Independent effects of cestode infection and simulated bacterial challenge on gene expression and microbiota in a social insect
Description
Animal immune systems exhibit distinct responses to different parasitic infections, often involving specific yet interconnected pathways. In both vertebrates and invertebrates these immune responses can be further modulated by symbionts such as the gut microbiome. In this study, we examine how the social ant Temnothorax nylanderi responds to infection by the tapeworm Anomotaenia brevis and to a simulated bacterial challenge using lipopolysaccharides (LPS) from gram-negative bacteria. Our experimental approach assesses both the individual and combined effects of these challenges on two key metrics: transcriptional activity in the fat body, an organ crucial for immune defence, and the composition of the gut microbiome. We observed significant immune responses to both the tapeworm infection and the simulated bacterial challenge, with tapeworm infection having a more pronounced impact. There was no evidence of an interaction between the two challenges in either gut microbiome composition or gene expression profiles. However, upon inspection of commonly differentially expressed genes, we demonstrate that the two different immune challenges cause opposing effects in gene expression. This suggests that the tapeworm may selectively suppress the host’s immune response against itself while allowing the host to defend against other pathogens. We identified involved immune pathways, such as the Toll pathway, involved in both basal and induced immune responses in this social insect. Overall, our findings demonstrate that the same immune genes can be activated by different types of pathogens, whose effects may not be synergistic and can even go in opposite directions.
Files
Steps to reproduce
Differential gene expression analysis We subjected the gene count matrix to two different analyses: first, comparing infected ants to uninfected ants from infected nests, and second, comparing ants from uninfected nests to uninfected ants from infected nests. We first used principal component analyses (PCAs) on both datasets to identify potential outliers. For further analysis, we used DESeq2 to fit the two subsets of the data against a negative binomial distribution [50]. For both subsets, we worked with three variables: cestode infection, LPS challenge and colony identity. We then performed a Likelihood Ratio Test (LRT) using DESeq2 separately on the first factors using the other ones as reduced factors. We also ran and LRT for the interaction term between cestode infection and LPS challenge using all three main factors as reduced variable. Overlapping differentially expressed genes (DEGs) from all three analyses were then visualized with InterActiVenn [51]. To statistically assess, whether the number of overlapping genes was higher than expected by pure chance, we employed a Monte-Carlo-Simulation of 10,000 distributions with the same number of genes as our dataset checking whether the number of overlapping DEGs found is likely to be random. For the functional annotation of DEGs, we divided them into up- and downregulated genes among the three factors and the overlapping genes (based on log2 fold change), we then performed Fisher’s exact test through TopGO with the InterProScan file as universe and reference [52]. We further investigated the DEG functions by running a KEGG enrichment analysis using our EGGNOG output and the ClusterProfiler R-package [53]. Finally, we BLASTED our DEGs using diamond BLAST [54]. We ran these BLAST terms through UniProt for Caenorhabditis elegans and Drosophila melanogaster for a more thorough understanding of the gene functions using a script from Stoldt et al. 2021 [38], we specifically looked for UniProt descriptions that had the word “immun” in it to locate known immune genes. Weighted Gene Co-expression Network analysis We used the WGCNA R package [55] in order to perform a Weighted Gene Co-expression Network Analysis (WGCNA). We first filtered out genes that had less than 10 counts for at least 5 samples. For the network construction steps we set a soft-thresholding power of 10 and the minimum module size to 100 based on visual confirmation through a Topological Overlap Matrix (TOM) plot (Fig. S1). From this, we identified 20 modules and performed GO and KEGG enrichment analyses. We then used the module eigengene values for each sample to build a linear mixed-effects model, using LPS challenge and cestode infection as predictors and colony identity as a random effect. This was done for each module, and the models were tested for significance using ANOVA, examining both fixed effects and the interaction between them. We corrected the p-values for multiple testing using the Benjamini-Hochberg method.
Institutions
- Johannes Gutenberg University MainzRheinland-Pfalz, Mainz