Global Patterns of Insect Resistance to Transgenic Bt Crops: The First 25 Years

Abstract

Crops genetically engineered to produce insecticidal proteins from the bacterium Bacillus thuringiensis (Bt) have improved pest management and reduced reliance on insecticide sprays. However, evolution of practical resistance by some pests has reduced the efficacy of Bt crops. We analyzed global resistance monitoring data for 24 pest species based on the first 25 yr of cultivation of Bt crops including corn, cotton, soybean, and sugarcane. Each of the 73 cases examined represents the response of one pest species in one country to one Bt toxin produced by one or more Bt crops. The cases of practical resistance rose from 3 in 2005 to 26 in 2020. Practical resistance has been documented in some populations of 11 pest species (nine lepidopterans and two coleopterans), collectively affecting nine widely used crystalline (Cry) Bt toxins in seven countries. Conversely, 30 cases reflect no decrease in susceptibility to Bt crops in populations of 16 pest species in 10 countries. The remaining 17 cases provide early warnings of resistance, which entail genetically based decreases in susceptibility without evidence of reduced field efficacy. The early warnings involve four Cry toxins and the Bt vegetative insecticidal protein Vip3Aa. Factors expected to favor sustained susceptibility include abundant refuges of non-Bt host plants, recessive inheritance of resistance, low resistance allele frequency, fitness costs, incomplete resistance, and redundant killing by multi-toxin Bt crops. Also, sufficiently abundant refuges can overcome some unfavorable conditions for other factors. These insights may help to increase the sustainability of current and future transgenic insecticidal crops.

sustainability, resistance management, practical resistance, Bacillus thuringiensis, corn

Insecticidal proteins from Bacillus thuringiensis (Bt) are toxic to some important pests but not to most nontarget organisms including vertebrates and arthropod natural enemies (Mendelsohn et al. 2003, Bravo et al. 2011, Comas et al. 2014, NASEM 2016, Krogh et al. 2020). The area planted globally to transgenic crops that produce Bt proteins increased from 1 million hectares (ha) in 1996 to 109 million ha in 2019 (ISAAA 2019, Fig. 1). In 2019, millions of farmers in 27 nations planted Bt crops including corn, cotton, soybean, sugarcane, and eggplant (ISAAA 2019). Bt crops can suppress pests, decrease reliance on conventional insecticides, and boost biological control (Hutchison et al. 2010, NASEM 2016, Dively et al. 2018, Romeis et al. 2019, Tabashnik et al. 2021). However, evolution of resistance by pests continues to reduce the benefits of Bt crops (Tabashnik and Carrière 2019, EPA 2021), despite widespread adoption of Bt crops called pyramids that produce two or more distinct Bt proteins that are toxic to each target pest (Carrière et al. 2015, 2016).

Fig. 1.

Annual planting of Bt crops worldwide.

Here we update the global status of field-evolved insect resistance to Bt crops. We classify 73 published cases of resistance monitoring data using three previously defined categories, review the evolutionary principles underlying resistance management, dispel some common misconceptions, compare field outcomes with predictions based on evolutionary principles, and conclude with a summary and look to the future.

Definitions of Three Categories of Resistance

We use previously published definitions for three categories of resistance to Bt crops: (1) practical resistance, (2) early warning of resistance, and (3) no decrease in susceptibility (Tabashnik and Carrière 2019). The first two categories are subsets of field-evolved resistance, which is a “genetically based decrease in susceptibility of an insect population to a Bt toxin caused by selection in the field” (Tabashnik and Carrière 2017). In addition to the criteria for field-evolved resistance, practical resistance means that more than half of the individuals in a population are resistant and the field efficacy of the Bt crop has decreased (Tabashnik et al. 2014). Thus, practical resistance has practical implications for pest control. As noted by Tabashnik and Carrière (2019): “Early warning of resistance includes all cases of field-evolved resistance where the two additional criteria for practical resistance are not met. For practical resistance and early warning of resistance, at least one population must meet the relevant criteria, whereas other populations may remain susceptible. No decrease in susceptibility indicates the monitoring data show no statistically significant decrease in susceptibility in any population tested after field populations have been exposed to a Bt crop.”

Global Status of Field-Evolved Resistance to Bt Crops

The 73 cases reviewed here are based on data from monitoring responses to 10 Bt toxins by 22 moth species and 2 beetle species in 12 countries, yielding 26 cases of practical resistance, 30 cases of no decrease in susceptibility, and 17 cases of early warning of resistance (Tables 1 – 5, Figs. 2 and 3). Each case conveys data on the resistance or susceptibility of one pest species in one country to one Bt toxin produced by one or more Bt crops (Bt corn, cotton, soybean, and/or sugarcane). This review includes 29 new cases as well as updates for 44 previously reported cases (Tabashnik and Carrière 2019). Nearly all the new cases summarized here are based on data published partly or entirely after we wrote the previous review. Also, the more thorough search we conducted for the current study identified a few papers reporting cases that we missed before (e.g., van Wyk et al. 2009; Nava-Camberos et al. 2010; Aguilar-Medel et al. 2017a,b). Our previous review is based on a literature search completed in May 2019. Because of the lag time between data collection and publication, those data are from 2016 or before. For this review, the literature search was completed on 3 September 2022 and the evidence is based on the first 25 yr of large-scale Bt crop cultivation: 1996–2020.

Table 1.

Practical resistance to Bt crops: 26 cases involving 11 pest species, nine Cry toxins, and seven countries (mean years to resistance = 6.6, SE = 0.7), updated from Tabashnik and Carrière (2019)

Insect	Crop	Toxin	Country	First yeara	Yearsb,c	Referenced
B. fusca	Corn	Cry1Ab	S. Africa	1998	8	Strydom et al. (2019)
C. aporema	Soybean	Cry1Ac	Brazil	2013	7	Horikoshi et al. (2021)
D. saccharalis	Corn	Cry1A.105	Argentina	2011	1c	Grimi et al. (2018)
D. saccharalis	Corn	Cry1Fa	Argentina	2008	4c	Grimi et al. (2018)
D. barberi	Corn	Cry3Bb	USA	2003	13	Calles-Torrez et al. (2019)
D. barberi	Corn	Cry34/35Ab	USA	2006	10	Calles-Torrez et al. (2019)
D. v. virgifera	Corn	Cry3Bb	USA	2003	6	Gassmann (2021)e
D. v. virgifera	Corn	Cry34/35Ab	USA	2006	7	Reinders et al. (2022)f
D. v. virgifera	Corn	eCry3.1Ab	USA	2014	0c	Gassmann (2021)
D. v. virgifera	Corn	mCry3A	USA	2007	4c	Gassmann (2021)
H. zea	Corn	Cry1Ab	USA	1996	8	Dively et al. (2016)
H. zea	Corn	Cry1A.105	USA	2010	6c	Yu et al. (2021)g
H. zea	Cotton	Cry1Ac	USA	1996	8	Yang et al. (2022)h
H. zea	Cotton	Cry1Fa	USA	2004	13c	Yang et al. (2022)i
H. zea	Cotton & corn	Cry2Ab	USA	2003	2c	Yang et al. (2022)j
O. nubilalis	Corn	Cry1Fa	Canada	2006	12	Smith and Farhan (2022)
P. gossypiella	Cotton	Cry1Ac	India	2002k	6	Naik et al. (2018)
P. gossypiella	Cotton	Cry2Ab	India	2006	8	Naik et al. (2018)
P. gossypiella	Cotton	Cry1Ac	Pakistan	2010k	6	Hassan et al. (2021)l
R. nu	Soybean	Cry1Ac	Brazil	2013	7	Horikoshi et al. (2021)
S. albicosta	Corn	Cry1Fa	Canada	2003	9	Smith et al. (2019)
S. albicosta	Corn	Cry1Fa	USA	2003	10	Coates et al. (2020)
S. frugiperda	Corn	Cry1Ab	Brazil	2008	2c	Omoto et al. (2016)
S. frugiperda	Corn	Cry1Fa	Argentina	2005	8	Vassallo et al. (2019)
S. frugiperda	Corn	Cry1Fa	Brazil	2009	2	Farias et al. (2014)
S. frugiperda	Corn	Cry1Fa	USA	2003	4	Huang et al. (2014)

^aFirst year of commercial planting of a Bt crop in the region monitored.

^bYears from the first commercial planting of a Bt crop in the region to the first sampling of field populations in the region yielding evidence of resistance.

^cCross-resistance suspected or known as a factor contributing to resistance.

^dFor cases supported by multiple references, a recent reference is cited above and selected additional references are cited for some cases below. The papers cited above and Tabashnik and Carrière (2017, 2019) provide more extensive lists of earlier references.

^eGassmann et al. (2011), Andow et al. (2016).

^fGassmann et al. (2014).

^gDively et al. (2016), Bilbo et al. (2019), Kaur et al. (2019).

^hLuttrell et al. (2004), Ali et al. (2006), Tabashnik et al. (2008a, b), Reisig et al. (2018).

ⁱRabelo et al. (2020).

^jAli and Luttrell (2007), Tabashnik et al. (2009), Dively et al. (2016), Bilbo et al. (2019).

^kIllegal planting of Bt cotton began in about 2000 in India (Tabashnik and Carrière 2019) and 2002 in Pakistan (Ahmad et al. 2021).

^lAkhtar et al. (2016, 2020).

Fig. 2.

Cumulative reported cases of practical resistance to Bt crops.

Fig. 3.

Global status of field-evolved pest resistance to Bt crops. Each symbol represents 1 of 73 cases indicating responses of one pest species in one country to one toxin in Bt corn, cotton, soy, and/or sugarcane (data from Tables 1–3).

Table 2.

No significant decrease in susceptibility to Bt crops: 30 cases involving five Bt toxins, 16 pest species, and 10 countries (mean documented years of susceptibility = 12.4 yr, SE = 1.1 yr), updated from Tabashnik and Carrière (2019)

Insect	Crop	Toxin	Country	First yeara	Yearsb	Reference
A. gemmatalis	Soy	Cry1Ac	Brazil	2013	8	Horikoshi et al. (2021)
C. includens	Soy	Cry1Ac	Brazil	2013	8	Horikoshi et al. (2021)
C. virescens	Cotton	Cry1Ac	Mexico	1996	11	Blanco et al. (2009)
C. virescens	Cotton	Cry1Ac	USA	1996	11	Blanco et al. (2009)
C. virescens	Cotton	Cry2Ab	USA	2003	2	Ali and Luttrell (2007)
D. grandiosella	Corn	Cry1Ab	USA	1999	6	Huang et al. (2007)
E. biplaga	Cotton	Cry1Ac	S. Africa	1998	15	Fourie et al. (2017)
E. insulana	Cotton	Cry1Ac	India	2002	13	Kumar et al. (2019)
E. vittella	Cotton	Cry1Ac	India	2002	9	Shera and Amora (2016)
H. armigera	Cotton	Cry1Ac	Australia	1996	16	Downes (2019)
H. armigera	Cotton	Cry2Ab	Australia	2004	15	Knight et al. (2021)
H. punctigera	Cotton	Cry1Ac	Australia	1996	21	Downes (2019)
H. punctigera	Cotton	Cry2Ab	Australia	2004	15	Knight et al. (2021)
H. zea	Corn	Vip3Aa	Brazil	2011	4	Leite et al. (2017)
H. zea	Cotton	Cry1Ac	Mexico	1996	19	Aguilar-Medel et al. (2017a)
H. zea	Cotton	Cry2Ab	Mexico	2004	11	Aguilar-Medel et al. (2017a)
O. nubilalis	Corn	Cry2Ab	Canada	2008	13	Smith and Farhan (2022)
O. nubilalis	Corn	Cry1Ab	Spain	1998	24	García et al. (2022)
O. nubilalis	Corn	Cry1Ab	USA	1996	15	Siegfried and Hellmich (2012)
O. nubilalis	Corn	Cry1Fa	USA	2003	8	Siegfried et al. (2014)
P. gossypiella	Cotton	Cry1Ac	China	2000	15	Wang et al. (2020)
P. gossypiella	Cotton	Cry1Ac	Mexico	1996	10	Nava-Camberos et al. (2010)
P. gossypiella	Cotton	Cry2Ab	Mexico	1996	10	Nava-Camberos et al. (2010)
P. gossypiella	Cotton	Cry1Ac	USA	1996	12	Tabashnik et al. (2010)
P. gossypiella	Cotton	Cry2Ab	USA	2003	5	Tabashnik et al. (2010)
S. calamistis	Corn	Cry1Ab	S. Africa	1998	8	Van Wyk et al. (2009)
S. nonagroides	Corn	Cry1Ab	Spain	1998	24	García et al. (2022)
S. exigua	Cotton	Cry1Ac	Mexico	1996	20	Aguilar-Medel et al. (2017b)
S. exigua	Cotton	Cry2Ab	Mexico	2004	12	Aguilar-Medel et al. (2017b)
S. frugiperda	Corn	Vip3Aa	Argentina	2011	6	Murúa et al. (2019)

^aFirst year of commercial planting of a Bt crop in the region monitored.

^bYears of documented susceptibility was calculated as the year of the most recent monitoring data showing no decrease in susceptibility minus the first year of commercialization in the region. Because susceptibility has probably persisted beyond the year of the most recent monitoring data, the years of documented susceptibility underestimate the total years of susceptibility.

Table 3.

Early warning of resistance to Bt corn, cotton, and sugarcane: 17 cases involving eight pest species, five Bt toxins, and eight countries, updated from Tabashnik and Carrière (2019)

Insect	Crop	Toxin	Country	First yeara	Yearsb	Referencesc
D. saccharalis	Sugarcane	Cry1Ab	Brazil	2017	2	Oliveira et al. (2022)
D. saccharalis	Sugarcane	Cry1Ac	Brazil	2017	2	Oliveira et al. (2022)
D. saccharalis	Corn	Cry1Ab	USA	1999	10	Huang et al. (2012)
E. vittella	Cotton	Cry1Ac	Pakistan	2010	8	Ahmad et al. (2021)
H. armigera	Cotton	Cry1Ac	China	1997	20	Zhang et al. (2019)d
H. armigera	Cotton	Cry1Ac	India	2002	12	Singh et al. (2021)
H. armigera	Cotton	Cry1Ac	Pakistan	2010e	4	Saleem et al. (2019)f
H. zea	Corn & cotton	Vip3Aa	USA	2010g	10	Yang et al. (2021)
O. furnacalis	Corn	Cry1Ab	Philippines	2003	6	Alcantara et al. (2011)
O. nubilalis	Corn	Cry1Ab	Canada	1997	21	Smith and Farhan (2022)
O. nubilalis	Corn	Cry1A.105	Canada	2008	10	Smith and Farhan (2022)
S. exigua	Cotton	Cry1Ac	Pakistan	2010	6	Shahid et al. (2019)
S. frugiperda	Corn	Cry1A.105	USA	2009	6	Gutierrez-Moreno et al. (2020)
S. frugiperda	Corn	Cry1Ab	Argentina	2008	9	Murúa et al. (2019)
S. frugiperda	Corn	Cry1A.105	Argentina	2011	6	Murúa et al. (2019)
S. frugiperda	Corn	Cry2Ab	Argentina	2011	6	Murúa et al. (2019)
S. frugiperda	Corn	Vip3Aa	Brazil	2011	5	Amaral et al. (2020)

^aFirst year of commercial planting of a Bt crop in the region monitored.

^bYears from the first commercial planting of a Bt crop in the region to the most recent monitoring data that yielded evidence of early warning of resistance.

^cFor cases supported by multiple references, a recent reference is cited above and selected additional references are cited for some cases below. The papers cited above and Tabashnik and Carrière (2017, 2019) provide more extensive lists of earlier references.

^dJin et al. (2018).

^eIllegal planting of Bt cotton began in about 2002 in Pakistan (Ahmad et al. 2021).

^fAlvi et al. (2012), Ahmad et al. (2019), Jamil et al. (2021).

^gIn the U.S., Vip3Aa-producing cotton and corn were registered in 2008 and 2009, respectively (Tabashnik et al. 2009), but their adoption has been slow. From 2017 to 2020, the percentage of the total crop that produced Vip3Aa rose from 3 to 6% for corn and from 4 to 37% for cotton (Yang et al. 2021).

Practical Resistance: 26 Cases

The total number of reported cases of practical resistance to Bt crops rose from 19 in 2016 (Tabashnik and Carrière 2019) to 26 in 2020 (Fig. 2). For these 26 cases, the mean time in a particular region from the first commercial planting of a Bt crop to the first sampling of field populations providing evidence of practical resistance is 6.6 yr (SE = 0.7 yr, Table 1). The time from first commercial planting to first evidence of practical resistance is shorter for the 8 cases where cross-resistance is known or suspected (median = 3 yr) than for the other 18 cases (median = 8 yr, Mann-Whitney U-test, U = 119, P = 0.01). From 2016 to 2020, the number of pest species with practical resistance increased from 7 to 11 (Table 1). The 11 pest species with practical resistance to at least one Bt crop includes nine moths (Busseola fusca, Crocidosema aporema, Diatraea saccharalis, Helicoverpa zea, Ostrinia nubilalis, Pectinophora gossypiella, Rachiplusia nu, Spodoptera frugiperda, and Striacosta albicosta) and two beetles (Diabrotica virgifera virgifera and Diabrotica barberi, Table 1). Eighteen of the cases of practical resistance are to Bt corn, five to Bt cotton, one to Bt corn and cotton, and two to Bt soybean (Table 1).

The seven new cases since 2016 include practical resistance to Cry1Ac soybean in Brazil by C. aporema and R. nu, which are the first cases of resistance to Bt soybean and the first documented cases of resistance to a Bt crop in secondary pests (Horikoshi et al. 2021). The other five new cases entail practical resistance of D. barberi to Cry3Bb and Cry34/35Ab corn in the United States (Calles-Torrez et al. 2019), O. nubilalis to Cry1Fa corn in Canada (Smith et al. 2019), H. zea to Cry1Fa cotton in the United States (Yang et al. 2022), and P. gossypiella to Cry1Ac cotton in Pakistan (Akhtar et al. 2016, 2020). The first five new cases mentioned above entail practical resistance to Bt corn in four pest species that had no previous cases of practical resistance. The two other new cases record H. zea resistance to an additional toxin in Bt cotton (Cry1Fa) in the United States and P. gossypiella resistance to Bt cotton producing Cry1Ac in Pakistan in addition to India (Table 1).

Each of the nine Cry toxins for which resistance is monitored in Bt crops (Cry1Ab, Cry1Ac, Cry1A.105, Cry1Fa, Cry2Ab, Cry3Bb, mCry3A, eCry3.1Ab, and Cry34/35Ab, which was renamed Gpp34/Tpp35Ab by Crickmore et al. 2021) is involved in at least one case of practical resistance (Table 1). The cases of practical resistance for each toxin are 10 for the three Cry1A toxins, eight for Cry1Fa, two for Cry2Ab, four for the three Cry3 toxins, and two for Cry34/35Ab (Table 1). We found no reports of practical resistance to Vip3Aa. Although Cry and Vip proteins share some structural similarities, they have different midgut receptors and cross-resistance between them is weak or nil (Chakroun et al. 2016, Tabashnik and Carrière 2020, Byrne et al. 2021, Tabashnik et al. 2022).

The 26 cases listed in Table 1 probably underestimate the current extent of practical resistance to Bt crops because of the lag between data collection and publication. Also, some cases are unlikely to be documented in publications. For example, Fatoretto et al. (2017) do not report the relevant monitoring data, but they imply that S. frugiperda has evolved practical resistance to Bt corn producing Cry1A.105 and Cry2Ab in Brazil (in addition to Cry1Ab and Cry1Fa as documented in Table 1). They also state that resistance of this pest to Bt corn producing the same toxins as in Brazil “emerged in Argentina, Paraguay, and Colombia evidenced by growers [sic] reports of unexpected damage (data not published).” If verified, these claims would add at least five cases of practical resistance to Bt crops to the list in Table 1.

No Significant Decrease in Susceptibility: 30 Cases

In striking contrast with the cases of practical resistance summarized above are 30 cases where monitoring data demonstrate no significant decrease in susceptibility after 2 to 24 yr of pest exposure to Bt crops (Table 2). We found 11 new cases of no decrease in susceptibility since our previous review (Tabashnik and Carrière 2019). Also, recent monitoring data extended the documented duration of susceptibility for several of the previously reviewed cases. In 67% of the 30 cases, susceptibility has been sustained for at least a decade. The mean documented duration of susceptibility for the 30 cases is 12.2 yr (SE = 1, Table 2). This mean underestimates the current duration of susceptibility because some pests have remained susceptible after the most recent monitoring data were reported. The 30 cases of no decrease in susceptibility occur in 16 lepidopteran species (Tables 2 and 4). Of these 16 species, 12 are not involved in any cases of practical resistance (Anticarsia gemmatalis, Chloridea virescens, Chrysodeixis includens, Diatraea grandiosella, Earias biplaga, Earias insulana, Earias vittella, Helicoverpa armigera, Helicoverpa punctigera, Sesamia calamistis, Sesamia nonagrioides, and Spodoptera exigua). The other four species (H. zea, O. nubilalis, P. gossypiella, and S. frugiperda) have evolved practical resistance in at least one case (Tables 1 and 4).

Table 4.

Field-evolved resistance to Bt crops for 24 pests (73 cases, data from Tables 1–3)

Pest	Practical resistance	Early warning	No decrease in susceptibility
H. zea	5	1	3
S. frugiperda	4	5	1
D. v. virgifera	4	0	0
P. gossypiella	3	0	5
D. saccharalis	2	3	0
D. barberi	2	0	0
S. albicosta	2	0	0
O. nubilalis	1	2	4
B. fusca	1	0	0
C. aporema	1	0	0
R. nu	1	0	0
H. armigera	0	3	2
O. furnacalis	0	1	0
E. vittella	0	1	1
S. exigua	0	1	2
A. gemmatalis	0	0	1
C. includens	0	0	1
D. grandiosella	0	0	1
E. biplaga	0	0	1
E. insulana	0	0	1
S. calamistis	0	0	1
S. nonagroides	0	0	1
H. punctigera	0	0	2
C. virescens	0	0	3

Early Warning of Resistance: 17 Cases

The 17 cases of early warning of resistance involve significantly decreased susceptibility to Vip3Aa by H. zea in the U.S. and S. frugiperda in Brazil, as well as to Cry toxins by D. saccharalis, E. vittella, H. armigera, Ostrinia furnacalis, O. nubilalis, S. exigua, and S. frugiperda (Table 3). The mean time from commercial introduction to detection of resistance in these 17 cases is 8.6 yr (SE = 1.3, Table 3). Two notable cases involve D. saccharalis resistance to Cry1Ab and Cry1Ac produced by Bt sugarcane in Brazil (Oliveira et al. 2022). F₂ screens conducted during the 2019-2020 season revealed significantly higher resistance allele frequency in São Paulo (0.0393 for Cry1Ab and 0.0245 for Cry1Ac) than in the neighboring states of Goiás and Minas Gerais (0.0034 for Cry1Ab and 0.0045 for Cry1Ac). Given that Bt sugarcane accounted for less than 0.1% of Brazil’s sugarcane in 2019–2020, Oliveira et al. (2022) conclude the resistance reflects previous selection from single-toxin Bt maize grown extensively in São Paulo with limited refuges.

Analysis of Cases by Pest Species and Country

The most cases of practical resistance recorded for any pest is five for H. zea, followed by four each for S. frugiperda and D. v. virgifera (Tables 1 and 4). Thus, these three pests account for half of all cases of practical resistance. For S. frugiperda, five cases of early warning of resistance are also reported (Tables 2 and 4). This yields 9 of 10 cases for this pest with some level of resistance, which is the most for any pest. At the other end of the spectrum, nine species have only cases with no decrease in susceptibility (i.e., no practical resistance or early warning of resistance, Table 4). Seven of these nine species each have one case of no decrease in susceptibility, whereas two such cases are documented for H. punctigera and three for C. virescens (Tables 2 and 4).

Among the 12 countries analyzed, the percentage of cases showing no decrease in susceptibility ranges from 0% (Pakistan and the Philippines) to 100% (Australia, Mexico, and Spain; Table 5). The U.S. has more than twice as many total cases reported as any other country, of which 30% (7 of 23) indicate no decrease in susceptibility and 57% (13 of 23) reflect practical resistance (Table 5). Thus, the U.S. accounts for half of the world’s 26 cases of practical resistance, which is consistent with its high adoption of Bt crops and extensive resistance monitoring. From 2009 to 2020, Bt cultivars accounted for a mean of over 75% of the hectares of corn and cotton in the U.S. (USDA ERS 2021).

Table 5.

Field-evolved resistance to Bt crops in 12 countries (73 cases, data from Tables 1–3)

Country	Practical resistance	Early warning	No decrease in susceptibility	Total cases	No decrease in susceptibility (%)a
Pakistan	1	3	0	4	0
Philippines	0	1	0	1	0
Argentina	3	3	1	7	14
Canada	2	2	1	5	20
Brazil	4	3	3	10	30
USAb	13	3	7	23	30
India	2	1	2	5	40
China	0	1	1	2	50
S. Africa	1	0	2	3	67
Australia	0	0	4	4	100
Mexico	0	0	7	7	100
Spain	0	0	2	2	100

^aPercentage of cases showing no decrease in susceptibility.

^bIncludes Puerto Rico.

Evolutionary Principles Underlying Resistance Management

The evolutionary principles underlying resistance management for Bt crops have been thoroughly elucidated in many previous articles (e.g., Tabashnik and Croft 1982; Gould 1998; Roush 1998; Caprio 2001; Andow and Ives 2002; Zhao et al. 2003; Carrière et al. 2004a, 2010; 2015; 2016; Sisterson et al. 2004; Tabashnik et al. 2008a, 2009; 2013; Ives et al. 2011; Onstad et al. 2011; Tabashnik and Carrière 2017). Here we summarize some key conclusions and dispel some common misconceptions.

Refuges

The refuge strategy, which was initially envisioned for delaying the evolution of resistance to insecticide sprays (Comins 1977, Georghiou and Taylor 1977, Tabashnik and Croft 1982) has been central for sustaining the efficacy of Bt crops since these crops were first introduced (EPA 1998; Gould 1998; Tabashnik and Carrière 2017, 2019). In the context of Bt crops, refuges consist of non-Bt host plants of the pests that attack Bt crops. Refuges can be non-Bt crop counterparts of Bt crops (e.g., non-Bt corn refuges for Bt corn). Refuges can also be other wild or cultivated non-Bt host plants, which are sometimes called “natural refuges” (e.g., non-Bt corn refuges for Bt cotton, Jin et al. 2015). Sometimes, even a Bt crop can be a refuge for a subset of the pests feeding on that crop. For example, Bt corn producing only lepidopteran-active toxins can be a refuge for beetles that attack Bt corn producing coleopteran-active toxins (Carrière et al. 2020a). Planting of refuges for resistance management is mandated in many countries and occurs voluntarily in others (Carrière et al. 2020b). Refuges also occur inadvertently for reasons unrelated to resistance management. In China, where refuges have not been mandated, such inadvertent refuges have been particularly important (Jin et al. 2015, Wan et al. 2017).

Eight Conditions Expected to Delay Evolution of Pest Resistance to Bt Crops

Analyses of population genetic models in the papers cited above indicate resistance is generally expected to evolve slower as the following occurs: 1) the abundance of non-Bt host plant refuges relative to Bt crops increases, 2) the inheritance of resistance becomes more recessive, 3) the frequency of alleles conferring resistance decreases, 4) the magnitude and dominance of fitness costs associated with resistance increase, and 5) resistance becomes less complete (i.e., more incomplete) (Carrière et al. 2016).

As explained by Tabashnik and Carrière (2017): “When fitness costs occur, fitness on non-Bt host plants is higher for susceptible insects than resistant insects.” This means that fitness costs can decrease the frequency of resistance in refuges. For homozygous resistant insects, incomplete resistance is a fitness disadvantage they suffer when feeding on Bt plants relative to the corresponding non-Bt plants (Carrière et al. 2010). For example, a resistant pest that survives on Bt corn might produce fewer offspring when reared on Bt corn than on non-Bt corn. Thus, incomplete resistance refers to the phenotype of individual insects, not the frequency of resistance in a population. Accordingly, one cannot evaluate incomplete resistance by testing a population that consists of a potentially heterogeneous mixture of resistant and susceptible individuals (Tabashnik and Carrière 2017).

The five conditions listed above are important for delaying pest resistance to Bt plants that produce a single toxin as well as those that produce two or more distinct toxins targeting the same pest, which are called pyramids (Carrière et al. 2016). Three additional conditions favor slower evolution of pest resistance to pyramids: 6) each toxin in the pyramid kills all or nearly all susceptible insects (i.e., redundant killing is complete or nearly complete), 7) little or no cross-resistance occurs between toxins in the pyramid, and 8) pyramids are not grown simultaneously with single-toxin plants that produce one of the toxins in the pyramid (Brévault et al. 2013; Carrière et al. 2015, 2016).

Interactions Among Factors Govern Evolution of Resistance

Modeling results predict that outcomes depend on interactions among the factors listed above (e.g., Tabashnik and Croft 1982, Ives et al. 2011, Carrière et al. 2015). For example, the resistance-delaying effect of fitness costs increases as the abundance of refuges relative to Bt crops increases. Conversely, without refuges, fitness costs have no effect on the evolution of resistance. Generally, the combined effects of all factors and their interactions determine the outcome. So, as detailed below, the ability of refuges to delay resistance does not always require one or more other factors to meet specific criteria.

Random Mating, Recessive Inheritance, and Low Resistance Allele Frequency Are Not Required

In the ideal scenario for success of the refuge strategy, resistance is recessively inherited and rare. The term “high dose refuge strategy” refers to the optimal condition where the dose of Bt toxin ingested by pests is high enough to kill all or nearly all individuals heterozygous for resistance, yielding functionally recessive resistance. Under ideal conditions, the extremely rare resistant homozygotes that survive on Bt crops mate almost entirely with the much more common homozygous susceptible insects from refuges. Such matings produce heterozygotes that cannot survive on the Bt crop. Many modeling studies have predicted that under these ideal conditions, evolution of resistance will be delayed substantially (e.g., Gould 1998, Carrière and Tabashnik 2001, Tabashnik et al. 2008a). Nonetheless, both modeling results and field outcomes show that refuges can delay resistance without random mating between resistant and susceptible pests, recessive inheritance of resistance, or a low resistance allele frequency.

Some mating between resistant and susceptible pests favors success of the refuge strategy and random mating is a convenient assumption for modelers. However, several detailed modeling studies have identified circumstances where the refuge strategy is more effective when mating is not entirely random. For example, resistance evolution was slower with nonrandom mating and nonrandom oviposition on Bt crops and refuges than with random mating and oviposition (Caprio 2001). Increasing movement of males from refuges to Bt crops while keeping females in refuges to oviposit and sustain susceptible populations, which entails some nonrandom mating, can be optimal for delaying resistance (Andow and Ives 2002, Carrière et al. 2004a, Ives et al. 2011).

Although refuges are most effective when resistance is recessive, modeling results show that sufficiently abundant refuges can substantially delay nonrecessive resistance (Tabashnik et al. 2008a, Jin et al. 2015). To achieve a given delay in resistance evolution, the abundance of refuges relative to Bt crops must be higher for nonrecessive resistance than for recessive resistance. For example, with a moderate recessive fitness cost and incomplete resistance, the refuge percentage needed to delay evolution of resistance >20 yr was predicted to be only 5% for completely recessive resistance versus 65% for completely dominant resistance (Tabashnik et al. 2008a). In a laboratory selection experiment for five generations with Plutella xylostella that had nonrecessive resistance to Bt canola producing Cry1Ac, the mean increase in the concentration killing 50% of larvae (LC₅₀) was 103-fold with no refuge versus 3-fold with a 50% refuge (Cerda et al. 2006).

Contrary to an often-repeated misconception, a resistance allele frequency of less than or equal to 0.001 is not essential for the success of the refuge strategy (Tabashnik et al. 2013). For any given rate of evolution of resistance and threshold resistance allele frequency (e.g., 0.50), the time to reach the threshold does increase as the starting frequency declines. The starting frequency (sometimes called initial resistance allele frequency) could be the frequency before a Bt crop is grown in the field or at any other time from which one is trying to predict the trajectory of resistance. Under most circumstances, the increase in resistance allele frequency per generation or per year is greater at higher resistance allele frequencies. Typically, this increase is expected to be roughly proportional to the logarithm of the initial resistance allele frequency such that the time for a 10-fold increase in resistance allele frequency is similar from 0.001 to 0.01 and 0.01 to 0.1.

An initial resistance allele frequency of 0.001 has been used as a default assumption in modeling when directly relevant data are not available. This is reasonable for pest populations that have had no previous exposure to Bt crops (Carrière et al. 2010). However, both modeling results and field outcomes show that refuges can delay resistance when the resistance allele frequency is >0.001. For example, refuges delayed resistance to a single-toxin Bt crop substantially in modeling with an initial resistance allele frequency of 0.3, fitness costs, and recessively inherited resistance (Carrière and Tabashnik 2001). Also, the initial frequency of alleles conferring resistance to Cry1Ac was estimated as 0.16 for P. gossypiella in the southwestern U.S. in 1997 (Tabashnik et al. 2000), yet this pest did not evolve resistance to Cry1Ac despite more than two decades of exposure to cotton producing this toxin alone or in combination with Cry2Ab (Tabashnik et al. 2021). Based on a two-locus model, fitness costs were especially effective for delaying resistance to two-toxin pyramids, even with initial frequencies of resistance substantially above 0.001 for one or both toxins (Gould et al. 2006).

Field Outcomes Confirm Evolutionary Principles Underlying the Refuge Strategy

Several analyses have demonstrated that global patterns of resistance to Bt crops are consistent with predictions from the evolutionary principles underlying the refuge strategy (e.g., Tabashnik et al. 2008a, 2013; Carrière et al. 2010; Tabashnik and Carrière 2017, 2019; Huang 2021). Empirical evidence that refuges can delay evolution of resistance to Bt toxins includes results from laboratory and greenhouse selection experiments with Plutella xylostella (Liu and Tabashnik 1997, Tang et al. 2001, Cerda et al. 2006, Liu et al. 2014, Zhou et al. 2018) and large-scale retrospective analyses comparing field outcomes with predictions from population genetic models (Tabashnik et al. 2005, 2008a; Jin et al. 2015; Wan et al. 2017). In related work, resistance of Bemisia tabaci to the insecticide pyriproxyfen was negatively associated with the area of unsprayed refuges of cotton in an 8-year field study of 84 populations (Carrière et al. 2012).

Another way to evaluate refuge effectiveness is to compare resistance evolution for the same pest, crop, and Bt toxin between regions that vary in the abundance of refuges relative to Bt crops. Although such comparisons are compelling, an important caveat is that factors other than refuge abundance may also differ among regions and affect outcomes in such natural experiments.

The remarkably different outcomes with P. gossypiella and Bt cotton in the world’s three leading cotton-producing countries (India, the U.S., and China) are consistent with variation in refuge abundance (Tabashnik and Carrière 2019). Because cotton is the predominant host of P. gossypiella in these countries, the abundance of non-Bt cotton refuges is key. In India, where mandated non-Bt cotton refuges were generally not implemented (Carrière et al. 2020b), practical resistance evolved to Bt cotton producing Cry1Ac in 6 yr and to Bt cotton producing Cry1Ac + Cry2Ab in 8 yr (Table 1). In the U.S., where most growers complied with requirements to plant non-Bt cotton refuges (Carrière et al. 2005), no decrease in susceptibility occurred for either Cry1Ac or Cry2Ab despite extensive exposure to both toxins (Table 2). In combination with mass releases of sterile moths and other tactics (Staten and Walters 2021), sustained efficacy of Bt cotton was critical for eradication of this invasive pest from the cotton-growing areas of the continental U.S., which was declared in 2018 (Tabashnik et al. 2021).

In China, P. gossypiella resistance to Cry1Ac cotton rose initially, then declined when non-Bt cotton refuge abundance increased (Wan et al. 2017). Planting of F₂ hybrids from crosses between Bt and non-Bt cotton increased the refuge percentage to about 25%. This change was not part of a resistance management strategy, but instead apparently reflects short-term economic advantages of the F₂ hybrids (Wan et al. 2017). The evidence from China implies that the increased non-Bt cotton refuge abundance over time within a region interacted with fitness costs to increase susceptibility, which complements and reinforces the conclusion from the comparisons among countries.

For pests that feed on a wide range of different host plants, such as H. armigera, the field outcomes suggest that natural refuges have also helped to delay resistance to Bt crops in some countries. In India, where P. gossypiella rapidly evolved practical resistance to Bt cotton producing Cry1Ac and Cry2Ab, practical resistance has not been recorded to either toxin for H. armigera, despite its long-term exposure to Bt cotton. These results are consistent with the hypothesis that the abundant non-Bt host plants of H. armigera other than cotton have acted as natural refuges in India (Ravi et al. 2005). Detailed analysis of H. armigera in China comparing field data with predictions from modeling also implies that non-Bt host plants other than cotton have delayed evolution of resistance there as well (Jin et al. 2015).

Another comparison consistent with the hypothesis that refuges can delay resistance involves the polyphagous pest H. zea, which attacks corn, cotton, and many other crops. Monitoring data reveal widespread practical resistance of H. zea to Cry1Ac and Cry2Ab in the U.S. (Table 1) versus susceptibility to these toxins sustained in neighboring Mexico through at least 2015 (Aguilar-Medel et al. 2017a). As in the U.S., Mexican growers first planted Bt cotton producing Cry1Ac in 1996, then gradually switched during the mid-2000s to Bt cotton producing Cry1Ac + Cry2Ab (Aguilar-Medel et al. 2017a). In Mexico, the percentage of all cotton planted to Bt cotton was 87% in 2011 and increased to 93% of the 239,000 ha planted in 2019 (James 2011, ISAAA 2019). Despite high adoption of Bt cotton in Mexico, susceptibility of H. zea to Cry1Ac did not decrease for at least 19 yr.

In Mexico, Bt cotton is the only extensively cultivated Bt crop, which means corn and other noncotton host plants of H. zea are potential refuges. Conversely, in the U.S., Bt plants comprised a mean of 81% of all corn and 87% of all cotton planted from 2016 to 2020 (USDA ERS 2021). The U.S. ended the refuge requirement for multi-toxin Bt cotton in 2007 for most states (EPA 2007) and compliance with planting refuges of non-Bt corn has been low in southern states (Reisig 2017). By contrast, Mexico still requires non-Bt cotton refuges (20% of total cotton hectares if sprayed or 4% if unsprayed; Aguilar-Medel et al. 2017a). More thorough analysis of refuge abundance and movement of H. zea in Mexico (e.g., Carrière et al. 2004b, Head et al. 2010) is needed to rigorously test the hypothesis that refuges of non-Bt corn and other non-Bt host plants have helped to sustain the efficacy of Bt cotton there.

Summary and Future Prospects

The first 25 yr of extensive Bt crop cultivation have produced some spectacular successes and disappointing failures in terms of durable efficacy. The successes include sustained susceptibility of P. gossypiella to Bt cotton that facilitated eradication of this invasive pest from the cotton-growing regions of the U.S. and northern Mexico (Table 2, Tabashnik et al. 2021). Sustained efficacy of Bt cotton against C. virescens is probably also the key factor that dramatically reduced this pest’s density in the U.S. and Mexico (Table 2, Blanco 2012; Nava-Camberos et al. 2018, 2019). Efficacy of Bt corn against O. nubilalis in the U.S. since 1996 has helped to reduce this pest’s density to historically low levels in several midwestern states including Wisconsin, where its abundance in 2018 and 2019 was the lowest in 78 yr (Hutchison et al. 2010, University of Illinois 2015, WDATCP 2019, Potter et al. 2022). For more than two decades, Bt cotton has remained effective against H. armigera and H. punctigera in Australia as well as H. armigera and P. gossypiella in China (Tables 2 and 3).

Whereas use of multiple Bt toxins has almost certainly extended the durability of Bt crops in Australia, Mexico and the U.S. in the examples listed above, Cry1Ac has been the only Bt toxin produced by Bt cotton in China. Likewise, Cry1Ab has been the only Bt toxin produced by Bt corn in Europe, where monitoring data show no decrease in susceptibility for more than two decades in the two major target pests, O. nubilalis and S. nonagrioides (García et al. 2022). This prolonged efficacy of single-toxin Bt crops has greatly exceeded expectations of most if not all experts.

Unlike the long-term efficacy of Bt crops in many cases, rapid evolution of practical resistance to Bt crops has reduced their efficacy in at least 26 cases (Table 1). For P. gossypiella in India and D. v. virgifera in the U.S., practical resistance has evolved to all Bt toxins produced by currently available Bt crops targeting each pest. For H. zea in the United States, widespread practical resistance has reduced the efficacy of the Cry toxins in Bt corn and cotton, leaving Vip3Aa as the only Bt toxin in current Bt crops that is fully or almost fully effective against this pest in some regions (Tables 1–4).

Although it will be difficult to implement new transgenic insecticidal crops that have safety and initial efficacy comparable to the most successful Bt crops deployed during the past quarter century, the potential rewards are great and prospects are emerging. Combining Bt toxins with RNA interference (RNAi) traits probably has some value, but the RNAi traits are generally not as potent as Bt toxins and are vulnerable to evolution of broad-spectrum pest resistance to RNAi (Khajuria et al. 2018). Two-toxin Bt corn producing the novel toxins Cry1B.868 and Cry1Da_7 is effective against some strains of S. frugiperda, H. zea, and D. saccharalis resistant to other Cry toxins (Wang et al. 2019; Horikoshi et al. 2021, 2022). New families of insecticidal proteins from non-Bt soil bacteria and ferns show promise against Bt-resistant D. v. virgifera and lepidopteran pests, respectively (Schellenberger et al. 2016, Liu et al. 2019). Bt toxin Vip4Da (renamed Vpb4Da by Crickmore et al. 2021) protects corn roots against D. v. virgifera that are resistant to Bt toxins Cry3Bb1, Cry34/35Ab, and RNAi (Yin et al. 2021). Other new insecticidal proteins such as Bt toxin Cry51Aa (Akbar et al. 2019, renamed Mpp51 by Crickmore et al. 2021) and Tma proteins from ferns (Shukla et al. 2016) are expanding the spectrum of pests targeted by transgenic crops to include hemipterans. Also, lepidopteran-active Bt toxins that have been used extensively in transgenic corn, cotton, and soybean are being deployed in other crops in Asia and Africa. Bt eggplant producing Cry1Ac to control Leucinodes orbonalis was approved in Bangladesh in 2013 (Shelton et al. 2020) and Bt cowpea producing Cry1Ab to control Maruca vitrata was approved in Nigeria in 2019 (Addae et al. 2020).

We view the first 25 yr of Bt crops as the dawn of the era of managing insect pests with transgenic crops. One key lesson is that refuges of non-Bt host plants can help to delay evolution of resistance by pests (Tabashnik and Carrière 2019). Both theory and field outcomes indicate that refuges are most effective when inheritance of resistance is recessive and redundant killing is complete or nearly complete for multi-toxin pyramids. However, sufficiently abundant refuges can delay the evolution of resistance even when resistance is not recessive or some of the other conditions favoring success of the refuge strategy are not met. Rather than relying on any single tactic such as transgenic crops with one or more insecticidal traits, integrated pest management (IPM) is optimal for sustainable pest suppression (Liu et al. 2014, Hutchison 2015, Anderson et al. 2019, Hurley and Sun 2019, Romeis et al. 2019, Carrière et al. 2020a, Tabashnik et al. 2021, Van den Berg et al. 2021). As we move forward to the next generation of transgenic insecticidal crops, the challenge is to implement what has been learned to enhance their sustainability for the benefit of all.

Acknowledgments

We thank Sotero Aguilar-Medel, Shakeel Ahmad, Sharon Downes, Kelly Estes, Gema P. Farinós, Krista Hamilton, William Hutchison, Kristen Knight, David Mota-Sanchez, Jocelyn Smith, and Joseph Spencer for sharing information about resistance monitoring and pest abundance. This work was supported by two grants from the United States Department of Agriculture (USDA) National Institute of Food and Agriculture: Agriculture and Food Research Initiative 2020-67013-31924 and Biotechnology Risk Assessment Research Grants Program 2020-33522-32268. All opinions expressed in this paper are the authors’ and do not necessarily reflect the policies and views of USDA or the individuals acknowledged above. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. USDA is an equal opportunity provider and employer.

Conflict of Interest

B.E.T. is coauthor of a patent on modified Bt toxins “Suppression of Resistance in Insects to Bacillus thuringiensis Cry Toxins, Using Toxins that do not Require the Cadherin Receptor” (patent numbers: CA2690188A1, CN101730712A, EP2184293A2, EP2184293A4, EP2184293B1, WO2008150150A2, and WO2008150150A3). J.A.F. is coauthor of a patent “Cadherin Receptor Peptide for Potentiating Bt Biopesticides” (patent numbers: US20090175974A1, US8354371, WO2009067487A2, WO2009067487A3). The Agricultural Biotechnology Stewardship Technical Committee (a consortium of biotechnology companies), BASF, Bayer CropScience, Corteva, and Syngenta did not provide funding to support this work but have funded other work by the authors.