Reproductive status alters transcriptomic response to infection in female Drosophila melanogaster

Abstract

Mating and consequent reproduction significantly reduce the ability of female Drosophila melanogaster to defend against systemic bacterial infection. The goal of the present study was to identify genes likely to inform the mechanism of this post-mating immunosuppression. We used microarrays to contrast genome-wide transcript levels in virgin vs. mated females before and after infection. Because the immunosuppressive effect of mating is contingent on the presence of a germline in females, we repeated the entire experiment by using female mutants that do not form a germline. We found that multiple genes involved in egg production show reduced expression in response to infection, and that this reduction is stronger in virgins than it is in mated females. In germline-less females, expression of egg-production genes was predictably low and not differentially affected by infection. We also identified several immune responsive genes that are differentially induced after infection in virgins vs. mated females. Immune genes affected by mating status and egg production genes altered by infection are candidates to inform the mechanism of the trade-off between mating and immune defense.

Keywords: Drosophila melanogaster; gene expression; immunity; microarray; reproduction.

Figure 1

Experimental design. To determine ways in which immune defense and reproduction may interact to cause post-mating immunosuppression, we compared genome-wide transcript abundance between virgin and mated, infected and uninfected females. In each contrast, the arrow conveys the treatment−control relationship, with the arrow emanating from the “control” condition and pointing toward the “treatment” condition in each analysis. We assayed for differential transcript abundance between virgin uninfected females and virgin infected females to identify infection-responsive genes in virgins (A) or mated females (B). By qualitatively comparing (A) with (B), we were able to establish differences in infection response that were dependent on mating status. By subtracting (A) from (B), we were able to ascertain which genes showed the largest quantitative differences in infection response between virgin and mated females. We also assayed for differential transcript abundance between virgin vs. mated females when infected (C) or uninfected (D) to determine which genes respond to mating and which differences depend on infection status. We independently performed this entire experimental design in triplicate for both egg-producing females and eggless females.

Figure 2

The effect of infection on transcript abundance in virgin and mated females. We assayed for genes that exhibited statistically significant 2-fold or greater differences in transcript abundance in virgin uninfected vs. virgin infected treatments and in mated uninfected vs. mated infected treatments. We then determined which genes change significantly in transcript abundance due to infection in both virgin and mated females, only in virgins, or only in mated females. Fold change values are in log₂ units and are expressed as uninfected minus infected signal; therefore, a negative logFC represents increased expression in response to infection whereas a positive logFC represents decreased expression in response to infection. In instances in which more than one probe showed significantly altered expression for a particular gene, only the probeset with the largest fold change is listed. GO term enrichment was determined using GOrilla and REVIGO was used to reduce lists of GO terms to those least redundant. Upward-pointing arrows indicate genes with increased expression and downward-pointing arrows indicate genes with depressed expression. A Benjamini-Hochberg correction (Benjamini and Hochberg 1995) was performed to correct for multiple tests, and only GO terms that were significant after controlling for a false-discovery rate of 5% were retained.

Figure 3

The effect of infection on transcript abundance in virgin and mated eggless females. We assayed for genes that exhibited significant 2-fold or greater differences in transcript abundance in virgin uninfected vs. virgin infected treatments and in mated uninfected vs. mated infected treatments. We then determined which genes change significantly in transcript abundance due to infection in both virgin and mated females, only in virgins, or only in mated females. Fold change values are in log₂ units, and are expressed as uninfected minus infected signal; therefore, a negative logFC represents increased expression in response to infection whereas a positive logFC represents decreased expression in response to infection. In instances in which more than one probe for a particular gene showed significant change in expression, only the probeset with the largest fold change is listed. GO term enrichment was determined using GOrilla and REVIGO was used to reduce lists of GO terms to those least redundant. Upward-pointing arrows indicate genes with increased expression and downward-pointing arrows indicate genes with depressed expression. A Benjamini-Hochberg correction (Benjamini and Hochberg 1995) was performed to correct for multiple tests, and only GO terms that were significant after controlling for a false-discovery rate of 5% were retained.

Figure 4

Vitelline membrane transcript abundances decrease after infection in egg-producing females. For all probesets that mapped to vitelline membrane genes, we determined averaged normalized signal intensity across all three biological replicates for each treatment. Only a single probeset exists on the array for Vm26Aa, Vm26Ab, Vm26Ac, and Vm32E, but Vm34Ca has three probesets and Vml has two. We then determined the change in mean signal intensity due to infection for virgin and mated females. These values are plotted to the left of each virgin line (solid) and to the right of each mated line (dashed) for each gene.

Figure 5

The effect of mating on transcript abundance in uninfected and infected females. We assayed for genes that showed significant twofold or greater differences in transcript abundance in virgin uninfected vs. mated uninfected treatments and in virgin infected vs. mated infected treatments. We then determined which genes change significantly in transcript abundance due to mating in both uninfected and infected females, only in uninfected, or only in infected females. GO term enrichment was determined for each set of genes using GOrilla, and REVIGO was used to reduce lists of GO terms to those least redundant. Upward-pointing arrows indicate genes with increased expression and downward-pointing arrows indicate genes with depressed expression. A Benjamini-Hochberg correction (Benjamini and Hochberg 1995) was performed to correct for multiple tests, and only GO terms that were significant after controlling for a false discovery rate of 5% were retained.

Figure 6

The effect of mating on transcript abundance in uninfected and infected eggless females. We assayed for genes that showed significant twofold or greater differences in expression in virgin uninfected vs. mated uninfected treatments and in virgin infected vs. mated infected treatments. We then determined which genes have significantly altered expression due to mating in both uninfected and infected females, only in uninfected females, or only in infected females. Fold change values are in log₂ units and are expressed as virgin minus mated signal; therefore, a negative logFC represents increased expression in response to mating whereas a positive logFC represents decreased expression in response to mating. In instances in which more than one probe indicated a significant change in expression for a particular gene, the probeset with the largest fold change is listed. GO term enrichment was determined using GOrilla. Upward-pointing arrows indicate genes with increased expression and downward-pointing arrows indicate genes with depressed expression. A Benjamini-Hochberg correction (Benjamini and Hochberg 1995) was performed to correct for multiple tests, and only GO terms that were significant after controlling for a false-discovery rate of 5% were retained.