The Drosophila brain on cocaine at single-cell resolution

Abstract

Whereas the neurological effects of cocaine have been well documented, effects of acute cocaine consumption on genome-wide gene expression across the brain remain largely unexplored. This question cannot be readily addressed in humans but can be approached using the Drosophila melanogaster model, where gene expression in the entire brain can be surveyed at once. Flies exposed to cocaine show impaired locomotor activity, including climbing behavior and startle response (a measure of sensorimotor integration), and increased incidence of seizures and compulsive grooming. To identify specific cell populations that respond to acute cocaine exposure, we analyzed single-cell transcriptional responses in duplicate samples of flies that consumed fixed amounts of sucrose or sucrose supplemented with cocaine, in both sexes. Unsupervised clustering of the transcriptional profiles of a total of 86,224 cells yielded 36 distinct clusters. Annotation of clusters based on gene markers revealed that all major cell types (neuronal and glial) as well as neurotransmitter types from most brain regions were represented. The brain transcriptional responses to cocaine showed profound sexual dimorphism and were considerably more pronounced in males than females. Differential expression analysis within individual clusters indicated cluster-specific responses to cocaine. Clusters corresponding to Kenyon cells of the mushroom bodies and glia showed especially large transcriptional responses following cocaine exposure. Cluster specific coexpression networks and global interaction networks revealed a diverse array of cellular processes affected by acute cocaine exposure. These results provide an atlas of sexually dimorphic cocaine-modulated gene expression in a model brain.

Figure 1.

Behavioral characterization of Canton S (B) flies after cocaine exposure. (A) Negative geotaxis. The 7.5-cm climb time for each fly was measured. n = 120 (♀control), 114 (♀cocaine), 141 (♂control), 128 (♂cocaine). Horizontal lines represent means with standard error. Male flies exposed to cocaine took longer on average to climb compared to controls. (*) P = 0.0042; two-tailed Student's t-test. (B) Startle response. The percent time out of 45 sec that each fly spent moving following a 42-cm drop was measured. n = 155 (♀control), 145 (♀cocaine), 120 (♂control), 123 (♂cocaine). Horizontal lines represent means with standard error. Flies exposed to cocaine spent less time moving on average than controls. For females, (***) P = 4.68 × 10⁻⁶; for males, (***) P = 7.62 × 10⁻¹³; two-tailed Student's t-test. (C) Seizures and grooming activity during negative geotaxis. The percent of flies that exhibited seizures or grooming activity during the negative geotaxis assay after exposure to cocaine was measured. n = 142 (♀control), 141 (♀cocaine), 166 (♂control), 161 (♂cocaine). Both females and males exposed to cocaine exhibited seizure activity more than controls. For females, (*) P = 0.0361; for males seizure, (***) P = 0.0007; Fisher's exact test. Males exposed to cocaine also exhibited excessive grooming activity more than controls (grooming: [***] P = 0.0007; Fisher's exact test), but females did not show statistically significant differences. (D) Seizures and grooming activity during the startle response. The percent of flies that exhibited seizures or grooming activity during the startle assay after exposure to cocaine was measured. n = 155 (♀control), 145 (♀cocaine), 120 (♂control), 123 (♂cocaine). Female flies exposed to cocaine exhibited more seizure activity than controls ([*] P = 0.0121; Fisher's exact test), whereas male flies exposed to cocaine exhibited more grooming activity than control ([***] P = 0.00001; Fisher's exact test).

Figure 2.

UMAP visualization and clustering of single-cell expression data. Cells were clustered based on their expression pattern using the unsupervised shared nearest neighbor (SNN) clustering algorithm. Individual dots represent each cell and the colors of the dots represent the cluster to which the cells belong. Identification of cell types from clusters was performed by cross-referencing cluster-defining genes across FlyBase (Thurmond et al. 2019) and published literature (see Supplemental Table S4).

Figure 3.

Distribution of differentially expressed genes across clusters in males (A) and females (B) exposed to cocaine, and Venn diagrams showing overlap between differentially expressed genes in males and females (C). To identify clusters with unique gene expression patterns following acute exposure to cocaine, we filtered the list of differentially expressed genes to only show the strongest responses (|log_eFC| > 1.0, Bonferroni-adjusted P-value < 0.05) to construct an expression matrix. Differentially expressed genes are listed on the top (columns) and cell clusters are represented by the rows. Magenta boxes show up-regulation and turquoise boxes show down-regulation of gene expression as a result of exposure to cocaine. Panel C shows Venn diagrams of clusters with sexually dimorphic responses to cocaine exposure. The numbers within each Venn diagram represent the unique and shared differentially expressed (|log_eFC| > 0.5, Bonferroni-adjusted P-value < 0.05) genes due to cocaine exposure from DGE analysis performed for the corresponding cluster in male and female data sets separately.

Figure 4.

Subnetworks from coexpression network analyses of DEGs from the male C16 and C22 clusters. Networks are constructed from Pearson coefficient-based coexpression values calculated from scaled data of genes that were differentially expressed (filtered for |log_eFC| > 0.5, Bonferroni-adjusted P-value < 0.05) due to cocaine exposure. Coexpressions have been filtered using Random Matrix Theory. (A–C) MCODE subnetworks derived from the full network of male cluster C16. The inset in A corresponds to a subset of genes within the subnetwork that have very strong correlation coefficient values with each other compared to the rest of the data set. Colors of the dots represent the connectivity index derived from MCODE scores. Colors of edges represent the positive (red) and negative (green) correlations. (D) Coexpression network analysis of DEGs from the male C22 cluster.

Figure 5.

Interaction network analysis of DEGs from all clusters in the male data set. Network constructed from interactions calculated using StringApp plugin within Cytoscape for genes that were differentially expressed (filtered for |log_eFC| > 0.1 and Bonferroni-adjusted P-value < 0.05) in all clusters from the male data set. Gray edges represent interactions. Genes that were differentially expressed in multiple clusters are depicted as pie charts with each color representing the respective cluster. Genes are grouped into circles based on their MCODE connectivity scores. Annotations of these circular groups represent the pathways that are enriched for the genes within these groups. Bonferroni-adjusted P-value < 0.05 was considered as significant for enrichment in the statistical overrepresentation tests.

Figure 6.

Interaction network analysis of DEGs from all clusters in the female data set. Network constructed from interactions calculated using StringApp plugin within Cytoscape for genes that were differentially expressed (filtered for |log_eFC| > 0.1 and Bonferroni-adjusted P-value < 0.05) in all clusters from the female data set. Gray edges represent interactions. Genes that were differentially expressed in multiple clusters are depicted as pie charts with each color representing the respective cluster. Genes are grouped into circles based on their MCODE connectivity scores. Annotations of these circular groups represent the pathways that are enriched for the genes within these groups. BH-FDR adjusted P-value < 0.05 was considered as significant for enrichment in the statistical overrepresentation tests.