Knockdown of RNA interference pathway genes impacts the fitness of western corn rootworm

Western corn rootworm (Diabrotica virgifera virgifera) is a serious agricultural pest known for its high adaptability to various management strategies, giving rise to a continual need for new control options. Transgenic maize expressing insecticidal RNAs represents a novel mode of action for rootworm management that is dependent on the RNA interference (RNAi) pathways of the insect for efficacy. Preliminary evidence suggests that western corn rootworm could develop broad resistance to all insecticidal RNAs through changes in RNAi pathway genes; however, the likelihood of field-evolved resistance occurring through this mechanism remains unclear. In the current study, eight key genes involved in facilitating interference in the microRNA and small interfering RNA pathways were targeted for knockdown in order to evaluate impact on fitness of western corn rootworm. These genes include drosha, dicer-1, dicer-2, pasha, loquacious, r2d2, argonaute 1, and argonaute 2. Depletion of targeted transcripts in rootworm larvae led to changes in microRNA expression, decreased ability to pupate, reduced adult beetle emergence, and diminished reproductive capacity. The observed effects do not support evolution of resistance through changes in expression of these eight genes due to reduced insect fitness.


Supplementary Methods
Expression analysis calculations Expression of core RNAi machinery genes was determined by first interpolating concentration values Q from the appropriate standard curve, then determining a ratio between the target gene (T) and each reference gene (R1 & R2) within each sample and calculating a geometric mean of the ratios using equation (1): Finally, the median and median absolute deviation of the geometric means for all samples within a treatment group were calculated ( Figure 2).

Larval growth
A generalized linear mixed model was fit to the data using a cumulative logit link function for the assumed multinomial distribution and the Laplace method of integral approximation. Treatment was modeled as a fixed effect, and bioassay day as a random effect. Statistical comparisons between the test and negative control treatments with Sidak's multiplicity adjustments were conducted (Figure 3).

Larval mortality
A generalized linear mixed model with a logit link function for the assumed binomial distribution and the Laplace likelihood approximation method was used to fit the data and estimate mortality rate for each treatment. Treatment was considered a fixed effect, and bioassay day a random effect. Statistical comparisons between the test and negative control treatments with Dunnett's multiplicity adjustments were conducted by testing on the odds ratio ( Figure 3).

Larval development
A generalized linear mixed model with a logit link function for the assumed binomial distribution and the Laplace likelihood approximation method was used to fit the data and estimate rate of live larvae in first or second and third developmental instars per treatment. Treatment was considered a fixed effect and bioassay day a random effect. Comparisons between the test and negative control treatments with Dunnett's multiplicity adjustments were conducted by testing on the odds ratio (Table 1).

Sample Mass
A one-way linear model with fixed treatment effects was used to fit the sample mass data for each sample type, and the restricted maximum likelihood estimation (REML) method was used to estimate treatment means. Pair-wise statistical comparisons were conducted with Tukey's multiplicity adjustments ( Figure  S1).

Total Adult Emergence
A generalized linear mixed model with a binomial distribution, a logit link function, and the Laplace likelihood approximation method was used to fit the data and estimate rate of adult emergence for each treatment. Treatment was considered a fixed effect and replicate within treatment a random effect. Comparisons between all treatments with Tukey's multiplicity adjustments were conducted by testing on the odds ratio ( Figure 4a).

Male and female adult emergence
A generalized linear mixed model with a binomial distribution, a logit link function, and the Laplace likelihood approximation method was used to fit the data and estimate either male or female proportion of adult emergence for each treatment. Treatment was considered a fixed effect. Comparisons between all treatments with Tukey's multiplicity adjustments were conducted by testing on the odds ratio ( Figure S2).

Adult mortality during oviposition
A generalized linear mixed model with a binomial distribution, a logit link function, and the Laplace likelihood approximation method was used to fit the data and estimate mortality rate for each treatment. Treatment, sex, and the interaction between them were considered fixed effects. Comparisons between all treatments with Tukey's multiplicity adjustments were conducted by testing on the odds ratio ( Figure  4d).

Fecundity
Four approaches were used to model egg counts relative to number of WCR adult females: 1. Eggs per starting female A linear mixed model was used to fit the ratio of eggs collected from replicates sampled at multiple time points to the number of females present at the beginning of the oviposition period. The REML method was used to estimate treatment means, and treatment, time point, and the interaction between them were considered fixed effects. A compound symmetry covariance structure was used for the error variance-covariance matrix to allow positive or negative error covariance among repeated samples from each replicate. Since the treatment and time point interaction was statistically significant, pair-wise statistical comparisons among all treatments within each time point were conducted with Tukey's multiplicity adjustments (Figures 5a and S3a).

Eggs per starting female at each time point
A linear mixed model was used to fit the ratio of eggs collected from replicates sampled at multiple time points to the number of females alive at the start of each oviposition time point. The REML method was used to estimate treatment means, and treatment, time point and the interaction between them were considered fixed effects. A compound symmetry covariance structure was used for the error variance-covariance matrix to allow positive or negative error covariance among repeated samples from each replicate. Since the treatment and time point interaction was statistically significant, pair-wise statistical comparisons among all treatments within each time point were conducted with Tukey's multiplicity adjustments ( Figure S3b).

Eggs per average live female
A linear mixed model was used to fit the ratio of eggs collected from replicates sampled at multiple time points to the average number of live females during each oviposition time period. The REML method was used to estimate treatment means, and treatment, time point and the interaction between them were considered fixed effects. A compound symmetry covariance structure was used for the error variance-covariance matrix to allow positive or negative error covariance among repeated samples from each replicate. Pair-wise statistical comparisons among all treatments across time points were conducted with Tukey's multiplicity adjustments for each time point ( Figure S3c).

Total eggs per starting female
A linear mixed model was used to fit the ratio of total eggs collected from replicates across all time points to the number of starting females at the beginning of the oviposition period, and the REML method was used to estimate treatment means. Treatment was considered a fixed effect. Pair-wise statistical comparisons among all treatments were conducted with Tukey's multiplicity adjustments ( Figure S3d).

Hatch Rate
A generalized linear mixed model with a binomial distribution, a logit link function, and the Laplace likelihood approximation method was used to fit the hatch rate data across time and estimate hatch rate for each treatment. Treatment, time point, and the interaction between them were considered fixed effects, while replicate within treatment, replication by treatment and time point, and replication by time point by aliquot within each treatment were considered random effects. Since the treatment and time point interaction was statistically significant, pair-wise statistical comparisons among all treatments within each time point were conducted with Tukey's multiplicity adjustments (Figure 5b).

Statistical analysis -qPCR assays
To estimate efficiency of the qPCR reaction for each assay (Supplementary File 2), the linear fixed effects model (2) was applied to standard curve data across multiple plates and days for each assay and target: where is the Cq of the k th sample of the i th concentration on the j th plate, Pj denotes the intercept of the j th plate, Xi denotes the logarithm of the i th concentration, denotes the slope for the j th plate, and εijk denotes residual where εijk ~ iid N(0, σ 2 Error).
Slope across plates was estimated for each standard curve type. Efficiency of qPCR was determined from the slope using equation (3): Confidence intervals of estimated efficiencies for each assay and target were then calculated from the confidence interval of the estimated slopes. The standard error of efficiency was calculated with the Delta method using equation (4): The 95% confidence intervals of the quantification cycle at each of the different concentrations tested were also estimated across plates to show viability at different concentrations along the standard curve. The Bonferroni method was applied for multiplicity adjustment of these confidence intervals.    Figure S4. Effects of core RNAi machinery knockdown on miRNA expression in WCR pupae. Insects collected 13 days post-treatment in the late-third-instar bioassays were analyzed for expression of miRNAs as described in Figure  6. Insects exposed to ago1 dsRNA were not available for analysis due to treatment effects. From left to right for each treatment, expression of miR-8 is shown in black, miR-276 in dark grey, miR-3761 in light grey, miR-1 in white, and miR-277 in gold.  Table S1. Primers and amplification conditions used for production of WCR RNAs