Low Dopamine D2 Receptor Expression Drives Gene Networks Related to GABA, cAMP, Growth and Neuroinflammation in Striatal Indirect Pathway Neurons

Background A salient effect of addictive drugs is to hijack the dopamine reward system, an evolutionarily conserved driver of goal-directed behavior and learning. Reduced dopamine type 2 receptor availability in the striatum is an important pathophysiological mechanism for addiction that is both consequential and causal for other molecular, cellular, and neuronal network differences etiologic for this disorder. Here, we sought to identify gene expression changes attributable to innate low expression of the Drd2 gene in the striatum and specific to striatal indirect medium spiny neurons (iMSNs). Methods Cre-conditional, translating ribosome affinity purification (TRAP) was used to purify and analyze the translatome (ribosome-bound messenger RNA) of iMSNs from mice with low/heterozygous or wild-type Drd2 expression in iMSNs. Complementary electrophysiological recordings and gene expression analysis of postmortem brain tissue from human cocaine users were performed. Results Innate low expression of Drd2 in iMSNs led to differential expression of genes involved in GABA (gamma-aminobutyric acid) and cAMP (cyclic adenosine monophosphate) signaling, neural growth, lipid metabolism, neural excitability, and inflammation. Creb1 was identified as a likely upstream regulator, among others. In human brain, expression of FXYD2, a modulatory subunit of the Na/K pump, was negatively correlated with DRD2 messenger RNA expression. In iMSN-TRAP-Drd2HET mice, increased Cartpt and reduced S100a10 (p11) expression recapitulated previous observations in cocaine paradigms. Electrophysiology experiments supported a higher GABA tone in iMSN-Drd2HET mice. Conclusions This study provides strong molecular evidence that, in addiction, inhibition by the indirect pathway is constitutively enhanced through neural growth and increased GABA signaling.

D2Rs are mostly expressed on indirect pathway medium spiny neurons (iMSNs) that mediate behavioral inhibition (nogo signal).Conversely, dopamine D 1 receptors (D1Rs) are mostly expressed on direct pathway MSNs (dMSNs) that mediate reward and the initiation of behavior (go signal) (23,24).In the context of addiction, sensitization and conditioned place preference to psychostimulants are generally elicited by dMSNs and inhibited by iMSNs (25)(26)(27)(28).Unbalanced striatal direct and indirect pathway signaling are hypothesized to be key to the pathophysiology of SUDs and other neuropsychiatric disorders such as schizophrenia and obsessive-compulsive disorder, and neurological conditions such as Parkinson's and Huntington's disease.Therapies for several of these conditions target dopamine neurotransmission and alter signaling balance, partly restoring function but at times leading to troubling side effects (29)(30)(31).For example, Parkinson's disease is treated with levodopa, a dopamine precursor, while dopamine D2R blockade is often used to treat both schizophrenia and Huntington's disease.Despite its centrality to SUDs, there are currently no effective treatments that target dopamine signaling.
We previously reported that iMSNs partially or fully deficient for Drd2 exhibit decreased dMSN excitability due to increased GABAergic collateral inhibition from iMSNs (32)(33)(34).This could contribute to addiction by increasing input needed for the striatum to select and maintain behavior, biasing the brain to pursue high-reward drug-associated cues and ignore lower intensity natural rewards.
In this study, we investigated gene expression changes in iMSNs that may explain the enhanced influence of the indirect pathway due to reduced Drd2 expression.For this, we used a genetically modified mouse with innate low expression of Drd2 selectively in iMSNs and analyzed its ribosome-bound transcriptome (translatome) with normal (2 functional alleles, iMSN-TRAP-Drd2WT) or heterozygous/low (1 functional allele, iMSN-TRAP-Drd2HET) expression of Drd2.iMSNs account for approximately 12% of all cells and approximately 47% of all neurons in the striatum, where the glia/neuron ratio is about 4:1 (35)(36)(37).Furthermore, iMSNs and dMSNs share most of their transcriptional signatures (38).Therefore, to reliably detect and quantify medium and low abundance transcripts specific to iMSNs, we used translating ribosome affinity purification (TRAP) (39) followed by a low-input RNA sequencing strategy and analysis.

Sequencing
The mRNA from 50 to 100 ng totRNA TRAP positive fractions was amplified and converted into double-stranded complementary DNA (cDNA)

Differentially Expressed Gene Analysis
An average of 27 million reads (68 SD) were obtained per sample.Reads were mapped to mouse reference genome (mm10), and gene expression was modeled by generalized linear model using CLC Genomics Workbench (version 10; Qiagen Bioinformatics) in default settings.Protein coding genes (21,950 genes) were filtered by expression of reads per kilobase of exon per million mapped reads (RPKM) $ 2, resulting in 8332 genes considered expressed.A batch effect driven by the TRAP purification was corrected using ComBat (42) (Figure S1 in Supplement 1) from Bioconductor.EdgeR (43) was used for differential gene expression (DEG) analysis.The expression of functional Drd2 mRNA in iMSN D2HET (Figure 1D) was calculated by exon2/exon3-6 ratio, exons 3 False discovery rate (= p adj ) was used for DEG analysis.

Gene Ontology Enrichment Analysis
The DEG analysis of the 8332 expressed genes was further analyzed with QIAGEN Ingenuity Pathway Analysis (IPA) (45).
Given the lack of nonsense-mediated decay for the knockedout allele of Drd2 mRNA (exon-2 deletion), its expression value was divided in half before import to account for expression of the functional isoform.An IPA core analysis (based on expression and fold change [FC]) was performed using default settings except for the following: reference set = user dataset (8332 genes) and set cutoff = .05p-value (unadjusted or nominal).Benjamini-Hochberg (= p adj ) was used for enrichment analysis.
An independent set of 24 iMSNs-TRAP-Drd2WT and -HET mice was used for qPCR validation (Figure S2 in Supplement 1) of RNAseq gene expression results.qPCR samples were processed similarly to RNAseq samples until the double stranded cDNA was obtained.Then, 10 ng of double-stranded cDNA per well were used in Custom TaqMan Array Fast 96well Plates (No. 4413261; Thermo Fisher Scientific), multiplexing with GAPDH-VIC and the gene target in FAM.TaqMan Fast Universal PCR Master Mix (No. 4352042; Thermo Fisher Scientific) and the QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific) were used to run the fast 96-well plates.Two samples were excluded because of low quality.
ComBat correction was used to address the qPCR TRAPbatch effect using DCt values.

Gene Expression From Postmortem Brain Tissue From Subjects With Cocaine Abuse History
The gene expression levels for DRD2 and FXYD2 from an independent transcriptomics study (yet to be published) on postmortem human brain from the anterior caudate and the nucleus accumbens (NAc) from subjects with a severe cocaine abuse history (n = 25) and age-matched unaffected control subjects (n = 25) were obtained from the University of Miami Brain Endowment Bank.Details on samples and psychiatric pathology of these individuals have been reported elsewhere (46).Briefly, neuropathological specimens were obtained during routine autopsy.Total RNA was extracted from approximately 100 mg frozen caudate or ventral striatum, including on-column DNase I treatment.Approximately 500 ng totRNA/ sample were sequenced at the Broad Institute (Cambridge, MA), using Illumina TruSeq library construction (Illumina, Inc.) including poly-A selection, run as 76-bp paired-end to a depth of approximately 50 million reads and aligned against Ensemble transcript reference.

Gene Ontology Analysis
To identify complex biological functions affected by innate, low Drd2 expression in iMSNs, we extended the list of genes of focus to include all those with suggestive or potential association (p # .05,n = 474 genes).Of these 474 potentially differentially expressed genes, 214 were upregulated and 260 were downregulated (Figure 2) and will be referred to as the extended DEG list.To identify molecules that could explain gene expression changes observed in iMSN D2HET , we performed upstream regulator analysis in IPA.Known Creb1 targets were strongly enriched (p adj = 4 3 10 210 , 52 out of 474 genes, i.e., 11%), and Creb1 function was strongly predicted to be increased (z = 2.6) (Figure 3; Table S3 in Supplement 2).Creb1 is a transcription factor well-recognized to be activated by cAMP/PKA signaling, facilitate synaptic plasticity, and play important roles in learning, long-term memory formation, and addiction (71)(72)(73).Other upstream regulators also predicted to be responsible for subsets of the observed differential expression pattern included beta-estradiol responsive to hormones, interferon gamma (IFNG), interleukin 2 (IL2) and interleukin 1b (IL1B), Ca 11 , and SNCA.
While the statistical significance for enrichment of canonical pathways did not survive correction for multiple testing (Table S4 in Supplement 2), glutamate dependent acid resistance (conversion of glutamate to GABA by Gad enzymes, p = .003)and cAMP-mediated signaling (p = .03)remain likely candidates to be affected in this low Drd2 model.See Figure S4 in Supplement 1 for iMSN D2HET DEGs in these signaling pathways.Further gene ontology analysis of the extended DEG list revealed strong enrichment in genes with diverse functions, including release of fatty acid and lipid and inflammation of the nervous system (Figure 4; Figures S5 in Suppement 1; Table S5 in Supplement 2).Dysregulated lipid metabolism in iMSNs has been associated with reward-related psychopathologies (74).Enriched disease and function terms with strong predicted activation (z score .2) include epilepsy or neurodevelopmental disorder, hypothermia, and secretion of catecholamine, while functions with strong predicted inhibition (z score , 22) include excitation of neurons, formation of neointima, and immune-mediated inflammatory disease.

The Gene Network of Drd2 in iMSN D2HET Is Enriched in Cell-to-Cell Communication Molecules
To identify genes whose expression covaried with Drd2 expression, we performed an independent weighted gene coexpression network analysis.In line with the IPA and further highlighting the impact of Drd2 expression on the functional output of iMSNs, the module containing Drd2 showed strong enrichment in genes involved in cell-to-cell communication, including ion transport, synaptic signaling and assembly, axon guidance, and matching molecular functions and cellular localizations (Figure S6 and Supplemental Methods in Supplement 1).Fxyd2 was one of the genes with strongest differential expression (both in magnitude and statistical significance) (Figure 2B) in the iMSN D2HET RNAseq and independent qPCR samples (Figure 5A).Fxyd2 encodes the modulatory g-subunit of the ATP-dependent Na/K pump that is essential for membrane potential and therefore neural excitability.Fxyd2 is a phosphorylation target of PKA ( 75) and a member of the cAMP pathway (Kyoto Encyclopedia of Genes and Genome).In the brain, Fxyd2 function is poorly understood but it modulates neuropathic pain through inhibition of the Na/K pump in nociceptive neurons (76,77).We found that FXYD2 and DRD2 mRNA expression was inversely correlated in both mouse purified iMSNs (Figure 5B) and human postmortem caudate (whole-tissue RNAseq) of individuals with severe cocaine abuse history and control subjects (Figure 5C).This correlation was validated by independent cohorts from public repositories (Figure S7A in Supplement 1).FXYD2 expression was not significantly different between cocaine and control groups (p = .11)(Figure S7B in Supplement 1); however, FXYD2 expression in both mouse iMSNs (Figure 2C) and human is relatively low, and several human postmortem whole tissue samples had nondetectable levels (Figure 5C; Figure S7B); note axis expression scales for DRD2 vs. FXYD2.RNAscope staining (Figure 5D) together with publicly available single-cell RNAseq data from mouse striatum (78) (Figure 5E) suggest Fxyd2 expression in striatal iMSNs, dMSNs and to a lesser extent in interneurons, with little or no expression in glial populations.
Finally, no association approaching genome-wide significance was observed (Figure S7C (32,82).Thus, electrophysiological recording of dMSNs provides a useful readout of GABA and D2R signaling in the striatum.We performed wholecell voltage-clamp recordings from tdTomato1 dMSNs from the NAc core in sagittal brain slices of iMSN-Drd2HET and control mice.The D2R-like agonist quinpirole (Figure 6) (green trace) decreased the holding current in WT dMSNs (compared with baseline, gray trace) but exerted no effect in dMSNs from iMSN-Drd2HET mice.This loss of quinpirole-mediated response (in both salineand cocaine-treated animals) provides further evidence of the functional impact of D2R reduction (83).Furthermore, the baseline average holding current of dMSNs in iMSN-Drd2HET was smaller than in control mice (224/218 pA vs. 238/252 pA in saline and cocaine-treated groups, respectively), indicating more hyperpolarized dMSNs in iMSN-Drd2HET mice compared with WT, consistent with a heightened inhibitory GABA tone.dMSNs in WT animals treated with repeated cocaine (15 mg/kg/day, 5 days) showed persistent depolarization at baseline compared with saline-treated control mice (252 vs. 238 pA).More depolarized dMSN membrane potential after repeated cocaine suggests reduced GABA inhibition.This cocaine-induced depolarization was not seen in dMSNs of iMSN-Drd2HET mice.This suggests that D2Rs are required for mediating the effect of repeated cocaine in dMSNs, likely via suppression of striatal GABA release (32).

DISCUSSION
Reduced D2R availability in the striatum is both a consequence of and an etiologic contributor to SUD.Innate low D2R expression in young/adult mice appears to simultaneously stimulate the no-go pathway and inhibit the direct go pathway (34).Here, we sought to better understand the long-term consequences of innately reduced Drd2 expression by analyzing the purified translatome (ribosomebound transcriptome) of iMSNs from mice with WT or heterozygous expression of Drd2 (selectively in iMSNs).
Consistent with our previous findings of increased striatal GABA signaling in mice with D2R deficiency (32,33), we found Gad1 and Gad2, key genes encoding GABA-synthesis enzymes, upregulated and a predicted inhibition in stimulation of neurons function (z score = 22) by enrichment analysis in iMSN D2HET .Electrophysiological recordings showed that dMSNs were more inhibited at baseline in iMSN-Drd2HET mice compared with WT control mice.
Possibly related to the increased GABA tone, Fxyd2, a potentially inhibitory subunit of the Na/K pump, was robustly upregulated in iMSN D2HET .Its modulatory role is cell-type and environment dependent (84) but was shown to be inhibitory in nociceptive neurons (76,77) where loss of Fxyd2 resulted in neural hyperpolarization (77).If Fxyd2 also inhibits the Na/K pump in iMSNs, its upregulation could result in increased iMSN excitability.This could be an adaptive response by iMSNs to counteract the heightened GABA tone.
An evolutionarily conserved negative correlation between FXYD2 and DRD2 mRNA expression was observed in both our mouse model and in human postmortem caudate with and without severe cocaine use history.This correlation could be explained by a genetic variant at the DRD2 locus affecting FXYD2 expression (i.e., expression quantitative trait locus), which we could not test due to the limited detectability of FXYD2 in the brain.
iMSN D2HET displayed differential expression of genes associated with cAMP signaling and cellular growth.Mapk3 (Erk) modulates processes that overlap with those of cAMPdependent Pka.We have previously observed a shift from Pka toward the Erk/Mapk signaling pathway in dMSNs from iMSN-Drd2KO mice (83).Here, iMSN D2HET showed downregulation of Mapk3 (Erk), suggesting an opposing shift in iMSNs.
Creb1, a transcription factor well-known for its role in growth, learning, and synaptic reinforcement, was predicted to be an upstream regulator (p adj = 4 3 10 210 ) with increased activity (z score = 2.6).There was no differential expression of the Creb1 gene at the time of sample collection, suggesting it exerted its effect at an earlier time point.Studies across development are needed to test the hypothesis of differential neural growth of iMSNs over dMSNs in iMSN-Drd2HET mice.
Several genes were associated with inflammation.Inflammation of the nervous system was among the most significantly enriched functions, and immune-mediated inflammatory disease was predicted to be inhibited (Figure 4; Table S5 in Supplement 2).There is a growing body of evidence for a role of neuroinflammation (not accompanied by infiltration of peripheral immune cells) in several psychiatric disorders, including SUDs (85)(86)(87).Transcription profiling of rhesus macaques following long-term (w100 days) cocaine selfadministration revealed upregulation of neuroinflammationrelated genes in the NAc but not in the ventral tegmental area (88), showing differential neuroinflammation response to drugs of abuse across brain regions.Anti-inflammatory strategies to treat SUDs show promising results including improvements in behavioral and cognitive outcomes (89-91) but with negative as well as positive outcomes in early clinical trials (92).iMSN D2HET also displayed upregulated Cartpt and downregulated S100a10, reproducing gene expression changes observed in mouse studies in cocaine paradigms.In the striatum, cocaine and amphetamine upregulate Cartpt mRNA expression and CART peptides, and co-administration of CART peptide and cocaine into the NAc reduced cocaineinduced locomotor activation (58,93,94), suggesting a compensatory role for Cartpt.Conversely, mice with reduced S100a10 (p11), a small calcium-binding protein involved in neurotransmitter transport, have an enhanced cocaine conditioned place preference, while p11 overexpression in the NAc reduced it (70).Mice with low p11 also exhibited depressionlike behavior, and its restoration in the NAc recovered the phenotype (95).
An intrinsic challenge of this study was the observed and expected small-magnitude changes between groups because they only partially differ in the expression of a modulatory signaling receptor (Drd2).Gene expression changes translate into function whether they are big or small in magnitude, but small-magnitude differences increase the number of samples needed to statistically resolve them.Regarding technical strengths, low-input sequencing of TRAP positive fractions allowed us to study the iMSNs (w12% of cells in the striatum) translatome and reliably detect transcripts with medium and low levels of expression that would have been lost via whole-tissue RNAseq.This was critical to our research, particularly given the functionally opposing roles of dMSNs and iMSNs, which also share a highly overlapping molecular profile.
In summary, we identified numerous differentially expressed genes in iMSNs driven by low D2R expression, modeling an observed trait of addiction in humans.We provided molecular evidence for enhanced GABA transmission.We identified enrichment in lipid metabolism, growth-related genes, cell-tocell communication, and synaptic components that may reflect neural growth and/or increased maintenance given previous active growth.Striatal development in iMSN D2HET and by extension in people with genetically encoded lower levels of D2R expression may feature an increased number of inhibitory iMSN-dMSN axon collaterals and an increased number of GABAergic presynaptic vesicles and release.Thus, our results further support an addiction model in which low D2R expression drives changes in the striatal microcircuitry and help explain its contribution to an enhanced indirect pathway and SUD-related behaviors.

Figure 1 .
Figure 1.Enrichment of iMSN translatome signal in Drd2-low expression mice.(A) Experimental design.Whole striatum was microdissected from mice with WT or HET expression of Drd2 and transgenic for a GFP-ribosomal subunit fusion protein conditionally expressed in Drd2 neurons (iMSNs) for cell-specific mRNA purification.Conditional expression of the TRAP system and the floxed Drd2 allele were driven by the Adora2a promoter.Isolation of actively translating mRNA (translatome) from iMSNs was achieved by pull down of GFPribosomes-associated mRNA followed by low input preamplification and sequencing.Differential gene expression from RNAseq samples was validated by qPCR (Figure S2 in Supplement 1) on an independent set of mice and samples.(B) Quantitative enrichment analysis of iMSN signal compared to total striatum by RNAseq shows a strong enrichment in iMSN translatome signal.(C) Drd2 exon signal coverage and enrichment in iMSN samples vs. total striatum, visualization on IGV (44).For comparison purposes, plots were set at the same scale.(D) Expression level of the full-length, functional Drd2 mRNA calculated by mRNA exon coverage signal from the RNAseq data.GFP, green fluorescent protein; HET, heterozygous; IGV, Integrative Genomics Viewer; iMSN, indirect medium spiny neuron; mRNA, messenger RNA; NAc, nucleus accumbens; RNAseq, RNA sequencing; qPCR, quantitative polymerase chain reaction; RPKM, reads per kilobase of transcript per million mapped reads; Tot Str, total striatum; TRAP, translating ribosome affinity purification; VTA, ventral tegmental area; WT, wild-type.

RESULTSDrd2-
Low iMSN Translatome From a Mouse Model of Addiction iMSN-specific, ribosome-bound mRNA (translatome) from a genetically diminished Drd2 expression mouse model was isolated by TRAP, amplified, and sequenced (Figure1A).Cre expression was driven by an Adora2a promoter in animals with one (HET) or no (WT) floxed alleles in exon 2 (containing the start codon) of the Drd2 gene.The amplified and sequenced TRAP positive fractions showed strong enrichment for iMSNspecific transcripts (Penk 7.8-fold, Drd2 6.8-fold, Adora2a 2.6-fold) and reduced levels of non-iMSN transcripts compared to total striatum (Figure1B, C; FigureS3in Supplement 1).Expression of functional Drd2 mRNA in iMSND2HET was approximately 60% of iMSN D2WT (Figure1D), while expression of Drd2 exons other than exon 2 was unaltered, indicating lack of nonsense-mediated decay of Drd2 mRNA transcribed from this construct.

Figure 3 .
Figure 3. Creb1 and others were predicted upstream regulators of the observed expression profile in iMSN D2HET .(A) Upstream regulators predicted to have driven the expression changes observed in overlapping and nonoverlapping subsets of genes (see also Table S3 in Supplement 2) from the extended DEG list in iMSN D2HET .(B) Gene network showing Creb1 and forskolin (exogenous drug) targets within our dataset.DEG, differentially expressed gene; iMSN, indirect medium spiny neuron.

Figure 4 .
Figure 4. Disease and function enrichment analysis in iMSN D2HET .(A) Top disease and function annotations significantly enriched in the extended DEG list in iMSN D2HET .(B) Gene network for behavior-related functions: learning, cognition, and long-term memory.(C) Gene network for lipid metabolism-related functions: release of fatty acid, release of lipid, release of eicosanoid, concentration of GABA, release of GABA, and concentration of lipid.Functions blue or orange indicate a predicted reduction or increase of function, respectively.For a complete list of genes, annotations, and values, see TableS5in Supplement 2. DEG, differentially expressed gene; GABA, gamma-aminobutyric acid.

Figure 6 .
Figure 6.Functional impairment of D2R mediates electrophysiologic response in striatum of mice with low Drd2 expression and with evidence for enhanced GABA tone.(A) Recordings from MSNs in control Adora2a-Cre mice show changes in holding current in response to application of the D2-likeagonist quinpirole 1 mM (green trace).This response is severely impaired in mice with low Drd2 expression (iMSNs-Drd2HET), which show no changes in holding current after quinpirole.Note that there are changes in baseline holding current in mice with low Drd2 that are suggestive of enhanced GABAergic tone in iMSN-Drd2HET mice.(B) Cocaine pretreatment (5 days, 15 mg/kg) enhanced the average holding current in control mice from 238 to 252 pA but not in mice with low Drd2 expression, which show persistently smaller holding currents than control mice consistent with increased GABAergic tone in mice with low D2Rs.D2R, dopamine type 2 receptor; GABA, gamma-aminobutyric acid; HET, heterozygous; iMSN, indirect medium spiny neuron.