient molecule since low doses of this insecticide (31.5 g a.i. ha-1) cause high mortality to *L. coffeella*; moreover, it is selective for wasps [26].

Selectivity is an important factor in managing resistance in pest insects [27]. Many studies with basic lines of susceptibility have been done with chlorantraniliprole insecticide and Lepidop‐ tera, and the observations are that populations show susceptibility with low variation in mortality [28,29]. The insecticide abamectin is not considered old and has been effective in controlling this pest insect, with no flaws detected in its control of *L. coffeella* as of yet. Despite the abamectin insecticide not being among those at risk of resistance in *L. coffeella*, this insecticide has not been studied. However, many arthropod pests have been classified as being at risk for resistance to this group. Among them are *Leptinotarsa decemlineata* (Say) [30], *Musca domestica* [31], *P. xylostella* [32], *Frankliniella occidentalis* [33], and *Tetranychus urticae* [34]. Abamectin resistance has been observed in populations of *F. occidentalis* [35] and *Liriomyza trifolii* [36]. Deltamethrin had surprising results, with low discrepancy between the resistant and the susceptible populations (5 times) compared to their insecticides such as thiamethoxam (10 times) that are less used in coffee plantation. In recent years, however, the number of pyrethroid applications in coffee production has been greatly reduced. Despite the low resistance to pyrethroids, however, the variation has been observed in Brazil for the moth *P. xylostella* [37] as well as with other pyrethroids (cypermethrin, β-cypermethrin, deltamethrin, and esfenvalerate) in Pakistan, India, China, and Korea [38,39]. Although deltamethrin has affected fewer Brazilian populations of *L. coffeella*, a difference of 5 times is cause for concern since it should have been more effective.

Insects usually have a resistance mechanism that confers nerve insensitivity, known as knockdown resistance (Kdr), as first reported in *M. domestica* (L.) (Diptera: Muscidae) [40]. This type of resistance is found in other agricultural pests based on patterns of cross-resistance and the absence of compound synergism that inhibits the activity of cytochrome P450 and esterase enzymes [41].

The insecticide thiamethoxam has been frequently used and can be applied as a spray or via the soil [42]. There are no studies of lepidopteran resistance to this insecticide. Control failures were observed depending on the time of application, however, for example [43] observed effectiveness of 4.1%, 50.6%, 62.1%, and 69.0%.

The grouping of populations from Rio Paranaíba, Carmo do Paranaíba, and Abaeté (Group I) and Santa Teresa with Guaraciaba (Group II), coupled with the significant response of the effect of distance on the LT50 of the chlorpyrifos, profenofos, and deltamethrin insecticides, showed that resistance was affected by the collection distance of these populations since more closely connected populations had similar resistance responses.

Studies have shown a strong relationship between collection distance and resistance patterns [44,10,45,12]. All of these studies showed significant association of resistance with distance, and nearby populations tended to show more similar responses, as is the case for *P. xylostel‐ la* (L.) (Lepidoptera: Plutellidae). Chen et al. [46] studied the resistance of pyrethroids to *Culex pipiens* (Diptera: Culicidae) and found different frequencies of resistance at different locations, ranging from 21.4% to 79.8%. Moreover, this type of response may be associated with the large dispersal capacity of adult *L. coffeella* and the sampling characteristics.

Adults of *L. coffeella* disperse easily between coffee crops and have different densities in different environments [47-49]. Moreover, there is a geographic corridor between the largestproducing Brazilian states (Figure 1). Isaaks and Srivastava [50] also found that in order to detect differences among geostatistical studies of spatial distribution, it was necessary to collect both near and distant samples.

We conclude that Brazilian populations of *L. coffeella* showed greater resistance to organo‐ phosphates. Furthermore, resistance may be associated with the distance between the pro‐ ducing regions, and local selection favored by dispersal seem important for insecticide resistance evolution among Brazilian populations of *L. coffeella* and should be considered in designing pest management programs. The insecticides that do not show mortality to *L. coffeella* should be sprayed in such conditions, and a higher variety of insecticides (out of the cross-resistance and multiple-resistance spectra) should be used in rotation to reduce the danger of evolution of resistance.
