On the functional relevance of spatiotemporally-specific patterns of experience-dependent long noncoding RNA expression in the brain

ABSTRACT

The majority of transcriptionally active RNA derived from the mammalian genome does not code for protein. Long noncoding RNA (lncRNA) is the most abundant form of noncoding RNA found in the brain and is involved in many aspects of cellular metabolism. Beyond their fundamental role in the nucleus as decoys for RNA-binding proteins associated with alternative splicing or as guides for the epigenetic regulation of protein-coding gene expression, recent findings indicate that activity-induced lncRNAs also regulate neural plasticity. In this review, we discuss how lncRNAs may exert molecular control over brain function beyond their known roles in the nucleus. We propose that subcellular localization is a critical feature of experience-dependent lncRNA activity in the brain, and that lncRNA-mediated control over RNA metabolism at the synapse serves to regulate local mRNA stability and translation, thereby influencing neuronal function, learning and memory.

KEYWORDS:

• Long noncoding RNA
• neuron
• synapse
• RNA metabolism
• learning and memory

Introduction

The mammalian transcriptome consists of a complex repertoire of RNA species. It is estimated that 98% of all RNAs transcribed from the mammalian genome are noncoding RNAs (ncRNAs)[1], which are classified based on their size, function and subcellular location [2]. Long noncoding RNAs (lncRNA), in particular, have emerged as crucial players in the regulation of gene expression. LncRNA, by definition, is any RNA that is longer than 200 nucleotides and does not have any protein-coding potential. LncRNAs were first identified in the early 1990s^3-[3]; however, due to technical limitations, their function could not be determined and the relevance of intragenic RNAs remained hypothetical [4]. After the FANTOM project was launched in the early 2000s, many more ncRNAs were identified and about one-third of cDNAs were annotated as lncRNAs [5–7]. Shortly after, a role for lncRNAs in directing chromatin modifiers and the regulation of gene expression in the nucleus was reported [7–10], firmly establishing lncRNAs as key players in cellular metabolism.

LncRNAs display a high degree of developmental and cell-type specificity, with approximately 40% being enriched in the mammalian brain [11,12]. The nuclear-enriched lncRNA Gm12371 has been shown to influence hippocampal dendritic morphology and synaptic plasticity [13], and the nuclear antisense lncRNA AtLAS regulates synapsin II polyadenylation and AMPA receptor trafficking [4]. These lncRNAs are expressed in a cell-type- and spatiotemporal-specific manner [14] and are therefore uniquely positioned to mediate rapid responses to environmental stimuli and potentially cognition [15]. In addition, many lncRNAs have been implicated in neurological disorders [16,17], and their role in neural development and in the ageing brain has also recently been investigated [18–21]. In this review, we highlight what is currently known about lncRNAs in the brain and discuss their potential role in regulating cognitive processes, as well as their involvement in neurological disorders. We address current technologies that will advance our understanding of lncRNA in the brain and propose that alternative splicing and subcellular localization are critically important features of experience-dependent lncRNA activity in neurons. Moreover, we suggest that lncRNA-mediated control over biomolecular condensates in the form of RNA reservoirs at the synapse serve to regulate local mRNA stability and translation, which may be critical for neural plasticity, learning, and memory.

LncRNA biogenesis

The biogenesis of lncRNA is very similar to that of protein-coding RNA. Although a subset of lncRNAs are RNA polymerase III-dependent [22–24], most lncRNAs are transcribed by RNA polymerase II in the opposite direction with overlapping coding exons (antisense), in the opposite direction without overlapping exons (divergent), from within introns (intronic), or between two coding genes (intergenic) [25]. In addition, lncRNAs can arise through alternative splicing of the host gene, and can themselves serve as a host for other ncRNAs [2,26–28]. Most lncRNAs possess a 5ʹ cap and a 3ʹ poly-A tail [10,29–31]; although lncRNAs without a poly-A tail have also been identified [30,31]. Because lncRNAs share many features with coding transcripts, it is not surprising that they are also subject to alternative splicing and degradation, and in some cases, can even undergo translation [32]. Similar to coding transcripts, such alternative splicing of lncRNAs could account for their cell type-specific expression as well as their subcellular localization, splice variants facilitating their interaction with a diverse array of RNA-binding proteins that are necessary for spatially and temporally resolved gene expression [33]. The initial definition of lncRNA included the inability to encode peptides. However, there is evidence that brain-specific lncRNAs such as BC1 or BC200 associate with ribosomes, suggesting the potential for translation [34]. Recent studies have shown that ribosomes bind to lncRNAs [35,36] and that lncRNAs often contain open reading frames (ORFs) [37,38]. Indeed, in addition to circular RNAs [39,40], several lncRNAs yield functional micropeptides [41–45]. Nevertheless, the occupancy of ribosomes on lncRNAs does not necessarily imply the capacity to encode functional peptides [46]. It is possible that these lncRNAs merely serve as guides or scaffolds for ribosomes [47,48]. Future research using more sensitive mass spectrometry analysis will help to delineate translating lncRNAs from the overall pool. Despite their coding potential, lncRNAs may still act independently of their peptide products via their conserved sequence modules [49,50].

LncRNAs show little sequence conservation

Although lncRNAs possess similar properties to coding RNAs, their sequence conservation and function remain the subject of intense debate. The ENCODE project showed that the vast majority of the transcripts arise from noncoding DNA regions with little sequence constrain, and concluded that lncRNAs are not under strong negative selection [51–53]. Indeed, lncRNAs display regions that are rapidly evolving [54]. Due to the lack of sequence constraint and imperfect fidelity of RNA polymerase II, proponents of the ‘transcriptional noise’ theory argue that transcripts derived from noncoding regions are merely by-products of transcriptional activity within a given locus [55]. The lack of primary sequence conservation and lower levels of expression seem to support this hypothesis. However, promoters, splice sites and exons of lncRNAs display functional conservation and are enriched in chromatin factors [12,29,56–59]. Moreover, lncRNAs are remarkably tissue-specific [60–65], and lncRNA variants and single nucleotide polymorphisms (SNPs) are also associated with disease phenotype [16,66–70]. For example, an SNP within the antisense lncRNA, DAOA-AS1, associates with bipolar disorder [71]. Therefore, a lack of primary sequence conservation does not necessarily indicate a lack of function [72,73].

LncRNAs are abundantly expressed in the brain

Half of all known lncRNAs are expressed in the brain [11,12], and some have been shown to be involved in neuronal development, neurogenesis and in the regulating cellular metabolism in the ageing brain [19,20,36,74–76]. LncRNAs exhibit cell-type and subregion-specific patterns of expression in the brain and are induced in an activity-dependent manner [11,13,18]. Moreover, lncRNAs are universally alternatively spliced, and with advances in sequencing technology, a myriad of novel isoforms, including those associated with neuropsychiatric disorders and ageing, are rapidly emerging [74,77,78]. Targeted enrichment and long-read sequencing have identified thousands of novel brain-specific lncRNA isoforms [77,78], including neuroligin-1 and regulating synaptic membrane exocytosis protein 2 (Rims2), both of which are directly involved in synaptic plasticity and memory [78–80]. Interestingly, several novel exons in the intron of Rims2 give rise to alternative transcripts with unknown function [78]. These studies illustrate the complexity of lncRNAome and highlight the need for targeted functional characterization of specific isoforms within different cellular contexts in both the healthy and diseased brain.

Brain-specific lncRNA activity may be critical for learning and memory

With respect to brain function, neuroscientists have historically focused on stimulus-induced nascent coding transcripts and protein synthesis [81,82], with the putative function of lncRNAs in cognition being completely overlooked. In 2015, the PsychENCODE project initiated a comprehensive study of lncRNAs in three neuropsychiatric diseases [83] and found the expression of some lncRNAs in different brain regions associated with synaptic transmission and memory [84–88]. More recently, these studies have been extended with the observation that LoNA-deficient mice display enhanced long-term potentiation (LTP) and long-term memory [84]. In addition, knockout of the dendritically localized lncRNA BC1 also impairs learning and memory [88]. Furthermore, the lncRNA Neat1 regulates gene expression by coordinating the deposition of repressive chromatin modifications, the dysregulation of which is associated with age-related memory impairment [89]. In addition, the dysregulation of a novel Gas5 isoform has been shown influence drug-associated learning and memory [90]. With respect to neurological and neurodegenerative disorders characterized by cognitive impairment, altered expression of lncRNAs has now been associated with schizophrenia [21,91–95], Autism [96,97], Parkinson’s disease (PD) [98,99], Alzheimer’s disease (AD) [100,101], Huntington’s disease (HD) [102], intellectual disability [103,104], neurodevelopmental disorders [105] and even personality disorder [106].

Taken together, these studies clearly implicate lncRNA activity in the regulation of memory processes in the healthy and diseased brain and, given the extraordinary breadth and complexity of lncRNA function, the challenge now is to elucidate the context- and cell-type-specific roles of individual lncRNA variants across many learning conditions.

Subcellular compartmentalization of lncRNAs

lncRNAs share many common RNA-binding proteins (RBPs) with other RNAs with interactions being highly dependent on their localization in space and time [107–109]. Indeed, lncRNAs can be trafficked to different subcellular locations within a cell [30] and the trafficking of lncRNA cargo to distal compartments from the nucleus may involve the kinesin motor and its regulating partners [110–112]. This capacity is dynamic and species-specific [113], with the dynamic localization of lncRNAs being structure- and state-dependent [114–118].

LncRNAs as guides, scaffolds and decoys in the nucleus

Nuclear lncRNAs can serve as guides to recruit transcription factors and epigenetic regulators for either initiation or inhibition of gene expression (Fig. 1A). Guide lncRNAs accumulate at the site of transcription, acting in cis (close to neighbouring genes), trans (distantly located genes), or in some cases, both. Cis-regulating natural antisense transcripts (NATs) regulate gene expression of their sense genes by recruiting factors that activate or repress transcription. For example, the antisense lncRNA Gdnfos regulates Gdnf isoform expression and is dysregulated in the temporal gyrus of AD and HD patients [119]. In AD, Bace1-AS lncRNA is upregulated, and its expression correlates with Bace1 gene expression and correlates with amyloid deposition [120]. An intronic cis lncRNA transcript, Nat51A, regulates alternative splicing, and dysregulation of this lncRNA greatly increases amyloid deposition in AD patients [22]. Conversely, the NAT Bdnf-AS represses brain-derived neurotrophic factor expression in the hippocampus by recruiting repressive chromatin modifier to the locus [121]. In addition to the brain, this mechanism also seems to be common in cancer [122,123]. Finally, in contrast to NATs, sense lncRNAs, such as utNgn1, can activate their proximal genes through a chromatin looping mechanism to regulate brain development [124].

Figure 1. lncRNA functions in the nucleus. In the nucleus, lncRNA interacts with mRNA and RNA-binding proteins (RBPs) to induce formation of membraneless condensates (top panel). lncRNA can act as guide to recruit DNA-binding proteins, such as transcription factors and chromatin modifiers, to the DNA to initiate transcription (A). lncRNA can also act as a scaffold or guide to bring two or more proteins (or DNA regions) into close proximity to maintain chromatin structure (B). lncRNA may also act as decoy to sequester DNA-binding proteins, such as RNA polymerases and chromatin modifiers, from interacting with DNA, and hence, leading to transcription inhibition (C). During active transcription, lncRNA can interact with RNA polymerases and chromatin modifiers at sites of transcription to induce formation of chromatin condensates (D)

The trans mechanism of guide lncRNAs is equally important for the simultaneous activation of a plethora of genes. During neurogenesis, the lncRNAs Tuna, Paupar and Rmst can modulate the expression of distal genes that regulate pluripotency. These lncRNAs recruit repressors to the promoter of pluripotent genes to inhibit transcription, and as a result, neuronal differentiation ensues [125–127]. Furthermore, the lncRNA Malat1 regulates synaptic density by recruiting splicing factors in trans to facilitate gene expression of its target transcripts [128], and Neat1 can repress global gene expression via recruitment of chromatin factors to c-Fos promoter during memory formation [89]. Guide lncRNAs can also regulate gene expression through both a cis and trans mechanism [129]. For example, the lncRNA Evf2 can recruit transcription factors to regulate Dlx5/6 and Gad1 gene expression in the GABAergic circuitry in cis and trans, respectively [130,131]. In dopaminergic neurons, antisense Uchl1 lncRNA plays a dual role: increased Uchl1 expression in the nucleus in cis, and shuttle to the cytoplasm to regulate Uchl1 translation in trans [116].

Scaffold lncRNAs serve as adaptors for forming discrete protein complexes and may or may not involve chromatin (Fig. 1B), and many lncRNAs carry out this important functional role [129,132–135]. A classic example is the X-chromosome interacting lncRNA, Firre. It acts as a scaffold for nuclear-matrix factor and chromosomes in neural-crest cells in the developing brain [136]. A recent study has also identified neuroLNC, a conserved neuron-specific nuclear lncRNA that regulates synaptic vesicle release, which serves as a scaffold for the TAR DNA-binding protein – 43 (TDP-43), where it stabilizes transcripts that are involved in synaptic transmission [137].

Decoy lncRNAs prevent proteins from accessing the chromatin or DNA (Fig. 1C). An excellent example of a decoy function is the Gas5 lncRNA [86]. Using its conserved glucocorticoid receptor (GR) motif, Gas5 binds to and prevents the GR from interacting with DNA, thereby repressing transcription of GR-targeted genes [138]. In addition, the Pnky lncRNA regulates neurogenesis in the postnatal brain by interacting with splicing factors [139]. NATs can also act as decoys to repress the transcription of their proximal sense genes. For example, the lncRNA Lrp1-AS sequesters Hmgb2 proteins from activation Lrp1 transcription, and Lrp1-As expression is elevated in AD patients [140]. The lncRNA gomafu is associated with distinct nuclear bodies [141], and one of the primary functions of this lncRNA appears to be the sequestering of splicing factors to regulate transcription and splicing in neuronal cells in a context-dependent manner [142].

It is also common for small RNA genes to be encoded within lncRNA loci. For example, Gas5 can host several small nucleolar RNAs (snoRNAs) within its introns [143]. The presence of these snoRNAs can then regulate the maturation of the Gas5 host lncRNA [144]. Because snoRNAs reside in lncRNAs, they are often categorized as sno-lncRNAs. In some cases, sno-lncRNAs regulate alternative splicing in the nucleus [27]. In contrast, host lncRNAs can function independent of their resident. Rmst, the host of miR-1251 and miR-135-a2, provides one example [126]. Knockdown experiments have shown that depleting the Rmst host gene has no effect on the expression of miR-1251 and miR-135-a2 during neurogenesis [126]. Future studies investigating alternative splicing of lncRNAs will no doubt reveal the existence of complex repertoire of isoforms with discrete functions in the brain.

LncRNAs as critical components of biomolecular condensates in the nucleus

Once lncRNAs come in close proximity with their lncRNA binding proteins (LBPs) they can oligomerise and recruit additional proteins. Using their low complexity prion-like domain, RNA-protein complexes form membraneless liquid-liquid phase-separated (LLPS) biomolecular condensates (Fig. 1). In the nucleus, these entities include paraspeckles, nuclear speckles, Cajal bodies, promyelocytic leukaemia (PML) bodies or simply nuclear bodies, which are characterized by the type of RNA present within them. The first lncRNA observed in the phase-separated compartment was Xist, which coats the inactivated X chromosome to form distinct architecture in the nucleus [145]. The lncRNA Neat1 is an essential component of paraspeckles and has a tendency to relocate between different compartments in response to neuronal stimuli [117,146]. In the motor neurons of amyotrophic lateral sclerosis (ALS) patients, the Neat1 isoform, Neat1_2, binds to the RNA-binding proteins Fused in Sarcoma (FUS) and TDP-43 and assembles them into paraspeckles [114,115]. Additional components of the paraspeckle have also started to emerge [147–150]. Malat1, for example, is a lncRNA that has been implicated in brain function and can also form a distinct nuclear speckle. The formation of this condensate correlates with the recruitment of factors during active transcription and the splicing of genes involves in synaptic formation and maintenance [128,151]. Taken together, these observations clearly suggest that LLPS involves the activity of lncRNAs, which lends to their complex, context-specific functions, and may therefore play an important role in normal cellular processes and in disease pathogenesis.

LncRNAs as guides, scaffolds and decoys along the axon and at the synapse

The transport of RNA along the axon or dendrites to the synapse serves to localize critically important RNAs to the synapse for local translation that is necessary for plasticity [110,152]. Although direct evidence of the involvement of lncRNAs in axonal cargo transport has yet to be demonstrated, it is possible that some localized granules can host lncRNAs. The axonally localized fragile X mental retardation protein (FMRP) binds and indirectly regulates lncRNA Tug1 expression in the regulation of axonal development [153]. In addition, a subset of NATs associate with kinesin motors in response to stimuli [112] and these kinesin motors are capable of transporting RNA granules and splicing ribonucleoprotein complexes (RNPs) along the axon [111,112,154]. The association of NATs with kinesin motors therefore further indicates that they are actively transported to the synapse.

Recent studies have also identified a population of lncRNAs that accumulate in the synaptic compartment during ageing[20]. BC1 was the first guide lncRNA discovered to modulate the translation of its target transcript at the synapse [34,155–158] (Fig. 2A). BC1 mediates translation repression by bridging the repressor FMRP and its target mRNAs [155–158]. In addition, NATs Bace1-AS and Uchl1-AS guide their sense transcripts to the synapse for protein synthesis [116,120,159,160] and guide lncRNAs in the cytoplasm have been shown to employ a similar mechanism [161,162]. This form of tuning and regulation by synaptic guide lncRNAs may be crucial for synaptic activity. For example, BC1 can bind to FMRP as a guide; it can also act as a bridge for FMRP and its target molecules [161]. In the synapse, BC1 can also bind to translation initiation factor (eIF4A) and polyA binding protein (PABP) and precludes them from interacting with target mRNAs to initiate translation (Fig. 2C) [163,164]. It has also been proposed that Meg3 functions as a miRNA decoy in the synaptic compartment to mediate AMPA receptor loading at the synapse [165]. Acting along similar lines, LncND reversibly sequesters miR-143-3p to regulate Notch signalling. The maintenance of the neuronal progenitor pool by this lncRNA–miRNA interaction is crucial for the expansion of the cerebral cortex in humans [166]. In addition, the lncRNA NORAD acts as decoy for dendritic-localized PUMILIO to prevent them from repressing translation [167,168]. Despite these interesting threads of evidence; however, the functional relevance of lncRNAs in axons and at the synapse is not completely understood, and future studies will likely yield new insight into the role of localized lncRNA function in the brain.

Figure 2. lncRNA mechanisms of action at the synapse. In the pre and post-synaptic compartments, lncRNAs may localize to membraneless condensates, such as stress granules, processing bodies (P-bodies) and silencing foci (S-foci) (top panel). These biomolecular structures may serve as reservoirs for RNAs and proteins. RNAs and proteins can travel in and out, or between these storage sites in response to learning-induced stimuli or stress. lncRNA may serve as guide to direct proteins, such as translation factors and helicases, to the mRNA for local translation (A). lncRNA may act as scaffold to bring 2 or more proteins (or RNAs) into close proximity to induce formation of condensates at sites of active translation (B). LncRNA-associated condensates may act as decoy to titrate away RBPs, such as translation initiation factors, from interacting with RNAs, and therefore, inhibits local translation at the synapse (C)

LncRNAs may be key functional elements associated with localized RNA granules

Sites of RNA storage within the synapse include stress granules (SG), processing bodies (P-bodies), and neuron-specific granules called silencing foci (S-foci) (Fig. 2). SG and P-bodies contain discrete RNPs [169] whereas P-bodies consist of mostly deadenylated transcripts, RNA decapping and RNA decay factors [170]. On the other hand, dendritic stress granules house distinct RNPs, including polyadenylated transcripts, translation initiation factors and small ribosomal subunits [171–174]. Neuron-specific S-foci are primarily comprised of deadenylases [175] and these membraneless subcellular compartments can move RNAs and proteins freely between cytosol and other granules in response to stress or stimuli [176–178]. The biogenesis and movement of these condensates is dynamic [179–183]. For example, P-bodies can travel along dendrites in response to stimuli [184]. Dendritic stress granules with distinct ribonucleoparticles (RNPs) can contribute to translational repression, adding another layer to posttranscriptional regulation [185]. lncRNAs may associate with RNPs to form granules [186], as BC1 associates with PABP, translation initiation factors and other components of the ribosome at the synapse [34,163,164] (Fig. 2B). Indeed, these granules have been shown to play a role in learning and long-term memory [187,188] and, quite extraordinarily, the formation of these membraneless condensates renders RNA in a ‘quiescent’ state where, upon stimulation, they serve to silence translation and promote RNA stability [186,189,190]. On the other hand, disassembly of SG can rapidly recover protein synthesis [175,191], which suggests that some of these biomolecular condensates may primarily function as centres for sorting or aggregation that promote RNA metabolism when translation is in high demand. This ‘reservoir’ hypothesis is supported by the fact that SG also stores small ribosomal subunits, N6-metheyladenosine (m6A) modified RNAs and translational initiation components [118,172,174]. Similar to transcription-dependent condensates in the nucleus [192] (Fig. 1D), dendritic condensates may perhaps coalesce to form translation-dependent condensates at sites of translation during protein synthesis (Fig. 2B). This alternative hypothesis is supported by the observation that transcripts containing internal ribosome entry sites (IRES) can recruit factors to promote translation [193,194]. Despite the existence of RNA reservoirs and translation-dependent condensates in neurons, the functional exchange of RNA between different membraneless compartments remains to be demonstrated in the brain, and the role of lncRNAs governing this process in the context of experience-induced plasticity warrants further investigation.

Current technologies

Short read sequencing of barcoded-cDNA fragments enables the detection of RNA between 45 and 400 nucleotides [195]. This technology has been incredibly been useful for generating reliable assemblies of protein-coding transcriptomes. However, short read sequencing cannot accurately determine the full repertoire of transcript isoforms, which is an important issue given that lncRNAs are universally alternatively spliced. Furthermore, because lncRNAs are often expressed at very low levels, detecting them remains a significant challenge. Single-molecule, real-time (SMRT) sequencing has the ability to yield a more complete picture of lncRNA expression patterns [77]. The advantage of this technology is that RNA or cDNA is sequenced in full without the need to fragment, and information about the 5ʹ and 3ʹ end of the transcript is retained. Critical features of RNA are now accessible, including the detection of their structure state and alternatively spliced isoforms. Recently, Oxford Nanopore has evolved this approach with ultra-long read sequencing, and this technology has been adapted to sequence lncRNAs [78]. Using lncRNA targeted biotin-conjugated oligonucleotides, a collection of dysregulated lncRNAs in the brain of neuropsychiatric patients has been captured and sequenced [78]. In addition, RNA modifications can also be accurately determined with specific base-calling algorithms [196]. The downside of Nanopore technology is that it is currently prone to a relatively high error-rate [197]. Nevertheless, nanopore technology is an important advance in the ability to detect alternative lncRNA sequence variants in their native state and in multiple biological systems, including neurological diseases.

The ability to manipulate lncRNA expression with spatiotemporal resolution in vivo remains challenging due to their dynamic nature. Traditionally, lncRNA knockdown has been achieved through the use of antisense oligonucleotide (ASO) [198]. This approach is quite efficient in silencing gene expression in-vivo. However, it is not possible to use ASO to pinpoint specific subcellular compartments for knockdown, which is a critical issue given that lncRNAs are functionally active depending on their location in the cell. An alternative approach is the recently discovered CRISPR-cas9 knockout technology [199]. CRISPR-cas9 generates mutations or deletions using guide RNAs (gRNAs) against the genomic DNA of interest. However, the knockout effect of cas9 is permanent and therefore not ideal for achieving temporal control in the brain. To circumvent this issue, a temporally controlled CRISPR-mediated gene interference system (CRISPRi) for gene inactivation in the brain has been developed [200]. The CRISPRi system utilizes a fusion peptide consisting of a catalytically inactive cas9 and a repressor protein. With gRNAs, the inactive cas9-repressor complex into close proximity to the gene hence repressing its expression; however, this method is also prone to off-target effects.

More recently, an optically controlled CRISPR-cas9 system has emerged [201]. This system uses a photoconvertible Dronpa domain with the Cas9 protein, which provides both spatial and temporal control. An RNA-targeted CRISPR-cas13 has also emerged as a tool specifically designed for RNA knockdown [202]. This is particularly useful as introduction of the guide in a specific time window can now be achieved. The CRISPR-cas13 system has also been adapted to RNA modifications to modulate gene expression [203]. Inspired by CRISPR-cas13 system, Bryan Dickinson’s group at the University of Chicago have developed a CRISPR-cas-inspired RNA targeting system (CIRTS) [204]. CIRTS is an engineered synthetic protein created by fusing different effector, RNA binding, and RNA guide recognition module into one assembly. It has the advantage of being small, which dramatically improves virus packaging and high titre production. Adapting the CRISPR-cas13, the CIRTS or the photocaged-system to the brain will enable the functional characterization of these lncRNAs, in particularly in the context of experience-dependent plasticity, learning and memory.

Overexpression of a transgene in the brain often employs a classic overexpression construct with a brain-specific promoter upstream of the transgene [205]. This approach requires careful consideration in choosing the gene isoform and promoter. Constitutively activating transgenes are not ideal in most cases, and an activity-dependent promoter might not be suitable in some learning contexts. Alternatively, overexpression of a gene of interest is attainable with the CRISPR-cas9 activation system (CRISPRa) [206]. Using the same concept as CRISPRi, CRISPRa uses a catalytically inactive cas9 and fuses it with a transcriptional activator. With promoter-targeted gRNAs, CRISPRa can activate genes in a targeted manner; however, the CRISPRa system is also not suitable for isoform-specific activation, and off-target gene activation remains a problem.

As indicated throughout this review, to better understand the role of lncRNAs in the brain, it is important to identify their pattern of subcellular localization. The simplest way to directly visualize lncRNA is using fluorescent in-situ hybridization (FISH) [117,207]. Fluorophores conjugated to probes target different regions of lncRNA without the need for a secondary substrate. However, this imaging technique might not be amenable in cases where target lncRNA expression is low. Using Z-probe technology, RNAscope has been applied to the brain for visualizing lncRNA [166]. Unfortunately, FISH and RNAscope are not yet amenable for live cell imaging. Molecular beacons are used for the imaging of RNA in live cells [208,209]; however, similar to the FISH approach, it may not be ideal for highly structured, lowly expressed lncRNAs. For live cell imaging, different fluorescent aptamers have been developed [210–215]. These aptamers emit light when bound to fluorogen, thus enabling the tagging of a specific RNA of interest. If fluorescent aptamers can be a target to lncRNAs, they will represent a powerful new tool for the tracking of lncRNA dynamics in real-time.

Finally, to understand the underlying mechanism of action of lncRNAs, identifying their functional interacting partners will be necessary. A straightforward way of purifying LBPs is by tagging the lncRNA of interest with an MS2 hairpin, immunoprecipitate the RNA-LBP complex, and followed by mass spectrometry analysis [216,217]. Although this approach is rapid, a caveat is that optimal binding cannot be reached due to competition with endogenous lncRNAs. An alternative is to purify the RNA-LBP complex with biotin tagged DNA oligonucleotides. With this approach, the interaction between lncRNA-LBP and chromatin can also be resolved [218–220]. A limitation of this approach is the amount of starting materials required for accurate detection. Future methods that consider the cell-type specific and temporal aspects of lncRNA-LBP in the brain will need to be required. Nonetheless, the current approaches are greatly improving our understanding of lncRNA interaction networks, particularly in the brain.

Final outlook

Neural plasticity requires not only precise spatiotemporal control over gene expression, but also a dynamic interaction between RNA and various proteins. Our understanding of how lncRNAs coordinate these processes is rapidly increasing. Elucidating their functional and mechanistic role in subcellular compartments will provide a better picture of how lncRNAs mediate behavioural adaptation. Central to this challenge will be a deeper understanding of how lncRNA can adopt various structure states and how this affects their ability to interact with specific protein complexes. Last but not least, more sensitive cell-type and compartment-specific proteomic approaches will enable a deeper understanding of the functional consequences of lncRNA activity in the brain. Undoubtedly, lncRNAs will soon be shown to play an important role in the transcriptional and local translational regulation underlying learning and memory.

Acknowledgments

The authors acknowledge grant support from the Brain and Behavioural Research Foundation National (NARSAD Independent Investigator Grant-TWB), the National Institutes of Health (NIH) (R21MH103812-TWB) and the National Health and Medical Research Council (GNT1145172-TWB). The authors thank Ms Rowan Tweedale for editing of the manuscript and Dr Nick Valmas for figure graphics.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the National Alliance for Research on Schizophrenia and Depression; National Health and Medical Research Council [GNT1145172]; National Institute of Mental Health [R21MH103812].