If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Single-nucleus transcriptional profiling of GAD2-positive neurons from mouse lateral habenula reveals distinct expression of neurotransmission- and depression-related genes
Corresponding author: Anne E. West, Duke University, Department of Neurobiology, Box 3209, 311 Research Drive, Bryan Research 301D, Durham, NC 27710 USA, , 919-681-1909
Glutamatergic projection neurons of the lateral habenula (LHb) drive behavioral state modulation by regulating the activity of midbrain monoaminergic neurons. Identifying circuit mechanisms that modulate LHb output is of interest for understanding control of motivated behaviors.
Methods
A small population of neurons within the medial subnucleus of the mouse LHb express the GABAergic synthesizing enzyme GAD2, and they can inhibit nearby LHb projection neurons; however, these neurons lack markers of classic inhibitory interneurons and they co-express the vesicular glutamate transporter VGLUT2. To determine the molecular phenotype of these neurons, we genetically tagged the nuclei of GAD2-positive cells and used fluorescence-activated nuclear sorting to isolate and enrich these nuclei for single nuclear RNA sequencing (FANS-snRNAseq).
Results
Our data confirm that GAD2+/VGLUT2+ neurons intrinsic to the LHb co-express markers of both glutamatergic and GABAergic transmission and that they are transcriptionally distinct from either GABAergic interneurons or habenular glutamatergic neurons. We identify gene expression programs within these cells that show sex-specific differences in expression and that are implicated in major depressive disorder (MDD), which has been linked to LHb hyperactivity. Finally, we identify the Ntng2 gene encoding the cell adhesion protein Netrin-G2 as a marker of LHb GAD2+/VGLUT+ neurons and a gene product that may contribute to their target projections.
Conclusions
These data show the value of using genetic enrichment of rare cell types for transcriptome studies, and they advance understanding of the molecular composition of a functionally important class of GAD2+ neurons in the LHb.
). The LHb receives inputs from hypothalamus, cortex, ventral tegmental area (VTA), dorsal raphe nuclei (DRN), and locus coeruleus (LC), and sends projections to modulate the firing of monoaminergic neurons in VTA, DRN, and LC (
). These observations have driven interest in identifying sources of inhibition that reduce the firing of LHb projection neurons. In addition to GABAergic afferents that project to the LHb (
). A small population of neurons expressing the GABA synthesizing enzyme GAD2 are found in the medial subnucleus of the LHb, though these cells do not express VGAT or GAD1 and they co-express the vesicular glutamate transporter VGLUT2 (
). Using Gad2-Cre mice in combination with Cre-dependent AAV-ChR2 expression in the LHb, optogenetic stimulation of these neurons resulted in inhibitory currents recorded from locally connected neurons within the LHb (
). The molecular mechanisms that allow this GAD2+ cell population to have properties of both local inhibitory interneurons and excitatory projection neurons has remained unknown.
Single cell sequencing has revolutionized neuronal classification by reducing traditional reliance on categorical marker genes and facilitating holistic analysis of gene sets that define cellular function (
). Cellular diversity has recently been characterized within the MHb and LHb, yielding insights into the gene expression programs that characterize downstream projection patterns (
Pandey S, Shekhar K, Regev A, Schier AF (2018): Comprehensive Identification and Spatial Mapping of Habenular Neuronal Types Using Single-Cell RNA-Seq. Curr Biol. 28:1052-1065 e1057.
Cerniauskas I, Winterer J, de Jong JW, Lukacsovich D, Yang H, Khan F, et al. (2019): Chronic Stress Induces Activity, Synaptic, and Transcriptional Remodeling of the Lateral Habenula Associated with Deficits in Motivated Behaviors. Neuron. 104:899-915 e898.
). Although a small number of Gad2 expressing cells were detected in these studies, there were insufficient numbers to drive these neurons into a cluster of their own, and thus the transcriptome of this population remains uncharacterized. To overcome this limitation, we and others have been using transgenes to enrich single cells (
Munoz-Manchado AB, Bengtsson Gonzales C, Zeisel A, Munguba H, Bekkouche B, Skene NG, et al. (2018): Diversity of Interneurons in the Dorsal Striatum Revealed by Single-Cell RNA Sequencing and PatchSeq. Cell Rep. 24:2179-2190 e2177.
Gallegos DA, Minto M, Liu F, Hazlett MF, Aryana Yousefzadeh S, Bartelt LC, et al. (2022): Cell-type specific transcriptional adaptations of nucleus accumbens interneurons to amphetamine. Mol Psychiatry. Online ahead of print. DOI: 10.1038/s41380-022-01466-1.
) of rare cell types by fluorescence activated sorting (FACS/FANS) from heterogeneous brain tissues prior to sequencing. Here, we provide and analyze a comprehensive single nuclear RNA-seq (snRNA-seq) dataset of nuclei from GAD2+ neurons enriched from the LHb. By integrating our Gad2-enriched dataset with previous total scRNA-seq data from the habenula, we describe the expression of key genes that define the transcriptional profiles of LHb GAD2+ neurons compared with other LHb neurons.
METHODS
Animals: We used adult (>P60) male and female mice, and all experiments were conducted in accordance with an animal protocol approved by the Duke University Institutional Animal Care and Use Committee. For nuclear isolation, we crossed homozygous Gad2-IRES-Cre mice (Gad2tm2(cre)Zjh/J, RRID: IMSR_JAX:010802) (
) with homozygous mice expressing a Cre-inducible Sun1-myc-sfGFP transgene, also known as INTACT (B6;129-Gt(ROSA)26Sortm5(CAG-Sun1/sfGFP)Nat/J, RRID: IMSR_JAX: -21039) (
) to generate dual Gad2-Cre/INTACT heterozygotes (HET).
Isolation and sequence of Gad2+ nuclei: LHb was dissected by punch biopsy bilaterally from Gad2-Cre/INTACT HET mice (n=7 male, n=3 female). Nuclei were isolated, pooled by sex, and incubated with MULTI-seq lipid-modified oligos (LMOs) (
) prior to Fluorescent-Activated Nuclear Sorting (FANS), gating on DAPI and GFP. We performed 10X Genomics 3' Gene Expression (v3 chemistry) library construction and sequenced all nuclei on a single 10X GEMwell. The count matrix was generated using CellRanger v3.0.2 and used as input to Seurat V417 (
) for downstream analysis. Differential expression was performed between clusters using a Wilcoxon rank Sums test using log fold change greater than 0.25 and p-values less than 0.05. Data are deposited at GEO datasets at GSE179198. Details of nuclear isolation and sequence analysis, including the integration of our sequencing data with a total LHb scRNA-seq dataset (
We genetically tagged and enriched for the nuclei of LHb Gad2+ neurons prior to performing snRNA-seq on the 10X Genomics platform (Fig. 1A; Fig. S1). We recovered transcriptomes from 3,651 GFP+ cells (Fig. S2) that fell in 11 clusters of cells. Almost all the clusters contained some nuclei with Gad2 mRNA (Fig. S2C) and variation in Gad2 expression between cells in a cluster could reflect RNA dropout. However, when we assessed the expression of known cell-type marker genes we found that many clusters were comprised of non-neuronal cells (Fig. S2B). This may reflect expression of the Gad2-Cre transgene in non-neuronal cells (Fig. S1B). To focus our analysis on neurons, we filtered clusters for the expression of the neuronal marker Rbfox3 (Fig. S2D). This resulted in five clusters of 1491 cells (Fig. 1B) with similar QC features between clusters (Fig S3). The expression of activity-inducible genes was uniform across the neuronal clusters supporting that these neurons clustered by cell type rather than activity state (Fig. S4).
Figure 1Isolation and snRNA-seq analysis of Gad2+ neurons from the LHb. (A) Schematic of the approach designed to enrich for Gad2+ neurons within the LHb and subsequent snRNA-seq. (B) UMAP clustering of the 5 neuronal clusters retrieved from snRNA-seq, filtered by expression of Rbfox3. (C) Heatmap showing differential expression (DE) of top 50 genes per cluster using a non-parametric Wilcoxon Rank Sum test, adjusted p-value ≤ 10e-6 and log2foldchange ≥ |0.25|. Columns represent each cluster with sub columns representing single cells. Each row represents a different gene. (D) Dot plot showing scaled expression of representative top DE genes per cluster. (E) Spatial expression from the Allen Institute Brain Atlas showing Gad2 expression enriched in the medial portion of the LHb. A reference map is shown for orientation. (F) Representative top DE genes per cluster to highlight the respective anatomical origins.
We assessed the most differentially expressed genes (DEGs) between clusters (Fig. 1C; Table S1) and used a subset of these DEGs (Fig. 1D) in combination with the Allen Brain Atlas to verify the original anatomical origins of these neuronal clusters (Fig. 1E). Clusters 2 and 4 are marked by expression of Megf11 and Pvalb respectively, both of which have expression patterns in the LHb that have similarities to Gad2. Pvalb+ neurons are also found in the dorsal thalamus, which immediately borders the LHb on the lateral side. Some of the clusters appear to have originated from regions outside of but in in close apposition to the LHb. Markers of clusters 1 (Ntng1) and 3 (Dab1) are enriched in the dorsal thalamus, whereas cluster 5 expressed high levels of Nwd2, which is highly expressed in the medial habenula (MHb).
Of the five neuronal clusters, clusters 2, 3, and 4 expressed the highest levels of Gad2 (Fig. 2A). Given our evidence (Fig. 1F) that clusters 2 and 4 are the most likely to represent Gad2+ neurons arising from the LHb, we focused on comparing gene expression between these two clusters. Unlike canonical cortical GABAergic interneurons (
Paul A, Crow M, Raudales R, He M, Gillis J, Huang ZJ (2017): Transcriptional Architecture of Synaptic Communication Delineates GABAergic Neuron Identity. Cell. 171:522-539 e520.
), Gad2-expressing cells in the medial LHb are known for their co-expression of Slc17a6 encoding the excitatory glutamate vesicular transporter VGLUT2 (
). Cluster 2 but not cluster 4 expressed Slc17a6 as well as Gad2 (Fig. 2B-C). When we graphed the levels of Gad2 expression against Slc17a6 expression for each single nucleus in each of the five clusters (Fig. S5) we saw that whereas most nuclei in clusters 1 and 5 showed Slc17a6 expression and most nuclei in clusters 3 and 4 showed Gad2 expression, only cluster 2 had a significant number of cells that showed expression of both transporter genes.
Figure 2Gad2+ LHb neurons in cluster 2 express both GABAergic and glutamatergic transmitter features. (A,B) Feature plots showing scaled expression and violin plots showing raw counts of Gad2(A) and Slc17a6(B) expression within the five neuronal Gad2+ clusters. (C) Dot plot showing scaled expression of genes related to synthesis, degradation, release, and reuptake of glutamate and GABA across the five neuronal Gad2+ clusters. (D) Dot plot and (E) violin plots showing scaled expression of GABAergic interneuron-associated markers in the five neuronal Gad2+ clusters.
Cluster 4 but not cluster 2 neurons expressed Gad1, encoding the GABA synthesizing enzyme GAD67 (Fig 2C). Consistent with cluster 4 representing a subtype of GABAergic inhibitory interneurons, we found that this was the only cluster strongly expressing the interneuron markers Pvalb and Sst (Fig. 2D,E) (
). By contrast, we observed little expression of any of the usual interneuron class markers in cluster 2, though these cells did express Calb1, encoding Calbindin and Calb2, encoding Calretinin (Fig. 2D,E). Consistent with prior studies of the Gad2+ neurons in the LHb, we found that neurons in cluster 2 did not express Slc32a1, encoding the vesicular GABA transporter VGAT, whereas this gene was strongly expressed in the Gad2+ neurons of cluster 4. Interestingly however, neurons in cluster 2 were seen to express Slc6a1, the gene encoding the plasma membrane GABA transporter GAT1 (Fig. 2C). This is important because GAT1 function has been identified as a possible mechanism for release of GABA by transporter reversal under membrane depolarized conditions even from neurons that lack VGAT to support vesicular GABA release (
To validate whether Gad2+ cells in the medial subnucleus of the LHb simultaneously co-express both the glutamate transporter Slc17a6 and the plasmid membrane GABA transporter Slc6a1, we used RNAscope fluorescent in situ hybridization (FISH) to quantify the expression and colocalization of these markers on brain sections (Fig. 3). Slc17a6 was expressed in a large percentage of cells across the MHb and LHb, consistent with the evidence that most neurons in this region are glutamatergic (Fig. S6; Fig. 3B, F). Slc6a1 was found in both MHb and LHb, though it was expressed at higher levels and in a greater percentage of cells in the LHb compared with the MHb (Fig. S6; Fig. 3C, F). Within the LHb, Slc17a6 and Slc6a1 were not limited to Gad2+ cells, but either one or both were co-expressed with Gad2 in the majority of Gad2+ cells. (Fig. 3D,E,G). These data confirm that cluster 2 represents the medial LHb Gad2+ cell cluster of functional interest for its coexpression of GABAergic and glutamatergic transmitter genes.
Figure 3Most Gad2+ neurons in the medial subnucleus of the LHb co-express Slc17a6 and/or Slc6a1. (A-C) Representative RNAscope ISH images of Gad2(A), Slc17a6(B), and Slc6a1(C) in the habenula. DAPI, nuclei (blue). Dotted white line, boundary between the MHb (left) and LHb (right). (D) Overlap of Gad2 with Slc17a6 and Slc6a1. Dotted white box is centered on the medial subnucleus of the LHb and expanded in (E). Yellow circles show Gad2+ cells evaluated for overlap. (F) Quantification of the percent of positive nuclei for each individual brain section from thresholding in Fig. S6. 40 cells/image/channel from 3 brains. (G) Percentage of the total of all Gad2+ positive nuclei from images of 3 brains that co-express Slc17a6, Slc6a1, or both. A total of 51 cells were quantified on 3 brains. Scale bar: 10 μm.
Gad2+/Slc17a6+ LHb neurons are distinguished by distinct neurotransmitter gene expression
We identified the strongest DEGs between cluster 2 and the other four Gad2+ neuron clusters (Fig. 4A, Table S2). GO analysis of the DE genes identified glutamatergic synapse genes and other components of chemical synaptic transmission as defining categories (Fig. 4B; Table S3). In cases where the same GO category was significant in both the genes that were significantly enriched and de-enriched in cluster 2, this is because different members of the same gene family were expressed in the different cell clusters. For example, in the molecular function (MF) category glutamate receptor activity (GO:0008066), Gria1, Gria2, and Grik4 were higher in cluster 2 relative to the other clusters, whereas Gria4, Grik2, and Grik3 were lower. Indeed, cluster 2 expressed a unique profile of genes for glutamatergic receptors in all the families, notably including low expression of genes for NMDA-type glutamate receptor subunits as well differential expression of subunits for AMPA-type, metabotropic, and kainate-type glutamate receptors (Fig. 4C). Among the GABA receptor genes, neither clusters 2 nor 4 showed high expression especially compared with cluster 1, but cluster 2 displayed higher expression of Gabrg1 and cluster 4 had highest expression of Gabra3, again suggesting distinctions in the composition of GABA receptors among different classes of Gad2+ neurons (Fig. 4D).
Figure 4Neurotransmitter and neuromodulator receptor profiles distinguish different Gad2+ neuronal populations in the LHb. (A) Volcano plot highlighting top DE genes for cluster 2 compared to the other 4 neuronal clusters in the dataset using a non-parametric Wilcoxon Rank Sum test. Dashed lines indicate cut-offs in the horizontal and vertical direction for log2foldchange ≥ |0.25| and adjusted p-value ≤ 10e-6, respectively. (B) GO analysis summary showing the top 5 GO terms in each category (molecular function, MF, biological process, BP, and cellular compartment, CC) for the up and down-regulated DE genes, as determined from A. (C) Heatmap displaying expression of glutamate receptor subunits and (D) GABA receptor subunits per cluster. Each sub column within a cluster represents a single nucleus. Scaled expression is shown. (E) Dot plot showing expression of the following gene categories within each of the five Gad2+ neuronal clusters: serotonin receptors, acetylcholine receptors, dopamine receptors, peptides and their receptors, and adrenergic receptors.
All Gad2+ clusters expressed genes encoding serotonin and acetylcholine receptors and they weakly expressed noradrenergic receptor subunits; however, we saw minimal expression of dopamine receptors in any of the clusters (Fig. 4E). The most strikingly differential expression between clusters is among the acetylcholine receptors, with Chrm3 highly expressed in cluster 2 versus Chrm2 in cluster 4. Chrm3 was identified in a previous scRNA-seq study of the habenula as a marker of cells in both the lateral oval as well as the medial subdivisions of the LHb (
). Given that Gad2 expression is restricted to the medial subdivision, our findings suggest these are distinct Chrm3+ cell types in the two regions.
Gad2+/Slc17a6+ LHb neurons are transcriptionally distinct from other Slc6a1+ LHb neurons
To compare the programs of gene expression we observed in our Gad2+ clusters against those of Gad2- neurons of the LHb, we integrated our dataset with a previously published scRNA-seq dataset from the MHb and LHb (
). Clustering of the integrated datasets confirmed our identification of microglia, astrocytes, oligodendrocytes, and endothelial cells from our preliminary clusters (Fig. S2, S7). After filtering the integrated dataset for only Rbfox3+ neuron-containing clusters, we obtained 12 clusters. We call these clusters I1-I12 (I for “Integrated”) to distinguish them from the 5 Gad2+ clusters in our data from Fig. 1. This clustering supports our conclusion that cluster 5 of our Gad2+ neurons (Fig. 1B) contained cells from the MHb, as these neurons colocalized in clusters I2, I4, and I7 with MHb cells from the Hashikawa et al. (2020) dataset (Fig. 5A,B; S8). As expected, the perihabenular Gad2+ cells from our clusters 1 and 3 failed to overlap with the habenular clusters. The highly Gad2+ cells in cluster 4 that we suggested were canonical interneurons (Fig. 2) most closely cluster with peri-habenular neurons in I9, suggesting either that they may not arise from the LHb, or that they may be too few in the larger dataset to permit them to cluster as LHb interneurons. The cells we identified as the Gad2/Slc17a6 dual-expressing cluster 2 from Fig. 1B clustered on their own in the integrated dataset as integrated cluster I7 (Fig 5A,B). Cluster I7 was the only cluster that showed significant co-expression of Gad2 and Slc17a6 in the entire integrated dataset (Fig. 5C).
Figure 5Gad2/Slc6a17 dual-positive neurons of the LHb are transcriptomically distinct from other LHb neurons. (A) UMAP clustering of all neurons after integration with Hashikawa et al., 2020 dataset. (B) UMAP clustering of all neuron clusters with neurons from our dataset highlighted according to original cluster number from Figure 1B versus neurons from Hashikawa et al., 2020 data set (grey). (C) Violin plot of expression of Gad2 and Slc17a6 across the 12 neuronal clusters in the integrated dataset. (D) Volcano plot highlighting top DE genes between the LHb clusters. DE was run using a non-parametric Wilcoxon Rank Sum test. Dashed lines indicate cut-offs in the horizontal and vertical direction for log2foldchange ≥ |0.25| and adjusted p-value ≤ 10e-32, respectively. (E) Dot plot showing expression of glutamate receptor subunits, GABA receptor subunits, and peptides between cluster I7 compared with I3, I4, and I8. Scaled expression is shown. (F) Dot plot showing the enriched (UP) and de-enriched (DOWN) genes in cluster I7 compared with clusters I3, I4, and I8 from cross-reference to a meta-analysis of 269 depression-linked genes. Scaled expression is shown.
We ran differential expression analysis between cluster I7 and the three clusters from the Hashikawa et al., 2020 dataset that contain LHb neurons (clusters I3, I4 and I8; Fig. 5A, Fig. S7-S8, Table S2) to determine which genes most strongly distinguish the Gad2/Slc17a6 double-positive neurons from Gad2- neurons in the LHb (Fig. 5D). Prior studies have shown that Gad2+ cells of the medial subnucleus express orexin receptors (Hcrtr2) that mediate the modulatory effects of orexin on aggression (
). We confirmed the preferential expression of the orexin receptor Hcrtr2 in cluster I7 relative to other LHb neurons (Fig. 5E), though we did not find that these neurons were particularly enriched for other peptide receptors (
). We confirmed the enrichment of Chrm3 in cluster I7 relative to the Gad2- LHb neurons and noted a relative paucity of neuromodulator receptors in any of the LHb neurons except the Gad2/Slc17a6 double positive cells in cluster I7 (Fig. S9) .
Finally, because the LHb has been implicated in MDD and antidepressant action (
), we sought to determine whether there were any significant DEGs preferentially expressed in cluster I7 versus other neurons in the LHb that were genetically associated with MDD. A recent meta-analysis (
) yielded a list of 269 genes with significant genetic risk for MDD. We cross-referenced this list against the set of genes that were preferentially expressed in cluster I7 Gad2/Slc17a6+ neurons relative to Gad2- LHb neurons from clusters I3, I4, and I8 (Table S4). We identified 25 MDD-associated genes that were preferentially expressed in cluster I7 relative to other LHb neurons (Fig. 5F). Only 12 of the 269 genes were more strongly enriched in any of the three other LHb clusters compared with I7. Of the I7-enriched genes, 10 were also significantly enriched in cluster 2 relative to other Gad2+ neurons from LHb (Dcc, Tcf4, Megf11, Cacna2d1, Sorcs3, Erbb4, Chd9, Sema6d, Zfp536, and Sdk125) (Fig S10).
Sex-specific differences in LHb Gad2+ neuron gene expression
A prior study has shown that Gad2/Slc17a6 double-positive LHb neurons express estrogen receptors and suggested that gonadal steroid-dependent modulation of these neurons might contribute to sex-specific and behaviorally relevant regulation of LHb activity (
). To determine whether Gad2+ LHb neurons show sex-specific differences in gene expression, we tagged the nuclei harvested from either male or female mice with distinct lipid-modified oligonucleotide barcodes (
). We confirmed that the Y-chromosome genes Eif2s3y, Uty, Kdm5d, and Ddx3y were expressed only in nuclei marked by the barcode we added to nuclei of male mice, whereas Xist and Tsix, which mediate X-chromosome inactivation, were found exclusively in nuclei containing the female barcode (Fig. 6A). After sex classification and filtering for expression of the neuronal marker Rbfox3, we unambiguously identified 816 male and 107 female nuclei.
Figure 6Sex bias in gene expression within Gad2+ populations from the LHb. (A) Expression of known female and male-specific genes within nuclei tagged by LMO barcodes added to nuclei isolated from female or male mice, respectively. (B,C) Differential expression of genes in Gad2+ LHb nuclei from female or male mice. Expression higher in female (B), expression higher in male (C). (D) Table of comparison between sex-biased genes and dbSNP genes in the disease categories shown. The p-values were calculated as described in the results section and were Bonferroni corrected for multiple hypothesis testing. Observed versus expected overlap between sex-biased genes and dbSNP genes for depression (E) and Alzheimer’s (F). Dotted red line, observed, black line, expected.
Excluding the 6 genes we used for sex classification, we found 17 additional genes that were significantly differentially expressed between Gad2+ neurons of male and female mice (Table S5, Fig. S11A). Sixteen are autosomal genes, five of which were expressed significantly more highly in Gad2+ nuclei from female mice (Fig. 6B) whereas the other eleven were expressed more highly in Gad2+ nuclei from male mice (Fig. 6C). We confirmed that the differences in gene expression did not arise from skewing of cell recovery across the five Rbfox3+ clusters and that both male and female classified nuclei expressed similar levels of Gad2 (Fig. S11). Because we pooled nuclei across male or female mice in this study, one limitation of our analysis is that we are not able to correct for individual variability when comparing differences in the expression of these genes between sexes (
). However at least six of these genes (Nav1, Nos1ap, Pde1a, Plekha5, Ptprd, and Zfp804c) were also reported as sex-biased in estrogen-receptor expressing neuronal populations from other brain regions (
Knoedler JR, Inoue S, Bayless DW, Yang T, Tantry A, Davis CH, et al. (2022): A functional cellular framework for sex and estrous cycle-dependent gene expression and behavior. Cell. 185:654-671 e622.
We next asked whether the genes we identified as sex-differentially expression in Gad2+ neurons showed genetic association with psychiatric or neurological disorders. Using all genes with average expression above 0.25 from our LHb neuronal clusters as background, a one-sided Fisher’s exact test was used to determine if our sex-biased genes had significant enrichment within the dbSNP database of genes linked to a variety of neuronal phenotypes (
). We computed the overlap of our 16 autosomal sex-biased genes with 6 categories of neuronal phenotypes from the dbSNP database, which we also filtered for autosomal genes, and calculated the statistical significance of this overlap. These analyses revealed significant overlap of LHb Gad2+ sex-biased genes with Depression and Alzheimer’s disease, which are both known to have female bias in the human population (Fig. 6D-E) (
Ntng2 as a marker of Gad2/Slc17a6 double positive LHb neurons
The gene for netrin-G2, Ntng2, appeared near the top of the DEG lists for up-regulated genes when we compared these cells within our dataset of Gad2+ neurons (Fig. 4A; Table S1) whereas Ntng1 was significantly down-regulated. The netrins-Gs are known to mediate axonal guidance and thus raised our interest in the possibility these molecules could define the projection targets of the Gad2/Slc17a6+ LHb neurons. Ntng2 was also significantly upregulated when compared to other LHb neurons from the Hashikawa et al., 2020 dataset suggesting it could define these cells regionally (Fig. 5D; Table S2).
We performed in situ labeling using RNAScope FISH probes for Gad2, Ntng1, and Ntng2 to determine the anatomical expression in LHb tissue and confirm the expression of Ntng2 within Gad2+ neurons (Fig. 7; Fig S12). We found distinct patterns of Ntng1 and Ntng2 expression in and around the habenular complex, with Ntng1 expressed predominantly in the surrounding regions including the paraventricular nucleus of the thalamus (PVT) (Fig. 7A-C, E). Interestingly, though Ntng2 was densely expressed in the MHb, it showed a scattered distribution into the medial part of the LHb (Fig. 7A-B, D, E) where nearly all the Gad2+ neurons were observed to co-express Ntng2 (Fig. 7D, F). These findings validate the results from snRNA-seq study and provide a new marker for this cell type within the LHb.
Figure 7Ntng2 is co-expressed in Gad2-positive neurons in the medial subnucleus of the LHb. (A) Representative RNAscope FISH images for Gad2, Ntng1, and Ntng2 as well as DAPI at 20X The medial subnucleus of the LHb on each image is centered within the dotted white box, which is shown at 40X in (B). The regions that correspond to the MHb, LHb, and PVT are labeled on the DAPI image. (C) Nonoverlap of Gad2 with Ntng1 and (D) overlap with Ntng2 in the LHb. (E) Quantification of the percent of positive nuclei for each individual brain section from thresholding in Fig. S12. 30 cells/image/channel from 3 brains. (F) Percentage of the total of all Gad2+ positive nuclei from images of 3 brains that co-express Ntng1, Ntng2, or both. Scale bar: 40 μm.
Single cell sequencing methods have advanced knowledge about brain complexity. However, rare cell types remain challenging to characterize because they are found in too low abundance to drive subclusters with sufficient power for DEG analysis. One approach to this limitation is to genetically enrich rare cells prior to single cell sequencing (
Munoz-Manchado AB, Bengtsson Gonzales C, Zeisel A, Munguba H, Bekkouche B, Skene NG, et al. (2018): Diversity of Interneurons in the Dorsal Striatum Revealed by Single-Cell RNA Sequencing and PatchSeq. Cell Rep. 24:2179-2190 e2177.
Gallegos DA, Minto M, Liu F, Hazlett MF, Aryana Yousefzadeh S, Bartelt LC, et al. (2022): Cell-type specific transcriptional adaptations of nucleus accumbens interneurons to amphetamine. Mol Psychiatry. Online ahead of print. DOI: 10.1038/s41380-022-01466-1.
). Here we used transgenic expression of Sun1-GFP to purify nuclei of Gad2+ neurons from the LHb for snRNA-seq. GFP+ nuclei comprised only ∼3% of all the DAPI+ nuclei in our samples. We recovered less than 700 of these cells from the LHb of each single mouse, which is consistent with Gad2+ cells comprising a few percent of the estimated 13,000 total LHb cells per mouse (
Our analysis resolved 5 clusters of FANS enriched neurons. Gad2 mRNA was only weakly expressed within some of these clusters (clusters 1 and 5; Fig. S5), however this was presumably sufficient to drive enough Sun1-GFP expression from the Cre-dependent transgene to allow these cells to be sorted by FANS (Fig. S1C). When we compared expression of the top DEGs in each cluster with in situ from the Allen Brain Atlas only cluster 2 markers strongly overlapped the distribution of the Gad2+/Slc17a6+ population in the medial subnucleus of the LHb (Fig. 1F, Fig. S1B). Clusters 1 and 3 appear to be derived primarily from Gad2-driven Sun1-GFP transgene expression (Fig. S1A) in the dorsal thalamic regions immediately flanking the LHb. Cluster 5 contains marker genes that are widely expressed in MHb as well as in neurons scattered through LHb. Given that the MHb shows no detectable Gad2 expression (Fig. 1E) or Gad2-driven Sun1-GFP transgene expression (Fig. S1A), cluster 5 is likely to come from cells in the LHb that share some similarities in gene expression with MHb neurons. Finally, cluster 4’s expression of canonical GABAergic genes suggest that this is a local inhibitory interneuron population likely to arise from the lateral LHb, where the inhibitory functions of Pvalb+ neurons have been previously characterized (
). Among our purified Gad2+ LHb neurons, the cells in cluster 2 were unique for their expression of Hcrtr2, the gene encoding the orexin receptor ORXR2, which supports these neurons as the same cells determined by Flanigan et al. (
) to be locally inhibitory within the LHb. We observed both by single cell sequencing (Fig. S5) analysis of cluster 2 and by in situ hybridization (Fig. 3) that Gad2+ in the medial subnucleus of the LHb co-express the vesicular glutamatergic transporter Slc17a6, encoding VGLUT2. We confirmed previous reports that these cells fail to co-express the vesicular GABA transporter VGAT, encoded by Slc32a1 (
), but we did find colocalization of Gad2/+Slc17a6+ cells in LHb with Slc6a1, encoding the plasma membrane GABA transporter (Fig. 3).
If LHb GAD2+ neurons release GABA, one possibility is that reversal of this plasma membrane GABA transporter could be used for non-vesicular GABA release, explaining how these neurons can drive inhibition (
). Alternatively, other transporters could package GABA in vesicles; for example, midbrain dopamine neurons use as Slc18a2, encoding VMAT2, to package GABA into synaptic vesicles for release (
). In some of the dual GABA/glutamate releasing neurons, GABA and glutamate transporters are segregated into separate populations of vesicles within a single terminal (
) currents measured upon activation of the dual Gad2/Slc17a6+ LHb neurons studied here might suggest that these neurons release GABA and glutamate from distinct projections.
One of the most powerful applications of scRNA-seq is linking molecular signatures of distinct cell types with the physiology assigned to a given brain region (
). Though many studies have focused on the importance of LHb projections in the control of dopamine systems in the brain, the Gad2+/Slc17a6+ neurons are thought to project primarily to serotonergic nuclei of the dorsal and median raphe nuclei (DR, MnR) (
). We find that all the Gad2+ neurons we isolated not only express multiple serotonin receptors (Fig. 4E) but also express them at higher levels compared with other types of LHb neurons (Fig. S9). The LHb receives a dense serotonergic projection back from the DR (
) in the Gad2+/Slc17a6+ population compared with other LHb neurons (Fig. 5F), and genes that showed sex-differential expression in LHb Gad2+ cells were more likely than chance to be associated with depression and Alzheimer’s risk genes (Fig. 6D,E). We also observed a segregation in the LHb and surrounding regions between cells that express Ntng1 and Ntng2, encoding GPI-anchored netrin-G proteins. Ntng2 is highly expressed in glutamatergic neurons of the MHb, whereas within the LHb it is selectively expressed in the GAD2+ population of the medial subnucleus. The netrin-Gs are axon guidance/cell adhesion molecules that play important functions in establishing specificity of excitatory synapse formation during development (
). What functions these proteins play in mature neurons, and whether their expression in GAD2+ neurons of the LHb contributes to the neurological phenotypes seen in humans with NTNG2 mutations (
Homozygous frameshift variant in NTNG2, encoding a synaptic cell adhesion molecule, in individuals with developmental delay, hypotonia, and autistic features.
We thank Mariah Hazlett for critical reading of the manuscript and Melyssa Minto for support with the bioinformatics. This work was supported by NIH grant R01DA047115 (A.E.W.). This study was posted as a preprint at bioRxiv with the identifier doi: 10.1101/2023.01.09.523312.
Pandey S, Shekhar K, Regev A, Schier AF (2018): Comprehensive Identification and Spatial Mapping of Habenular Neuronal Types Using Single-Cell RNA-Seq. Curr Biol. 28:1052-1065 e1057.
Cerniauskas I, Winterer J, de Jong JW, Lukacsovich D, Yang H, Khan F, et al. (2019): Chronic Stress Induces Activity, Synaptic, and Transcriptional Remodeling of the Lateral Habenula Associated with Deficits in Motivated Behaviors. Neuron. 104:899-915 e898.
Munoz-Manchado AB, Bengtsson Gonzales C, Zeisel A, Munguba H, Bekkouche B, Skene NG, et al. (2018): Diversity of Interneurons in the Dorsal Striatum Revealed by Single-Cell RNA Sequencing and PatchSeq. Cell Rep. 24:2179-2190 e2177.
Gallegos DA, Minto M, Liu F, Hazlett MF, Aryana Yousefzadeh S, Bartelt LC, et al. (2022): Cell-type specific transcriptional adaptations of nucleus accumbens interneurons to amphetamine. Mol Psychiatry. Online ahead of print. DOI: 10.1038/s41380-022-01466-1.
Paul A, Crow M, Raudales R, He M, Gillis J, Huang ZJ (2017): Transcriptional Architecture of Synaptic Communication Delineates GABAergic Neuron Identity. Cell. 171:522-539 e520.
Knoedler JR, Inoue S, Bayless DW, Yang T, Tantry A, Davis CH, et al. (2022): A functional cellular framework for sex and estrous cycle-dependent gene expression and behavior. Cell. 185:654-671 e622.
Homozygous frameshift variant in NTNG2, encoding a synaptic cell adhesion molecule, in individuals with developmental delay, hypotonia, and autistic features.
00043-5&id=gr1.jpg){kind=link}
00043-5&id=gr2.jpg){kind=link}
00043-5&id=gr3.jpg){kind=link}
00043-5&id=gr4.jpg){kind=link}
00043-5&id=gr5.jpg){kind=link}
00043-5&id=gr6.jpg){kind=link}
00043-5&id=gr7.jpg){kind=link}