A Disease in Search for a Biomarker: MicroRNAs in Parkinson’s Disease

: This review discusses the current research data on using microRNAs as biomarkers for the diagnostics and screening of Parkinson’s Disease (PD). We provide a comprehensive, critical analysis of the overwhelmingly diverse data on circulant microRNAs associated with PD. We also highlight the possible underlying molecular pathogenesis-related circulant microRNAs in the context of the natural history of PD


Introduction
With a worldwide yearly incidence of up to new cases/ , individuals, Parkinson's Disease (PD) is the second most common neurodegenerative disease after Alzheimer's.Its global burden is predicted to double by .PD is a disease of the elderly (> years of age), its prevalence increasing abruptly up to % in individuals > years of age [ , ].The incidence of PD varies depending on sex (more common in men), race (less common in African Americans), ethnicity (more frequent in Ashkenazi Jews, Inuit, and Alaska Natives), and environment (e.g., exposure to pesticides) [ ].
The diagnosis of PD is essentially clinical, based on non-motor (hyposmia, constipation, sleep disorder, cognitive decline, and psychiatric disturbances) and motor (rigidity, rest tremor, postural instability, and bradykinesia) symptoms [ ].The therapeutic response (rapid improvement of motor symptoms) to levodopa, the assessment of cardiac sympathetic denervation (by myocardial scintigraphy with iodine--meta-iodobenzylguanidine), and the dopamine transporter single-photon emission computed tomography (DaT SPECT) are valuable tests for de ning PD diagnostics.However, they cannot di erentiate PD from parkinsonism involving the dysfunction of dopamine transport [ , ]. Usually, at the time of diagnosis, half of the dopaminergic neurons are already lost in the substantia nigra in patients with motor signs [ ].
Most patients with PD respond well to dopamine substitution therapy, the o periods and dyskinesia appearing within to years after the onset of levodopa therapy [ ].There are no biochemical or molecular bona , ( ), de biomarkers for PD diagnosis, therapy monitoring, or prognostic evaluation.Intensely investigated in the last decade, blood, saliva, and cerebrospinal uid (CSF) α-synuclein have proven to lack speci city and sensitivity [ , ].
MicroRNAs are endogenous small non-coding RNAs involved in post-transcriptionally regulating gene expression.A single microRNA modulates the stability of hundreds of mRNAs, while one mRNA might interact with multiple target mRNAs, a biunivocal relationship that explains the ability of microRNAs to regulate the expression of almost half of the human transcriptome [ ]. Due to their outstanding stability in various biological uids, microRNAs are ideal biomarker candidates in a wide variety of pathologies, including neurodegenerative diseases [ ].
The quest for a PD-related microRNA biomarker has led to an impressive amount of data, characterized by a bewildering lack of consistency/overlap, due to di erences in methodologies, analytical platforms, target tissues (CSF, whole blood, plasma, and serum), and the ethnicity and staging of the included patients [ ]. Very few studies ful ll the requirements of a biomarker study in terms of design, the stringency of analysis, and the scale of patients' enrolment, which could explain why microRNAs have not reached the clinical setting.A basic analytical pipeline would include a screening step (usually next-generation sequencing (NGS), microarray, or RT-PCR array) followed by a validation step on a di erent technology (e.g., by qRT-PCR) and in an independent cohort; in both steps, stringent criteria of inclusion and the exclusion of both patients and controls and proper bioinformatics and statistical analysis are essential.
Here, we analyze the current published data describing the association of microRNA with PD and their application in diagnosis and therapy monitoring.We performed a systematic search on PubMed and Google Scholar using the search terms "Parkinson's" AND "microRNA" AND "human", further re ned by "cerebrospinal uid", "whole blood", "plasma", "serum", and "PBMC".We restricted our search to human patients and included only statistically signi cant data from original research papers (thus excluding review papers and meta-analyses).We recorded the microRNA ID, the method of screening/detection, the direction of deregulation, the size of the cohorts analyzed, and the DOI.

Cerebrospinal Fluid
There are several arguments in favor of using CSF to analyze di erentially expressed microRNAs in PD: it is in direct communication with the extracellular space of the brain and, given the blood-brain barrier (BBB), it re ects, almost exclusively, the brain physiology and pathology.

Burgos et al. (
) compared postmortem microRNA pro les of CSF and serum in PD patients and identi ed a speci c, stage-and time-dependent microRNA PD signature.Interestingly, the authors found that CSF microRNAs are slightly more stable and consistent than serum in discriminating PD versus controls.Unexpectedly, none of the microRNAs identi ed after miRDeep analysis with DESeq normalization were validated by qRT-PCR [ ].Of note, there is a limited overlap between the CSF and serum microRNAs, which raises questions regarding the utility of serum microRNA in exploring PD biology, given that BBB alterations do not in uence the (lack of) correlation between the CSF and the blood compartments [ ].
Gui et al. isolated CSF exosomes and used low-density TaqMan arrays to describe a profound alteration of the microRNA pro le in PD patients: upregulated and downregulated.Furthermore, eight microRNAs (miR-, miR-, miR--p, miR-b-p, miR-a-p, miR--p and miR-and let-g-p) were further validated by individual qRT-PCR assays in independent samples, arguing for the validity of the data [ ].
Marques et al. used a targeted approach to monitor the expression of microRNAs in the CSF of PD patients.They found that miR-(age-independently) and miR-(age-dependently) discriminate with a relatively modest accuracy between PD and control samples [ ].
In a (very) preliminary study, Qin et al. ( ) describe a consistent downregulation of hsa-miR-in the CSF of sporadic PD patients versus Alzheimer's disease (AD) and controls; however, the small size of the cohorts and the control group consisting of encephalitis and Guillain-Barre syndrome patients in uence the validity of this microRNA as a diagnostic biomarker [ ].
Caldi et al. sequenced small RNAs isolated from extracellular vesicles in the CSF of PD patients.They demonstrated, using a machine learning approach, that miR--p (the most discriminative), miR-a-p, and miR--p (the least discriminative) could di erentiate between PD and control samples, a feature validated in an independent cohort of probands [ ]. Dos Santos et al. identi ed a panel of microRNAs (Let-f-p, miR-a-p, miR-a-p, miR-a-p, and miR--p) with over % sensitivity in discriminating between early PD and healthy controls; interestingly, when combined with α-synuclein, the discriminating power rose to % sensitivity, % speci city, and % AUC [ ].
Overall, unique microRNAs were found to be di erentially regulated in the CSF of PD patients, of which only overlap: let-b-p, let-g-p, miR-a-p, miR--p, miR--p, miR-, miR-b-p, miR--p, and miR-a-p.Furthermore, except for let-g-p (upregulated), miR-(upregulated), and miR-a-p (downregulated), the other overlapping microRNAs showed divergent changes (Table ).All these discrepancies might be due to di erences in the PD cohorts analyzed (e.g., postmortem vs. vivo, stage, length of disease, and exosomes vs. whole CSF) and/or the stringency of the normalization procedures.It is currently di cult to discern whether exosome-related microRNAs are superior markers compared to whole CSF, given that the permeability of the BBB for microRNAs is a phenomenon that is currently not fully understood; brain-derived microRNAs can reach the periphery, while astrocytes and microglia take up peripheral microRNAs [ ].

Whole Blood
Margis et al. were the rst to show that whole blood microRNAs can be used to di erentiate between PD and control samples: miR-, miR-*, and miR-distinguished untreated PD from controls, while miR--*, miR-a *, and miR a identi ed treated from untreated patients.Interestingly, the authors demonstrate that Levodopa therapy also increased the expression of miR-, miR-*, and miR-, which decreased in untreated patients, suggesting that miR upregulation in response to PD substitutive therapy might be a rather large phenomenon.To our knowledge, this is the rst study investigating the possible role of Levodopa in modulating the expression of microRNAs in biological uids [ ].
A very interesting exploratory study by Grossi et al. found a signi cant upregulation of miR-a-p in a subset (small size) of extracellular vesicles (EV) from the plasma of PD patients; notably, although the discrimination power is rather modest (AUC = .), the change is associated with the length and staging (Hoehn and Yahr, H-Y) of the disease (suggesting accumulation along with evolution), and the Beck Depression Inventory score.Surprisingly, however, there are no signi cant correlations with the patients' age at disease onset, the dosage of levodopa, or uni ed Parkinson's disease rating (UPDR) scale.Of signi cant interest is the authors' nding that small EVs contain more microRNAs than the other miR--p was found to be strongly overexpressed in PD patients and could discriminate (although with relatively modest accuracy, AUC = .) PD from healthy controls but did not correlate with the disease's length or severity [ ].
Bai et al. were among the few to consistently explore sex di erences in the expression of miR-family microRNAs in the serum of PD patients; the authors describe a signi cant downregulation of miR-a, b, and c, and show that all three microRNAs are expressed at higher levels in female samples.Interestingly, the expression of miR-a and miR-c is inversely correlated with disease severity, suggesting they might serve as biomarkers for disease progression [ ].In an exciting follow-up on miR-family expression, the group of Jian Wang analyzed the expression of miR-in relation to cognitive impairment in PD patients; mir-b seems to be the best performer: it discriminates the PD patients with dementia from non-dementia probands, and is associated with the global cognitive status and the z-scores of memory, language, and executive function.Surprisingly, none of the miR-family members were associated with any of the clinical features analyzed: age, sex, duration, staging, and severity of the disease [ ]. Overall, unique microRNAs were found to be di erentially regulated in the serum of PD patients, of which overlap between at least two studies: miR-, miR-b-p, miR-a-p, miR-, miR-b, miR-, miR-, miR-a, mir-b, and miR-c.All the changes are concordant: miR-, miR-expression (upregulated), miR-, miR-b-p, miR-a-p, miR-b, miR-, miR-a, mir-b, and miR-c (downregulated) (Table ).

Peripheral Blood Mononuclear Cells (PBMCs)
Behbahanipour et al. used a qRT-PCR targeted approach to identify the deregulated PBMC expression of three microRNAs usually associated with aging and cellular senescence: miR-and miR-(increased), and miR-(decreased).The combined analysis of all three microRNAs led to an astonishing level of discrimination between PD and controls, with an AUC of .[ ].
Martins et al. used microarrays to identify an intriguing, strong downregulation of microRNAs in PBMCs from PD patients; interestingly, the authors describe as the primary source of variation the a ection status of the individuals included in the study [ ].
A targeted RT-PCR array approach focused on the expression of miR-, miR-a, miR-a, and miR-in PBMCs of PD patients under substitutive therapy identi ed miR-as a putative responder (downregulation) to Levodopa therapy [ ].Of note, miR-is known for its role as a modulator of alpha-synuclein-induced in ammatory phenomena in PD [ ].
Soreq et al. investigated the changes in microRNAs expression in the leucocytes of PD patients before and after one hour of deep brain electrical stimulation; when compared to healthy controls, the sets of microRNAs pre-( microRNA) and poststimulation ( microRNAs) have a surprisingly consistent overlap ( ve microRNAs), although with the inverse direction of change.It is thus surprising that pre-and poststimulation datasets show only one single overlap (although, again, the change is in opposite directions), miR-, the signi cance of which remains obscure.In a complex network analysis of microRNA-mRNA interactions pre-and postelectrical stimulation, the authors show that the interactions pattern change, with miR-, miR-, and miR-occupying central positions [ ].
Sera n et al. designed a targeted approach to identify the microRNAs di erentially expressed in the PBMCs of treated vs. naïve (untreated) PD patients [ ].
Fazeli et al. investigated the association of miR-a-p and miR-b-p with SRRM (an RNA splicing factor) in the early diagnostics of PD.Interestingly, both miR-a-p and miR-b-p show an age-dependent upregulation in healthy controls, which is abolished in PD miR-b-p (but not miR-a-p) and is inversely correlated with disease severity, and thus, might predict disease progression [ ].
Baghi et al. analyzed clinical samples and MPP+-treated SHSY Y cells and showed a strong upregulation of miR-a in PBMCs, while MPP+-treated SHSY Y cells showed a biphasic response: initial downregulation in acute MMP exposure followed by upregulation in chronic exposure.miR-a is correlated with disease severity and has an acceptable AUC ( . ); thus, it might serve as a diagnostic and prognostic biomarker for PD [ ].
Overall, unique microRNAs were di erentially regulated in the plasma of PD patients, none overlapping between at least two di erent studies (Table ).

Discussion
Assaying microRNAs in the blood and its components is relatively simple, fast, inexpensive, and, above all, minimally invasive.Given the outstanding stability of microRNAs in the various biological uids, it is tempting to search for microRNA signatures associated with a pathology known for its lack of a bona de biomarker.
As mentioned in the introduction, the data from all these studies show a surprising lack of overlap among them, and a speci c miRNA signature is basically impossible to describe.In our opinion, the main reason for this resides in the rather small size of cohorts and the heterogeneity of the probands (and controls) included in the analyses, which is not only due to the disease itself but also to the subjectivity of diagnostics.Next, the diverse source of biological uids analyzed also plays an important role: CSF is richer in brain-derived microRNAs, while the serum content of small RNAs is signi cantly skewed by microRNA species associated with activation of platelets during coagulation [ ]. From a technical point of view, the puri cation methods are known to in uence both the yield and quality of the (small) RNA isolated.Of note, few data sets were obtained after checking for hemolysis (at least by monitoring oxyhemoglobin absorbance at nm), a factor known to signi cantly in uence extracellular microRNA levels [ ]. Next, the method for analysis of microRNA levels in biological uids also a ects the results.Used in basically all data sets analyzed, qRT-PCR (and, by extension, qRT-PCR arrays) represents (due to its sensitivity and speci city) the gold standard in assessing the expression of a microRNA in a tissue.Although they cannot match the sensitivity and speci city of qRT-PCR yet, NGS and microarray compensate by breadth and depth of investigation.Furthermore, there are issues related to the reproducibility and replicability of the high-throughput approaches (especially when dealing with low-input RNA samples); hence, validation by qRT-PCR is obligatory [ ]. Finally, and often overlooked or scarcely approached, there is a (overly) broad spectrum of normalization procedures, including endogenous microRNA, endogenous non-microRNA molecules, and exogenous spiked-in microRNAs.Thus, it is of paramount importance to identify a (set of) microRNA with stable expression across all samples, independent (as much as possible) of age, sex, and physiological/pathological condition.In our opinion, and based on our experience, combining spike-ins with endogenous microRNAs produces the best results.
Overall, of all unique microRNAs found to be deregulated in biological uids sampled from PD patients, .% were upregulated, arguing (should other regulatory mechanisms, such as transcriptional activation by transcription factors of CpG methylation, be excluded) for a general, post-transcriptional inhibitory e ect on gene target networks in PD pathology.This ratio (upregulated microRNAs/downregulated microRNAs > ) is maintained in CSF ( .% upregulated) and plasma ( .%) PD samples; unexpectedly, the ratio is reverted in the serum ( %) and PBMC ( .%), indicating signi cant di erences in the microRNA's sources for the three blood components analyzed.
The similarities between CSF and plasma microRNA pro les continue when comparing the upregulated and downregulated microRNA lists; there are three overlaps between the upregulated lists of microRNAs (miR--p, miR-b-p, and miR--p), and three between the downregulated lists of microRNAs (miR--p, miR--p, and miR--p).Thus, it is possible that in plasma, only the changes in miR-b-p, miR--p, miR--p, and miR--p levels accurately re ect the brain-derived CSF microRNAs alteration in PD (Figure ).

Conclusions
How useful are these microRNA biomarkers in discriminating between PD and non-PD patients?The sensitivity and speci city of the microRNAs identi ed/analyzed vary widely and, especially when combining several microRNAs, surpasses %.We consider the analysis of subsets of circulant extracellular vesicles, which have the potential to re ne the diagnostics and augment discrimination accuracy, to be of particular interest.Of note, very few studies have managed to validate their results in independent cohorts and with a di erent platform.
Despite the poor reproducibility between the studies included in our analysis, the identi cation of a few microRNAs showing concordant changes in CSF and plasma components is encouraging.More studies combining NGS (for screening), qRT-PCR (for validation), and techniques to isolate subsets of circulant extracellular vesicles would eventually lead to a PD-speci c microRNA signature.
subsets [ ].This result contrasts with and complements the data from Cosin-Tomas et al. describing no change in miR-a-p in the plasma of PD patients [ ]. Cardo et al. analyzed (by PCR array followed by qRT-PCR validation) a small cohort of patients and described a signi cant upregulation of miR--p [ ].In yet another targeted approach looking into microRNAs modulating α-synuclein expression, miR-and miR-b were found to be downregulated in the plasma of PD patients, but with no correlation with neither the age nor stage of the disease [ ]. Two microRNAs highly enriched in adult brain structures and known for their involvement in PD pathogenesis were found to be dysregulated in PD plasma by Li et al.: miR-(upregulated) and miR-(downregulated) distinguished between PD and controls, although no correlation with neither the UPDRS score nor H-Y stage could be evidenced [ ].In a rather unusual analysis, Khoo et al. combined k-Top Scoring Pairs (k-TSP) and signi cance analysis of microarrays (SAM) to identify a set of di erentially regulated microRNAs, of which miR-/miR-b-p), miR-, and miR-have the highest predictive power and highest sensibility and sensitivity in the discovery cohort.Unfortunately, none of these characteristics hold up after the analysis of the validation lot [ ].In one of the very few published analyses of plasma from naïve PD patients, Chen et al. describe a set of ve deregulated microRNAs able to discriminate between PD and controls with surprisingly good accuracy (AUC > .) [ ]. Uwatoko et al. used microarray to comparatively analyze microRNA expression in PD and MSA and describe a signi cant (correlated) upregulation of miR-b-p and miR--p and downregulation of miR--p, which has discriminatory power between the two related pathologies [ ].

Serum
Serum miR-b, which is strongly downregulated in PD patients, is correlated with serum ceruloplasmin, a molecule believed to be involved in PD pathogenesis [ ]. Dong et al. used Solexa sequencing to analyze a rather large set of PD patients and identi ed a set of four downregulated microRNAs (miR-, miR-, miR-b-p, and miR-a-p), which could di erentiate between HY stage and [ ]. Shu et al. show a strong downregulation of miR--p and miR--p in PD samples; both microRNAs are positively correlated with disease severity and negatively correlated with Braak staging [ ].There is a surprising concordance between serum and CSF microRNAs in the data of Burgos et al. ( ); except for miR-, all serum microRNAs (miR---p, miR-a-p, miR-e-p, and miR--p) were dysregulated in the same direction in the CSF.None of them correlated with Braak stages, neuro brillary tangle score, or plaque density score [ ]. Vallelunga et al. aimed to di erentiate between PD and MSA and found that when compared to controls, PD serum was enriched in mir-, miR-, and miR--p and had lower levels of miR-c and miR-b.At the same time, in comparison with MSA, miR-, miR-b, and miR-b are upregulated [ ].
Cao et al. describe a downregulation of miR-b and upregulation of miR-and miR-in the exosome-like microvesicles puri ed from the serum of PD patients [ ].The expression of none of these microRNAs correlated with age, sex, smoking or drinking habits, or H-Y scale.miR-showed the best diagnostic value (AUC = .) and the best sensitivity and speci city values ( .% and .%, respectively) [ ]. Ding et al. describe the deregulation of ve microRNAs (miR-, miR-b, miR-, miR-a, and miR-) in PD samples; although the individual sensitivity and speci city values do not surpass .% and .%, respectively (the case of miR-b), altogether, this set of microRNAs can correctly classify .% of PD cases [ ]. Ma et al. tested the ability of serum PD-associated microRNAs to predict PD and found four microRNAs (miR-c, miR-a, miR-, and miR-) to be downregulated when compared to healthy controls.Out of these four candidates, miR-performed the best: AUC = .and is positively correlated with UPDRS-III and UPDRS-V (although with modest scores) [ ]. Jin et al. demonstrated elevated levels of miR-d-p in PD patients vs. controls and Alzheimer's disease or MSA patients and speculated upon the possible impact on alteration of ceruloplasmin levels; however, miR-d-p correlated with neither the severity nor motor phenotype of the disease [ ].
, controls in discovery lot; idiopathic PD, LRRK PD, and controls in validation set ; idiopathic PD, controls in validation set [ 's disease.

Figure .
Figure .Venn diagram depicting microRNAs in CSF and plasma from PD patients.

Figure .
Figure .Venn diagram depicting upregulated (left) and downregulated (right) microRNAs in plasma, serum, and PBMCs from PD patients.

Table .
Altered microRNA expression in CSF from PD patients.

Table .
Altered microRNA expression in plasma from PD patients.

Table .
Altered microRNA expression in serum from PD patients.

Table .
Altered microRNA expression in PBMCs from PD patients.