Genetic dissection of Anopheles gambiae gut epithelial responses to Serratia marcescens

Abstract

Genetic variation in the mosquito Anopheles gambiae profoundly influences its ability to transmit malaria. Mosquito gut bacteria are shown to influence the outcome of infections with Plasmodium parasites and are also thought to exert a strong drive on genetic variation through natural selection; however, a link between antibacterial effects and genetic variation is yet to emerge. Here, we combined SNP genotyping and expression profiling with phenotypic analyses of candidate genes by RNAi-mediated silencing and 454 pyrosequencing to investigate this intricate biological system. We identified 138 An. gambiae genes to be genetically associated with the outcome of Serratia marcescens infection, including the peptidoglycan recognition receptor PGRPLC that triggers activation of the antibacterial IMD/REL2 pathway and the epidermal growth factor receptor EGFR. Silencing of three genes encoding type III fibronectin domain proteins (FN3Ds) increased the Serratia load and altered the gut microbiota composition in favor of Enterobacteriaceae. These data suggest that natural genetic variation in immune-related genes can shape the bacterial population structure of the mosquito gut with high specificity. Importantly, FN3D2 encodes a homolog of the hypervariable pattern recognition receptor Dscam, suggesting that pathogen-specific recognition may involve a broader family of immune factors. Additionally, we showed that silencing the gene encoding the gustatory receptor Gr9 that is also associated with the Serratia infection phenotype drastically increased Serratia levels. The Gr9 antibacterial activity appears to be related to mosquito feeding behavior and to mostly rely on changes of neuropeptide F expression, together suggesting a behavioral immune response following Serratia infection. Our findings reveal that the mosquito response to oral Serratia infection comprises both an epithelial and a behavioral immune component.

Figure 1. Gut infection with S. marcescens varies between individual An. gambiae mosquitoes.

Mosquitoes were antibiotic treated for 5 days and subsequently fed on sugar containing the Db11-GFP strain of S. marcescens. Bacteria-fed mosquitoes were selected 2 days post infection and the prevalence of fluorescent bacteria in their gut was monitored from day 2 to 6 post infection. 1A: The level of S. marcescens infection in the mosquito gut showed considerable variation: mosquitoes with intense fluorescence in most of the gut were characterized as highly infected (left panel), mosquitoes in which fluorescence was evident but confined to a part of the gut were characterized as lowly infected (middle panel) and mosquitoes with no sign of fluorescence were characterized as non-infected (right panel). 1B: S. marcescens infected mosquitoes were dissected each day, from day 2 to 6 post infection, and the proportions of highly, lowly and non-infected mosquitoes were determined over 4 independent infections. The average percentage ±SEM for each level of infection is indicated for each day post infection, with the total number of mosquitoes dissected each day in all 4 infections shown over each bar. 1C: In 2 independent infections used for SNP genotyping, mosquitoes were dissected 5 days post infection and the percentage of highly, lowly and non-infected mosquitoes, pooled from both infections, can be seen beside the respective part of the bar representing each level of infection.

Figure 2. Mapping of An. gambiae genetic variation associated with the S. marcescens infection phenotype.

SNPs with MAF difference >0.5 and 10-SNP windows with Bonferroni-corrected significance (p-value<10⁻⁵) are shown in their respective chromosomal position as red X crosses and dots, respectively. Non-significant 10-SNP windows are shown as blue dots. Genomic areas with highlighted SNPs and/or significant 10-SNP windows in close proximity are referred to as peaks and are numbered. Each peak is referred to using the chromosomal arms it resides on and its respective assigned number. The genomic positions of genes of interest found within a 5 kb radius of highlighted SNPs or within genomic areas delineated by 10-SNP windows with a significant p-value are indicated by vertical arrows.

Figure 3. Silencing of FN3D1–3 increases Serratia levels in orally infected mosquitoes or mosquitoes retaining their natural gut microbiota.

Antibiotic treated and subsequently orally infected with S. marcescens (Ab+Sm+) or non-treated mosquitoes retaining their natural midgut microbiota (Ab−Sm−), were dsRNA treated to silence FN3D1 (3A), FN3D2 (3B) or FN3D3 (3C) or treated with the LacZ dsRNA control. The bacterial load in the midguts of surface sterilized mosquitoes was determined 5 days post S. marcescens infection for Ab+Sm+ mosquitoes, or 5 days post dsRNA treatment for Ab−Sm− mosquitoes. Bacterial load was determined using broad range bacterial 16S or Serratia-specific primers using qRT-PCR, in which relative to the endogenous AgS7 control bacterial abundance was determined for each sample and then normalized to the relative abundance of the dsLacZ treated control. For Ab+Sm+ mosquitoes, the average ±SEM of the fold-change in bacterial load is shown as determined over 7 independent infections for FN3D1 (3A) and FN3D2 (3B), or 8 independent infections for FN3D3 (3C), with the qRT-PCR in each infection replicated at least twice. For Ab−Sm− mosquitoes, the average ±SEM of the fold-change in bacterial load is shown as determined over 4 independent assays for FN3D1 and FN3D3, or 5 independent assays for FN3D2. Asterisks indicate significance in an one-sample t-test against zero using the log2-transformed fold-change values so that zero corresponds to no difference from dsLacZ treatment. Two asterisks indicate a p-value<0.005 while three asterisks indicate a p-value<0.0005.

Figure 4. FN3D1–3 silencing changes the composition of the mosquito gut microbiota in favor of Enterobacteriaceae.

The 16S V4–V6 hypervariable regions of gut bacterial populations from mosquitoes retaining their natural gut microbiota without antibiotic treatment or S. marcescens infection (Ab−Sm−, Figure S3) were sequenced using 454 pyrosequencing (Table S6). cDNA pools from guts of FN3D1–3 dsRNA treated mosquitoes or dsLacZ treated controls, surface sterilized and dissected 5 days post dsRNA treatment, were PCR amplified and sequenced over 3 independent assays (panels A to C). The gut microbiota composition of the FN3D1–3 dsRNA treated pools or the dsLacZ-treated control in each independent assay can be seen in the respective pie charts, with the dsRNA treatment indicated below each pie chart. The color legend indicates the bacterial family corresponding to each pie chart color. Modulation of total bacteria or Serratia abundance can be seen for each sequenced pool in Figure S3, with FN3D1 kd corresponding to replicate 1 in panel 4A, FN3D2 kd corresponding to replicate 1 in panel 4A and 3 in panel 4B and FN3D3 kd corresponding to replicate 2 in panel 4B and 4 in panel 4C.

Figure 5. Gr9 silencing increases S. marcescens levels in orally infected mosquitoes.

5A: The bacterial load of antibiotic treated mosquitoes orally infected with S. marcescens (Ab+Sm+), treated either with Gr9 dsRNA or the dsLacZ control was determined at day 5 post infection either with broad range 16S or Serratia-specific primers. The average ±SEM of the bacterial fold-increase is shown, compared to dsLacZ treated mosquitoes over 5 independent infections, with the qRT-PCR performed at least twice for each infection. 5B: Antibiotic treated mosquitoes were starved overnight and then offered a sugar meal through a 5 µl capillary. Sugar meal consumption was determined 16 hours later for 38 LacZ and 55 Gr9 dsRNA treated mosquitoes. The average ±SEM percentage of sugar consumption through the capillary for each mosquito is shown for LacZ and Gr9 dsRNA treatments. In panel 5A, asterisks indicate significance in a one-sample t-test against zero using the log2-transformed fold-change values while in panel 5B asterisks indicate significance in the non-parametric Mann-Whitney test. Two asterisks indicate a p-value<0.005 while three asterisks indicate a p-value<0.0005.

Figure 6. Transcriptional regulation following S. marcescens infection using DNA microarrays.

Antibiotic treated mosquitoes were orally infected with S. marcescens and, 3 days post infection, transcriptional regulation in the gut of bacteria-fed mosquitoes was determined using DNA microarrays, compared to uninfected mosquitoes further antibiotic treated for 3 days. 6A: Volcano plot of transcriptional regulation as determined over 3 independent infections. The log2-transformed fold-change values for each transcript, as determined by two probes for each of the three arrays, were used for a one-sample t-test against zero, where zero corresponds to no regulation. Transcripts with more than 1.75-fold regulation are indicated either by black dots if the p-value of the t-test is >0.05 or red dots if the p-value is <0.05. Transcripts corresponding to LYSC2, PGRPLC, CLIPE6, CLIPB14 and NPF are indicated by arrows. 6B: Functional classification of more than 1.75-fold regulated genes. The 97 genes with more than 1.75-fold regulation were assigned to a functional class based on assigned GO terms, InterPro-predicted domains or Drosophila orthologs. The pie chart shows the proportion of genes assigned to each functional class. Functional classes corresponding to significantly overrepresented GO terms are indicated by asterisks.

Figure 7. The Gr9 antibacterial effect mostly relies on changes in NPF expression.

7A: Antibiotic treated mosquitoes orally infected with S. marcescens (Ab+Sm+) were dissected 5 days post infection and the NPF levels of mosquitoes treated either with LacZ or Gr9 dsRNA were determined in their guts. The average ±SEM of the NPF fold-increase as determined over 8 independent infections, with the qRT-PCR performed at least twice for each infection, can be seen. 7B: The bacterial load of antibiotic treated mosquitoes orally infected with S. marcescens (Ab+Sm+) treated either with LacZ dsRNA or with a 50–50% mix of either Gr9 and LacZ or Gr9 and NPF dsRNA was determined and normalized to the levels of dsLacZ treated mosquitoes. The average ±SEM of the bacterial fold-increase as determined over 3 independent infections can be seen, with the qRT-PCR performed at least twice for each infection. In panel 7A asterisks indicate significance in a one-sample t-test against zero using the log2-transformed fold-change values while in panel 7B asterisks indicate significance in the non-parametric Mann-Whitney test. Two asterisks indicate a p-value<0.005 while three asterisks indicate a p-value<0.0005.