Genome-Wide Association Analyses Point to Candidate Genes for Electric Shock Avoidance in Drosophila melanogaster

Abstract

Electric shock is a common stimulus for nociception-research and the most widely used reinforcement in aversive associative learning experiments. Yet, nothing is known about the mechanisms it recruits at the periphery. To help fill this gap, we undertook a genome-wide association analysis using 38 inbred Drosophila melanogaster strains, which avoided shock to varying extents. We identified 514 genes whose expression levels and/ or sequences co-varied with shock avoidance scores. We independently scrutinized 14 of these genes using mutants, validating the effect of 7 of them on shock avoidance. This emphasizes the value of our candidate gene list as a guide for follow-up research. In addition, by integrating our association results with external protein-protein interaction data we obtained a shock avoidance-associated network of 38 genes. Both this network and the original candidate list contained a substantial number of genes that affect mechanosensory bristles, which are hair-like organs distributed across the fly's body. These results may point to a potential role for mechanosensory bristles in shock sensation. Thus, we not only provide a first list of candidate genes for shock avoidance, but also point to an interesting new hypothesis on nociceptive mechanisms.

Fig 1. Shock avoidance of 38 inbred Drosophila melanogaster strains.

A. For the shock avoidance assay, flies (represented by black dots) were loaded into the setup using ‘shock tubes’ (coloured yellow) at 0:00 min. At 1:00 min, they were transferred to a movable ‘mid-compartment’ (coloured orange). At 4:00 min, the mid-compartment was moved to the choice point of a maze with shock tubes as the two arms. After 2-min dispersal time, one of the maze-arms was applied with four pulses of electric shock (represented by yellow lightening symbols). 10 s after the last pulse, the maze-arms were sealed and the flies in each arm were counted to calculate a shock avoidance score. Negative values indicated avoidance of the shocked maze-arm. Orange and black arrows represent the movement of the mid-compartment and of the flies, respectively. B. The 38 tested inbred strains had significantly different shock avoidance scores. Box plots show the median as the midline, 25 and 75% as the box boundaries and 10 and 90% as the whiskers. Sample sizes were from left to right N = 32, 16, 22, 24, 24, 16, 16, 24, 26, 28, 16, 24, 28, 34, 16, 32, 24, 22, 18, 18, 22, 32, 16, 16, 24, 16, 28, 18, 22, 16, 30, 16, 20, 24, 16, 20, 32, 24.

Fig 2. Genome-wide association analyses for shock avoidance.

On the left, the gene expression level—shock avoidance association analysis is sketched. After pre-processing the raw expression microarray data, for each of the 18 769 probe-sets, we tested for a linear relationship across strains between the mean expression levels and the median shock avoidance scores. This analysis suggested 588 shock avoidance-associated probe-sets (linear regression P< 0.05; see S1 Table for a list with full statistical reports), corresponding to 356 candidate genes (see S4 Table for a list). On the right, the single nucleotide polymorphism (SNP)—shock avoidance association analysis is shown. We narrowed down our analysis to pre-selected SNPs with favourable minor allele frequencies and call rates. Testing for relationships between the allele types and the shock avoidance scores suggested 607 shock avoidance-associated SNPs (linear regression P< 0.0005; see S2 and S3 Tables for a list with full statistical reports), pointing to 169 candidate shock avoidance-genes (see S4 Table for a list), 11 of which were already suggested by the expression level-associations.

Fig 3. A shock avoidance-associated gene network.

Each of the 38 nodes in the network represents a gene associated with shock avoidance in terms of expression level (see S6 Table for a list). Each edge indicates a pair-wise physical interaction between the proteins encoded by the respective genes, based on independent empirical evidence. Shades of green mean that the higher the respective gene’s expression level, the stronger the shock avoidance was. Shades of red mean the converse, i.e., the higher the expression level the weaker the shock avoidance. The darker the shading, the greater the estimated effect of expression level on shock avoidance was. Circles represent genes with a statistically strong association with shock avoidance resulting in a positive network score. Potential functionally related genes with less significant association (negative scores; represented by squares) were included to form connections between the more strongly associated genes. Genes implicated in bristle-function are haloed blue.

Fig 4. Independent validation of candidate shock avoidance genes using transposon insertion mutants.

Each panel shows, for a selected candidate gene, the shock avoidance scores of a respective transposon insertion mutant vs. those of the corresponding control (see S7 Table for full genotypes). The colour of the font indicates the direction of the gene expression level—shock avoidance association (i.e. green: the higher the expression level, the stronger the shock avoidance; red: the higher the expression level, the weaker the shock avoidance). In 6 out of 14 cases, shock avoidance scores significantly differed between the genotypes. *: FDR< 0.05, ns: FDR≥ 0.05. Sample sizes are given in S7 Table. Box plots are as explained in Fig 1B.

Fig 5. Validation of candidate shock avoidance gene CG13397 using a deletion mutant.

A. We sketch the organization of the splice variant RA of the CG13397 in: a transposon insertion mutant (M1) and corresponding control strain (C0) as well as three independent control genotypes obtained by precise excision of the mutagen transposon (C1, C2, C3) and a deletion mutant obtained by imprecise excision of the mutagen transposon (M2). Please note that C0 and M1 are the same genotypes as used in the respective panel of Fig 4 (see S7 Table for full genotypes). Boxes represent exons (black and white filling for coding and non-coding regions, respectively), whereas the fat grey arrows represent the transposon. Arrows 1, 2 and 3 indicate the binding sites for the PCR primers. Expected amplification products are marked by dashed lines. Please note that in M2, in addition to a 574 nt-long deletion, a 79 nt-long residue of the transposon remained inserted. Expected fragments were obtained in each genotype using either primers 1 and 2, or 1 and 3 in single-fly PCR experiments. B. The deletion mutant (M2) performed worse than each of the controls (C1, C2, C3) in shock avoidance. *: P< 0.05/ 3. Sample sizes are N = 28 for each genotype. Box plots are as explained in Fig 1B.

Fig 6. Effects of validated shock avoidance candidate genes on locomotion.

Each panel shows, for the indicated gene, the locomotion scores of a respective transposon insertion mutant vs. those of the corresponding control. Please note that the genotypes are identical to those used in the respective panels of Fig 4 (see S7 Table for full genotypes). We used a locomotion assay that mimicked the shock avoidance assay except for the presentation of shock. Thus, using the setup and temporal schedule depicted in Fig 1A, at 6:00 min, instead of delivering shock pulses, the setup was vigorously shaken to force flies to the end of one maze-arm. The setup was then immediately put back to its horizontal position, letting the flies disperse towards the opposite arm in the absence of shock. The maze was sealed 25 s later and the locomotion score was calculated to reflect the ratio of flies that had travelled more than a shock tube-length in the given time. In 3 out of 7 cases, locomotion scores significantly differed between the genotypes. *: FDR< 0.05, ns: FDR≥ 0.05. Sample sizes are given in S7 Table. Box plots are as explained in Fig 1B.