MicroRNAs in Neurodegenerative Diseases and Their Therapeutic Potential

Abstract

MicroRNAs (miRNAs) are abundant, endogenous, short, noncoding RNAs that act as important post-transcriptional regulators of gene expression by base-pairing with their target mRNA. During the last decade, substantial knowledge has accumulated regarding the biogenesis of miRNAs, their molecular mechanisms and functional roles in a variety of cellular contexts. Altered expression of certain miRNA molecules in the brains of patients with neurodegenerative diseases such as Alzheimer and Parkinson suggests that miRNAs could have a crucial regulatory role in these disorders. Polymorphisms in miRNA target sites may also constitute an important determinant of disease risk. Additionally, emerging evidence points to specific miRNAs targeting and regulating the expression of particular proteins that are key to disease pathogenesis. Considering that the amount of these proteins in susceptible neuronal populations appears to be critical to neurodegeneration, miRNA-mediated regulation represents a new target of significant therapeutic prospects. In this review, the implications of miRNAs in several neurodegenerative disorders and their potential as therapeutic interventions are discussed.

Neurodegenerative disorders are characterized by the progressive loss of specific neuronal groups in the central nervous system, resulting in significant motor and cognitive disability, such as Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), amyotrophic lateral sclerosis (ALS), and Prion diseases. Development of neuroprotective and disease modifying therapies for these disorders has been hindered by our limited understanding of their basic pathogenic mechanisms. Molecular studies about these disorders have often focused on rare monogenic mutations which are not necessarily informative of the pathogenic mechanisms of the common sporadic forms of the respective diseases, where combinations of many genetic loci and non-genetic environmental factors appear to play critical roles. As gene regulatory networks are multi-layered, greater knowledge of these networks will be required to fully comprehend the pathophysiological complexity underlying neurodegenerative diseases. An important and newly emerging layer of biological complexity comes from miRNA-based gene regulatory networks. A number of studies have revealed that some miRNAs are differentially expressed in the human brain and regulate the expression of genes associated with specific neurodegenerative disorders (Eacker et al., 2009; Provost, 2010; Sonntag, 2010). Here, we present an overview of our current understanding of miRNA-mediated functions in these diseases and the therapeutic implications of this knowledge.

1. The basic biology of MicroRNA

Many types of small, non-coding RNA molecules have been found in diverse eukaryotes (Lee & Ambros, 2001; Ambros et al., 2003; Bartel, 2004). One of the most studied of these is the class of miRNAs that are involved in post-transcriptional regulation of target gene expression. As of the writing of this review, 1424 human miRNAs have been identified and listed in the official miRNA database (miRBase). Mature miRNAs are single stranded RNA molecules of about 20 -25 nucleotides that are derived from a 70 - 100 nucleotide long hairpin-precursor called pre-miRNA. miRNAs are commonly transcribed from intragenic or intergenic chromosomal DNA regions by RNA polymerase II into primary transcripts of various lengths (usually 1-3 kb) named pri-miRNAs (Lee et al., 2004a; Garzon et al., 2010). The pri-miRNA is cleaved endonucleolytically by an RNase complex composed of the RNase III Drosha and its cofactor, the double-stranded RNA-binding protein, DiGeorge syndrome critical region gene 8 (DGCR8), producing a hairpin structure (Lee et al., 2003; Landthaler et al., 2004). The pre-miRNA is then transported to the cytoplasm by an exportin-5 dependent mechanism. Once in the cytosol, pre-miRNAs are further processed by a different RNAse III enzyme, known as Dicer, into a mature double-stranded miRNA of variable length (∼20–25 nucleotides) (Lee et al., 2004b). Cleavage results in an imperfect duplex that is unwound, and the strand with the weakest base pairing at the 5′ end is preferentially loaded into the RNA-induced silencing complex (RISC). Therefore, the guide strand or mature miRNA is joined into a RISC, while the passenger strand, marked as miRNA* is commonly destroyed (Hutvagner & Zamore, 2002; Chendrimada et al., 2005). Binding of the miRNA to its target mRNA takes place within the RISC, and target silencing entails the presence of the Argonaute (AGO) family proteins (argonaute 1– argonaute 4) (Bartel, 2009). miRNA targeting is specified by base-pairing between positions 2 to 8 in the miRNA (the seed region) and usually the 3′-untranslated region (UTR) of its target mRNAs (Bartel, 2009). The binding of miRNAs to the 3′-UTR of their target mRNAs leads to a decrease in the expression of the target proteins by either degradation of the mRNA or repression of translation. When there is perfect complementarity between the sequences of miRNA and its target mRNA, the RISC complex induces mRNA degradation, whereas imperfect miRNA–mRNA target pairing induces translation inhibition (Bartel, 2009). The resulting translationally blocked mRNAs and miRNA complex are then translocated to cytoplasmic processing bodies (P-bodies) where untranslated mRNAs are stored and eventually degraded (Liu et al., 2005) through other enzymes such as Dcp1/Dcp2 (mRNA decapping enzymes) and Xrn1 (exonuclease) (Sheth & Parker, 2003; Cougot et al., 2004). In addition to the complexity of the steps involved in miRNA biogenesis, recent findings suggest that the maturation process of miRNA is highly regulated, and that miRNAs themselves are controlled post-transcriptionally as well (Farajollahi & Maas, 2010). As evidence accumulates that abnormal miRNA expression may play a role in the pathogenesis of neurodegenerative diseases, alterations in the biogenesis of miRNA will provide important insights into molecular mechanisms of disease, and could eventually generate novel targets for therapeutic intervention.

2. miRNAs in neurodegenerative disorders

2.1 Genetic models with Dicer mutants

Experiments employing genetic manipulations of the RNase III Dicer in animal models raised the possibility that miRNAs are important in the pathogenesis of neurodegenerative disorders. Ablating this key enzyme, which is required for early steps of miRNA biogenesis, eliminates the generation of all mature miRNAs in the specific cell or tissue explored. For example, spinocerebellar ataxia type 3 (SCA3), which is characterized by progressive dysfunction of the cerebellum and brain stem, is caused by polyglutamine (polyQ) expansion in the ataxin-3 gene. Expression of mutant ataxin-3 in Drosophila bearing dicer-1 mutation exacerbated polyQ mediated toxicity compared with dicer-1 wild-type flies. Additionally, depletion of R3D1, a dsRNA-binding protein, which forms a stable complex with Dicer-1, causes accumulation of precursor miRNA and enhances ataxin-3-induced toxicity in flies and in HeLa cells (Bilen et al., 2006). A systematic screen for factors that modulate polyQ toxicity in the fly model identified the Drosophila miRNA bantam as a potent repressor of polyglutamine-induced toxicity. Thus, bantam prevents the toxicity induced by expanded ataxin-3, although the mechanism remains uncertain. Notably, bantam also reduces tau-induced toxicity associated with AD, suggesting that this particular miRNA targets downstream apoptotic pathways in protecting cells from toxicity rather than by regulating the expression of specific disease-causing proteins (Brennecke et al., 2003). This raises the possibility that single miRNA species might improve cell survival without directly targeting mRNAs encoding specific pathogenic proteins. It should be emphasized that bantam is not conserved in mammals and no homologue has yet been identified in higher organisms, raising the possibility that other miRNAs could have corresponding roles in mammals. However, expression of bantam in HeLa cells resulted in diminished aggregation of the pathogenic form of ataxin-3 (Bilen et al., 2006), indicating that bantam can be functional in heterologous mammalian cells.

The role of miRNAs in neuronal cells has also been investigated in the mouse brain. Using the Cre-LoxP recombination system, models have been developed whereby Dicer protein and/or activity is diminished in specific neuronal cell types including cortical/hippocampal neurons, Purkinje cells and midbrain dopaminergic neurons (Kim et al., 2007; Davis et al., 2008; Kawase-Koga et al., 2009; Hebert et al., 2010). The timing of Dicer deletion during embryogenesis is critical in determining the extent of abnormalities in neuronal migration and differentiation as well as in oligodendroglial expansion and differentiation (Kawase-Koga et al. 2009). Generally, models of Dicer deletion in select neuronal populations exhibit a progressive neuronal loss associated with behavioral deficits reflecting the cell types that were targeted by the Cre-LoxP system. For example, mice lacking Dicer specifically in dopamine neurons are born alive but develop a progressive neurodegeneration with 90% loss of these neurons associated with reduced locomotion in an open-field (Kim et al., 2007), which is reminiscent of the phenotype of PD. Such findings suggest that Dicer is essential for neuronal survival, and loss of miRNAs may be involved in the development and/or progression of PD. Although this study supports the view that dysfunction of the miRNA network in the brain is involved in the pathogenesis of neurodegenerative diseases, these findings as well as others obtained using a similar approach should be interpreted with caution since Dicer is also involved in other functions such as the production of small interfering RNA (siRNA) and the formation of the heterochromatin structure (Fukagawa et al., 2004; Tam et al., 2008). However, transfection of dicer-deleted mouse embryonic stem (ES) cells with small RNAs of < 200 bp in length (including miRNAs) derived from mouse embryonic midbrain partially protects the dopaminergic phenotype whereas large RNAs (> 200 bp) cannot (Kim et al., 2007). This finding raises the likelihood that the absence of miRNAs rather than the lack of other potential Dicer-related functions contributes to the neurodegenerative process when Dicer is depleted. It remains to be elucidated whether some of these neuronal defects are attributable to the loss of a specific miRNA species or are due to the generalized loss of the miRNA machinery. As an example of loss of a specific miRNA leading to a neurodegenerative phenotype upon conditional Dicer knock-out, a recent study demonstrated that deletion of Dicer in the adult forebrain results in neuronal loss in the hippocampus and cellular shrinkage in the cortex accompanied with hyperphosphorylation of tau at several epitopes associated with neurofibrillary tangle pathology. In these mice, ERK1 was identified as one of the candidate kinases for tau (Hebert et al., 2010), a finding that led to the discovery of the miR-15 family as potent regulators of ERK1 expression. Adding credence to these findings, miR-15a was found to be down-regulated in AD brains (Hebert et al., 2010).

Conditional deletion of Dicer has also been used to model spinal muscular atrophy. This has been demonstrated by crossing a Dicer1 conditional allele with a Cre-recombinase transgene driven by a cholinergic-specific promoter, which directs transgene expression in postmitotic spinal motor neurons (SMNs) as early as postnatal day 7 but is not expressed during development (Haramati et al., 2010). Thus, these mutant mice lose the ability to make functional miRNA in a subset of SMNs postnatally. A phenotype simulating spinal muscular atrophy emerges in these mice with sclerosis of the spinal cord ventral horns, considerable decline in motor neuron numbers throughout the lumbar region, abnormal end plate architecture, myofiber atrophy and denervation compared to control animals. Behaviorally, these patholgoical features correlate with progressive locomotor dysfunction and a median survival of just 29 weeks. These observations indicate that the miRNA machinary is required for the survival of SMNs after birth.

The critical role of miRNA homeostasis in neuronal function is further substantiated by the development of behavioral and neuronal deficits in mice lacking another component of the miRNA-generating machinery encoded by the DGCR8 gene (Stark et al., 2008). These mutant mice show an obvious reduction in miRNA levels associated with abnormal dendritic spine development and complexity, corroborating the central role of miRNAs in this neuronal phenotype. Taken together, these studies demonstrate the significance of miRNAs in neuronal maintenance and survival, and suggest that failure of miRNA function during aging could contribute to a neurodegenerative phenotype in humans. The on-going questions are which miRNAs are impaired functionally in specific disease states and what are the roles of individual as well as networks of miRNAs in these diseases.

2.2 miRNA profiling studies

2.2.1. Parkinson's disease

Analysis of postmortem brain tissue from disease affected brains and their comparison to non-affected controls has been one strategy to understand the role of miRNAs in disease states. Accordingly, several reports demonstrate that changes in miRNA networks in the brain play a part in the pathogenesis of neurodegenerative diseases. Profiling the expression of 224 different miRNAs obtained from the midbrains of patients with PD and control subjects showed distinct alterations in several miRNAs, including miR-133b (Kim et al., 2007), although the relative number of patients investigated in this study was too small to draw definitive conclusions about their clinical significance. Normally, miR-133b is enriched in the midbrain but has been found to be less abundant in the brains of patients with PD. miR-133b appears to function as a negative regulator of dopaminergic neuron development, as its overexpression results in reduced number of dopamine neurons, while its inhibition has the opposite effect in primary embryonic rat midbrain cultures (Kim et al., 2007). Pitx3, which is one of the physiological targets of miR-133b, is a transcription factor involved in dopaminergic neuron differentiation (Hwang et al., 2003; Nunes et al., 2003; Martinat et al., 2006). A negative-feedback loop consisting of Pitx3 and miR-133b has been proposed as the Pitx3 gene transcribes miR-133b, which in turn suppresses Pitx3 expression. However, the discovery that miR-133b inhibits the differentiation of dopaminergic neurons while its expression is decreased in the brains of PD patients appears on the surface to be incongruent. Therefore, whether miR-133b might have additional functions in dopaminergic neuronal maturation beyond silencing Pitx3 expression remains to be elucidated.

More recent miRNA profiling of PD brains uncovered decreased expression of miR-34b and miR-34c, but not miR-133b, in affected brain areas such as the amygdala, frontal cortex, substantia nigra and cerebellum (Minones-Moyano et al., 2011). Additionally, knock-down of miR-34b/34c in differentiated SH-SY5Y neuroblastoma cells resulted in cell death associated with altered mitochondrial activity and oxidative stress, which are consistently recognized biochemical abnormalities in PD. However, it remains to be proven whether the decreased expression of these miRNAs is due to their specific down-regulation in surviving neurons or simply secondary to dopaminergic neuron degeneration. Identifying the physiological targets of miR-34b and miR-34c is also needed to better understand the role of these miRNAs in cellular pathways that they might modulate.

2.2.2. Alzheimer's disease

Similar profiling studies have been performed in other neurodegenerative diseases. Expression profiling of 328 microRNAs in anterior temporal cortex from five sporadic AD patients and five age-matched controls showed the expression of 13 microRNAs to be decreased in AD (Hebert et al., 2008). Notably, among them, miR-29a, -29b-1, and -9 can decrease the expression of β-amyloid cleavage enzyme 1 (BACE-1), which is one of two critical enzymes required for the cleavage of amyloid precursor protein and the generation of toxic Aβ species (Vassar et al., 2009). The same study also found that the expression of the cluster miR-29a/b-1 was significantly down-regulated in a subgroup of AD patients in whom BACE-1 protein was aberrantly enhanced while BACE-1 mRNA levels were unaffected (Hebert et al., 2008). Further, experimental evidence is provided for a causal link between miR-29a/b-1 expression and Aβ generation in a cell culture model, suggesting that loss of miR-29a/b-1 can contribute to increased BACE-1 and consequently Aβ levels in sporadic AD. Down-regulation of miR-29a in AD brains has been confirmed in a separate miRNA profiling study (Shioya et al., 2010). Using microarray analysis and in situ hybridization, another study showed that the expression of miR-107 is decreased in the brains of AD patients at the earliest stages of pathology in parallel with increased BACE-1 mRNA levels (Wang et al., 2008b). The same group has since confirmed miR-107 down-regulation in a new set of human AD samples (Nelson & Wang, 2010; Wang et al., 2011). The functional relevance of this finding from postmortem brain analysis is supported by reporter assays in cell culture using a subset of the predicted miR-107 binding sites and demonstrating the presence of a physiological miR-107 target sequence in the 3′-UTR of human BACE-1 mRNA. The complementary application of miRNA profiling, in situ hybridization and biochemical validation suggests that miR-107 may be involved in AD progression through regulating BACE-1 expression. Another profiling study of a small number of samples (4 AD and 4 control brains) found that, among 48 miRNAs that were differentially expressed, miR-29b but not miR-107 was decreased in AD (Nunez-Iglesias et al., 2010). Yet another profiling study reported miR-9, miR-125b and miR-128 to be increased in the hippocampal region of AD affected brains (Lukiw, 2007). Thus, there is little overlap to date in the miRNA profiles among various studies. Small samples sizes and differences in experimental approaches and assay sensitivities likely account for these disparate results.

2.2.3. Huntington's disease

Huntington's disease (HD) is the most common hereditary neurodegenerative disorder due to polyglutamine expansion characterized by motor, cognitive, and psychiatric symptoms (Gil & Rego, 2008). miRNA expression profiling in the cerebral cortex of HD affected subjects has shown considerable decrease of miR-9 and miR-9* as the disease progresses (Packer et al., 2008). miR-9/miR-9* targets two components of repressor element 1 (RE1) silencing transcription factor (REST) complex whereby miR-9 regulates REST and miR-9* regulates CoREST. REST suppresses the expression of neuronal genes in non-neuronal cells. In healthy neurons, REST is sequestered in the cytoplasm partly via complexing with the protein huntingtin (Zuccato et al., 2003). The pathogenic polyglutamine-expanded huntingtin abolishes the REST-huntingtin interaction, and consequently REST translocates to the nucleus, suppresses neuronal gene expression and might lead to neuronal death. Thus, the decreased expression of miR-9 / miR-9* in HD affected brains would cause increased REST protein, promote its nuclear translocation and contribute to neurodegeneration. The regulatory mechanism between miR-9/miR-9* and REST operates in the reverse direction as well since the promoter of the gene for miR-9/miR-9* has RE1 repressor sequence and is down-regulated by this transcription factor. The translocation of REST to the nucleus in HD brains explains the reduced level of miR-9/miR-9* in this disease. The decreased amount of miR-9/miR-9* can in turn lead to increased REST/CoREST expression, which is again toxic to neuronal cells, thereby creating a positive feedback loop that contributes to the relentless neurodegeneration of HD. Another microRNA implicated in HD is miR-132, which is down-regulated in the brains of HD mouse models as well as in postmortem brains of HD patients (Johnson et al., 2008). Accordingly, its mRNA target p250GAP, encoding a member of the Rac/Rho family of GTPase-activating proteins, which inhibits neurite growth (Vo et al., 2005), is increased. Identifying additional target mRNAs of miR-9/miR-9* and miR-132 can lead to further insight into the molecular pathogenesis of HD

2.2.4 Amyotrophic lateral sclerosis (ALS)

ALS, also known as Lou Gehrig's disease, is an aggressive and deadly motor neuron disease caused by the degeneration of neurons located in the ventral horn of the spinal cord and in the brainstem. The clinical presentation includes muscle atrophy and weakness of limbs and respiratory muscles as well as bulbar signs and symptoms (Bruijn et al., 2004). Two percent of these cases are due to mutations in the superoxide dismutase 1 (SDO1) gene. To identify miRNAs that are important in the progression of this disease, Williams et al. carried out miRNA profiling studies (Williams et al., 2009) using an ALS mouse model, which expresses a mutant form (G93A) of human SOD1 (Son et al., 2007). miRNA expression in skeletal muscles from the lower limbs of normal adult mice and G93A-SOD1 transgenic mice were compared. Among 320 miRNAs tested, the muscle-specific miRNA miR-206 was the most dramatically up-regulated in G93A-SOD1 muscles. Since the increased expression of miR-206 in this ALS model could be either a contributor to disease progression or a compensatory attempt, miR-206-deleted mice were generated to directly examine the role of miR-206 in this ALS mouse model. While miR-206-deficient mice formed normal neuromuscular synapses during development, lack of miR-206 in the ALS mouse model accelerated progression of the disease phenotype, suggesting that miR-206 is protective in ALS, and that the increased miR-206 expression in the ALS mouse muscles represents a compensatory effort. In normal mice, miR-206 is required for efficient regeneration of neuromuscular junctions after nerve injury, which is a likely explanation for its beneficial effects in the ALS model. The downstream beneficial effects of miR-206 might be partly through histone deacetylase 4 (HDAC4) and fibroblast growth factor (FGF) signaling pathways. miR-206 suppresses the expression of HDAC4 in muscle cells in which HDAC4 generally obstructs reinnervation by preventing the expression of fibroblast growth factor binding protein 1 (FGFBP1). Secreted FGFBP1 is believed to boost the activities of FGF-7, FGF-10 and FGF-22 on distal terminals of motor neurons and subsequently enhance innervation (Beer et al., 2005). These observations suggest that miR-206 could delay ALS progression by recognizing motor neuron injury and promoting the compensatory regeneration of neuromuscular synapses. Thus, interventions that promote innervation of muscle, such as by delivering miR-206 itself, small-molecule mimetics of miR-206, or inhibitors of HDAC4 pathways, could be exploited for the development of potentially disease modifying therapies for ALS.

2.2.5 Prion diseases

Transmissible spongiform encephalopathies (TSE) encompass a set of neurodegenerative diseases, such as bovine spongiform encephalopathy (BSE) in cattle, Scrapie in sheep, and Creutzfeldt-Jakob disease (CJD) in humans (Chesebro, 2003). The causative agent is a pathogenic isoform of the cellular prion protein (Cohen et al., 1994) and, therefore, these conditions are collectively called prion diseases. To determine if miRNA dysfunction is involved in prion disease pathogenesis, Montag et al. analyzed the differential expression of miRNAs in the brains of BSE-infected cynomolgus macaques as a model for Creutzfeldt-Jakob disease (CJD) in humans (Montag et al., 2009). Through miRNA-microarrays and qRT-PCR assays, miR-342-3p and miR-494 were found to be significantly up-regulated in the brains of BSE-infected macaques compared to non-infected animals. In addition, miR-342-3p was also found to be up-regulated in brain samples of human type 1 and type 2 sporadic CJD. Given that miR-342-3p is also up-regulated in Scrapie-infected mice (Saba et al., 2008), the increase in miR-342-3p levels could be a common event in late stage prion disease and might be used as a biomarker for these conditions. Although experimental validation regarding target mRNA(s) for miR-342-3p has not been done, putative target genes predicted by the Targetscan algorithm include, among others, ataxin7, huntingtin interacting protein 1 (HIP1) and tau tubulin kinase 2, which are genes involved in protein aggregation disorders. Further validation of the relevance of miR-342-3p in prion diseases is needed considering the small number of experimental animals and patient brains (two sCJD and one healthy control) included in this study.

Although miRNA profiling studies in diseased brains vs age-matched controls can be a starting point for understanding how these small RNAs relate to the pathogenic mechanisms of neurodegenerative diseases, this approach has several drawbacks. One shortcoming is the fact that there is little consistency in the results among different studies as shown in Table 1. A number of factors can potentially explain these diverse results including the methods used for detecting miRNAs ranging from microarray analyses, quantitative PCR, or northern blotting, postmortem handling of brain samples and autolysis, the technique of miRNA isolation and, importantly, the neuropathological and clinical stage of the disease. Additionally, since some miRNAs are expressed specifically in certain brain regions, cell types and even specific subcellular locations (Farh et al., 2005; Junn et al., 2009), caution should be exercised in interpreting results from brain tissue homogenates. Variability in miRNA expression among different brains is another obstacle which is circumvented in cancer research where both cancerous and control tissue samples are obtained from the same individual allowing internal normalization. Considering these limitations with current miRNA profiling studies in the brain, improvements in the precision of the approach are needed to generate reliable data that can be replicated. For example, the multiplicity of various cellular phenotypes within the brain and the diverse miRNA expression patterns in different neuronal populations hinder the analysis of results obtained from homogenized brain lysates. These difficulties can be overcome by capturing the appropriate target cells from dissected tissues using UV-laser microdissection techniques (Pietersen et al., 2011).

Table1. Differentially expressed miRNAs in neurodegenerative disorders.

Disease	Decreased miRNAs	Increased miRNAs	mRNA target	Reference
AD	miR-29a/29b-1, miR-9		BACE1	Hebert et al. 2008
				Shioya et al. 2010
				Nunez-Iglesias 2010
AD	miR-107, miR-103, miR-23b		BACE1	Wang et al. 2008b
				Nelson & Wang 2010
				Wang et al. 2011
AD		miR-9, miR-138, miR-125b	N/A	Lukiw 2007
AD	miR-9	miR-29a/29b-1	N/A	Cogswell et al. 2008
PD	miR-133b		Pitx3	Kim et al. 2007
PD	miR-34b/34c		N/A	Minones-Moyano et al. 2011
HD	miR-9/9*		REST/CoREST	Packer et al. 2008
HD	miR-132	miR-29a, miR-330	p250GAP	Johnson et al. 2008
ALS	miR-206		HDAC4	Williams et al. 2009
Prion		miR-342-3p, miR-494	N/A	Montag et al. 2009
Prion		miR-342-3p, miR-320, let-7b/7d	N/A	Saba et al. 2008
		miR-328, miR-191, miR-370
	miR-338-3p, miR-337-3p

Profiling studies pertinent to ALS and Prion diseases were carried out in animal models.

2.3 Disease-causing genes and miRNAs

In several neurodegenerative disorders there is a clear gene-dosage effect that correlates with disease severity. For example, multiplication of the α-synuclein gene locus is linked to familial PD with each extra copy shifting symptom onset age by about a decade earlier (Singleton et al., 2003; Farrer et al., 2004). Individuals with amyloid precursor protein (APP) gene duplication or trisomy 21 (Down syndrome) develop young onset AD (Podlisny et al., 1987; Rovelet-Lecrux et al., 2006). These clinical genetic correlates suggest that augmented protein expression of α-synuclein or APP is sufficient to cause the neuropathology associated with these respective diseases. Therefore, it is conceivable that depletion of certain miRNA species targeting these mRNAs is responsible for upregulated expression of these genes, subsequently influencing the pathogenesis of neurodegenerative disorders. Several recent reports support this notion. For example, miR-7, which is expressed mainly in neurons, represses α-synuclein expression post-transcriptionally and protects cells against oxidative stress (Junn et al., 2009). Further, the expression of miR-7 is decreased in the MPTP intoxicated mouse model of PD, which may contribute to the increased α-synuclein expression in this model. The role of miR-7 in down-regulating α-synuclein has been confirmed in several cellular systems (Doxakis, 2010). miR-153 is another miRNA shown to repress α-synuclein expression (Doxakis, 2010). Despite these functional studies in experimental systems, it remains to be determined whether the expression of these miRNAs is altered in PD brains compared to normal age-matched controls.

The functional interaction between specific miRNAs and a disease-causing gene has also been demonstrated in AD. The 3′-UTR of APP mRNA is a potential target for several microRNAs. Experimental evidence has been developed showing that miR-106a and miR-520c decrease the expression of reporter genes containing the APP 3′-UTR in cultured cells and reduce APP levels by 50% (Patel et al., 2008). Other microRNAs reported to repress APP expression include miR-16 (Liu et al., 2010) and miR-101 (Vilardo et al., 2010; Long & Lahiri, 2011). However, the expression levels of these miRNAs in AD affected brains are presently unknown and, therefore, their participation in the pathological mechanisms of the disease remains preliminary at this time.

Another line of evidence suggesting a role of miRNA in regulating the expression of a disease-causing gene comes from observations in spinocerebellar ataxia type 1 (SCA1). This polyglutamine expansion disorder is characterized by the demise of cerebellar Purkinje cells linked to the expansion of a translated CAG repeat in ataxin1 (ATXN1) (Orr & Zoghbi, 2007). The level of polyQ-expanded ataxin1 is one of the factors contributing to the severity of the phenotype in this disorder. Based on experiments in transfected cultured cells, three miRNAs, miR-19, miR-101 and miR-130, have been shown to directly target the expression of ATXN1 mRNA (Lee et al., 2008). miR-101 and miR-130 are expressed at substantial levels in Purkinje cells. Rescuing the ATXN1 transcript from miRNA-mediated repression by mutating the target sites within its 3′-UTR exacerbates the cytotoxicity of ATXN1, suggesting that these miRNAs expressed at constitutive levels reduce cytotoxicity by lowering the levels of the pathogenic ATXN1. Although a number of genes that are critical in neurodegenerative diseases are clearly subject to miRNA regulation in cell culture-based experiments, evidence for the role of specific miRNAs in the pathogenesis of these disorders remains somewhat elusive, unless their expression proves to be altered in diseased brains.

In addition to the plethora of reports showing that specific miRNAs target and regulate the expression of pathogenic genes, a recent study provided an example of the reverse whereby a disease-linked protein can regulate the miRNA pathway (Gehrke et al., 2010). Autosomal dominant mutations in leucine-rich repeat kinase 2 (LRRK2) is the commonest known genetic cause of PD (Paisan-Ruiz et al., 2004; Zimprich et al., 2004). In the Drosophila model, expressing the pathogenic mutant forms of LRRK2 (I1915T or G2019S), which exhibit increased kinase activity, de-repressed miRNA-mediated translational repression of the transcription factors E2F1 and DP and, thereby, increased dopaminergic neuronal death. On the other hand, kinase dead mutant LRRK2 did not affect gene repression. The mode of action of pathogenic LRRK2 mutants on miRNA regulation might be through their interaction with components of the RNA-induced silencing complex (RISC), such as the Drosophila homologues of Arabidopsis Argonaute 1 (Ago1) and Dicer1 (Dcr1). The translation of E2F1 and DP are repressed by let-7 and miR-184*, respectively, and increasing the levels of these two miRNAs partially rescued the toxic effect of LRRK2. Although this study suggests that translational regulation of E2F1 and DP by miRNAs is involved in LRRK2-mediated neurodegeneration, the exact mechanism by which pathogenic forms of LRRK2 inhibit the miRNA pathway still needs to be delineated. Further, whether this mechanism is operating in the mammalian system, including human PD patients, awaits further investigation.

Additional examples of disease gene products that might influence RNA metabolism come from certain types of ALS. Mutations in the genes encoding two DNA/RNA binding proteins, TAR DNA binding protein-43 (TDP-43) (ALS10 locus) and fused in sarcoma/translocated in liposarcoma (FUS/TLS) (ALS6 locus), have been identified in familial ALS patients (Sreedharan et al., 2008; Kwiatkowski et al., 2009; Vance et al., 2009). In addition, TDP-43 is a major component of the ubiquitinated inclusions in sporadic ALS (Arai et al., 2006; Neumann et al., 2006). Interestingly, both these disease-related RNA-binding proteins are found in neuronal RNA granules and with miRNA-associated complexes, providing additional support to the notion that the miRNA pathway could be involved in ALS.

Tauopathies, a group of progressive neurological disorders associated with atypical parkinsonism and cognitive deficits, are characterized by unusual intracellular deposits of the microtubule-associated protein tau. Aberrant splicing of tau exon 10 produces an imbalance between three-repeat (3R) and four-repeat (4R) tau isoforms that appear to be important in disease pathogenesis. Recently, a number of brain miRNAs including miR-124, miR-9, miR-132 and miR-137 were found to regulate 4R:3R-tau ratios in neuronal cells (Smith et al., 2011). miRNA expression analysis of temporal cortex from patients with sporadic progressive supranuclear palsy (PSP), a 4-repeat tauopathy, revealed miR-132 to be specifically down-regulated compared to controls. Further, miR-132 was shown to directly target the 3′-UTR of the neuronal splicing factor polypyrimidine tract-binding protein 2 (PTBP2), whose protein levels were augmented in PSP afflicted brains. These results demonstrate that alterations in the miR-132/PTBP2 pathway could account for the aberrant splicing of tau exon 10 in PSP.

2.4 miRNA target site polymorphisms and neurodegeneration

Single nucleotide polymorphisms (SNPs) are small genetic variations in chromosomal DNA sequences in which a single nucleotide is substituted by one of the other three nucleotides. SNPs are located frequently outside of coding regions and occur with >1% prevalence in the human population. There are ∼10-30 million SNPs in the human genome occurring on average every 100-300 bases. Sequence data deposited into the dbSNP database is assigned a unique SNP number ID, and identical records are compiled into one reference SNP cluster (rs#). SNPs can affect transcription of the primary transcript, processing of the primary and precursor microRNAs, and/or microRNA-mRNA interactions (Sayers et al., 2010).

Examples of genetic variations in miRNA target sites, which affect the interactions between miRNAs and their targets and can potentially lead to disease, are emerging. For example, a miRNA target site mutation in myostatin (GDF8) causes muscular dystrophy in sheep (Clop et al., 2006). SNPs in miRNA target sites are also associated with various human diseases ranging from cancer to PD (Sethupathy & Collins, 2008). A study of 644 families identified significant association of PD with polymorphic variations in the fibroblast growth factor 20 gene (FGF20) including in highly conserved sequences of the 3′-UTR (van der Walt et al., 2004). FGF20 is expressed in the substantia nigra and supports the survival of dopaminergic neurons (Ohmachi et al., 2000; Ohmachi et al., 2003). Recently, another SNP (rs12720208 [C/T]) located in the 3′-UTR of FGF20 mRNA was found to be strongly associated with increased risk of having PD. The latter SNP lies within a predicted binding site for miR-433, which is highly expressed in the brain (Wang et al., 2008a). The risk allele rs12720208 inhibits the binding of the mRNA with miR-433 and increases FGF20 expression. Such variation appears to increase the vulnerability to PD through augmented α-synuclein expression, which is mediated by FGF20. However, it is difficult to reconcile the pro-survival effect of FGF20 on dopamine neurons and its ability to increase α-synuclein expression. In addition, subsequent reports have failed to find evidence for an increased association between FGF20 polymorphism and PD risk, or a relationship between the rs12720208 genotype, FGF20 and α-synuclein protein levels (Wider et al., 2009; de Mena et al., 2010).

The association of AD with genetic variability in miRNA binding sites in APP and BACE-1 has been studied by direct sequencing of Belgian cohorts including 358 AD patients and 462 controls. Six variants in the 3′-UTR of APP and 29 variants in the 3′-UTR of BACE1 were found; few of these variants were limited to patients (Bettens et al., 2009). However, association studies with all common variants detected in the 3′-UTR of BACE-1 failed to identify an association with AD risk. The precise role of patient-specific variants within the 3′-UTR region of APP and BACE-1 awaits further studies.

Polymorphisms linking miRNA to neurodegeneration have also been found in frontotemporal dementia (FTD), which is a neurodegenerative disease characterized by cognitive and memory impairment with behavioral changes. FTD can be divided into two main neuropathological subtypes: frontotemporal lobar degeneration (FTLD) with neuronal and glial tau inclusions (FTLD-tau), and FTLD with neuronal cytoplasmic inclusions (NCIs) that are positive for ubiquitin and TAR DNA-binding protein (TDP-43) (FTLD-U). Loss-of function mutations in the progranulin gene (GRN), encoding a secreted growth factor on chromosome 17q21, have been found as a major cause of familial FTLD-U (Baker et al., 2006; Cruts et al., 2006; Gass et al., 2006). The identification of SNP rs5848 at a binding-site for miR-659 in the 3′-UTR of the progranulin gene as a major risk factor for FTLD-U presents another example for miRNA target site polymorphism (Rademakers et al., 2008). Support for the functional relevance of this genetic finding comes from experimental evidence showing that miR-659 targets progranulin mRNA suppressing its translation. Additionally, progranulin protein levels are decreased by 30% in the brains of FTLD rs5848 homozygous TT carriers compared to CC carriers. As expected, the amount of progranulin protein in FTLD-U patients heterozygous for rs5848 (C/T) is midway of the amount found in the two homozygous groups. Although the presence of miR-659 has been verified in the human brain using qRT-PCR, the relative expression level of this miRNA in the brains of FTLD-U patients has not been determined yet.

3. The therapeutic potential of miRNAs

RNA interference (RNAi) has been exploited with great success in knocking down the expression of selected target genes (Bumcrot et al., 2006; de Fougerolles et al., 2007). This has led to the expectation that RNAi-based therapies will be one of the major classes of drugs in the future (Blow, 2007). RNAi has helped support the notion that cells naturally use miRNA to regulate endogenous gene expression. Recently, miRNAs are also gaining attention for their therapeutic potential with far-reaching roles in neurodegenerative diseases (Mack, 2007).

Two miRNA-based therapeutic strategies are being explored in vivo: microRNA mimics and anti-microRNAs. miRNA mimics, which are small RNA molecules that resemble miRNA precursors, are used to down-regulate the expression of specific target proteins. The target protein can be any gene involved in disease pathogenesis, or the disease gene itself having a gain-of-function pathogenic mutation. The goal of this strategy is based on the hypothesis that decreasing the specific protein level itself is a protective therapeutic strategy.

As with any other novel intervention, miRNA-based therapeutics have their own set of challenges that must be overcome before assessing their efficacy in humans. One of the downfalls of manipulating the levels or function of a specific miRNA is its potential off-target effects. The possibility of overwhelming the endogenous RNAi processing machinery and, therefore, interfering with the normal functioning of the cell is another concern. The half-lives of miRNA mimics in in vivo conditions have not been studied extensively. Presumably, repetitive administration of the miRNA mimic will be required to create persistent effects. This issue can be overcome by using virally expressed miRNA. Another major challenge for this approach is the delivery of the molecule to the right cells in the body. Delivery across the blood-brain barrier poses a particular technical hurdle for brain disorders. Recently, Alvarez-Erviti et al. showed that exosomes, which are endogenous nanovesicles that transport RNAs and proteins (Valadi et al., 2007), can deliver siRNA to the brain of mice when injected intravenously (Alvarez-Erviti et al., 2011). To reduce immunogenicity, exosomes were purified from dendritic cells isolated from the same animal. Targeting to the brain was attained by engineering the dendritic cells to express Lamp2b, an exosomal membrane protein, fused to the central nervous system-specific rabies viral glycoprotein (RVG) peptide (Kumar et al., 2007) that binds to the acetylcholine receptor. After loading exosomes with BACE-1 siRNA (or GAPDH siRNA as control) by electroporation, intravenously injected RVG-targeted exosomes delivered the siRNA load specifically to neurons, microglia and oligodendrocytes in the brain, resulting in a specific gene knockdown without eliciting an immune response. This approach as a future therapeutic vehicle to deliver siRNA/miRNA is worthy for several reasons including the specific targeting of exosomes to the brain following systemic delivery, the ability to load the desired siRNA/miRNA into the exosomes, and the ability to escape an immune response. However, the possible effects of nucleic acids and proteins derived from dendritic cells and carried with the exosomes on the target cell will have to be explored further.

The second approach to miRNA-based therapeutics is to exploit anti-miRNA molecules in order to create a loss-of-function in the miRNA of interest (Meister et al., 2004). In certain conditions in which miRNAs are over-expressed, the aim would be to block these miRNAs by injecting a complementary RNA sequence that binds to and inactivates the target miRNAs. To achieve this, Stoffel's group designed “antagomirs,” which are RNA snippets conjugated to cholesterol molecules that help the RNA enter cells (Krutzfeldt et al., 2005). Intravenous administration of antagomirs resulted in an apparent decline of corresponding miRNA levels in many organs, such as liver, lung, kidney, heart, intestine, fat, skin, bone marrow, muscle, ovaries and adrenals. The silencing of endogenous miRNAs by this method was specific, efficient and long-lasting. However, antagomirs cannot cross the blood-brain barrier, but can penetrate brain cells if injected directly into the brain (Kuhn et al., 2010). Although antagomirs have been used in animal experiments without apparent side effects, it remains unclear whether they can be the delivery vehicle of choice for microRNA silencers particularly since the cholesterol moiety they contain might be harmful to humans, especially the brain.

The development of miRNA-targeting oligonucleotides with improved pharmacological activity and optimized pharmacokinetic properties holds promise as therapeutic agents against disease-associated miRNAs (Orom et al., 2006). Locked nucleic acids (LNAs) comprise a class of bicyclic conformational analogues of RNA, which exhibit high binding affinity to complementary RNA target molecules and high stability in vivo (Fluiter et al., 2003; Hutvagner et al., 2004). The backbone modifications of oligonucleotides in LNAs are used to enhance the specificity and reduce effective doses of anti-miRNA molecules (Orom et al., 2006). A striking feature of LNA-modified anti-miRNAs is their dual ability to hinder miRNA binding to its targets and reduce miRNA levels. Intravenous injection of LNA-modified anti-miRNAs targeting a liver-specific miRNA had long-lasting effects for five to seven weeks. Such studies are required to assess specifically the dosage, duration of activity and clearance of anti-miRNA molecules in in vivo models of neurodegeneration. The efficacy of LNA-based anti-miRNAs can be enhanced by the incorporation of an endonucleolytic DNAzyme motif (Jadhav et al., 2009). These antagomiRzymes result in the cleavage of the targeted miRNA, ensuring more efficient clearance. However, the latter strategy has not been tested in in vivo models.

Another experimental strategy developed in recent years to inhibit miRNA function is to exploit synthetic sponge mRNA, which contains complementary binding sites to a miRNA of interest (Care et al., 2007; Ebert et al., 2007). When the sponge is expressed from transgenes, it specifically hampers the activity of a family of miRNAs having a common seed sequence, resulting in the sequestration and blocking of a whole family of related miRNAs. The miRNA binding sites in a sponge construct either have perfect complementary or contain mismatches in the middle positions. Sponges containing sites with perfect complementary to the miRNA show less inhibitory activity than those with mismatches (Care et al., 2007; Ebert et al., 2007; Scherr et al., 2007; Bonci et al., 2008; Sayed et al., 2008; Gentner et al., 2009; Haraguchi et al., 2009), because perfect base-pairing would be vulnerable to Ago2-mediated endonucleolytic cleavage leading to the destruction of the sponge mRNAs. In contrast, sponges containing bulged sites that are mispaired opposite miRNA positions 9–12 (Ebert et al., 2007; Gentner et al., 2009) presumably form a more stable interaction with the miRNA, including miRNA complexed with Ago2 without destroying sponge mRNA. Typical sponge constructs comprise four to ten binding sites separated by a few nucleotides each. The efficiency of a miRNA sponge depends not only on the affinity of binding sites, but also on the high amount of sponge RNAs relative to the amount of the miRNA, which can be achieved by expressing the sponge from strong promoters such as the CMV promoter. In experimental paradigms, a reporter gene is placed directly upstream of the miRNA binding sites in a single mRNA, such that the expression of the reporter protein directly parallels that of the sponge mRNA. Although deletion of the gene encoding a miRNA is the ultimate method to clear its activity, the sponge method presents several advantages. First, making a sponge to miRNA is easier than knocking it out. Second, many miRNAs have the same seed sequence but are encoded by multiple distant loci and, therefore, each locus will have to be knocked out individually and the animals cross-bred to generate a strain with complete knockout. Further, as some miRNA precursors are transcribed in clusters, the deletion of a miRNA within a cluster will be difficult to achieve without affecting the others in the same cluster. Since sponges interact with the mature miRNA, their effectiveness is unaffected by the latter two situations. Although sponge technology has advantages in experimental settings, further investigations will be required to develop it into a therapeutically viable strategy.

4. Conclusion

The implications of miRNA in the pathogenesis of neurodegenerative diseases offers a novel approach for elucidating hitherto unknown molecular mechanisms underlying these diseases as well as for these molecules to be used as potential therapeutic interventions. It will be crucial to understand further the molecular/cellular mechanisms of disease initiation and progression and the detailed role of miRNA network in the sporadic forms of neurodegenerative diseases. Such knowledge will lead to opportunities to modulate miRNA activity and function to attain disease modification for these incurable disorders. Therapies for complex multigenic diseases like most neurodegenerative disorders will likely require modulation of multiple targets or cellular disease pathways. Considering that a single miRNA molecule inhibits the expression of a number of target genes, which can work together in a network within the same cellular pathway, a whole disease phenotype can potentially be altered by modulating a single miRNA molecule. In contrast, it may be challenging to modulate complex diseases with a single siRNA, which is mostly specific for only one gene. Rapid advances in the microRNA field in recent years and the anticipated acceleration of this knowledge are expected to lead to new prospects in tackling the fundamental causative factors in incurable neurodegenerative diseases.

Abbreviations

AD

Alzheimer's disease

AGO

Argonaute

ALS

Amyotrophic lateral sclerosis

APP

amyloid precursor protein

BACE-1

β-amyloid precursor protein

BSE

bovine spongiform encephalopathy

CJD

Creutzfeldt-Jakob disease

DGCR8

DiGeorge syndrome critical region 8

FGF

fibroblast growth factor

FGFBP1

FGF binding protein 1

FTD

frontotemporal dementia

FTLD

frontotemporal lobar degeneration

FUS/TLS

fused in sarcoma and in liposarcoma

HD

Huntington's disease

HDAC4

histone deacetylase 4

HIP1

huntingtin interacting protein 1

miRNA

microRNA

LNA

locked nucleic acid

LRRK2

leucine-rich repeat kinase 2

PD

Parkinson's disease

PSP

Progressive Supranuclear Palsy

RE1

repressor element 1

REST

RE1 silencing transcription factor

RISC

RNA-induced silencing complex

SCA

spinocerebellar ataxia type

SMN

spinal motor neuron

SNP

single nucleotide polymorphism

SOD1

superoxide dismutase 1

TDP-43

TAR DNA binding protein

TSE

transmissible spongiform encephalopathies

UTR

untranslated region