Fundamental Neurochemistry Review: At the intersection between the brain and the immune system: Non‐coding RNAs spanning learning, memory and adaptive immunity

Non‐coding RNAs (ncRNAs) are highly plastic RNA molecules that can sequester cellular proteins and other RNAs, serve as transporters of cellular cargo and provide spatiotemporal feedback to the genome. Mounting evidence indicates that ncRNAs are central to biology, and are critical for neuronal development, metabolism and intra‐ and intercellular communication in the brain. Their plasticity arises from state‐dependent dynamic structure states that can be influenced by cell type and subcellular environment, which can subsequently enable the same ncRNA with discrete functions in different contexts. Here, we highlight different classes of brain‐enriched ncRNAs, including microRNA, long non‐coding RNA and other enigmatic ncRNAs, that are functionally important for both learning and memory and adaptive immunity, and describe how they may promote cross‐talk between these two evolutionarily ancient biological systems.

Among the non-coding RNAs that have been identified, those in the brain exhibit the most diversity with respect to their cell type and spatiotemporal patterns of expression, which suggests an important role in the regulation of neurological processes.Indeed, non-coding RNAs have been implicated in critical aspects of neuronal development and metabolism and, in the brain, control subcellular activity and are involved in communication within and between neurons, and between neurons and other cell types.
Moreover, recent studies have identified a role for ncRNAs in adaptive immunity, particularly with respect to T-cell regulation in brain inflammation and psychiatric disorders (Dantzer, 2018;Marshall & Bredy, 2019).
Given the similarities and cross-talk between the regulatory processes associated with neural plasticity, learning and memory, and those related to adaptive immunity, it would come as no surprise that the same mechanisms that drive the immune response have been co-opted by the brain to promote adaptation and the formation of long-term memories.Furthermore, accumulating evidence suggests that behavioural memory and adaptive immunity may be regulated by common ncRNAs.This is caused by the similar operational paradigms between the two systems, and because many ncRNAs interact with pathways integral to both.
In this review, we discuss the evidence that several classes of non-coding RNA, namely long non-coding RNA (lncRNA) and several small ncRNA (sncRNA) families, including microRNA (miRNA), Y-RNA and vault RNA (vRNA), are involved in the regulation of both learning and memory and the adaptive immune response.
We describe how ncRNA could regulate each system, particularly with respect to their structure and subcellular localisation, and consider how these features underlie their functional influence in different cells of the central nervous system, particularly those associated with adaptive immunity which, itself, represents an ancient form of memory.

| THE ADAP TIVE IMMUNE SYS TEM
Interactions between the central nervous system (CNS) and the immune system (IS) are required for brain homeostasis.However, they are tightly regulated and thought to be partitioned in order to limit the extent of damage to post-mitotic neurons by immunerelated activity (Buckley & McGavern, 2022).Here, we focus on the cells and molecules that are involved in the adaptive immune system (AIS), as mechanisms of 'adaptive immunity' that can function similarly in learning and memory.In the healthy brain, a variety of activated T-cells-typically 'effector-memory'-circulate between the gut and parenchyma (Prinz & Priller, 2017).These memory T-cells are activated by antigen exposure and persist in the body until antigen re-exposure occurs, at which point they mount a rapid immune response, making them the backbone of the AIS (Prinz & Priller, 2017).
T-cells are also implicated in the progression of neurodegenerative disorders such as Parkinson's disease and neuropsychiatric disorders, including those related to autism (Filiano et al., 2015).
Both neurological disorders have a high pro-inflammatory bias, with an abundance of activated microglia and pro-inflammatory cytokines causing an increased vulnerability to brain damage (Filiano et al., 2015).Conversely, memory T-cell enrichment reduces proinflammatory responses in immunocompromised mice, leading to increased neurogenesis and a reversal of the cognitive and spatial memory deficits associated with brain inflammation (Filiano et al., 2015).
The primary interface between the AIS and the brain appears to be microglia, which are usually considered 'innate' immune cells (Pape et al., 2019).However, this may be an oversimplification as they participate as part of an axis with astrocytes, expressing toll-like receptors and cell surface receptors for ATP and glutamate (Garland et al., 2022;Pape et al., 2019).This, combined with their branching processes allows microglia to sense neurons in their immediate environment and perform processes such as synaptic pruning when synapses are tagged with complement factors (Pape et al., 2019).
Unlike microglia, astrocytes express a very select range of tolllike receptors, which suggests that they cannot respond to many forms of immune challenge.However, they do respond to chemokines released by microglia and can uptake ncRNA from microglia extracellular vesicles and release their own (Garland et al., 2022).
The AIS therefore interacts with the brain via memory T-cells that release signalling molecules upon detecting a previously encountered antigen, which then activate microglia.Microglia respond by transitioning to a pro-inflammatory state and migrate towards the source of immune challenge, where they release chemokines and extracellular vesicles that can be either neuroprotective or neurotoxic (Pape et al., 2019;Prinz & Priller, 2017).Astrocytes responding to these signals increase their proliferation, alter their morphology and function and appear to wrap around the damaged neurites of neurons under immune attack (Garland et al., 2022).
Although we still do not fully understand how the AIS influences the brain, it is evident that the two systems frequently interact, and that the AIS is involved in determining the severity of neuropsychiatric and neuroinflammatory disorders (Fernandez-Egea et al., 2016).Indeed, severe combined immunodeficient (SCID) mice exhibit profound deficits in spatial learning and memory, which is caused by broad-scale activated memory T-cell depletion (Kipnis et al., 2004).Studies in wild-type animals support this observation, with behavioural training increasing the presence of activated T-cells in the CNS.Both depleting and inhibiting CNS penetration of T-cells impairs spatial learning and memory, and this can be rescued by the injection of activated T-cells (Derecki et al., 2010).Furthermore, ln-cRNA NEAT1 and microRNA miR-32 are implicated in T-cell plasticity and signalling (Shui et al., 2019;Soreq et al., 2014).This is particularly evident during neuroinflammation and in neurodegenerative diseases such as Parkinson's disease (Ren et al., 2020;Sethi et al., 2013;Soreq et al., 2014).These findings strongly support the idea that the learning systems of the brain and AIS are functionally connected and may therefore share common signalling molecules and mechanisms of action.

| Chemical and electrical neurotransmission
The brain's immune system can influence synaptic plasticity, and hence learning, by way of cross-talk among neurons, microglia and astrocytes.This integration occurs at the neuronal pre-andpost-synapses, forming tripartite synapses with nearby glia, which increase intercellular communication within the synaptic microenvironment (Bondy, 2020;Kastellakis et al., 2015;Morimoto & Nakajima, 2019).The most common and well-characterised form of intercellular communication is chemical signalling, but electrical signalling, tunnelling nanotubes and secreted proteins and vesicles are also involved.
Both chemical and electrical signalling occur in synapse-, cell type-and brain region-specific patterns.Chemical signalling operates through vesicular bodies that are locally synthesised and, upon neuronal stimulation, fuse with the plasma membrane and release their contents into the synaptic cleft.This includes comparatively rare growth factors, cytokines and major neurotransmitters such as glutamate and gamma-aminobutyric acid.ncRNA are also stored in the same vesicles, and therefore likely released into the synaptic cleft to act as neuromodulators (Gumurdu et al., 2017;Li et al., 2015).In the cholinergic system, miRNA and lncRNA perform a variety of functions, including miRNA-mediated transcriptional repression of key factors like cholinesterase, and lncRNA-mediated sponging of miRNA (Madrer & Soreq, 2020).Chemical signalling typically defines synapse function, as the major type of chemical ligand released from the pre-synapse, and major receptor at the post-synapse determines whether the information received is excitatory or inhibitory (Durkee & Araque, 2019;Nusbaum et al., 2001).Little is known about how chemical signalling influences the glial contribution to the tripartite synapse.
Electrical signalling uses connexin polymerisation, forming hexameric channels to either allow passive diffusion from a neuron's cytoplasm to the extracellular medium (i.e.hemichannels), or to a neighbouring neuron or astrocyte (i.e.gap junctions).Hemichannels and gap junctions exist as both connexin hetero-and homodimers, with 11 CNS-specific isoforms that have developmental state-, brain region-and cell type-specific expression.Post-translational modifications can also occur inside the channel to alter its permeability (Lapato & Tiwari-Woodruff, 2018;White & Bruzzone, 1996).
Functionally, gap junctions propagate calcium transients across the tripartite synapse, which can be altered by long-term potentiation or depression to both control connexin opening and release neurotransmitters.In this manner, electrical signalling regulates synaptic transmission and plasticity (Lapato et al., 2017;Takeuchi & Suzumura, 2014).
Signalling between neurons and astrocytes is not as well understood as that between neurons.Astrocytes exist in mutually exclusive 'spongiform domains' that vary in size and shape across brain region, with each astrocyte forming peri-synaptic 'nodes' as part of tripartite synapses (Arizono et al., 2020;Durkee & Araque, 2019).

| Extra-synaptic intercellular communication
Extra-synaptic intercellular communication, performed by secreted extracellular vesicles (EVs) and tunnelling nanotubes (TNTs), also contributes to communication between the AIS and the CNS.Both EV secretion and extracellular targeting are multifaceted, involving both lipid-protein and protein-protein interactions that differ between EV classes (Losurdo & Grilli, 2020;Lukiw & Pogue, 2020).
Most neural EV research to date has focused on small EVs (100-1000 nm), their miRNA cargo and the glia that secrete them, as miR-NAs are key to many processes involved in adult neurogenesis and synaptic plasticity.The cargo of each is distinct from the parent cell, contains many ncRNA families and is modified according to the surrounding environment (Antoniou et al., 2023;Jeppesen et al., 2019).
One example is the BDNF-stimulated synaptic concentration and release of miRNA-containing EVs, which increases excitatory synapse density and burst-firing activity in the hippocampus (Antoniou et al., 2023).Furthermore, in multiple sclerosis, chronic activation of microglia leads to the secretion of EVs containing miR-146a-5p, which is transferred to nearby neurons.This results in the inhibition of several genes involved in the maintenance of synaptic integrity (Prada et al., 2018).Aberrant miRNA function is common to most neuropathologies; however, how EV-derived miRNA and glia, in general, interact during learning remains relatively uncharacterised.Similarly, the function of other EV ncRNAs and how their presence is modified according to the environment and differs on a ncRNA speciesspecific basis are unknown.
Studies of TNTs have indicated that they begin as groups of transient, actin-based cytoplasmic tubes that connect two cells, possibly based on receptor-ligand interactions (Henderson et al., 2022;Sartori-Rupp et al., 2019).As a result of the static position of mature neurons, TNT formation would also require a density gradient in the extracellular milieu, although which intercellular messengers mediate this is unknown.Neuronal activation or cellular attack can stabilise the TNT cytoskeleton by suffusing it with microtubules, which appear to be critical for vesicular trafficking (Saenz-de-Santa-Maria et al., 2017;Wang & Gerdes, 2015).Cargo passes through TNTs bidirectionally either in a native state, as in the case of organelles and large molecules, or encapsulated in vesicles for nucleic acids.Unlike gap junctions, TNTs do not allow passive diffusion so calcium transients do not propagate, and membrane-embedded proteins cannot diffuse laterally (Yamashita et al., 2018).TNTs might therefore serve as a faster, more direct method of intercellular communication than exo-or endocytosis, although it is unknown whether EVs released into the extracellular milieu have different cargo compared to TNT EVs (Driscoll et al., 2022).Supporting this idea, TNTs have been shown to establish connections between immune cells during both cancer and inflammation, and are even able to transmit pathological tau aggregates between afflicted neurons (Driscoll et al., 2022;Tardivel et al., 2016).

| ncRNA structure confers function
The majority of the mammalian genome is devoted to producing non-coding RNA (ncRNA), which correlates with organism complexity (Abugessaisa et al., 2021;The ENCODE Project Consortium et al., 2020).NcRNAs possess a high degree of secondary and tertiary structure, and accordingly, their interactions with binding partners and resulting functions are highly complex.Secondary structures are created by Watson-Crick-Franklin (WCF) or 'canonical' base-pairing: A-U or C-G hydrogen bonds form between sequential nucleotides, creating a variety of structures (Rauch & Dickinson, 2018).These structures change in response to environmental stimuli and the presence of binding partners.Studies have also indicated that ncRNA splicing is modified by disease state and during learning (Meydan, Bekenstein, & Soreq, 2018;Soreq et al., 2014).Although divided into many families based on their interacting partners and structure, all ncRNAs have several characteristics in common.First, ncRNAs are transcribed using all three RNA polymerases, although most ncRNAs also have a 5′ 7-methyl-guanosine cap and 3′ poly-A tail, and contain 5′ and 3′ untranslated regions (Ransohoff et al., 2018).Second, ncRNAs have multiple protein or RNA-/DNA-binding sites and can therefore function in subcellular localisation, intercellular communication, as molecular sponges and scaffolds, or in transcriptional and epigenetic regulation.Third, many ncRNAs, especially long ncRNA, tend to have low-sequence conservation compared to the common microRNA (miRNA) family, but high-structural conservation.These characteristics are thought to be as a result of more evolutionary constraints on larger ncRNA (Pang et al., 2006;Qu & Adelson, 2012).
Consequentially, even though the sequence can change between species, the structure and charge profile of ncRNAs, and therefore their interactome, are mostly consistent.In addition, as a result of their highly plastic structure, ncRNAs can be said to have individual 'fold hyperspaces' in which they behave predictably between species.Simply put, ncRNA structure equals function (Mattick, 2018;Pang et al., 2006).
The high degree of functional plasticity found in ncRNA makes brain-specific ncRNAs attractive targets for studies aimed at understanding learning and memory, and how they intersect with the brain's immune system (Irwin et al., 2021;Karapetyan et al., 2013;Samaddar & Banerjee, 2021).Below we describe some examples of ncRNA that are known to functionally intersect learning and the AIS, including Gas5, Gomafu, Malat1 and Meg3.

| Gas5
The lncRNA Gas5 is down-regulated in systemic lupus erythematosus (lupus) patients, specifically in plasma and activated memory T-helper cells (Table 1).Conversely, over-expressing Gas5 in these T-cells inhibits their activation via the E4BP4/CD40L pathway.Gas5 sponges miR-92, which leaves E4BP4 active to inhibit the cells (Zhu et al., 2020).A similar association has been discovered in allergic rhinitis (AR), with Gas5 being up-regulated in AR exosomes.These exosomes decrease interferon gamma, a key signalling molecule for activating T-helper cells, biasing their differentiation towards humoral (rather than cellular) immune responses.Gas5 application reverses this by translationally and transcriptionally inhibiting EZH2, leading to increased T-bet, and therefore activated T-helper cells and interferon gamma (Zhu et al., 2020).
Gas5 is also involved in a variety of learning-related behaviours, particularly those involving the hippocampus and the prefrontal cortex, and is up-regulated in neural stem cells (NSCs) where it promotes neuronal differentiation (Table 1; Zhao et al., 2020).In a hippocampal model of cholinergic injury, Gas5 over-expression in NSCs rescues the spatial learning and memory deficits caused by partial restoration of the cholinergic system (Zhao et al., 2020).This is also seen in aged mouse studies, with Gas5 expression positively correlating with both age and stress levels, and negatively correlating with spatial learning (Meier et al., 2010).Furthermore, synapse-enriched Gas5 interacts with large, distinct classes of proteins before and after fear extinction learning-particularly those involved in RNA metabolism and RNA granules.Knocking down synaptic Gas5 impairs the trafficking and clustering of RNA granules, suggesting that this lncRNA may be critical for processing the synaptic RNA pool associated with memory formation (Liau et al., 2023).

| Gomafu
The lncRNA Gomafu has been extensively studied in the context of innate immunity and in the inflammatory response (Table 1).
Using a mouse model of obesity, it has been demonstrated that treatment with berberine (BBR) is broadly anti-inflammatory and leads to decreased Gomafu expression (Han et al., 2020).Gomafu knockdown produces similar effects to BBR treatment, whereas its over-expression partially reverses this effect, and leads to a reduction in key anti-inflammatory factors interleukin-4 (IL-4) and − 10 (IL-10; Han et al., 2020).These factors are required for adaptive immunity, with IL-4 inducing T-helper cell proliferation and IL-10 activating T-regulatory cells (Coyle et al., 1995;Yoshimoto et al., 2009).As such, Gomafu may partially repress adaptive immunity.Gomafu has also been shown to be dysregulated in a brain region-specific manner in schizophrenia.During development, it is up-regulated in neural progenitor cells (NPCs) in the subventricular zone, which suggests that Gomafu is also linked to adult neurogenesis (Teng et al., 2023).Supporting this idea, inhibiting Gomafu in neuron-like and NPC-derived neurons produces changes in alternative splicing of genes related to neurogenesis and schizophrenia, including QKI, DISC1 and ERBB4 (Barry et al., 2014;Teng et al., 2023).Gomafu knockdown also leads to an increase in the activity of several immune pathways implicated in schizophrenia and autism spectrum disorders, particularly interferon-gamma signalling, and the adaptive response to viral infection (Teng et al., 2023).
In addition, several studies have linked Gomafu to behavioural hyperactivity, a key symptom of schizophrenia and similar behaviours related to anxiety disorder (Table 1).Although Gomafu knockout mice display no obvious morphological alterations in the hippocampal CA1 region, they do exhibit increased locomotion and rearing (exploratory) behaviour (Ip et al., 2016).Moreover, these mice do not exhibit depressive or social anxiety-like behaviours, suggesting that the phenotype may be caused by hyperactivity.Supporting this idea is the fact that Gomafu knockouts exhibit increased sensitivity to methamphetamines, manifesting as increased locomotion and dopamine synthesis (Ip et al., 2016).Gomafu knockout also alters the expression of a select number of genes, including Xlr3b and Cebpb, which are involved in cognition and hippocampal neurogenesis (Ip et al., 2016).Interestingly, endogenous Gomafu is down-regulated after fear learning, which increases freezing behaviour, the nominal opposite of hyperactivity.Knocking down Gomafu exacerbates this, increasing distance travelled in an open field and decreases the time spent in the field's centre (Spadaro et al., 2015).Collectively, these data suggest that Gomafu negatively regulates anxiety-related behaviour, rather than hyperactivity.

| Malat1
The lncRNA Malat1 promotes autophagy in macrophages by serving as a decoy for miR-23, which belongs to the phosphatidylinositol 3-kinase/Akt pathway, and is a negative regulator of autophagy by preventing autophagosome/lysosome fusion (Table 1; Ma et al., 2019).Malat1 itself has also been shown to up-regulate fusion (Ma et al., 2019).With respect to adaptive immunity, Malat1 is up-regulated in breast cancer.In these tissues, Malat1 inhibits both natural killer cells and CD8+ (cytotoxic) T-cell activation by inhibiting miR-34 and -17, respectively, thereby impairing the AIS (Mekky et al., 2023).In keeping with Malat1 acting through regulation of miRNA activity, it also inhibits miR-9 in hippocampal neurons in a model of a vascular dementia (Wang et al., 2023).In this case, Malat1 appears neuroprotective as its over-expression aids recovery from ischaemic reperfusion (Wang et al., 2023).
A neuroprotective phenotype has also been observed in other models of ischaemia, with Malat1 serving as a sponge for miR-142 to inhibit the activity of SIRT1 deacetylase.In this manner, Malat1 over-expression reduces the pro-inflammatory response, leading to reduced infarct volume and rescues spatial learning and memory (Meng et al., 2023).Moreover, exercise further reduces apoptosis TA B L E 1 ncRNAs involved in the adaptive immune system, and learning and memory.after reperfusion, an effect that is strengthened by Malat1 overexpression, which improves both spatial learning and memory (Shang et al., 2018).
Finally, Malat1 has been implicated in the formation of fear extinction memory (Table 1).While this lncRNA was thought to be primarily expressed in the nucleus, a methylated version has been detected at the synapse after behavioural training (Madugalle et al., 2022).Methylated Malat1 interacts with CYFIP2 and DPYSL2, which belong to the WAVE complex and are key to actin dynamics and dendritic spine formation.DPYSL2 requires m6A to bind Malat1, which is necessary for fear extinction memory (Madugalle et al., 2022).

| Meg3
The lncRNA Meg3 is involved in multiple aspects of the miRNA/ immune response axis (Table 1).In a mouse model of arthritis, the Meg3 promoter is methylated, and the gene is thereby silenced (Liu et al., 2019).Usually, the pro-inflammatory factors TNFα and IL-6 are increased in the fibroblasts of these mice; however, this effect is reversed following Meg3 over-expression with fibroblast proliferation also being inhibited (Liu et al., 2019).The protective effect of Meg3 is also seen in diabetic retinopathy (DR), which includes aberrant vascularisation of the retina and eyeball (Chen et al., 2021).
Meg3 acts as part of an axis with miR-6720, and its expression negatively correlates with pro-inflammatory factor expression, leading to inhibition of DR, a reduction in apoptosis and increased migration of retinal microvascular cells (Chen et al., 2021).Meg3 is similarly anti-inflammatory in oesophageal cancer.Normally, this cancer progresses via activated T-helper cells and is accompanied by both pro-inflammatory and malignancy markers (Xu et al., 2021).In Thelper cells, Meg3 inhibits miR-9 inhibition, leading to its indirect up-regulation, suppression of regulatory T-cell proliferation and reduction in the immune escape of cancer cells (Xu et al., 2021).
In a rat model of diabetes, a pro-inflammatory phenotype is associated with reduced Meg3 expression and up-regulation of Rac1 and its activated form Rac1-GTPase (Wang et al., 2021).Over-expression of Meg3 normalises the inflammatory response, which leads to increased neuron survival and alleviates spatial memory impairments (Wang et al., 2021).Meg3 is similarly down-regulated in Alzheimer's disease mouse models, with its over-expression rescuing their proinflammatory skew, amyloid-beta plaque presence and spatial learning and memory (Yi et al., 2019).Finally, Meg3 has been shown to be transiently up-regulated during fear learning acquisition, with its knockdown reducing functional AMPA receptor presence at the membrane after neuronal activation (Tan et al., 2017).This would impair learning behaviour, as affected neurons would be unable to undergo long-term potentiation (Tan et al., 2017).

| Summary
Although the studies on lncRNA discussed above mostly indirectly link memory T-cells and various learning behaviours, consider that each lncRNA is key to various aspects of T-cell function, and affect memory when knocked down.Given that T-cells, microglia, neurons and astrocytes all inter-communicate, and that lncRNAs are extremely functionally plastic, it is likely that common lncRNAs act throughout this cellular axis.The difference, simply, would be that the interacting partners of the lncRNA are different, as are the cellular functions (Figure 1).pathway (Wei, Luo, et al., 2017).Canonically, these include Drosha, and precede processing by Dicer into mature miRNAs, which are 20-25 nt and have a very basic hairpin structure that includes a 'seed sequence', which is required for target binding (Wei, Batagov, et al., 2017).miRNA exert different functions based on their complementarity at target transcript 5′ and 3' UTRs, promoters, or even within the coding sequence, although typically they inhibit their target transcript (Wei, Batagov, et al., 2017).There are several silencing pathways, and even some evidence for the opposite, that is, up-regulation, but all appear to use miRNA as part of an RNAinduced silencing complex (RISC; Wang et al., 2022;Wei, Batagov, et al., 2017).The RISC can include different proteins depending on cell type, subcellular localisation and environmental signals, which provide flexibility that mature miRNAs lack as a result of their size (Wang et al., 2022).Partial miRNA complementarity leads to physical translation blockage of the target, which can be compounded by multiple miRNA/RISC complexes binding the same transcript.Full complementarity marks the transcript for degradation by the RISC (Wang et al., 2022;Wei, Batagov, et al., 2017).

| MicroRNA in memory and adaptive immunity
MiRNAs have been consistently implicated in regulation of the immune system, particularly in the context of cancer and neurodegeneration (Table 1).Among these, miR-34, miR-132 and miR-9 are the best characterised examples, with miR-34 generally acting as a tumour suppressor in cancers, inhibiting the cell cycle to minimise proliferation and metastasis (Rokavec et al., 2014).In adenocarcinoma, miR-34 is inhibited by cancer-induced circRNAs, leading to exhaustion of CD8+ T-cells, another part of the AIS (Wang et al., 2020).Conversely, Zika virus-induced gliomas arising from NSCs have up-regulated miR-34, particularly the miR-34c isoform (Iannolo et al., 2019).This appears to be a protective effect, as miR-34c over-expression reduces glioma severity by pushing cells towards differentiation and adhesion (Iannolo et al., 2019).MiR-34 also has an interesting effect on age-dependent neurodegeneration in Drosophila: miR-34 knockouts present with increased protein synthesis and higher autophagy, evidence of the lack of cell cycle inhibition seen in cancer, as discussed above (Srinivasan et al., 2022).
MiR-132 is similarly involved with the brain's immune system.
In astrocyte models of inflammation, miR-132 expression negatively correlates with pro-inflammatory cytokine expression (Kong et al., 2015;Zhang et al., 2017).This astrocytic activation mirrors that seen in medial temporal lobe epilepsy (Zhang et al., 2022).MiR-132 knockdown in epilepsy models inhibits IL-1β-induced astrocyte activation, reducing epileptic symptoms and neuronal damage (Zhang et al., 2022).This anti-inflammatory effect of miR-132 is also seen in multiple sclerosis models, with over-expression rescuing motor deficits and neuronal death by microglial activation (Zhang et al., 2017).
Unlike miR-34 and miR-132, miR-9 appears to be proinflammatory, particularly in some viral infections, where viral replication positively correlates with miR-9 expression.During Zika virus infection in embryogenesis, both the cortex and corpus callosum are reduced as a result of neuron and NPC apoptosis (Zhang et al., 2019).
The same effects occur with miR-9 over-expression, apparently caused by targeted inhibition of glial cell-derived neurotrophic factor (Zhang et al., 2019).This may indicate that the pro-inflammatory effects of miR-9 are due an imbalance between microglia/astrocyte signalling that prevents microglia from acting as an immune buffer for neurons.Similar effects are seen in both HIV-generated and mouse-modelled immunodeficiency.HIV infection depletes the immune system's resources over time, leading to a dearth of activated T-cells able to effectively fight recurrent infection (Yang et al., 2018).
After infection, miR-9 is up-regulated in astrocyte EVs near infected neurons, and is sent to microglia, inducing migration (Yang et al., 2018).This either creates or is mirrored by miR-9 up-regulation in effector T-cells, leading to inflammation in the brain.The presence of up-regulated miR-9 in activated effector T-cells in multiple sclerosis patients supports all (Jin et al., 2018;Majd et al., 2018).High microglial miR-9 is also seen in gliomas, where it leads to increased glioma proliferation and migration (Tan et al., 2012).
The three miRNAs above have also been heavily implicated in learning.Specifically, miR-34 in cortical function, miR-132 in synaptic plasticity and spatial memory and the circadian rhythm and miR-9 in hippocampal function.The tumour suppressor p53 and p73 proteins regulate several miRNAs, including miR-34a, and when knocked out inhibit brain development, including hippocampal malformation (Agostini et al., 2011).Although knocking down miR-34a does not produce the same developmental malformations, it does increase neuronal arborisation (Agostini et al., 2011).Over-expressing miR-34a reverses this effect, resulting in a decreased number of inhibitory synapses (Agostini et al., 2011).Considering that removing the regulators of miR-34a affect the hippocampus, it is likely that this decrease in neural plasticity with miR-34a over-expression alters a variety of learning and memory-related behaviours.
BDNF stimulation leads to the secretion of MiR-132 in small EVs at hippocampal excitatory synapses, and simultaneously increases vesicular pools, spine density and burst firing (Antoniou et al., 2023).
Over-expressing miR-132 has been shown to increase spine density to the point of creating deficits in novel object recognition and spatial memory (Aten et al., 2018;Hansen et al., 2016).Rather than miR-132 regulating or being regulated by one or two proteins, this over-expression appears to dysregulate multiple kinase pathways to offset neural plasticity (Aten et al., 2018).Finally, MiR-9-3p is implicated in learning and long-term memory (Sim et al., 2016).
Knocking down the 3p, but not the 5p fragment, is sufficient to impair long-term potentiation (LTP) in the CA1 hippocampus without altering baseline excitability (Sim et al., 2016).This was caused by dysregulation of several genes, including up-regulation of dystrophin (Dmd) and SAP97.These decreased LTP in the CA1, and produced consistent deficits in spatial, novel object and fear learning (Sim et al., 2016).

| Summary
miRNAs are well defined as translational inhibitors but also appear to act as intercellular communication molecules under specific circumstances.This is because they are ubiquitously expressed in EVs which, as discussed, are key to intercellular communication and are therefore critical for the AIS and synaptic plasticity.This has been seen in exosomes derived from animal models of MDD, which causes a pro-inflammatory phenotype and anhedonia.It was also seen with miR-132 in astrocytes, which use EVs to communicate.
Considering that circRNAs are frequently found to sponge miRNA, it appears that miRNAs may act as intercellular inhibitors in a balanced manner similar to that between pro-and anti-inflammatory cytokines (Figure 2).Indeed, a recent study has shown that dendritic cells, the key adaptive immune cells present in cerebrospinal fluid, can detect and respond to gene deficiency in nearby cells (Herbst et al., 2023).This process, which has been termed intercellular monitoring, appears to involve TNTs and has been shown in similar cells like macrophages (Herbst et al., 2023).When functionally characterising miRNA, especially by manipulating expression, it is important to consider whether the target cell type interacts with cells capable of this process directly or indirectly via EVs.This is because altering the outgoing EV RNA pool will likely have unintended effects that could be severely detrimental, depending on their environment (Figure 2).Among their binding partners, HuD, ELAV4 and La autoantigens are the best characterised, although Y-RNA can also function independently.As stated, La binding protects and sequesters Y-RNA in the nucleus.This complex is also required for transcription initiation, potentially as a result of helicase-like activity (Boccitto & Wolin, 2019;Fabini et al., 2000).While in the nucleus, Y-RNA can also sequester HuD, a neuron-specific translational enhancer that binds and stabilises mRNA 3' UTRs.This aids processes such as axonogenesis and is therefore involved in many neurodegenerative disorders (Driedonks & Nolte-'t Hoen, 2018).

| Y-RNAs in memory and adaptive immunity
Although Y3 is the best-characterised Y-RNA, no published studies have examined their function in learning, with most, instead, focusing on cancer (Table 1).A study on patient-derived glioblastoma cultures found that the extracellular environment has several characteristics that are distinct from their parent cells.These characteristics, shared among the microvesicle (MV), EV and naked RNP/liponucleoprotein (R/LNP) fractions, include depletion of whole mRNA exons (although truncated transcripts are present) and enrichment of miRNA, and whole and precisely processed Y-RNA fragments.Y-RNAs were most common in naked R/LNP fractions, although the identity of these RNPs was not identified (Wei, Batagov, et al., 2017).It is also unknown what percentage of Y-RNAs, if any, were glycosylated.
Studies on tumour biopsies have supported this Y-RNA abundance in EVs, with 30% of brain tumours but 80% of malignant tumours being enriched in Y-RNAs (Wei, Batagov, et al., 2017).
Y-RNA enrichment in EVs is also tumour and environmental stress specific.Medulloblastoma (MB) and diffuse infiltrative pontine glioma (DIPG), two forms of malignant paediatric tumours, have distinct RNAomes from their parent cells, with some commonly enriched RNAs, namely Y-RNAs (Magaña et al., 2022).While EV-derived whole and fragmented Y-RNAs have an unknown function, it is likely that they simply latch on to their binding partners in the target cell.This could be why Ro60 and the ELAVL family are linked to many pathologies.For example, Ro60 is involved in neonatal lupus, with mouse knockouts having multiple lupus-like phenotypes.When Y-RNA-bound, the resulting RNP can activate toll-like receptors, which are important mediators of inflammation (Boccitto & Wolin, 2019;Clancy et al., 2010).ELAVL/Y-RNA RNPs are also upregulated in both Alzheimer's disease and neuroblastoma (Scheckel et al., 2016).
Y-RNAs are the first example of an RNA that is glycosylated, and that are deposited on the cell membrane.Given this, and that they  vRNAs are named for the RNP with which they were first found to be associated: the 'Vault' (Table 1).The vault and its protein components, major vault protein (MVP), TEP1 and VPARP, are highly structurally conserved, although also exhibit moderate sequence conservation.Structurally, the vault is a hollow barrel formed by two reversibly dimerising halves, with unknown cargo (Eichenmuller et al., 2003;Suprenant, 2002).The vault's function is largely uncharacterised, although experimental evidence suggests that some of the vault's components could function separately from the complex.
Most literature around the vault and its components focuses on the vault's role in the immune system, in particular up-regulation of MVP in response to cancer, infection and inflammation (Teng et al., 2017).
vRNAs are structurally, although not sequentially, conserved and are approximately 140 nt in length.vRNAs are transcribed by RNA polymerase 3 in both mice and humans, and can be further processed by the Dicer enzyme into miRNA-like small-vault RNAs (svtRNAs; Trabucchi et al., 2009).vRNA isoform number differs between species, with one to two in mice and three to four in humans.However, whereas mice have a functional transcript located inside a long (ca.350 nt) pseudogene, humans have three functional vRNAs and one pseudogene, all of which structurally resemble the functional mouse vRNAs.vRNAs have a basic predicted stem-loop structure, with one large main stem and one to two smaller stems around a central loop (Kickhoefer, 2019).The mouse pseudo-vRNA is the exception to this rule as a result of its length, as it comprises many stem loops and has predicted Sam68-and Ksrp-binding domains (Trabucchi et al., 2009).
Both vaults and vRNA are found throughout the body, with high levels of vRNA being found in the synaptic compartment.vRNAs are also highly abundant in EVs, although most studies have focused on vRNAs in cancer and cancer-derived EVs (Latowska et al., 2020;Spinelli et al., 2018).Furthermore, human vRNAs are differentially expressed in response to environmental cues such as infection and inflammation, which indicates that each human vRNAs may have a different functional complement (Li et al., 2013).This has also been observed with svtRNAs, which appear to function like miRNA in mRNA inhibition (Miñones-Moyano et al., 2013).unknown physiological consequences (Wakatsuki et al., 2021).
Considering that the vault itself has previously shown involvement with the MAPK pathway, it is possible that these data are at least partially caused by impaired vault function via Vaultrc5 knockdown (Steiner et al., 2006).
Studies have directly implicated the vRNA in neurodegeneration via two roles, svtRNA in Alzheimer's disease and sequestosome 1 function in autophagy.Following NSUN2 methylase knockout studies producing intellectual disability and Alzheimer's pathology, RNAseq in human fibroblasts after NSUN2 was found to methylate HVG1-3.
Interestingly, NSUN2 had a dose-dependent effect on Dicer-mediated svtRNA production, with the HVG1 fragment svtRNA4 being progressively decreased in response to NSUN2.SvtRNA4 binds RISC and inhibits a voltage-gated calcium channel which could be partially responsible for the intellectual disability phenotype seen in NSUN2 knockouts (Hussain et al., 2013).The apparent main vRNA (HVG1 in humans and MVG1 or Vaultrc5 in mice) is required, as mentioned, for Sequestosome 1 (SQSTM1/p62) function in autophagy.P62 is an autophagic scaffold protein that undergoes self-oligomerisation and binds ubiquitin, then, in concert with TNF6, regulates the NF-kβ pathway, a major autophagic process (Horos et al., 2019).Oligomerisationdeficient p62 is dysfunctional in autophagy, and therefore implicated in many neurodegenerative diseases.HVG1 binds p62 at its stemloop regions, inhibiting its function and therefore attenuating p62's control over autophagy (Horos et al., 2019).
Given that vRNAs are responsive to infection and inflammation, an investigation of this inhibition would be central to our understanding of neurodegenerative disease, particularly as both human and mouse vRNAs interact with p62.Huntington's disease (HD) is one example of vRNAs' prominence in neurodegenerative disease.
HD mouse models were able to show an age-dependent cortical decrease in both sno-and vRNA (Chanda et al., 2018).Whether this is caused by differential expression of vRNA or increased secretion into vRNA-containing EVs as often seen in inflammation, likely for intercellular communication, remains to be determined.
Despite being a sncRNA family, vRNAs appear to act like Swiss army knives, displaying multifunctionality on par with the much larger lncRNAs.This comes primarily from the fact that the 'major' vRNA, MVG1/HVG1, is part of HuD, SQSTM1 and the vault RNP.This is despite a lack of common protein motifs between them, with only the vRNA HuD and SQSTM1 functions identified.Furthermore, vRNAs are environmentally responsive, and are consistently secreted in EVs.Their processed forms when methylated, svtRNA, also seem to behave like miRNA, with their own subset of functions.
vRNAs may therefore be small 'effector hubs', able to bind multi-

| SUMMARY AND CON CLUS ION
It is evident that the activity of many ncRNAs will not be specific for either the brain's immune system or in support of learning and memory.There are two main reasons that these two systems could Instead, re-purposing pre-existing ncRNA molecules, which have flexible structures that can be chemically modified, would be the least energy-intensive way of transmitting, and likely encoding, information.This is because although RNA throughout the body have a short half-life, ncRNAs in the brain have much longer 3' UTRs, markedly extending their half-lives (Holdt et al., 2018).
ncRNAs evolved primarily via class I retrotransposition (Ganesh & Svoboda, 2016).In this process, a transcript is duplicated using its internal machinery via an RNA intermediate, and then inserted at a different point in the genome (Ferrari et al., 2021;Ganesh & Svoboda, 2016).This can be accompanied by flanking sequences in a process known as 5′ or 3′ transduction, and can produce gene and exon duplication, shuffling and deletion throughout the genome (Ferrari et al., 2021).These transduced transcripts were be co-opted to become part of or regulate existing pathways or give rise to entirely new ones (Ferrari et al., 2021).Indeed, this is evident on a cellular level as fusogens are able to couple different cell subtypes.When this occurs in neurons, it leads to information exchange that alters the function of the fused neuron which likely involves a myriad of ncRNA (Giordano-Santini et al., 2016;Giordano-Santini et al., 2020).
Interestingly, if an engram and non-engram neuron were connected by fusogens, or by a similar avenue such as TNTs, they could exchange information and therefore expand the engram even after learning.Combined, these characteristics provide a putative pathway by which pre-existing ncRNA involved in responding to, and protecting cells from, their environment, that is, the immune system, could be co-opted.Initially, they may have been co-opted to act as a 'record' for highly valent environmental stimuli and, over time, have grown complex enough to develop into organ we know today as the brain.
In summary, there are three potential avenues by which ncRNAs may functionally intersect learning and the brain's AIS: as effectors, scaffolds and messengers (Figure 3).Effectors are those ncRNAs that act alone, like those involved in tRNA processing, or form part of larger complexes to modify other macromolecules, such as Y-RNA (Ishida et al., 2020).Scaffolds include lncRNAs, which act as hubs for miRNA networks in a myriad of processes.Perhaps the most interesting is the largely uncharacterised potential of small ncRNAs to act as messengers, both between cells and across subcellular compartments.
Particularly considering that ncRNA structure changes depending on their location as a result of the interacting partners and surrounding environment, ncRNA could provide flexibility in passing information between cells of the adaptive immune system and the CNS.
Although the organelle content of these nodes is unknown, it may include mitochondria which control local calcium transients (Noriega-Prieto & Araque, 2021).Each astrocyte's spongiform domain contains neurotransmitter G-protein-coupled receptors to detect and individually respond to multiple neuronal signals by increasing intracellular calcium, which releases gliotransmitters (Noriega-Prieto & Araque, 2021; Santello et al., 2012).The tripartite synapse therefore exists in a feedback loop, using neuro-and gliotransmitters to respond to and regulate synaptic activation and plasticity.

|
lncRNAs may typically function as 'scaffolds' and bind with and regulate a broad variety of RNA and proteins.This supposition is driven by the observation that Gomafu, Gas5 and Malat1 interact with distinct protein partners, and Malat1 and Meg3 have the ability to act as decoys for miRNAs (Figure1).Considering their unique complex structures and long sequences, it is likely that all four ln-cRNAs have a plethora of RNA-and protein-binding sites.The major ramifications of lncRNA containing such highly plastic modular domains are as follows: first, all functionally annotated lncRNA are likely not limited to a single pathway or system, and instead exert their influence across multiple systems in a state-and structuredependent manner.Second, functional changes may occur because the lncRNA can bind many different molecules at once.This means that targeting part of a lncRNA or one of its binding partners could F I G U R E 1 LncRNA Meg3 at the interface between the adaptive immune system and neuronal plasticity associated with learning and memory.In the retina (left), (1) inflammatory neovascularisation up-regulates Meg3, which (2) inhibits miR-6270, (3) inhibiting further inflammation.In the neuron (right), (1) Meg3 is upregulated at the synapse during fear learning and, (2) when knocked down, leads to reduced AMPA receptor surface expression, implicating Meg3 in synaptic trafficking and learning.Created with BioRe nder.com.have downstream effects on the nominally unaffected binding partners of that lncRNA.Structure and function miRNAs are highly conserved and abundant in the CNS, and comprise the best-characterised ncRNA family in both learning and adaptive immunity.Beginning as large 'pri-miRNAs' with a 5′ cap and 3′ poly-A tail, they are processed into approximately 70-nucleotide (nt) 'pre-miRNAs' by different enzymes depending on the biogenesis RNAs that may bridge the gap between memory and the immune system5.4.1 | Y-RNAsStructure and broad function Y-RNA is a sncRNA (80-120 nt) family that is both structurally and sequentially conserved in vertebrates.Y-RNA numbers vary across species, with two in mice and four in humans, although there are hundreds of Y-RNA pseudogenes in both(Valkov & Das, 2020).Y-RNAs are transcribed by RNA polymerase 3, producing a poly(U) tail that allows binding to the La autoantigen, which protects Y-RNA from exonucleases and promotes their nuclear retention(Valkov & Das, 2020).From there, most Y-RNAs associate with Ro60 and are exported to the cytoplasm, although some remain in the nucleus and/or are exported via different pathways, such as HuD or La autoantigen(Driedonks & Nolte-'t Hoen, 2018).Shaped as their name suggests, Y-RNA typically have two main stems, with hairpins formed by 3′ to 5′ end base pairing and joined by a pyrimidine-rich single-stranded RNA loop.While the stem base is required for nuclear export, the main stem binds Ro60, forming the Y-RNA/Ro60 ribonucleoprotein complex (Y/Ro RNP) for cytoplasmic export(Valkov & Das, 2020).The variable sequence at the top of the stem can also bind other proteins to form specialised RNPs.Similar to most of the ncRNAs discussed thus far, the expression pattern of Y-RNAs differs with respect to tissue type and disease state, although their function is not well characterised.Broadly, they can be split into those which are Ro60 bound and those bound to other proteins.Ro60 itself is a small (60 kDa), highly conserved autoantigen that, based on its ring-like structure and extensive RNA binding regions, may sequester misfolded RNA(Driedonks & Nolte- 't Hoen, 2018).The Y /Ro RNP has three proven modes of affecting cellular stress.First, its nuclear accumulation, which enhances survival after UV or oxidative stress.Second, that Y-RNA binding is F I G U R E 2 MicroRNA miR-9 at the interface between the adaptive immune system and neuronal plasticity associated with learning and memory.In astrocytes (left), (1) neuronal infection up-regulates miR-9, which is (2) secreted in extracellular vesicles to create or mirror miR-9 up-regulation in T-cells.(3) This increases inflammation.In hippocampal neurons (right), (1) miR-9-3p is present at the synapse, and (2) knocking it down up-regulates SAP97 and Dmd to impair learning.Created with BioRe nder.com.required for Y/Ro exocytosis in EVs, and to simulate TNFα release from macrophages.Finally, Y-RNA secretion in EVs from macrophages in response to cancer, apoptosis, infection and inflammation(Boccitto & Wolin, 2019;Gulìa et al., 2020).

Functioning
solo, or at least with unknown binding partners, Y-RNAs are ubiquitously expressed in EVs and are the most abundant ncRNA in the extracellular environment.They are also seen during cancer, as tumour cells produce EVs containing both whole and precisely fragmented Y-RNAs, which circulate between tumour cells and healthy tissue (Driedonks & Nolte-'t Hoen, 2018; Gulìa et al., 2020).Moreover, Y-RNAs have the unique distinction of being the first glycosylated RNAs embedded in the external cell membrane.These glycans are deposited on the N-terminus and are highly fucosylated and sialylated.On the cell surface, Y-RNAs appear to bind sialic acid-binding immunoglobin-like (Siglec) receptors, which are required for the recognition of self-versus non-self in immune cell function (Flynn et al., 2021).Therefore, Y-RNAs are key to immune function when present in the nucleus, in EVs and on the external cell membrane.
are required for Ro60-mediated immune responses and bind HuD, it appears that they behave like keys.With Ro60, Y-RNAs appear to sequester misfolded mRNA to aid in cell survival, although this function varies depending on the isoform.Binding to HuD/ELAV4 produces an opposing effect, as this protein stabilises mRNA 3' UTRs, but is sequestered by Y-RNAs.Although whether these functions occur with glycosylated Y-RNA is unknown, these modified Y-RNAs are exposed to the extracellular environment and bind Siglec receptors.These receptors regulate adaptive immunity, and may be used by neurons to distinguish self from non-self.Y-RNAs therefore function as both non-specific mRNA inhibitors, and as ligands or receptors, depending on whether Siglec binding also leads to the bound Y-RNA-inducing changes in the host cell.Furthermore, Y-RNAs are the most abundant ncRNA in plasma and EVs, and so may fulfil another function there.Alternatively, these Y-RNAs could be present on the EV surface to activate target cell Siglecs, or form part of a cascading inhibition of mRNA through HuD and Ro60.In unpublished studies, we have found that both Y-RNAs are highly enriched in the synaptic compartment.Their levels fluctuate with neuronal activation -both in vitro and in vivo -in response to fear learning, where they exhibit a significant decrease at the synapse without corresponding increases in the cytoplasm.This suggests that Y-RNA may be secreted, which would fit with their presence in plasma EVs, and as glycosylated RNAs on the cell membrane.In addition to regulating the immune response and potentially recognition of self versus non-self.Ongoing experiments are investigating the functional relevance of Y-RNA at the synapse, and how this relates to learning and memory.Using a targeted RNA degradation construct coupled with single-molecule tracking, this will allow us to determine not only the behavioural effect of isoform-specific Y-RNA knockdown but also its effect on synaptic morphology and intercellular communication.

5. 5 |
Vault-associated RNA (vRNA) 5.5.1 | Structure and broad function vRNAs have at least three functions, only two of which are known for the mouse vRNAs: sequestosome 1 function in autophagy and synaptic MAPK signalling.Otherwise, the vRNAs could affect the vault's cargo, or its trafficking across the neuron; and finally, given that vRNAs are released in EVs, they could be simply part of the vault's cargo and go on to fulfil other functions in intercellular communication.Whether the MAPK function overlaps with this possibility has not been explored.5.5.2 | vRNAs in memory and adaptive immunityLike Y-RNAs, vRNAs are ubiquitously expressed, so it is surprising that no studies have focused on vRNA (or svtRNA) in learning and memory.Most have focused on pathology, with the remainder only tangentially noting vRNA's presence.The sole exception to this is a recent study involving MVP, the mouse vRNA and the MAPK pathway.In addition to MVP being phosphorylated by Aurora A kinase, both it and Vaultrc5 were found to co-immunoprecipitated with MEK, part of the MAPK pathway(Wakatsuki et al., 2021).This was not, however, seen with the human vRNA, which could indicate a species-specific function.Antisense oligonucleotide-mediated knockdown of Vaultrc5 also decreased the number of synapses in cortical cultures, potentially through the MAPK pathway, with ple different proteins and RNA to affect their interactions.For example, vRNAs may facilitate the vault's transit around the neuron as they are bound to the vault cap, which is where the vault binds microtubules to transit the neuron.Alternatively, vRNA may package any vault cargo for transport to the synapse, as vRNAs require another vault component called TEP1 to associate with MVP, and vaults co-immunoprecipitated with La autoantigen, which has several RNA and protein-binding sites.In either case, vRNAs appear to fulfil a unique function at the synapse, as they are packaged into EVs, whereas the full vault is not.This synaptic function could, however, incorporate vRNA binding to SQSTM1.Additional unpublished experiments in our lab have shown that Vaultrc5 is highly enriched at the synapse and, like Y-RNA, appears to be secreted from the neuron.Likewise, this fits with the vRNAs being consistently detected in EVs.Furthermore, transcriptionally degrading Vaultrc5 attenuates fear extinction learning, and single molecule tracking of the vault itself detected extensive pausing along dendrites, indicative of a role in regulating the dendrite.Supporting this role, immunoprecipitation of the vault has detected specific RNA cargo relating to neuron development and synaptic maintenance and plasticity.Therefore, in addition to vRNAs responding to immune signalling, and functioning in autophagy, they, and likely the vault itself, are integral to learning and memory formation.Next steps are to determine whether transcriptionally degrading Vaultrc5 affects vault movement and cargo within the neuron, and what the consequences of this are at the synapse.
be intertwined: conservation of energy and ncRNA flexibility.First, the brain is the body's most energy-consuming organ, requiring 20% of daily glucose intake and transcribing the majority of ncRNA, although this number fluctuates with context.The major ramification of this is that molecules that do not function in the immediate context are only transcribed at very low copy numbers.Considering the abundance of ncRNA species and families that are frequently differentially expressed in response to learning, despite only having known functions in, for example, glioblastoma, it makes little sense for these ncRNAs to be expressed in only one context.

F
Bridging the gap between memory and the immune system.LncRNA scaffolds bind with, and regulate, a broad variety of RNA and proteins.miRNA can be intercellular communicators secreted in vesicles in response to environmental signals, and which inhibit intracellular translation.Y-RNAs can be used to activate DNA replication and act as receptors on adjacent cells.Vaults, putative trafficking granules, can carry coding and non-coding RNA.vRNA effector hubs bind different proteins and RNA in response to environmental signals.Created with BioRe nder.com.