Abstract
The striatum plays a central role in directing many complex behaviors ranging from motor control to action choice and reward learning. In our study, we used 55 male CFW mice with rapid decay linkage disequilibrium to systematically mine the striatum-related behavioral functional genes by analyzing their striatal transcriptomes and 79 measured behavioral phenotypic data. By constructing a gene coexpression network, we clustered the genes into 13 modules, with most of them being positively correlated with motor traits. Based on functional annotations as well as Fisher's exact and hypergeometric distribution tests, brown and magenta modules were identified as core modules. They were significantly enriched for striatal-related functional genes. Subsequent Mendelian randomization analysis verified the causal relationship between the core modules and dyskinesia. Through the intramodular gene connectivity analysis, Adcy5 and Kcnma1 were identified as brown and magenta module hub genes, respectively. Knock outs of both Adcy5 and Kcnma1 lead to motor dysfunction in mice, and KCNMA1 acts as a risk gene for schizophrenia and smoking addiction in humans. We also evaluated the cellular composition of each module and identified oligodendrocytes in the striatum to have a positive role in motor regulation.
Introduction
The striatum plays a central role in guiding numerous complex behaviors, ranging from motor control to action selection and reward learning (Prager and Plotkin, 2019). The diverse responsibilities of the striatum are reflected by the complexity of its organization (Prager and Plotkin, 2019). It is composed of subregions with different functions that receive signals from different brain regions. In rodents, the dorsal striatum and the dorsolateral striatum receive excitatory signals from the limbic and sensorimotor cortex, respectively, while the intermediate regions can be activated by signals from the axons of neurons in the relevant cortex (Eisinger et al., 2018; Prager and Plotkin, 2019). Structural abnormalities or genetic mutations in the striatum lead to a variety of neurodegenerative diseases and movement disorders. For instance, abnormal development of the striatal circuit leads to Huntington's disease (Tereshchenko et al., 2020), and gene deletion (e.g., Shank3B−/−) leads to behavioral and striatal abnormalities (Peixoto et al., 2019).
A fundamental process in biology is the flow of genetic information from DNA to RNA and then to proteins to mediate physiological functions (Crick, 1970). At the cellular level, the expression of functionally related genes is often tightly coregulated to enable temporal control of molecular complex formation and to avoid conflicting molecular processes (Wang et al., 2022). Gene coexpression network (Zhang and Horvath, 2005; van Dam et al., 2018) is a systems biological method used to describe patterns of gene associations between different samples. Instead of focusing solely on differentially expressed genes, coexpression network information from all or thousands of genes is used to identify gene sets with high covariation, thus helping in identifying candidate biomarker genes or therapeutic targets based on the endogeneity of the gene set and the association between the gene set and phenotype. This strategy has been applied in many research fields, including neuronal system-related diseases (Miller et al., 2010; Swarup et al., 2019; Haghani et al., 2020).
The CFW mice are an outbred mouse population that exhibits rapid linkage disequilibrium decay compared with other mouse populations, which is particularly important for finding accurate quantitative trait locus (QTL) by genetic mapping (Keane et al., 2011; Svenson et al., 2012). CFW mice are currently used in many research fields, including hearing impairment (Du et al., 2022) and drug reaction (Ovary et al., 1975; Basante-Romo et al., 2021; Du et al., 2022). Especially, Parker et al. (2016) pioneered the use of genotyping by sequencing to obtain CFW mice genotypes at 92,734 single-nucleotide polymorphisms (SNPs) and a large number of behavioral traits, including conditioned fear, schizophrenia, motor, and methamphetamine sensitivity, as well as the transcriptome of the striatum. By applying QTL and expression QTL (eQTL) mapping, Azi2 and Zmynd11 were identified as QTL candidates related to methamphetamine sensitivity and anxiety-like behavior, respectively.
Although the study showed that CFW mice are a superior model for rapid identification of QTL loci and functional genes, they did not conduct further systematic analysis in terms of coexpression network. Here, we reanalyzed their striatal transcriptome data and behavioral phenotypes using a gene coexpression network and finally found the key gene set that could explain the main function of the striatum, as well as the active role of oligodendrocytes in the striatum for motor regulation.
Materials and Methods
Data source
The data used in this study came from C. C. Parker et al.'s study on CFW mice (http://dx.doi.org/10.5061/dryad.2rs41; Parker et al., 2016; Carbonetto, 2017). Behavioral phenotypes were collected from 1,219 adult male CFW mice. A total of 79 types of phenotypes were tested, including nine conditioned fear traits, 30 motor phenotypes, 15 methamphetamine (MA) sensitivity phenotypes, and 25 prepulse inhibition (PPI) phenotypes (Extended Data Table 1-1). Finally, the striatum transcriptome sequencing was performed from 55 randomly selected mice and was normalized by the fragments per kilobase of transcript per million mapped reads method.
Construction of coexpression network and module–trait relationship analysis
Striatal mRNA expression data was used to construct a coexpression network using the weighted gene coexpression network analysis (WGCNA) package (version 1.71; Zhang and Horvath, 2005; Langfelder and Horvath, 2008). Through sample-level clustering analysis, the outlier samples with a height greater than 150 were removed, and finally 41 samples remained. The edge properties of the network were calculated as follows:
To identify the correlations between the phenotypes and modules, we performed Pearson's correlation coefficient analysis between the traits and module eigengenes, which are defined as the first principal component of a given module. P < 0.05 was considered as a significance threshold (Extended Data Table 2-1).
Table 2-1
Module grouping information and module phenotype correlation analysis. Download Table 2-1, XLSX file.
Gene function enrichment analysis of coexpression modules
In order to identify the functional significance of each module, we performed Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes pathway, and Mammalian Phenotype Ontology (MPO) analyses using WEB-based Gene Set Analysis Toolkit (Liao et al., 2019; http://www.webgestalt.org/). The annotations with p < 0.05 were considered significant.
Association between coexpression module genes and striatal-related diseases
To identify the coexpression modules enriched in striatal-related diseases, we identified all genes associated with schizophrenia, anxiety, memory, and locomotory behavior based on genome-wide association studies (GWAS) from GWAS Catalog (Sollis et al., 2023) and International Mouse Phenotyping Consortium (IMPC) (Dickinson et al., 2016) databases. The conversion of all GWAS genes to mouse homologous genes was performed using the R package homologene (version 1.4.68.19.3.27; Extended Data Table 4-1).
Further, we used Fisher's exact and hypergeometric tests to measure the statistical significance of the representation of GWAS and IMPC genes in each module (Sabik et al., 2021). The modules significant with p < 0.05 were prioritized.
Meanwhile, we searched for stroke-related genes from GWAS Catalog, including cerebrovascular disorder, cardioembolic stroke, and ischemic stroke, and retained 804 genes after homozygous transformation to serve as a control.
Mendelian randomization (MR) analysis
Theories and methods for MR analysis have been well established (Smith and Ebrahim, 2003; Davey Smith and Hemani, 2014). In our study, we used MR analysis to verify the causal relationship between modules and traits. Firstly, we obtained the TopEffects eQTL (containing the top association per gene and permutation thresholds) of the basal ganglia (BG) region from MetaBrain (de Klein et al., 2023; https://download.metabrain.nl/files.html) and then matched them with the transformed module genes by gene symbol. We then obtained GWAS summary statistics data for schizophrenia (ieu-b-42), memory loss (finn-b-MEMLOSS), movement disorders (finn-b-G6_XTRAPYR), Alzheimer's disease (ieu-b-5067), and anxiety (ukb-d-KRA_PSY_ANXIETY) from the IEU OpenGWAS project (Hemani et al., 2018; Elsworth et al., 2020). In the present study, the TwoSampleMR R package (Hemani et al., 2018; version 0.5.6, https://mrcieu.github.io/TwoSampleMR/) was used for MR inverse variance weighted (IVW) analysis. Genetic variants (SNPs) associated with the gene expression in modules (MetaBrain cis-eQTL data) were used as instrumental variables (Extended Data Table 5-1), and the genome-wide summary statistics from the IEU OpenGWAS project were used as the outcome. Calculated odds ratio (OR) values between gene expression and diseases were calculated using the generate_odds_ratios function in the TwoSampleMR package. We used the R package homologene (version 1.4.68.19.3.27) for the conversion of module genes to human homologous genes. This analysis involved the European population.
Identification and functional analysis of the hub genes
For identifying the hub genes in the core module, we first selected the genes with the strongest connectivity within the module and then analyzed their detailed functions. Here, we used the Mouse Genome Informatics (MGI) database (https://www.informatics.jax.org/) to obtain information about mutations, alleles, and phenotypes of the hub genes, whereas the association of the hub genes with diseases was obtained from GWAS Catalog (Sollis et al., 2023). MGI (Baldarelli et al., 2021) is the international database resource for the laboratory mouse, providing integrated genetic, genomic, and biological data to facilitate the study of human health and disease. Correlation analysis between the eigengenes of hub genes and phenotypes was performed using Pearson’s correlation coefficient analysis. A p < 0.05 was considered significant. Using the WGCNA exportNetworkToCytoscape function, we obtained the gene coexpression network of the module harboring the hub genes and visualized the network using Cytoscape (v. 3.9.1; Shannon et al., 2003).
Cell-type identification in modules
Mouse Cell Atlas is a database (Han et al., 2018; Fei et al., 2022; Wang et al., 2023; https://bis.zju.edu.cn/MCA/index.html) that provides different cell-type marker genes in various tissues of mice at different ages. We used the adult mouse brain marker genes obtained from MCA2.0 (Extended Data Table 8-1) to overlap with all our module genes and evaluate the enrichment of different brain cell types in each module.
Results
Overview of the study design
In this study, we first constructed a gene coexpression network using the striatal transcriptome data. Next, we performed module trait association analysis, module functional enrichment, and GWAS and IMPC gene enrichment analyses to identify the core modules. MR analysis was used to further verify the accuracy of the core modules and finally to identify the key functional genes in the striatum. At the same time, we evaluated the presence of different cell types in the modules through the enrichment of cell marker genes (Fig. 1).
Overview of the study. Gene coexpression network was constructed using the striatal transcriptome from 55 CFW mice, followed by module–trait association, functional enrichment, and GWAS and IMPC gene enrichment to pinpoint the core modules. MR analysis was further employed to validate the core modules in humans. The cell composition was also assessed for each module. By integrating the above results, the key functional genes related to motor were finally elucidated.
Table 1-1
Detailed description of the phenotype abbreviation. Download Table 1-1, XLSX file.
Construction of the striatal gene coexpression network
Functionally related genes theoretically work closely together in time and space to ensure orderly biosynthesis, which is similarly reflected in modules of the coexpression network. We applied WGCNA and grouped the striatum transcriptome (14,245 genes) into 13 modules, with the number of genes ranging from 85 to 5,389 in different modules (Extended Data Table 2-1). The gray module refers to a set of genes that are not fully categorized into other modules; hence, it was not included in the downstream analysis. Furthermore, module–trait correlation analysis (Fig. 2) revealed that the pink module was positively correlated with the motor phenotypes. The magenta module was negatively correlated with the PPI phenotypes and positively correlated with the motor phenotypes. Both brown and green–yellow modules were positively correlated with the motor phenotypes, while purple modules negatively correlate with multiple MA phenotypes (Fig. 2).
Module–trait correlation analysis. Heatmap of correlation of phenotypes with module eigengenes. Heatmap color and numbers indicate correlation coefficients and p values. At the bottom is the categorized of phenotypes. Only phenotypes with significant correlation with modules were shown, and all phenotypes–module associations were provided in Extended Data Figure 2-1.
Figure 2-1
Correlation analysis of modules with all phenotypes. Heatmap of correlation of phenotypes with module eigengenes. Heatmap color and numbers indicate correlation coefficients and p-values. At the bottom is the categorized of phenotypes. Download Figure 2-1, TIF file.
Functional analysis of the modular genes
We performed functional enrichment analyses to determine the biological functions of the coexpression networks (Fig. 3). The results showed green and blue modules to be enriched with genes involved in nucleotide metabolism and repair (Fig. 3) and Alzheimer's and Huntington's diseases (Extended Data Fig. 3-1B). These findings are in agreement with a study by Wang et al. (2022), which is also based on gene coexpression network construction using mouse striatum profiles. The pink module was found to be enriched with genes associated with the differentiation and growth of oligodendrocytes (Fig. 3). Interestingly, brown, turquoise, and magenta modules were enriched in learning and memory, drug addiction, behavior, and dopamine reward mechanisms, which are closely related to the function of the striatum (Fig. 3). It is worth mentioning that the purple module was significantly enriched in MPO for cocaine addiction and nicotine addiction (Extended Data Fig. 3-1A), which corresponds to the fact that the purple module correlates with the MA phenotype (Fig. 2).
Gene ontology enrichment analysis of the module genes. Different color bars represent different coexpression modules. The y-axis is bisected, with enrichment significance shown as −log10 (p-value) on the bottom and the number of genes enriched on the top.
Figure 3-1
Functional enrichment analysis of all module genes. (A) KEGG pathway, (B) Mammalian Phenotype Ontology. (C) Genetic ontology. Different color bars represent different co-expression modules. The x-axis is bisected, with enrichment significance shown as -log10 (p-value) on the left and the number of genes enriched on the right. Download Figure 3-1, TIF file.
Magenta and brown modules are closely related to striatal function
When certain diseases or phenotypic gene expressions significantly enrich within a module, genes within that module often possess important functionalities. To identify modules of coexpressed genes informative of GWAS and IMPC-related striatal phenotypes, we collected striatum-related phenotype genes from the GWAS Catalog (henceforth referred to as GWAS) and IMPC databases. The GWAS gene list yielded 1,245, 179, 332, and 419 genes for schizophrenia, memory, anxiety, and locomotor phenotypes, respectively. Similarly, 380, 324, and 736 genes related to memory, anxiety, and locomotion, respectively, were obtained from IMPC. We did an intersection of these disease-related gene sets and found no significant overlap between them. Particularly, the most critical set of motor-related genes accounted for only eight of the genes in both databases (Fig. 4A). The most intersected gene set was the anxiety and schizophrenia gene set from GWAS with 126 (Fig. 4A), followed by the anxiety and motor gene set from IMPC with 94 in both gene sets (Fig. 4A). They retain more than half of the unique genes that can be used for enrichment analysis.
Identification of the striatal core modules. Enrichment of coexpression modules in behavioral phenotype genes obtained from GWAS Catalog and IMPC databases. A, Upset plots showing the number of intersecting genes between different striatal-associated diseases from GWAS and IMPC and the number of their unique genes. B–E, Bubble plots showing the enrichment of each striatal-associated disease gene in the module, with the size of the dots representing the number of genes and the color representing the significance and with the x-axis representing Fisher's exact test of the odds ratio and the y-axis representing the modules.
Figure 4-1
Identification of the striatal core modules. Hypergeometric test enrichment of co-expression modules in behavioral phenotype genes obtained from (A-D) GWAS Catalog and (E-F) International Mouse Phenotyping Consortium (IMPC) databases. The modules above the dotted line [-log10(p-value) > 1.31] were considered significant. (H) Enrichment of the list of stroke genes from GWAS in the module (fisher’ exact test), the red line indicates the threshold of significant enrichment. Download Figure 4-1, TIF file.
Table 4-1
GWAS/IMPC gene lists. Download Table 4-1, XLSX file.
We determined whether any of the 12 modules were enriched for genes that overlapped with GWAS- and IMPC-retrieved gene lists. Our analysis revealed that three modules (green–yellow, brown, and magenta) were significantly enriched in the GWAS gene list (Fig. 4). The green–yellow module corresponded to anxiety, memory, and schizophrenia, whereas the brown module corresponded to anxiety, memory, and locomotory-related genes (Fig. 4B–E), and the magenta module corresponded to schizophrenia (Fig. 4E). However, only the magenta module was enriched in the IMPC-retrieved memory and locomotor genes (Fig. 4B–D). The enrichment results were significant with p < 0.05 based on Fisher's exact test. The significance of the above results was further verified using the hypergeometric test (Extended Data Fig. 4-1A–G). Together, gene enrichment and functional enrichment (Fig. 3; Extended Data Fig. 3-1) analyses identified brown and magenta as the core modules for striatal function. Meanwhile, we also found the genes implicated in stroke, a nonstriatal-related disorder, are not overrepresented in any modules, demonstrating the modules we constructed were only associated with striatum-related diseases (Extended Data Fig. 4-1H).
MR further validated the accuracy of the core modules
To evaluate the potentially causative effects of the genes in the core module on striatal function in humans, we obtained the summary statistics data of cis-eQTL corresponding to these module genes in the human BG brain region. Finally, in the brown module, 849 genes (65%) were retained and could be used as cis-eQTL summary statistics data for downstream analysis. Similarly, 125 genes (81%) were retained in the magenta module (Extended Data Table 5-1). We then performed MR analysis by the IVW method with cis-eQTL genes in modules as exposure and multiple striatal-related GWAS phenotypes as outcomes and found that the core modules we identified were significantly associated with extrapyramidal and movement disorders (p < 0.05; Fig. 5). No horizontal pleiotropy (MR-Egger, Q_p = 0.72; IVW, Q_p = 0.72) and heterogeneity (MR-Egger intercept, p = 0.33) were observed. Gene expression in the magenta module was protective against extrapyramidal movement disorders, while the opposite was true for the brown module. Extrapyramidal movement disorders include hypokinetic rigid and hyperkinetic or mixed forms, most of them originating from the dysfunction of the BG and their information circuits (Jellinger, 2019); the striatum is a key part of the BG (Graybiel, 1995; Groenewegen, 2003). These results provide evidence that regulation of gene expression within the core module may be critical to the motor regulatory function of the striatum. Meanwhile, MR analysis on other modules did not obtain significant results.
Results of Mendelian randomization analysis. Forest plots of Mendelian randomization analysis of the brown and magenta modules with multiple striatal-related phenotypes. The exposure instrument variable was derived from the basal ganglia, and the population used was European.
Table 5-1
Mendelian randomization analysis of exposure instrumental variables. Download Table 5-1, XLSX file.
Adcy5 is the hub gene of the brown module
Next, we analyzed the intramodular gene connectivity in the core module in detail and found that Adcy5 had the strongest connectivity in the brown module. Adcy5 and eigengene showed a strong positive correlation (r = 0.9760; p < 1 × 10−4; Fig. 6A). Furthermore, Adcy5 was positively correlated with “mean freezing ratio on Day 1 during the pretraining period (30–180 s) before exposure to auditory cues and shocks (PreTrainD1)” (Fig. 6B), “locomotor response observed during the 10–15 min interval on Day 2 (D2TOTDISD15)” (Fig. 6C), “reaction to the novel environment within the initial 0–15 s following MA injection (D3ctrtime0to15)” (Fig. 6D), and “reaction to the novel environment within the initial 0–30 s following MA injection (D3ctrtime0to30)” (Fig. 6E) phenotypes.
Adcy5 is the hub gene of brown module. A, Correlation between Adcy5 and the brown module eigengenes. B–E, Correlation between Adcy5 and traits. PreTrainD1, mean freezing ratio on Day 1 during the pretraining period (30–180 s) before exposure to auditory cues and shocks; D2TOTDISD15, locomotor response observed during the 10–15 min interval on Day 2; D3ctrtime0to15, reaction to the novel environment within the initial 0–15 s following MA injection; D3ctrtime0to30, reaction to the novel environment within the initial 0–30 s following MA injection. F, Interaction network of Adcy5 with GWAS and IMPC locomotor genes in the brown module. G, Interaction network of Adcy5 with GWAS and IMPC anxiety genes in the brown module. H, Interaction network of Adcy5 with GWAS memory genes in the brown module. I, Phenotypes resulting from Adcy5 knock-out model. The colors of the squares represent the phenotypes affected by the knock-out of the gene in mice, with blue indicating affected and white indicating unaffected. Zoom in to show the specific phenotypes affected by this phenotype.
In the brown module, 44, 48, and 25 GWAS genes, respectively, were implicated in anxiety, locomotor, and memory traits, whereas 28 and 55 IMPC genes were implicated in anxiety and locomotor traits. Notably, Adcy5 was associated with >93% of these genes (Fig. 6F–H). In addition, Adcy5 knock-outs were associated with decreased locomotor activity, impaired coordination, impaired behavioral response to xenobiotics, and decreased vertical activity. These phenotypes are closely related to the main function of the striatum (Fig. 6I).
Kcnma1 is the hub gene of the magenta module
Similarly, we looked at the magenta module genes. Kcnma1 had the highest connectivity in the magenta module, and it showed a strong positive correlation with the module eigengene (r = 0.9513; p < 1 × 10−4; Fig. 7A). Furthermore, Kcnma1 was positively correlated with multiple PPI phenotypes including “mean ratio of the prepulse response to the amplitude of the pulse-alone startle in the 6 dB prepulse test (pp6PPlavg),” “ratio of the prepulse response to the amplitude of the pulse-alone startle in the 12 dB prepulse test within Block 1 (pp12PPlb1),” “ratio of the prepulse response to the amplitude of the pulse-alone startle in the 6 dB prepulse test within Block 2 (pp6PPlb2),” “average across all amplitudes (PPlavg),” “mean ratio of the prepulse response to the amplitude of the pulse-alone startle in the 12 dB prepulse test (pp12PPlavg),” and “ratio of prepulse response to pulse-alone startle amplitude in the 12 dB prepulse test in Block 2 (pp12PPlb2)” (Fig. 7B), which are related to schizophrenia.
Kcnma1 is the hub gene of magenta module. A, Correlation between Kcnma1 expression and magenta module eigengenes. B, Correlation between Kcnma1 and the phenotype traits: pp6PPlavg, mean ratio of the prepulse response to the amplitude of the pulse-alone startle in the 6 dB prepulse test; pp12PPlb1, ratio of the prepulse response to the amplitude of the pulse-alone startle in the 12 dB prepulse test within Block 1; pp6PPlb2, ratio of the prepulse response to the amplitude of the pulse-alone startle in the 6 dB prepulse test within Block 2; PPlavg, average across all amplitudes; pp12PPlavg, mean ratio of the prepulse response to the amplitude of the pulse-alone startle in the 12 dB prepulse test; pp12PPlb2, ratio of prepulse response to pulse-alone startle amplitude in the 12 dB prepulse test in Block 2. C, Gene interaction network between Kcnma1 and memory IMPC and schizophrenia GWAS genes. D, Interaction network between Kcnma1 and the locomotor IMPC gene located in the magenta module. E, Phenotypes resulting from Kcnma1 knock out. The colors of the squares represent the phenotypes affected by the knock out of the gene in mice, with blue indicating affected and white indicating unaffected. Zoom in to show the specific phenotypes affected by this phenotype. F, Results of the analysis of the Kcnma1 human ortholog in GWAS.
In the magenta module, 11 and 14 memory and locomotor IMPC genes were implicated, 91 and 79% of which showed association with Kcnma1, respectively. Furthermore, 22 schizophrenia GWAS genes overlapped with the magenta module, and 91% of these were correlated with Kcnma1 (Fig. 7C,D). Kcnma1 knock-out has been shown to result in impaired coordination and abnormal locomotor behavior in mice (Fig. 7E). Kcnma1 was subsequently identified as a risk gene for schizophrenia, smoking, and educational attainment in the GWAS analysis (Fig. 7F).
Assessment of module cell-type composition
To assess the cell-type composition of each module, we retrieved the marker genes of 11 brain cell types in the adult mouse from Mouse Cell Atlas and performed intersection statistics with modular genes (Fig. 8). We found all modules contain astrocyte and oligodendrocyte progenitor cell marker genes. Notably, 48.1% of the genes in the pink module were oligodendrocyte marker genes, which was in line with our GO enrichment results, that is, genes in the pink module were significantly enriched in oligodendrocyte growth and differentiation (Fig. 3). Oligodendrocytes mainly wrap axons by producing myelin sheath, which provides nutrition, protection, and rapid signal transmission for axons (Kuhn et al., 2019). Moreover, the pink module was positively correlated with the motor phenotype (Fig. 2), suggesting oligodendrocytes may contribute to striatal motor regulation.
The percentage of brain cell marker genes in each module. The percentage of different cell types within each module is the percentage of marker genes for that particular cell type within the module, out of all genes in that module. For example, the percentage of astrocytes in the pink module is 48.1%. This is calculated by subtracting the percentage corresponding to the leftmost end of the x-axis from the percentage corresponding to the rightmost end of the x-axis in the pink module.
Table 8-1
Adult mouse brain cell marker genes. Download Table 8-1, XLSX file.
In the tan module, 16% of genes were microglial marker genes (Fig. 8). Microglia play a central role in immune response in the central nervous system (CNS; Fourgeaud et al., 2016; Ferrero et al., 2018). In fact, many risk genes of CNS-related disorders revealed by GWASs, including Alzheimer's disease, Parkinson's disease, schizophrenia, autism, and multiple sclerosis, are expressed in microglia (Prinz et al., 2019). The tan module also showed antigen processing and presentation and increased apoptosis of these immune pathways (Fig. 3; Extended Data Fig. 3-1).
The core modules that we identified (brown and magenta) contained multiple cell types (Fig. 8); however, the proportion of cell markers was not clearly distinguished in these modules.
Discussion
In this study, we first constructed a gene coexpression network using WGCNA and performed module–trait correlation analysis. We found that most modules correlated with motor and MA sensitivity phenotypes, while correlations with PPI phenotypes and conditioned fear phenotypes were weak. One possible explanation for this is that the striatum is primarily involved in motor regulation and dopamine reward mechanisms (Kreitzer and Malenka, 2008).
In the subsequent MR analysis, extrapyramidal and movement disorders were significant; nevertheless, this phenotype is directly related to the main functions of the striatum (Jellinger, 2019). Although both core modules were significant in the MR analysis, interestingly, their OR values were opposite (the magenta module, OR < 1; the brown module, OR > 1). The lack of movement in Parkinson's disease results from overactivity of the indirect pathway, whereas excess movement in disorders like Huntington's disease represents overactivity of the direct pathway (Graybiel, 1995). This suggests that one of our core modules may play an excitatory role in motor signals and the other may inhibit the movement. Furthermore, abnormal regulation of striatal motor signals may contribute to the development of neurodegenerative diseases.
Adcy5, as the brown module hub gene, encodes a protein that regulates adenylyl cyclase activity, adenylyl cyclase binding activity, and scaffold protein binding activity (Ferrini et al., 2021). Loss of adenylate cyclase has been shown to cause multiple dysfunctions in the striatum, including learning and movement (Kheirbek et al., 2009). This is consistent with our study showing that Adcy5 was associated with >93% of known motor and memory-related genes within the module. As reported previously, ADCY5-related diseases comprise a spectrum of hyperkinetic disorders involving chorea, myoclonus, and/or dystonia, often with paroxysmal exacerbations (Hisama et al., 1993; Ferrini et al., 2021). In the current study, Adcy5 was positively associated with multiple motor phenotypes, in addition to anxiety, and showed dyskinesia in mouse knock-outs (Casado-Sainz et al., 2022). We suggest that Adcy5 may have a critical role in the regulation of movement in the striatum of CFW mice, which may be equally applicable to humans.
We identified Kcnma1 as the hub gene for the magenta module. The protein encoded by Kcnma1 is a part of the voltage-gated potassium ion channel complex and is located in various cellular components, including the apical plasma membrane, the outer surface of the plasma membrane, and the postsynaptic membrane, which is an integral part of the membrane (Singh et al., 2013). In humans, KCNMA1 mutations are primarily defined by brain and muscle dysfunction (Bailey et al., 2019). Several KCNMA1 mutants have been identified and are more closely related to generalized epilepsy and paroxysmal dyskinesia (Yao et al., 2021). This is consistent with our findings that Kcnma1 was associated with multiple genes related to locomotory traits from IMPC within the magenta module. Notably, recent large cohort studies have shown that people with schizophrenia have seizures 4–5 times more frequently than the general population (Adachi and Ito, 2022). In our analysis, Kcnma1 was positively associated with multiple schizophrenia phenotypes (e.g., PPI phenotype) and was associated with most of the schizophrenia GWAS genes in the magenta module, suggesting that Kcnma1 mutations may further induce epilepsy or muscle dysfunction in addition to triggering schizophrenia. Therefore, Kcnma1 can be used as an important target for clinical treatment. In a study by Cadet et al. (2017), Kcnma1 mutation in mice was found to be associated with drug addiction, which is consistent with nicotine addiction (smoking initiation and cessation) in our GWAS analysis.
In module cell composition assessment, our study concurs with prior findings that certain brain gene coexpression modules are enriched in markers of cell class or cell types (Oldham et al., 2008; Hawrylycz et al., 2012). Interestingly, nearly half of the genes in the pink module were markers for oligodendrocytes. Notably, the tan module consisted of a total of 85 genes, 16% of which were microglial marker genes. Furthermore, functional enrichment analysis revealed a representation of immune pathways, which is consistent with previous functional studies related to microglia (Nayak et al., 2014). Both oligodendrocytes and microglia are known to be involved in the pathological development of multiple neurodegenerative diseases (Yeung et al., 2019; Guerrero and Sicotte, 2020; Wang et al., 2021). In addition, microglia are also associated with brain homeostasis (Kwon and Koh, 2020). Thus, tan and pink modules may be used as potential targets for treating neurodegenerative diseases. Nevertheless, the core modules that we identified were enriched for multiple cellular marker genes, including GABAergic, oligodendrocyte, and astrocyte. The execution of precise movements and action-based sequences requires the convergence of inputs from cortical and subcortical structures onto the striatum and the primary input nucleus to the BG (Gerfen and Surmeier, 2011; Jin and Costa, 2015). These inputs target GABAergic spiny projection neurons (SPNs), the predominant striatal neuronal class, as well as a menagerie of interneurons, in a topographically organized manner (Ebrahimi et al., 1992; Assous and Tepper, 2019), while oligodendrocytes and astrocytes may provide support for the efficient signaling of SPNs. These cellular interactions may be essential for the striatum to perform its major functions.
The complexity of striatal compartmentalization and its role in shaping behavior are only starting to become clear (Prager and Plotkin, 2019). While decades of research have led to an overall model of BG function where limbic and sensorimotor loops are compartmentalized to guide unique aspects of behavior (Prager and Plotkin, 2019), the genes that play critical functions in the striatum remain elusive. Our study identified the gene set that can basically explain the main function of the striatum, further providing two key genes that may play a major role in the striatum and may be associated with the development of a variety of diseases. The findings from our study will help in subsequent research on revealing a detailed picture of striatum function.
Footnotes
This research was funded by Taishan Scholars Construction Engineering, National Natural Science Foundation of China (32170989), Major Basic Research Project of Shandong Provincial Natural Science Foundation (ZR2019ZD27), Key Research and Development Program of Shandong Province (2023CXPTO12), Natural Science Foundation of Shandong Province (ZR2021MH141, ZR202212010032), Shandong Province Higher Educational Youth Innovation Science and Technology Program (2019KJE013), and Binzhou Medical University Research Start-up (50012305190).
The authors declare no competing financial interests.
- Correspondence should be addressed to Jia Mi at jia.mi{at}bzmc.edu.cn or Fuyi Xu at xufuyiphd{at}gmail.com.
