Circulating microRNAs as potential biomarkers for psychiatric and neurodegenerative disorders

Circulating microRNAs (cimiRNAs) are a class of non-encoding RNAs found in bodily fluids such as blood, cerebrospinal fluid (CSF) and tears. CimiRNAs have been implicated as promising biomarkers for central nervous system (CNS) disorders because they are actively secreted as messengers and are profoundly involved in fine-tuning of developmental and differentiation processes. Furthermore, they are attractive biomarkers because they are extremely stable, tissue enriched and can be determined in a quantitative manner. This review aims to provide a comprehensive assessment on the current progress regarding the potential value of cimiRNAs as CNS biomarkers. Within this framework five CNS disorders are explored which share a common pathological hallmark namely cognitive impairment. The CNS disorders include Major depression disorder (MDD), Bipolar disorder (BD), Schizophrenia (SZ), Alzheimer's disease (AD) and Parkinson disease (PD). The similarities and differences between altered cimiRNAs in the different disorders are described. The miR-29 family, miR-34a-5p and miR-132-3p are discussed as common dysregulated cimiRNAs found in the CNS disorders. Furthermore, it is shown that the type of bodily fluid used for measuring cimiRNAs is important as inconsistencies in cimiRNAs expression directions are found when comparing CSF, blood cell-free and blood cell-bound samples.

the complementary 3'UTR region on the messenger RNA (mRNA) (Bartel, 2004;Ha and Kim, 2014;Shao et al., 2010). The RISC-mRNA interaction causes gene silencing through translational inhibition or mRNA degradation, and thereby depending on the impact, may lead to fine-tuning but even to crisp changes in gene expression (Bartel, 2009;Ha and Kim, 2014). MiRNAs are important regulators and modulators of developmental and differentiation processes, such as; cell proliferation, differentiation, maturation and apoptosis. As example of the large impact on gene expression, miRNAs are heavily involved in early developmental stages, when pluripotent cells switch into lineage-specific cells. This switch is caused, among other factors, by the upregulation of miRNAs. Specifically, miR-9 and miR-124a are of vital importance for neural development. They are highly expressed by neurons and astrocytes in the brain and modulate the phosphorylation of STAT3, an important molecule for neuronal differentiation. There are many more examples of how miRNAs influence developmental processes and we refer the reader to recent reviews on this matter (Hollins and Cairns, 2016;Ivey and Srivastava, 2015;Rajman and Schratt, 2017).

Circulating microRNAs (cimiRNAs)
MiRNAs have not only been detected in tissue, but also in bodily fluids such as blood plasma and serum, cerebrospinal fluid (CSF), urine, saliva, breast milk, seminal plasma and tears (Sohel, 2016). These miRNAs are collectively known as circulating miRNAs (cimiRNAs). CimiRNAs are produced within the nucleus, transported to the cytoplasm and are then actively or energy-free passively excreted into bodily fluids. The latter occurs after cell death as a result of e.g. apoptosis, metastasis or inflammation (Kichukova et al., 2015). Ci-miRNAs are active and selective excreted through extracellular vesicles (e.g. exosomes, micro-particles, apoptotic cell bodies) or excreted as micro-vesicle free miRNAs in the circulation ( Fig. 1) (Hunter et al., 2008;Mause and Weber, 2010;Sohel, 2016). The latter are associated with diverse proteins for stability and protection, such as Argonaute2 (Ago2) and high-density lipoprotein (HDL) (Arroyo et al., 2011;Kichukova et al., 2015;Sohel, 2016;Vickers et al., 2011;Wang et al., 2010). Arroyo et al. (2011) and Turchinovich et al. (2011) showed that 90 % of the cimiRNAs are associated with the Ago2 in plasma and serum. However, other studies indicated that cimiRNAs are predominantly concentrated in exosomes e.g. in serum, saliva and follicular fluid (Gallo et al., 2012;Sohel et al., 2013). These opposing results may indicate that cimiRNAs are differently secreted in various bodily fluids. Subsequently, it is not clear yet whether cimiRNAs associated with Ago2 are actively excreted, or represent a by-product of the physiological cellular activity or apoptosis (Sohel, 2016;Turchinovich et al., 2011).
CimiRNAs in exosomes are relatively more intensively studied compared to the other extracellular vesicles mentioned above (Gallo et al., 2012;Osier et al., 2018). Exosomes are formed via an endocytic multi-vesicular body that fuses with the plasma membrane and subsequently excreted to the surrounding extracellular microenvironment (Fig. 1) (Osier et al., 2018;Prada and Meldolesi, 2016;Raposo and Stoorvogel, 2013;Valadi et al., 2007;Yoon et al., 2014). These exosomes may fuse with other cells leading to the transfer of their load, thereby facilitating cell-to-cell communication and cellular processes (Osier et al., 2018;Prada and Meldolesi, 2016;Yoon et al., 2014). In 2007, Valadi et al. showed for the first time that miRNAs could be transferred between cells through exosomes. Other studies showed the relevance of this exosome-mediated genetic transfer by revealing that certain cancers and neuropathological diseases are dependent on exosome-mediated transport, e.g. serving as cell-to-cell communication between the brain and the periphery (Osier et al., 2018;Sohel, 2016;Yoon et al., 2014). Fig. 1. Biogenesis of miRNA and the release of miRNAs into the circulation by extracellular vesicles or as micro-vesicle free miRNAs. RNA polymerase II transcribes pri-miRNAs from miRNA genes in the nucleus. The long pri-miRNAs are thereafter processed by DROSHA into a shorter pre-miRNAs. The pre-miRNAs are transported out the nucleus to the cytoplasm mediated by Exportin5. Dicer cleaves the pre-miRNAs into mature miRNAs in the cytoplasm. The mature miRNAs are released into the cytoplasm by extracellular vesicles such as exosomes (inward budding) or micro-particles (outward budding). Alternatively, the miRNAs associated themselves with high-density lipoproteins (HDL) or Argonaute 2 (AGO2) and are then released into the circulation. Additionally, miRNAs can passively enter the circulation after changes in permeability or damage of the cell membrane in case of cell death or as a result of apoptosis, metastasis or inflammation.

Blood brain barrier (BBB)
Specifically in relationship to central nervous system (CNS) disorders, for many years it was discussed whether cimiRNAs are capable of crossing the blood brain barrier (BBB), but not many studies demonstrated this yet. During membrane fusion process, the exomes integrate membrane proteins from the cell of origin (Prada and Meldolesi, 2016), so determining these membrane proteins enables to identify the source of origin of the exosomes (Osier et al., 2018). For instance, Ko et al. (2016) characterized the ionotropic glutamate receptor (GluR2) in serum exosomes after a traumatic brain injury (TBI). GluR2 is highly abundant in the membrane of neurons and developing oligodendrocytes in the brain, and is low expressed in the periphery. Because of this, it was suggested that cimiRNA from the brain are able to cross the BBB into the peripheral circulation using exosomes (Ko et al., 2016). Moreover, Xu et al. (2017) recently demonstrated that elevated neuronal miR-132 expression also elevated miR-132 expression in peripheral endothelial cells. It was also shown that endothelial cells take up exosomes derived from neurons, suggesting that neuronal exosomes with their cargo can cross the BBB and are capable of transferring their cargo to the periphery (Xu et al., 2017;Zhao and Zlokovic, 2017). Vice versa, in a mice model of systemic inflammation, peripheral miRNAs cross the BBB (Balusu et al., 2016). This study indicated that choroid plexus-derived extracellular vesicles can cross the BBB, enter the brain parenchyma and are taken up by astrocytes and microglia. Subsequently, causing miRNA target repression and up-regulation of inflammatory genes. However, more research must be done to elucidate the exact mechanism of cimiRNA excretion, and in particular how micro-vesicle free cimiRNAs cross the BBB, especially under milder conditions than severe inflammation or TBI.

Characteristics and detection
MiRNAs have been shown to be tissue-enriched, including different regions in the brain. For instance, Cogswell et al. (2008) examined that the miRNA profiles were different for the hippocampus, cerebellum and medial frontal gyrus in Alzheimer's disease (AD) patients in terms of expression direction and concentration. They revealed region-specific miRNAs as well as Braak stage-specific dysregulation of miRNAs (Cogswell et al., 2008). Correspondingly, Bekris et al. (2013) investigated the relationship between brain miRNAs, CSF and plasma at different Braak stages of AD patients. The study indicated, likewise Cogswell et al. (2008), that the miRNAs are brain region-specific. Collectively, these results indicate that changes in cimiRNA levels might be related to changes in brain tissue miRNA levels. Hence, it has been hypothesized that cimiRNA signatures reflect physiological and pathological conditions of diverse CNS disorders, such as AD .
CimiRNAs are stable and resistant to degradation from endogenous RNAs activity Tsui et al., 2002). They are able to withstand severe environmental conditions such as extreme pH levels, boiling, long-term storage and multiple freeze-thaw cycles in multiple different types of bodily fluids Glinge et al., 2017;Sanz-Rubio et al., 2018;Tsui et al., 2002). In addition, cimiRNAs can be determined in a quantitative manner using different techniques, such as next generation sequencing (NGS), quantitative real time polymerase chain reaction (qRT-PCR) and microarrays. The most widely applied method is qRT-PCR because of its simplicity, high sensitivity and capability to detect low levels cimiRNAs, but it can only be applied for a limited number of selected miRNAs. Microarrays enables measurement of about 2000 selected cimiRNAs, but analysis requires larger amounts of RNA, has lower reproducibility and represents problems with crosshybridization (Git et al., 2010;Jin et al., 2013;Krauskopf et al., 2015). NGS allows throughput quantification of global cimiRNAs levels, including low abundant cimiRNAs. Additionally, it also allows the detection of novel cimiRNAs. However, NGS is appreciably more expensive than the other methods; hence, it may be used to explore disease-specific cimiRNAs fingerprints but is not suitable for clinical diagnostic purposes yet (Jin et al., 2013). For all methods the limiting step is the amount of input material needed, consequently, one wants to extract preferably all the cimiRNAs from a sample. This is challenging since cimiRNA concentrations are relatively low in bodily fluids compared to e.g. proteins (Jin et al., 2013). This limitation, in combination with the current absence of standardized methodological and normalization approaches causes cimiRNAs still to be in its infancy of becoming widely used biomarkers.

CimiRNAs as potential biomarkers for CNS disorders
CimiRNAs may have the potential to be used to accurately diagnose CNS disorders already at an early stage. This is beneficial for patients in order to receive earlier and more individual targeted treatment therapies. Although the number of studies investigating the potential value of cimiRNAs as biomarkers for diseases has exponentially increased, still a gap in knowledge of cimiRNAs in relation to CNS disorders exists. The development of biomarkers for CNS disorders is lagging behind because of complex mechanisms involved, inaccessibility of brain tissue, and complex neuroanatomy and functioning of the brain. Moreover, many confounding factors influence experimental study designs such as environment-gene interactions, heterogeneity of CNS disorders and stratification of population sample. This review focusses on studies that have investigated the potential value of cimiRNAs serving as biomarkers for acquired psychiatric and neurodegenerative disorders. Moreover, it discusses the implications of the altered cimiRNAs on cellular processes and etiologies of the different CNS disorders, as some cimiRNAs may reveal to be risk factors or modulate the disorder. This review focuses on five main CNS disorders, which all have cognitive impairment implications, namely, Major depression disorder (MDD), Bipolar disorder (BD), Schizophrenia (SZ), AD and Parkinson disease (PD). We investigated the similarities and differences of dysregulated cimiRNAs within and between the different disorders. In particular, we discuss the effects of measuring cimiRNAs in different bodily fluids. Consequently, this review aims to provide a comprehensive assessment on the potential value of cimiRNAs as CNS biomarkers for brain pathology.

Literature search
We performed a Pubmed search in 2018 and explored all articles. Only articles that explored cimiRNAs involved in one or more of the 5 selected CNS disorders were considered for inclusion. Exclusion and inclusion criteria for the CNS disorders are described and specified in each article. Articles were included when cimiRNAs were measured in (blood)cell-free samples including serum, plasma and CSF and blood cell-bound cimiRNAs including whole blood, peripheral blood mononuclear cells (PBMC), leukocytes and blood-derived monocytes (BDM). Studies were excluded if no significant altered cimiRNAs were reported. In total 99 articles were included in the overview (Tables 1-5). Ci-miRNAs names were manually updated to the current names found at www.mirbase.org for analyses purposes, and only current names are used in this review. In Tables 1-5 the original cimiRNAs names are depicted.

Major depression disorder (MDD)
Major depression disorder (MDD) is the most prevalent psychiatric disorder and it has been estimated that about 30-40 % of the risk for depression is caused by genetic factors (Association, 2013;Bigdeli et al., 2017). In 2019, worldwide more than 264 million people suffered from MDD and about 800 thousand people die due to suicide yearly (World  (Belzeaux et al., 2012) (continued on next page) M.M.J. van den Berg, et al. Progress in Neurobiology 185 (2020) 101732 Health Organization, 2019). Although there are many treatments for depression, this is not fully deployed in all. One of the reasons is an inappropriate assessment of the type of depression, so patients often do not receive the appropriate type of antidepressant (e.g. only approximately 50 % of the depressed patients respond to the first-line treatment with a selective serotonin reuptake inhibitor (Kennedy, 2013)). The diagnosis is predominantly evaluated based on the patient's subjective complaints leading to a very heterogeneous disorder and therefore patients are often misdiagnosed (World Health Organization, 2019). Despite many years of research, no specific biomarker for MDD has been identified yet. Several articles investigated cimiRNAs in MDD patients and collectively 98 different cimiRNAs were identified, 50 of these were upregulated, 45 downregulated and 3 cimiRNAs were bidirectional (Fig. 2, supplementary A).  are the most prevalent upregulated and miR-636 is most downregulated cimiRNAs. Li et al. (2013) suggested that the brain-derived neurotrophic factor (BDNF) gene is likely the most essential gene involved in the pathophysiology for MDD. It was shown that miR-182-5p and miR-132-3p, which both decrease BDNF production, were upregulated in MDD patients. Accordingly, low serum BDNF level is suggested to be a marker of MDD (Molendijk et al., 2011). Moreover, miR-132-3p levels are positively correlated to scores on the self-rating depression scale (SDS) Su et al., 2015), anxiety scores and impaired visual memory dysfunction in MDD patients . Likewise, Su et al. (2015) observed increased peripheral blood miR-132-3p levels and decreased peripheral blood BDNF levels.
Another important cimiRNA involved in MDD is miR-124-3p, which is neuron-specific and involved in synaptic plasticity (Fischbach and Carew, 2009). Roy et al. (2017) detected that serum miR-124-3p was elevated in MDD patients compared to matched controls. Moreover, it was observed in post mortem brain tissue that several genes, which can be linked to the pathophysiology of MDD, were dysregulated by the increased levels of miR-124-3p (Roy et al., 2017). Moreover, miR-124-3p might be therapeutically relevant as it was shown that PBMC miR-124-3p was significantly downregulated after different antidepressant treatments .

Bipolar disorder (BD)
Bipolar disorder (BD) has a lifetime prevalence of around 1 % in the general population (Rowland and Marwaha, 2018). BD is a complex chronic mental disorder characterized by recurrent episodes of mania and depression, as well as impairments in cognitive performance (Anderson et al., 2012;Association, 2013;Schloesser et al., 2008). It has been suggested that BD arises from abnormalities in synaptic and neuronal plasticity, and that several candidate genes are involved in the pathophysiology of BD (Schloesser et al., 2008). A growing body of data suggests that cimiRNAs regulate neuroplasticity, and therefore it has been hypothesized that cimiRNAs can be sensitive biomarkers for BD . Unfortunately, as revealed by our literature research, only a limited number of studies have explored cimiRNAs in BD. Collectively, 16 cimiRNAs were dysregulated in BD, 11 were upregulated and 5 were downregulated (Fig. 3, supplementary B).
It has been suggested that miR-134 is involved in synaptic plasticity, as it represses the translation of LIM Domain Kinase 1 (Limk1)-mRNA,  which encodes for a protein kinase heavily involved in dendritic spine development (Schratt et al., 2006;Ye et al., 2016). Moreover, it has been suggested that BDNF mediates the release of miR-134 (Schratt et al., 2006). Given the close relationship of miR-134 with neuroplasticity, and thus likely with the pathophysiology of BD, Rong et al. (2011) hypothesized that miR-134 would be altered in BD patients. Indeed, the study revealed that plasma miR-134 levels were significantly lower in BD patients compared to age/gender matched controls. After 2 weeks of a combined antipsychotic and mood stabilizer treatment, plasma miR-134 levels increased in BD patients, but did not differ significantly from pretreatment baseline. Subsequently, after 4 weeks of treatment, the plasma miR-134 levels were significantly increased in BD patients compared to baseline, but remained significantly lower compared to control subjects. These results suggest that aberrant expression levels of miR-134 might be involved in the pathophysiology of BD, through disturbing the neural and synaptic plasticity processes in the brain (Rong et al., 2011). Consequently, miR-134 plasma levels may serve as a diagnostic biomarker for BD, and prognostic biomarker for antipsychotic and mood stabilizer treatment.

Schizophrenia (SZ)
Schizophrenia (SZ) is a highly prevalent occurring neuropsychiatric disorder, affecting approximately 1 % of the population worldwide (Saha et al., 2005). Research findings showed that genetic factors are strong mediators for the development of SZ, as it has been estimated that the heritability of SZ can range between 24-80 % depending on the endophenotype (Sun et al., 2015a). SZ is characterized by positive (e.g. hallucinations, delusions), negative (e.g. anxiety, depression) and cognitive (e.g. memory impairments) symptoms, resulting in a heterogeneous disease that is difficult to diagnose (Association, 2013;Owen et al., 2016). The diagnosis of SZ remains symptom based, relying on the mental state examination, self-reports and clinical interviews, which sometimes leads to misdiagnosis (Owen et al., 2016;Wakefield, 2010). Physiological and biomedical biomarkers for diagnosing SZ are needed and recent studies have proposed the potential value of ci-miRNAs as early and accurate biomarkers for SZ.
27 Aberrant cimiRNAs in SZ patients were found, of which 17 were upregulated, 9 were downregulated and 1 cimiRNA was shown to be both up-and downregulated in SZ patients (Fig. 4, supplementary C). The upregulated cimiRNAs including miR-181-5p, miR-30e-5p, miR-34a-5p and miR-7-5p are the most prevalent occurring cimiRNAs and have been detected in 3 or more independent studies.
Increased levels of miR-181b-5p have been detected in plasma (Song et al., 2014;Sun et al., 2015b) and serum (Shi et al., 2012), as well as in brain regions such as the superior temple gyrus and dorsolateral prefrontal cortex of SZ patients (Beveridge et al., 2008). Song et al. (2014) explored plasma miR-181b-5p levels of SZ patients before and after a 6-month treatment of different atypical antipsychotics. The study revealed that miR-181b-5 plasma levels decreased after treatment and were positively correlated with the improvement of negative symptoms. Consequently, Beveridge et al. (2008) suggested that miR-181b-5p targets genes for glutamate, serotonin and cannabinoid receptors involved in synaptic transmission, as well as genes involved in brain development and neurodevelopmental disorders (Beveridge et al., 2008). Regarding the latter, miR-181b-5p especially targets glutamate ionotropic receptor AMPA type subunit 2 (GRIA2) and Visinin-like protein 1 (VSNL1), which are genes that have been associated to the     (Lugli et al., 2015) (continued on next page)  (Zhu et al., 2015) (continued on next page)   (Beveridge et al., 2008;Eastwood et al., 1995;Vawter et al., 2002). The GRIA2 protein is involved in fast excitatory neurotransmission, which is an important component for synaptic plasticity including the establishment of long-term potentiation (LTP) and long-term depression (LTD), which are supposedly electrophysiological substrates of memory (Luscher and Malenka, 2012), and for stimulating growth and density of dendritic spines (Beveridge et al., 2008;Carroll et al., 2001;Passafaro et al., 2003). VSNL1 codes for VILIP-1, which is a calcium sensor protein and a BDNF receptor (tropomyosin receptor kinase B (trkB)) mRNA binding protein. It has been suggested, that the depleted expression of VSNL1 elucidates the altered calcium concentrations and reduced expression of trkB found in the brain of SZ patients (Bernstein et al., 2002). Moreover, aberrant expression of VSNL1 has been observed in phencyclidine and ketamine rodent models of SZ (Bernstein et al., 2003;Kajimoto et al., 1995). Collectively, these results indicate the possible contribution of miR-181b-5p and its target genes in the development of the complex pathophysiology of SZ. Additionally, multiple studies have linked aberrant expression of miR-30e-5p to SZ. Increased miR-30e-5p levels have been found in plasma (Song et al., 2014;Sun et al., 2015a;Sun et al., 2015b), peripheral leukocytes  and PBMC (Sun et al., 2015a), as well as in the prefrontal cortex of SZ patients (Perkins et al., 2007). Furthermore, Xu et al. (2010) revealed that the polymorphism

Fig. 2. Venn diagram depicting the direction of expression of cimiRNAs detected in MDD.
Comparing all samples including serum, plasma, PBMC, whole blood, and CSF.  ss178077483, located in the precursor of miR-30e-5, has been associated with SZ . However, it remains unclear how miR-30e-5p links to the pathophysiology of SZ.

Alzheimer's disease (AD)
Alzheimer's disease (AD) is a progressive chronic neurodegenerative disease characterized by cognitive impairment. The pathophysiological hallmarks of AD is the development of extracellular β-amyloid (Aβ) plaques, which are aggregates of oligomers of the Aβ protein, and intracellular neurofibrillary tangles, which are aggregates of hyperphosphorylated tau protein (Dos Santos Picanco et al., 2018;Swerdlow, 2007). It is the most common type of dementia, accounting for 60-80 % of the dementia cases (Kljajevic, 2015) and it has been estimated that in 2030 65.7 million people will be diagnosed with AD (Prince et al., 2013). AD attributes to extremely high medical costs due to medication, diagnostic procedures and caretaking (Sopina et al., 2019). AD is therefore not only a social but also an economic burden. An early and accurate diagnose for AD could reduce in trillions of medical costs and permits possible earlier therapy interventions for AD patients.
There are different serum and CSF protein biomarkers to evaluate AD, such as Aβ and tau protein levels (Bekris et al., 2013). Although very promising biomarkers, its sensitivity and specificity for AD are still questioned (Rosa et al., 2014;Sorensen et al., 2016). Moreover, Aβ and tau levels are late response biomarkers. CimiRNAs appear to be early and more sensitive biomarkers in this regard (see below).
It was hypothesized that cimiRNAs which target AD-related genes, such as amyloid precursor protein (APP) and Presenilin -1 and -2 (PSEN1&2), play a crucial role in the pathophysiology of AD (Bekris et al., 2013;Wu et al., 2016). For instance, miR-29a-3p and miR-29b-3p target β-site APP-cleaving enzyme (BACE1), which contributes to the process of amyloid protein accumulation, forming Aβ plaques (Holsinger et al., 2002). The downregulation of these miRNAs may explain the abnormal high levels of BACE1 proteins detected in AD patients (Hebert et al., 2008). The miR-29 family will be discussed more extensively in the Section 3.1.
MiR-146a-5p has an increased expression level in both the peripheral circulation and brains of AD patients. Subsequently, miR-146a-5p has been associated with neuro-inflammation, which promotes degeneration of neurons Denk et al., 2015;Muller et al., 2014). Inflammation of the brain occurs relatively early in the onset of AD, and therefore Muller et al. (2014) speculated that miR-146a-5 plays an important role in initiating AD. Their research study showed that miR-146a-5p was upregulated in the hippocampus and CSF of AD patients in Braak stages III/IV, but downregulated in Braak stage VI compared to controls, thereby supporting their initial hypothesis (Muller et al., 2014). Furthermore, Denk et al. (2015) showed an inverse correlation of miR-146a-5p with tau and Aβ in the brain, suggesting that miR-146a-5p might have an inhibitory activity on tau production (Denk et al., 2015). Finally, Alexandrov et al. (2012) showed that the above mentioned miR-9-5p, miR-125-5p, miR-146a-5p and miR-155-5p were upregulated in AD CSF before any significant correlations were seen between Aβ levels and the AD pathophysiology. This underlines the potential usefulness of cimiRNAs as early diagnostic biomarkers for AD.

Parkinson disease (PD)
Parkinson disease (PD) is the second most prevalent neurodegenerative disease after AD, affecting approximately 1-2 % of people over 60 years old worldwide (Tysnes and Storstein, 2017). The main neuropathological hallmark of PD is the composition of α-synuclein-containing Lewy bodies and degeneration of dopaminergic neurons in the substantia nigra (Lees et al., 2009). Deficiency of the nigro-striatal pathways in PD patients lead to motor dysfunction such as tremor, rigidity, postural instability and bradykinesia, and non-motor symptoms such as pain, fatigue, psychiatric problems and impaired cognition (Chen et al., 2018;Kalia and Lang, 2015;Sveinbjornsdottir, 2016). The disease starts in the brainstem, which makes it hard to early diagnose and monitor progression of PD. Diagnosis for PD relies largely on impaired motor functions, which only occurs a few years later after the onset of the disease (Kalia and Lang, 2015). Considerable research has been performed to explore early, specific and sensitive biomarkers for PD. However, despite the efforts no clinical biomarkers have been proven yet (Kalia and Lang, 2015).

CimiRNAs involved in multiple CNS disorders
The miR-29 family, miR-34a-5p and miR-132-3p are dysregulated in 2 or more of the selected CNS disorders and a plethora of studies have investigated their role in the development of the CNS disorders. These cimiRNAs seem to play a central role in pathways that lead to cognitive deficits. The cimiRNAs and their associated pathways are described below.
3.1.  The miR-29 family seems to play a predominant role in neurodegenerative disorders such as AD, PD and Huntington disease (HD) . The miRNAs included in this family are miR-29a-3p, miR-29b-1, miR-29b-2 and miR-29c, which are transcribed from chromosome 7 and 1 of the human genome, respectively (Bai et al., 2017;Mott et al., 2010). MiR-29b-1 and miR-29b-2 have an identical mature sequence, whilst miR-29a and miR-29c differ by one nucleotide (Bai et al., 2017). Studies imply that the miR-29 family are associated with neuronal survival, maturation proliferation, plasticity and dendritic spine and synapse morphology (Bai et al., 2017;Kole et al., 2011;Lippi et al., 2011). Bai et al. (2017) showed that the expression level of the miR-29 family in serum of PD patients was significantly downregulated. Interestingly, serum miR-29b was to a lesser extent downregulated in PD patients compared to miR-29a and miR-29c. It has been hypothesized that a duplication of miR-29b in the human genome may affect its expression in another way than the other miRNAs of its family. The expression level of serum miR-29a and miR-29c were negatively correlated to disease severity (Hoehn & Yahr stage) but not to disease duration, Unified Parkinson's Disease Rating Scale (UPDRS) scores or Ldopa therapy. Similarly, Margis et al. (2011) showed significant decreased levels of peripheral blood miR-29a in PD patients but no significant changes between L-dopa treated patients and de novo patients. The authors propose that miR-29a expression level should not be related to motor symptoms of PD, since miR-29a levels remain low despite the improved motor symptoms during L-dopa treatment (Margis et al., 2011). Furthermore, in contrast to gender, age does not affect the miR-29 family expression level in controls and PD patients (Bai et al., 2017;Botta-Orfila et al., 2014). Bai et al. (2017) and others showed that serum miR-29a and miR-29c levels in females were markedly higher than in male controls and PD subjects (Bai et al., 2017;Botta-Orfila et al., 2014). After gender stratification, cimiRNA differences were mainly found in PD males (Botta-Orfila et al., 2014), collectively suggesting that the expression level of the miR-29 family in PD is genderspecific. It is well-known that the prevalence of PD is higher for males than females (Wooten et al., 2004). Thus, it warrants further investigation what the role is of the miR-29 family in the pathophysiology and onset of PD and its association with gender-specific miRNA regulation (Bai et al., 2017;Botta-Orfila et al., 2014;Wooten et al., 2004).
Although, Bai et al. (2017) and Botta-Orfila et al. (2014) propose that dysregulated expression levels of serum miR-29 family are PD specific, many others have also shown dysregulated levels of miR-29 s levels in AD patients. As mentioned before, miR-29a and miR-29b are associated with BACE1, contributing to AD pathology (Holsinger et al., 2002). Increased brain and serum ceramide levels have been associated with AD (Filippov et al., 2012;Mielke et al., 2012). MiR-29a and miR- 29b-1 post transcriptionally regulate SPT long chain 2 (SPTLC2), which is a subunit of serine palmitoyltransferase (SPT), the rate limiting enzyme involved in ceramide synthesis (Geekiyanage and Chan, 2011). Linked to this, it was shown by Geekiyanage et al. (2011) that decreased miR-29a and miR-29b-1 levels were correlated with increased SPT and Aβ levels in the brain of an AD mouse model. Collectively this suggests that SPT and the regulatory miR-29 s play a role in AD through ceramide production (Geekiyanage and Chan, 2011). Another target of miR-29b is specificity protein 1 (Sp1), which is a transcription factor abundant in the brain responsible for neuronal survival and regulates AD-related proteins such as BACE1, COX-2 and tau related genes (Citron et al., 2008;Villa et al., 2013). It was shown that downregulated mir-29b levels in AD PBMC samples were negatively correlated with Sp1 levels in AD brains and PBMC samples (Villa et al., 2013).
Of note, in PD downregulated miR-29 s appear to be associated with gender, however to our knowledge this has not been shown and investigated in AD patients. Contrary to PD, AD is higher prevalent in females than in males (Podcasy and Epperson, 2016). Interestingly, gender stratification in a mouse model of AD indicated that serum miR-29bs are significantly more downregulated in female than in male mice (Geekiyanage et al., 2012). It is peculiar that the miR-29 family appears to be differently gender-specific regulated between AD and PD. There are many factors contributing to the susceptibility for developing neurodegenerative diseases, and it has yet to be elucidated what the associations are between certain cimiRNA expression levels, such as the miR-29 family, and cofactors such as gender but also hormones, ethnicity and dietary intake (Geekiyanage et al., 2012;Villa et al., 2013).

MiR-34a-5p
Upregulation of miR-34a-5p in blood samples were observed in AD, MD and SZ patients. MiR-34a-5p has been associated with cell survival/ apoptosis and neuroprotection signaling though B-cell lymphoma 2 (Bcl2) and silent information regulator 1 (SIRT1) deacetylase, respectively. MiR-34a-5p suppresses SIRL1, leading to increased acetylated p53, a regulator of the cell cycle/apoptosis and functions as a tumor suppressor. It has also been shown that miR-34a-5p is a transcriptional target of p53, thereby establishing a positive feedback loop between miR-34a-5p, p53, and SIRT1 (Li et al., 2011;Yamakuchi et al., 2008). Oxidative stress induces upregulation of p53 activity, consequently increasing miR-34a-5p expression levels (Bhatnagar et al., 2014;Li et al., 2011). Oxidative stress has been suggested to contribute to ageing, but also to neurological and psychiatric disorders. Thus, oxidative stress augmented by disease and ageing further enhances the upregulation of miR-34a-5p and contributes to weakening of the oxidative defense signaling pathways, leading to a positive loop between oxidative stress and the severity of the disease. AD, MDD and SZ are all correlated with high oxidative stress levels, which could clarify the upregulated miR-34a-5p found in the serum and brain tissue of these patients (Bhatnagar et al., 2014;Bitanihirwe and Woo, 2011;Kim et al., 2010;Li et al., 2011;Palta et al., 2014;Zhao and Zhao, 2013). Moreover, AD, MDD, SZ and anxiety disorders have been genetically linked to the SIRT1 gene (Herskovits and Guarente, 2014;Kim et al., 2007;Kishi et al., 2011Kishi et al., , 2010Libert et al., 2011;Lu et al., 2018;Palta et al., 2014). The link between oxidative stress, miR-34a, SIRT1 and CNS disorders is an interesting topic that requires further investigation.

MiR-132-3p
Neuronal plasticity and its associated pathways have shown to be dysregulated in SZ, MDD, AD, PD and BD (Liu et al., 2017b;Miller et al., 2012;Qian et al., 2017). MiR-132-3p targets important pathways involved in neuronal plasticity such as LTP, DNA methylation and neuronal cAMP response element-binding protein (CREB), N-methyl-Daspartate receptor (NDMAr) and BDNF signaling (Qian et al., 2017). The latter seems to play a predominant role in the pathophysiology of CNS disorders and is involved in neurodevelopment and synapse regulation (Nieto et al., 2013). MiR-132-3p is closely related to MDD through BDNF signaling. Low BDNF levels were also detected in serum, plasma and CSF in AD and SZ patients (Laske et al., 2011;Pillai et al., 2010). Likewise MDD, miR-132-3p levels are upregulated in SZ and AD and correlated with disease severity and cognitive deficits (Xie et al., 2015). Accordingly, it is plausible to suggest that BDNF signaling through miR-132-3p plays a central role in the pathophysiology of the selected CNS disorders leading to common dysregulated pathways, eventually causing cognitive deficits, a hallmark shared between the selected CNS disorders (Laske et al., 2011;Nieto et al., 2013;Tanila, 2017).
Additionally, it has been shown that miR-132-3p is sensitive to treatment therapies. It was shown by Sun et al. (2015b) that miR-132-3p levels decreased significantly after 3 and 6 weeks of psychotropic treatment therapy (i.e olanzapine) in SZ patients. Furthermore, the plasma levels of miR-132-3p were highly correlated with symptomatic improvements, especially those in the high effect group of SZ patients. Interestingly, Alieva et al. (2015) observed that miR-132-3p expression levels were 3 times higher in the peripheral blood lymphocytes of treated PD patients compared to the untreated PD patients. It was concluded that miR-132-3p could be a potential indicator of treatment response and prognosis of CNS disorders.

Effects of bodily fluids
Numerous different bodily fluids have been used to detect cimiRNAs in. Based on this review the highest percentage of studies have either used serum (23 %), plasma (21 %) or whole blood (14 %) samples. Only rarely studies compared cimiRNAs within and between different bodily fluids. CSF is rarely collected because it is more invasive than blood sampling, except for AD, because patients allow CSF sampling for diagnostic purposes.

CSF versus blood cell-free and cell-bound samples
Using CSF is an advantage due to its close interaction with the brain and therefore it is expected that cimiRNAs in the CSF resemble brain miRNAs (Stoicea et al., 2016). Of note, Wei et al. (2015a) declares that CSF-based cimiRNA profiles are more sensitive and specific than bloodbased miRNA profiles for diagnosing cancers of the CNS. Their metaanalysis revealed that CSF-based profiling for CNS cancers was 7 % more specific and detected 3 % more CNS cancers in CSF-profiling than blood-profiling (Wei et al., 2015a). In addition, Stoicea et al. (2016) pointed out that blood volume is greater than CSF volume, therefore the concentration of cimiRNAs in blood is diluted compared to CSF. Furthermore, Burgos et al. (2014) mentioned that a high percentage of the cimiRNAs are both detectable in CSF and blood, however, they are often poorly correlated, or even appear contradicting to each other in terms of expression direction, as further described below.
We observed inconsistencies in the direction of expression of certain cimiRNAs between different studies. For example, in AD 38 cimiRNAs were shown to be both upregulated and downregulated (Fig. 5), for MDD 3 cimiRNAs (Fig. 2) and PD 8 (Fig. 6). The differences in expression direction may have been caused by detecting the cimiRNAs from different bodily fluids. Interestingly, we observed that often the CSF-cimiRNA signature profile appears contradicting to the peripheral blood-cimiRNA signature profile in terms of expression direction ( Fig. 5  and 6, supplementary D and E). To further analyze this, we have split the differentially expressed cimiRNAs into three categories; blood cellfree (serum, plasma), blood cell-bound (whole blood, PBMC, leukocytes, BDM) and CSF samples (supplementary A-E). This clearly indicated that that the majority of the inconsistence between the directions of expression of cimiRNAs are between the CSF samples and the blood cell-free and cell-bound samples. Examples of inconsistencies between CSF and blood based samples are described below.
A few exploratory studies have compared CSF and blood cimiRNA profiles directly in the same cohort. Song et al. (2015) investigated the involvement of miR-16-5p in MDD through its regulation of serotonin transporters. Their study revealed that CSF miR-16-5p was significantly lower in MDD patients and was positively correlated with CSF serotonin. However, no difference was detected in blood miR-16-5p levels between MDD patients and controls . Likewise, Wan et al. (2014) indicated that more cimiRNAs were significantly altered in CSF than in serum of MDD patients compared to controls. Their results indicated that 11 miRNAs were significantly upregulated in CSF of MDD patients, whilst only 3 of these 11 miRNAs were upregulated in serum of the same individuals. Subsequently, 5 miRNAs were downregulated in CSF whilst only 1 of these miRNAs was also downregulated in serum (Wan et al., 2015). Burgos et al. (2014) compared post-mortem collected cimiRNAs profiles of CSF and serum samples of AD and PD patients. The study observed that only 2 miRNAs were overlapping between the two different bodily fluids for AD samples, and the expression directions were contradicting. Interestingly, the authors observed that the aberrant ci-miRNAs in the CSF samples were all downregulated, whilst in the serum samples 85 % of the aberrant miRNAs were upregulated for AD patients (Burgos et al., 2014). Sorensen et al. (2016) further analyzed the correlation between CSF and blood cimiRNAs and measured the functioning of the BBB with albumin ratios. The study indicated that there were no correlations between the expression of individual miRNAs in CSF, blood and the leakage of the BBB. The authors also concluded that the individual expression of cimiRNAs in CSF and blood are individually regulated and are unaffected by small disturbances in the BBB (Sorensen et al., 2016).
Notably, the study by Gui et al. (2015) and Burgos et al. (2014) show contradicting results regarding the direction of expression of three miRNAs, even though they both use the same bodily fluid, CSF. MiR-409-3p, miR-136-3p, miR-10a-5p are upregulated in the study by Burgos et al. (2014) but downregulated in Gui et al. (2015) for PD patients. The difference in expression direction may be caused by comparing post-mortem samples with ante-mortem samples. Further, Gui et al. (2015) isolated miRNAs from the exosomes in the CSF, whilst the other study used whole CSF. CimiRNA release to the circulation might be differently regulated between exosomal and free-floating cimiRNAs. Interestingly, Liu et al. (2014a) supports this hypothesis as the authors found a correlation between exosomal miR-193b-3p in CSF, serum, and plasma of AD patients. In contrast, no correlations were observed between the total blood and CSF miR-193b-3p levels.

Blood cell-free versus blood cell-bound
Although CSF-based miRNA profiles seem to have higher sensitivity and specificity for detecting CNS disorders, it is often preferred to detect cimiRNAs in blood samples because it is less invasive. A few exploratory studies examined the differences and similarities between cimiRNA profiles in different blood samples such as serum, plasma, PBMC and whole blood in the same cohorts. This review discriminates between blood cell-free and blood cell-bound samples, and it seems that often discrepancies in cimiRNA directions are between these two fractions (supplementary A-E).
Literature suggests avoiding whole blood samples for cimiRNA profiling because of high background noise. MiRNAs derived from red and white blood cells may contaminate the sample and coagulation of platelets could result in false positive results (Duttagupta et al., 2011;Wang et al., 2012). Pritchard et al. (2012) indicated that perturbations of blood cell counts and hemolysis alter plasma cimiRNAs levels by a 50-fold increase, impairing cimiRNAs profiles greatly (Pritchard et al., 2012). A study by Sun et al. (2015a) supported this argument, as the authors indicated that plasma cimiRNA expression profiles were more sensitive than PBMC profiles for discriminating SZ from healthy controls.

Plasma versus serum
Lastly, also differences have been found between serum and plasma cimiRNA expression profiles. Wang et al. (2012) observed in cancer patients, although not statically significant, higher cimiRNA concentrations in serum samples compared to plasma samples of the same individuals. The authors suggested that additional RNA was released during the coagulation process of platelets. During this process, blood cells are exposed to stress, which may lead to the stimulation of ci-miRNA release causing higher miRNA concentrations in serum than in plasma . The coagulation process is a complex physiological process that varies among individuals and increases sample-to-sample variation, hence increases cimiRNAs sample variation . Considering these observations, it is plausible to conclude that plasma samples are a better choice for studying cimiRNA profiles.

Future perspective and concluding remarks
Using cimiRNAs as biomarkers is a fastly developing field. CimiRNAs have the potential to become useful biomarkers for the detection of cancers (i.e. gastric and breast cancers), diabetes and cardiovascular diseases. Nonetheless, up-to-date cimiRNAs do not yet fulfill the specificity and sensitivity criteria as clinical biomarkers, and therefore have not been applied clinically yet (Filipow and Laczmanski, 2019;Grimaldi and Incoronato, 2019;Jaeger et al., 2018;Poel et al., 2018;Saliminejad et al., 2019;Zhang et al., 2018). Only recently, cimiRNAs have explored the field of CNS disorders as they have the potential to become highly valuable biomarkers at an early stage. Moreover, cimiRNAs may help to unravel the complex pathophysiological mechanisms of CNS disorders, as explored in this review. However, cimiRNAs are still in its infancy of becoming reliable biomarkers for CNS disorders, and the above mentioned other diseases. Many factors contribute to or affect the expression levels of cimiRNAs and these have to be taken into account. This exploratory review showed in particular that the type of bodily fluid used for measuring cimiRNAs is important as inconsistencies in cimiRNAs expression directions are seen when comparing CSF, blood cell-free and blood cell-bound samples. Moreover, there are other factors contributing to ci-miRNA expression such as ethnicity, age, dietary intake and gender. Therefore, we would need larger sample sizes and dedicated studies enabling the validation of these cimiRNAs in multiple disease cohorts. Also, methodological and technical implications such as the sequencing platform (NGS vs. qRT-PCR), isolation and purification of the samples and data normalization need to be considered. For instance, applying standard operating procedures (SOPs) for isolation and processing, 2) focusing more on NGS for detecting all cimiRNAs instead of applying qRT-PCR, 3) focus on cimiRNA fingerprints rather than a single ci-miRNAs, and 4) setting up interlaboratory validation studies.
Subsequently, building and using databases indicating regulation and (raw) expression values of cimiRNAs in different CNS disorders is essential, allowing cimiRNAs to be compared in a quantitative and a more advanced statistical manner, rather than comparing lists of differentially expressed cimiRNAs as done in this review. Such platforms specific for cimiRNAs have only been recently constructed. The exRNA database (https://exrna-atlas.org) contains cimiRMA datasets for a diversity of diseases and miRondola database (http://mirandola.iit.cnr.it/ ), provides comprehensive information on extracellular circulating noncoding RNAs for various diseases (Ainsztein et al., 2015;Russo et al., 2012). Recently, Li et al. (2019) launched a new platform called the Circulating MicoRNA Expression Profiling (CMEP) database that contains large-scale cimiRNA datasets from different platforms including NGS, qRT-PCR and microarrays and from a wide range of sample types including cell-free, cell-bound and urine cimiRNAs (Li et al., 2019). These platforms are crucial in order to compare datasets across different disorders. However, unfortunately, these platforms are incomplete, not functional or still contain little CNS disorder datasets. Moreover, as discussed by Li et al. (2019) these databases still have limitations since the datasets are conducted by different research groups on different sequencing platforms, subsequently, leading to inevitable batch effects. Using statistical advanced approaches to compare cimiRNA between the different CNS disorders is therefore challenging, but it is a crucial next step in order to improve cimiRNAs biomarker research.
Taken the above mentioned factors into consideration, the field of cimiRNAs can develop further and possibly in the near future cimiRNAs have the potential to early diagnose CNS disorders and possible evaluate prognosis of therapies. However, it is essential that technical and methodological approaches improve before cimiRNAs will operate as clinical biomarkers.