Dysregulation of miR‐543 in Parkinson's disease: Impact on the neuroprotective gene SIRT1

Parkinson's disease (PD) is a progressive and age‐dependent neurodegenerative disease characterised clinically by a variety of motor symptoms and cognitive impairment. PD was initially considered to be a grey matter disease; however, recently, evidence has emerged that white matter changes in PD precede the neuronal loss seen in the grey matter. The cause of these initial white matter changes is yet to be elucidated. Here, we explored whether dysregulated miRNAs and their target mRNA could provide insight into the underlying mechanisms of early white matter changes in PD.


INTRODUCTION
Parkinson's disease (PD) is a progressive and age-dependent neurodegenerative disease [1]. PD is a disease that steadily increases in incidence as individuals age. Among women aged 60 to 69, this incidence Lucia Mesarosova, James D. Mills and Eleonora Aronica Shared last authorship. is 30 out of 100,000 and for men, it stands at 58 out of 100,000. For those over 80, the incidence among women is 80 out of 100,000 and for men, 258 out of 100,000 [2]. PD manifests with both motor and non-motor symptoms, with the former being the most common. The motor symptoms include resting tremor, bradykinesia, postural instability and rigidity [3]. Furthermore, non-motor symptoms have become more acknowledged over time as they can be similarly debilitating as the motor symptoms. Non-motor symptoms in PD include cognitive decline, depression, anxiety, dysautonomia and sleep disturbances [4]. With this broad spectrum of symptoms, PD has wide-ranging social and economic impacts resulting in a decrease in the quality of life of patients and their surroundings.
The main neuropathological hallmarks of PD include increasing accumulation and aggregation of alpha-synuclein (α-Syn) protein and the loss of the nigrostriatal dopaminergic neurons [5]. The resulting motor abnormalities mark the progression of PD, whereas non-motor symptoms can already be identified in the prodromal stages of the disease. Due to this characterisation, PD was originally classified as a grey matter disorder. However, recent studies have introduced the idea that white matter changes occur prior to the degenerative loss of neuronal cell bodies [6]. Axon degeneration and alterations in axonal transport are considered to be the earliest and most predominant features of PD [7]. Further evidence of early white matter changes comes from imaging studies implying that white matter integrity is affected in PD patients. These studies have shown that microstructural changes in the white matter could reflect the loss of white matter because of demyelination and axonal damage [8,9]. The cause of these initial white matter changes is yet to be elucidated; however, initial axonal damage may be attributed to the synaptic accumulation of toxic α-Syn.
Recently, there has been an increasing number of studies showing that microRNAs (miRNAs) play an important role in posttranscriptional regulation of genes that are associated with the pathogenesis of several diseases [10]. miRNAs are small non-coding RNAs that interact with 3 0 untranslated regions (UTRs), 5 0 UTRs, gene promoter regions and coding sequences of their target mRNAs, resulting in degradation and regulation of transcription and translation [11][12][13].
In PD, multiple miRNAs are differentially expressed compared to healthy controls. Dysregulated miRNAs in PD are related to the regulation of PD-associated genes such as synuclein alpha (SNCA), parkin RBR E3 ubiquitin protein ligase (PRKN), PTEN Induced Kinase 1 (PINK1) and more. Furthermore, miRNAs related to neuroinflammation and dopaminergic neuron survival have also been found to be dysregulated in PD [14]. Hence, disturbance of miRNA expression could potentially contribute to the pathogenesis of PD through modulation of PD-associated gene and protein expression. Previous studies have identified several differentially expressed miRNAs in PD brain tissue; however, their role in disease pathogenesis has not been fully elucidated [15,16].
Thus, we hypothesise that dysregulated miRNAs may be involved in the establishment and progression of PD through the targeting and regulation of PD-associated genes. Therefore, this study set out to determine differentially expressed miRNAs in the middle frontal gyrus (MFG) of PD patients. Tissue was selected from the MFG as it is affected in the later stages of PD, predominately in Braak stages 5 and 6, allowing the determination of the molecular events prior to the serious cellular loss in PD pathology. Next, we looked to identify and functionally characterise promising miRNA targets. Moreover, we also assessed the feasibility of using miRNAs as cerebral spinal fluid (CSF)-based biomarkers for PD.

Cohort
The middle frontal gyrus was selected and obtained from the Netherlands Brain Bank (NBB), Netherlands Institute for Neuroscience, Amsterdam. The material was collected from donors with written informed consent for the use of the material and clinical information for research purposes. For all samples (Table 1), brain tissue was either frozen, separated into grey and white matter using the cryostat and kept at À80 (for RNA isolation) or fixed in 4% paraformaldehyde and embedded in paraffin (FFPE, for staining experiments). Moreover, the CSF of patients was obtained from the same patients. All samples and the experiments they used are listed in supporting information

Read quality and alignment
Read quality was assessed using FastQC version 0.11.8 software produced by the Babraham Institute (Babraham, Cambridgeshire, UK), and Trimmomatic v0.36 was used to filter low-quality base calls and any adapter contamination [20] Low-quality leading and trailing bases were removed from each read, a sliding window trimming using a window of four and a phred33 score threshold of 15 was used to assess the quality of the read body. Any reads of < 17 nucleotides were discarded.
Reads were aligned to the human reference genome, GRCh38 using Bowtie2 version 2.2.6 [21]; no mismatches between the seed sequence and the reference genome were allowed, and reads were allowed to align a maximum of 10 times. Using the featureCounts programme from the Subread package version 1.6.4, the number of reads that aligned to the miRNAs, according to miRBase22 [22,23] and other short RNA species extracted from Gencode v31 were calculated [24,25]. miRNAs with ≥ 1 read count in at least one of the samples were kept for analysis. Differential expression analysis was performed using the R package DESeq2 [26]. As RNA-Seq was performed in two batches, the batch was included in the DESeq2 design formula, this allow for the statistical inferences to be adjusted for the batch. The false discovery rate was controlled for using the Benjamini-Hochberg correction, and gene expression changes with an adjusted p-value < 0.05 were considered statistically significant.

miRNA target prediction
Potential targets of differentially expressed miRNAs were explored using the target gene prediction algorithm TargetScan [27]. The correlation between the expression of the miRNA and the expression of the predicted target genes thought to bind in the 3 0 UTR was determined. Target genes that were negatively correlated (<0.6) with their respective miRNA were used for further analysis. To further narrow down the list of potential target genes, pathway enrichment analysis was performed. miRNA regulatory target gene enrichment analysis was performed through gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) [28], with a cut-off criterion of adjusted p-value < 0.05.

Assessment of biomarker usage
To determine the potential biomarker properties of miRNAs, the expression profile of chosen miRNAs was used for a Receiver Operation Characteristic (ROC) curve analysis. This method was used to display the discriminatory accuracy of the miRNA to distinguish between controls and PD with an Area Under the Curve (AUC). The analyses were carried out using the 'pROC' package in R studio [29].
To assess the robustness of miR-543 as a classifier, a permutation analysis was performed. For each permutation, the samples were randomly assigned to groups, and an ROC analysis was performed for miR-543. The AUC for each permutation was recorded, and this was repeated 30,000 times to produce a test statistic distribution. Finally, by observing where the original AUC fell within this distribution, the p-value was calculated.
T A B L E 1 Demographic and details of PD patients and non-demented controls Note: Donors were classified based on their final clinical diagnosis as PD [17]. The autopsy was performed using a standardised protocol by the NBB (open access: www.brainbank.nl and the presence of neuropathological features were assessed following consensus criteria for diagnosis of PD [18,19]. PD, Parkinson's disease; NBB, Netherlands Brain Bank.
For RNA isolation from tissue, 700 μl Qiazol Lysis Reagent Antibodies used can be found in supporting information Table S4.
Images taken (magnification: 20X) with a microscope were loaded into QuPath, and the image type was set to Brightfield (HDAB) for all images. In the annotation menu, the entire image was selected for anal- Data are expressed as a fold-change compared to the control group.

Western blot
For protein analysis, cell cultures were lysed using radioimmunoprecipitation assay (RIPA) buffer supplemented with protease inhibitors. Homogenates were centrifuged at 13,000 g for 10 min at A p-value < 0.05 indicated a statistically significant difference. Data for expression validation are shown as box plots for grey and white matter with mean AE SEM.

Differential expression and validation of miRNAs in PD
A total of 2303 miRNAs were identified as expressed in the MFG of controls and PD patients. To determine differentially expressed miRNAs in PD patients, differential expression (DE) analysis was performed. After batch correction and normalisation, a total of 12 differentially expressed miRNAs were identified in PD patient samples. These miRNAs consisted of nine downregulated and three upregulated miRNAs (supporting information Figure S1). Cell-type composition can function as a potential confounding factor when analysing expression patterns. Therefore, we also assessed the cohort for the expression of specific cell-type markers (supporting information Figure S2) compared to the PDD5/6, and similar results were found in the white matter when comparing the PD5/6 and PDD5/6 group ( Figure 1A and B). For miR-424-3p, upregulation was determined in PD4 in the grey matter, and in PD5/6 upregulation was found in both grey and white matter compared to the controls. Moreover, in the grey matter, PD4 and PD5/6 showed significant upregulation compared to the PDD5/6 group ( Figures 1C and D). Finally, miR-543 was only found to be upregulated in the white matter in PD4 and PD5/6, whereas no significant differences were found in the grey matter when comparing PD groups to the controls ( Figure 1E  for distinguishing between controls and PDD5/6 ( Figure 2B). To test the robustness of miR-543, a permutation analysis (n = 30,000) was performed. The permutation analysis demonstrated that miR-543 performed better than would be expected by chance alone (p < 0.05).
Finally, miR-543 did not show high diagnostic accuracy for PD4 and PD5/6 (supporting information Figure S3). For let-7e-3p, a low diagnostic accuracy was found for distinguishing between groups, eliminating this miRNA as a possible biomarker for PD (supporting information Figure S4). Target prediction of let-7e-3p, miR-424-3p and miR-543 To increase our knowledge of the role of let-7e-3p, miR-424-3p and miR-543, we continued with further bioinformatic analysis through pathway enrichment analysis. The target genes of the differentially expressed miRNAs were predicted bioinformatically using TargetScan [27]. mRNA targets for let-7e-3p were predicted through TargetScan and assessed for enriched pathways (supporting information Table S5).
Pathways related to the regulation of synaptic plasticity and modulation of synaptic transmission were among the most significantly enriched pathways for the targets of let-7e-3p (supporting information

SIRT1 expression in PD tissue
As SIRT1 is only a validated target of miR-543 in organs and tissue outside the brain, we were initially interested in determining the relationship between SIRT1 and miR-543 in PD brain tissue. We investigated both the mRNA and protein expression of SIRT1. SIRT1 mRNA expression was determined through both RNA-Seq and RT-qPCR.
Although RNA-Seq showed no dysregulation of SIRT1 mRNA in either the grey or white matter at any stage of PD, RT-qPCR was used to validate SIRT1 mRNA expression in both grey and white matter. The mRNA level of SIRT1 was shown to be stable across all four groups in both white and grey matter, with only a slight upward trend for F I G U R E 2 Receiver operating characteristic (ROC) curve analysis and violin plots for PD and PDD5/6. (A) ROC curve analysis showed an 88.9% specificity and 69.2% sensitivity for miR-543 in discriminating between the control group and all PD groups combined, corresponding to an AUC of 0.883. (B) For the discrimination between the control group and PDD5/6, miR-543 showed an 88.9% specificity and 100% sensitivity corresponding to an AUC of 0.978. X axis ROC curve: 100-specificity in percentage (%); Y axis ROC curve: sensitivity in percentage (%). X axis violin plot: groups; Y axis violin plot: normalised expression. Control: n = 9, PD4: n = 6, PD5/6: n = 10, PDD5/6: n = 10 PDD5/6 in the white matter (supporting information Figure S5A) and the grey matter (supporting information Figure S5B).
More importantly, we looked at the expression of SIRT1 protein by immunohistochemistry, as we predicted that differential expres- Interestingly, we also found downregulation of SIRT1 protein in PDD5/6 in white and grey matter ( Figure 3B/D). Immunohistochemical staining for SIRT1 ( Figure 3E) showed a clear intensity difference between the control group and PD groups. Moreover, it showed that SIRT1 is expressed in both grey and white matter, whereas in PD without dementia only changes in expression were observed in the white matter.

SIRT1 is a direct target of miR-543 in SH-SY5Y cells and foetal astrocytes
To further validate that SIRT1 is a target of miR-543 in the brain and determine whether this targeting only occurs in the white matter, as tissue experiments indicate, we transfected both SH-SY5Y cells and foetal astrocytes with a miR-543 mimic. miR-543 and SIRT1 expression was determined by RT-qPCR. We found that transfection with miR-543 mimics significantly increases the expression of miR-543 in both SH-SY5Y cells and foetal astrocytes ( Figure 4A and C, respectively). Furthermore, after transfection with miR-543, SIRT1 mRNA was significantly decreased in both SH-SY5Y cells and foetal astrocytes ( Figure 4B and D, respectively). The much stronger downregulation found in foetal astrocytes might suggest that targeting of SIRT1 by miR-543 is more pronounced in the white matter compared to the grey matter.
Additionally, to investigate whether the downregulation of mRNA after miR-543 mimic transfection also leads to the downregulation of SIRT1 protein, we determined the protein expression through western blot. A visible downregulation of protein was observed in foetal astrocytes, but not in SH-SY5Y cells, indicating that upregulation of miR-543 downregulates SIRT1 in the white matter but not in the grey matter which is in line with the data from PD tissue (supporting information Figure S6).

DISCUSSION
Identifying differentially expressed miRNAs in PD increases our knowledge of the underlying mechanisms of the prodromal PD T A B L E 2 Enrichment analysis of target genes of miR-543. Gene ontology (GO) biological process and molecular function  Although there are limited studies identifying miR-543 as a dysregulated miRNA in PD, a recent study found that the inhibition of miR-543-3p can possibly relieve dyskinesia in a PD model through the rescue of glutamate transporter type 1 (GLT-1) expression and function [36]. This is in line with the observed upregulation of miR-543 that could contribute to PD pathology. Previous studies have found miR-424 to be dysregulated in PD [37], and this miRNA was reported to be significantly altered in a study in which the expression levels of miR-424 pre-and post-deep brain stimulation were compared [38]. Previously, studies have found miR-543 to be negatively correlated with SIRT1 expression in gastric cancer tissues [39]. Moreover, a study focused on insulin resistance showed that overexpression of miR-543 lowered SIRT1 mRNA and protein levels in different gastric cell types [40], indicating a negative correlation between miR-543 and SIRT1 expression. However, none of these studies focused on the potential role of miR-543 and its effect on the expression of SIRT1 in the brain. Therefore, we first looked at the expression of miR-543 and SIRT1 in patient tissue in grey and white matter separately to determine whether targeting is cell type specific. Next, by identifying SIRT1 as a target of miR-543 in both SH-SY5Y cells and foetal astrocytes on mRNA level and in foetal astrocytes on protein level, we have provided evidence of a similar relationship between this miRNA and gene in the brain, with this relationship appearing to be most common in the white matter.
SIRT1, a member of the sirtuin family, is a nicotinamide adenine dinucleotide (NAD)-dependent histone deacetylase [41]. Together with other members of the sirtuin family, SIRT1 has been found to act as a deacetylase for numerous protein targets involved in pathways including axonal degeneration [42]. Therefore, dysregulation of SIRT1 expression in PD is hypothesised to contribute to PD pathology. Previous studies have shown that the enzymatic activity of SIRT1 is decreased in patients with PD, which may reduce their ability to resist neuronal damage caused by various cytokines, neurotoxins and α-Syn aggregation [43]. Over the years, SIRT1 has been described to be possibly protective against the toxic α-Syn aggregation through the F I G U R E 4 Expression of miR-543 and SIRT1 mRNA after transfection with miR-543 mimic.
(A) miR-543 is increased significantly after transfection with mimic in SH-SY5Y cells. (B) The expression of SIRT1 mRNA is significantly decreased after transfection with miR-543 mimic.
(C) In foetal astrocytes, miR-543 mimic transfection increases the expression of miR-543 significantly. (D) A strong downregulation of SIRT1 mRNA is observed after the transfection with miR-543 mimic. X axis: groups, Y axis: relative expression compared to control group. Error bars indicate SEM. *p < 0.05, ***p < 0.001, ****p < 0.0001 activation of molecular chaperones such as heat shock protein Hsp70.
SIRT1 overexpression studies showed a reduction in α-Syn aggregates, whereas deletion of SIRT1 increased α-Syn aggregation [44]. Furthermore, SIRT1 was also described to reduce inflammation, apoptosis, and even the activation of astrocytes [45]. Finally, oxidative stress and dysfunction of mitochondria can be caused by the dysregulation of SIRT1 [43]. Thus, with the upregulation of miR-543 resulting in the downregulation of SIRT1, it can be hypothesised that this influences the neuroprotective role of SIRT1 in the brain. Considering the fact that both miR-543 and SIRT1 seem to be most affected in the white matter, the downstream cascade of SIRT1 downregulation may contribute to the early white matter changes observed in PD. Therefore, while a lot more knowledge is required concerning the role of miR-543 and SIRT1 in PD and other neurodegenerative diseases, both can be considered potential neuroprotective therapeutic targets. This would be possible through either the downregulation of miR-543 in early PD or the upregulation of SIRT1 on its own. However, for the potential use of miR-543 as a therapeutic target, it would be necessary to investigate other targets of miR-543 and how they are regulated in PD. The regulation of SIRT1 is very promising and is in line with the studies focused on SIRT1 activator resveratrol. A recent study showed that resveratrol displayed neuroprotective efficacy in several animal models of PD [46]. Furthermore, resveratrol can enhance mitochondrial function via SIRT1 pathways [47].
Although we have clearly shown that SIRT1 is a target of miR-543, the identification and evaluation of other miRNAs involved in the regulation of this gene deserve further investigation. starBase v2.0 was used to look into this further into this issue and identify miR-NAs targeting SIRT1 [48,49]. For this search, a very high stringency was used, which meant that more than five CLIP-seq experiments supported the predicted miRNA target site. This resulted in a total of 29 other miRNAs that have been reported to target SIRT1 (supporting information Table S10). Moreover, a recent study has shown that miR-384-5p targets SIRT1 in mice and SH-SY5Y cell lines. This study found that this regulation possibly promotes the progression of PD through cell apoptosis. However, as this study was performed in mice and SH-SY5Y cell lines, the results are mostly applicable to the grey matter [50]. Therefore, the involvement of miR-543 in the regulation of SIRT1 in the white matter further supports the role of early white matter changes found in PD.
Although the expression of miR-543 and SIRT1 in tissue indicates that miR-543 only targets SIRT1 in the early stages of PD in the white matter, miR-543 was also looked at for its biomarker potential. Initially, it was hypothesised that miR-543 could be a potential biomarker for PD in general or early PD. Although the overall biomarker potential discriminating the control group from the three PD groups combined, showed promise with an AUC of 0.83, a more interesting finding was discovered when separating the three PD groups. miR-543 showed strong potential as a biomarker for PDD5/6 (AUC: 0.978). This result suggests that while miR-543 is dysregulated in PD without dementia, it can also be implicated in PD with dementia. As this biomarker was determined in CSF at the final stage of the disease, it would be interesting to look further into the biomarker potential of miR-543. The expression patterns and prognostic accuracy of miR-543 should be investigated in more readily available biofluids such as blood/serum, and this could lead to the development of a simple, lessinvasive biomarker for cognitive impairment in PD.

CONCLUSION
In conclusion, analysis of differentially expressed miRNAs in PD has led to the identification of miR-543 as a dysregulated miRNA. Furthermore, miR-543 is identified as a regulator of SIRT1 in the white matter of the brain, which possibly implicates both the miRNA and the gene in the early white matter changes in PD, and miR-543 has also shown potential diagnostic accuracy for dementia in PD patients.
With the discovery of a potential biomarker, it is important that future studies investigate the possibility of applying biomarker research in the clinic. The work presented here shows a possible framework that could lead to the discovery of dysregulated pathways in PD and the identification of novel therapeutic targets. Nonetheless, further investigation is still required to establish the role of miR-543 and SIRT1 in PD and miR-543 as potential biomarker or possible therapeutic target.