Resistance status of lepidopteran soybean pests following large-scale use of MON 87701 × MON 89788 soybean in Brazil

Widespread adoption of MON 87701 × MON 89788 soybean, expressing Cry1Ac Bt protein and glyphosate tolerance, has been observed in Brazil. A proactive program was implemented to phenotypically and genotypically monitor Cry1Ac resistance in Chrysodeixis includens (Walker). Recent cases of unexpected injury in MON 87701 × MON 89788 soybean were investigated and a large-scale sampling of larvae on commercial soybean fields was performed to assess the efficacy of this technology and the distribution of lepidopteran pests in Brazil. No significant shift in C. includens susceptibility to Cry1Ac was observed eight years after commercial introduction of this technology in Brazil. F2 screen results confirmed that the frequency of Cry1Ac resistance alleles remains low and stable in C. includens. Unexpected injury caused by Rachiplusia nu (Guenée) and Crocidosema aporema (Walsingham) in MON 87701 × MON 89788 soybean was detected during the 2020/21 season, and studies confirmed a genetically based alteration in their susceptibility to Cry1Ac. MON 87701 × MON 89788 soybean remains effective against Anticarsia gemmatalis (Hübner), C. includens, Chloridea virescents (Fabricius) and Helicoverpa armigera (Hübner) in Brazil. However, there is evidence of field-evolved resistance to MON 87701 × MON 89788 soybean by the secondary soybean pests R. nu and C. aporema.


Widespread adoption of MON 87701 × MON 89788 soybean, expressing Cry1Ac Bt protein and glyphosate tolerance, has been observed in Brazil. A proactive program was implemented to phenotypically and genotypically monitor Cry1Ac resistance in Chrysodeixis includens (Walker). Recent cases of unexpected injury in MON 87701 × MON 89788 soybean were investigated and a large-scale sampling of larvae on commercial soybean fields was performed to assess the efficacy of this technology and the distribution of lepidopteran pests in Brazil. No significant shift in C. includens susceptibility to Cry1Ac was observed eight years after commercial introduction of this technology in Brazil. F 2 screen results confirmed that the frequency of Cry1Ac resistance alleles remains low and stable in C. includens. Unexpected injury caused by Rachiplusia nu (Guenée) and
, and tolerance to glyphosate is conferred by the event MON 89788, which expresses CP4 EPSPS. Brazil was the first country to adopt MON 87701 × MON 89788 soybean in 2013/14. Eight years after its first commercial cultivation, more than 30 million hectares have been cultivated with this technology in Brazil (80% of the country's soybean acreage) in the 2020/21 crop season 3 .
The development of resistance in populations of the target pests to transgenic plants expressing Bt insecticidal proteins has been the main threat to the sustainability of these technologies 8,9 . The refuge strategy has been recommended as the major insect resistance management (IRM) strategy for delaying resistance in target pest populations of MON 87701 × MON 89788 soybean 1,4,6,10 . The assumption is that a Bt plant expressing www.nature.com/scientificreports/ a high dose of the Bt protein would control most or all of the individuals heterozygous for resistance alleles in natural populations, making resistance functionally recessive. When a sufficient refuge of non-Bt plants is available near a Bt field, the rare homozygous resistant insects that survive on Bt plants mate with the relatively abundant homozygous susceptible insects in the refuge. As a result, the heterozygous offspring produced would not survive on the high-dose Bt crop, substantially delaying the evolution of resistance 11 . The technical refuge recommendation for MON 87701 × MON 89788 soybean in Brazil is planting at least 20% of the soybean acreage as a structured refuge formed by non-Bt soybean and located within 800 m of the Bt crop area 12 . However, low compliance with structured refuge recommendations seems to be a common theme across most of the cases of insect resistance to Bt crops documented globally [13][14][15][16][17][18][19][20][21] . This highlights the importance of programs to monitor the development of insect resistance to MON 87701 × MON 89788 soybean and assess the adoption of refuge planting according to the technical recommendations for this technology in Brazil.
In addition to protecting against the insect species already mentioned, MON 87701 × MON 89788 soybean also offers protection against larval injury caused by Rachiplusia nu (Guenée, 1852) (Lepidoptera: Noctuidae) and Crocidosema aporema (Walsingham, 1914) (Lepidoptera: Tortricidae) 1 . Rachiplusia nu and C. aporema are historically considered key soy lepidopteran pests in Argentina [22][23][24] ; in Brazil, however, the occurrence of these species in soybean fields has historically been restricted to reasonably low levels in the mid-south region of the country [25][26][27] . Cases of unexpected injury to the MON 87701 × MON 89788 soybean caused by "loopers" were identified during the 2020/21 crop season. We investigated these cases and here we report that the insects causing unexpected injury were R. nu and C. aporema.
The goals of this paper are to (a) report data from the proactive resistance monitoring of C. includens to MON 87701 × MON 89788 soybean in Brazil; (b) characterize the level of Cry1Ac resistance in R. nu and C. aporema samples collected during the 2020/21 season; and (c) assess the efficacy of the MON 87701 × MON 89788 technology and abundance of key lepidopteran pests in the 2020/21 season.

Methods
Proactive MON 87701 × MON 89788 soybean resistance monitoring of C. includens. Insect samples and permits. To proactively monitor resistance to Cry1Ac protein expressed in Bt soybean, C. includens larvae were sampled from commercial plantings of non-Bt soybeans in distinct geographic regions of Brazil and sent to a laboratory in our monitoring network, where they were reared on artificial diet 28 to obtain F 1 and F 2 generation larvae for use in bioassays. All C. includens field populations sampled in Brazil are listed in Tables 1, 2, 3 of the Supporting Information (SI). All insect collections in Brazil were done in accordance with the approval granted by the System of Authorization and Information on Biodiversity (SISBIO) of the Ministry of Environment to a contracted company responsible for the field sampling (PROMIP, Permit for scientific purpose activity: 61826 and 61824).
The research did not involve the collection of plant material in nature. All plants used in the study were grown from commercially available seeds. The study complies with relevant institutional, national, and international guidelines and legislation.
Phenotypic resistance monitoring of C. includens using a diagnostic concentration of Cry1Ac protein. Dietincorporated bioassays were performed with a Cry1Ac formulated product (MVP II, Pseudomonas encapsulated Cry1Ac from Dow Chemicals, San Diego, CA, containing 11.14% of active Cry1Ac protein). As described in Yano et al. 10 , 5.6 µg of Cry1Ac protein/mL of diet was used as the diagnostic concentration for resistance monitoring in C. includens populations (SI Table 1). For the bioassays, the Cry1Ac protein was diluted in distilled water to 56 µg/mL and 4 mL of this solution was poured into a 50-mL Falcon tube. The tube was then filled with 36 mL of artificial diet 28 . The mixture was homogenized in a vortex mixer for ∼ 40 s. Then, 1 mL of diet containing Cry1Ac protein was poured into each well of a 128-well bioassay tray (BIO-BA-128; CD International Inc., Pitman, NJ). After the diet had dried, one neonate larva (< 24 h old) from the F 1 generation was placed into each well with a fine brush. Trays were sealed with a plastic adhesive (BIO-CV-16; CD International Inc.) that allowed air exchange and kept in a climatic chamber at 25 ± 2 °C, 70 ± 10% RH and 14:10 h (L:D) photoperiod. Mortality was recorded at 7 days. A total of 1,024 neonates/population were tested. Mortality data for each population were plotted on a time scale from 2015/16 (2016) to 2020/21 (2021). Data from 2009/10-2014/15 were previously reported in Yano et al. 10 and plotted herein.
Genotypic Cry1Ac resistance monitoring of C. includens using F 2 screens. To estimate the Cry1Ac resistance allele frequency in Brazilian populations of C. includens, we utilized the F 2 screen method proposed by Andow and Alstad 29 . From 2016/17 (2017) to 2020/21 (2021), a total of 74 populations of C. includens were screened (SI Table 2). The collected larvae were transported to a laboratory and kept on artificial diet 28 . After pupation, pupae were separated by sex and used to establish multiple single-pair mating couples under laboratory conditions. The offspring (F 1 progeny) of each single-pair mating (isoline) were reared in artificial diet, and pupae were transferred to a PVC cylindrical cage (20 cm height × 10 cm diameter) lined with paper (oviposition substrate) and covered with a mesh fabric until adult emergence. Adults were fed with a 10% honey solution provided on cotton inside a plastic cup. Eggs were collected and kept in plastic cups with filter paper moistened with water. F 2 generation neonates (≤ 24 h old) were then used for screening. For the bioassays, leaves from the upper third of greenhouse-grown soybean plants were collected and kept in a refrigerator until use. Each isoline was tested in 128-well bioassay trays (BIO-BA-128; CD International Inc.) containing a 2% agar solution and one soybean leaf disc of 1.7 cm diameter. Two neonate larvae (< 24 h old) were placed in each well with a fine brush, with a target number of 128 neonates tested/isoline in 2016/17 (2017) and 256 neonates tested/isoline for 2017/18 to 2020/21 (2018-2021). Then, plates were sealed and placed under the same environmental conditions described above.   Table 3). A susceptible R. nu laboratory colony was not available, so a susceptible laboratory population of C. includens was used as a standard for comparison of tolerance levels between the two species. The purified Cry1Ac insecticidal protein was produced by Bayer Crop Science US (Chesterfield, MO, USA). Proteins were isolated from fermentation broths of recombinant B. thuringiensis strains transformed to express individual toxins, like described in Chen et al. 35 Bioassays were performed in 96-well bioassay trays with 200 μL of artificial diet per well. Seven concentrations of Cry1Ac were prepared by dilution in TX buffer (0.005% Triton X-100, 10 mM Tris-HCl, pH 7.4). The control treatment was composed of TX buffer. After preparation, 20-µl protein samples were overlaid on the diet surface of each well and ventilated until the excess moisture dissipated. After drying, each well was infested with a single neonate larva (< 24 h old) using a fine brush. Mortality was recorded at 6 days. Lethal concentrations (LC 50 ) and 95% confidence intervals were estimated using probit analysis in SAS 9.1 36 . Leaf-disc bioassays were also performed with C. includens and R. nu field populations sampled from Brazil and Argentina (SI Table 3). Bioassays were conducted in the respective country of sampling with neonates from the F 1 to F 2 generations in Argentina and from the F 1 generation in Brazil. The MON 87701 × MON 89788 soybean and non-Bt soybean were grown in a greenhouse. Seeds were sown on the ground within a greenhouse in Brazil, whereas in Argentina they were cultivated in 5 L plastic pots with 1 plant/pot. Nutrients and water were provided according to the necessity during plant development. Bioassays were performed with leaves of V6-to R3-growth-stage plants 37 following the leaf-disc bioassay method previously described, but with the difference that a single neonate (< 24 h old) was placed in each well. A total of 128 neonates was tested for each population/ treatment combination. Bioassay were carried out with 8 replicates of 16 larvae in Brazil and 4 replicates of 32 larvae in Argentina. Mortality was recorded at 4 days counting the number of dead larvae in each replicate. The percent mortality on MON 87701 × MON 89788 soybean and non-Bt soybean. The total number of tested and dead insects on MON 87701 × MON 89788 soybean and non-Bt soybean were used to estimate the 95% confidence interval (CI) for the probability of mortality, according to binomial distribution. The statistical analysis of the data from these bioassays was made using the function binom.probit from the package binom in R statistical software-R version 4.0.2 31 . Percent mortality on MON 87701 × MON 89788 soybean and non-Bt soybean were considered significantly different when their 95% CI did not overlap.
To determine whether the same resistance allele was present in populations of R. nu sampled from distinct locations in Brazil, we also ran a complementation test for allelism. For this, the field population of R. nu from Paranapanema sampled in 2020 was crossed with four other field populations (Uberaba 2020, Taquarituba 2020, Taquarituba 2021, Perdizes 2021) (SI Table 3). At least 15 pairs were used for each crossing. The F 1 progeny were tested using leaf-disc bioassays as previously described, with 12 replicates of 16 larvae/cross, totaling 192 neonates tested on MON 87701 × MON 89788 soybean and on non-Bt soybean. Mortality was recorded after 4 days, and data were compared as earlier described.  Table 3). Neonates from these populations (F 1 generation) were used in leaf-disc bioassays of MON 87701 × MON 89788 soybean and non-Bt soybean as described in the previous section. Because only a limited number of eggs were available from the mass mating of field insects, only 13 neonates from the Itararé population and 32 neonates from the Cristalina, Perdizes and Tibagi populations were tested in each treatment. Mortality was recorded at  Tubes containing larvae were then sent to the laboratory and kept in a freezer (− 20 °C) until identification. Identification of lepidopteran larvae were based on Herzog 32 , Sosa-Gómez et al. 38 , Navarro et al. 33 and Gilligan and Passoa 34 . Sampling fields were grouped according to the location within Embrapa's (Embrapa Soybean) soybean variety regionalization named "edaphoclimatic regions" and "soybean macroregions" 39 . Statistical modeling was used to characterize soybean geographic variation in pest abundance in non-Bt soybean and MON 87701 × MON 89788 soybean. Specifically, random effects for edaphoclimatic regions were estimated separately for each pest and field type (non-Bt soybean and MON 87701 × MON 89788 soybean) using a linear mixed-effects model for larval count data with Poisson link, with edaphoclimatic region and soybean macroregion treated as random effects, and edaphoclimatic region nested within soybean macroregion. An advantage of random effects estimates, compared to fixed effects estimates or simply using average counts by region, is that they properly account and adjust for sample size differences between regions, providing more accurate predictions of true pest density. The random effects estimates of pest abundance by edaphoclimatic region were summarized using choropleth maps. Analyses were performed with R statistical software-R version 4.0.2 31 using the glmer function in R package "lmer".

Proactive MON 87701 × MON 89788 soybean resistance monitoring of C. includens. Phenotypic resistance monitoring in C. includens using a diagnostic concentration of Cry1Ac protein.
There was high mortality in all field populations of C. includens collected from 2015/16 to 2020/21 (42 field populations) when exposed to the Cry1Ac diagnostic concentration of 5.6 µg of Cry1Ac protein per mL of diet (Fig. 1). The neonate mortality in this period ranged from 96 to 100%, not differing over time. Compared with data from 2009/10-2014/15, there was also no obvious change in the susceptibility to Cry1Ac protein in populations of C. includens from Brazil over the evaluated years (Fig. 1).  Table 2). The estimated ratio of tolerance to Cry1Ac protein was > 2,709 for the R. nu population tested. Significant differences in susceptibility to Cry1Ac protein between species were also verified by the equality (χ 2 = 230.0; df = 2; P < 0.001) and parallelism (χ 2 = 130.3; df = 1; P < 0.001) tests, which indicated that the mortality curves had distinct parameters (intercepts and slopes). www.nature.com/scientificreports/ When neonates of the R. nu population tested above and four other populations obtained in MON 87701 × MON 89788 soybean fields in Brazil were exposed to MON 87701 × MON 89788 soybean leaf discs, there was no higher than 9.4% mortality, not differing significantly from those verified on non-Bt soybean controls (Fig. 3A). In contrast, R. nu populations from Argentina presented 100% mortality on MON 87701 × MON 89788 soybean leaf discs and no higher than 12.5% mortality on non-Bt soybean. Similar results to those for R. nu from Argentina were obtained for C. includens populations from Brazil and Argentina exposed to MON 87701 × MON 89788 soybean (> 99.2% mortality) and non-Bt soybean (< 15% mortality) (Fig. 3B).
In the complementation study, there were no significant differences in the mortality of progeny from four crosses involving five populations of R. nu from distinct MON 87701 × MON 89788 soybean fields in Brazil when exposed to MON 87701 × MON 89788 soybean (3.6 to 10.9%) and non-Bt soybean (2.6 to 7.8%) ( Table 3). These results provide evidence against the hypothesis that the resistance alleles in any of the populations were at different loci and therefore suggest that the resistance alleles are probably at the same locus.
Overall, our results suggest a genetically based decrease in susceptibility to Cry1Ac protein expressed in soybean in the Brazilian populations of R. nu tested.

Abundance of lepidopteran pests on commercial fields of MON 87701 × MON 89788 soybean and non-Bt soybean in Brazil.
A random effects statistical model was used to estimate pest abundance for each edaphoclimatic region. Separate analyses were conducted for each pest and field type. Pest abundance estimates were summarized using choropleth maps, where each edaphoclimatic region is color-coded according to its estimated pest abundance (different colors are assigned for estimates in the ranges < 0.10, 0.1-0.25, 0.25-0.5, 0.5-1, 1-2.5, 2.5-5, 5-10, and > 10 larvae/10 m). In all maps that show more than one color, there is statistically significant variation across edaphoclimatic regions (P < 0.05) (Fig. 5).
The visualization of geographic variation in lepidopteran pest abundance in MON 87701 × MON 89788 soybean and non-Bt soybean fields shows that C. includens was present at high density in 301 of the 395 sampling locations in non-Bt soybean (76.2% of fields) but nearly absent (1.01%) from MON 87701 × MON 89788 soybean fields (Fig. 5). Anticarsia gemmatalis, Helicoverpa spp. and C. virescens were also nearly absent (< 0.51%) from MON 87701 × MON 89788 soybean in almost all samples but were detected in 25.3, 11.1 and 2.3% of the non-Bt soybean fields, respectively (Fig. 5). Among the species analyzed, A. gemmatalis had the second-highest density on non-Bt soybean across the regions examined (Fig. 5). Crocidosema aporema larvae was observed at low density in non-Bt soybean fields in some areas of south and central Brazil (Fig. 5); however, the beat-cloth method is not the best method to sample this species so its incidence might have been underestimated. In contrast to other species, R. nu larvae appeared in a moderate number of non-Bt soybean fields from south to central Brazil. Its occurrence was also detected in 5.5% of the MON 87701 × MON 89788 soybean fields sampled, mainly in growing areas located in Paraná and São Paulo states (Fig. 5).

Discussion
The proactive phenotypic and genotypic monitoring of Cry1Ac resistance did not show shifts in the susceptibility of C. includens that would lead to unexpected injury on MON 87701 × MON 89788 soybean in the field. Although an initial shift was observed from the genotypic resistance baseline monitoring carried out in 2014/15 10 , our findings indicated that the resistance allele frequency did not increase significantly over the following years. Even though some C. includens larvae tested in our F 2 screening survived on MON 87701 × MON 89788 soybean leaves, most of the larvae that developed into adults did not produce fertile eggs, suggesting an incomplete resistance to Cry1Ac in C. includens. If a major recessive trait confers resistance, 1/16 of the F 2 larvae are expected to be homozygous resistant and therefore able to complete their life cycle on a Bt plant 29,42 . Because the percentage of resistant larvae was much smaller than 1/16, we hypothesized that minor genes were responsible for the survival of C. includens on MON 87701 × MON 89788 soybean in our F 2 screening. The F 2 screen method proposed by Andow and Alstad 29 might not be the best way to estimate the resistance allele frequency of minor alleles involved with resistance, since it was developed to estimate the frequency of major resistance alleles. Although the resistance allele frequency might be biased with that method, it is still a good estimation of resistance allele frequency over the time, as we are using the same method over years. So far, no unexpected injury on MON 87701 × MON 89788 soybean was reported by C. includens, confirming the low resistance allele frequency detected in the F 2  www.nature.com/scientificreports/ screens. We further hypothesized that the resistance alleles detected in our genetic monitoring might be strongly associated with fitness, since the C. includens adults did not produce viable offspring. The high-dose/refuge strategy used for management of resistance to MON 87701 × MON 89788 soybean seems to have contributed to maintaining a low frequency of resistance in natural populations of C. includens. In addition to the results previously published by MacRae et al. 1 and Bernardi et al. 4 , further evidence supporting the functional recessiveness of Cry1Ac resistance in C. includens exposed to MON 87701 × MON 89788 soybean was obtained when a laboratory strain of C. includens selected on Cry1Ac Bt cotton was crossed with susceptible insects and the F 1 progeny showed complete mortality on MON 87701 × MON 89788 soybean leaves 43 .
In contrast to previous results, unexpected injury caused by secondary target pests R. nu and C. aporema was detected in MON 87701 × MON 89788 soybean fields during the 2020/21 season. Although susceptible reference populations of these species were not available, differences were detected in the susceptibility to MON 87701 × MON 89788 soybean among R. nu populations from Brazil and Argentina. Diet-overlay bioassays using Cry1Ac protein also indicated a high tolerance ratio (> 2,709-fold) for R. nu sampled from MON 87701 × MON 89788 soybean fields compared to C. includens from non-Bt soybean. A study conducted before the commercial launch of MON 87701 × MON 89788 soybean showed that the LC 50 of Cry1Ac was 1.53 and 0.70 µg/ml of artificial diet for C. includens and R. nu, respectively 44 . These results indicated a tolerance ratio of R. nu in relation to C. includens of 0.45-fold, evidencing that these Plusiinae had similar susceptibility to Cry1Ac in Brazil. The complementation test results suggested that the resistance alleles are probably at the same locus in different field populations of R. nu sampled from MON 87701 × MON 89788 soybean fields in Brazil. Similarly, offspring of C. aporema sampled from MON 87701 × MON 89788 soybean fields presented similar mortality on Bt and non-Bt soybean, suggesting a genetically based decrease in the susceptibility to Cry1Ac protein.
The development of resistance of R. nu and C. aporema to MON 87701 × MON 89788 soybean may have also been influenced by the biological characteristics of these species and of the cropping systems in Brazil. Historically, R. nu has been a soybean pest in the southern part of South America (Rio Grande do Sul state in Brazil, Uruguay and Argentina) 24,26,45 . In a prior field survey in 2019 and 2020, R. nu occurred at a low level in Brazil 7 . However, R. nu has increased in abundance in soybean areas at lower latitudes, which may indicate an adaptation of R. nu populations to warmer climate conditions before evolving resistance to MON 87701 × MON 89788 soybean. Low genetic distances between populations from different South American countries also suggest the absence of geographical isolation amongst these natural populations 45 . For C. aporema, the short lifecycle (12 days) allows the development of several generations per season, therefore increasing the selection pressure for resistance. The planting of common bean (Phaseolus vulgaris L.) combined with the presence of volunteer MON 87701 × MON 89788 soybeans after crop harvest in field areas where these pests were found in southern Brazil creates a "green bridge" during the soybean off-season and increases the sequential exposure of these pests to MON 87701 × MON 89788 soybean. Furthermore, the low adoption of structured refuges together with the overuse of insecticides in non-Bt areas 46 likely played an important role in establishing field-evolved Cry1Ac resistance in these species. Reductions in refuge compliance might compromise the high-dose/refuge strategy recommended for MON 87701 × MON 89788 soybean in Brazil. The common understanding is that the high yield of MON 87701 × MON 89788 soybean varieties relative to most of the commercially available non-Bt soybean varieties and the relative ease of pest management relative to non-Bt soybean have led some growers to not comply with the refuge recommendation in Brazil. In addition, the planting of refuges is not mandatory in Brazil, and recommendations are mainly promoted by technology providers, adding more challenges for compliance to  Our results also demonstrate that C. includens and A. gemmatalis remain the most abundant primary lepidopteran pests in non-Bt soybean fields, confirming the results from previous research 7 . These species were widely distributed across the soybean-growing regions of Brazil, whereas R. nu, Helicoverpa spp. and C. aporema were less prevalent. It is important to note, however, that beat-cloth sampling does not provide an accurate count of C. aporema, so its occurrence might have been underestimated. From 1980 to the 2000s, A. gemmatalis was the main lepidopteran pest of soybean in Brazil 48,49 . However, from the early 2000s, C. includens emerged as the key pest, which was likely influenced by the expansion of the soybean crop in Brazil 50 . In our results, R. nu and C. aporema were restricted to soybean fields in the Mid-South of Brazil, probably due to their better adaptation to subtropical and temperate climate of this region [25][26][27] , whereas Helicoverpa spp. was present in soybean fields in the Central regions of Brazil. Chloridea virescens was practically absent from both MON 87701 × MON 89788 soybean and non-Bt soybean fields.
Overall, we can conclude that MON 87701 × MON 89788 soybean remains effective against A. gemmatalis, C. includens, Helicoverpa spp. and C. virescens after eight years of commercial plantings in Brazil. However, we also report the first evidence of field-evolved resistance to MON 87701 × MON 89788 soybean in the secondary soybean pests R. nu and C. aporema. Our results indicate the MON 87701 × MON 89788 soybean continues to be a highly valuable and efficacious IPM tool that soybean growers in Brazil can rely on to manage the major pests C. includens and A. gemmatalis despite the localized cases of Cry1Ac resistance documented in secondary soybean lepidopteran pests. However, the increasing adoption of Bt soybean technology and the decrease in structured refuge compliance, along with these first and still localized cases of Cry1Ac resistance documented in secondary lepidopteran pests, highlights the importance of IRM practices.
To maintain the benefits of MON 87701 × MON 89788 soybean against primary target pests such as C. includens and A. gemmatalis, it is important to follow the refuge recommendation. The next generation of Bt soybean will continue to use Cry1Ac protein, now pyramided with Cry1A.105 and Cry2Ab2 51 . Therefore, the implementation of resistance management strategies is essential for the sustainability of both current and new Bt soybean varieties for managing lepidopteran pests in South America.  Figure 5. Incidence of lepidopteran pests in MON 87701 × MON 89788 soybean and non-Bt soybean in Brazilian fields sampled during the 2020/21 cropping season. A random effects statistical model was used to estimate pest abundance for each edaphoclimatic region. Separate analyses were conducted for each pest and field type. Pest abundance estimates were summarized using choropleth maps, where each edaphoclimatic region is color-coded according to its estimated pest abundance (different colors are assigned for estimates in the ranges < 0.10, 0.1-0.25, 0.25-0.5, 0.5-1, 1-2.5, 2.5-5, 5-10, and > 10 larvae/10 m). In all maps that show more than one color, there is statistically significant variation across edaphoclimatic regions (p < 0.05). Maps were generated using R statistical software-R version 4.0.2 (https:// www.R-proje ct. org/).