VTA dopamine neuron population transcriptomics

Stereotactic surgery and injection of AAVs were conducted as described above for experiments 1 and 2. To enable identification of VTA DA neurons, mice were injected with a cocktail of 2 AAV vectors, each in a volume of 300 nl per hemisphere: ssAAV-9/2-mTH-EGFP-WPRE-SV40p(A) (AAV mTH-EGFP, 7.0 × 10¹¹ vg/ml; Viral Vector Facility, ETH and University of Zurich), to achieve EGFP expression in DA neurons; ssAAV-9/2-hGAD67-chI-mScarlet-I-SV40p(A) (AAV hGAD67-mScarlet-I, 8.0 × 10¹¹ vg/ml; Viral Vector Facility, ETH and University of Zurich), to achieve m-Scarlet-I expression in GABA interneurons. In each vector, expression of a specific fluorescent protein was therefore dependent on a promoter-region sequence of a neuron type-specific marker gene: EGFP under the control of tyrosine hydroxylase (Th) promoter for DA neurons, and monomeric bright red fluorescent protein under the control of glutamate decarboxylase 67 (Gad1) promoter for GABA (inter)neurons. Stereotactic coordinates were set to inject into VTA at AP −3.1 mm, ML ± 0.5 mm, DV −4.9 mm⁴⁹. Mice were weighed and wound healing was controlled for 10 days post-surgery.

To validate the specificity of the AAV vectors, pilot mice were injected with AAV mTH-EGFP and/or AAV hGAD67-mScarlet-I, and brains were perfused-fixed with PBS and then ice-cold paraformaldehyde (PFA, 4%). Brains were extracted and post-fixed in PFA, and then transferred into 30% sucrose solution for 48 h prior to freezing. Using a freezing microtome (Leica), brains were sectioned coronally at 40 µm from bregma −2.8 to −3.5 mm for VTA sections, and stored in tissue collection solution (TCS; glycerine and ethylene glycol in 0.2 M phosphate buffer; Sigma-Aldrich) at −20 °C. Using a 24-well plate, sections were placed free-floating in Tris-Triton buffer (pH 7.4) and then underwent immunofluorescence staining for TH or GAD67. For TH, a primary antibody of rabbit anti-TH (1:2500; AB152, Chemicon) and a secondary antibody of donkey anti-rabbit IgG-Alexa Fluor 647 (1:1000; A31573, Invitrogen) were used. For GAD67, a primary antibody of mouse anti-GAD67 (1:200; ab26116, Abcam) and a secondary antibody of donkey anti-mouse IgG-Alexa Fluor 647 (1:1000, A31571, Invitrogen) were used. Images including the VTA and surrounding regions were acquired using a confocal laser scanning microscope (Leica SP8) at x20 magnification. Separate laser channels were used for DAPI (405 nm), EGFP (488 nm), mScarlet-I (552 nm) and Alexa Fluor 647 (638 nm).

Littermate pairs were allocated to CSS (n = 6 mice) and CON (n = 6 mice) by counterbalancing on body weight. Mean cumulative duration of daily attack experienced by CSS mice was 49.6 ± 5.4 s (range: 43.0-55.5 s); all CSS mice displayed submissive behaviour and vocalisation during the proximal stressor. From day 15 of CSS until the end of the experiment, each CSS mouse remained in the same divided cage with the same CD-1 mouse without further attacks.

At 3 days after completion of CSS/CON, mice were deeply anaesthetized and then perfused with PBS (20 mL) at RT. The brain was removed and placed in a cryo-mould (E6032-ICS, Sigma) with embedding medium (Tissue-TEK OCT Compound). The cryo-mould was then placed on dry ice, wrapped in aluminium foil and a polythene bag and stored at −80 °C.

Frozen brains were processed using RNA- and RNAse-free conditions throughout. Using a cryostat set at −18 °C, coronal sections that included the VTA at AP −2.9 to −3.3 mm were cut at 10 µm and mounted (3 sections/slide) on RNAse-free PET membrane slides (50102, Molecular Machines & Industries, MMI). Sections then underwent fixation and dehydration: 100% ETOH at RT for 20 s and xylene at RT for 20 s. Slides/sections were placed on their edge in a covered box at RT for 10 min or until completely dried, and then in a capped 50 ml Falcon tube for storage at −80 °C for 3 days maximum. Tissue samples that were EGFP⁺ were collected from these coronal sections using a laser capture microdissection (LCM) system (CellCut, MMI). Fluorescence settings were optimised for visualisation of EGFP⁺ tissue (channel FITC) or mScarlet-I⁺ tissue (channel TRITC). The membrane slide was positioned and using 4x magnification, VTA tissue areas that were EGFP⁺ were each encircled at Ø=35 µm using the MMI CellTools software. Selected EGFP⁺ areas that were also mScarlet-I⁺ were deselected. There were 20-30 EGFP⁺/m-Scarlet-I^- samples per VTA hemisphere/section; these were encircled for both hemispheres for each of the 3 sections on the membrane slide. An MMI Universal UV laser (355 nm, 2 µJ, 4 kHz frequency, 500 pico-s pulse-duration) at 88% laser power was activated (velocity = 51 µm/s, focus = 2233 µm) and the designated tissue areas were collected on the adhesive cap of an MMI isolation tube (0.5 ml). The procedure was conducted with 3 membrane slides (7-9 sections) and isolation tubes per mouse, to yield a total of 500 EGFP⁺/mScarlet-I^- tissue samples per mouse; this was with the exception of one CSS mouse in which the EGFP/mScarlet-I signals were weak (likely due to misplaced injection), and this mouse was excluded from the experiment. Following tissue collection, tissue lysis was conducted by adding QIAzol (100 µl) to the tube, triturating the tissue on the cap with 20 µl volumes and returning this volume to the tube; the tube was closed, inverted for 15 min at RT, and vortexed for 1 min, inverted for 5 min and centrifuged for 5 s. The tube was then sealed with Parafilm and frozen at −80 °C until RNA extraction.

Per mouse sample, lysate aliquots (3 × 100 µl per sample) were pooled to give a final lysis volume of 300 µL. Samples were transferred to 2 mL PhaseLock tubes (QuantaBio). A half volume of chloroform:isoamyl alcohol (24:1 v:v) was added before shaking, 3 min RT incubation and centrifugation at 4 °C. The aqueous phase was then transferred to a 1.5 mL Eppendorf tube and mixed with a 1.5 volume of isopropanol (Sigma). After thorough pipette mixing, the isopropanol mixture was applied to a RNeasy MinElute spin column and total RNA was extracted using the miRNeasy Micro Kit (Qiagen) with a DNase treatment. Samples were eluted in 14 µL nuclease-free water. RNA samples were assessed both quantitatively and qualitatively using the High Sensitivity Total RNA 15nt Analysis DNF-472 Kit on a 48-channel Fragment Analyser (Agilent). Total RNA yield was 1.14 ± 0.20 ng; RNA integrity could often not be computed due to low input.

Up to 1.4 ng of total RNA was used for cDNA synthesis, conducted with the SMART-Seq® v4 Ultra Low Input RNA kit (Takara Bio); 12 amplification cycles were conducted. After clean-up, up to 10 ng of cDNA was used to generate the final sequencing libraries with the tagmentation-based DNA Prep Kit (#20018705) and the IDT® DNA/RNA UD Indexes Set A (#20026121), both Illumina®. The index PCR was performed with 9 cycles, while the final library was eluted in 30 µL EB Buffer. Low input mRNA libraries were then quantified using the High Sensitivity dsDNA Quanti-iT Assay Kit (ThermoFisher) on a Synergy HTX (BioTek). Library molarity averaged 42 nM. Libraries were also assessed for size distribution and adapter dimer presence (10, Rd3: 10, Rd4: 101), reaching an average depth of 26 million Pass-Filter reads per sample (14.2% CV).

Sequencing reads were mapped to the Mus musculus reference genome (mm10) using STAR v2.5.2b allowing for soft clipping of adapter sequences. An average of 20 million reads per sample was obtained, from which approximately 10 million reads were assigned to genomic features. Transcript quantification was conducted with RSEM v1.3.0 and feature Counts v1.5.1. QC and downstream bioinformatics analyses were performed with R v4.1.0 and Bioconductor v3.12 tools, respectively. Briefly, we identified expressed genes based on the distribution of median log2 raw counts across samples, and this yielded a median of 12,250 expressed genes per sample in the experiment. A Gaussian mixture model was fitted to the distribution with mclust v5.4.7 to identify two clusters: genes with median expression values belonging to the cluster with the mean closest to 0 were filtered out from the expression matrix. Then, we normalised the expression matrix using the variance stabilising transformation from package DESeq2 v1.32.0 and identified the 500 highest variable genes (HVGs). Principal component analysis (PCA) was performed with these 500 HGVs using PCAtools 2.4.0. Using brain cell type-specific marker genes to identify the relative contribution of different cell types to the RNA sample (mouse visual cortex⁵⁵), the DA neuron gene marker Th, as well as the pan-neuronal gene marker Snap25, displayed consistent and markedly higher expression than marker genes for GABA (inter)neurons (Gad1, Gad2) and each of the glial cell types (astrocyte: Aqp4, oligodendrocyte progenitor cell: Pdgfra, myelinating oligodendrocyte: Opalin, microglia: Ctss). Differential gene expression analysis (DGEA) was conducted for CSS vs CON with DESeq2 v1.32.0, using an absolute log2 fold-change of at least 0.5 and a raw p-value of ≤0.001. Functional enrichment analysis of differentially expressed genes was performed with enrichR v3.0 against the mouse-specific pathway collection from KEGG 2019.