Insecticide resistance reduces the profitability of insect-resistant rice cultivars

Graphical abstract

Highlights

• •Imidacloprid-resistant N. lugens population rapidly adapted to resistant rice IR36.
• •Insecticide resistance-related genes nAChR-7-like and CYP4C61 contributed to the adaptation to resistant rice in N. lugens.
• •One avirulent/susceptible genotype and two virulent/resistant genotypes could be inferred from the corresponding alleles of genes nAChR-7-like and CYP4C61.
• •The virulent/resistant genotypes of N. lugens already existed in the wild in China with increasing frequencies along with insecticide usage.

Abstract

Introduction

Preventing crop yield loss caused by pests is critical for global agricultural production. Agricultural pest control has largely relied on chemical pesticides. The interaction between insecticide resistance and the adaptation of herbivorous pests to host plants may represent an emerging threat to future food security.

Objectives

This study aims to unveil genetic evidence for the reduction in the profitability of resistant cultivars derived from insecticide resistance in target pest insects.

Methods

An experimental evolution system encompassing resistant rice and its major monophagous pest, the brown planthopper Nilaparvata lugens, was constructed. Whole genome resequencing and selective sweep analysis were utilized to identify the candidate gene loci related to the adaptation. RNA interference and induced expression assay were conducted to validate the function of the candidate loci.

Results

We found that the imidacloprid-resistant population of N. lugens rapidly adapted to resistant rice IR36. Gene loci related to imidacloprid resistance may contribute to this phenomenon. Multiple alleles in the nicotinic acetylcholine receptor (nAChR)-7-like and P450 CYP4C61 were significantly correlated with changes in virulence to IR36 rice and insecticide resistance of N. lugens. One avirulent/susceptible genotype and two virulent/resistant genotypes could be inferred from the corresponding alleles. Importantly, we found that the virulent/resistant genotypes already exist in the wild in China, exhibiting increasing frequencies along with insecticide usage. We validated the relevance of these genotypes and the virulence to three more resistant rice cultivars. Knockdown of the above two genes in N. lugens significantly decreased both the resistance to imidacloprid and the virulence towards resistant rice.

Conclusion

Our findings provide direct genetic evidence to the eco-evolutionary consequence of insecticide resistance, and suggest an urgent need for the implementation of predictably sustainable pest management.

Introduction

Agriculture is the foundation of the industry and economy of self-sustaining and developing nations. It provides raw material for food and feed industries, which aim to meet the increasing demand of a rapidly growing global population of nearly 7.5 billion people [1], [2]. Modern agricultural practices encounter multiple challenges, such as loss of soil fertility, fluctuating climatic factors, and in particular, increasing pathogen and pest attacks. Crop pests pose great threats to global agricultural production [3]. Annual yield losses accounting for 8% in wheat, 10% in maize, and 15% in rice production have been incurred due to feeding and pathogen transmission by arthropod pests worldwide [4], and these losses are estimated to be more severe in food security hotspots with rapidly growing populations [5]. Therefore, efficient pest control is urgently needed in agroecological systems to support global human population growth.

Since the introduction of synthetic pesticides in the 1940s, the control of agricultural pests has largely relied on their use, reducing yield losses due to insects by 39% on average [4]. However, the side effects of this pest control strategy on the environment and human health soon became a critical issue [6]. In addition, excessive pesticide application in turn increases resistance in pest populations and often leads to a resurgence [7]. Statistics indicate that over 600 pest species have developed resistance to one or more pesticides [8].

In recent decades, sustainable means to prevent crop losses, such as developing insect-resistant cultivars that possess specific physiological or biochemical properties against pests, have been proposed and implemented, with the aim of fulfilling the increasing demand for food in a resource-conserving and environmentally friendly manner [9], [10], [11]. However, pesticide usage may accelerate the adaptation of pests to resistant cultivars and lead to the failure of their deployment [12], [13]. Actually, plants can generate diverse defensive compounds to protect themselves from pest damage, and most of the plant allelochemicals operate with similar modes of action to synthetic pesticides [14]. Thus, it has been suggested that the evolution of pest resistance to toxic pesticides may interact with the capacity of pests to cope with host plant defenses [14], [15], and lead to the consequence of cross-resistance between host plant adaptation and pesticide resistance development [16]; however, direct genetic evidence is lacking.

Exposure to insecticides poses selective pressure on insect individuals carrying genetic variations that help them to resist insecticides [17]. We hypothesize that some of these loci may also be beneficial to herbivores in adaptation to resistant cultivars. To test this hypothesis, we generated an experimental evolution system encompassing resistant rice and its major monophagous pest, the brown planthopper Nilaparvata lugens. We showed that a laboratory N. lugens population resistant to the insecticide imidacloprid, a widely used neonicotinoid for pest control, could rapidly adapt to the resistant rice variety IR36 that possesses the insect-resistance gene bph2. Whole genome resequencing and selective sweep analyses were performed to unveil the genetic basis responsible for the phenotypic changes. Furthermore, a field survey was conducted to test for the relevance between these genetic factors and the threats of this insect pest species in the wild. A causal relationship between the candidate loci in regulating both phenotypes was confirmed.

Materials and methods

Rice varieties and insects

Two rice varieties were used in this study. Variety IR36, which carries the bph2 resistance gene, was used as the resistant rice to conduct the experimental evolution. Variety TN1, which does not express any resistance gene and is susceptible to N. lugens, was used as the control rice. Three additional resistant rice varieties, Rathu Heenati (R.H, mainly carrying the Bph3 resistance gene), R476 (carrying the Bph14 resistance gene), and ASD7 (carrying multiple resistance genes), were used for testing the potential threat of virulent N. lugens genotypes.

The insecticide-resistant N. lugens population was reared on TN1 rice in a controlled environment (25 ± 2 °C, 14 h light/10 h dark photoperiod) at Guangdong Academy of Agricultural Sciences without contact with any resistant cultivar for >15 y. For selection, 1500 mated females from the initial population (F0) were randomly divided into two subpopulations, and fed on IR36 (P-IR36) and TN1 (P-TN1) (Fig. 1A). The offspring of these females were deemed F1 and continued to be maintained on the corresponding rice varieties for >28 generations. No <1000 individuals contributed to each generation, and each generation was kept discrete from the previous generation. Bioassays of virulence and insecticide resistance were performed on generations F0, F5, F15, and F25 for each subpopulation, and the fitness of each subpopulation was assessed for generations F1 and F28 (Fig. S1). As the early generations of P-IR36 showed low fitness and the bioassays would consume plenty of nymphs, and to ensure that a sufficient number of insect individuals was available for the next generation, samples for genotyping were only collected from generations F0, F11, F16, and F21; the F0 and F21 generation samples were also used for genomic resequencing.

Fig. 1.

Adaptation of N. lugens populations to a resistant rice variety (IR36) and its genetic architecture. (A) Scheme used for the selection of adaptive lineages to the resistant rice variety IR36. The starting population (F0) was randomly divided into two subpopulations, and fed on IR36 (resistant rice variety, P-IR36) and TN1 (susceptible rice variety as control, P-TN1), respectively, and the degree of virulence was determined according to the criteria proposed by the International Rice Research Institute. (B) Comparison of the reproductive fitness of the selected populations. Reproductive fitness was indicated by oviposition per female, and the data are mean values ± SEM (n = 10). Values sharing the same letter are not significantly different at P < 0.05 (one-way ANOVA and Duncan’s multiple range test). (C) Distributions of population branch statistics (PBS) in P-IR36 were calculated over 20 kb sliding windows with 2 kb steps. Outlier genomic regions are shown as the points above the dashed line (Z-test, P < 0.01). (D) GO enrichment of genes within the outlier loci was estimated using Fisher’s exact test (adjusted P-value < 0.05). (E) Resistance ratio to the insecticide imidacloprid of P-IR36 versus P-TN1 at different generations.

For temporal studies, we selected insect samples that were previously collected from fields in three provinces (Guangdong, Guizhou, and Zhejiang) of China over the past few decades. Field populations from Guangdong were collected from Shaoguan City in 2009 and 2017; field populations from Guizhou were collected from Zunyi City in 2010 and 2017; and field populations from Zhejiang were collected from Hangzhou City in 2010 and 2017.

A further three wild populations from Guangdong Province were collected from rice fields in August 2018 in Shaoguan City (GD-P1), in September 2018 in Maoming City (GD-P2), and in June 2018 in Zhongshan City (GD-P3). All these wild populations were reared on TN1 rice in a greenhouse at 25 ± 2 °C with a 14 h light/10 h dark photoperiod prior to tests.

Phenotypic bioassays

The adaptation to plant resistance was determined by the virulence of the insects. Here, we used the bulk seedling test method to evaluate the virulence of N. lugens population [18], [19]. Ten TN1 and IR36 rice seedlings at the two-leaf stage were planted in a line under a transparent cover, and each seedling was then infested with 10 s-third instar N. lugens nymphs. A positive seedling reaction was recorded when >90% of the TN1 seedlings (used as the susceptible strain) were clearly observed to wither. The damage conferred by N. lugens was then evaluated according to the criteria proposed by the International Rice Research Institute (IRRI) [20]. Four biological replicates were tested for each treatment, and the average damage level was considered as the virulence level for N. lugens population against the corresponding rice variety.

To assess insecticide resistance, we performed a bioassay and analysis as described by Pang et al. [21]. Briefly, the insecticide imidacloprid was dissolved in acetone and then diluted to a series of concentrations (300, 100, 33.3, 11.1, 3.7, 1.23, and 0.41 mg/L). Rice stems rooted in culture cups were then dipped into these insecticide solutions for 30 s. After the rice stems had been air-dried, 20 third-instar nymphs were placed into each culture cup. Each treatment was performed in triplicate. After 96 h, nymph mortality was assessed and used for calculating the lethal concentration–mortality probit regression line (LC-P line).

Here, the fitness of P-IR36 and P-TN1 was determined from the life-history traits of the corresponding insects [22]. For testing, multiple mated female adults of generations F0 and F27 were randomly chosen from each of the two subpopulations and maintained in two separate cages to lay eggs on their corresponding rice varieties. When the eggs started to hatch, we transferred the newly hatched nymphs from each subpopulation (F1 and F28 generations) to >10 vials (20 nymphs/vial) containing fresh rice plants to record their growth, development, and nymph survival rate. When the adults emerged, we randomly paired 10 virgin females and 10 virgin males from the 10 vials and transferred each of the 10 pairs from each subpopulation to a plastic cage containing fresh rice plants for oviposition. The total number of offspring from each pair was counted and recorded after 15 days.

Pool sequencing

As the DNA content in a single N. lugens individual is very low, pooling individuals from the same lineage is an alternative way to obtain sufficient DNA levels to allow for an estimation of population allele frequencies by next-generation sequencing (NGS) [23]. Genomic DNA was extracted from multiple female individuals at F0 (n = 48), P-TN1-F21 (n = 95), and P-IR36-F21 (n = 96) by using an E.Z.N.A.® Insect DNA Kit (Omega, Norcross, GA, US), and mixed in equal quantities to create one pooled DNA sample for each group. Each pooled-DNA sample was fragmented to 350 bp using a Covaris sonicator and then used to generate Illumina libraries. Each library was sequenced on the Illumina Hiseq4000 platform at BGI-Shenzhen (Shenzhen, China), producing 150-bp paired-end reads.

Single nucleotide polymorphism (SNP) calling and selective sweep analysis

Clean reads from each sample were aligned to the N. lugens reference genome (GCA_000757685.1) using the BWA tool V.0.7.13 [24]. Picard tools V.1.119 (https://broadinstitute.github.io/picard/) was used to identify duplicate reads and create BAM indexes. The best practices guide recommended by the Genome Analysis Toolkit (GATK) was followed to call and refine the SNP variants [25], [26]. SNP sites with a genotyping quality lower than 30 and a depth <10 were removed.

To detect genomic signatures for selection, we performed selective sweep analyses using a sliding window approach (20 kb windows sliding in 2 kb steps). The population branch statistic (PBS) approach was employed due to its ability to detect incomplete selective sweeps over short divergence times [27]. Our approach was to search for genomic regions that separated the resistant rice variety-selected populations from the control samples (P-TN1 and F0). The analysis was carried out according to the formula described by Yi et al. [27]: PBS_I = (T^I-T + T^I-F0 – T^T-F0) / 2, where T^A-B is the log-transformed F_ST between populations A and B, I represents the resistant rice variety-selected population (P-IR36), and T represents the control population (P-TN1). We then Z-transformed the PBS values, and outliers with a P-value below 0.005 were selected as adaptation-associated genomic regions. The heterozygosity of each sample was also calculated based on 20 kb windows sliding in 2 kb steps. The predicted genes located in the outlier genomic regions were annotated with a BLAST search of the NCBI Non-redundant (NR) protein database. The Blast2GO software was used to obtain Gene Ontology (GO) annotation of these genes [28]. GO enrichment analysis was performed according to Fisher’s exact test with Benjamini and Hochberg adjustment.

RNA extraction and quantitative real-time PCR analysis

Total RNA was extracted from N. lugens samples using a Total RNA extract kit (Omega, Norcross, GA, USA), and was then reverse transcribed to cDNA using PrimeScript reverse transcriptase (Takara, Kyoto, Japan). The primers for the amplification of the CYP4C61, CYP4C76, and CYP4C77 genes are listed in Table S1. Quantitative real-time (qRT) PCR was performed on LightCycler480 (Roche, Indianapolis, IN, USA) using a GoTaq qPCR Master Mix (Promega, Madison, WI, USA) according to the manufacturer’s protocol. The amplification protocol was as follow: 95 ˚C for 10 min, followed by 45 cycles of 95 ℃ for 10 s, 60 ℃ for 20 s, and 72 ℃ for 20 s. The housekeeping gene β-actin was used as endogenous controls [29]. Three biological replicates of three technical replicates each were performed for each sample, and relative expression levels were calculated according to the 2^−△CT method.

CYP4C61 promoter activity assay

Fragments of the CYP4C61 promoter region were PCR-amplified from P-TN1 and P-IR36 samples respectively. The primers used for PCR are listed in Table S1. The resultant PCR products were purified using a QIAquick PCR Purification Kit (Qiagen, Hilden, Germany), double cleaved with the restriction enzymes KpnI and XhoI, and cloned into the luciferase reporter plasmid pGL3-basic (Promega), yielding the CYP4C61 promoter constructs of P-TN1 alleles and P-IR36 alleles. The obtained constructs were sequenced to validate that one contained most of the P-TN1 alleles and the other contained most of the P-IR36 alleles.

Hi5 cells were transfected with a mixture of 0.1 ng internal control plasmid (pRL-CMV vector, Promega), 0.6 μL Fugent HP (Promega), and 0.2 μg construct in Grace’s medium without FBS [21]. After 48 h, the cells were lysed in 1 × passive lysis buffer (Promega). Finally, the luciferase activities were measured using a Dual-Luciferase Assay System (Promega). All transfections were independently repeated three times and three technical replicates were measured.

SNPs genotyping and phenotypic association analysis

To genotype the candidate SNPs, we extracted genomic DNA from a single female from each lineage and used it as the template for PCR amplification. The primers used for PCR are listed in Table S1. After amplification using the KAPA2G Fast Genotyping mix according to the user’s guide (KAPA, Boston, MA, USA), the PCR products were sent for Sanger sequencing by IGE Biotechnology Ltd (Guangzhou, China). The sequencing chromatograph was analyzed using MUTATION SURVEYOR v.3.30 [30]. After genotyping, we calculated r² values between pairs of SNPs using Haploview V.4.2 to estimate the linkage disequilibrium patterns of SNPs from the candidate genes [31].

Pairwise Goeman’s Bayesian scores were used for an association analysis between phenotypic changes in the N. lugens populations and combination allele frequencies of candidate SNPs from the paired samples using the AssotesteR package in R [32]. Permutations were set to 100 for each pairwise comparison.

Genotype-based phenotypic bioassay

To estimate the population structure, a model-based method implemented in the program STRUCTURE V.2.3.4 [33] was used to infer the coancestry of individuals within both P-IR36 and P-TN1 at different generations. The STRUCTURE algorithm was run using a 10,000 burn-in period and 10,000 MCMC replicates under the admixture model. This was repeated 10 times for each K, ranging from 2 to 7. The best K value was estimated using Structure Harvester (https://taylor0.biology.ucla.edu/structureHarvester/) [34]. Based on the best K value, a cluster matching and permutation analysis was performed using CLUMPP software [35]. The outputs from CLUMPP were used directly as inputs for the DISTRUCT program for cluster visualization [36].

Three germlines with the homozygous genotypes Genotype-1, Genotype-2, and Genotype-3 were then generated from the tested N. lugens populations. Briefly, each newly emerged female adult was randomly paired with a male adult, and the pairs were transferred to a plastic cage with fresh rice plants for oviposition. Genotyping was performed on each pair of insects after 7 days. All the offspring of paired insects belonging to the same genotype (paired female and male were both Genotype-1, Genotype-2, or Genotype-3) were pooled together and considered as a homozygous germline. After the Genotype-1, Genotype-2, and Genotype-3 lines were obtained from each population, individual virulence and imidacloprid resistance were both assessed.

Individual virulence was evaluated according to the excretion of honeydew, the amount of which was weighed (to an accuracy of 0.1 mg) for every single female adult at 24 h after the start of the experiment. More than five individuals were tested for each group. For imidacloprid resistance assessment, 30 female adults were placed on rice stems dipped with 50 mg/L imidacloprid solutions. Mortality was calculated 72 h after treatment. A least three replicate experiments were conducted for the imidacloprid-resistance assessment. Statistical differences between two groups were assessed using Student’s t-test.

RNAi assay

The full coding sequences of nAChR-α-7-like, CYP4C61, CYP6AY1, and CYP6ER1 were cloned into the pMD-18T vector (Takara), and sequenced by IGE Biotechnology Ltd before double-stranded (ds) RNA synthesis. The verified plasmid was used to amplify the template for synthesizing the dsRNA. The fragment of the green fluorescent protein (GFP) gene (ACY56286) was amplified with the primers reported by Chen et al. [29] (Table S1) and used as the negative control. The dsRNAs of nAChR-α-7-like, CYP4C61, CYP6AY1, CYP6ER1, and GFP were synthesized using the T7 RiboMax Express RNAi System (Promega, Madison, WI, USA) following the manufacturer’s instructions.

A 100-nL aliquot of the purified dsRNA (1 ng/nL) was injected into each two-day-old brachypterous female adult of GD-P1 according to a previously described method [37]. The injected female adults were reared in a cage containing fresh TN1 rice at 26 ± 2 °C with 80% ± 10% humidity and a light/dark cycle of 14/10 h. At 24 h post-injection, five female adults from each of the three replicates were randomly sampled for subsequent RNA extraction and qRT-PCR detection of gene expression. After confirming the RNAi efficiency, injected female adults were used for phenotypic bioassays at 24 h post-injection. The phenotypic bioassays were conducted as described above.

Induced expression assay

The insecticide exposure experiment was performed according to the method of Pang et al. [21]. The stem of TN1 rice was dipped in 12.5 mg/L imidacloprid (approximate to the 30% lethal concentration (LC₃₀) for N. lugens population GD-P1). Brachypterous female adults of GD-P1 were introduced into either the insecticide-treated or control rice. Five individuals form each of the three replicates were collected at 0, 24, 48, and 72 h post-treatment. For the IR36 rice feeding experiment, female GD-P1 adults were starved for 4 h and then placed on TN1 rice (control) and IR36 rice. Five individuals from each of the three replicates were collected after continuous feeding for 24 and 48 h. All collected samples were used for RNA extraction and qRT-PCR detection of gene expression as described above.

Statistical analysis

The half-lethal concentration (LC₅₀) for the survival curves was calculated by probit regression analysis. Comparisons of the gene expression, promoter activity, honeydew content, and survival rate results were analyzed using Student’s t-test. P < 0.05 and P < 0.01 were considered as significant and highly significant, respectively. The reproductive fitness and induced expression across different time points were compared by one-way analysis of variance (ANOVA) and Duncan’s multiple range test. The statistical analyses were conducted using SPSS 21.0 (SPSS Inc., Chicago, IL).

Results

Insecticide-resistant N. lugens population rapidly adapts to rice resistance

We randomly divided a laboratory N. lugens population (F0, reared on the susceptible rice variety TN1 without contact with any resistant cultivar previously) into two subpopulations, which were then fed on resistant variety IR36 (P-IR36) and TN1 (P-TN1), respectively (Fig. 1A). The N. lugens population used in this study was shown to have a high level of resistance to the insecticide imidacloprid (resistance ratio ∼402.5-fold higher than the susceptibility baseline formulated by Nanjing Agricultural University, Table S2) [18]. Surprisingly, a medium virulence level (i.e., the ability to survive, develop, and damage a plant host, ranging from 1 to 9 according to the criteria proposed by the International Rice Research Institute) [20] towards the resistant rice IR36 was observed in the F0 population (virulence level ≈ 5; Fig. 1A), indicating that virulent individuals were already present in the initial population. Following this, the virulence of this population evolved to a relatively high level (virulence level > 7) after only five generations of feeding on IR36 rice (Fig. 1A). Ultimately, an extremely virulent population (virulence level 9) was obtained after continuous selection for 25 generations. The adaptation of N. lugens to this resistant rice variety was further assessed by examining the changes in the fitness of N. lugens populations. After one generation of selection (F1), a high fitness cost was found for P-IR36, resulting in a significantly longer developmental duration, a lower nymph survival rate, and a lower fecundity on IR36 rice (one-way ANOVA, n = 10, P < 0.05; Fig. 1B and Table S3) compared with those for P-TN1 on TN1 rice. However, the fitness cost was largely alleviated in P-IR36 after selection for 28 generations (one-way ANOVA, P < 0.05).

Whole genome resequencing was performed on pooled insects from the 21st generation of P-IR36, P-TN1, and the starting population (F0) to identify the genomic architecture of the adaptation. Based on the genome-wide SNP variations (approximately 32 million SNPs), we calculated the PBS [27] to identify genomic regions related to the adaptation of N. lugens to IR36 rice. A total of 8,050 putative selective sweep regions (1.78% of the whole genome) encompassing 1,524 gene loci were then obtained (Z-test, P < 0.005; Fig. 1C, Fig. S2, and Dataset S1). In contrast, only 787 regions (0.17% of the whole genome) could be obtained for P-TN1 using the same cutoff of the PBS value (Dataset S1). In addition, the heterozygosity of the putative selective sweep regions significantly declined in P-IR36 compared to the control samples (Table S4), indicating a strong signal for orientation selection. Intriguingly, these loci were significantly enriched for genes related to monooxygenase activity, extracellular ion channel activity, membrane composition, and transport (adjusted P-value ≤ 0.05; Fig. 1D). These functional categories have typically been attributed to insecticide-resistance mechanisms in insects, such as xenobiotic metabolism by monooxygenase and target insensitivity for ion channels that are involved in the imidacloprid resistance of insects [39], [40]. Indeed, multiple genes belonging to commonly known insecticide resistance-related gene families were found in the candidates (Table S5).

The levels of imidacloprid resistance in P-IR36 and P-TN1 were compared. The LC₅₀ to imidacloprid of P-IR36 was 2.07-fold higher than that of P-TN1 at F5, and the resistance ratio to P-TN1 then increased by 2.66-fold at F15 (Fig. 1E and Table S2). At F25, the resistance ratio to P-TN1 increased by 4.13-fold in P-IR36. Notably, imidacloprid resistance was strongly positively correlated with virulence towards resistant rice in the tested populations (Pearson’s coefficient of correlation [ρ] = 0.89, P = 0.007; Fig. S3). These findings revealed a noticeable interaction between insecticide resistance and adaptation to host plant resistance in this insect.

Genetic basis of the changes in virulence and insecticide resistance

We hypothesized that some of the selective sweep loci related to insecticide resistance contributed to the adaptation of the N. lugens population to resistant rice. Based on the GO enrichment data and previous studies on imidacloprid resistance in this insect species [21], [39], [41], genes encoding the nicotinic acetylcholine receptor (nAChR) and cytochrome P450 were selected for subsequent analysis.

Two nAChR genes located on adjacent loci in the same scaffold were identified (Fig. S4). The corresponding scaffold was found to be entirely under selection in P-IR36, showing lower heterozygosity and significantly higher PBS values compared with the controls (Fig. S4). Considering the potential correlation between target-site mutations and insecticide resistance, we then surveyed significant mutations in the exons of these two genes. Only four missense SNPs were identified in one of the nAChR genes and none of them were present in the conserved functional domain; meanwhile the other nAChR gene (XLOC_013900, annotated as nAChR-α-7-like) harbored 16 significant missense SNPs in its exons, of which two were located in its conserved functional domain (Table S6). Coincidentally, three of the 10 outlier P450 genes were located in an adjacent scaffold that was entirely under selection (Fig. S5). The overexpression of P450 genes is a major mechanism for metabolic resistance in pests [21], [42]. Thus, we compared the mRNA levels of these genes in P-TN1 and P-IR36. Among them, one P450 gene (CYP4C61) showed significantly higher expression levels in P-IR36 than in P-TN1 (Fig. S6). Notably, CYP4C61 harbored multiple significant SNPs in its promoter region (Table S7), which we proved were responsible for its overexpression in P-IR36 (Fig. S7). Based on these findings, we further investigated the roles of the nAChR-α-7-like and CYP4C61 genes in N. lugens.

According to the linkage disequilibrium determined by PCR-based genotyping, there were two significant SNPs in exon-5 of the nAChR-α-7-like gene (Fig. 2A, left). All of the significant SNPs in the CYP4C61 promoter were combined into four LD blocks, from which four representative SNPs were selected (Fig. 2B, left). Notably, the adaptive trait was significantly associated with the SNP combination frequencies of the nAChR-α-7-like and CYP4C61 genes in P-IR36, but not in P-TN1 (Pairwise Goeman’s Bayesian score test, P ≤ 0.01; Fig. 2A and 2B). In addition, the transition in the SNPs in the nAChR-α-7-like gene had been fixed in P-IR36 from the 11th generation to some extent (Fig. 2A, right), whereas significant transitions in the SNPs of CYP4C61 could still be observed in F21 of P-IR36 compared to F11 and F16 of this population (P ≤ 0.05) (Fig. 2B, right), suggesting that CYP4C61 may be sequentially under selection by IR36 rice until at least the 21st generation.

Fig. 2.

Identification of selective sweep loci associated with the adaptive traits in P-IR36. (A) Linkage disequilibrium (LD) analysis of single nucleotide polymorphisms (SNPs) in exon-5 of the nAChR-α-7-like gene. (B) LD analysis of SNPs in the promoter of the CYP4C61 gene. The generational trends for the allelic frequencies of two representative SNPs in nAChR-α-7-like (A) and four representative SNPs in CYP4C61 (B) are shown. The frequencies are shown according to the minor alleles in F0 compared with other generations of P-IR36. The correlation between phenotypic changes and the combination allele frequencies of mutations in nAChR-α-7-like (A) and CYP4C61 (B) are presented as heatmaps. Each square represents a comparison between paired vertical and horizontal samples. The color bar indicates the scale for the Pairwise Goeman’s Bayesian score. One star in a square means that the comparison is significant at P < 0.05, and two stars in a square means that the comparison is significant at P < 0.01.

It appeared that the genetic basis for adaptive evolution in P-IR36 was polygenic, or at least mediated by both nAChR-α-7-like and CYP4C61 genes. Therefore, we inferred the genetic structures for the P-TN1 and P-IR36 in different generations based on the combined SNPs (two representative SNPs from nAChR-α-7-like and four representative SNPs from CYP4C61) of the two loci. According to the best estimated K value (Fig. S8), a total of 210 tested individuals could be categorized into three subpopulations (Fig. 3A). Pop1 (marked by a yellow bar) comprised most of the individuals from F0, and both F11 and F21 of P-TN1, and may represent avirulent individuals. In contrast, Pop2 (marked by purple) and Pop3 (marked by blue) were more enriched in samples of P-IR36 (Fig. 3A). Additionally, Pop2 may represent virulent individuals that harbored resistant alleles of the nAChR-α-7-like gene that contributed to a larger proportion of the virulence of P-IR36 at F11; meanwhile, Pop3 may represent CYP4C61-mediated virulent individuals that were distinguished from Pop2 individuals (Fig. 3A).

Fig. 3.

Association between N. lugens phenotypes and the combined genotypes based on the nAChR-α-7-like and CYP4C61 genes. (A) Population structure of 210 individuals from two populations (P-IR36 and P-TN1) at different generations (F0, F11 and F21) estimated by the STRUCTURE software based on a K value of 3. The yellow, purple, and blue colors represent three subgroups. (B) Combined alleles of the nAChR-α-7-like and CYP4C61 genes responsible for Genotype-1, Genotype-2, and Genotype-3 individuals in N. lugens. (C) Scheme used to screen for Genotype-1, Genotype-2, and Genotype-3 lines from a specific N. lugens population. (D) Mean content of honeydew secreted by a one-day-old female N. lugens adult (n = 20) after feeding on IR36 rice for 24 h was assessed for the Genotype-1, Genotype-2, and Genotype-3 lines that were screened from P-TN1 and P-IR36. (E) Corrected survival of 30 one -day-old female N. lugens adults exposed to 50 mg/L imidacloprid for 72 h (n = 8) was assessed for the Genotype-1, Genotype-2, and Genotype-3 lines that were screened from P-TN1 and P-IR36. The data are the mean values ± SEM. Statistical analysis was performed using Student’s t-test between two groups (*P < 0.05; **P < 0.01). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

We named the genetic features in Pop1 as Genotype-1, those in Pop2 as Genotype-2, and those in Pop3 as Genotype-3 (Fig. 3B). For example, Genotype-2 mainly carried heterozygous or homozygous major alleles in the two SNPs of nAChR-α-7-like, homozygous major alleles in representative SNPs 1 and 3 and homozygous minor allele in representative SNP 4 of CYP4C61. The N. lugens lines that were respectively defined as homozygous Genotype-1, Genotype-2, and Genotype-3 were screened from P-TN1 and P-IR36 (Fig. 3C). For both populations, insects defined as Genotype-2 and Genotype-3 showed significantly higher virulence to IR36 rice, as well as higher degrees of insecticide resistance, than those defined as Genotype-1 (Fig. 3D and E). In addition, insects defined as Genotype-3 showed superior insecticide resistance to those defined as Genotype-2 in P-IR36 (P = 0.047). The contributions of the target site and metabolic resistance mechanisms to insecticide resistance differ among insect populations [43]; however, in the present study, they appeared to make similar contributions towards N. lugens virulence.

Field relevance of the genotypes and phenotypes responsible for virulence of N. lugens to resistant rice

The involvement of the nAChR-α-7-like and CYP4C61 genes and their related genotypes in the crosstalk between insecticide resistance and adaptation to resistant rice in N. lugens was validated in our experimental system. However, the field relevance of this phenomenon remains unknown because limited studies have been conducted on this aspect [14], [15]. Therefore, N. lugens populations sampled from fields in three provinces (Guangdong, Guizhou, and Zhejiang) of China in 2009/2010 and 2017/2018 were used to test the prevalence and temporal patterns of these genetic features in the wild (Fig. 4A). Remarkably, N. lugens individuals of Genotype-2 were present at relatively high proportions in these field populations (33.61–39.92%) in 2009/2010, and further increased in proportion (40.74–60.44%) in 2017/2018 (Fig. 4A). Genotype-3 was found at a relatively low proportion (<15.00%) in the different provinces over the various periods except for the populations collected from Guangdong Province in 2017 and 2018. Alternatively, the proportion of Genotype-1 in field N. lugens populations was clearly lower in 2017 than that in 2009/2010.

Fig. 4.

Prevalence of Genotype-1, Genotype-2, and Genotype-3 and their correlation with phenotypes in the wild. (A) Temporal distributions of Genotype-2 and Genotype-3 in field populations collected from three different provinces (Guangdong, Guizhou, and Zhejiang) of China. The mean honeydew content secreted by a one -day-old female N. lugens adult after feeding on (B) IR36 rice and (C) other insect-resistant varieties for 24 h was assessed for the Genotype-1, Genotype-2, and Genotype-3 lines that were screened from wild population(s) in Guangdong Province. (D) Reproductive fitness of the Genotype-1, Genotype-2, and Genotype-3 lines. Reproductive fitness was indicated by oviposition per female. The data are the mean values ± SEM. Statistical analysis was performed using Student’s t-test between two groups (n = 5 for panel B, n = 6 for panel C, and n = 10 for panel D; **P < 0.01).

According to our survey, insect-resistant rice cultivars have not been previously planted in the rice fields of Guangdong Province; rather, imidacloprid and other insecticides have been employed intensively over the past 10 years (Table S8). Thus, we further assessed the phenotypes of N. lugens lines with specific genotypes generated from isofemale lines from three wild populations (GD-P1, GD-P2, and GD-P3) in Guangdong Province in 2018 (Fig. 4A). All Genotype-2 and Genotype-3 lines showed significantly higher insecticide resistance than Genotype-1 lines, with only one exception, the case of which (Genotype-3 line to Genotype-1 line of GD-P1) showed the expected trend, but was not statistically significant (Fig. S9). Importantly, the Genotype-2 and Genotype-3 lines showed significantly higher virulence towards IR36 rice than Genotype-1 lines in all tested populations (Fig. 4B). Additionally, a certain degree of high virulence toward rice varieties carrying other resistance genes [44], [45], [46] was observed for the Genotype-2 and Genotype-3 lines of GD-P1 (Fig. 4C), implying that these genotypes may confer a potential broad-spectrum advantage to N. lugens against resistant rice cultivars.

We further measured the reproductive fitness of the three genotype lines for both the TN1 and IR36 rice varieties. The Genotype-3 line showed the lowest oviposition on TN1 rice (Fig. 4D). The higher reproductive fitness of the Genotype-2 line on susceptible rice than that of the Genotype-3 line may explain why this genotype spread more widely in the wild (Fig. 4A). However, on IR36 rice, the Genotype-3 line demonstrated significantly higher oviposition than the Genotype-1 and Genotype-2 lines, suggesting a greater fitness advantage of this genotype on resistant rice.

Functional validation of nAChR-α-7-like and CYP4C61 in N. lugens

To further validate the causal relationship between the candidate loci and the phenotypes, the expression of nAChR-α-7-like and CYP4C61 genes was down-regulated in GD-P1 insects through RNA interference (RNAi). The results showed that the mRNA levels of both target genes were significantly decreased in all three genotype lines at 24 h; however, they had no impact on each other (Fig. S10), indicating the lack of interplay between the expressions of nAChR-α-7-like and CYP4C61 genes. After insect individuals were injected with dsnAChR-α-7 or dsCYP4C61 and then treated with imidacloprid, the survival rates of all three genotype lines significantly decreased compared with those of the control insects (dsGFP) (Fig. 5A). No significant change in virulence towards TN1 rice was found among those individuals treated with dsRNA of two target genes and dsGFP (Fig. S11). However, the RNAi knockdown of both target genes significantly decreased the virulence towards IR36 rice in the Genotype-2 and Genotype-3 lines (P < 0.05, one-way ANOVA), equivalently to the levels in the Genotype-1 line (Fig. 5B). In addition, the knockdown of CYP4C61 led to lower virulence in the Genotype-3 line than the knockdown of nAChR-α-7-like; meanwhile, the effect of nAChR-α-7-like knockdown was equivalent to that of CYP4C61 knockdown in the Genotype-2 line.

Fig. 5.

Functional validation of nAChR-α-7-like and CYP4C61 genes in N. lugens. (A) Impacts of RNAi knockdown of nAChR-α-7-like and CYP4C61 on the survival rates of female N. lugens adults exposed to 50 mg/L imidacloprid for 72 h (n = 3). (B) Impacts of RNAi knockdown on the honeydew content secreted by a one-day-old female N. lugens adult after feeding on IR36 rice (n = 5), which were assessed for Genotype-1, Genotype-2, and Genotype-3 lines that were screened from GD-P1. Time course of (C) nAChR-α-7-like and (D) CYP4C61 expression in three genotype lines of GD-P1 in response to 12.5 mg/L imidacloprid were assessed (n = 3). nAChR-α-7-like (E) and CYP4C61 (F) expression in three genotype lines of GD-P1 fed with TN1 and IR36 rice was assessed at 24 h (n = 3). The data are the mean values ± SEM. *: P < 0.05; **: P < 0.01 (t-test). For panels C and D, the values for each time point sharing the same letter are not significantly different at P < 0.05 (one-way ANOVA and Duncan’s multiple range test were conducted for each genotype separately).

The induction of the expression of nAChR-α-7-like and CYP4C61 in response to imidacloprid exposure and IR36 rice feeding was also tested in GD-P1 individuals. After exposure to imidacloprid, the mRNA levels of nAChR-α-7-like and CYP4C61 significantly increased in all three genotype lines over time (Fig. 5C and D), except for the CYP4C61 expression in the Genotype-3 line, which reached a peak at 48 h post-treatment. In contrast, after being challenged by IR36 rice for 24 h and 48 h, all insects in three genotype lines showed no significant change in the mRNA levels of both genes (Fig. 5E and 5F; Fig. S12).

Discussion

As a sustainable and economical strategy, the breeding of crop cultivars with resistance to pests has attracted considerable attention. However, host plant resistance is frequently broken down by the unexpected rapid adaptation of target pests [47], [48]. The durability of host plant resistance is determined by the strength of pest-resistance genes and the features of the pest population, particularly the prevalence of pre-adapted individuals in the pest population [49]. This category of pre-adaptation includes adaptation to biotic factors (such as host plant resistance) and abiotic factors (such as insecticides). In the present study, we focused on the pre-adaptation of N. lugens to a commonly used insecticide (imidacloprid) and found that mutations in two potential insecticide resistance genes, nAChR-7-like and CYP4C61, accelerated the adaptation of N. lugens to resistant rice IR36. CYP4C61 was previously proven to play a key role in the ability of BPH to feed on the resistant rice variety YHY15 [50], and we further validated its function in the insecticide resistance of N. lugens here. Additionally, the role of another new gene nAChR-7-like in the cross-resistance between adaptation to the resistant rice variety and insecticide resistance was also confirmed in this study.

nAChR and P450 genes represent target site insensitivity and enhanced metabolism respectively, two major insecticide-resistance mechanisms [40], [51]. Here, the nAChR-7-like and CYP4C61 genes tended to exhibit higher expression in resistant N. lugens genotypes than in the susceptible genotype (Fig. 5E and F), and the RNAi knockdown of these two genes showed their contribution to insecticide resistance in N. lugens. However, these two genes may be minor genes responsible for imidacloprid resistance in N. lugens laboratory and field populations, as no previous study reported the direct selection of insecticides on their loci. In contrast, some other genes (in particular, P450s CYP6ER1 and CYP6AY1) have previously been proven as the major genes for imidacloprid resistance in N. lugens populations [21], [41], [52], [53]. However, CYP6ER1 and CYP6AY1 may not be involved in the adaptation of N. lugens to IR36 rice, because of that no change was found in the gene loci during IR36 rice selection, and the RNAi knockdown of these two genes had no impact on the virulence against IR36 rice in N. lugens (Fig. S13). These results suggest the division of labor for insecticide resistance-related genes in this species. The major genes were more likely to be obligate to insecticides, while minor genes may participate in cross-resistance between adaptation to resistant rice varieties and insecticide resistance, increasing the difficulty of pest control using both resistant rice varieties and insecticides.

Although no studies have reported how N. lugens adapts to resistant rice IR36, the overexpression of P450 genes is frequently associated with the adaptation of N. lugens to other resistant rice varieties. For instance, the expression of CYP6CS1 and CYP6CW1 was induced by the resistant rice variety MH63 [54]. Additionally, the RNAi of CYP4C61 inhibited the ability of N. lugens to feed on the resistant rice YHY15, indicating a potential role of CYP4C61 in catalyzing YHY15 plant allelochemicals [50]. IR36 rice also generates various allelochemicals in response to N. lugens infestation [55], [56]. Therefore, CYP4C61 may have the ability to detoxify rice allelochemicals homologous to imidacloprid. Further investigation is needed to unveil the catalytic mechanism of CYP4C61.

Remarkably, the resistance ratio to imidacloprid in N. lugens field populations in China increased from 135.3 to 301.3-fold in 2006/2007 to 233.3–2029-fold in 2012, and further reached over 2000-fold in 2016 [38], [52], [57]. Combined with our survey data (Table S7) and the previous report showing that BPH-resistant rice varieties are rarely used across China [58], the accumulation of resistant genotypes of nAChR-7-like and CYP4C61 in field populations is likely the result of long-term insecticide usage (Fig. 4A). This is consistent with previous studies that showed that alleles conferring resistance are expected to increase in frequency and become fixed in insect populations under steady selection [42], [59], [60], [61]. The fact that the expression of nAChR-7-like and CYP4C61 was induced by insecticide exposure, but not by resistant rice feeding in a short period, suggests that these two genes are likely challenged by insecticide usage, rather than resistant rice under field conditions (Fig. 5C–F). Thus, the increased virulence against IR36 rice in N. lugens populations in our case should be at least partially attributable to the pre-adapted loci of insecticide-resistance genes, and may be the eco-evolutionary consequence of insecticide-resistance development [16].

The two resistant genotypes (Genotype-2 and Genotype-3) inferred from the alleles of genes nAChR-7-like and CYP4C61 may have made different contributions to the adaptation of N. lugens to IR36 rice due to their differences in conferring fitness advantages (Fig. 4D) and gene expression (Fig. 5). Therefore, we could figure out the adaptive lineage to IR36 rice in our case based on the population structure of different generations of P-IR36. From F0 to F11, Genotype-2 offspring largely accumulated in the population and the population’s virulence to IR36 was significantly elevated, but may have led to a certain degree of fitness cost (Fig. 3A). Then, from F11 to F21, the Genotype-3 offspring gradually became dominant in the population as less fitness cost was observed (Fig. 1B). Meanwhile, the Genotype-3 offspring exhibited higher basic expression of CYP4C61, thus conferring a higher level of imidacloprid resistance (Fig. 1E).

It is noteworthy that the adaptation of N. lugens to resistant rice is known to be due to polygenic inheritance [47]. For example, the laboratory selection of N. lugens populations against a rice variety carrying the Bph1 resistance gene revealed that the virulence of N. lugens to the resistant rice cultivar Mudgo could be derived from a single recessive gene or from several major quantitative trait loci [62], [63], suggesting redundancy in virulence-related genes against the same resistant rice variety. In this study, the adaptation of the imidacloprid-resistant N. lugens population to the insect-resistant rice variety IR36 may also be attributed to the selection of multiple loci, including other insecticide resistance-related genes (eg, cuticle protein, acetylcholinesterase, ATP-binding cassette transporter, and esterase) in addition to nAChR-7-like and CYP4C61 (Fig. 1D; Table S5). Some of these genes were also previously associated with the adaptation to host plant resistance. For instance, the ATP-binding cassette transporter was involved in the secretion of both insecticides and plant chemicals [64], [65]. Besides, esterase in insects is reported to be associated with the degradation of both insecticides and plant secondary defense compounds [66], [67]. The roles of these genes in the interaction between insecticide resistance and adaptation to resistant rice plants in N. lugens deserve further investigation.

Conclusion

Collectively, we provide experimental and genetic evidence for the reduction in the profitability of insect-resistant rice cultivars mediated by insecticide resistance. Despite the limited experimental populations used in this study, the related genotypes we identified using this approach are widely applicable in the field, indicating that our findings are reliable in reflecting the eco-evolutionary effects conferred by insecticide resistance. To include the ecological and evolutionary consequences of pesticide resistance development in future pest management practices, we propose the implementation of conceptually predictable agriculture, a phase of agriculture with predictive outputs for measures. Based on a genomic sequencing strategy, we elucidated the genetic basis of agricultural pests in response to host plant resistance and other abiotic factors, such as insecticides, and used these data to predict the potential phenotypes of pest populations in the field. Only the crop varieties that are most compatible with the current (even in the near future) population structure of pests should be approved to be planted in the field to ensure adequate income from the varieties. As a result, pest control practices should be upgraded to predictably sustainable management, which will support modern agricultural systems to meet the global food demand.

CRediT authorship contribution statement

Rui Pang: Methodology, Investigation, Data curation, Writing – original draft, Funding acquisition. Shihui Li: Validation, Investigation, Resources. Weiwen Chen: Validation, Investigation, Resources. Longyu Yuan: Validation, Resources. Hanxiang Xiao: Resources. Ke Xing: Resources. Yanfang Li: Resources. Zhenfei Zhang: Resources. Xionglei He: . Wenqing Zhang: Conceptualization, Methodology, Supervision, Writing – original draft, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

The following are the Supplementary data to this article: