Resistance to Bacillus thuringiensis Cry1Ac toxin requires mutations in two Plutella xylostella ATP-binding cassette transporter paralogs

The diamondback moth, Plutella xylostella, is a cosmopolitan pest and the first species to develop field resistance to toxins from the gram-positive bacterium Bacillus thuringiensis (Bt). Although previous work has suggested that mutations of ATP-binding cassette transporter subfamily C2 (ABCC2) or C3 (ABCC3) genes can confer Cry1Ac resistance, here we reveal that P. xylostella requires combined mutations in both PxABCC2 and PxABCC3 to achieve high-level Cry1Ac resistance, rather than simply a mutation of either gene. We identified natural mutations of PxABCC2 and PxABCC3 that concurrently occurred in a Cry1Ac-resistant strain (Cry1S1000) of P. xylostella, with a mutation (RA2) causing the mis-splicing of PxABCC2 and another mutation (RA3) leading to the premature termination of PxABCC3. Genetic linkage analysis showed that RA2 and RA3 were tightly linked to Cry1Ac resistance. Introgression of RA2 and RA3 enabled a susceptible strain (G88) of P. xylostella to obtain high resistance to Cry1Ac, confirming that these genes confer resistance. To further support the role of PxABCC2 and PxABCC3 in Cry1Ac resistance, frameshift mutations were introduced into PxABCC2 and PxABCC3 singly and in combination in the G88 strain with CRISPR/Cas9 mediated mutagenesis. Bioassays of CRISPR-based mutant strains, plus genetic complementation tests, demonstrated that the deletion of PxABCC2 or PxABCC3 alone provided < 4-fold tolerance to Cry1Ac, while disruption of both genes together conferred >8,000-fold resistance to Cry1Ac, suggesting the redundant/complementary roles of PxABCC2 and PxABCC3. This work advances our understanding of Bt resistance in P. xylostella by demonstrating mutations within both PxABCC2 and PxABCC3 genes are required for high-level Cry1Ac resistance.

Introduction Bacillus thuringiensis (Bt) is a ubiquitous gram-positive bacterium that can produce insecticidal protein toxins used for the control of many lepidopteran, coleopteran and dipteran pests [1,2]. Bt toxins often show high insect specificity and are safe for non-target organisms [1], however, like many synthetic agrochemicals, evolution of resistance among pests is threatening their efficacy [3]. By 2019, practical field-evolved resistance to Bt has been documented in at least 22 cases and nine insect species [3][4][5][6][7]. Complete understanding of the mechanisms causing Bt resistance remains limited, although mutations or expression changes in toxinbinding proteins, such as cadherin-like protein, alkaline phosphatase (ALP), aminopeptidase N (APN), and ABC transporters, appear to play important roles in mediating the resistance level [8,9].
The diamondback moth, Plutella xylostella, is a notorious pest of cruciferous crops worldwide. Field resistance to Bt spray formulas first arose in this species within a population from Hawaii, USA [5,26]. Baxter et al. [17] have identified a 30-bp deletion in ABCC2 within a Cry1Ac-resistant strain (NO-QAGE) derived from this Hawaiian population. Reduced expressions in ABCC2 and ABCC3 have also been observed in resistant P. xylostella strains from other countries, including the DBM1Ac-R strain collected from Loxahatchee (Florida, USA) [22]. A recent gene-editing-based investigation provides further evidence of the roles of ABCC2 and ABCC3 in Bt resistance: knockout of either of the two P. xylostella genes, PxABCC2 and PxABCC3, confers high levels of resistance to Bt Cry1Ac (724 and 413 folds compared to the reference strain DBM1Ac-S) [16]. However, these resistance levels to Cry1Ac were more than 4,000-fold lower than a selected resistant strain [16].
In this study, we tested the hypothesis that mutations in both P. xylostella ABCC2 and ABCC3 are required to achieve a high-level Cry1Ac resistance in natural populations, while mutations in either gene confer low-level resistance. To do so, the sequences of PxABCC2 and PxABCC3 were compared between a Cry1Ac-resistance strain (Cry1S1000) and a susceptible strain (G88) of P. xylostella to identify the potential mutations. We then tested the genetic linkage of these two genes with Cry1Ac resistance and analyzed their contribution to Bt resistance by introgressing PxABCC2 and PxABCC3 from Cry1S1000 into G88. Finally, we introduced mutations into these ABCC genes separately and together in a reference strain with the CRISPR/Cas9 system, and then assessed their capacity to confer resistance to Cry1Ac using bioassays and genetic complementation tests.

Resistance of the Cry1S1000 strain to Cry1Ac protoxin
Bioassays performed with artificial diet overlay assays indicated that the concentration of Cry1Ac protoxin capable of killing 50% of resistant Cry1S1000 larvae (LC 50 ) was higher than 1,000 μg/ml (S1 Table). The Cry1S1000 strain was > 8,000 fold more resistant to Cry1Ac than the susceptible reference, G88 (S1 Table). In subsequent generations without Bt selection, Cry1S1000 maintained stable resistance to Cry1Ac (LC 50 > 1,000 μg/ml) and showed 100% survival at a diagnostic concentration of 0.5 μg/ml of Cry1Ac protoxin (S2 Table).

ABCC2 and ABCC3 sequence comparison between P. xylostella strains
Full-length PxABCC2 transcripts were sequenced using midgut cDNA template generated from Bt-susceptible G88 larvae. The coding sequence was 4,044 bp and encoded a predicted 1,347 aa protein (S1 and S2 Figs) that displayed a typical structure among known lepidopteran ABCCs [11], including two transmembrane domains (TMD1 and TMD2), each with six transmembrane helices, and two nucleotide-binding domains (NBD1 and NBD2) ( Fig 1A) containing conserved Walker A, B and C motifs (S2 Fig).
Five independent PxABCC2 transcript variants from the resistant Cry1S1000 strain were found to contain a range of insertions or deletions, however, all lacked exon 7 (ABCC2_R1-5 in Fig 1B, Table 1, S1 Fig). Two isoforms (ABCC2_R1 and ABCC2_R5) had in-frame deletions in exon 5 (Fig 1B), disrupting the portion of PxABCC2 protein from intracellular loop 2 to NBD1 ( Fig 1A,  All Cry1S1000 PxABCC2 isoforms lacked exon 7, however, genomic DNA amplicon sequencing revealed the coding sequence was present in the genome and not simply a chromosomal deletion. A point mutation (G>T) in the 3 0 splice junction of intron 6 was identified, which would disrupt pre-mRNA processing and prevent the incorporation of exon 7 into the mature mRNA transcript. Hereafter, we refer to this genomic DNA mutation as "R A2 " (Fig 1C  and S3 Fig).
Alignments between gDNA and cDNA sequences also found an 8-bp deletion in ABCC2_R1 was caused by an alternative 5 0 splice site in intron 5, and a 29-bp deletion in ABCC2_R2 was caused by an alternative 3 0 splice site in intron 14 (Table 1, S3 and S4 Figs). The insertions were found in two other isoforms, which were caused by the retention of introns in the mature mRNA (intron 12 in ABCC2_R3 and introns 11, 15 and 16 in ABCC2_R4, S4-S6 Figs). Finally, an 804-bp deletion in ABCC2_R5 was caused through skipping exons 6-10 ( Fig 1B).
Intron 6 varied in size between G88 (581 bp, S A2 ) and Cry1S1000 (429 bp, R A2 ), which enabled development of an allele-specific PCR assay (AS-PCR) capable of quickly differentiating between susceptible and resistant alleles (S7 Fig). PxABCC3 cDNA from the G88 (ABCC3_S) contained an open reading frame (ORF) of 4,047 bp and encoded a predicted protein of 1,348 aa (S8 and S9 Figs). Similar to PxABCC2,

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella the PxABCC3 protein also consisted of two transmembrane domains (TMD1 and TMD2) and two nucleotide-binding domains (NBD1 and NBD2) (Fig 2A and S9 Fig). A point mutation (C>T) identified at nucleotide 2131 in Cry1S1000 PxABCC3 cDNA (ABCC3_R) (Fig 2B and  S8 Fig) introduced a premature stop codon just prior to transmembrane helix 7, truncating the protein (Fig 2A and S9 Fig). This allele was referred as "R A3 ", and the susceptible PxABCC3 allele "S A3 ". Size variation was again used to develop an AS-PCR in PxABCC3 using primers spanning intron 13 (480 bp in G88 and 546 bp in Cry1S1000) (S10 Fig).

Genetic linkage between R A2 /R A3 and Cry1Ac resistance
We tested for genetic linkage between R A2 /R A3 alleles and Cry1Ac resistance using 10 backcross families, which were obtained from single-pair crosses between Cry1S1000 (RR) and an F 1 (RS, from Cry1S1000♀ × G88♂) ( Fig 3A). Bioassays were performed on 50 larvae from each of 10 backcross families (n = 500) and 224 survived after exposed to 0.5 μg/ml Cry1Ac protoxin. A total of 447 untreated control larvae were also collected (S3 Table).
We first genotyped PxABCC2 and PxABCC3 in all cross parents using AS-PCRs (Fig 3B  and 3C) and confirmed Cry1S1000 individuals were homozygous for the R A2 and R A3 alleles and only detected S A2 and S A3 alleles among G88 individuals. All F 1 progeny from Cry1S1000 × G88 mating were heterozygous for the R A2 /S A2 and R A3 /S A3 alleles (Fig 3B and  3C). Among untreated control backcross progeny, we expected half the individuals to be heterozygous (R A2 /S A2 , R A3 /S A3 ) and half to be homozygous for ABCC mutations (R A2 /R A2 , R A3 / R A3 ) and observed numbers did not significantly differ (216 heterozygous, 231 homozygous, Fisher's exact test P > 0.41 for each control family) (S3 Table). Among insecticide treated larvae, no heterozygous R A2 /S A2 , R A3 /S A3 individuals were detected, indicating the Cry1Ac discriminating dose was sufficient to kill carriers of ABCC2 and ABCC3 G88 alleles (Fisher's exact test for each treated family, P < 0.0005) (S3 Table). Therefore, Cry1Ac resistance was genetically linked to R A2 and R A3 alleles in the Cry1S1000 strain.

Introgression of R A2 /R A3 alleles into the genetic background of G88
To eliminate potential differences in fitness between strains, and assess whether loci that were not linked to ABCC mutations contributed to Cry1Ac resistance, R A2 and R A3 alleles were introgressed into G88. Cry1S1000 males were crossed to G88 females, then F 1 progeny backcrossed to G88 for a further six generations. Selection of the R A2 allele was performed in each generation backcrossing using AS-PCR. Random mating between progeny of backcross six

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella with heterozygous genotypes (R A2 S A2 ) were used to establish a near-isogenic line to the G88 strain (G88-R A2 ) that was homozygous for both the R A2 and R A3 alleles (S4 Table). Diet bioassays confirmed the G88-R A2 strain was equally resistant to Cry1S1000 (>1,000 μg/ml Cry1Ac protoxin, S1 Table).
Reciprocal crosses between resistant strains (Cry1S1000 and G88-R A2 ) and G88 were performed and susceptibility of F 1 progeny to the diagnostic concentration (0.5 μg/ml) of Cry1Ac protoxin was assessed. The mass crosses between each of the two resistant strains with G88 resulted in no survival of F 1 progeny (i.e., Cry1S1000♀ × G88♂, G88♀ × Cry1S1000♂, G88-R A2 ♀ × G88♂, and G88♀ × G88-R A2 ♂) (S5 Table). This indicated that the Cry1S1000 and G88-R A2 strains exhibited autosomal and completely recessive (h = 0) resistance to Cry1Ac at the diagnostic concentration. Third-instar larval progeny from ten single-pair crosses between strain. The PxABCC3 protein is predicted to contain two transmembrane domains with 6 transmembrane helices (dark green cylinders) for TMD1 and 6 transmembrane helices (light blue cylinders) for TMD2, and two nucleotidebinding domains (NBD1 and NBD2). Transmembrane helices, except for TM1 and TM2, were predicted by Phobius. The dotted lines, TMD2 in light blue and NBD2 in yellow represent the predicted region missing in the PxABCC3 from Cry1S1000. (B) The partial cDNA sequences and the deduced protein sequences from the G88 and Cry1S1000 strains show the point mutation of C2131T (the letter in red) in the Cry1S1000 strain that leads to the premature termination of PxABCC3. Amino acid residues highlighted in light blue are the portion of transmembrane helix 7. https://doi.org/10.1371/journal.ppat.1008697.g002

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella Cry1S1000 and G88-R A2 were assayed and showed 100% survival after a 5-day bioassay (n = 385), confirming the strains shared a common, recessive resistance locus (Index of commonality = 1) (S5 Table).

Mutagenesis of PxABCC2 and PxABCC3 mediated by CRISPR/Cas9
To knock out PxABCC2 and PxABCC3, we injected mixtures of sgRNA and Cas9 protein (or mRNA) into fresh eggs from the wild type G88 strain. Successful site-specific mutations within PxABCC2 and PxABCC3 were identified within mosaic G 0 moths (S6 Table).
To further determine the contribution of recessive mutations in two ABCC loci to Cry1Ac resistance, we performed complementation tests for PxABCC2 and PxABCC3 between resistant strains and CRISPR-based mutants. The average survival rates when exposed to 0.5 μg/ml Cry1Ac protoxin were 11.7% (7/60) for G88-ABCC2 --1, 11.7% (0.0% -20.0%) for the F 1 progeny from single-pair crosses between G88-ABCC2 --1 and Cry1S1000, and 2.6% (0.0% -21.7%) for the F 1 progeny from single-pair crosses between G88-ABCC2 --1 and G88-R A2 (Fig 5B). The results suggested that two mutations within PxABCC2 conferred a low level of resistance to Cry1Ac. Similarly, no survival of F 1 progeny from single-pair crosses between G88-ABCC3 --1 and each of two resistant strains was observed, indicating that two mutations at the PxABCC3 locus failed to confer Cry1Ac resistance. However, the F 1 progeny produced from crossing the double mutant strain G88-A2 -A3 --1 and Cry1S1000 or G88-R A2 were highly resistant to Cry1Ac like their parent strains ( Fig 5B). These results showed that two mutations at each of PxABCC2 and PxABCC3 loci were required for high levels of resistance.

Discussion
The complex mode of action of Bt insecticides requires activated toxins to interact with receptors on the brush border membrane of the insect midgut [27]. In this study, we provide rigorous support for the duel involvement of two ABC transporter subfamily genes, PxABCC2 and PxABCC3, in mediating Bt Cry1Ac protoxin susceptibility in P. xylostella. CRISPR/Cas9-mediated mutagenesis verified that high-level resistance to Cry1Ac only occurred after introducing mutations into both PxABCC2 and PxABCC3 alleles, while mutations in one gene alone had little effect. Therefore, PxABCC2 and PxABCC3 appear to function as redundant Cry1Ac receptors for determining susceptibility of Cry1Ac in P. xylostella. This finding supports recent research by Wang et al. [25] demonstrating that the knockout of ABCC2 and ABCC3 are required for Cry1Ac resistance in H. armigera.

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella be directly caused by the R A2 splice-site mutation of intron 6. One of these transcripts, ABCC2_R1 was previously identified by Guo et al. and reported as isoform VIIa from a resistant strain (DBM1Ac-R) of P. xylostella [22], and the loss of exon 7 from sequenced transcripts was a common feature of both these studies [22].
The PxABCC2 R A2 splice-site mutation described here in Cry1S1000 is independent from another 30-bp deletion in exon 20 of another Bt resistant P. xylostella NO-QAGE strain [17]. This suggests that multiple mutations can occur in Bt-selected populations. Expressing PxABCC2 in Drosophila midgut confirmed the transmembrane protein can act as a Cry1Ac toxin receptor [21], however, mutations in PxABCC3 have never been reported in NO-QAGE nor any P. xylostella Bt resistant strain to the best of our knowledge. A premature stop codon predicted in PxABCC3 (R A3 ) of Cry1S1000 here appeared to prevent translation.
Reducing expression levels of either PxABCC2 or PxABCC3 mRNA through RNAi knockdown has previously been shown to reduce mortality in wild type P. xylostella larvae when challenged with discriminating doses of Cry1Ac protoxin [22]. Simultaneous reduction of PxABCC2, PxABCC3 and membrane-bound ALP (PxmALP) transcripts using RNAi reduces mortality more than ABCC2 or ABCC3 alone [22], providing some support for a combined role of receptors in Cry1Ac mode of action. Reducing or eliminating PxABCC2 and PxABCC3 through transcriptional regulators has also been proposed as a mechanism for Cry1A resistance development without requiring genetic mutations within the genes themselves [22].
Although many lepidopteran insects are susceptible to Cry1Ac toxins, it is currently unclear whether mutations in ABCC2 and ABCC3 are required for high-level resistance in all species. Some of the first examples with Heliothis virescens and Bombyx mori [11,12] provide functional support that ABCC2 is required for Cry1A susceptibility. The B. mori Cry1Ab/c resistant strain, "C2", was rescued to a susceptible phenotype through piggyback transformation that introduced a wild-type copy of BmABCC2 [11]. Mutations in BmABCC3 have not been reported, however, only rescue of one receptor may be required for function. CRISPR/Cas9mediated double knockout of H. armigera ABCC2 and ABCC3 confers a >15,000-fold resistance to Cry1Ac [25]. Knockout of ABCC2 in Ostrinia furnacalis causes fairly high level of resistance to Cry1F, but not Cry1A toxins [28]. In Spodoptera exigua, however, ABCC2 mutation appears to be sufficient to confer resistance to Cry1Ac [15].
Expressing ABCC2 in cell lines of both H. virescens and B. mori cause cells to swell or lyse and the effect is synergized when co-expressed with another well-characterized Bt toxin receptor, a cadherin-like protein (CaLP) [19,20]. However, the involvement of P. xylostella CaLP in Cry1Ac toxicity is still controversial [29][30][31]. Synergism between ABCC2 and CaLP has also been demonstrated in vivo. Mutations in either the Trichoplusia ni cadherin or ABCC2 only slightly reduce larval susceptibility to a variant of Cry1Ac, while deletion of both proteins confers much higher levels of resistance [32]. Deficiencies of H. virescens 12-cadherin-domain protein (HevCaLP) and ABCC2 (HevABCC2) offer higher resistance to Cry1Ac than a defect in either one [12].
The results here support redundant roles between PxABCC2 and PxABCC3 in conferring Cry1Ac susceptibility, and differ from two previous studies with diamondback moth [16,22]. In their cases, reduced expression of either PxABCC2 or PxABCC3 using RNAi decreased Cry1Ac susceptibility of P. xylostella larvae [22] and knockout of either PxABCC2 or PxABCC3 using CRISPR/Cas9 caused high levels of resistance to Cry1Ac [16]. Although we used artificial-diet bioassay and the previous studies used leaf-dip bioassay, this seems unlikely to be responsible for such differences in these results. Transcripts of ABCC2 and ABCC3 sequenced from our susceptible reference strain (G88) all contained full-length open reading frames, however, Guo et al. [22] reported a low frequency of alternatively spliced isoforms from the control strain in their study. Control strains that express different ABCC2 and ABCC3 alleles,

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella or control strains that produce different mRNA isoforms could potentially explain the discrepancies between our work and the earlier works of Guo et al. [16,22].
Generating PxABCC2 mutation in the G88 wild-type strain led to a slightly higher resistance to Cry1Ac protoxin, yet PxABCC3 mutations caused higher susceptibility. Similarly, cultured cells of B. mori and S. exigua exhibit higher Cry1Aa susceptibility when expressing ABCC2 than ABCC3 [24,33]. Contribution of ABCC2 in determining toxicity of Cry1Ac may therefore be more significant than that of ABCC3. In B. mori, the extracellular loop 1 (ECL1) in BmABCC2/3 transporters determines receptor specificity for Cry1A toxins [33]. The loop 1 Q125 residue of SfABCC2 is also the key factor for the high toxicity of Cry1Ac to Hi5 cells expressing this protein [34]. Structure-based evidence therefore is necessary to further verify the roles of ECL1 in determining the specificity of PxABCC2 and PxABCC3 for Cry1Ac toxicity, and to better address their potential synergism in terms of functional complementation or redundancy. In addition, PxABCC2 has been associated with Cry1Ac oligomerization and oligomer membrane insertion [35]. Our knockout mutants of PxABCC2, PxABCC3 or both genes provide new opportunities to test this hypothesis and investigate whether PxABCC3 plays a similar role in Cry1Ac oligomerization.
Our work indicates that P. xylostella has evolved resistance to Cry1Ac protoxin through mutations in both PxABCC2 and PxABCC3. DNA-based screening may be a useful tool for monitoring the frequency of P. xylostella field resistance to Cry1A toxins, caused by mutations in ABCC genes. However, additional allelic variants causing mutations that affect toxin binding are likely to exist in field populations, and this needs to be considered. The present finding provides strong evidence for understanding the roles of ABC transporters in Bt mode of action and the genetic mechanisms that lead to Bt resistance in P. xylostella.

Insect sample
The Plutella xylostella strains were established by Dr. Anthony M. Shelton (Cornell University, USA) and provided to the Institute of Applied Ecology, Fujian Agriculture and Forestry University, in 2016. The insecticide susceptible reference strain, Geneva 88 (G88), was collected from the New York State Agricultural Experiment Station in 1988 and maintained on artificial diet without exposure to insecticides [36]. A population collected from Loxahatchee, Florida, in 1992 was resistant to B. thuringiensis subsp. kurstaki spray formula [37]. It was subsequently reselected with transgenic broccoli that only expressed the Cry1Ac toxin [38] and referred to as Cry1Ac-R [39]. The resistant strain Cry1Ac-R was then selected using 1000 μg/ml Cry1Ac protoxin in 2016 and reared on the artificial diet for over 50 generations without further insecticide exposure. We refer to this resistant strain as Cry1S1000. Both of the P. xylostella strains were kept at 26±1˚C, 65% RH, and photoperiod of 16: 8 h (L: D). Adults were provisioned with 10% honey solution.

Bioassays
Cry1Ac protoxin was prepared from B. thuringiensis subsp. kurstaki strain HD-73 (provided by Key Laboratory of Biopesticide and Chemical Biology, Ministry of Education, Fujian Agriculture and Forestry University) as previously described by Kain et al. [40] and serially diluted in ultrapure water. An artificial diet overlay assay, similar to that described by Zhao et al. [39] and Kain et al. [40], was used to determine the susceptibility of P. xylostella larvae to Cry1Ac. Bioassays were performed on five replicates for five different concentrations of Cry1Ac protoxin, plus an insecticide free control. For each replicate, pre-warmed liquid artificial diet (5 mL) was poured into a 30 mL plastic cup, and 0.3 mL aliquot of Cry1Ac protoxin was then evenly

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella distributed over the surface of the diet (surface area � 7 cm 2 ). The assay was maintained at room temperature and after 1.5-2 hours of air-drying, ten early-molted 3 rd -instar larvae were placed in the cup. Larval mortality was recorded at day 5, and mortality rate was calibrated according to Abbott's formula [41].

Cloning and sequencing of cDNA and genomic DNA
Midguts were dissected from G88 (n = 5) and Cry1S1000 (n = 5) 4 th -instar larvae and total RNA was extracted from each individual using the TransZol Up Plus RNA Kit (TranGen, Beijing, China). cDNA synthesis was performed with the TransScript One-Step gDNA Removal and cDNA Synthesis Super Mix (TranGen, Beijing, China) according to the manufacturer's instructions. PCR amplification of full length PxABCC2 and PxABCC3 transcripts were performed in a 25-μl reaction system containing 0.5 μl of cDNA template, 0.4 μM of primer (S11 Table), 12.5 μl of 2 × Phanta Max Buffer, and 0.5 μl of Phanta Max Super-Fidelity DNA Polymerase (Vazyme, Nanjing, China). Reactions were carried out with an initial denaturation at 95˚C for 3 min and followed by 35 cycles of denaturation at 95˚C for 15 s, annealing at 55˚C for 15 s and extension at 72˚C for 5 min. After examination by agarose gel electrophoresis, the amplified products were purified using the Gel Extraction Kit (Omega, Morgan Hill, GA, USA), cloned into the pJET1.2/blunt Cloning Vector using the CloneJET PCR Cloning Kit (Thermo Scientific, Waltham, MA, USA), and sequenced by Biosune Biotech Company (Fuzhou, China).
Genomic DNA (gDNA) was isolated from each of the individual carcasses remaining after midgut dissection using the Tissue DNA Kit (Omega, Morgan Hill, GA, USA). The gDNA was used as template for PCR amplification of four PxABCC2 fragments (gDNA_PF1-PF4). Amplification reactions (25 μl) contained 50 ng of gDNA template, 0.4 μM of primer (S11 Table), 12.5 μl of 2 × Phanta Max Buffer, and 0.5 μl of Phanta Max Super-Fidelity DNA Polymerase (Vazyme, Nanjing, China). All fragments were amplified with the extension time adjusted according to the fragment size, then cloned and sequenced as previously described.

Allele-specific PCR assay
Based on the size variation of PxABCC2 intron 6 and PxABCC3 intron 13 between Cry1S1000 and G88, to detect resistant (R A2 or R A3 ) alleles, we used allele-specific PCR and two pairs of primers (S11 Table) that flank either R A2 or R A3 mutations. All adults that needed to be genotyped for R A2 or R A3 allele were sacrificed and gDNA was isolated individually using the Tissue DNA Kit (Omega, Morgan Hill, GA, USA). PCR amplification was performed in a 25-μl reaction system under the conditions described in section [Cloning and sequencing of cDNA and genomic DNA]. PCR products and DL1000 DNA marker (TAKARA, Dalian, China) were separated using 3% agarose gel electrophoresis to differentiate the size of DNA bands of resistant (R A2 and R A3 ) and susceptible (S A2 and S A3 ) alleles. To validate allele-specific PCR, amplified products were cloned and sequenced from Cry1S1000 (n = 5) and G88 (n = 5) as described in section [Cloning and sequencing of cDNA and genomic DNA].

Genetic linkage analysis of R A2 and R A3 with Cry1Ac resistance
To test for genetic linkage between Cry1Ac resistance in the Cry1S1000 strain and R A2 and R A3 , a single-pair cross was performed between a Cry1S1000 female and G88 male to generate F 1 progeny. Ten single pair backcrosses (BC) were performed by crossing an F 1 male and Cry1S1000 female (BCa1-BCa5) or an F 1 female and Cry1S1000 male (BCb1-BCb5). For each backcross family, 36-48 3 rd -instar larvae were reared on the artificial diet (control), while 50 larvae were reared on artificial diet supplemented with 0.5 μg/ml of Cry1Ac protoxin for

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella five days. Mortality was recorded and all surviving larvae were allowed to complete development before gDNA was isolated.

Introgression of the R A2 and R A3 alleles from Cry1S1000 into G88
The Cry1S1000 Cry1Ac resistance locus was introgressed into the genetic background of G88. First, males from Cry1S1000 (R A2 R A3 /R A2 R A3 ) were mass crossed with G88 females (S A2 S A3 / S A2 S A3 ) to produce heterozygous F 1 's (R A2 R A3 /S A2 S A3 ) and the male progeny were then mass crossed with G88 females to generate BC1 progeny. BC1 males were individually crossed with G88 females (single-pair mating) to generate BC2 progeny and sacrificed after mating for genotyping using allele-specific PCR assays. Only lines containing an R A2 allele (R A3 was not genotyped during introgression crosses as the two genes are tightly linked) were retained. Additional four single-pair backcrosses (BC3-BC6) and corresponding molecular assays were performed using the same method of backcrossing between BC1 males and G88 females. Subsequently, BC6 siblings were crossed in single pairs to generate BC6F 1 progeny and lines containing heterozygous parents (R A2 S A2 ) were retained after genotyping. Similarly, BC6F 1 siblings were crossed (single-pair mating) to generate BC6F 2 progeny and genotyped. Only lines containing homozygous parents (R A2 R A2 ) were retained and constituted a homozygous strain of G88-R A2 (S16 Fig). Finally, the G88-R A2 strain was genotyped to confirm it was fixed for the R A3 allele.

Determination of inheritance mode of Cry1Ac resistance
Reciprocal crosses were made between each of two resistant strains (Cry1S1000 and G88-R A2 ) and G88 to generate F 1 heterozygotes using 15 virgin females from one strain and 15 males from the other. Bioassays were performed using 0.5 μg/ml Cry1Ac protoxin on Cry1S1000, G88-R A2 , G88 and their F 1 progeny from reciprocal crosses. To evaluate maternal effects and sex linkage of resistance in Cry1S1000 and G88-R A2 , we compared the survival of F 1 progeny from reciprocal crosses between Cry1S1000 and G88, and between G88-R A2 and G88. Dominance (h), ranging from 0 (completely recessive resistance) to 1 (completely dominant resistance), was estimated based on the survival of each resistant strain (Cry1S1000 and G88-R A2 ), G88 and their F 1 progeny using the single-concentration method [42].

Preparation of sgRNA and Cas9
sgRNA target sites containing 5'-N20NGG-3' (with the PAM sequence underlined) were selected in PxABCC2 exons 1, 3, and 20 and in PxABCC3 exon 1. A template for in vitro transcription of sgRNA's was obtained using PCR that included two oligonucleotides. The first contained the T7 polymerase binding site and sgRNA target sequence indicated by twenty underlined N's (5 0 -TAATACGACTCACTATAGGNNNNNNNNNNNNNNNNNNNNGT TTTAGAGCTAGAAATAGCAAGTTAA-3 0 ), and the second contained the remaining sequence common to all sgRNAs (5 0 -AAAAGCACCGACTCGGTGCCACTTTTTCAAGTT GATAACGGAC TAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAA-3 0 ). All sgRNA target sequences for PxABCC2 and PxABCC3 are listed in S12 Table. PCR amplification was performed in a 50-μl reaction system containing 0.4 μM of primer, 5 μl of 10× PCR buffer for KOD-Plus-Neo, 5 μl of 2 mM dNTPs, 3 μl of 25 mM MgSO 4 , and 1 μl of KOD-Plus-Neo (TOYOBO, Osaka, Japan) with the following program: 98˚C for 2 min, 35 cycles at 98˚C for 10 s, 55˚C for 30 s, and 68˚C for 30 s. The PCR products were examined by agarose gel electrophoresis and purified using the Gel Extraction Kit (Omega, Morgan Hill, GA, USA). sgRNAs were generated via in vitro transcription using the HiScribe T7 Quick High Yield RNA

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella Synthesis Kit (New England Biolabs, Ipswich, MA, USA) and purified through phenol-chloroform extraction and ethanol precipitation. sgRNAs were then stored at -80˚C until usage.

Microinjection of P. xylostella embryos
A parafilm sheet (5 cm × 15 cm) precoated with dry radish seedling powder was provided to G88 females for 15-20 minutes to facilitate egg collection. The parafilm sheet was retrieved and fixed to a glass slide to inject the embryos with a mixture containing Cas9 (300 ng/μl of Cas9 mRNA or 600 ng/μl of Cas9 protein) and 100 ng/μl of sgRNA. Injections were performed within 1-2 hours of lay egg collection using the IM 300 Microinjector (Narishige, Tokyo, Japan) mounted on the SZX16 Stereo Microscope (Olympus, Tokyo, Japan). Injected embryos were incubated without light at 26±1˚C in a Petri dish containing a moist tissue paper until hatching. Hatched 1 st -instar larvae were transferred to the artificial diet (representing the G 0 ).

Screening and development of homozygous mutant strains
Virgin G 0 adults that survived microinjection were individually sexed and coupled with virgin G88 adults for single-pair mating. After successful oviposition, G 0 individuals were sacrificed and gDNA isolated. PCR amplicons spanning sgRNA target sites were generated then sequenced from the G 0 's, and only lines containing mutations were retained. G 1 siblings were crossed in single pairs to produce G 2 progeny and only lines that contain heterozygous parents sharing the same allelic mutation based on molecular identification were retained. Finally, G 2 siblings were crossed (single-pair mating) again to generate G 3 progeny and genotyped. Only lines containing homozygous mutant parents were kept to generate the homozygous mutant strains. If homozygous individuals could not be obtained from the G 2 progeny, additional sibling crosses were performed (S17 Fig). Identification of CRISPR/Cas9 mediated mutations first required isolating gDNA from whole adults after mating using the Tissue DNA Kit (Omega, Morgan Hill, GA, USA). Amplicons spanning the sgRNA target sites for PxABCC2 and PxABCC3 were generated using primers listed in S11 Table. PCR amplification was performed as described in section [Cloning and sequencing of cDNA and genomic DNA] and products were Sanger sequenced (Biosune Biotech Company, Fuzhou, China).

Preparation of midgut brush border membrane vesicles (BBMV)
BBMV were prepared using the differential magnesium precipitation method [44]. The 4 thinstar larvae were dissected longitudinally in cold buffer A (300 mM mannitol, 5 mM EGTA, 17 mM Tris-HCl, pH 7.5) on ice to isolate the midgut epithelium. For each strain analyzed, 10 dissected midguts were pooled and homogenized in 750 μl of buffer A containing 1 mM PMSF using the TissueLyser II mixer (Qiagen, Dusseldorf, Germany) and a 3 mm stainless-steel bead at 22 Hz for two 1-min periods, separated by a 1-min cooling interval on ice. Subsequently, an equal volume of 24 mM MgCl 2 solution was added to the midgut homogenate and mixed. The mixture was incubated on ice for 15 min, and then centrifuged at 2,500 g for 15 min at 4˚C. The supernatant was collected and stored on ice. A second extraction was performed by resuspending the pellet using half the volumes of cold buffer A solution, then the same process as the first extraction with equal volume of 24 mM MgCl 2 . The second 2,500 g supernatant was combined with the first one and centrifuged at 30,000 g for 30 min at 4˚C. The final pellet was resuspended in 50 μl of cold half-strength buffer A. Protein concentrations were measured by the BCA assay kit (Solarbio, Beijing, China) with bovine serum albumin (BSA) as a standard.
PxABCC2 protein was further confirmed by nano-liquid chromatography-tandem mass spectrometry (nano-LC-MS/MS). Since PxABCC3 lacked primary antibody for the western blotting, its protein from the midgut BBMV was identified by nano-LC-MS/MS. Total BBMV proteins (30 μg) were separated on 7.5% SDS-PAGE. The regions expected to contain PxABCC2 and PxABCC3 proteins (~150 kDa) were excised from Coomassie blue stained gels, preserved in PBS buffer and shipped with ice packs to Huada Protein Research Center (Shenzhen, China) for tryptic digestion and analysis by nano-LC-MS/MS.

Allelic complementation test
Allelic complementation tests were performed between Cry1S1000 and introgression strain G88-R A2 using 10 single-pair reciprocal crosses and challenged their F 1 progeny (25 to 50 larvae per family) with 0.5 μg/ml Cry1Ac protoxin. To determine whether the homozygous mutations in PxABCC2 or PxABCC3 alone or combined together confer Cry1Ac resistance, we performed 8 single-pair reciprocal crosses between three CRISPR/Cas9 lines and Cry1S1000, G88-R A2 or G88. Subsequently, the F 1 progeny from each of these 8 single-pair crosses (20 to 45 larvae per family), along with six parental strains (three CRISPR/Cas9 lines, Cry1S1000, G88-R A2 and G88; 50 to 60 larvae per strain), were assayed with 0.5 μg/ml Cry1Ac protoxin. We then used the surviving individuals to calculate the index of commonality (C), which varies from close to or < 0 (resistance conferred by alleles at different loci) to close to or > 1 (resistance conferred by alleles at a shared locus or loci) [45].
In another additional experiment, each of three CRISPR/Cas9 mutant lines were mass crossed with G88-R A2 (mass mating, 15 pairs) to generate F 1 progeny. These F 1 progeny individuals were tested with at least five concentrations of Cry1Ac protoxin to get the LC 50 using artificial diet overlay assay as described in section [Bioassays].

Data analyses
All statistical analyses were performed using IBM SPSS Statistics 22. LC 50 values for Cry1Ac larval bioassays were calculated using probit analysis. Significant differences in the LC 50 values were defined by non-overlapping 95% confidence limits [46]. Fisher's exact test was used to

PLOS PATHOGENS
Genetic basis of Bt resistance in Plutella xylostella determine significant differences between observed and expected genotypes among genetic crosses.
Supporting information S1  Table. A total of 1,179 peptides identified from the G88-ABCC2 --1 mutant strain. (XLS) S8 Table. A total of 2,553 peptides identified from the G88-ABCC3 --1 mutant strain. (XLS) S9 Table. A total of 1,968 peptides identified from the G88-ABCC3 --2 mutant strain. (XLS) S10 Table.  peptides (red arrows) specific to PxABCC2 identified from the G88 strain. Black arrow indicates where the protein of G88-A2 -A3 --1 is expected to be truncated. The predicted tryptic fragments (no detection from the mutant strain using nano-LC-MS/MS analysis) for expected truncated PxABCC2 proteins from G88-A2 -A3 --1 are indicated with blue arrows. Numbers indicate the position of amino acid residues. (D) Details of the 40 peptides specific to PxABCC3 identified from the G88 strain. None of PxABCC3 peptides were detected from the G88-A2 -A3 --1 strain. (E) Map of the full-length PxABCC3 protein showing the position of 40 peptides (red arrows) specific to PxABCC3 identified from the G88 strain. Black arrow indicates where the protein of G88-A2 -A3 --1 is expected to be truncated. The predicted tryptic fragments (no detection from the mutant strain using nano-LC-MS/MS analysis) for expected truncated PxABCC3 proteins from G88-A2 -A3 --1 are indicated with blue arrows. Numbers indicate the position of amino acid residues. (TIF) S16 Fig. Schematic strategy of continuous backcrossing for introgression of R A2 and R A3 alleles from the resistant Cry1S1000 strain into the genetic background of the susceptible G88 strain of P. xylostella. Square and circle represent a male adult and a female adult, respectively. BC1-6: 1-6 generations of backcrossing. BC6F1 and BC6F2 are two additional generations for sibling crosses in single pair to generate a homozygous strain with R A2 R A2 (G88-R A2 ). S A2 S A2 /R A2 S A2 underlined with the red line represent the filtered individuals of each generation after molecular identification using AS-PCR. (TIF) S17 Fig. Schematic strategy for screening of homozygous mutant strains. Each rectangle represents a paired female and male. Red rectangles represent the single-pair families kept for the following generations whose parent(s) harbored the desire frameshift mutations as determined by PCR and direct sequencing. "A" shown in the rectangles is the wild type allele; "a 1 ", "a 2 " and "a 3 " are mutant alleles with different type of indel mutations.