Transcriptional signatures of fentanyl use in the mouse ventral tegmental area

Abstract

Synthetic opioids such as fentanyl contribute to the vast majority of opioid‐related overdose deaths, but fentanyl use remains broadly understudied. Like other substances with misuse potential, opioids cause lasting molecular adaptations to brain reward circuits, including neurons in the ventral tegmental area (VTA). The VTA contains numerous cell types that play diverse roles in opioid use and relapse; however, it is unknown how fentanyl experience alters the transcriptional landscape in specific subtypes. Here, we performed single nuclei RNA sequencing to study transcriptional programs in fentanyl‐experienced mice. Male and female C57/BL6 mice self‐administered intravenous fentanyl (1.5 μg/kg/infusion) or saline for 10 days. After 24 h abstinence, VTA nuclei were isolated and prepared for sequencing on the 10× platform. We identified different patterns of gene expression across cell types. In dopamine neurons, we found enrichment of genes involved in growth hormone signalling. In dopamine‐glutamate‐GABA combinatorial neurons, and some GABA neurons, we found enrichment of genes involved in Pi3k‐Akt signalling. In glutamate neurons, we found enrichment of genes involved in cholinergic signalling. We identified transcriptional regulators for the differentially expressed genes in each neuron cluster, including downregulated transcriptional repressor Bcl6, and upregulated transcription factor Tcf4. We also compared the fentanyl‐induced gene expression changes identified in mouse VTA with a published rat dataset in bulk VTA, and found overlap in genes related to GABAergic signalling and extracellular matrix interaction. Together, we provide a comprehensive picture of how fentanyl self‐administration alters the transcriptional landscape of the mouse VTA that serves as the foundation for future mechanistic studies.

1. INTRODUCTION

Drug overdose deaths are at an all‐time high in North America, and over the last several years, the primary cause of opioid overdose has shifted from heroin to synthetic opioids. ¹ Increasingly, illicit and counterfeit drugs contain fentanyl, ² a potent synthetic opioid, which has led to an increase in fentanyl misuse and overdoses. Despite the increased presence of fentanyl in the drug supply, our understanding of how fentanyl alters the brain remains incomplete. Most of our mechanistic knowledge is derived from decades of work on natural and semisynthetic opioids like morphine and heroin. ³

Opioid exposure causes lasting molecular adaptations throughout the brain, especially in mesolimbic reward regions such as the ventral tegmental area (VTA). The VTA is especially important for mediating the rewarding aspects of opioids, ⁴ , ⁵ and rats will self‐administer intracranial morphine or fentanyl directly into the midbrain. ⁶ , ⁷ Acutely, opioids elevate extracellular dopamine concentrations, particularly in the downstream nucleus accumbens. ⁸ Activation of Gi coupled mu opioid receptors on VTA GABAergic interneurons causes hyperpolarization and subsequent disinhibition of VTA dopamine neurons. Chronic opioid exposure also increases VTA dopamine neuron firing, ⁹ , ¹⁰ , ¹¹ which is accompanied by transcriptional and epigenetic changes that support synaptic plasticity. ¹² , ¹³ , ¹⁴ However, the VTA contains numerous neuron subtypes aside from dopaminergic, including glutamatergic, and GABAergic projection neurons, along with more recently discovered combinatorial neurons that possesses synthesis and release machinery for multiple neurotransmitters. ¹⁵ , ¹⁶ , ¹⁷ There is increasing evidence that both GABAergic ¹⁸ and glutamatergic ¹⁹ neurons in the VTA help orchestrate reward‐related behaviours, and opioid exposure drives both glutamatergic and GABAergic synaptic plasticity in the VTA. ⁴ , ²⁰ However, most studies investigating the molecular adaptations in VTA relied on whole VTA homogenates, and it is difficult to attribute changes in bulk gene expression to a specific neuronal subtype. This is further exacerbated by a lack of true neuronal subtype marker, as tyrosine hydroxylase, the rate limiting enzyme for dopamine synthesis and traditional dopamine neuron marker, is expressed by multiple neuron populations. ²¹ , ²² , ²³ VTA homogenates also contain numerous glial cell types that may transcriptionally respond to opioids and contribute to the behavioural effects as recently demonstrated in other brain regions. ²⁴ , ²⁵ , ²⁶ Understanding the changes that happen in specific cell types is important, as we recently showed fentanyl exerts cell type‐specific effects on neurons in the nucleus accumbens that can be targeted to alleviate opioid abstinence induced negative affect. ²⁷

Here, we address this gap and ask how fentanyl alters the transcriptional landscape in different ventral tegmental area cell types. We used single nuclei RNA sequencing in male and female mice that self‐administered intravenous fentanyl. We integrated our data with the rat transcriptional atlas ¹⁷ to separate into known cell types, and found divergent transcriptional responses dependent on cell type. We further compared our findings in mouse with those from a recently published bulk VTA RNAseq study using a food versus fentanyl choice procedure in rats. ²⁸ Together, this work highlights the similarities and differences across cell type and species and lays the foundation for subsequent mechanistic investigations.

2. METHODS

2.1. Experimental subjects

All experiments were approved by the Institutional Animal Care and Use Committee at the Pennsylvania State University College of Medicine (PSUCOM) or University of Maryland School of Medicine (UMSOM) and performed in accordance with NIH guidelines for the use of laboratory animals. Mice were given food and water ad libitum and housed in the PSUCOM or UMSOM vivarium on a 12:12 h light: dark cycle with lights on at 07:00. Mice were 8‐ to 9‐week‐old male and female C57BL/6J mice bred at PSUCOM or UMSOM with original breeding pairs obtained from Jackson Labs. All mice were pair housed in corn‐cob bedding, provided with nestlets, and separated by a perforated acrylic divider.

2.2. Intravenous surgery

At 7–8 weeks of age, mice were anaesthetised with ketamine (100 mg/kg) and xylazine (12 mg/kg) and implanted with long‐term indwelling jugular catheters constructed from Micro Renathane tubing (Braintree Scientific, Braintree, MA) connected to a 24 gauge cannula (P1 Technologies, Roanoke, VA, USA) as in our previous work. ²⁹ Mice were flushed daily with 30 μL of heparinized saline containing enrofloxacin (400 IU/mL heparin, 0.227% enrofloxacin), and allowed to recover from surgery for >5 days.

2.3. Intravenous fentanyl self‐administration

One day prior to the start of self‐administration, mice were habituated to the operant chambers for 30 min. Mice then underwent 10 consecutive days of fentanyl or saline self‐administration, using procedures modified from our studies on cocaine. ²⁹ The first 5 days were under a fixed‐ratio 1 (FR1) schedule of reinforcement, followed by 5 days under FR2 (3 h/day, 1.5 μg/kg/infusion). Operant chambers (MED Associates, Saint Albans, VT, USA) had two nose‐poke holes on one wall, a cue‐light above each nose‐poke, and a house‐light in the middle of the opposite wall. At the start of the session, the house‐light and active nose‐poke were illuminated. Sufficient responses in the active nose‐poke triggered a 10 μL fentanyl infusion, turned off the house‐light, the active nose‐poke light, and illuminated the cue light above the active nose‐poke for 1 s. Any additional active responses during the 1 s infusion period were recorded but did not result in further infusions (‘timeout’). Responses on the inactive nose‐poke were recorded but were without programmed consequences. No prior training was used for FR1 acquisition. Mice used for single nuclei RNA sequencing were trained at UMSOM. Mice used for qRT‐PCR were trained at PSUCOM and also underwent a seeking test 24 h after the last self‐administration. The 1 h seeking test was performed under extinction conditions in which responses in the active nose‐poke resulted in cue presentations but no drug delivery. All behavioural testing was done during the light cycle between 08:00–12:00.

2.4. Tissue collection

Mice were euthanized by cervical dislocation, and brains were rapidly removed and chilled in ice cold PBS. Cold brains were cut into 1 mm coronal sections, and one tissue punch containing VTA (14 gauge) was flash frozen on dry ice and stored at −80 until processing. All tissue was collected at an equivalent fentanyl abstinence timepoint of ~24–26 h between 12:00 and 13:00. Tissue from the qRT‐PCR cohort was collected 4 h after the seeking test to minimize the influence of handling and behavioural testing on gene expression.

2.5. Statistics

Behavioural data were analysed using multiway repeated measures ANOVAs in Graph Pad Prism (v.10) and JASP (jaspstats.org), using drug and sex as ‘between factors’, and day and nose‐poke as ‘within factors’ where applicable. As there were no significant effects of sex, data were collapsed across sex. Sidak's multiple comparison correction was used unless noted otherwise. qRT‐PCR data were first analysed by two‐way ANOVA. As there were no significant effects of sex, data were collapsed by sex and analysed with unpaired, two‐tailed t‐tests.

Procedures for nuclei isolation, library preparation, sequencing, snRNA‐seq analysis, bulk RNA extraction, and quantitative RT‐PCR are detailed in Supplemental Methods.

3. RESULTS

To determine how fentanyl experience alters the transcriptional landscape in the ventral tegmental area, we trained male and female mice to self‐administer intravenous fentanyl or saline on a fixed ratio (FR) 1, then a FR2 schedule over 10 days (timeline in Figure 1A). Mice self‐administering fentanyl learned to discriminate between the active and inactive nose‐poke. (Figure 1B RM‐ANOVA: Day × Drug, F _{9, 108} = 3.96, p < 0.001; nose‐poke F _{1, 108} = 26.98, p < 0.001; Day × Nose‐Poke × Drug, F _{9, 108} = 3.66, p < 0.001, fentanyl active versus inactive, p = 0.003 Holm–Sidak post‐hoc), while mice self‐administering saline did not have a statistically significant difference between responses on active and inactive nosepoke (Figure 1C, saline active versus inactive, p = 0.07). There was no significant sex effect on the number of infusions earned (three‐way RM‐ANOVA, Day × Drug × Sex, F _{9, 108} = 0.6, p = 0.78), nor a significant difference in total fentanyl intake across the sexes (two‐way RM‐ANOVA, Day × Sex F _{9, 54} = 1.1, p = 0.38; female average total intake: 0.43 ± 0.14 mg/kg, male: 0.36 ± 0.04 mg/kg). We thus combined sexes for subsequent analysis. Mice self‐administering fentanyl earned more infusions relative to saline under FR2 (Figure 1D, two‐way RM‐ANOVA, Day × Drug F _{9, 126} = 2.8, p = 0.005. Fentanyl versus saline p < 0.05 on day 7, 8, 10, Sidak's post‐hoc; Total infusions under FR2: Fentanyl: 73 ± 9, Saline: 32 ± 5).

FIGURE 1.

Intravenous fentanyl self‐administration in mice. (A) Experimental timeline. Male and female mice (n = 4/group) underwent surgery for indwelling jugular vein catheter placement. Following recovery, mice administered fentanyl (1.5 μg/kg/10 μL infusion) or saline (10 μL) on a fixed‐ratio (FR)‐1 schedule for 5 days, then on a FR2 schedule for 5 days. Twenty‐four hours after the last self‐administration session, brains were removed and fresh tissue punches containing the ventral tegmental area (VTA) were flash frozen, then nuclei isolated and prepared for single nuclei sequencing on the 10× platform. (B) Mean ± SEM responses on the active and inactive nose‐poke in male (triangles) and female (circles) mice self‐administering fentanyl or (C) saline. (D) Mean ± SEM fentanyl or saline infusions earned collapsed across sex. *p < 0.05 fentanyl versus saline, Holm–Sidak post‐hoc after two‐way RM‐ANOVA.

After 10 days of self‐administration, we collected fresh tissue punches containing the ventral tegmental area 24 h after the last self‐administration session. We pooled dissociated nuclei based on sex and drug group, then prepared libraries for single nuclei RNA sequencing (snRNAseq) on the 10× platform. We captured a similar number of nuclei per sample across sex (Figure S1). We took two approaches to look at changes in gene expression in fentanyl experienced mice. First, using the FindMarkers function in Seurat, ³⁰ we looked for differentially expressed genes collapsed across all nuclei, in an attempt to mimic a bulk RNAseq experiment (Data S2, the 194 transcripts that reached a relaxed cut‐off of > 0.25 log2 fold change and the 794 < −0.25 are highlighted in the volcano plot in Figure 2A). In total, we found 1180 downregulated, and 220 upregulated transcripts with adjusted p < 0.05, and used this list for Gene Ontology (GO) analysis with Metascape. ³¹ We found overrepresentation of GO terms related to neuronal structure, cell‐adhesion, and synaptic transmission (top 10 GO terms in Figure 2B, all annotations in Data S1; gene lists in Data S2).

FIGURE 2.

Differentially expressed genes in VTA collapsed across cell type. (A) Volcano plot showing differentially expressed genes as determined with the Seurat FindMarkers function, collapsed across cell type. Genes with Bonferonni corrected p < 0.05 and ±0.25 fold change are highlighted in blue (downregulated) and yellow (upregulated). (B) Top 10 biological process gene ontology terms identified with Metascape. (C) Timeline for qRT‐PCR cohort. After recovering from jugular catheter surgery, mice (4–6/sex/group) self‐administered fentanyl or saline for 10 days. On day 11, mice were tested for drug seeking under extinction conditions in a 1 h test. Four hours after drug seeking, fresh VTA tissue punches were collected for subsequent bulk RNA extraction. (D) Mean ± SEM infusions earned by male and female fentanyl or saline self‐administering mice. Drug, F _{1, 19} = 6.6, p = 0.018. (E) Mean ± SEM responses in the active and inactive nose‐poke during the seeking test. Active saline versus active fentanyl, **p = 0.006, active fentanyl versus inactive fentanyl ***p = 0.0005, Sidak's post‐hoc. (F) Mean ± SEM fold change expression relative to GAPDH for selected gene targets. Hcn1: t _{1, 17} = 2.3, *p = 0.036; Pitx2: Welch‐corrected t _{1, 10.4} = 2.1, p = 0.064; Clk1 t _{1, 19} = 2.71 *p = 0.014; Neat1 t _{1, 19} = 2.35, *p = 0.02;Kdm6a t _{1, 19} = 4.38 ***p = 0.0003. Triangles denote datapoints from male mice and circles from female.

We next sought to confirm changes in bulk gene expression in an independent cohort of mice with qRT‐PCR (timeline in Figure 2C). In the qRT‐PCR cohort, mice self‐administering fentanyl earned more infusions compared with saline mice (Figure 2D, two‐way RM ANOVA, Drug, F _{1, 19} = 6.6, p = 0.018) and exhibited more drug seeking behaviour during a non‐reinforced seeking test (Figure 2E, two‐way RM ANOVA, nosepoke × drug F _{1, 19} = 4.6, p = 0.045, active saline versus active fentanyl: p = 0.006, fentanyl active versus fentanyl inactive p = 0.005, Sidak's post‐hoc). We selected four upregulated and four downregulated genes for validation based on (1) enriched GO terms, (2) moderate expression based on the Allen Brain Atlas, and (3) reliable qRT‐PCR primers with 100% primer efficiency. We reproduced several gene expression changes from snRNASeq: fentanyl experienced mice had upregulation of the hyperpolarization‐activated cyclic nucleotide gated channel Hcn1, trending upregulation of homeobox gene Pitx2, downregulation of the lysine demethylase Kdm6a, the kinase Clk1, and the long non coding RNA Neat1 (Figure 2F). Increased expression of Meg3 and Kcnq5, and decreased expression of Grm3 failed to replicate with this approach.

We next performed Uniform Manifold Approximation and Projection (UMAP) clustering to separate nuclei into specific cell types. Because half of our samples were from fentanyl exposed animals, we chose to integrate our data from mouse VTA with the 21 600 rat VTA nuclei data from the transcriptional atlas ¹⁷ (Figure S2A,C). This approach helped us prevent spurious clustering as the rat VTA atlas contains nuclei from drug‐naïve animals, and the cell type clusters were rigorously validated. To examine changes happening specifically in neuron subtypes, we removed non‐neuronal cell types and re‐clustered the rat and mouse data. (Figure 3). We separated the clusters into similar neurotransmitter identity clusters as in the transcriptional atlas by looking at co‐expression of synthesis and transport genes, as well as previously identified markers Gch1 and Slc26a7 (selected clusters in Figure 3C, all others in Figure S3. Cluster marker genes in Data S2). To ensure the rat data did not considerably alter cell type cluster assignments for mice, we looked at the UMAP with only mouse data (Figure S2B), and also performed independent clustering using solely mouse data (Figure S4). While both approaches produced a similar proportion of neuronal to non‐neuronal clusters, the mouse and rat integration was better at identifying multi‐neurotransmitter clusters (Figure S5, Table S2), likely due to the greater number of nuclei in the combined dataset. We thus performed all downstream analysis using clusters determined by rat and mouse integration.

FIGURE 3.

Nuclei cluster into single and multi‐neurotransmitter phenotypes. (A) UMAP of integrated mouse and rat neuronal nuclei. (B) Number of nuclei per cluster for rat and mouse data. Violin plots showing enrichment of cell type markers. Dopaminergic markers: Th, tyrosine hydroxylase, Slc18a2, vesicular monoamine transporter 2, Drd2, dopamine D2 receptor, Ddc, dopa decarboxylase, glutamatergic markers, Slc17a6, vesicular glutamate transporter, Grm2, metabotropic glutamate receptor 2, GABAergic markers, Slc6a1, GABA transporter protein type 1 (GAT1), Slc32a1, vesicular GABA transporter, Gad1, glutamate decarboxylase 1, Gad2, glutamate decarboxylase 2. Oprm1, mu opioid receptor. (C) Heatmaps of mouse nuclei co‐expressing genes involved in dopamine, glutamate, and GABA synthesis and transport.

We looked at changes in gene expression in each cluster using Libra ³² (Data S2). Here, we highlight the findings from four cell type clusters, namely, cluster 16 dopamine neurons, cluster 18 dopamine‐glutamate‐GABA neurons (‘combinatorial neurons’), cluster 13 glutamate neurons, and cluster 9 GABA neurons. We selected the latter two clusters based on low co‐expression of GABAergic and glutamatergic markers, respectively. In cluster 16 dopamine neurons, we found 407 upregulated and 403 downregulated genes in fentanyl‐experienced mice relative to saline (Figure 4A, Data S2). In cluster 18 combinatorial neurons, we found 381 downregulated, and 361 upregulated genes (Figure 4B; Data S2). We identified 44 shared differentially expressed genes between dopamine cluster 16 and combinatorial cluster 18, 13 of which exhibited opposite changes in expression (Figure 4C). We next performed gene set enrichment analysis in each cluster list with Metascape. ³¹ 565 differentially expressed genes in cluster 16 dopamine neurons were assigned to eight KEGG pathways (top 3 pathways and associated genes in Figure 4D, all annotations in Data S1), which indicated fentanyl altered expression of genes involved in ‘growth hormone action’, ‘calcium signalling’, and ‘cAMP signalling’, among others.

FIGURE 4.

Two dopamine clusters have divergent differential expression profiles. (A) Volcano plot showing differentially expressed genes in cluster 16 dopamine neurons and (B) cluster 18 dopamine‐glutamate‐GABA combinatorial neurons. Genes with Benjamini–Hochberg corrected p < 0.05 are highlighted in blue (downregulated) and yellow (upregulated). (C) Heatmap showing log fold change expression for the 44 shared differentially expressed genes between cluster 16 dopamine and cluster 18 combinatorial neurons. (D) Sankey plot showing the top 3 KEGG pathway terms for cluster 16 and (E) cluster 18 differentially expressed genes, the genes that are contained in that enriched pathway, and the log fold change expression in fentanyl versus saline within that cluster is encoded in colour, using the same scale as (C).

The top GO term for cluster 16 dopamine neurons was ‘transmembrane receptor protein tyrosine kinase signaling pathway’. Five hundred thirty‐five differentially expressed genes in cluster 18 combinatorial neurons were assigned to five KEGG pathways (top 3 in Figure 4E, all in Data S1), which indicated fentanyl altered expression of genes involved in ‘RNA polymerase activity’ and ‘phosphoinositide 3‐kinase/protein kinase B activity (Pi3k‐Akt)’. There was also enrichment for ‘hypertrophic cardiomyopathy’, which contains genes important for cell adhesion. The top GO term for cluster 18 combinatorial neurons was ‘DNA damage response’. Seven hundred eight genes in cluster 9 GABA neurons were assigned to four KEGG pathway terms (Data S1). Like dopamine neurons, this included ‘PI3K‐AKT signalling’. There was also enrichment for ‘Neuroactive ligand receptor interaction’, ‘axon guidance’, and the ‘Hippo signalling pathway’, which modulates cell survival. The top GO term for cluster 9 GABA neurons was ‘regulation of nervous system development’. Eight hundred eighty‐four differentially expressed genes in cluster 13 glutamate neurons were assigned to nine KEGG pathway terms, among them the notable ‘morphine addiction’, ‘cholinergic synapse’, and ‘adrenergic signalling’. The top GO term for cluster 13 glutamate was ‘monoatomic ion transmembrane transport’ (Data S1). Differentially expressed genes and enrichment analysis for the other six neuron clusters are available in Data S1 and S2.

We next took the list of differentially expressed genes in each cluster and looked for common transcription factor binding sites in gene regulatory regions (20 kb around TSS) with iRegulon ³³ (top 5 regulators for the selected clusters in Figure 5A–D, all in Data S1). We compared the lists of transcriptional regulators with the lists of differentially expressed genes in our selected neuron clusters, and selected differentially expressed transcriptional regulators for qRT‐PCR validation. Although not statistically significant, we found a trend towards similar decreased expression of cluster 9 regulator Bcl6, and increased expression of cluster 18 regulator Tcf4 (Bcl6: t _{1, 13} = 1.7, p = 0.1, Tcf4, t _{1, 15} = 2.0, p = 0.06, Figure 5E). We next looked for concordance in the patterns of expression of the transcription factor target genes by counting the number of differentially expressed target genes within the corresponding cluster (Example network of Bcl6 target genes in cluster 9 in Figure 5F).

FIGURE 5.

Predicted transcriptional regulators of differentially expressed genes differ across cluster. Top five predicted transcriptional regulators of differentially expressed genes in (A) cluster 9 GABA, (B) cluster 13 glutamate, (C) cluster 16 dopamine, and (D) cluster 18 combinatorial neurons. The iRegulon normalized enrichment score is on the y‐axis, and the number of predicted gene targets on the x‐axis. (E) Mean ± SEM fold change expression relative to GAPDH for selected predicted transcription factors in bulk VTA. Bcl6: t _{1, 13} = 1.7, p = 0.1, Tcf4, t _{1, 15} = 2.0, p = 0.06. (F) Network of genes regulated by cluster 9 regulator Bcl6. Log fold change fentanyl versus saline within cluster 9 is encoded in colour. (G) All predicted transcriptional regulators also differentially expressed in snRNASeq, and the number of their targets that are downregulated or upregulated within that cluster. Bcl6, Gata2, Ikzf1, Nfib, and Pou3f1 are cluster 9 regulators, the remainder are cluster 18. The log fold change fentanyl versus saline of that transcriptional regulator is encoded in colour, using the same scale as (F).

Most patterns were concordant with the expected direction, i.e. increased expression of Pou3f1, a positive regulator of transcription, was linked with a greater number of gene targets upregulated in cluster 9; decreased expression of transcriptional repressor Bcl6, was linked with a greater number of gene targets upregulated in cluster 9. Many of the predicted transcriptional regulators, including Tcf4, can both promote and suppress transcription. Increased expression of Arnt, Sox11, Tcf4, or decreased expression of Mitf, which have context dependent transcriptional regulatory effects, were associated with a mostly equivalent number up and downregulated target genes in cluster 18 (Figure 5G).

We next compared the genes identified in our mouse snRNASeq dataset with a recently published bulk RNAseq dataset from fentanyl self‐administering rats. ²⁸ We compared our differentially expressed genes to both the male and female rat VTA gene lists using three approaches. We first used the list we generated in Figure 2A, using the FindMarkers function and adjusted p < 0.05, and compared these genes to those reported as p < 0.05 differentially expressed in male or female rats. This resulted in only 11 species‐shared upregulated, and 49 downregulated genes, with the remainder of overlapping genes exhibiting opposite changes (Figure S6A). We next performed a threshold‐free, rank‐rank hypergeometic (RRHO) overlap analysis using the ‘bulk’ list generated in Figure 2A, and found greater concordance between our mouse data and the male rat data relative to female rat data (Figure S7B,C, Data S3, downregulated male: 168, female: 1; upregulated male: 36, female 4). We performed gene set enrichment with Metascape on the RRHO‐identified gene list, and the common up and downregulated genes were assigned to 18 and 19 KEGG pathways, respectively, including ‘morphine addiction’ and ‘long term depression’ (top 3 in Figure S7D‐E). We next compared our cluster specific lists determined by pseudo‐bulking (Figure S7A). This resulted in 124 species‐shared upregulated, and 83 downregulated. For the common upregulated and downregulated genes in this list, we performed gene set enrichment analysis with Metascape³¹. The common upregulated genes were assigned to five KEGG pathways (top 3 shown in Figure S7B), including ‘gabaergic synapse’, ‘neuroactive ligand receptor interaction’, and ‘ferroptosis’. The common downregulated genes were assigned to seven KEGG pathways (top 3 in Figure S7C), including extracellular matrix ‐receptor interaction, focal adhesion, and dilated cardiomyopathy, the latter pathway containing genes important for calcium signalling. The common up and downregulated genes between fentanyl experienced mouse and rat were spread across mouse cell type clusters. The top GO terms for rat and mouse common genes were all related to development (upregulated: ‘positive regulation of neuroblast proliferation, positive regulation of neurogenesis’, downregulated: ‘head development’, ‘brain development’).

4. DISCUSSION

Here, we identified different transcriptional responses to fentanyl self‐administration in the mouse VTA. Collapsed across cell type, we found downregulation of Kdm6a, Clk1, and Neat1, and upregulation of Hcn1 that replicated with bulk qRT‐PCR. We used a transcriptional atlas to separate cells into previously described types, and found many cell type‐specific transcriptional programs that were altered by fentanyl exposure. The cell type‐specific changes were in turn predicted to be regulated by different transcription factors. Finally, we showed many genes are differentially expressed in both fentanyl experienced rats and mice using a recently published dataset. Together, this work provides a detailed look into the transcriptional landscape of the fentanyl experienced ventral tegmental area and will serve as a launching point for many future mechanistic investigations.

Our primary goal in this work was to uncover cell type‐specific changes in gene expression in the VTA as a function of fentanyl experience. We chose to conduct this work in mice as they are the preferential model for transgenic lines, and any identified gene targets could be directly tested in genetically defined cell types. Mice learned to self‐administer intravenous fentanyl, as evidenced by discrimination of the active versus inactive nose‐poke, a greater number of infusions earned compared with saline control. While rats have been used more extensively for intravenous self‐administration studies and for more complex behavioural tasks, our mice earned a comparable number of saline versus fentanyl infusions as rats in a 3 h FR1 session, ³⁴ and the low number of drug infusions is typical of opioids versus psychostimulants. ³⁵ Mice also exhibited drug seeking behaviour, as evidenced by an increased responding under extinction conditions in the active but not inactive nose‐poke. One limitation in our study is that the transcriptional effects here are a combination of the pharmacologic actions of fentanyl, in addition to the act of self‐administration. Because mice only have fentanyl access for 3 h/day, the transcriptional changes here also reflect acute withdrawal effects that are absent in other models using 24 h access, or uninterrupted exposure such as through implanted morphine pellets or osmotic minipumps. Future studies could employ a yoked control to examine which changes in gene expression can be attributed to volitional intake versus mu opioid receptor activation, and provide continuous access to reduce the impacts of withdrawal.

Another limitation of this work is the low sample number. Due to the high cost of snRNAseq, we elected to pool our biological replicates such that each sample contained multiple mice from the same sex and drug condition. Pooling samples is common for snRNAseq studies ¹⁷ and allowed us to achieve a greater depth of sequencing coverage per sample. Because each sample contains four mice, any one potential outlier sample is diluted by the other three. Nevertheless, to strengthen confidence in our findings, we used bulk qRT‐PCR in a separate cohort of mice. The qRT‐PCR cohort underwent a single seeking session under extinction conditions and is thus not an identical replication. However, both groups were collected at similar pharmacologic timepoints to ensure no differences in fentanyl withdrawal time, and several differentially expressed genes replicated despite these differences. In our experience, a single 1 h session is insufficient to extinguish responding and mice have high rates of responding during a second extinction session. ²⁹ We thus do not believe the qRT‐PCR cohort is capturing the molecular signatures of extinction learning. However, it is possible the genes that did not replicate between qRT‐PCR and snRNAseq are due to the seeking session, which will require additional investigation. There are two additional potential sources for the discrepancy. First, snRNAseq only captures nuclear transcripts, while bulk also captures cytoplasmic RNA. Second, the bulk qRT‐PCR lacks cell type specificity, which we suspect drives the lack of statistical significance for the cell type transcriptional regulators in Figure 5. Our future work will use more cell type‐specific approaches such as fluorescence in situ hybridization and translating ribosome affinity purification in Cre/Flp driver lines to further validate cell type‐specific findings. However, even with this additional layer of validation, the gene candidates will still have to be tested for causal roles in fentanyl use and relapse behaviour, so we have chosen to disseminate this data in its present form so as to facilitate more expedient mechanistic inquiries. It is also worth noting that we were inadequately powered for sex differences in our snRNAseq, and thus, potential sex‐specific genes may not be detected. While we did not see sex differences with our qRT‐PCR experiments, this does not rule out sex‐specific transcriptional programs like those found in rat, ²⁸ and does not indicate that similar gene expression changes in males and females have identical physiologic and behavioural consequences.

With our cluster specific analysis using Libra, we found several interesting patterns of gene expression changes that both converge and diverge from other published studies on the transcriptional consequences of opioids. In dopamine neurons, we found mostly upregulation of genes associated with calcium signalling, mostly downregulation of genes associated with growth hormones, and both up‐ and down‐regulation of genes involved in tyrosine kinase receptor signalling. Of the differentially expressed genes in these categories, decreased expression of Igf1 (insulin‐like growth factor 1) is a standout, as recent work in the prefrontal cortex showed decreased IGF1 protein in mice that self‐administered oral fentanyl, and IGF1 replacement in the PFC attenuated fentanyl seeking behaviour. ³⁶ Decreased expression of Calcr (calcitonin receptor) is another intriguing target, as Calcr expression has bidirectional effects on opioid self‐administration in nucleus accumbens neuron subtypes. ³⁷ The top upregulated genes in dopamine neurons was the extracellular matrix protein Ecel1, and the actin polymerization regulator Diaph3, both of which are implicated in synaptic function and structural remodelling. These will be interesting gene targets to explore, as chronic morphine is known to induce structural plasticity of VTA dopamine neurons, ¹¹ , ³⁸ , ³⁹ and extracellular matrix proteins are an area of interest in the opioid use disorder field. ⁴⁰ We found increased expression of Insulin receptor substrate Irs1 and Irs2, opposite of the decreased Irs proteins after chronic morphine in rat VTA. ⁴¹ Other similarities and differences with our data in fentanyl self‐administration dopamine neurons and that of chronic morphine mouse total VTA ⁴² (see Data S3) are a concordant increases in the zinc finger protein Bcll1b, the kinase Dclk3, the postsynaptic protein Synpo, and a discordant decrease in the kinase Sgk1, synaptic vesicle protein Synpr, and increase in kinase Hipk2. Of our selected neuron clusters, cluster 9 GABA neurons had the most overlap with genes in Heller et al ⁴² , with concordant downregulation of transcription factor Alx4, and upregulation of ankyrin domain protein Ankrd33b, transcriptional regulators Arx, Ddn, Lhx6, adapter protein Baiap2, long noncoding RNA Dlx6os1, deaminase Gda, GTPase regulators Ngef, Rab40b, and cytoskeletal regulator Wipf3. There was discordant decrease in kinase anchoring protein Akap5, cAMP phosphoprotein Arpp21, and discordant decreases and increase in transcription factors Rarb, and Tcf712, respectively. There were only three common genes in cluster 13 glutamate (concordant down actin protein Actg2, concordant up Wipf3, discordant down phosphodiesterase Pde10a), and two in cluster 18 combinatorial (discordant down Baiap2, concordant up G protein subunit Gng7). It will be important to determine if these common up‐and down‐regulated genes play a causal role in volitional opioid intake and relapse, as recent work showed that deleting Sgk1 in dopamine neurons, upregulated after chronic morphine and cocaine, does not disrupt drug reward behaviours, and only a catalytically inactive mutant decreases cocaine conditioned place preference, but not self‐administration. ⁴³ It will also be important to determine if the discordant changes in expression are due to differences in the opioid, its route of administration (volitional vs. experimenter), or repeated episodes of spontaneous withdrawal.

We found some overlap in the differentially expressed genes within the two dopamine neuron populations, including concordant up‐regulation potassium channel Hcn1, and cAMP binding protein Creb5. These targets are of particular interest as chronic opioids are known to disrupt expression of potassium channels, ¹¹ and cAMP‐mediated upregulation of HCNs in dopamine neurons was recently shown to promote cocaine self‐administration. ⁴⁴ We also found enrichment for Pi3k‐Akt signalling pathway genes in the combinatorial neurons and our selected GABA neurons cluster. This included upregulation of neurotrophic factors Ngf and Gdnf. However, this upregulation in Gdnf is contrary to the decreased Gdnf mRNA in rat VTA after an equivalent abstinence period from heroin self‐administration. ⁴⁵ One of the most interesting findings was an upregulation of mRNA for the mu opioid receptor, Oprm1, in cluster 13 glutamate neurons. Recent work indicates that mu opioid receptors are in fact, expressed on VTA glutamate neurons, and that they modulate excitatory transmission onto neurons that release dopamine into the nucleus accumbens core. ⁴⁶ We also found decreased expression of adrenergic receptor Adra1b. Activation of α1 receptors increases glutamate release onto VTA dopamine neurons. ⁴⁷ Together, increased Gi signalling from mu opioid receptors and decreased signalling through α‐1 receptors would reduce glutamatergic input to dopamine neurons and may be a potential source of decreased dopamine release seen after opioid experience. ⁴⁸ The numerous gene targets that control cellular excitability will be interesting candidates to pursue as the field of non‐dopaminergic mechanisms in the VTA continues to expand.

To further explore the patterns of differential expression in the cell types, we identified transcription factors that have binding sites upstream of differentially expressed gene promoters, as altered transcription factor expression likely contributes to the differential patterns of gene expression within each cell type cluster. Many of the transcription factors we identified were also differentially expressed in our sequencing data, and in a direction concordant with their known regulatory effects. One of these differentially expressed transcription factors, Sox11, is also upregulated in the nucleus accumbens after chronic morphine. ¹⁰ Identifying the common transcriptional regulators is a useful tool in a candidate gene discovery pipeline, as it is impossible to test how every individual differentially expressed gene mediates behaviour. Targeting a transcription factor allows one to manipulate expression of many different genes at once, which is a strategy we have successfully used to change transcriptional programs in the fentanyl abstinent nucleus accumbens. ²⁷

We also sought to identify gene expression changes that persist across multiple fentanyl self‐administration models. We found a number of common differentially expressed genes in the VTA from fentanyl self‐administering mice and rats. ²⁸ Many of the commonly upregulated genes were involved in GABAergic signalling, and included two genes for GABA(A) receptor alpha subunits, Gabra1, Gabra5, and GABA synthesis enzyme Gad2. VTA GABA signalling is augmented by both mu opioid receptor activation and morphine withdrawal, ⁴⁹ and chronic morphine produces a similar upregulation of α1 and α5 subunits in hippocampus, ⁵⁰ but their role in mediating opioid use and relapse has not yet been investigated. The top KEGG pathway for shared downregulated genes in rats and mice was extracellular matrix receptor interaction. Downregulation of the extracellular glycoprotein Reln is especially interesting, as global reduction in Reelin protein is associated with increased locomotor sensitization to cocaine. ⁵⁰ We can find no mention of most of the common differentially expressed genes between rats and mice regarding a role in opioid related behaviour, but as they are preserved across species and fentanyl intake paradigms, they warrant future follow up.

Overall, we have provided a comprehensive look at the transcriptional consequences of fentanyl use in the mouse ventral tegmental area, including diverse changes across cell types. It will be important to follow up on the top gene targets and their common transcriptional regulators in a cell type‐specific manner to establish a causal role for the gene targets in driving opioid use and relapse behaviour. By highlighting the similarities and differences across cell type and species, we have laid the foundation for numerous future mechanistic investigations.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to disclose.

Supporting information

Data S1 Supporting Information.

Data S2 Supporting Information.

Data S3 Supporting Information.

Table S1: Primer sequences for qRT‐PCR.

Figure S1. Uniform approximation and projection (UMAP) of integrated mouse nuclei, coloured by sex.

Figure S2. (A) Uniform approximation and projection (UMAP) of integrated mouse and rat nuclei (‘mouse‐rat integration’). (B) UMAP of mouse only subset. (C) Number of nuclei per cluster for rat and mouse data. Violin plots showing enrichment of cell type markers. Neuronal markers: Slc32a1, vesicular GABA transporter, Gad, glutamate decarboxylase 1 and 2, Oprm1, mu opioid receptor, Grm2, metabotropic glutamate receptor 2. Th, tyrosine hydroxylase, Slc18a2, vesicular monoamine transporter 2, Slc6a3, dopamine transporter. Endothelial marker, Eng, endoglin. Polydendrocyte marker Pdgfra, platelet derived growth factor receptor a. Mural cell marker, Pdgfrb. Oligodendrotyce precursor markers, Dock6, dedicator of cytokinesis 6, Hist1h2an, histone cluster 1, H2an, Oligodendrocyte marker, Mobp, myelin associated oligodendrocyte basic protein; Microglia marker, Arhgap15, Rho GTPase activating protein 15, Astrocyte marker, Aqp4, aquaporin 4.

Figure S3. Heatmaps of mouse nuclei clusters co‐expressing genes involved in synthesis and transport of GABA, glutamate, and dopamine. Clusters are from ‘mouse‐rat integration’. Clusters 9,13,16 and 18 are in Figure 3.

Figure S4. UMAP clustering of nuclei from mouse without rat data integration (‘mouse‐only clustering’).

Figure S5. Heatmaps of mouse nuclei clusters co‐expressing genes involved in synthesis and transport of GABA, glutamate, and dopamine. Clusters numbers are from ‘mouse‐only clustering’.

Table S2. Number of nuclei in clusters as determined by the mouse and rat integration clustering, and the mouse only clustering.

Figure S6. (A)Venn Diagram showing overlap between differentially expressed genes in the mouse dataset in this paper, and the rat data in Townsend et al(2021) with mouse gene lists taken from the FindMarkers padj<0.05. Rank‐rank hypergeometric overlap analysis showing threshold free overlap between the ‘bulk’ mouse gene lists taken from FindMarkers and the Townsend et al (B) male rat and (C) female rat data. Arrow legends indicate expression directionality in mouse and rat data (e.g. ↑↓ is up mouse, down rat). Sankey plots showing Top 3 KEGG pathway terms in the mouse and male rat common (D) upregulated and (E) downregulated genes. Note some genes appear to express in opposite directions in Supplemental Figure 6 due to differences in cell type specific vs cell type independent gene lists.

Figure S7. (A)Venn Diagram showing overlap between differentially expressed genes in the mouse dataset in this paper, and the rat data in Townsend et al(2021) with mouse gene lists taken from pseudobulking with Libra. Sankey plots showing Top 3 KEGG pathway terms in the mouse cluster‐specific and rat common upregulated genes (B) and (C) downregulated genes. The common genes are shown as belonging to numbered mouse neuron clusters, and if they were differentially expressed in male or female rats.

Fox ME, Montemarano A, Ostman AE, et al. Transcriptional signatures of fentanyl use in the mouse ventral tegmental area. Addiction Biology. 2024;29(5):e13403. doi: 10.1111/adb.13403

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in the supplementary material of this article.