A systems level approach to temporal expression dynamics in Drosophila reveals clusters of long term memory genes

Abstract

The ability to integrate experiential information and recall it in the form of memory is observed in a wide range of taxa, and is a hallmark of highly derived nervous systems. Storage of past experiences is critical for adaptive behaviors that anticipate both adverse and positive environmental factors. The process of memory formation and consolidation involve many synchronized biological events including gene transcription, protein modification, and intracellular trafficking: However, many of these molecular mechanisms remain illusive. With Drosophila as a model system we use a nonassociative memory paradigm and a systems level approach to uncover novel transcriptional patterns. RNA sequencing of Drosophila heads during and after memory formation identified a number of novel memory genes. Tracking the dynamic expression of these genes over time revealed complex gene networks involved in long term memory. In particular, this study focuses on two functional gene clusters of signal peptides and proteases. Bioinformatics network analysis and prediction in combination with high-throughput RNA sequencing identified previously unknown memory genes, which when genetically knocked down resulted in behaviorally validated memory defects.

Fig 1. Wasp exposure causes ethanol seeking and long-term memory formation in fruit flies.

Flies are cohabitated with wasps for either 3 days (acute assay) or 4 days (memory assay) (A). At the end of the exposure period flies are given two food options, ethanol and control food, either in the presence of wasps (B) or in the absence of wasps (C). CS flies as well as memory mutants have an acute wasp response where the ethanol food is preferred as an oviposition substrate (D). However, only CS flies maintain ethanol preference during the wasp memory assay (E). Error bars indicate bootstrap 95% confidence intervals.

Fig 2. Signal peptide and protease genes are differentially regulated following wasp memory formation.

Volcano plot displays sequencing results from wasp-exposed fly heads. Differentially regulated genes with FDR = < 0.05 are shown in dark blue points, genes with significant FDR and a log₂ fold change magnitude of 2 or more are shown in light blue (A). Signal peptide and protease gene clusters were identified as enriched from a DAVID analysis. Bar shading indicates FDR, all genes shown have FDR = < 0.05 (B). IMP network analysis was preformed for the signal peptide cluster (C) and protease cluster (D) to generate a network of interactions amongst genes. In panels C and D, each node represents a gene; a green node indicates an input gene and grey node is a predicted interacting gene within the network, size of the node reflects the number of interactions. Known interactions are shown in red, blue edges are predicted interactions—the darker the edge the higher the predicted interaction score. A subset of genes identified from the sequencing was validated with qPCR. Samples were normalized to unexposed, the lower error bars represent SE of control group, upper error bars show SE of the exposed samples (E).

Fig 3. Differential gene expression occurs through the memory formation process.

Ethanol preference depends on the length of wasp exposure, 2.5 and 7 hours of exposure shows no significant ethanol preference change (A). Volcano plots display sequencing results from heads of flies exposed to wasps for 2.5 hours (B) and 7 hours (C). Genes with significant FDR (= < 0.05) are shown in dark blue points, genes with significant FDR and a log₂ fold change magnitude of 2 or greater are shown in light blue. Error bars indicate bootstrap 95% confidence intervals.

Fig 4. Gene expression is dynamic across memory formation time course.

Differentially expressed gene totals are shown for each time point 4 days (blue) 7 hours (grey) 2.5 hours (red). Genes with significant FDR are displayed as a Venn diagram (A). Alternatively, genes with significant FDR and minimum log₂ fold change magnitude of 2 are shown with overlap (B). Heat map illustrates gene expression across time, blue indicating down regulation and red up regulation of transcript (C). All genes from panel B are included in the heat map.

Fig 5. Temporal gene expression points to interacting networks during memory formation.

Heat map illustrates gene expression as quantified by qPCR following 2.5, 7, and 96 hours of wasp exposure. A fourth time point, 24 hours of recovery following wasp exposure, was included in the cluster analysis; shading indicates direction and magnitude of gene expression change (A). IMP network analysis was preformed for differentially regulated signal peptide genes from the 7-hour time point and a subset of genes from the 4-day signal peptide cluster (B). Each node represents a gene, a green node indicates a gene from the 4-day signal peptide cluster. Orange and blue nodes are genes from the 7-hour signal peptide cluster, color indicates that the gene is up or down regulated (orange and blue respectively). Grey nodes represent predicted interacting genes within the network, size of the node reflects the number of interactions. Known interactions are shown in red, blue edges are predicted interactions the darker the edge the higher the predicted interaction score.

Fig 6. Conditional knockdown in neurons identifies important genes for memory formation.

Gal4-Switch system expresses inactive Gal4 transcription factor, resulting in no RNAi hairpin (A). When flies are fed RU486, the Gal4-switch becomes active and can bind DNA, driving RNAi expression (B). Acute wasp exposure experiment shows that flies have ethanol seeking behavior when fed vehicle (C) or RU486 (D). All genotypes maintain ethanol seeking following wasp exposure when treated with vehicle only (E). Ethanol seeking in the memory assay is disrupted for flies expressing RNAi to a subset of genes (F). Error bars indicate bootstrap 95% confidence intervals.

Fig 7. Conditional knockdown in mushroom body neurons identifies genes acting specifically within the learning and memory center of the brain.

Flies have an ethanol seeking behavior in the presence of wasps during vehicle feeding (A) and RU486 feeding (B). Ethanol preference is maintained following wasp removal when flies are treated with vehicle only (C). Ethanol seeking is not maintained for all RNAi lines when fed RU486, indicating disruption in memory formation (D). Error bars indicate bootstrap 95% confidence intervals.

Fig 8. Conditional knockdown in mushroom body does not permanently inhibit memory formation.

Flies treated with RU486 that are transitioned back to normal food are able to form memory following recovery period (A). Mushroom body neurons expressing GFP under the MB switch inducible driver, dashed white line outlines the brain (B).