Genome-Wide Expression Profiles Drive Discovery of Novel Compounds that Reduce Binge Drinking in Mice

Abstract

Transcriptome-based drug discovery has identified new treatments for some complex diseases, but has not been applied to alcohol use disorder (AUD) or other psychiatric diseases, where there is a critical need for improved pharmacotherapies. High Drinking in the Dark (HDID-1) mice are a genetic model of AUD risk that have been selectively bred (from the HS/Npt line) to achieve intoxicating blood alcohol levels (BALs) after binge-like drinking. We compared brain gene expression of HDID-1 and HS/Npt mice, to determine a molecular signature for genetic risk for high intensity, binge-like drinking. Using multiple computational methods, we queried LINCS-L1000 (Library of Integrated Network-Based Cellular Signatures), a database containing gene expression signatures of thousands of compounds, to predict candidate drugs with the greatest potential to decrease alcohol consumption. Our analyses predicted novel compounds for testing, many with anti-inflammatory properties, providing further support for a neuroimmune mechanism of excessive alcohol drinking. We validated the top 2 candidates in vivo as a proof-of-concept. Terreic acid (a Bruton's tyrosine kinase inhibitor) and pergolide (a dopamine and serotonin receptor agonist) robustly reduced alcohol intake and BALs in HDID-1 mice, providing the first evidence for transcriptome-based drug discovery to target an addiction trait. Effective drug treatments for many psychiatric diseases are lacking, and the emerging tools and approaches outlined here offer researchers studying complex diseases renewed opportunities to discover new or repurpose existing compounds and expedite treatment options.

Figure 1

Prioritization of candidate compounds from the LINCS-L1000 (Library of Integrated Network-Based Cellular Signatures) search results. Using the constructed input gene expression signatures (see Supplementary Table S1) with various algorithm parameter settings, a total of 32 input queries were submitted to LINCS using the sig_gutc tool via C3 (Compute Connectivity on the Cloud, see Materials and Methods). Queries represented the top differentially expressed genes or the differentially expressed landmark genes (those directly measured in the LINCS-L1000 database) between ethanol-naive, male High Drinking in the Dark (HDID-1), and HS/Npt (the founder population) mice across eight brain areas. The sig_gutc tool reports a summary connectivity score (aggregating across doses, time points, and cell lines) between the input signature and signatures of more than 3000 chemical compounds in the LINCS-L1000 touchstone data set (comprised the most reproducible signatures for well-characterized compounds). Each compound had 32 scores (one for each input query). To identify compounds that consistently had negative scores across brain areas irrespective of the query parameter settings, we rank ordered compounds according to the 75th percentile score (out of the 32 scores) (a), and the number of significant connectivity scores (ie, ⩽−90, see text) (out of the 32 scores) (c). The boxplots summarizing the connectivity scores across the 32 input queries for the top 15 compounds from each prioritization method are shown in b and d. Pergolide and terreic acid, circled in red in a and c, were the top 2 hits from both selection approaches (e). A full color version of this figure is available at the Neuropsychopharmacology journal online.

Figure 2

Effects of the candidate drug, terreic acid, on binge drinking in High Drinking in the Dark (HDID-1) mice. Terreic acid significantly reduced binge-like ethanol (20%) intake (a) and blood alcohol levels (BALs) (b). NIAAA defines binge drinking as a pattern of drinking that produces BALs of 80 mg%, designated in b by a dashed line (at y=80). Although the highest dose of terreic acid tested reduced water intake (c), none of the doses tested significantly reduced intake of saccharin (8.5 mM) (d). Furthermore, doses of terreic acid that reduced ethanol intake and BAL did not alter locomotor activity (e). Results of Tukey’s post-hoc analysis are indicated above the SEM bars. In a and b, *P<0.05, **P<0.01, and ***P<0.001 vs the 0-dose groups and in c %=P<0.0001 for 15 vs 0 mg/kg, ^P<0.001 for 15 vs 5 mg/kg, and &P<0.01 for 15 vs 7.5 mg/kg terreic acid. Values represent mean±SEM (Drinking assays: n=12-24/sex/dose; data shown are collapsed on sex, thus each bar represents 24–47 mice; Activity assay: n=7–8/sex/dose).

Figure 3

Effects of the candidate drug, pergolide, on binge drinking in High Drinking in the Dark (HDID-1) mice. Pergolide significantly reduced binge-like ethanol (20%) intake (a) and blood alcohol levels (BALs) (b). Pergolide also significantly reduced water (c) and 8.5 mM saccharin (d) intake but did not alter locomotor activity (e). Results of Tukey’s post-hoc analysis are indicated above the SEM bars (*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001vs the 0-dose groups). Values represent mean ±SEM (Drinking assays: n=12–24/sex/dose; data shown are collapsed on sex, thus each bar represents 24–47 mice; Activity assay: n=7–8/sex/dose).