Pyrethroid exposure alters internal and cuticle surface bacterial communities in Anopheles albimanus

Abstract

A deeper understanding of the mechanisms underlying insecticide resistance is needed to mitigate its threat to malaria vector control. Following previously identified associations between mosquito microbiota and insecticide resistance, we demonstrate for the first time, the effects of pyrethroid exposure on the microbiota of F₁ progeny of field-collected Anopheles albimanus. Larval and adult mosquitoes were exposed to the pyrethroids alphacypermethrin (only adults), permethrin, and deltamethrin. While there were no significant differences in bacterial composition between insecticide-resistant and insecticide-susceptible mosquitoes, bacterial composition between insecticide-exposed and non-exposed mosquitoes was significantly different for alphacypermethrin and permethrin exposure. Along with other bacterial taxa not identified to species, Pantoea agglomerans (a known insecticide-degrading bacterial species) and Pseudomonas fragi were more abundant in insecticide-exposed compared to non-exposed adults, demonstrating that insecticide exposure can alter mosquito bacterial communities. We also show for the first time that the cuticle surfaces of both larval and adult An. albimanus harbor more diverse bacterial communities than their internal microbial niches. Together, these findings demonstrate how insecticide pressure could be selecting for certain bacteria within mosquitoes, especially insecticide-metabolizing bacteria, thus potentially contributing to insecticide resistance.

Fig. 1

Sampling sites of field-caught An. albimanus and details of bioassays on resulting F₁ progeny. This figure shows the map of Guatemala indicating the Department of Escuintla (right) and an expanded view of La Gomera Municipality indicating the four sites from which field-collected mosquitoes were sampled (left). F₁ larvae (n = 132) and adult (n = 135) progeny originating from these sites and tested for insecticide resistance, were sequenced. For each collection site, the figure shows the developmental stage of F₁ progeny tested, color-coded by the type of insecticide they were tested for. See Suppl. 2 for a summary of the number of larvae and adults processed for each insecticide per site

Fig. 2

Gneiss linear regression outputs showing variables that contributed to shifts in balance (abundance between subsets) of amplicon sequence variants (ASVs). a Hierarchical clustering of ASVs used to create gneiss balances. b–e are based on this clustering. Red vertical bars indicate the first 10 of 1976 balances (y) that were used to build the linear regression model. ASVs that were included as balance numerators are highlighted in light red, while denominators are highlighted in dark red. The first balance, y0, comprised a majority of the ASVs and was the focus of all gneiss analysis. b–e show heatmaps of ASV abundance across variables that contributed the most to the regression model. The regression model predicted 51% of the shifts in balance between microbial communities, with location (b) contributing 9%, insecticide type (c) 1%, developmental stage (d) 24%, and microbial niche (e) 4%. These outputs informed downstream analysis. Alpha alphacypermethrin, Delta deltamethrin, Perm permethrin

Fig. 3

Principal coordinate analysis (PCoA) plots of Bray–Curtis distances between pyrethroid-exposed and non-exposed F₁ adult An. albimanus. PCoA plots, based on Bray–Curtis dissimilarity distances, show clustering patterns of the internal and cuticle surface microbiota in adult F₁ mosquitoes with respect to pyrethroid exposure. Separate plots are presented for each insecticide tested. Each point on the plots represents the bacterial composition of a pool of three mosquitoes, with the axes representing the first two dimensions of the PCoA, along with the proportion (%) of variation in bacterial composition explained by pyrethroid exposure. Results of overall non-pair-wise beta diversity (Bray–Curtis) comparison using permutational multivariate analysis of variance (999 permutations) tests, presented above each plot, were used to determine whether PCoA patterns were significant. The test statistic value (pseudo-F) for each comparison is presented, with significance set to p value < 0.05. In general, the plots show distinct separation of pyrethroid-exposed (resistant and susceptible) mosquitoes from non-exposed mosquitoes. These clustering patterns were statistically significant for alphacypermthrin and permethrin but not for deltamethrin in the internal microbiota. The cuticle surface microbiota, however, showed varying clustering patterns with only patterns of alphacypermethrin-tested mosquitoes being consistent with those of the internal microbiota

Fig. 4

Principal coordinate analysis (PCoA) plots of Bray–Curtis distances between pyrethroid-exposed and non-exposed An. albimanus F₁ larvae. PCoA plots, based on Bray–Curtis dissimilarity distances, show clustering patterns of the internal and cuticle surface microbiota in An. albimanus F₁ larvae with respect to pyrethroid exposure. Separate plots are presented for each insecticide tested. Each point on the plots represents the bacterial composition of a pool of three mosquitoes, with the axes representing the first two dimensions of the PCoA, along with the proportion (%) of variation in bacterial composition explained by pyrethroid exposure. Results of overall non-pair-wise beta diversity (Bray–Curtis) comparison using permutational multivariate analysis of variance (999 permutations) tests, presented above each plot, were used to determine whether PCoA patterns were significant. The test statistic value (pseudo-F) for each comparison is presented, with significance set to p value < 0.05. For the internal microbiota, the plots show distinct separation of pyrethroid-exposed (resistant and susceptible) mosquitoes from non-exposed mosquitoes, particularly within each location. These clustering patterns, although consistent for both insecticides, were not statistically significant. The cuticle surface microbiota, however, showed a general separation of exposed mosquitoes away from those that were not exposed. This clustering pattern was statistically significant for both insecticides

Fig. 5

Bar plots showing the relative abundance of taxonomically annotated amplicon sequence variants (ASVs) from F₁ adult An. albimanus. ASVs were taxonomically annotated to the genus level, and only taxa with relative abundance >0.1% are shown, all other taxa are collapsed and presented as “rare taxa.” The bar plots show the relative abundance of annotated ASVs across all sites, sub-categorized by insecticide type and resistance status, indicating that the adult internal (a) and cuticle surface (b) microbiota is dominated by Asaia. Across all insecticides tested, the relative abundance of identified taxa differed between resistant, susceptible, and unexposed mosquitoes in both internal (a) and cuticle surface (b) microbiota. ASVs that were not identified to the genus level are presented in square brackets, indicating the lowest possible level annotated. Identified taxa are organized by phylum, with phylum name indicated in bold. Res resistant, Sus susceptible, Unexp non-exposed

Fig. 6

Bar plots showing the relative abundance of taxonomically-annotated amplicon sequence variants (ASVs) from An. albimanus F₁ larvae. ASVs were taxonomically annotated to the genus level, and only taxa with relative abundance >0.1% are shown, all other taxa are collapsed and presented as “rare taxa.” The bar plots, grouped by location and sub-grouped by insecticide type and resistance status, show variable relative abundance of annotated ASVs across all locations, indicating that the larval internal (a) and cuticle surface (b) microbiota is mostly dominated by bacteria belonging to the phylum Proteobacteria. The genus Leucobacter was dominant in the internal microbiota, especially in El Terrero and Las Cruces 4 (a), while Acidovorax dominated the cuticle surface microbiota (b). Across all insecticides tested, the relative abundance of identified taxa differed between resistant, susceptible, and non-exposed mosquitoes in both internal (a) and cuticle surface (b) microbiota. ASVs that were not identified to the genus level are presented in square brackets, indicating the lowest possible level annotated. Identified taxa are grouped and color-themed by phylum, with phylum name indicated in bold. Res resistant, Sus susceptible, Unexp non-exposed

Fig. 7

Log ratios of amplicon sequence variants (ASVs) in gneiss balance y0 and the number of unique taxa that contributed to shifts in bacterial composition between insecticide exposed and non-exposed adults. Box plots show the distribution of ASV log ratios between insecticide exposed and non-exposed adult F₁ progeny. Corresponding horizontal bar plots (right hand side of each panel) show the top unique bacterial taxa in y0 numerator and denominator that contributed to shifts in y0 balance. Separate plots are presented per microbial niche and for each insecticide wherein significant differences in bacterial composition were observed between exposed and non-exposed mosquitoes. The bacterial composition of non-exposed mosquitoes was used as a reference to determine any shifts caused by pyrethroid exposure. Across each insecticide tested, y0 log ratio was lower in insecticide-exposed mosquitoes compared to those that were not exposed, indicating that the abundance of the unique taxa in y0 denominator was higher in insecticide-exposed mosquitoes compared to those that were not and vice versa for the taxa in numerators. This was seen in both internal and cuticle surface microbiota. Taxa were annotated to the genus level where possible or otherwise included in square brackets

Fig. 8

Log ratios of amplicon sequence variants (ASVs) in gneiss balance y0 and the number of unique taxa that contributed to shifts in cuticle surface bacterial composition between insecticide exposed and non-exposed larvae. Box plots show the distribution of ASV log ratios between insecticide-exposed and non-exposed larval F₁ progeny. Corresponding horizontal bar plots (right hand side of each panel) show the top unique bacterial taxa in y0 numerator and denominator that contributed to shifts in y0 balance. Separate plots are presented for each insecticide wherein significant differences in bacterial composition were observed between exposed and non-exposed mosquitoes. The bacterial composition of non-exposed mosquitoes was used as a reference to determine any shifts caused by pyrethroid exposure. y0 log ratio for deltamethrin-exposed larvae was lower than that of non-exposed larvae, indicating that the abundance of unique taxa in y0 denominator was higher in exposed vs non-exposed larvae. In permethrin-exposed larvae, however, y0 log ratio was higher than those of non-exposed mosquitoes, indicating that the abundance of unique taxa in y0 numerator was higher in exposed compared to non-exposed larvae and vice versa for unique taxa in the denominator. Taxa were annotated to the genus level where possible or otherwise included in square brackets