Identification of key genes and signaling pathways associated with dementia with Lewy bodies and Parkinson's disease dementia using bioinformatics

Objective Dementia with Lewy bodies (DLB) and Parkinson's disease dementia (PDD) are collectively known as Lewy body dementia (LBD). Considering the heterogeneous nature of LBD and the different constellations of symptoms with which patients can present, the exact molecular mechanism underlying the differences between these two isoforms is still unknown. Therefore, this study aimed to explore the biomarkers and potential mechanisms that distinguish between PDD and DLB. Methods The mRNA expression profile dataset of GSE150696 was acquired from the Gene Expression Omnibus (GEO) database. Differentially expressed genes (DEGs) between 12 DLB and 12 PDD were identified from Brodmann area 9 of human postmortem brains using GEO2R. A series of bioinformatics methods were applied to identify the potential signaling pathways involved, and a protein–protein interaction (PPI) network was constructed. Weighted gene co-expression network analysis (WGCNA) was used to further investigate the relationship between gene co-expression and different LBD subtypes. Hub genes that are strongly associated with PDD and DLB were obtained from the intersection of DEGs and selected modules by WGCNA. Results A total of 1,864 DEGs between PDD and DLB were filtered by the online analysis tool GEO2R. We found that the most significant GO- and KEGG-enriched terms are involved in the establishment of the vesicle localization and pathways of neurodegeneration-multiple diseases. Glycerolipid metabolism and viral myocarditis were enriched in the PDD group. A B-cell receptor signaling pathway and one carbon pool by folate correlated with DLB in the results obtained from the GSEA. We found several clusters of co-expressed genes which we designated by colors in our WGCNA analysis. Furthermore, we identified seven upregulated genes, namely, SNAP25, GRIN2A, GABRG2, GABRA1, GRIA1, SLC17A6, and SYN1, which are significantly correlated with PDD. Conclusion The seven hub genes and the signaling pathways we identified may be involved in the heterogeneous pathogenesis of PDD and DLB.


. Introduction
Lewy body dementia (LBD) is the second most prevalent form of neurodegenerative dementia after Alzheimer's disease (AD) in patients older than 65 years (1). Parkinson's disease dementia (PDD) and dementia with Lewy bodies (DLB), collectively known as LBD, are synucleinopathies morphologically characterized by neuronal loss, inclusions containing Lewy body/α-synuclein and βamyloid, and tau pathologies, often reported as part of the same spectrum (2,3). Cognitive decline in the LBD may, in part, be due to a general loss of synapses and related functional failure (4,5). There is an average of 30-40% loss of synapses in the frontal and the temporal cortex in DLB (6), and in PDD, a reduction of the synaptophysin immunoreactivity of the cortical neuropil was 8.2% (7). Synaptic functional failure happens in the early stages of synucleinopathies due to altered transport of vesicles, synaptic proteins, and mitochondria, which lead to presynaptic terminal loss, dendritic damage, axonal dystrophy, and eventually degeneration of selective neuronal populations within the striatonigral and cortico-limbic systems, among others (6). Clinical distinctions between the two refer to the "so-called 1-year rule" (1,8,9), that is, the term DLB is used if dementia occurs before or concurrently with parkinsonism or within 1 year of onset of the motor symptoms; PDD describes dementia starting 1 year or more after Parkinson's disease (PD) becomes well-established (1). This mode of distinction is clearly arbitrary and based on the distinction between the time of onset of cognitive and motor symptoms (3). The mechanisms underlying these differences in clinical manifestations are unclear, and it is necessary to explore the mechanisms by which differences occur to differentiate from the early stages of the disease or even from differences at the genetic or molecular level, and hopefully to provide targeted treatments for these differences. Therefore, it is important to further study the differences in the pathogenesis between the two dementias (10). To this end, we used bioinformatics methods to delve deep into the mechanisms of their heterogeneity.
. Materials and methods . . Data source Gene expression datasets were obtained from the GEO database. After a careful review of the datasets, we chose the series of mRNA expression profile datasets of GSE150696 (https:// www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE150696) (11). Consensus criteria used for clinical diagnoses of PDD and DLB with neuropathologic confirmation have been previously described in detail (12). The samples processed in each group were matched for age, sex, and postmortem interval. The Brodmann area 9 from human postmortem brains was chosen for analysis. A total of 12 (6 women/6 men) PDD samples and 12 (6 women/6 men) DLB samples were retrieved from GSE150696 and published on 24 May 2021. All brain samples were provided by the Brains for Dementia Research, UK. Data were freely available online, and our study did not involve any experiments in the lab performed by any of the authors.

. . Data processing of di erentially expressed genes
The GEO2R online analysis tool (https://www.ncbi.nlm.nih. gov/geo/geo2r/) was used to detect differentially expressed genes (DEGs) between PDD and DLB samples, and the P-value and |log (FC)| (FC-fold change) were calculated. Genes that met the cutoff criteria, P < 0.05 and |log FC| ≥ 1.0, were considered DEGs (11). Genes with P < 0.05 and log FC ≥ 1.0 were considered upregulated genes, and genes with P < 0.05 and log FC ≤ −1.0 were considered downregulated genes (11). GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA; www.graphpad.com), graphing software that can perform data analysis and data visualization, was used to visualize volcano maps of all identified DEGs and a heat map of the top 50 genes (11).

. . GO and KEGG pathway analysis
The R software (version 4.2.1) was used for the GO annotation, the KEGG pathway enrichment analysis, and the visualization of DEGs (13). An online analysis tool Metascape website (http:// metascape.org) was used for GO and KEGG analyses of gene modules selected by weighted gene co-expression network analysis (WGCNA) (14).

. . Gene set enrichment analysis
Gene Set Enrichment Analysis (GSEA) is a promising and widely used software package (15) that derives gene sets to find the different biological functions of the whole genes between PDD and DLB. The potential contribution of the whole altered genes to LBD was explored using the GSEA software (version 4.2.3). A normalized enrichment score (NES) was calculated, and NES is the enrichment score for the gene set after it has been normalized across analyzed gene sets. The gene set was deemed to be significantly enriched when the P-value was <5% and |NES| was >1 for each analysis (16).

. . Weighted gene co-expression network analysis
Weighted gene co-expression network analysis (WGCNA) can cluster genes with higher co-expression levels, assemble them into modules, and establish connections between their modules and phenotypes to find the hub genes of the phenotype. We selected a WGCNA package of the R software to filter the top 6,000 median absolute deviation genes to construct a representation matrix and the scale-free network (17). The β-value was selected as long as R 2 was >0.8. The β-value was a soft threshold. The algorithm introduces an approximate scale-free topology to accurately calculate the soft threshold and then replaces the hard threshold of the previous traditional algorithm; Scalefree topologies are more realistic when compared with random networks (18). Based on the selected soft threshold, network .
/fneur. . modules were constructed by clustering the gene topology matrix using the dynamic shear tree algorithm. The minimum number of genes included in the network module was set to 20. The module color was established by using the degree of dissimilarity automatically by WGCNA software (18). The relationships between modules and LBD haplotypes are shown with a heatmap. We measured the module membership (mM) and gene significance (GS) of individual genes, and the hub genes in the selected modules with |mM| >0.8 and |GS| >0.2 were screened for further analysis (19,20).

. . Protein-protein interaction network construction and hub gene identification
We used the online Search Tool for the Retrieval of Interacting Genes database tool (STRING-DB) (http:// string-db.org/) to analyze protein-protein interaction (PPI) information (21). PPI pairs were extracted with a combined score of >0.4, and the results were calculated with their automatically cited parameters. Subsequently, the PPI network was visualized using the CytoHubba plug-in in Cytoscape software (version 3.9.1; https://www.cytoscape.org/) was used to calculate the degree of each protein node (11). We considered the top 30 identified genes as the hub genes in our study. Finally, we screened the hub genes by intersecting them in the selected modules, mentioned earlier, and in the DEG-PPI network. .

. . Identification of DEGs
The samples of PDD and DLB obtained from Brodmann area 9 of postmortem brains were selected for the present study from the GSE150696 series. On the basis of the criteria of P < 0.05 and |log FC| ≥ 1.0, 1,864 DEGs between PDD and DLB were filtered by the online analysis tool GEO2R. It included 1240 upregulated and 624 downregulated DEGs in patients with PDD (Supplementary Table S1). A volcano map of all identified DEGs is shown in Figure 1A. In addition, a heat map of the top 50 DEGs is shown in Figure 1B.

. . Functional and pathway enrichment . . . Analyses for DEGs
The GO and KEGG enrichment analyses were performed on 1,864 DEGs between PDD and DLB, and the findings were visualized with the cluster profiler package of R software (22). The mainly enriched biological process of GO analysis included the establishment of vesicle localization, vesicle-mediated transport in the synapse, and learning and memory. The cellular component of GO analysis included the presynapse axon part and the glutamatergic synapse. The molecular function of GO analysis included ATPase activity and motor activity (

. . . Gene set enrichment analysis
The GSEA analysis was used to filter unique pathways involved in the pathogenesis of PDD or DLB. When the green line plot was in the negative direction, the gene on the right side of the maximum enrichment score value was the core gene, the pathway was positively correlated with the PDD group, contrarily, the pathway was positively correlated with the DLB group. As shown in Figures 3A-D, the pathways of glycerolipid metabolism and viral myocarditis were positively correlated with the PDD group. The pathways of the B-cell receptor signaling pathway and one carbon pool by folate signaling pathways were positively correlated with the DLB group.

. . Weighted gene co-expression network analysis
We used WGCNA software to identify the associations between the key gene modules related to PDD and DLB. As shown in Figures 4A, B, the power was set as 6 for further analysis and satisfied the scale-free co-expression network relationships, with the mean value of the adjacency function gradually approaching 0. According to the module-trait relationships, eight modules were identified by the average linkage hierarchical clustering method from the co-expression network, and the colors were defined by the software automatically ( Figure 4C). Based on the correlation between different modules and subtypes of LBD shown in the heatmap, we found the green module was significantly positively associated with the PDD group (cor = 0.75, P-value < 0.01); the yellow module was significantly positively associated with the DLB group (cor = 0.75, P-value < 0.01) ( Figure 4D); and the gray module represented genes that were not assigned to each network. Other modules such as black, turquoise, red, brown, and blue also suggest a clear correlation.

. . Enrichment analyses of module genes identified by WGCNA
We used the Metascape tool to perform GO annotation and KEGG pathway enrichment analyses to analyze the features of the module genes. The number of genes within the blue and yellow modules was 2,530 and 182, respectively. As shown in Figures

. . Identification of hub genes
In the present study, because we found there was an ideal overlap in the blue module with the top 30 hub DEGs, the yellow module was significantly positively associated with the DLB group, and we finally chose blue and yellow modules for further study. The scattered plots of blue and yellow modules (Supplementary Figures S4A, B) present significantly positive correlations (P < 0.01) between PDD and DLB. We identified 815 and 3 hub genes in the blue and yellow modules, respectively (Supplementary Table S2). Protein interactions among the 1,864 DEGs were predicted using the STRING-DB tool. The top 30 hub genes in DEGs were evaluated by the maximal clique centrality method with the Cytohubba plugin. The network of the top 30 hub genes and expanded DEGs included 234 nodes and 878 edges, visualized by Cytoscape software ( Figure 6A). We intersected the hub genes of the blue module and the top 30 hub genes in DEGs ( Figures 6B, C), and we identified seven overlapping hub genes. There was no overlap between the top 30 hub genes in DEGs and the hub genes of the yellow module. As shown in Table 1, the information about these seven overlapping hub genes included in the blue module (SNAP25, 163 GRIN2A, GABRG2, GABRA1, GRIA1, SLC17A6, and SYN1) is mentioned in more detail.

. Discussion
The significantly enriched entries for GO and KEGG enrichment analyses demonstrated that the 1,864 DEGs  mainly enriched the functions of vesicles and synapses in neurodegenerative diseases. This study supports the mechanistic role of disturbed vesicle trafficking in neurodegenerative diseases (23). Positron emission computed tomography-based study suggests that the loss of synaptic density contributes to dysfunction and cognitive decline in patients with LBD (4). Presynaptic and postsynaptic proteins modulate axonal/dendritic growth and remodeling, thus representing likely key players in the synaptic dysfunction in neurodegenerative diseases (5). Synaptic disruption is a key pathophysiological mechanism leading to neurodegeneration (24). DLB and PDD differ in terms of not only the time of onset of cognitive deficits but also variability in affected functions (2). Patients with DLB present more severe and widespread cognitive dysfunction than those with PDD, Frontiers in Neurology frontiersin.org . /fneur. . particularly in attentive and visuospatial domains, executive functions, constructional tasks (25,26), and episodic verbal memory (27). Hence, the Mini-Mental State Examination score is lower in patients with DLB than in those with PDD (28). The percentage of patients with DLB who fail to finish the Montreal Cognitive Assessment subitem analysis on the Digit Span Forward was higher than that of patients with PDD, possibly because the former is associated with a more severe attentive deficit than the latter (2). This may explain the enrichment of DEGs. Due to acting in contrast to each other, DLB and PDD have opposite correlations to enriched pathways as shown in Figure 3. The GSEA data suggested that glycerolipid metabolism and viral myocarditis were positively correlated with PDD. Lipid metabolic dysregulation is involved in the pathogenesis of PDD; α-synuclein may induce dementia in patients with PD possibly through lipid metabolism (29). Coxsackievirus B3 is considered the dominant etiological agent of viral myocarditis. Coxsackievirus B3 infection can induce α-synuclein-associated inclusion body formation in neurons, which might act as a trigger for PD. Transgenic mice that express α-synuclein showed enhanced Coxsackievirus B3 replication and exhibited dopaminergic neuronal death in the substantia nigra (30). A B-cell receptor signaling pathway and one carbon pool by folate signaling pathways were positively correlated with the DLB subset. Accumulating evidence suggests the involvement of immune mediators in DLB (31, 32). Immunization of mice with different B-cell epitopes of human α-synuclein vaccines produced high titers of anti-human α-synuclein antibodies that bound to Lewy bodies and Lewy neurites in the brain tissue of patients with DLB and induced robust helper T-cell expression. Immunotherapeutic approaches that reduce α-synuclein deposits may provide therapeutic benefits for patients with DLB (33). Folate is an essential factor involved in nucleotide synthesis, one-carbon metabolism, and DNA methylation, which have been linked to cognitive impairment and dementia (34). Homocysteine is a central metabolite formed as an intermediate product of one-carbon metabolism, following transmethylation (35). Elevated plasma total homocysteine levels were independently associated with DLB (34).
To further investigate the relationship between co-expressed genes and different LBD subtypes, we performed WGCNA. The results of enrichment for the blue module included axon, postsynapse, presynapse, and neuron projection development. Axonal and synaptic pathology is an important feature of LBD (36). During the early prodromal phase of PD, synaptic alterations happen before cell death, and these alterations are linked to the synaptic accumulation of toxic α-synuclein, specifically in the presynaptic terminals, which affects neurotransmitter release (37).
. /fneur. . Generalized synaptic degeneration and loss of synaptic density and connectivity may contribute to dysfunction and cognitive decline in patients with neurodegenerative diseases (4,5,38). Reduced expression of synaptic proteins could be an index of the degree of synaptic degeneration in the central nervous system (38). Synaptic proteins reliably discriminated PDD and DLB from controls with high sensitivity and specificity (39). The particular synaptic proteins have an important predictive and discriminative molecular fingerprint in neurodegenerative diseases and could be a potential target for early disease intervention (39). The results of the enrichment for the yellow module included positive regulation of macrophage activation and integrin binding. α-synuclein expressed in neurons is released into the extracellular space and taken up by macrophages and microglia; α-synuclein fibrils are considered to be formed from monomers in macrophages and to spread to neurons to induce α-synuclein aggregation in PD model (40). Our data support the difference among axonal and synaptic, inflammation, and neurodegeneration-multiple diseases.
After taking the intersection of the top 30 hub genes of DEGs and the hub genes in blue modules, seven key genes were identified, including SNAP25, GRIN2A, GABRG2, GABRA1, GRIA1, SLC17A6, and SYN1. SNAP25 and SYN1 as presynaptic proteins are markers of functional synapses (38,41,42). SYN1 is a phosphoprotein that coats the cytoplasmic side of synaptic vesicles and regulates their trafficking within nerve terminals. Inhibition or knockout of SYN1 can reduce the density of excitatory and inhibitory synapses and impair both glutamatergic and GABAergic synaptic transmission (43). SNAP25 is a key adhesion molecule for vesicle docking, trafficking, fusion of membranes, and exocytosis, and it has also been implicated in axonal outgrowth and neurite elongation (5). It has been suggested that SNAP25 could be an effective and accessible biomarker reflecting synaptic integrity and degeneration in the brain (38,44). In previous studies, SNAP25 levels were negatively correlated with cognitive functions (38), and SNAP25 expression was low in patients with PDD and DLB; however, the decrease was more pronounced in the DLB patient group (45), which is consistent with our study so that SNAP25 is a known gene that is more relevant to LBD (44). The GRIN2A gene encodes a member of the glutamate-gated ion channel protein family. GRIN2A was found to play important roles not only in synaptic plasticity but also in learning and memory (46), and spatial or discrimination learning impairments have been observed in mice with GRIN2A subunit knockout (47,48). Studies have also shown that the suppression of GRIN2A expression impairs .
/fneur. . Seven key genes were obtained from the intersection of 30 identified genes in 1,864 differentially expressed genes and 815 hub genes in the blue module. The seven key genes were upregulated in the Parkinson's disease dementia group.
the learning of complex motor skills (49). Dendritic branch pruning along with maturation is accompanied by an elevation in GRIN2A levels (50). GRIN2A deletion was shown to decrease the total dendritic length and dendritic complexity in the dentate gyrus neurons of the hippocampus located in the inner granular zone (51). GRIA1 is a subunit in the α-amino-3-hydroxy-5methylisoxazole-4-propionic acid subtype of ionotropic glutamate receptors, which is a primary receptor that mediates excitatory synaptic transmission at glutamatergic synapses in the central nervous system and plays key roles in synaptic plasticity, neuronal development, and neurological diseases (52). The α-amino-3hydroxy-5-methylisoxazole-4-propionic acid subtype of ionotropic glutamate receptors mediates most of the fast postsynaptic responses at glutamatergic synapses (53). Synaptic plasticity relies on the normal integration of glutamate receptors at the postsynaptic density (54). The increased translation of GRIA1 facilitates certain forms of hippocampus-dependent synaptic plasticity and memory (55)(56)(57). GO biological process enrichment analysis showed that SNAP25, GRIN2A, GRIA1, and SYN1 were significantly enriched in learning and memory in our study, further suggesting that they participate in the neurobiological basis of pathogenesis by affecting synapse function. KEGG enrichment analysis of DEGs showed that SNAP25 and SLC17A6 were enriched in the synaptic vesicle cycle in our study. SLC17A6 is responsible for the uploading of glutamate into presynaptic vesicles, while SLC17A6 is utilized by a majority of cortical and subcortical glutamatergic excitatory neurons (58, 59). Kashani et al. observed that SLC17A6 downregulation was correlated with the degree of cognitive impairment in Brodmann area 9 in patients with Alzheimer's disease (60); the lower the decline in SLC17A6 expression, the greater the degree of cognitive impairment. Studies suggest that patients with DLB present more severe and widespread cognitive dysfunction than those with PDD (25)(26)(27). Consistent with the conclusions of our study, GRIA1, GRIN2A, and SLC17A6 expressions were more upregulated in patients with PDD than in those with DLB, further genetically explaining the cognitive dysfunction is graver in patients with DLB than in those with PDD. The GABRA1 and GABRG2 genes encode subunits of the γ-aminobutyric acid (GABA) type A receptor family (61). GABRA1 was found to be significantly more downregulated in the postmortem frontal cortices of patients with DLB than in those with neuropathological examination normal control (62). Similarly, GABRG2 expression was found to be significantly low in symptomatic mouse models of tauopathy (63). Both play an important role in the maintenance of normal synaptic function. RNA-sequencing analysis of mutation of GABRA1 in zebrafish larval brains identified a marked downregulation of genes encoding inhibitory synaptic components as well as proteins involved in axon guidance. Immunocytochemical analysis revealed a marked decrease in the accumulation of GABA synaptic markers; consistently, transgene GABRG2 mutation resulted in postsynaptic and presynaptic defects (64, 65). GABRG2 was found to be associated with suicidal behavior and major depressive disorder (66). Studies indicated that low brain levels of GABA may be related to schizophrenia and psychosis (67-69). Fluctuations in core clinical features of DLB are typically delirium-like, occurring as spontaneous alterations in cognition, attention, and arousal (70), which is different from PDD; similarly, in our study, GABRA1 and GABRG2 genes were downregulated in DLB.
These findings suggest the potential roles of SNAP25, GRIN2A, GABRG2, GABRA1, GRIA1, SLC17A6, and SYN1 as biomarkers to distinguish PDD from DLB. The function of the seven hub genes participated in the neurobiological basis of pathogenesis by affecting synaptic function and the GABAergic/glutamatergic neurotransmitters. GO enrichment analysis showed that SNAP25 was the core gene participating in the neurobiological basis of pathogenesis by affecting synapse function in our study; KEGG enrichment analysis of DEGs showed that SNAP25 was enriched in the synaptic vesicle cycle. SNAP25 may be a more significant gene distinguishing between PDD and DLB by the affected synaptic vesicle cycle. In addition, GRIN2A, GABRG2, GABRA1, GRIA1, and SYN1 are involved in regulated GABAergic/glutamatergic functions. Hence, synaptic transmission impairment and GABAergic/glutamatergic dysfunction may be the more outstanding difference between PDD and DLB.

. Supplementary content
We also constructed sets of gene maps based on the online resource OMIM.org (https://omim.org/) (71) associated with DLB and PDD, compared the two lists of gene maps with the top 30 hub genes in DEGs and hub genes in the blue module, respectively, the results suggested that the three lists had no intersection. There were 15 (GDAP1, ATP1A1, DNAJC6, SNCA, TUBA4A, VCP, GYG1, TRIM2, BMPR2, NEFL, C9orf72, DHX16, FIG4, and COPA) overlap hub genes with the blue module and gene map in .

Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

Ethics statement
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the patients/participants or patients/participants' legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

Author contributions
J-jC was the senior author of the report. JX made the material preparation and the first draft of the manuscript. All authors contributed to the study's conception, design, read, and approved the final manuscript.

Funding
This study was supported by the Jilin Science and Technology Department Project (nos. 20200602045ZP and 20200201451JC).