Regulatory miRNA–mRNA Networks in Parkinson’s Disease

Parkinson’s disease (PD) is the second-most common neurodegenerative disease, and its pathophysiology is associated with alpha-synuclein accumulation, oxidative stress, mitochondrial dysfunction, and neuroinflammation. MicroRNAs are small non-coding RNAs that regulate gene expression, and many previous studies have described their dysregulation in plasma, CSF, and in the brain of patients with PD. In this study, we aimed to provide a regulatory network analysis on differentially expressed miRNAs in the brain of patients with PD. Based on our systematic review with a focus on the substantia nigra and the putamen, we found 99 differentially expressed miRNAs in brain samples from patients with PD, which regulate 135 target genes. Five genes associated with neuronal survival (BCL2, CCND1, FOXO3, MYC, and SIRT1) were modulated by dysregulated miRNAs found in the substantia nigra and the putamen of patients with PD. The functional enrichment analysis found FoxO and PI3K-AKT signaling as pathways related to PD. In conclusion, our comprehensive analysis of brain-related miRNA–mRNA regulatory networks in PD showed that mechanisms involving neuronal survival signaling, such as cell cycle control and regulation of autophagy/apoptosis, may be crucial for the neurodegeneration of PD, being a promising way for novel disease-modifying therapies.


Introduction
Parkinson's disease (PD) is the second-most common neurodegenerative disease, affecting approximately 6 million individuals worldwide, with a growing incidence in the last few decades [1]. Furthermore, the disease reduces life expectancy and increases disability-adjusted life years, and apparently, these negative impacts have not been mitigated by the advance of new therapies [2]. Some authors compare the recent expansion of PD to a pandemic and suggest a substantial increase in the funding of new research on its pathophysiology [1].
Multiple mechanisms are associated with the pathophysiology of PD, such as the accumulation of α-synuclein, mitochondrial dysfunction, oxidative stress, calcium homeostasis, and neuroinflammation [3]. These epigenetic mechanisms are influenced by microRNAs (miRNAs), small non-coding RNAs that regulate gene expression at a posttranscriptional level by binding to their target messenger RNAs (mRNAs) [4]. Several studies analyzed the differentially expressed miRNAs in biological samples from patients with PD; however, the low sample size and high methodological heterogeneity compromise the interpretation of these combined results [5]. A recent meta-analysis on miRNAs in PD identified 13 miRNAs that are consistently differentially expressed in the blood and brain of patients with PD, such as hsa-miR-133b, hsa-miR-221-3p, and hsa-miR-214-3p [5].
For instance, it was demonstrated that hsa-miR-34b and hsa-miR-34c are downregulated in PD, reducing the levels of DJ-1 and Parkin in the brain, two proteins involved in the ubiquitin-proteasome system in neurons, causing cell death. Furthermore, hsa-miR-4639-5p is upregulated in PD and inhibits DJ-1, also promoting cell death [6]. The dysregulation of these miRNAs shows how these molecules can modulate the pathophysiology of PD.
For a better understanding of the role of these miRNAs in PD pathophysiology, regulatory miRNA-mRNA networks, followed by their topological analysis and functional enrichment of the hub genes, are important to provide a broad view of the PD-related biological processes and signaling pathways [7,8]. The objective of this study was to explore PD pathophysiology through regulatory network analyses based on differentially expressed miRNAs (DE-miRNAs) in the brain of patients with PD described in previous studies, with a special focus on substantia nigra and putamen. These data can be useful for proposing miRNA-based therapies capable of slowing disease progression [9].

Screening of Candidates Differentially Expressed Brain-Related miRNAs Based on a Systematic Review
To screen differentially expressed miRNAs (DE-miRNAs) in the brain of patients with PD, we conducted a systematic literature search on MEDLINE, EMBASE, and Web of Science (from inception to December 2020) using the following algorithms: MEDLINE-"Parkinson's disease" AND "microRNA" AND "brain"; EMBASE-("Parkinson disease"/exp OR "Parkinson disease") AND ("microrna"/exp OR "microRNA") AND ("brain"/exp OR "brain"); Web of Science-ALL = ("Parkinson's disease" AND "microRNA" AND "brain"). Reference lists of the studies included were checked to identify new studies missed in the primary search (cross-reference search).

Study Selection and Data Extraction
We aimed to select all original research studies describing DE-miRNAs in the brain of patients with PD. Two rounds of selection were performed. In the first round, titles and abstracts were screened and filtered following these exclusion criteria: (1) studies not conducted in patients with PD, (2) studies not conducted in human subjects, and (3) duplicate articles. In the second round, full texts were evaluated and excluded following other exclusion criteria: (1) review studies, (2) studies assessing different conditions from PD (such as atypical parkinsonism and dementia with Lewy bodies), (3) conference abstracts, and (4) full text not found. A single reviewer performed each selection round.
We extracted the following data: (1) the first author's name, (2) year of publication, (3) brain region, (4) sample size, sex, and age of the study population (patients and controls), (5) dysregulated DE-miRNAs associated with PD, and (6) DE-miRNAs up-or downregulation in PD.

Prediction of the Target Genes of the Differentially Expressed Brain-Related miRNAs
After that, we predicted the target genes of the DE-miRNAs using miRTargetLink (https://ccb-web.cs.uni-saarland.de/mirtargetlink/ accessed on 7 January 2021), a tool for automating miRNA-targeting gene analysis procedures [10], considering only the strong evidence type of experimental validation. To filter the target genes, we downloaded RNA-Seq data from GTEx (https://gtexportal.org/home/ accessed on 7 January 2021) and selected only those that presented median gene-level TPM > 1 in all brain tissues (amygdala, anterior cingulate cortex, caudate-basal ganglia, cerebellar hemisphere, cerebellum, cortex, frontal cortex, hippocampus, hypothalamus, nucleus accumbens-basal ganglia, putamen, spinal cord, and substantia nigra). The filtered target genes were used in the following analyses.

Regulatory Networks and Their Topology Analysis
Regulatory networks of miRNA-mRNA interactions were constructed and visualized using Cytoscape software version 3.8.0 (http://www.cytoscape.org/ (accessed on 7 January 2021) [11]. We analyzed the networks' centrality (degree, betweenness, and closeness) and identified the hub genes using the CytoNCA plugin [12] in Cytoscape [13]. Hub gene expressions in GTEx brain tissues were plotted in heatmaps using the pheatmap package in R (Version 1.2.5033).

Functional Enrichment Analysis
Functional enrichment analysis of the target genes was performed using clusterProfiler and org.Hs.eg.db packages in R (Version 1.2.5033) [14]. The enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were plotted using the clusterProfiler package in R (Version 1.2.5033).
Particularly, we analyzed samples from substantia nigra and putamen, which comprise the nigrostriatal pathway, a brain circuit with relevant importance to PD. From the substantia nigra samples, 38 DE-miRNAs were reported, while 14 DE-miRNAs were reported related to putamen (Table 2). Two DE-miRNAs, hsa-miR-34b and hsa-miR-95, were dysregulated in both substantia nigra and putamen.

Analysis of the Differentially Expressed Brain-Related miRNAs' Target Genes
Target genes prediction was performed using an experimentally validated microRNAtarget interactions database. The predicted targets were filtered according to the GTEx data by considering only those that presented median gene-level TPM > 1 in the brain. This approach resulted in 58 target genes for the upregulated brain-related miRNAs and 79 genes for the downregulated miRNAs (Table 3). Especially for the DE-miRNAs found in the substantia nigra and putamen, we found 22 and 18 target genes, respectively (Table 3). When comparing these results, we found some common target genes between the four sets ( Figure 1). For instance, we identified that three genes (CCND1, FOXO3, and SIRT1) are in common between all sets, while five genes are shared between substantia nigra and putamen (BCL2, CCND1, FOXO3, MYC, and SIRT1) (Table S1). Table 3.

Regulatory Networks and Their Topology Analysis
We constructed four miRNA-mRNA regulatory networks: (1)

Regulatory Networks and Their Topology Analysis
We constructed four miRNA-mRNA regulatory networks: (1)   As shown in Figure 3, some genes potentially have a central role in the regulatory networks, such as CCND1 and MYC. To better identify these hub genes, we analyzed the degree, betweenness, and closeness centrality of the nodes (Table 4). After identifying the hub genes of each network, we analyzed their expression across the brain regions using GTEx RNA-Seq data (Figure 4). Overall heatmaps evidence the high expression of PTEN and CCND1 in substantia nigra and putamen, respectively, suggesting the potential role of these genes in the brain. As shown in Figure 3, some genes potentially have a central role in the regulatory networks, such as CCND1 and MYC. To better identify these hub genes, we analyzed the degree, betweenness, and closeness centrality of the nodes (Table 4). After identifying the hub genes of each network, we analyzed their expression across the brain regions using GTEx RNA-Seq data (Figure 4). Overall heatmaps evidence the high expression of PTEN and CCND1 in substantia nigra and putamen, respectively, suggesting the potential role of these genes in the brain.

Discussion
Based on a systematic review, we found a total of 99 DE-miRNAs (including 60 upregulated miRNAs and 39 downregulated miRNAs) in brain samples from patients with PD compared to healthy controls. Among them, hsa-miR-144 is the only one found as both up-and downregulated in PD-there is some evidence showing that this miRNA modifies the expression of three genes associated with monogenic forms of PD (SNCA, PRKN, LRRK2) [27]. Cho et al. showed that an inverse correlation between hsa-miR-205 and LRRK2 in PD was previously described, with high LRRK2 protein expression and low hsa-miR-205 levels in the frontal cortex of patients with PD, probably due to the 3 -UTR region of LRRK2 being an hsa-miR-205 target site [18]. A former study showed that hsa-miR-7, which was downregulated in the substantia nigra according to our review, is a direct regulator of SNCA, reducing its expression in a cell model and in an MPTP PD murine model [34]. Considering the miRNAs associated with both substantia nigra and putamen, we found that hsa-miR-34b and hsa-miR-95-hsa-miR-34b are associated with a reduction in the expression of alpha-synuclein [35], DJ-1, and Parkin [6], while hsa-miR-95 regulates the lysosomal function through the enzyme sulfatase-modifying factor 1 [36], and it was downregulated in pregnant women with multiple sclerosis [37]. To compare with another prevalent neurodegenerative disease, miRNAs such as hsa-miR-132 and hsa-miR-339-5p could be found in the brain of both PD and Alzheimer's disease patients [38].
Together, the DE-miRNAs regulate 135 genes. From these, five genes are regulated simultaneously by the dysregulated sets of miRNAs found in the substantia nigra and the putamen of patients with PD (BCL2, CCND1, FOXO3, MYC, and SIRT1) ( Table 4 and Figure 3). These genes have central roles in the miRNA-mRNA regulatory networks, and some of them have high expression in the brain, particularly in the substantia nigra and the putamen (Figures 3 and 4).
Cyclin D1 (CCND1) is a regulator of the cell cycle progression mediated by extracellular stimulation, and its overexpression results in neoplastic growth [39], or apoptotic-related cell death in postmitotic neurons [40]. The re-expression of cyclins and cyclin-dependent kinases in neurons from patients with Alzheimer's disease suggests that the failure of cell cycle arrest in adults may be associated with late-onset neurodegenerative diseases [40]. In PD, there is an overexpression of mitotic-associated proteins, such as cyclins and cyclindependent kinases, in the substantia nigra of postmortem patients with PD and an MPTP mouse model of PD, resulting in apoptosis of dopaminergic neurons [41,42]. Recently, some cell cycle genes were found enriched in a cell model of PD, and CCND1 was reported as upregulated and involved in alpha-synuclein cell death. It was shown that the knockdown of CCND1 reduces cell death [43], reinforcing our results of upregulation of miRNAs that regulate CCND1 in the brain.
Forkhead box protein O3 (FOXO3), comprising the Forkhead family, is a transcription factor associated with longevity in humans, and it is expressed in dopaminergic neurons of the substantia nigra. In a rat model of PD, FOXO3 was essential in the neuronal survival of the substantia nigra, and it may also reduce alpha-synuclein accumulation and its toxicity [44]. Also extending longevity, the silence information regulator 1 (SIRT1) is a member of the sirtuin family, which regulates DNA stability and controls gene expression and cell cycle progression. Enzymatic activity of SIRT1 is reduced in the temporal and frontal cortex of patients with PD [45], playing a critical role in the pathophysiology of PD through induction of autophagy, regulation of mitochondrial function, inhibition of neuroinflammation, and increasing degradation of alpha-synuclein [46].
MYC (or c-myc) is a transcription factor that regulates cell growth, division, differentiation, and death, and despite having a classic role in brain cancer progression and brain development, MYC expression is increased in neurodegenerative diseases, such as Alzheimer's disease, and like CCND1, its role is based on cell cycle control [47]. BCL2 is a suppressor of autophagy and apoptotic cell death, and its expression is decreased in cell models of PD [48].
Despite being mostly related to cancer, two of the pathways associated with PD are closely associated with neuronal survival and neurodegenerative diseases: FoxO and PI3K-AKT signaling pathways. The Forkhead box class O (FoxO) family of transcription factors has an essential role in multiple cellular processes in the nervous system, such as neural development and neuronal survival, promoting a proapoptotic effect [49]; otherwise, the PI3K-AKT pathway is associated with neuroprotection and is a major regulator of the FoxO pathway, inhibiting FoxO-induced neuronal death [49].
Previous studies have explored pathways involved in PD pathogenesis. Song et al. [50] found 21 different pathways associated with PD, based on GWAS datasets. In another study, data from Gene Expression Omnibus from patients with PD were used to perform regulatory network and functional and enrichment analysis, showing that distinct pathways, such as amoebiasis and MAPK signaling, might be related to PD [51]. More recently, another study also used a dataset from Gene Expression Omnibus and revealed new 12 pathways associated with PD [52].
These results suggest that, in PD, the expression of genes involved in cell survival is dysregulated by miRNAs. Therefore, besides alpha-synuclein accumulation, oxidative stress, and neuroinflammation, the neurodegeneration of PD may include competing mechanisms over neuronal survival, such as cell cycle control and regulation of autophagy/apoptosis, particularly in the substantia nigra. Neuronal survival signaling may become the target of new disease-modifying treatments for PD, including the use of miRNA-based therapies [9].

Conclusions
In conclusion, our analysis of miRNAs associated with PD based on a systematic review showed a multitude of differentially expressed genes in the brain of these patients, especially in the substantia nigra. This expression dysregulation is linked to several pathways, including neuronal survival signaling. The role of these genes and pathways must be explored in further studies and can be used by future studies on miRNAbased therapies.