Introduction

Major depressive disorder (MDD) is one of the most prevalent psychiatric disorders, it affects about 17% of Americans during their lifetime1 and is associated with psychosocial impairment, poor quality of life, and significant disability,2 morbidity, and mortality.3-5 MDD is being diagnosed at early ages, and about 25% of people diagnosed with MDD are under 19 years old. Although much work has been done to characterize MDD, about 40% of MDD patients do not respond to the currently available medications.6 This is partially a result of poor understanding of the molecular pathophysiology underlying MDD.

As is well known, compromised neural and structural plasticity are intimately associated with MDD.7,8 This is evident from studies in MDD subjects showing altered structural and functional plasticity,9-12 changes in the synaptic circuitry,13 decreased dorsolateral prefrontal cortical activity,14,15 impaired synaptic connectivity between the frontal lobe and other brain regions,16,17 changes in number and shape of dendritic spines,18,19 the primary location of synapse formation, altered dendritic morphology of neurons in the hippocampus, decreased length and number of apical dendrites,20 neuronal atrophy and decreased volume of the hippocampus,21,22 decreased number of neurons and glia in cortical areas,23 and spatial cognition deficits.24. In addition, MDD is associated with a negative impact on learning and memory,25,26 while stress, a major factor in depression, hinders performance of hippocampal-dependent memory tasks and impairs the induction of hippocampal long-term potentiation (LTP).27

The cellular mechanisms that underlie such compromised neural plasticity and structural impairments in MDD are not clearly understood, and no single mechanism appears to be responsible for its etiopathogenesis; however, it is becoming increasingly evident that MDD may result from disruptions across whole cellular networks, leading to aberrant information processing in the circuits that regulate mood, cognition, and neurovegetative functions.7 In fact, evidence demonstrating impaired cellular networks that regulate neural plasticity has reshaped our views about the neurobiological underpinnings of MDD.28

In recent years, the emergence of small noncoding RNAs as a mega-controller and regulator of gene expression has gained much attention in various disease pathophysiologies. These small noncoding RNAs regulate gene expression by several mechanisms, including ribosomal RNA modifications, repression of mRNA expression by RNA interference, alternative splicing, and regulatory mechanisms mediated by RNA-RNA interactions. The group of small noncoding RNAs include: miRNAs, small nucleolar RNAs, small interfering RNAs, piwi-interacting RNAs, splisosomal RNAs, and p/MRP RNase genes.29,30 Among them, miRNAs are the most studied and well characterized; they have emerged as a major regulator of neural plasticity and higher brain functioning,31,32 regulate about 60% of total mammalian RNAs, and are involved in virtually all biological functions. By modulating translation and/or stability of mRNA targets in a coordinated and cohesive fashion, they are able to regulate entire genetic circuitries.33 It has been shown that a combination of miRNAs is a much more powerful regulator than individual miRNAs. Interestingly, differential coexpression of a group of miRNAs has not only been shown to play a direct role in human disease pathogenesis, but can also help in identifying the nature of disordered pathways implicated in such pathogenesis.34-37

miRNAs are expressed highly in neurons, and because they can regulate the expression of a large number of target mRNAs, neuronal miRNA pathways can create an extremely powerful mechanism to dynamically adjust the protein content of neuronal compartments, even without the need for new gene transcription.38,39 miRNAs have been extensively studied in cancer biology; however, a large body of evidence demonstrates their role in several neuropsychiatric diseases, such as schizophrenia, autism, Parkinson's disease, Huntington's disease, Tourette's syndrome, Fragile X syndrome, DiGeorge syndrome, Down syndrome, and Alzheimer's disease. Studies are now being geared to examine if mutations in genes that encode miRNAs or various components of miRNA biogenesis machinery can lead to aberrant miRNA synthesis and target genes that can be linked to specific disease pathophysiology. Knowledge of the role of miRNAs in MDD is still in its infancy; however, several lines of evidence clearly demonstrate that miRNAs may play a major role in the development of stress-related disorders, including MDD. The aim of this review is to critically evaluate the role of miRNAs in MDD pathogenesis and examine whether miRNAs can be developed as biomarkers for depression.

miRNA biogenesis and regulation of target mRNA expression

An overview of miRNA biogenesis is depicted in Figure 1. As shown, miRNA biogenesis occurs in the nucleus. miRNAs are encoded within primary miRNA (primiRNA) gene transcripts that may be intergenic (away from known protein-coding genes) or may be located within introns of protein-coding host genes (intragenic).

miRNA genes are transcribed to long primary miRNA by RNA polymerase II or III. The pri-miRNAs are then processed further within the nucleus to form one or a series of small hairpin miRNA precursors, or precursor miRNAs (pre-miRNAs), that are generally 60 to 100 nt long and fold into a stem-loop structure. The protein that converts pri-miRNA into pre-miRNA is an RNase III enzyme, Drosha. Generally, Drosha requires the DiGeorge syndrome critical region 8 (DGCR8) protein as a cefaclor for activation. Together with DGCR8, Drosha forms a large complex known as the “microprocessor complex.” Drosha removes the flanking segments and ≈ 11 base pair (bp) stem region of the pri-miRNA. The pre-miRNAs are then transported out of the nucleus via the exportin transfer system, which consists of Exportin 5 and guanosine triphosphate -bound Ran (RanGTP). Pre-miRNA is released into the cytoplasm upon hydrolysis of GTP to GDP. The premiRNAs are further processed in the cytoplasm by the RNase III enzyme Dicer, which coverts pre-miRNA into double-stranded mature small RNA (miRNA/miRNA* duplexes) of approximately 22 nucleotides (nt) long.40 Dicer requires cofactors such as HIV-1 transactivating response (TAR) RNA-binding protein (TRBP) or protein kinase R (PKR)-activating protein (PACT). One of the miRNA/miRNA* duplexes is loaded onto an Argonaute (Ago) homologue protein (isoform of the eukaryotic translation initiation factor [eIF] 2C) to generate the effector complex, known as RNA-induced silencing complex (RISC). The other miRNA* strand is degraded.

Figure 1.
Figure 1.: miRNA biogenesis. microRNAs (miRNAs) are encoded in the genome (inter or intragenis) and transcribed by RNA polymerase II (RNA pol II) to generate primary microRNA (pri-miRNA). These pri-miRNAs are taken up by the RNA II enzyme Drosha/DiGeorge syndrome critical region 8 (DGCR8) protein, which catalyzes the formation of precursor miRNA (pre-miRNA). Pre-miRNA is then exported to cytoplasm by ExportinB in conjunction with Ras-related nuclear protein, RanGTR In the cytoplasm, pre-miRNA is cleaved into miRNA:miRNA* duplex by the Rnase III enzyme Dicer/Tar RNA-binding protein (TRBP) or PKR activating protein (PACT). One of these miRNA/miRNA* duplexes is discarded and the other one is loaded onto an Argonaute (Ago) homologue protein (isoforms of eukaryotic translation initiation factor [eIF] 2c) to generate the effector complex, known as RNA-induced silencing complex (RISC). The RISC complex directs miRNA to specific “short-seed” sequences located predominantly within the 3' UTR region of the target messenger RNA (mRNA). This leads to degradation or translational inhibition. nt, nucleotides

miRNA-mediated regulation of target mRNAs and expression

RISC binds to specific “short-seed” sequences located predominantly within the 3' untranslated region (3' UTR) of target mRNAs, and can interfere with the translation of mRNA and/or reduce mRNA levels. miRNA-mediated translational inhibition also depends upon the 5' cap region of the target mRNA. Ago proteins can stimulate miRNA-dependent translation inhibition by competing with efF4E for the 5' cap binding site, thus preventing circularization of mRNA and lowering initiation efficiency.41 Although miRNAs target transcripts through imperfect base-pairing to multiple sites in 3' UTRs, Watson-Crick base-pairing to the 5' end of miRNAs, especially to the so-called “seed” that comprises nucleotides 2 to 7, is also crucial for targeting.42 This provides a mechanism by which one miRNA can target several mRNAs. RISC can also associate with both the 60S ribosome and eIF6.43 eIF6 regulates the formation of the translationally active SOS subunit. By regulating eIF6, miRNAs can modify polysome formation and expose target mRNAs for degradation.43 In addition to the direct sequence-specific interaction of RISC with mRNAs, other proteins that bind nearby sites within the 3' UTR (eg, fragile X mental retardation protein [FMRP] homologues, Hu protein B [HuB] family members, and other adenylate-uridylate-rich element [ARE]-binding proteins) may control the magnitude and even the direction of miRNA effects. In certain circumstances (eg, depending on the phase of the cell cycle in dividing cells, which possibly reflects reversible phosphorylation or methylation of FMRP homologues), miRNAs may actually enhance, rather than inhibit, translation.44 miRNA-mediated regulation of mRNA stability is another mechanism by which miRNAs suppress expression of specific mRNA. Using miR-125b and let-7 as representative miRNAs, Wu et al45 showed that in mammalian cells the reduction in mRNA abundance is a consequence of accelerated deadenylation, which leads to rapid mRNA decay. Besides regulating translational processes, it has been shown that miRNA can also regulate gene transcription by targeting transcription factors. In this case, levels of transcription factors are downregulated by miRNAs, which in turn cause less expression of mRNA, leading to reduced protein synthesis.46,47

Recent evidence suggests that miRNA biogenesis can be regulated at the epigenetic level.48 For example, inhibitors of DNA methylation and histone deacetylases can affect expression of several miRNAs.49 On the other hand, a subset of miRNAs can control the expression of epigenetic regulators, such as DNA methyltransferases, histone deacetylases, and polycomb group genes, leading to transcriptional activation of numerous protein coding gene sequences, thereby contributing to gene expression. This network of feedback between miRNAs and epigenetic pathways appears to form an epigenetics-miRNA regulatory circuit, and to organize the whole gene expression profile.50

The expression of miRNAs is tissue-specific51-53 and, in some cases, even cell-type-specific.54-56 In addition, some of the miRNAs are expressed specifically at the developmental stages.57,58 Approximately 20% to 40% of miRNAs in the brain are developmentally regulated.59,60 For example, miR-124a, which is conserved at the nucleotide level and is important for neuronal differentiation, neurite outgrowth, and glucocorticoid receptor (GR)—mediated functions, is expressed throughout embryonic and adult brain.61,62 There are studies which suggest that miRNAs, such as miR-124 and miR-128, are primarily expressed in neurons, whereas miR-23, miR-26, and miR-29 are expressed in high amounts in astrocytes.63 A recent study by He et al64 suggests that a large number of miRNAs show distinct profiles in glutamatergic and GABAergic neurons and subtypes of GABAergic neurons. Even within neurons, it has been demonstrated that some of the pre-miRNAs are highlyexpressed in the dendrites where they can be locally transcribed into mature miRNAs65 and can locally regulate mRNA translation. These include synaptically enriched miRNAs: miR-200c, miR-339, miR-332, miR-318, miR-29a, miR-7, and miR-137.65,66 Several of the miRNAs are also expressed in the exons and presynaptic nerve terminals; some of them (miR-16, miR-221, miR-204, miR-15b) are highly expressed in distal axons compared with cell bodies.67 Moreover, a number of miRNAs encoded by a common pri-miRNA were differentially expressed in the distal axons, suggesting that there is a differential subcellular transport of miRNAs derived from the same coding region of the genome.67 Interestingly, the entire transcriptional machinery involved in the synthesis of mature miRNA has been shown to be localized in the synaptic fraction.65,68,69

Given their differential nature of expression and because of the local regulation of mRNA translation, the role of miRNAs in numerous biological phenomena, including neuronal development, cell proliferation, cell cycle, neurogenesis, synaptic development, axon guidance, and neuronal plasticity, have been studied.70-73

miRNAs: potential regulators of neurogenesis and neural plasticity

miRNAs in neurogenesis

Adult neurogenesis, the process of generating new neurons from neural stem cells, plays a critical role in synaptic plasticity, learning and memory, and mood regulation.74,75 In the mammalian brain, neurogenesis occurs throughout adulthood in the hippocampus (subgranular zone [SGZ] of the dentate gyrus) and olfactory bulb (subventricular zone [SVZ]). Neurogenesis in the SVZ is important for olfactory learning,76 whereas hippocampal neurogenesis is involved in memory and spatial learning.77 Numerous studies suggest that stress and MDD are associated with decreased hippocampal neurogenesis.78-80 Additionally, disturbed adult neurogenesis, possibly resulting in a malfunctioning of hippocampus, may contribute to cognitive deficits.81 Conversely, hippocampal neurogenesis buffers stress responses and depressive behavior.82 Enriched environment, exercise, electroconvulsive therapy, deep brain stimulation, and antidepressants increase hippocampal neurogenesis.83 The regulatory factors that control adult neurogenesis are currently under investigation; however, recent studies demonstrate that miRNAs play a role in both embryonic as well as adult neurogenesis.84 For example, Choi et al56 demonstrated that olfactory tissues express more than 100 distinct miRNAs, the most abundant being the miR-124a and let-7 variants and the family of miR-200. To determine whether miRNAs are required during olfactory neuronal development, these investigators analyzed embryonic tissues in which Dicer function was specifically ablated in olfactory progenitor cells. They showed that the loss of miRNA function from olfactory progenitor cells produced no alterations in patterning. In contrast, they noted that terminal differentiation of the olfactory progenitor pool into mature olfactory neurons does not occur and that the olfactory precursor cell population is not maintained. Dicer depletion also impacts proliferation and cell death, migration, and differentiation during corticogenesis as assessed by McLoughlin et al85 in the developing brain. Using markers for proliferation and in vivo labeling, they showed reduced numbers of proliferating cells, altered cell cycle kinetics from embryonic day 15.5 (E15.5), distributed progenitor cells throughout the cortex (rather than restricted to the SVZ and ventricular zones), and increased cortical cell death as early as E15.5. DGCR8 heterozygous mice also show reduced cell proliferation and neurogenesis in adult hippocampus.86 Pathania et al87 recently showed that timely directed overexpression of miR-132 at the onset of synaptic integration using an inducible approach leads to a significant increase in the survival of newborn neurons in SVZ. These effects are mirrored with respective changes in the frequency of γ-aminobutyric acid (GABA)ergic and glutamatergic synaptic inputs reflecting altered synaptic integration. The results suggest that miR-132 forms the basis of a structural plasticity program seen in SVZ-olfactory bulb postnatal neurogenesis. Cheng et al88 showed that miR-124 is an important regulator of the temporal progression of adult neurogenesis. They found that knockdown of endogenous miR-124 maintained purified SVZ stem cells as dividing precursors, whereas ectopic expression led to precocious and increased neuron formation in mice. They identified the SRY-box transcription factor Sox9 as a physiological target of miR-1 24 during the transition from the transit-amplifying cell to the neuroblast stage. The overexpression of Sox9 abolished neuronal differentiation, whereas Sox9 knockdown led to increased neuron formation. Thus, miR-124-mediated repression of Sox9 is important for progression along the SVZ stem cell lineage to neurons. Bruno et al89 identified brain-specific miRNA: miR-128, that represses nonsense-mediated RNA decay machinery, which controls transcripts of a battery of target genes to regulate neurogenesis and neural differentiation. More recently, Åkerblom et al,90 using a transgenic reporter mouse, found that miR-124 expression is initiated in the rapidly amplifying progenitors and remains expressed in the resulting neurons. Inhibition of miR-124 in vivo results in the blockade of neurogenesis, leading to the appearance of ectopic cells with astrocyte characteristics in the olfactory bulb. Conversely, neural stem cells are not maintained in the SVZ, when miR-124 is overexpressed, resulting in a loss of neurogenesis. These results suggest that miR-124 is a neuronal fate determinant in the SVZ. miR-137, which is epigenetic ally regulated by DNA methyl-CpG-binding protein, can modulate proliferation and differentiation of adult neural stem cells such that overexpression of miR-137 promotes, whereas its reduction enhances proliferation of adult neural stem cells.91 Recently, Zhang et al92 identified a new cerebellum-enriched rno-miR-592, which plays an important role in embryonic neurogenesis and/or astrogliogenesis. By using gain-/loss-of-function approaches, they demonstrated that rno-miR-592 could change the balance between neuron- and astrocyte-like differentiation and neuronal morphology. miR-592 could induce astrogliogenesis differentiation arrest and/or enhance neurogenesis in vitro, whereas silencing of miR-592 was not beneficial for neuronal maturation. They also identified LRRC4C and NFASC as miR-592 target genes, suggesting that these two target genes may be involved in miR-592 regulative function in neuronal progenitor cell differentiation and neuronal maturation.

miRNAs in neural plasticity

As mentioned earlier, miRNAs play a critical role in regulating synaptic and neural plasticity. It has been shown that by knocking down components of the miRNA synthesis machinery such as Dicer leads to a reduction in neuronal size and branching as well as aberrant axonal pathfinding.93-95 On the other hand, DGCR8 knockout mice show loss of synaptic connectivity and reduced number and size of the dendritic spines.96,97 At the behavioral level, these mice display impaired spatial working memory-dependent tasks.97 FMRP, which regulates protein synthesis in dendritic spines after binding to specific sites within the 3'UTR of certain mRNAs in concert with RISC components Agol and Dicer,98,99 is associated with learning, and LTP. Both FMRP and RISC complex components are localized in the somatodendritic compartment100,101 and FMRP associates strongly with several components of the miRNA machinery, including mature miRNAs, pre-miRNAs, Dicer, and eIF2c. One of the RISC proteins, armitage, is essential for LTP and synaptic protein synthesis and is cleaved during the learning process.102 FMRP is also associated with miR-125b and miR-132 in the brain. miR-132 overexpression increases dendritic protrusion as well as branching,103 whereas miR-125b targets NR2A mRNA and regulates synaptic plasticity in a negative fashion.104 Similar negative regulation of the size of dendritic spines in rat hippocampal neurons has been shown to be associated with miR-138 as well as miR-134. miR-138 controls the expression of acyl-protein thioesterase 1 (APT1), an enzyme regulating the palmitoylation status of proteins that are known to function at the synapse.66 On the other hand, miR-134 inhibits translation of Lim-domain-containing protein kinase 1 (Limkl),105 a protein that regulates dendritic spine growth.106 Exposure to brain-derived neurotrophic factor (BDNF) relieves Limkf translation suppression caused by miR-134. miR-134 can also promote dendritogenesis by inhibiting the translational repressor Pumilio 2.107 Interestingly, BDNF is lower in the brain of depressed subjects.108,109 Recently, Cohen et al110 showed that miR-485 regulates dendritic spine number in an activity-dependent manner, in conjunction with synaptic vesicle protein SV2A. miRISC protein Mov 10, an RNA helicase that associates with the Argonaute protein, is present at synapses and regulates synaptic expression of calmodulin (CaM) kinase II and Limkl.111

Several studies demonstrate that Creb, a key transcription factor regulating synaptic plasticity and whose expression is lower in specific brain regions of MDD subjects,112 is one of the major targets of a number of miRNAs. Conversely, many miRNAs have binding sites in their promoter regions for Crebl.114 Expression of miR-132, which enhances neurite outgrowth, dendritic morphogenesis, and spine formation,103,113,115 is induced by BDNF via Creb. Another miRNA, miR-124, which plays a critical role in maintaining neuronal cell identity, is a major target of Creb. Interestingly, miR-124 is rapidly and robustly regulated by serotonin, which selectively affects mature miR-124 levels, without affecting its precursor, suggesting that the miR-1 24 level may be regulated during Dicer processing or RISC incorporation and stabilization by Ago.116 miR-124 responds to serotonin by de-repressing Creb and thereby enhances serotonin-dependent long-term facilitation.116 More recently, it has been shown that expression of SIRT1, which modulates synaptic plasticity and memory formation, is regulated via Creb, which itself is translationally repressed by miR-134.117 On the other hand, SIRT1 inhibits the expression of miR-134 via a repressor complex containing the transcription factor YY1. Unchecked miR-134 expression after SIRT1 deficiency may result in reduced expression of Creb and BDNF, whereas knocking down miR-134 rescues LTP and memory impairment caused by SIRT1 deficiency.117

Impey et al115 have shown that CREB- and activity regulated miR-132 is necessary and sufficient for hippocampal spine formation. Expression of the miR-132 target, p250GAP, is inversely correlated with miR-132 levels and spinogenesis. Furthermore, knockdown of p250GAP increases spine formation while introduction of a p250GAP mutant that is unresponsive to miR-132 attenuates this activity. Inhibition of miR-132 decreases both miniature excitatory postsynaptic current (mEPSC) frequency and the number of glutamate receptor 1 (GluRl)-positive spines, while knockdown of. p250GAP has the opposite effect. Additionally, the miR-132/p250GAP circuit regulates Ras-related C3 botulinum toxin substrate 1 (Rac1) activity and spine formation bymodulating synapse-specific Kalirin7-Rac1 signaling. These results suggest that neuronal activity regulates spine formation, in part, by increasing transcription of miR-132, which in turn activates a Rac1-Pak actin remodeling pathway.115 Behaviorally, it has been shown that overexpression of miR-132 in the rat perirhinal cortex impairs short-term recognition memory, which is associated with a reduction in both long-term depression and long-term potentiation, and could be predicted from the excitatory and dendritogenic effects of mir132.118

Long-lasting changes at synapses are at the core of brain activities, which primarily rely on enduring changes in synaptic efficacy.119,120 Such synaptic efficacy is critically dependent upon the regulation of specific protein synthesis near or within the synapse.121-123 In this regard, miRNAs provide fascinating mechanisms to modulate synaptic efficacy and plasticity by regulating translational control of protein synthesis locally at the synapse. Lugli et al65 determined that there was a synaptic enrichment of miRNAs in mouse forebrain synaptosomes and found that significant subsets of forebrain-expressed miRNAs are highly enriched in synaptic fractions relative to total forebrain homogenate. These synaptic enriched miRNAs were biologically distinct from synaptic-depleted miRNAs, both in their expression patterns (synaptic-enriched miRNAs are expressed predominantly in pyramidal neurons, whereas synaptic-depleted miRNAs tend to have widespread and abundant tissue expression) and in their evolutionary histories (synaptic-enriched miRNAs tend to be evolutionary new and often mammalian-specific, whereas the synaptic depleted miRNAs tend to be highly conserved across vertebrates). Interestingly, they found that the synaptic enrichment was not simply related to the specificity of miRNA expression within neurons, but they arise from precursors that are expressed in the synaptic fractions and associated tightly with postsynaptic density (PSD).65 Furthermore, the synaptic enrichment of miRNAs was related to structural features of their precursors, suggesting a basis by which pre- or pri-miRNA may be selectively and stably transported to dendrites.124 Since both Dicer and pre-RNAs are expressed in synaptic fractions and are strongly associated with PSD, it suggests that at least a portion of the mature miRNAs are locally processed near synapses. Dicer is released from PSD and its RNase III activity is markedly enhanced following stimuli such as N-methyl-D-aspartate (NMDA) that can cause an increase in local calcium and activation of calpain. Dicer is expressed in PSDs, but is enzymatically inactive until conditions that activate calpain cause its liberation.65,68 These findings suggest that miRNAs are formed, at least in part, by the processing of miRNA precursors locally within dendritic spines, and synaptic stimulation may lead to local processing of miRNA precursors in proximity to the synapse. Synaptic efficacy can be regulated by modulating miRNA functions at the synapse and consequently synaptic plasticity due to the critical feature of miRNAs to regulate gene circuitry locally at the synapse in an activity-dependent fashion. This may provide a unique opportunity at the therapeutic level, where regulation of miRNA can be used to control plasticity at the synapse.

miRNAs in MDD pathogenesis and treatment

The diagnostic and prognostic values of miRNA have been established in various types of cancer.125 The potential of miRNAs as diagnostic markers for psychiatric and neurodegenerative diseases has been advancing rapidly.31,126-128 Both preclinical and clinical evidence demonstrates that miRNAs can be extensively involved in stress-related disorders and MDD, as well as the antidepressant response.

Coping response to stress and miRNAs

An individual's ability to cope with stress is critical in the development of MDD. We recently examined miRNA expression in both the frontal cortex of rats who developed behavior (learned helpless [LH]) that resembles stress-induced depression and those who did not develop depression (nonlearned helpless [NLH]), even though they received similar inescapable shocks (Table I).129 In this manner, we were able to distinguish the factors associated with the development of the depression phenotype. We found that NLH rats showed a robust adaptive miRNA response to inescapable shocks whereas LH rats showed a markedly blunted miRNA response. One set of miRNAs showed large, significant, and consistent alterations in NLH rats, consisting of miR-96, miR-141, miR-182, miR-183, miR-183*, miR-198, miR-200a, miR-200a*, miR-200b, miR-200b*, miR-200c, and miR-429. All were downregulated in NLH rats relative to tested controls (no shock group), and all showed a blunted response in LH rats (more like tested controls). These miRNAs were encoded at a few shared polycistronic loci, suggesting that their downregulation was coordinately controlled at the level of transcription. Most of these miRNAs have previously been shown to be enriched in synaptic fractions.65 Moreover, almost all of these miRNAs share 5'-seed motifs with other members of the same set, suggesting that they will hit similar or overlapping sets of target mRNAs. Interestingly, half of this set are predicted to hit Crebl as a target, and binding sites for CREB lie upstream of miR-96, miR-182, miR-183, miR-200a, miR200b, miR-200c, miR-220a*, and miR-200b*. This suggests that a similar feedback loop arrangement may also exist for Creb, similar to what has been described for other Creb-stimulated miRNAs and target genes.114 Since these miRNAs are downregulated in NLH rats, but not LH rats, this can be interpreted as a homeostatic response intended to minimize the repressive effects on Crebl. In addition, we identified a large core coexpression module, consisting of miRNAs that are strongly correlated with each other across individuals of the LH group, but not with either the NLH or tested control group. The presence of such a module implies that the normal homeostatic miRNA response to repeated inescapable shock is not merely absent or blunted in LH rats; rather, gene expression networks are actively reorganized in LH rats, which may support their distinctive persistent phenotype.

Another piece of evidence comes from studies of stress-sensitive F344 rats (which show a higher stress response to restraint stress) compared with Sprague-Dawley rats (which show lower hypothalamic-pituitary adrenal axis activity over a period of time). In this context, it is important to mention that glucocorticoids regulate the hypothalamic-pituitary adrenal axis through a negative feedback mechanism while binding to soluble GRs in the pituitary and the hypothalamus and inhibit the release of corticotropin-releasing factor and adrenocorticotropic hormone. Several studies have reported that the GR expression of is downregulated in depressed individuals.130 The GR protein is under constant miRNA regulation.131 More specifically, miR-124a and miR-18a bind to the 3' UTR of GR and downregulate its expression.131 Overexpression of miR-18a attenuates the glucocorticoid -induced leucine zipper, a gene induced by stress-like levels of glucocorticoid. It is possible that higher expression of miR-18a and consequent downregulation of GR could be responsible for the genetic susceptibility to stress in F344 rats.132 Turner et al133 recently predicted several possible miRNA binding sites within the GR first exon, suggesting further regulation of GR genes by miRNAs.

miRNAs implicated in stress and depression.

miRNAsEffectsRéférences
Restreint stressFrontal cortex: miR-9, miR-9*, miR-26b, miR-29b, miR-30b, miR-30c, rniR-30e, miR-125a, miR-126-3p, miR-129-3p, miR-207, miR-212, miR-351, miR-423, miR-487b, miR-494, miR-690, miR-691, miR-709, miR-711, and Let-7 a-e let-7a, miR-9, miR-26a/b, miR-30b/c, miR-125aRinaldi et al124
Immobilization stressHippocampus CA1, amygdala: miR-134, miR-183, miR-132, Let-7a-1, miR-9-1, miR-124a-1Meerson et al135
Unpredictable chronic mild stressHippocampus: miR-298, miR-130b, miR-135a, miR-323, miR-503, miR-15b, miR-532, and miR-125a and up-regulating miRNAs miR-7a, miR-212, miR-124, miR-139, miR-182Cao et al136
Early life stressMediaI prefrontal cortex: pre-miRs 132, 124-1, 9-1, 9-3, 212, 29aUchida et al140
Animal model of dépressionLearned helpless vs control frontal cortex: mmu-miR-184, mmu-miR-197, mmu-miR-107, mmu-miR-329, mmu-miR-125a-5p, mmu-miR-872, mmu-miR-181c, mmu-miR-18a*, mmu-miR-29b*, mmu-let-7a*, rno-Iet-7e*, rno-miR-20a*Smalheiser et al129
Postmortem brain studiesPrefrontal cortex: hsa-miR-142-5p, hsa-miR-33a, hsa-miR-137, hsa-miR-489, hsa-miR-148b, hsa-miR-101, hsa-miR-324-5p, hsa-miR-301a, hsa-miR-146a, hsa-miR-335, hsa-miR-494, hsa-miR-20b, hsa-miR-376a*, hsa-miR-190, hsa-miR-155, hsa-miR-660, hsa-rniR-552, hsa-miR-453, hsa-miR-130a, hsa-miR-27a, hsa-miR-497, hsa-miR-10a, hsa-miR-20a, hsa-miR-142-3pSmalheiser et al156
Peripheral mononuclear cellshas-miR-107, miR-133a, miR-148a, miR-200c, miR-381, miR-425-3p, miR-494, miR-517b, miR-579, miR-589, miR-636, miR-652, miR-941, miR-1243Belzeaux et al179
Whole blood cells (12 weeks of treatmentwith escitalopram)hsa-miR-130b, hsa-miR-505, hsa-miR-29b-2 , hsa-miR-26b, hsa-miR-22, hsa-miR-26a, hsa-miR-664, hsa-miR-494, hsa-let-7d, hsa-let-7g, hsa-let-7e, hsa-miR-34c-5p, hsa-let-7f, hsa-miR-629, hsa-miR-106b, hsa-miR-103, hsa-miR-191, hsa-miR-128, hsa-miR-502-3p, hsa-miR-374b, hsa-miR-132, hsa-miR-30d, hsa-miR-500, hsa-miR-770-5p, has-miR-589, hsa-miR-183, hsa-miR-574-3p, hsa-miR-140-3p, hsa-miR-335, hsa-miR-361-5pBocchio-Chiavetto et al180

miRNAs in response to stress

Several types of stressors have been utilized to examine how miRNAs respond to stressors. Interestingly, acute and chronic restrained stress cause differential changes in miRNA expression in a brain region-specific manner (Table I). 134 For example, acute stress induces a transient increase in the expression of selected miRNAs (miR-9, miR-9*, miR-26b, miR-29b, miR-30b, miR-30c, miR-30e, miR-125a, miR-126-3p, miR-129-3p, miR-207, miR-212, miR-351, miR-423, miR-487b, miR-494, miR-690, miR691, miR-709, miR-711, and Let-7 a-e) in the frontal cortex, but not the hippocampus. Some of them (let-7a, miR-9, miR-26a/b, miR-30b/c, and miR-125a) show an increase in their expression 5 days after acute stress; however, their expression levels are not altered after repeated restraint. These results suggest that acute stress modulates miRNA expression quickly to external stimuli by changing their synaptic efficacy through regulation of localized mRNA translation.

Using psychological stress (acute or chronic immobilization), Meerson et al135 examined miRNA expression in the central amygdala and the Cornu Ammonis area 1 (CA1) region of the hippocampus of rats subjected to acute or chronic immobilization stress (Table I). They found that the expression of several miRNAs was differentially altered during acute and chronic stress, with chronic stress causing much larger changes than acute stress. Some of the miRNAs that were altered during acute and chronic stress include: miR-134, miR-183, miR-132, Let-7a-l, miR-9-1, and miR-124a-l. Interestingly, except for miR-Let-7a-l, the expression of stress-responsive miRNAs were different in the two analyzed brain regions. In the CA1 region, miR-376b and miR-208 increased whereas miR-9-1 decreased under both acute and chronic stress conditions. Stressresponsive miR-134 and miR-183 target many splicing factors, such as SC35, SRP46, and SFRSll. SC35 promotes the alternative splicing of acetylcholinesterase (AChE) from the synapse-associated isoform AChE-S to soluble AChE-R protein and the expression of SC35 is increased during stress. Thus, by regulating splicing factors and their targets, miR-183 and miR-134 may modify both alternative splicing and cholinergic neurotransmission in the stressed brain. In addition, one of the targets of miR-183 is profilin 2 mRNA, which regulates dendritic spine morphology in neurons. Interestingly, neurotransmitter homeostasis and behavior are severely affected in profilin 2 knockdown mice136 and prolifin 2 (PFN2) expression is increased in lymphoblastoid cell lines of monozygotic twin pairs discordant for bipolar disorder.137

Using unpredictable chronic mild stress combined with separation, Cao et al found changes in 13 specific miRNAs in the rat hippocampus (Table I).138 These include downregulating miRNAs (miR298, miR-130b, miR-135a, miR-323, miR-503, miR-15b, miR-532, and miR-125a) and upregulating miRNAs (miR7a, miR-212, miR-124, miR-139, and miR-182). Among these, miR125a and miR-182 recovered to normal after intervention with traditional herbal antidepressant medication.

The effect of early-life stress, which is one of the critical factors in the development of affective disorders,139 on miRNA expression, has also been studied. Uchida et al found that maternal separation enhanced stress vulnerability to repeated restraint stress exposure in adulthood (Table I).140 At the molecular level, maternal separation increased the expression of repressor element-f silencing transcription factor (REST) 4. Transient overexpression of REST4 in the medial prefrontal cortex of neonatal mice produced depression-like behaviors in adults after repeated exposure to restraint stress, suggesting that REST4 may play a role in the development of stress vulnerability. REST regulates the expression of several miRNAs141,142 that are involved in brain development and plasticity.46,113,116 For example, maternal separation increased the expression of REST-associated premiRs 132, 124-1, 9-1, 9-3, 212, and 29a in rat medial prefrontal cortex.140 miR-132 and miR-124 are involved in synaptic plasticity113,116 and cause decreased dendritic length,143 high synaptic density,144 and altered basal neuronal activity.145 Interestingly, Bai et al146 demonstrated that maternal deprivation not only led to the development of depressive behavior in rats and a subsequent decrease in BDNF expression, but decreased BDNF expression was negatively correlated with the expression of miR-16. miR-16 has been shown to be an important effector of antidepressant action in serotonergic raphe and noradrenergic locus coeruleus as well as adult neurogenesis in the hippocampus.147

miRNAs and antidepressants

Several recent studies show that miRNAs may be involved in the mechanisms of action of antidepressants. Baudry et al148 showed that the serotonin transporter is a target of miR-16, which has a higher expression pattern in noradrenergic neurons than in serotonergic neurons. A reduction of miR-f 6 in noradrenergic neurons caused de novo serotonin transporter expression.

Interestingly, long-term treatment with fluoxetine to mice increased miR-16 levels in serotonergic raphe nuclei, which, in turn, reduced serotonin transporter expression. Furthermore, raphe responded to long-term fluoxetine treatment by releasing S100β, which, in turn, acted on the noradrenergic neurons of the coeruleus locus. Thus, by lowering miR-16 levels, S100β unlocked the expression of serotonergic functions in the noradrenergic brain area. This study suggests that miR-16 acts as a central effector in regulating serotonin transporter expression and mediates the adaptive response of serotonergic and noradrenergic neurons to fluoxetine treatment. These investigators further showed that fluoxetine treatment induced the secretion of various signaling molecules such as BDNF, Wnt2, and 15d-PGJ2 from raphe serotonergic neurons and acted cooperatively on the hippocampus, whereas S100β controlled the locus coeruleus-dependent hippocampal response to fluoxetine.147 These signals relayed the action of fluoxetine on adult hippocampal neurogenesis by downregulating miR-16 in the hippocampus in a region-specific manner. In a complementary fashion, these signaling molecules were found to be increased in the cerebrospinal fluid of depressed patients upon fluoxetine treatment.147

O'Connor et al149 investigated whether early-life stress in rats induced changes in hippocampal miRNA levels and whether the rapidly acting NMDA receptor antagonist ketamine, electroconvulsive therapy (ECT), or fluoxetine treatment could reverse these changes. They found that early-life stress affected expression of multiple hippocampal miRNAs. Antidepressant treatments reversed some of these effects including a stress-induced change to miR-451. Ketamine and ECT possessed the highest number of common targets, suggesting convergence on common pathways. Interestingly, all three treatments possessed miR-598-5p as a common target. This demonstrates that changes to hippocampal miRNA expression may represent an important component of stress-induced pathology, and antidepressant action may reverse these. In this context, Ryan et al150 examined ECT-induced BDNF expression and BDNFassociated miRNAs. Following acute or chronic electroconvulsive stimulation (ECS), they found that the level of selective miR-212 was significantly increased in dentate gyrus. miR-2f 2 level was also increased in parallel in whole blood following chronic ECS and was positively correlated with miR-212 level in the dentate gyrus, suggesting that miR-212 may be crucial in the mechanism of action of ECT and that it can be used as a biomarker.

Using genome-wide expression profiling of human lymphoblastoid cell lines (LCLs), Morag et al151 identified CHL1 as a selective serotonin reuptake inhibitor (SSRI) sensitivity biomarker. The same group reported that specific miRNAs may be implicated in SSRI sensitivity of LCLs.152 They examined genome-wide expression profiling with miRNAs in LCLs exhibiting high or low sensitivities to paroxetine. They found that miR-1513p had a 6.7-fold higher basal expression in paroxetinesensitive LCLs, which was correlated with lower expression of CHL1, a target of miR-151-3p. The additional miRNAs miR-212, miR-132, miR-30b*, let-7b, and let-7c also differed by > 1.5-fold between the two LCL groups. These results suggest a possible therapeutic value of miRNAs in responders vs nonresponders to antidepressant treatment in MDD patients.

Schmidt et al153 found that the therapeutic action of fluoxetine is associated with a reduction in prefrontal cortical mir-1971 expression levels in a mouse model of post-traumatic stress disorder. Long-term lithium and valproate treatment have also been shown to alter a number of miRNAs; however, 9 miRNAs (let-7b, let-7c, miR-105, miR-128a, miR-24a, miR-30c, miR-34a, miR221, and miR-144) were regulated by both lithium and valproate. The most significant signaling pathways that are targeted by these miRNAs are the PKC, PTEN, ERK-MAP kinase, Wnt/β-catenin, and β-adrenergic pathways. Some of these have been shown to be altered in MDD and bipolar disorder.154 Lithium- and valproate induced downregulated miR-128a, miR-24, and miR-34a significantly upregulated hippocampal protein levels of DPP10, GRM7, and THRB. Among these proteins, GRM7 has emerged as a novel candidate gene for bipolar disorder.155

miRNAs in MDD subjects

One of the approaches that have been taken to directly assess the status of miRNAs in psychiatric illnesses is to examine the postmortem brain. Using this approach, we recently profiled miRNA expression in the prefrontal cortex of depressed subjects who had died by suicide (Table I).156 We took several different approaches to analyzing the data. When we analyzed miRNA expression globally, we found that 21 miRNAs were significantly downregulated in the MDD group. We also found that 24 miRNAs were downregulated by 30% (although not statistically significant), suggesting a global downregulation of miRNA levels in the MDD group. When analyzed individually, we found that almost half of the downregulated miRNAs were encoded at chromosomal loci near other miRNAs and are possibly transcribed by the same pri-miRNA gene transcripts (mir-1 42 -5p and 142-3p; mir-494, 376a*, 496, and 369-3p; mir-23b, 27b and 24-1*; mir-34b* and 34c; mir-17* and 20a). In addition, three pairs of miRNAs were encoded at distances greater than 100 kb, but still lie within the same chromosomal region (mir-424 and 20b at Xq26.2-3, 377 kb apart; mir-142 and 301a at 17q22, 820 kb apart; mir-3245p and 497 at 17pl3.1, 205 kb apart). This suggests that at least some of the downregulated miRNA expression is due to decreased transcription. Many of the downregulated miRNAs also shared 5' seed sequences that are involved in target recognition. For example, identical seed sequences are shared by: (i) mir-20a and 20b; (ii) mir-301 a and 130a; and (iii) mir-424 and 497. In addition, a 6-mer nucleotide motif is shared by mir-34a, 34b*, and 34c, and strikingly, a 5-mer motif (AGUGC) within the 5' seed is shared by 5 of the affected miRNAs (mir-148b, 301a, 130a, 20a, and 20b) that is predicted to bind Alu sequences within the 3* UTR region of target mRNAs. This suggests that the downregulated miRNAs should exhibit extensive overlap among their mRNA targets. When we did pair-wise correlation (a complementary method of analyzing the miRNA expression data is to identify pairs of miRNAs that are coregulated in their expression, up or down, across individuals within a single group) we found that a set of 29 miRNAs, none of whom were pairwise correlated in the normal control group, formed a very extensive interconnected network in the MDD group. Several of the miRNAs (let-7b, mir132, 181b, 338-3p, 486-5p, and 650) were “hubs” correlated with four to nine other miRNAs in the network. Target analysis revealed that many of the targets are transcription factors, and nuclear, transmembrane, and signaling proteins. Intriguingly, four different downregulated miRNAs target VEGFA (mir-20b, 20a, 34a, and 34b*), a molecule implicated in depression in both humans and animal models. Other validated targets include BCL2 (mir-34a), DNMT3B (mir-148b), and MYCN (mir-101, 34a). Among predicted targets, estrogen receptor a, ESR1, was predicted to be targeted by three different downregulated miRNAs (mir-148b, 301a, 496). Others targeted by three or more affected miRNAs include ubiquitin ligases (UBE2D1 and UBE2W); signal transduction mediators (CAMK2G, AKAP1); the splicing factor NOVA1 that regulates brain-specific alternative splicing; the GABA-A receptor subunit GABRA4; calcium channel CACNA1C; and brain-active transcription factors including SMAD5, MITF, BACH2, MYCN, and ARID4A. Several of these predicted targets interact with validated targets; for example, ARIA4A binds E2F1; SMAD5 binds RUNX1; and estradiol treatment decreases E2F1 levels in the prefrontal cortex.157 BACH2 transcription factor binding sites have been identified upstream of many brainexpressed miRNAs.114 Retinoblastoma binding protein 1 (ARIA4A) is of interest because it recruits histone deacetylases and regulates gene expression via chromatin-based silencing.

Recently, He et al158 studied an association between miRNA processing gene variants and depression. They genotyped three polymorphisms from three miRNA processing genes (DGCR8, AG01, and GEMIN4) in a case-control study including 314 patients and 252 matched healthy controls. Frequencies of genotypes and alleles showed a significant difference between patients with depression and healthy controls in DGCR8 rs3757 and AGOl rs636832. An allele frequency was significantly higher in rs3757 and lower in rs636832, respectively. Variant allele of DGCR8 rs3757 was associated with increased risk of suicidal tendency and improvement response to antidepressant treatment, whereas the variant of AGOl rs636832 showed decreased risk of suicidal tendency, suicidal behavior, and recurrence. Besides, allele frequency showed significant difference when comparing patients with remission with controls; no significant differences were found in GEMIN4 rs7813 between patients and healthy controls. DGCR8 rs3757 and AGOl rs636832 were found to have a significant association with depression, and GEMIN4 rs7813 did not affect susceptibility to depression. These observations suggested that miRNA processing polymorphisms may affect depression risk and treatment.

Circulating miRNAs: potential biomarker in depression and antidepressant response

miRNAs can be detected in circulating biological fluids such as serum, plasma, urine, saliva, and cerebrospinal fluid (CSF).159-161 Interestingly, it has been shown that miRNAs are endogenously expressed in plasma and their expression profile is similar in blood cells.162 More interestingly, under healthy conditions, these miRNAs are stably expressed in blood cells; however, under pathological conditions, the profile of miRNAs changes significantly, suggesting the possibility that peripheral miRNAs can be used as a reliable biomarker under disease conditions. Although the source of miRNAs in peripheral cells is not clear, it has been shown that miRNAs can be released actively or passively from tissues into the circulating blood.162,163 The actively secreted miRNAs are enclosed in exosomes164 and protected by RNA binding proteins including NPM1,165 HDL,166 or Argonaute2.167 Therefore, measuring miRNAs in blood cells is highly reproducible. Interestingly, it has been shown that exosomes containing miRNAs are excreted physiologically in response to stress or brain injury, suggesting that these miRNAs can be ideal biomarker candidates.168-170 Exosomal miRNAs are processed by the same machinery used in miRNA biogenesis and thus have widespread consequences within the cell by inhibiting the expression of target protein-coding genes.169

Over the past several years, attempts have been put forward towards developing circulating miRNAs as a potential biomarker for many diseases. These include cancer and cardiovascular, inflammatory, and neurodegenerative diseases.128,171-173 In cancer patients, miRNAs have not only been shown to be useful indicators of various types of cancer, but based on miRNA profiling, it can be shown which patient group responds better to a particular treatment regimen.125 In neurological disorders, there is a significant correlation between circulating miRNAs and brain disease and injury.174 For example, miR-9 is decreased in blood cells, and in the cortex and hippocampus of Alzheimer's disease patients.175 Similarly, circulating miR-34 serves as a novel biomarker for Huntington's disease.176 Identifying biomarkers in these diseases has been a challenging task due to the heterogeneity and complex nature of psychiatric illnesses. Nonetheless, emerging studies provide evidence that miRNAs can be used successfully as biomarkers in these illnesses. For example, in schizophrenia patients, several circulating miRNAs (miR-181b, miR-219-2-3p, miR-1308, let-7g, and miR-195) have been identified as potential disease biomarkers.177 Plasma miR-134 levels in drug-free, 2-week medicated, and 4-week medicated bipolar mania patients were significantly decreased when compared with controls, and its level increased following medication.178 Decreased circulating miR-134 level, both in drug-free and medicated patients, showed a negative correlation with the clinical scales, suggesting that a decrease in plasma miR-134 levels may be directly associated with the pathophysiology and severity of manic symptoms in bipolar disorder.

The use of miRNAs as a peripheral biomarker in MDD is gaining momentum. Belzeaux et al examined miRNA expression profiles in peripheral blood mononuclear cells (PBMCs) collected from 16 severe MDD patients and 13 matched controls at baseline, and 2 and 8 weeks after treatment (Table I). 179 A comparison of miRNA expression between MDD patients and controls at baseline and at 8 weeks showed a similar number of dysregulated miRNAs (14 miRNAs, with 9 miRNAs upregulated and 5 downregulated). miRNAs that showed changes between MDD and controls at base line included: has-miR-107, miR-133a, miR-148a, miR-200c, miR-381, miR-425-3p, miR-494, miR-517b, miR-579, miR-589, miR-636, miR-652, miR-941, and miR-1 243. Only two miRNAs showed stable overexpression in MDD patients during the 8-week follow-up compared with controls (miR-941 and miR-589). They also identified miRNAs exhibiting significant variations of expression among patients with clinical improvement (7 upregulated and 1 downregulated). Fourteen dysregulated miRNAs had putative mRNA targets that were differentially expressed in MDD, suggesting that a common RNA regulatory network functions in MDD. These results suggest the potential utility of miRNA signatures as markers of major depressive episode evolution.

Bocchio-Chiavetto et al conducted a whole-miRNA ome quantitative analysis in the blood of 10 MDD subjects after 12 weeks of treatment with escitalopram (Table I).180 They found that 30 miRNAs were differentially expressed after the escitalopram treatment: 28 miRNAs were upregulated, and two miRNAs were downregulated. Thirteen (let-7d, let-7e, miR-26a, miR-26b, miR-34c-5p, miR-103, miR-128, miR-132, miR-183, miR-192, miR-335, miR-494, and miR-22) play a role in neural plasticity and stress response and in the pathogenetic mechanisms of several neuropsychiatric diseases. miR-132 exerts critical functions in the biological circuits implicated in neurogenesis and synaptic plasticity, stimulating axonal and dendritic outgrowth in different brain areas.181 This miRNA, together with miR-26a, miR-26b, and miR-183, widely contributes to the action of the neurotrophin BDNF in the brain.103,134,182,183 miR-132, miR-26a, miR-26b, miR-183, let-7d, let-7e, miR-26b, miR-103, miR-128, miR-494, and miR-22 have been reported to play a role in the pathogenesis of psychiatric disorders and in the mechanism of action of antipsychotic drugs and mood stabilizers. Moreover, postmortem studies on the brains of bipolar disorder patients showed increased levels of miR-22* in the prefrontal cortex.184 On the other hand, miR-494 and miR-335 are downregulated in the prefrontal cortex of MDD patients.156 The target genes of these altered miRNAs include: BDNF, GR, NR3C1 and nitric oxide synthase NOS1, growth factors (IGF1, FGF1, FGFR1, VEGFα, and GDNF), calcium channels (CACN41C, CACNB4, SLC6A12, and SLC8A3), and neurotransmitter receptors (GABRA4 and 5-HT4), some of which have been implicated in MDD and in the mechanism of action of antidepressants. These studies suggest that miRNAs can not only be used to diagnose, but can also be used for treatment response.

Conclusion and future directions

Based on studies showing the involvement of miRNAs in neural plasticity, neurogenesis, and stress response, it is clear that miRNAs may participate in the pathogenesis of MDD. More direct evidence comes from human postmortem brain studies showing aberrant expression of miRNAs in the prefrontal cortical area. From these, as well as animal studies showing a blunted response in NLII rats, one can assume that miRNAs may actively participate in developing the MDD phenotype.

Despite these findings, one needs to find an integrated view of miRNA networks and the pathways that are affected by these miRNAs. It is well established that a combination of miRNAs is a much more powerful regulator than individual miRNAs. Interestingly, differential c-expression of a group of miRNAs has not only been shown to play a direct role in human disease pathogenesis, but they also help in identifying the nature of disordered pathways implicated in such pathogenesis.34-37 A set of miRNAs that are significantly affected in MDD, and the corresponding set of mRNAs that are affected in the same samples, will help resolve this issue. The affected miRNAs and mRNAs are likely to interact with and regulate each other, either directly as targets or indirectly as part of larger regulatory networks. One can also identify sets of miRNAs that are not correlated in expression across individuals in the control group, yet are positively correlated in the MDD group and vice versa. There is a possibility that the correlated miRNAs and mRNAs are likely to be driven in their expression by the same (possibly overlapping) set of transcription factors or epigenetic influences. If a given miRNA is driven by one transcription factor in the control group and by a different transcription factor in the MDD group, this may result in no change in its mean expression levels across groups, yet may be detectable by observing shifts in the miRNAs that are correlated across individuals. In addition, it is important to determine whether the changes in miRNA/mRNA network are similar or different across different brain areas and more so, whether they are cell type-specific and are reversible. Also important is to examine the potential reasons for altered miRNA expression. Is it because of genetic changes in the promoter region upstream of primary miRNA gene transcripts, the pre-miRNA hairpin, or the mature miRNA, or due to RNA editing of transcripts or epigenetic suppression of the chromosomal region encoding the miRNAs? A variety of enzymes are responsible for processing miRNAs. These include Drosha, Dicer and cofactors DGCR8, TRBR and PACT. Several of these proteins have been shown to be modified post-translationally in a dynamic manner. For example, altering the relative expression of eIF2c may change the efficiency of translational arrest produced by a given miRNA.185 Recently, it has been shown that Dicer is activated by proteolytic cleavage under conditions of elevated calcium levels68,124 and eIF2C undergoes reversible phosphorylation within cells, which is required for its translocation to processing bodies.186 The phosphorylation of eIF2C appears to be due to activation of ERK1/2.186 Since we have shown abnormalities in calcium-sensing proteins and ERK1/2 signaling in the brain of MDD subjects,187-190 it will be worthwhile to ask whether Dicer cleavage patterns or eIF2C phosphorylation are altered in the MDD group. One can also examine whether there is any genetic link between miRNA and MDD. Such genetic linkage has been reported in specific miRNAs in schizophrenia.191

The development of miRNAs as biomarkers for MDD pathogenesis will be an important step in detecting MDD pathogenesis. In this regard, blood cell studies will be critical. The miRNA findings in these cells of MDD subjects so far have been very exciting, and indicate the possibility of developing biomarkers in a noninvasive manner. Also, the blood expressed miRNAs can be expanded to monitor antidepressant response. Much work, however, is needed in this direction, since miRNAs may be expressed differently in various blood cell types. In addition, whether miRNAs in a specific blood cell-type truly represent brain-derived miRNAs, is unclear at the present time. In this regard, measuring miRNAs in exosomes may be an alternate and better approach.192,193