An atlas of transcriptionally defined cell populations in the rat ventral tegmental area

SUMMARY

The ventral tegmental area (VTA) is a complex brain region that is essential for reward function and frequently implicated in neuropsychiatric disease. While decades of research on VTA function have focused on dopamine neurons, recent evidence has identified critical roles for GABAergic and glutamatergic neurons in reward processes. Additionally, although subsets of VTA neurons express genes involved in the synthesis and transport of multiple neurotransmitters, characterization of these combinatorial populations has largely relied on low-throughput methods. To comprehensively define the molecular architecture of the VTA, we performed single-nucleus RNA sequencing on 21,600 cells from the rat VTA. Analysis of neuronal subclusters identifies selective markers for dopamine and combinatorial neurons, reveals expression profiles for receptors targeted by drugs of abuse, and demonstrates population-specific enrichment of gene sets linked to brain disorders. These results highlight the heterogeneity of the VTA and provide a resource for further exploration of VTA gene expression.

Graphical Abstract

In brief

The ventral tegmental area (VTA) is a heterogeneous brain region implicated in motivated behavior and psychiatric disorders. Phillips et al. use snRNA-seq to comprehensively define the molecular architecture of the rat VTA, identifying novel markers for key cell types and providing a resource for exploration of VTA gene expression.

INTRODUCTION

Dopaminergic signaling in the mammalian brain is essential for reward learning, motivation, and motor coordination. The brain circuits critical for dopamine neurotransmission are highly conserved across vertebrate evolution, and dysregulation of this circuitry has been linked to substance abuse (Savell et al., 2020; Volkow et al., 2007a), neurodevelopmental disorders, and neuropsychiatric disease (Cousins et al., 2009; Hamilton et al., 2013; Perez-Costas et al., 2010; Volkow et al., 2007b). For decades, the ventral tegmental area (VTA) has been studied for its role in dopaminergic neurotransmission and reward processes. The VTA is a complex brain region that sends dense projections of dopamine (DA) neurons to other reward-related areas of the brain, such as the nucleus accumbens (NAc) and prefrontal cortex. DA neuron firing and release in terminal regions dynamically encodes rewarding stimuli as well as cues that predict rewards (Stuber et al., 2010; Coddington and Dudman, 2018; Day et al., 2007; Kim et al., 2020; Phillips et al., 2003; Saunders et al., 2018b) and is thought to provide a learning signal consistent with reward prediction error computational functions (Kim et al., 2020; Lerner et al., 2020; Schultz et al., 1997). Consistent with this role, DA neurotransmission is critical for reward-related behavioral conditioning (Di Ciano et al., 2001; Wise, 2004), and phasic activation of VTA DA neurons is sufficient for reward learning (Tsai et al., 2009).

Despite a heavy focus on dopaminergic neurotransmission in the VTA, there is also substantial evidence for the importance of VTA γ-aminobutyric acid (GABA)ergic and glutamatergic neurotransmission in reward-related behaviors. For example, the activation of μ-opioid receptors on GABAergic interneurons in the VTA results in the depolarization of DA neurons (Gysling and Wang, 1983; Johnson and North, 1992). In the absence of DA release, optical stimulation of VTA glutamate neurons is sufficient for positive reinforcement (Zell et al., 2020). This evidence suggests a role for multiple neurotransmitter systems in VTA function, which is further supported by molecular characterization studies that have identified heterogeneous populations of dopaminergic, glutamatergic, and GABAergic neurons (La Manno et al., 2016; Li et al., 2013; Morales and Margolis, 2017; Saunders et al., 2018a; Stamatakis et al., 2013; Tiklová et al., 2019; Viereckel et al., 2016; Yamaguchi et al., 2007). Specifically, DA neurons (neurons expressing genes involved in both the synthesis and transport of DA) are concentrated in the lateral area of the VTA, while glutamate neurons (expressing VGLUT2 mRNA), but not protein for tyrosine hydroxylase (TH, a common DA neuron marker), are concentrated in the medial VTA (Yamaguchi et al., 2011). Interestingly, parallel lines of evidence have demonstrated the presence of “combinatorial” neurons that co-express genes involved in the synthesis and transport of multiple neurotransmitters (Granger et al., 2017; Morales and Margolis, 2017; Stamatakis et al., 2013; Tritsch et al., 2012, 2014; Yamaguchi et al., 2015; Yoo et al., 2016), suggesting an additional layer of complexity in VTA cellular and synaptic function (Granger et al., 2017; Kim et al., 2015).

Previous studies investigating the molecular heterogeneity of cell types within the VTA have typically relied on low-throughput methods that do not provide comprehensive information on transcriptional diversity within this brain region. Recent advances in single-cell RNA sequencing (scRNA-seq) technologies have allowed for the interrogation of whole transcriptomes from many cells within the VTA, circumventing previous technical limitations. These studies have provided key insights into transcriptomic diversity within VTA cell types (Saunders et al., 2018a), DA neuron development (Tiklová et al., 2019), and VTA cell type composition similarities between species (La Manno et al., 2016). However, these studies have either focused on a single-cell type using a previously defined marker gene or have combined other midbrain regions together for analysis. Furthermore, these studies used mice and have not included sex as a biological variable. Here, we performed single-nucleus RNA sequencing (snRNA-seq) on 21,600 VTA nuclei from both male and female adult Sprague-Dawley rats, a commonly used model organism for studies of reward function and substance abuse. This dataset confirmed the presence of combinatorial neurons and identified marker genes for combinatorial neurons as well as classically defined DA neurons. Moreover, we demonstrate cell-type-specific enrichment for gene sets implicated in multiple genome-wide association studies (GWASs) for neurodegenerative, neuropsychiatric, and neurodevelopmental diseases as well as addiction-related phenotypes. This atlas, available as a searchable online resource (www.day-lab.org/resources), highlights additional avenues for regulation and identification of selected VTA cellular populations.

RESULTS

Identification of transcriptionally distinct cell populations in the rat VTA

To investigate cellular heterogeneity within the VTA, we used the 10x Chromium platform to perform snRNA-seq on 21,600 individual VTA nuclei from male and female adult Sprague-Dawley rats (Figures 1A and S1). Following data integration, unsupervised dimensionality reduction techniques revealed 16 transcriptionally distinct cell populations (Figures 1B–1H and S2; Table S1) that were equally represented in both sexes (Figure S3). Additionally, calculation of the local inverse Simpson’s index (Korsunsky et al., 2019) demonstrated that these populations were well mixed and contained cells from each sample, indicating successful integration (Figure S2). These cell populations included well-characterized glial cell types as well as VTA DA neurons that selectively express Th and Slc18a2 mRNA, genes encoding TH and the vesicular monoamine transporter 2 (VMAT2) (Figures 1B–1D and Table S2). Th and Slc18a2 were chosen as markers of DA neurons because cells that selectively express these genes have the molecular machinery required to both synthesize and package DA into vesicles for release (Morales and Margolis, 2017; Poulin et al., 2020).

Figure 1. Transcriptional atlas of cell types in the rat VTA.

(A) snRNA-seq workflow. Tissue was harvested from naive rats (n = 6/sex) prior to nuclei isolation and snRNA-seq.

(B) Uniform approximation and projection (UMAP) of cell types in the rat VTA (no. of cells/cluster listed in parentheses).

(C–G) Feature plots of expression values for markers of dopaminergic, glutamatergic, and GABAergic neurons within cell types in the VTA.

(H) Bar graph indicating number of cells per cluster (top) and violin plots of major marker genes for identified clusters (bottom).

To compare the VTA DA neuron population identified here against subtypes previously described in scRNA-seq datasets generated from adult mouse midbrain neurons (La Manno et al., 2016; Saunders et al., 2018a; Tiklová et al., 2019), we performed unsupervised dimensionality reduction exclusively on DA neurons. While canonical DA neuron markers Ddc, Th, Slc6a3, and Slc18a2 were ubiquitously expressed within this cluster (Figure S4C, top row), we also observed a localized subset of Sox6⁺/Aldh1a1⁻ cells, a population previously identified in the parabrachial pigmented region and retrorubral area of the VTA (Poulin et al., 2020) that also co-expresses Lypd1, Sncg, and Tacr3 (Figure S4C, second row). We also find Aldh1a1⁺ cells co-enriched with Slc17a6, Grp, and Neurod6, similar to the Aldh1a1⁺/Slc17a6⁺/Otx2⁺ population previously described in the ventromedial VTA (Poulin et al., 2020) (Figure S4C, third row), with the exception that Otx2 was not detected in our dataset. Although limited, we also find a discrete pocket of Vip⁺ cells within the DA neuron cluster co-enriched for Calb1 and Gipr, similar to that described by Poulin et al. (2020) (Figure S4C, bottom row). However, Vip expression is modest in the present dataset, likely because these cells primarily reside in a region slightly dorsal to the VTA (Poulin et al., 2014, 2020). Similarly, we found few cells representative of the Aldh1a1⁺/Sox6⁺ subtype, a DA neuron population that is restricted to the substantia nigra (Poulin et al., 2020).

A wealth of literature has elucidated mechanisms by which subsets of VTA GABAergic neurons modulate DA neuron activity, and this comprehensive snRNA-seq approach also enabled simultaneous study of the unique transcriptional architecture of these populations in a single dataset. Unsupervised dimensionality reduction identified three populations of GABAergic neurons marked by high levels of Gad1, the gene encoding glutamate decarboxylase 1 (GAD1), and Slc32a1, the gene encoding the vesicular GABA transporter (VGAT) (Figures 1E and 1F). In addition to the expression of canonical GABAergic marker genes, GABA-Neuron-1 and GABA-Neuron-2 express Htr2c, the gene encoding the serotonin 2C receptor, and likely represent previously identified cell populations within the VTA (Bubar and Cunningham, 2007) (Figure 1H). The GABA-Neuron-3 cluster exhibited high expression of Oprm1, the gene encoding the μ-opioid receptor (Figure 1H). Agonists at this receptor produce hyperpolarization of these cells and subsequent disinhibition of dopaminergic neurons, which may contribute to the reinforcing nature of opioids (Gysling and Wang, 1983; Johnson and North, 1992). Two of these GABAergic populations, GABA-Neuron-2 and GABA-Neuron-3, also exhibited expression of canonical interneuron markers Kit and Sst. To investigate these putative interneuron populations we again performed subclustering analysis, which identified five transcriptionally distinct cell populations (Figures S5A and S5B). Whereas all populations expressed Nos1, a gene found to mark non-projecting GABAergic neurons in the VTA (Paul et al., 2018) (Figures S5F and S5G), population 3 exhibited selective expression of Sst (Figures S5D and S5G). Similarly, population 2 exhibited selective expression of Vip (Figures S5E and S5G), and populations 0, 1, and 4 expressed Kit (Figures S5D and S5G), a marker of parvalbumin interneurons. Notably, these populations also differed in their expression of Oprm1 and Htr2c (Figure S5D).

Similar to the GABAergic heterogeneity identified in the VTA, unsupervised clustering techniques also identified three populations of glutamatergic neurons, each of which highly expressed Slc17a6, the gene encoding the vesicular glutamate transporter 2 (VGLUT2) (Figures 1G and 1H). However, only one of these populations (Glut-Neuron-2) expressed Grm2, a group II metabotropic glutamate receptor that acts as an autoreceptor at excitatory synapses (Muguruza et al., 2016; Schoepp, 2001; Shigemoto et al., 1997) (Figure 1H). The co-expression of Slc17a6 and Grm2 within this cluster suggests that these cells represent classically defined VTA glutamatergic neurons. Another one of these glutamatergic neuron populations (Glut-Neuron-3) exhibited high expression of Pnoc, the gene encoding a pre-proprotein for nociceptin (Figure 1H). This neuronal population likely represents a recently characterized cell type that negatively regulates motivation for reward (Parker et al., 2019). The third glutamatergic population expressing Slc17a6 (Glut-Neuron-1) did not express Pnoc or Grm2. However, a small subset of cells within this population exhibited expression of Th, Gad1, Slc18a2, and Slc32a1 (Figures 1C–1F). The co-expression of Slc17a6 and genes involved in other neurotransmitter systems suggests that this subpopulation may be capable of co-release or co-synthesis of multiple neurotransmitters.

Finally, as a confirmation that our VTA subdissections did not include other nearby structures, we examined expression patterns of genes that are known to mark nearby regions such as the substantia nigra pars compacta, interpeduncular nucleus (IPN), and rostromedial tegmental nucleus (RMTg). As noted above, DA neurons identified in our dataset lacked co-expression of Aldh1a1 and Sox6, which are both expressed in substantia nigra DA neurons (Poulin et al., 2020). Likewise, we failed to observe robust expression of IPN markers such as Epha8, Sox11, Cyp26b1, and Efna2 (García-Guillén et al., 2021), or RMTg markers such as Tph2 (Smith et al., 2019).

Sex comparison of gene expression across cell populations in the rat VTA

Increasing evidence suggests that biological sex plays a critical role in VTA function, with numerous dopamine-dependent behavioral phenotypes displaying sex differences (Becker and Chartoff, 2019; Diekhof et al., 2021; Rivera-Garcia et al., 2020). We next split the merged dataset by sex to examine potential baseline differences in gene expression across 16 cell populations. Similar representation from male- and female-isolated nuclei was identified within each cell population (Figures S3A–S3C). Next, to identify differentially expressed genes (DEGs) between males and females for each cell cluster, we performed Wilcoxon ranked-sum tests on log-normalized gene counts (Table S3). To strengthen confidence in DEG testing and to avoid potential false-positive or false-negative errors due to insufficient sample size for smaller populations, we completed power calculations for each cluster. Cell populations insufficiently powered for DEG testing (i.e., clusters containing low cell counts that would not allow for 80% power in detecting log₂ fold changes >1) were not included in subsequent DEG analyses. For clusters that met power analysis criteria, significant sex-linked DEGs were observed on the X, Y, and autosomal chromosomes in all clusters (Figure S3D). DEG analyses confirmed sexually dimorphic expression patterns for several genes previously reported to exhibit a sex bias. For example, expression of the transcription factor Peg3 was biased in the male direction across nearly all clusters (Figure S3E, left), a finding consistent with previous work identifying this autosomal gene as paternally imprinted (Kaneko-Ishino et al., 1995; Kim et al., 1997). Expression of Kdm6a, a histone lysine demethylase expressed on the X chromosome, was biased in the female direction in all cell populations, consistent with previous work suggesting that this gene escapes X inactivation in several eutherian species (Greenfield et al., 1998) (Figure S3E, center). Interestingly, DEG testing also revealed some genes with population-specific sex differences in expression, including the orphan nuclear receptor Esrrg (Figure S3E, right). This transcription factor closely related to the estrogen receptor family exhibited a unique, cluster-specific pattern of expression, biased toward females in DA neuron, astrocyte, and Glut-Neuron-1 and −2 populations, but biased toward males in the GABA-Neuron-1 population. These results suggest a potential avenue for sex-specific transcriptional regulation at estrogen response elements in these populations.

Subclustering of neuronal cells identifies populations of combinatorial neurons

Emerging evidence suggests that within the VTA, small subsets of neurons can synthesize and/or release more than one neurotransmitter (Li et al., 2013; Morales and Margolis, 2017; Root et al., 2020; Tritsch et al., 2012, 2014; Yamaguchi et al., 2015; Yoo et al., 2016). To explore neuronal populations more thoroughly, we first subclustered all neuronal cells within the dataset and identified 11 transcriptionally distinct neuronal populations (Figure 2A). With increased resolution to detect smaller neuronal populations, this analysis revealed several clusters expressing markers corresponding to more than one neurotransmitter system (Figure 2B). To provide direct evidence for the presence of combinatorial neurons, we examined the proportion of cells within a single cluster co-expressing unique genes involved in the synthesis or release of GABA, glutamate, or DA (Figure 2C). In this analysis, a neuronal population that uses a single neurotransmitter is expected to contain a significant number of cells co-expressing genes critical to the synthesis, packaging, and release of that neurotransmitter. We detected five neuronal clusters which appear to represent selective GABAergic (clusters 1, 4, and 6), glutamatergic (cluster 3), and dopaminergic neurons (cluster 5) (Figures 2B and 2C). However, ~20% of the cells within each “selective” cluster co-express at least a single gene involved in another neurotransmitter system, suggesting inherent promiscuity in the co-expression of genes important for neurotransmitter synthesis, transport, and release. In contrast to these classically defined “single neurotransmitter” populations, we also detected additional clusters in which markers for multiple neurotransmitter systems were more abundantly co-expressed (clusters 2, 8, 9, and 10), which may represent neurons capable of combinatorial neurotransmitter release. For example, cluster 2 represents a population of glutamatergic neurons that co-expresses genes involved in the synthesis of DA and transport of GABA. Likewise, cells in cluster 8 expressed markers of GABAergic neurons, but also co-expressed Slc17a6 mRNA. While clusters 2 and 8 primarily co-express marker genes involved in the synthesis or transport of two neurotransmitters, few cells within these clusters also express marker genes that would allow for the synthesis and transport of three neurotransmitter systems. Interestingly, cells in cluster 9 expressed at least one gene involved in the synthesis and transport of glutamate, GABA, and DA. However, these cells are devoid of Grm2 and Gad1 mRNA, two genes that are highly expressed in classically defined glutamatergic and GABAergic cell populations. Previous research has identified populations of neurons that co-express Th and Slc17a6 (Li et al., 2013; Root et al., 2016) Th and Gad1/2 (Stamatakis et al., 2013), and Th and Slc32a1 mRNA (Stamatakis et al., 2013). Consistent with this evidence, a previous report also identified VTA neurons that displayed co-expression of VGLUT2 mRNA, GAD mRNA, and TH at the protein level as well (Root et al., 2014).

Figure 2. Subclustering of VTA neurons reveals co-release and co-synthesis populations.

(A) Subclustering of neuronal populations identified 11 distinct cell types (no. of neurons/cluster listed in parentheses).

(B) (Top) Dendrogram and bar graph of population numbers. (Bottom) Violin plots for genes involved in the synthesis and transport of DA, glutamate, and GABA.

(C) Heatmaps of cells co-expressing genes involved in synthesis and transport of DA, glutamate, and GABA.

Neuronal subcluster analysis identifies cell-type-specific marker genes

To identify markers of well-studied and recently characterized cell types in the VTA, we focused on neuron subcluster 5 (classically defined DA neurons) and neuron subcluster 9 (combinatorial neurons). Using a Wilcoxon ranked-sum test (Mann and Whitney, 1947), we conducted differential expression testing to identify enrichment of genes in one neuronal subcluster versus all other neuronal clusters (Table S4). Classically defined DA neurons were enriched for genes involved in the synthesis and release of DA, such as Th, Slc18a2 (VMAT2), and Slc6a3 (DAT) (Figures 3A and S6). While Th is widely used as a marker of DA neurons, only36.8% of all Th-expressing neurons are found in cluster 5 (classically defined DA neurons). DA neurons were also enriched for Dlk1, a gene that encodes a transmembrane protein involved in the differentiation of DA neurons (Bauer et al., 2008; Christophersen et al., 2007), and Gch1, a gene encoding the enzyme guanosine triphosphate (GTP) cyclohydrolase 1. Mutations in Gch1 have been implicated in genetic predisposition to dystonia and Parkinson’s disease (Lewthwaite et al., 2015; Pan et al., 2020; Rudakou et al., 2019; Yoshino et al., 2018), two diseases in which dysregulation of DA has been extensively studied. Comparisons of DEGs for clusters 5 and 9 found that classic markers of DA neurons, such as Th, Slc18a2 (VMAT2), and Slc6a3 (DAT) mRNA, were shared between the two clusters (Figure 3B). While DEG testing allows for the identification of specific marker genes of neuronal subclusters, it does not provide any information for the inherent selectivity of gene expression patterns across all clusters. Thus, we employed the Gini coefficient, a statistic previously used to measure wealth inequality(Gini, 1921), to characterize the selectivity of all detected genes. Here, a Gini coefficient of 0 indicates that gene is expressed in every neuronal subcluster equally, and a Gini coefficient of 1 indicates exclusive expression in a single neuronal subcluster. The Gini coefficient for Gch1 is 0.76 (Figure 3B), indicating that this gene exhibited a population-selective expression pattern. Furthermore, plotting the distribution of expression values for Gch1 across all neuronal subtypes revealed an enrichment within cluster 5 neurons (Figure 3C), suggesting that it is a selective marker of classically defined DA neurons. Similarly, differential expression testing identified Slc26a7, a gene encoding an anion transporter, as a marker for cluster 9 neurons (Figure 3D). Slc26a7 is required for chloride homeostasis in the stomach and kidney, and has been implicated in chloride transport in neurons in the olivocerebellar system (Rahmati et al., 2018). Comparison of DEG lists identified that Slc26a7 is a cluster 9-specific DEG (Figure 3E). Furthermore, the high Gini coefficient for Slc26a7 (0.79), combined with the high expression of this gene almost exclusively in cluster 9 neurons (Figure 3F), indicated that this gene is a selective marker for this combinatorial neuron population.

Figure 3. Identification of marker genes for classical dopamine and combinatorial neurons.

(A) Volcano plot of all cluster 5 DEGs. X axis is log₂ fold change between cluster 5 neurons and all other neurons. Top 15 genes by adjusted p value are labeled.

(B) Scatterplot of Gini coefficient and log-transformed gene expression values for cluster 5 neurons. Colored points are cluster 5-specific DEGs, dark-gray points are shared DEGs between clusters 5 and 9, and light-gray points are non-DEGs.

(C) UMAP depicting location of cluster 5 neurons and distribution of expression for Gch1.

(D) Volcano plot of all cluster 9 DEGs. X axis is log₂ fold change between cluster 9 neurons and all other neurons.

(E) Scatterplot of Gini coefficient and log-transformed gene expression values for cluster 9 neurons. Colored points are cluster 9-specific DEGs, dark-gray points are shared DEGs between clusters 5 and 9, and light-gray points are non-DEGs.

(F) UMAP depicting location of cluster 9 and distribution of expression for Slc26a7.

Expression of DA synthesis, transport, and degradation genes

Although genes such as Th, Slc6a3 (DAT), and Slc18a2 (VMAT2) have all been used as markers of DA neurons, prior studies have not had the ability to systematically explore expression of all genes that regulate DA levels and localization in specific cell types. Therefore, we next focused on a set of genes known to play important roles in DA synthesis, transport, and degradation (Figure 4). In the VTA, DA synthesis likely begins with tyrosine, a non-essential amino acid that crosses the blood-brain barrier. In the rate-limiting step in DA synthesis, tyrosine is converted to L-DOPA by TH, using tetrahydrobiopterin (BH4) as a co-factor (Daubner et al., 2011; Molinoff and Axelrod, 1971). Notably, the GTP cyclohydrolase Gch1, identified in Figure 3 as a selective marker of cluster 5 DA neurons, performs the rate-limiting step in BH4 synthesis by converting GTP to dihydroneopterin triphosphate (Kapatos, 2013; Kaufman, 1959) (Figure 4B). This intermediary metabolite is further converted in a variety of de novo and salvage pathways to BH4, involving reductases such as sepiapterin reductase (Spr), aldo-keto reductase family 1, member B1 (Akr1b1), and dihydrofolate reductase (Dhfr) (Kapatos, 2013; Yang et al., 2006). Surprisingly, in addition to Gch1, we observed significant enrichment of several other genes in the BH4 synthesis pathway in cluster 5 DA neurons, but not in cluster 9 “combinatorial” neurons that expressed other commonly used markers of DA synthesis and transport (e.g., Th, Ddc, Slc6a3, and Slc18a2). Taken together, these results suggest that combinatorial or “co-release” DA neurons may have a paradoxically limited capacity for DA synthesis, despite the enrichment of Th and Ddc in this neuron cluster. Consistent with elevated DA synthesis capacity in cluster 5 DA neurons, we also observed enrichment of two genes involved in DA degradation (monoamine oxidase A, encoded by Maoa, and the previously identified DA neuron marker Aldh1a1, which codes for an aldehyde dehydrogenase) within this cluster (Figures 4A and 4B).

Figure 4. Expression of DA synthesis, transport, and degradation machinery in VTA neurons.

(A) Heatmaps of mean expression values for genes involved in BH4 synthesis and DA synthesis, transport, and degradation.

(B) Pathway for genes involved in BH4 and DA synthesis.

Validation of DA and combinatorial population marker genes

While snRNA-seq provides unprecedented resolution of thousands of genes within single cells, more targeted techniques such as fluorescence in situ hybridization (FISH) allows for visual confirmation of computational findings as well as assessment of the spatial distribution of specific transcripts. To confirm Gch1 and Slc26a7 as selective markers for classically defined DA and combinatorial neurons, respectively, we used a multiplexed RNAscope approach with hybridization probes specific for Th, Gch1, and Slc26a7 mRNA. As expected, Th mRNA signal was abundant in the VTA as well as the neighboring substantia nigra (Figure 5A). Further quantification of each probe within the VTA revealed distinct cell populations, including Th⁺/Gch1⁺ and Th⁺/Slc26a7⁺ cells predicted by our snRNA-seq dataset. Furthermore, a significant number of Th⁺ cells detected with RNAscope lacked co-expression of Gch1 or Slc26a7, which is also consistent with snRNA-seq results (Figures 1 and S6). Overall, Th⁺ only or Th⁺/Gch1⁺ cells were much more abundant than Th⁺ cells that also expressed Slc26a7 (Figures 5D and 5F). Interestingly, as observed with “combinatorial” cluster 9 neurons co-expressing Th and Slc26a7 detected with snRNA-seq (Figure 2B), we observed that Th⁺/Slc26a7⁺ cells identified with RNAscope expressed less Th than Th⁺ only and Th⁺/Gch1⁺ cells (Figure 5E).

Figure 5. Validation of DA and combinatorial marker genes.

(A) Representative image of Th in the VTA. Scale bar, 1 mm.

(B) Identification of DA and combinatorial cells using probes targeting Th, Gch1, and Slc26a7. Arrows mark cells expressing Th only, Th⁺ Gch1, and Th⁺ Slc26a7. Scale bar, 50 μm.

(C) 10× images of Th, Gch1, and Slc26a7, and anatomical coordinates. IF, interfascicular nucleus; IPR, interpeduncular nucleus, rostral subdivision; PBP, parabrachial pigmented nucleus of the VTA; PIF, parainterfascicular nucleus of the VTA; PN, paranigral nucleus of the VTA; RMC, red nucleus, magnocellular part; vtgx, ventral tegmental decussation; RLi, rostral linear nucleus of the raphe; ml, medial lemniscus; tth, trigeminothalamic tract. Scale bar, 500 μm.

(D) Analysis of location of cells expressing selected marker genes. Scale bar, 500 μm.

(E) Quantification and comparison of marker intensity in Th⁺ only, Th⁺/Gch1⁺, and Th⁺/Slc26a7⁺ populations (Kruskal-Wallis test with Dunn’s multiple comparisons test: **p < 0.01, ***p < 0.001, ****p < 0.0001).

(F) Quantification of identified cell types in the VTA (n = 4 sections).

VTA cell types also display distinct spatial distribution patterns (Morales and Margolis, 2017), a characteristic that can only be interrogated with in situ approaches. Using the position of nucleus-specific regions of interest (ROI) identified from the DAPI channel, we plotted the location of Th⁺ only, Th⁺/Gch1⁺, and Th⁺/Slc26a7⁺ cells within the VTA. Whereas Th⁺ and Th⁺/Gch1⁺ cells were distributed throughout the VTA, Th⁺/Slc26a7⁺ cells were located exclusively near the midline of the VTA (Figure 5D). These findings are consistent with previous studies demonstrating that most DA + glutamate “combinatorial” neurons are also located near the midline of the VTA (Morales and Margolis, 2017).

Opioid neuropeptide marker analysis within neuronal subclusters of the VTA

Endogenous opioid neuropeptides and their receptors participate in a host of physiological functions, including pain modulation, neuroprotection, euphoria, dysphoria, and maintenance of ionic homeostasis. However, our understanding of the cell-selective nature of opioid effects in the VTA remains poorly understood. To examine cell-type-specific expression of opioid neuropeptide markers in the present dataset, we plotted the distribution of expression for several endogenous opioid peptide precursors (Penk, Pomc, Pdyn, Pnoc, and Tac1) as well as the genes encoding their respective receptors (Oprd1, Oprm1, Oprk1, Oprl1, Tacr1, and Tacr3; Figure 6A). The pro-enkephalin gene Penk, a precursor for met- and leu-enkephalins implicated in analgesia, euphoria, and stress resilience, was highly enriched in glutamatergic neuronal subcluster 3. In contrast, high-affinity opioid receptors for met- and leu-enkephalins, Oprd1 and Oprm1, exhibited limited expression in this glutamatergic subcluster and were instead preferentially expressed in GABAergic cells in cluster 4 (Figures 5A and 5B). This pattern of μ-opioid receptor expression is consistent with the excitatory actions of opioids in the VTA being mediated via GABAergic disinhibition of DA and glutamatergic neurons (Johnson and North, 1992; Zell et al., 2020). Pomc, the precursor for several pro-opiomelanocortin derivatives including endorphins, melanotropins, and adrenocorticotropins, was absent from the vast majority of neurons. The nociceptin precursor Pnoc was also sparsely expressed, with its highest enrichment in classically defined GABA and Glu/GABA combinatorial neurons (clusters 1 and 10, respectively; Figures 6A and 6B). Similarly, the nociceptin receptor Oprl1 was expressed in all neuronal subclusters, but to a limited extent. Tachykinin precursor gene Tac1, which is required for the synthesis of substance P and neurokinins, was preferentially expressed in combinatorial subclusters 0 and 2. Substance P receptor Tacr1 was present in all clusters but expressed at modest levels compared with neurokinin B receptor Tacr3, which was robustly expressed in classical DA neurons (cluster5), combinatorial neurons (cluster 10), and GABAergic cells (clusters 4 and 6).

Figure 6. Endogenous opioid neuropeptides and receptors are expressed in discrete neuronal subpopulations in the VTA.

(A) Violin plots of expression distribution of opioid precursor genes and receptors.

(B) Feature plots of opioid neuropeptide marker enrichment across neuronal subclusters.

Cell-type-specific gene set enrichment for brain disorders and addiction-related phenotypes

GWASs have identified risk genes that confer susceptibility to brain disorders and related phenotypes. Single-cell sequencing enables further contextualization of these variants by identifying cell types that harbor enriched expression of risk genes. We noticed that expression of several genes implicated by GWAS to contribute to schizophrenia risk was unequally distributed across neuronal and non-neuronal cell types. For example, Rims1, a schizophrenia risk gene that encodes a protein responsible for the control of synaptic vesicle endocytosis, demonstrated pan-neuronal enrichment (Figure 7A). Therefore, we next sought to identify cell-type-specific enrichment for brain disorders and diseases with graphical analysis of risk gene enrichment in VTA cell types, as well as MAGMA (Multi-marker Analysis of GenoMic Annotation) gene set analysis (de Leeuw et al., 2015). Specifically, we identified risk genes and mined summary statistics from GWASs and associated meta-analyses for Alzheimer’s disease (Jansen et al., 2019), Parkinson’s disease (Nalls et al., 2019), bipolar disorder (Stahl et al., 2019), schizophrenia (Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014), attention deficit hyperactivity disorder (ADHD) (Demontis et al., 2019), autism spectrum disorders (Grove et al., 2019), and specific traits associated with alcohol and tobacco use (Liu et al., 2019).

Figure 7. snRNA-seq from VTA reveals cell-type-specific enrichment for genes implicated in human neuropsychiatric and brain disorders by GWAS.

(A) Expression plots for rat orthologs of selected human GWAS hits.

(B) Expression profiles of all rat homologs for human GWAS sets from Alzheimer’s disease, schizophrenia, ADHD, and smoking initiation. Scaled expression data were downsampled to 75 cells per cell type (columns), and rows represent individual genes.

(C) Heatmap of MAGMA analysis shows cell-type-specific gene set enrichment for GWAS phenotypes. p values are represented in color scale; significant false discovery rate-corrected observations are marked by asterisks (adjusted *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Notably, MAGMA gene set analysis identified a pan-neuronal association with SNPs identified in schizophrenia GWAS (Figure 7B). This enrichment of schizophrenia risk genes is expected given that the dysregulation of dopaminergic, glutamatergic, and GABAergic neurotransmission is heavily implicated in the development and progression of schizophrenia (Brisch et al., 2014; de Jonge et al., 2017; Moghaddam and Javitt, 2012). This analysis also identified a pan-neuronal enrichment for “smoking initiation,” a binary phenotype in which study participants were surveyed for whether they had ever been a regular smoker (Figures 7B and 7C). Specifically, Scn2a, a gene which encodes a voltage-gated sodium channel, was highly expressed in all VTA neurons. However, some genes such as Cadps2, a gene encoding a calcium-binding protein responsible for the regulation of synaptic vesicle exocytosis, was most highly expressed in Glut-Neuron-1, DA neuron, and interneuron populations. The pan-neuronal enrichment of gene sets with SNPs identified in schizophrenia and smoking GWAS are interesting, as upward of 60% of schizophrenia patients are also smokers (Figure 7C and Table S5). These findings suggest a shared genetic susceptibility within neuronal cell types of the VTA.

Consistent with the reported dysregulation of glutamatergic neurotransmission and signaling in ADHD (Courvoisie et al., 2004; Miller et al., 2014; Naaijen et al., 2017), we also found that ADHD risk genes were enriched in the Glut-Neuron-1 and Glut-Neuron-3 populations (Figure 7C). Although dopaminergic signaling is also implicated in ADHD (Madras et al., 2005; Wu et al., 2012) and the DA neuron gene set was significantly enriched before multiple hypothesis testing correction, many of the combinatorial neurons that express genes involved in the synthesis and transport of DA originate from the Glut-Neuron-1 parent cluster. Taken together, these results demonstrate that specific populations of neurons within the VTA display an inherent genetic susceptibility to brain disorders and diseases that is most likely linked to their involvement in the synthesis and transport of specific neurotransmitters. These results also highlight the utility in using single-nucleus transcriptomic technology to investigate the cell-type-specific expression of disease-linked genes.

DISCUSSION

Although it is well established that the VTA is composed of heterogeneous cell types (La Manno et al., 2016; Li et al., 2013; Morales and Margolis, 2017; Parker et al., 2019; Root et al., 2016; Saunders et al., 2018a; Tiklová et al., 2019; Viereckel et al., 2016; Yamaguchi et al., 2007, 2015), the majority of studies focused on this topic have relied on more targeted techniques, such as immunohistochemistry, in situ hybridization, and single-cell qRT-PCR (Kawano et al., 2006; Li et al., 2013; Morales and Margolis, 2017; Paul et al., 2019; Yamaguchi et al., 2007, 2015). More recently, single-cell sequencing approaches have also been employed to investigate molecular heterogeneity within the VTA (Hook et al., 2018; La Manno et al., 2016; Poulin et al., 2020; Saunders et al., 2018a; Tiklová et al., 2019). The present dataset expands on these studies in several key ways. For example, previous single-cell sequencing studies were conducted exclusively in the mouse brain (Hook et al., 2018; La Manno et al., 2016; Poulin et al., 2020; Saunders et al., 2018a; Tiklová et al., 2019) and have relied primarily on sequencing a subset of fluorescence-activated cell sorting (FACS)-isolated midbrain dopaminergic populations, rather than sampling all VTA cell types (Hook et al., 2018; La Manno et al., 2016; Poulin et al., 2020; Tiklová et al., 2019). Here we used an unbiased approach to conduct transcriptomic profiling of all cell types in the VTA, identifying 16 transcriptionally distinct cell populations and multiple neuronal subtypes. Notably, this sequencing dataset focuses exclusively on VTA subregions, unlike other studies that have focused on pooled cells from mouse substantia nigra and VTA (Saunders et al., 2018a) or a subset of fluorescently tagged FACS-isolated cells from general midbrain regions (Hook et al., 2018; La Manno et al., 2016; Poulin et al., 2020; Tiklová et al., 2019). Finally, with over 21,000 nuclei from both male and female rats, our dataset also constitutes the largest and most comprehensive single-cell transcriptomic analysis focused exclusively on the composition of the VTA.

Single-cell sequencing enables robust DEG testing to identify markers for viral targeting of distinct VTA DA neuron subclasses. Of the 21,600 nuclei captured, 399 exhibited gene markers for classically defined DA neurons. This proportion of DA neurons is in agreement with a previous scRNA-seq study that captured 991 DA neurons from a combined substantia nigra and VTA dataset of 44,091 cells (Saunders et al., 2018a). Here, we focused on identifying marker genes for classically defined DA and combinatorial neurons. Identification of additional marker genes is important because genes widely used for the targeting of single neurotransmitter cell types, such as Th and Slc17a6, are also expressed in combinatorial neurons. Differential expression analysis identified Gch1 as a selective marker of classically defined DA neurons. Gch1 encodes GTP cyclohydrolase 1, the rate-limiting enzyme in tetrahydrobiopterin synthesis, which acts as a co-factor for TH and is essential for DA production (Ichinose et al., 1994; Müller et al., 2002; Yoshino et al., 2018). Thus, cells lacking Gch1 may not have the requisite machinery for de novo DA synthesis. RNA FISH experiments also demonstrated that a significant number of Th⁺ cells do not express Gch1 and may not represent bona fide DA neurons, suggesting that exclusive use of the Th promoter to drive transgenes may result in targeting of DA and non-DA neuronal populations. While a very small subset of Th⁺/Slc26a7⁺ cells express some Gch1, this marker gene may provide selectivity for true, classically defined DA neurons. Similarly, differential expression analysis also identified Slc26a7, a gene encoding an anion transporter, as a selective marker for combinatorial neurons. RNA FISH experiments confirmed the presence of this Th⁺/Slc26a7⁺ cell population at the VTA midline, consistent with the interpretation that these represent combinatorial neurons. Notably, both snRNA-seq and RNA FISH experiments demonstrated that these cells express less Th than DA neurons. While Gch1 and Slc26a7 represent markers of DA and combinatorial neurons, respectively, additional work is required to identify the distinct functional and release properties of Th⁺/Gch1⁺ and Th⁺/Slc26a7⁺ neurons. Together, these findings highlight the utility of leveraging snRNA-seq for molecular characterization and identification of selective target genes.

In addition to classical DA, GABA, and glutamate neurotransmitter systems, the VTA is also a key site for opioid neuropeptides and their receptors. Research in this area has further intensified in response to the current opioid crisis. Here, we found that μ-opioid receptors were most abundant in GABAergic populations, which is consistent with studies demonstrating that opioid-induced increases in DA and glutamatergic activity in the VTA are mediated by the disinhibition of GABAergic neurons (Johnson and North, 1992; Matsui and Williams, 2011). In addition to GABA populations, Oprm1 is also moderately expressed in classically defined glutamate (Figure 5, subcluster 3) and Glu/DA combinatorial neuron populations (Figure 5, subcluster 2). This aligns with recent reports that μ-opioid receptors may also regulate VTA neurotransmission by directly engaging glutamatergic neurons (Zell et al., 2020). The presence of these receptors in Glu/DA combinatorial neurons presents yet another means of fine-tuning VTA neural activity and behavioral responses to opioids. Notably, expression of several opioid receptors (e.g., Oprd1, Oprm1, and Oprk1) was low or absent in classically defined DA neurons, suggesting that direct opioid modulation of DA neurons may be limited. Similarly, the pre-pronociceptin precursor Pnoc and its receptor Oprl1 were also minimally expressed in DA neurons. Instead, subclustering revealed that Pnoc is most enriched in GABAergic subclusters (Figure 5, clusters 1, 4, and 6) and Glu/GABA combinatorial neurons (Figure 5, cluster 10). This aligns with findings suggesting that the VTA may contain two Pnoc⁺ subpopulations: one glutamatergic population anatomically centered in the ventromedial IPN and one GABAergic population in the lateral parabrachial pigmented nucleus (Parker et al., 2019). Tachykinin precursors, including substance P, neurokinins, and their respective receptors, mediate inflammatory response, vasodilation, neuronal function, and the perception of pain (De Felipe et al., 1998; Xia et al., 2010; Zubrzycka and Janecka, 2000). More recently, these neuropeptides have also been implicated in morphine-induced conditioned place preference and self-administration behaviors (Murtra et al., 2000; Ripley et al., 2002), suggesting a role for tachykinins in opioid reinforcement. However, the mechanism through which these actions are regulated in the VTA remains unclear. For example, substance P has excitatory effects on both DA and GABAergic neurons in the VTA (Korotkova et al., 2006), but whether DA neuron activity is modulated directly or indirectly is not well established (Guo et al., 2009; Schank, 2020). Subclustering in the present dataset suggests that these effects may be mediated through GABAergic or glutamate/DA combinatorial populations, rather than via direct modulation of DA neurons, where Tac1 and Tacr1 are largely absent. Overall, these findings confirm and extend previous research highlighting the complexity of neuropeptide modulation in multiple neuron populations within the VTA.

In conclusion, this dataset provides comprehensive molecular characterization of cell types in the VTA. The use of snRNA-seq allowed us to investigate the heterogeneity of previously identified cell types, leading to the identification of a rare subset of combinatorial neurons. Furthermore, the ability to assay thousands of genes enabled identification of selective marker genes, which may be used for more accurate viral targeting for functional studies. Finally, the delineation of VTA neuron subtypes and cellular heterogeneity serves as a resource for future work involving characterization and manipulation of neuronal subpopulations that contribute to neuropsychiatric and neurodevelopmental disorders.

Limitations of the study

Although snRNA-seq provides unprecedented resolution in detecting thousands of transcripts in diverse cell types in the VTA, it does not provide single-cell-level resolution regarding protein levels. Detection of protein levels within specific cell types will be critical for determining a cell’s ability to synthesize and transport one, or many, neurotransmitters. Furthermore, a cell’s ability to engage in multiple neurotransmitter systems is far more complicated than ensuring that the cell expresses the genes necessary for synthesis and release. For example, previous research has demonstrated that VMAT2, the protein responsible for vesicular release of DA, is also involved in non-canonical release of GABA (Tritsch et al., 2012). However, limitations in translating gene expression data into functional consequences at the cellular level are not specific to this study but rather are inherent to all snRNA-seq research. Future studies will be required to not only examine protein levels of gene markers identified for distinct cell subclasses, but also to confirm that Slc26a7⁺ neurons have the capacity to synthesize and transport multiple neurotransmitters and that Gch1⁺ neurons are selectively involved in DA transmission. Ex vivo and in vivo experiments harnessing these marker genes for viral targeting of distinct subpopulations will provide essential insights into the putative functional differences of these populations at the physiological and behavioral levels.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

All relevant data that support the findings of this study are available by request from the lead contact, Jeremy J. Day (jjday@uab.edu).

Materials and availability

This study did not generate new unique reagents or materials.

Data and code availability

• Sequencing data that supports the findings of this study are available in Gene Expression Omnibus. Accession number for the study is listed in the key resources table.

• All original code has been deposited at Zenodo and is publicly available as of the date of publication. DOIs are listed in the key resources table.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
snRNA-seq data	This paper	GEO : GSE168156
Software and algorithms
Cell Ranger v3.0.2	10x Genomics	https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/what-is-cell-ranger
Seurat v3.2.2	Stuart et al. (2019)	https://satijalab.org/seurat/
MAGMA	de Leeuw et al. (2015)	https://ctg.cncr.nl/software/magma
ImageJ2	Schindelin et al. (2015) & Rueden et al., 2017	https://imagej.net/software/imagej2/
StarDist	Schmidt et al. (2018)	https://imagej.net/plugins/stardist
Other
Rscripts	This paper	https://gitlab.rc.uab.edu/day-lab/rat-vta-snrna-seq, https://doi.org/10.5281/zenodo.6205342
Th RNAscope probe	ACD Bio	Catalog no. 314651-C3
Gch1 RNAscope probe	ACD Bio	Catalog no. 1094971-C4
Slc26a7 RNAscope probe	ABD Bio	Catalog no. 1094961-C1
Chromium NextGem Single Cell Chip	10x Genomics	catalog no. 10000121
Chromium Next GEM single cell 3′ library and gel bead kit, v3.1 single index	10x Genomics	catalog no. 10000121

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Approximately 8–12 week old male and female Sprague-Dawley rats (200–250g) were purchased from Charles River Laboratories and individually housed for 1 week prior to tissue collection. Experimentally naïve rats were maintained on a 12 h light/dark cycle with ad libitum access to food and water and were briefly handled for 3 days prior to tissue collection (1–2min/day) to acclimate them to the experimenters. All experimental procedures were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee and were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

METHOD DETAILS

Tissue collection from adult VTA

Rats were euthanized by live decapitation, and brains were rapidly removed and chilled in ice cold Hibernate A media without Ca²⁺ or Mg²⁺ (BrainBits, HACAMG500) for 30–60 sec. Chilled brains were then blocked into 1-mm-thick coronal sections on wet ice. Sections containing the VTA (ranging from −4.92–6.46 anterior-posterior (AP) to Bregma; Figure S1) were transferred to glass petri dishes on dry ice, and the VTA was microdissected from neighboring brain regions (n = 6 rats per sex). Dissected VTA tissue was then transferred to ice-cold centrifuge tubes, and stored at −80°C until the day of snRNA barcoding and cDNA library preparation.

Single-nuclei dissociation

Frozen VTA tissue was briefly thawed on wet ice before being chopped by scalpel 100 times in two orthogonal directions. Tissue from three rats per sex was combined and transferred to 5 ml of ice cold lysis buffer (10 mM tris-HCl, 10 mM NaCl, 3 mM MgCl2 and 0.1% Igepal in nuclease-free water (Sigma-Aldrich, 18896-50ML) for 15 min, mixing the tissue by inversion every 2 min. The lysis was quenched after 15 min with 5 ml of complete Hibernate A (Thermo Fisher Scientific, A1247501) supplemented with B27, GlutaMAX (Life Technologies, 35050-061), and NxGen RNase Inhibitor (0.2 U/μ; Lucigen, 30281-2). Tissue was then triturated by fire-polished Pasteur pipette (three pipettes of decreasing diameter; 8 to 10 passes per pipette) for 35 min and passed through a 40-μm prewet filter. The samples were then pelleted at 500 rcf for 10 min at 4°C followed by a wash in 10 ml of nuclei wash and resuspension buffer [13 phosphate-buffered saline (PBS), 1% bovine serum albumin (BSA), and NxGen RNase Inhibitor. Supernatant was then removed, and the pellet was gently resuspended in 800 μl of wash buffer before 7-aminoactinomycin D (Thermo Fisher Scientific, 00-6993-50) staining and fluorescence-activated cell sorting (FACS) to further purify the nuclei for sequencing (BD FACS Aria, 70-μm nozzle, BD Biosciences). Immediately after FACS, nuclei were washed a final time at 250 rcf in 10 ml of supplemented Hibernate A containing 1% BSA and RNase Inhibitor for 10 min at 4°C to remove any remaining fine debris and nascent RNA. Nuclei were then brought to a concentration of 1400 nuclei/μL. A total of 11,367 nuclei pooled from three rats per sex and treatment group were loaded into individual wells of the Chromium NextGem Single Cell Chip (10x Genomics, catalog no. 10000121) using four of the eight available wells.

snRNA-seq library prep and sequencing

Libraries were constructed according to manufacturer’s instructions for Chromium Next GEM single cell 3′ library and gel bead kit (10x Genomics, v3.1 single index, catalog no. 10000121), which utilizes version 3 chemistry. A total of 21,606 nuclei were captured across 4 GEM (gel bead in emulsion) wells (2 wells per genetic sex) using the 10x Chromium Controller. Each GEM well contained nuclei from 3 separate male or female animals for a total of 6 rats per genetic sex. Libraries were then sequenced on the Illumina NextSeq500 at the Heflin Genomics Core at UAB to an average read depth of ~26,197 reads per nuclei.

Tissue preparation and imaging for RNAscope

Fresh tissue from was frozen in 2-methylbutane at −80°C until the day of sectioning. After 30 min equilibration at −20°C, 10 μm coronal sections were obtained using a Leica cryostat. Sections containing the VTA (ranging from −4.92–6.46mm anterior-posterior (AP) to Bregma) were mounted onto room temperature slides, air dried for 60 min at −20°C, and stored at −80°C. The RNAscope Multiplex Fluorescent v2 assay kit (Advanced Cell Diagnostics, 323110) was used to stain sections, following the manufacturer’s recommended protocol. Probe sets used include DAPI (320858), Rn-Th-C3 mRNA (314651-C3), Rn-Gch1-C4 mRNA (1094971-C4), and Rn-Slc26a7-C1 mRNA (3 Channels were matched with genes according to expression level and channel background: the green channel (520 nm dye, recommended for high expressors) was assigned to Th, the red channel (570 nm dye, recommended for low expressors) was assigned to Slc26a7, and the white channel (690 nm dye, recommended for low expressors) was assigned to Gch1. After staining, sections (4 total sections, 2 each from one male and one female rat) were imaged using the Keyence BZ-X700 microscope. Images were taken using OP-87767 filter cubes for GFP (Channel 1), Texas Red (PE) (Channel 2), DAPI (Channel 3), and Cy5 (Channel 4) at 4x (PlanFluor NA 0.13 PhL) and 20x (PlanFluor NA 0.45 Ph1) magnification.

QUANTIFICATION AND STATISTICAL ANALYSIS

snRNA-seq analysis

A CellRanger reference package was generated using a custom gene transfer file (GTF) from the Ensembl Rn6 genome (version 95) to ensure accurate mapping to both pre-mature, unspliced RNAs as well as mature mRNA. CellRanger filtered outputs were then analyzed with Seurat v3.2.2(Butler et al., 2018; Stuart et al., 2019) using R v4.0.2. Nuclei with <200 genes and >5% of reads mapping to the mitochondrial genome were removed from the dataset. Molecular count data from each GEM (gel bead in emulsion) well were then log-normalized with a scaling factor of 10,000. To ensure that the identified clusters were representative of the cell type heterogeneity known to exist in the adult rat VTA, uniform approximation and projection (UMAPs) were generated following data integration for every combination of 21 principal components (10 to 30) and 10 resolution values (0.1 – 1.0). Data from each GEM well were then integrated with FindIntegrationAnchors() and IntegrateData() using 25 principal components and a resolution value of 0.2 as these values produced clusters representative of known cell types in the adult rat VTA. Local Inverse Simpson’s Index was used to ensure successful integration of 4 GEM wells. LISI was calculated in R using lisi v1.0 with default parameters. The cell type identity of each cluster was validated with a two-step approach. First, we utilized a Wilcoxon ranked sum in which the log-normalized gene expression values for one neuronal subtype are tested against log-normalized gene expression values for all other cell types. Resulting p-values were adjusted using the Bonferroni correction based on the number of genes identified in each cluster. DEGs were then compared to previously identified marker genes to identify populations. Secondly, we investigated the selectivity of expression of DEGs, as well as known marker genes, by plotting the distribution of expression values by cluster with FeaturePlot(), VlnPlot(), Dot-Plot(). Investigating the selectivity of DEGs allowed for the identification of genes that marked several populations, such as Slc17a6 and Mobp, as well as those genes that marked specialized populations of neurons. Neuronal subtypes were identified by removing all non-neuronal cells and reclustering with 15 principal components and a resolution value of 0.2. Low confidence clusters with <50 cells were not included in further analyses as they did not represent any specific, transcriptionally defined cell type. To investigate cells that may have the ability to synthesize and transport multiple neurotransmitters, we first identified those cells that co-express two genes involved in separate neurotransmitters. To do this we first extracted a count matrix, containing log-normalized gene counts for Th, Ddc, Slc6a3, Slc18a2, Slc17a6, Grm2, Slc17a7, Gad1, Gad2, Slc32a1, and Slc6a1, for every neuronal subcluster. Then, the number of cells with log-normalized count values > 0 for two of the above genes was divided by the total number of cells within the cluster. The resulting value represented the proportion of cells within a given neuronal subtype that co-express two genes involved in two different neurotransmitter systems. Marker genes for 7 neuronal subtypes were identified were calculated as described above. Gini coefficients were calculated using the average log-normalized gene expression values for each cluster with the gini() command from the reldist R package (Handcock and Morris, 1999; Handcock, 2016). To investigate GABAergic cell type heterogeneity within the VTA, GABAergic neurons were subclustered using 11 principal components and a resolution value of 0.1. To compare DA neurons to previously identified DA neuron subtypes, the DA neuron cluster was subclustered using 11 principal components and a resolution value of 0.12. R code for this analysis is available at https://gitlab.rc.uab.edu/day-lab.

MAGMA analysis

Cell type specific gene lists were generated by first identifying marker genes for the 16 transcriptionally distinct cell types expressed in > 10% of cells within the cluster that also had a Bonferroni adjusted p-value < 1×10–12 and a log2 fold change value greater than 0.25. Human homologs for these rat genes were then identified by mapping the “HomoloGene.ID” value for each marker gene from rat to human (Tran et al., 2021). These “HomoloGene.ID” values can be found at http://www.informatics.jax.org/downloads/reports/HOM_AllOrganism.rpt. Following the generation of these gene sets, Multi-marker Analysis of GenoMic Annotation (MAGMA) (de Leeuw et al., 2015) v1.08b was used to identify cell type specific enrichment, and susceptibility, for Alzheimer’s disease(Jansen et al., 2019), Parkinson’s disease(Nalls et al., 2019), bipolar disorder(Stahl et al., 2019), schizophrenia(Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014), attention deficit hyperactivity disorder(Demontis et al., 2019), autism spectrum disorders (ASDs) (Grove et al., 2019), and specific traits associated with alcohol and tobacco use (Liu et al., 2019). SNPs obtained from summary statistics for each GWAS were mapped to hg19 gene coordinates with a +/−10kb window. Following annotation, these same summary statistics files for each GWAS were then used to perform gene-level analyses with the default snp-wise=mean gene analysis model. Finally, the default competitive gene-set analysis was performed for all gene sets corresponding to the 16 transcriptionally distinct cell types. Resulting empirical p-values were then adjusted with false discovery rate (FDR) for the 16 tests performed for each GWAS.

Image analysis

Original 20X images were stitched using the BZ-X800 Analyzer software. 8-bit single probe images were then generated and pseudocolored using ImageJ2 v2.3.0/1.53f (Rueden et al., 2017; Schindelin et al., 2015). To identify cell types with RNAscope, VTA specific ROIs were generated and applied to the DAPI channel for each sample. Nuclei specific ROIs were generated using the StarDist (Schmidt et al., 2018) plugin using the “Versatile (fluorescent nuclei)” model with a probability/score threshold of 0.60 and an overlap threshold of 0.15. Nuclei specific ROIs were then applied to images containing Th, Gch1, or Slc26a7 probe signal and probe intensity was measured. Nuclei specific ROI positions were used to identify locations of Th+ only, Th+/Gch1+, and Th+/Slc26a7+ cells. Normalized intensity scores were generated by rescaling raw max intensity scores within each ROI to range between 0 and 100.

Highlights.

• Transcriptional atlas of rat ventral tegmental area (VTA) generated using snRNA-seq

• Computational prediction and validation of novel dopamine neuron marker genes

• Population-specific enrichment of gene sets linked to psychiatric disorders

• Searchable and interactive database to query genes of interest in VTA