piRNAs and endo-siRNAs: small molecules with large roles in the nervous system

Abstract

Since their discovery, small non-coding RNAs have emerged as powerhouses in the regulation of numerous cellular processes. In addition to guarding the integrity of the reproductive system, small non-coding RNAs play critical roles in the maintenance of the soma. Accumulating evidence indicates that small non-coding RNAs perform vital functions in the animal nervous system such as restricting the activity of deleterious transposable elements, regulating nerve regeneration, and mediating learning and memory. In this review, we provide an overview of the current understanding of the contribution of two major classes of small non-coding RNAs, piRNAs and endo-siRNAs, to the nervous system development and function, and present highlights on how the dysregulation of small non-coding RNA pathways can assist in understanding the neuropathology of human neurological disorders.

1. Introduction

The landmark discovery of microRNAs (miRNAs) in the nematode Caenorhabditis elegans and their evolutionarily conserved post-transcriptional modulatory roles during development was groundbreaking in that it launched a new era of tiny RNAs as unorthodox potent gene regulators (Lee et al. 1993; Pasquinelli et al. 2000; Reinhart et al. 2000). Since then, two additional major categories of small noncoding RNAs (sncRNAs) have been identified: PIWI-interacting RNAs (piRNAs) and endogenous siRNAs (endo-siRNAs). Each of these sncRNAs associate with specific Argonaute (Ago) proteins to target complementary RNAs to regulate gene expression at the transcriptional or post-transcriptional levels in a process called RNA interference (RNAi) (Billi et al. 2014). To date, much has been learned about the biology of sncRNAs that make them highly appealing gene therapy tools for the treatment of human diseases such as cancer, viral pathologies, and a number of genetic disorders (Saulnier et al. 2006; Lares et al. 2010; Setten et al. 2019).

In this review, we present an overview of the roles of piRNAs and endo-siRNAs on neuronal development and function in model organisms, what insights have been gathered from their misregulation, and their application on human neurodegenerative diseases. The founding class of sncRNAs, miRNAs, are amongst the best studied small RNAs with neural modulatory roles. Their extensive role in the nervous system have been covered recently in a series of excellent reviews (Bushati and Cohen 2007; Busto et al. 2017; Carthew et al. 2017; Ambros and Ruvkun, 2018; Bartel 2018). Other types of non-coding RNAs, such as long non-coding RNAs (lncRNAs), circular RNAs (circRNAs), and small nucleolar RNAs (snoRNAs), are also reported to have roles in the nervous system and are reviewed elsewhere (Salta and De Strooper 2017; Constantin 2018; Leighton and Bredy 2018; Kristensen et al. 2019; Li et al. 2019).

1.1. PIWI interacting RNAs (piRNAs)

Although all the different types of sncRNAs regulate gene expression, the biogenesis and regulatory mechanisms of piRNAs, miRNAs, and endo-siRNAs are unique to each RNA species. piRNAs range between 21-32 nucleotides (nt) long with an evolutionarily conserved role in silencing mobile genetic elements in the germ line. Distinct features differentiate piRNAs from siRNAs and miRNAs including a bias for a uridine (U) at the 5’ end and a 2’-O-methylated 3’ end (Ruby et al. 2006; Brennecke et al. 2007; Kirino et al. 2007; Saito et al. 2007). piRNAs are by far the most abundant and diverse class of sncRNAs, numbering >68 million in mice, >8 million in humans, nearly 42 million in Drosophila melanogaster, and >28,000 in Caenorhabditis elegans (Wang et al. 2019). piRNA genes are derived from transposons, protein coding genes, or intergenic regions and are clustered in large genomic regions (Ruby et al. 2006; Brennecke et al. 2007; Batista et al. 2008; Das et al. 2008; Wang and Reinke 2008; Li et al. 2013). piRNA cluster loci are transcribed as long single-stranded precursors that are processed into primary piRNAs, which initiate the “ping-pong” loop, a piRNA amplification mechanism in Drosophila, mice and zebrafish (Aravin et al. 2008; Brennecke et al. 2007; Houwing et al. 2007). However, in C. elegans, piRNA amplification is achieved by the recruitment of a RNA-dependent RNA polymerase (RdRP) by the PRG-1 Piwi protein instead of the ping-pong cycle (Das et al. 2008; Gu et al. 2009). Readers who want more information on the mechanism of piRNA biogenesis are invited to consider several excellent reviews that cover the topic in more detail (Weick and Miska 2014; Huang et al. 2017; Czech et al. 2018; Rojas-Rios and Simonelig 2018; Ozata et al. 2019; Parhad and Theurkauf 2019).

piRNAs associate with the PIWI (p-element induced wimpy testis) protein that was originally discovered in Drosophila melanogaster as being essential for stem cell maintenance in the male and female germ lines (Lin and Spradling 1997). Its function in stem cell self-renewal appears to be an evolutionarily conserved role as piwi genes have been recognized in mammals, worms, plants, and aquatic invertebrates (Cox et al. 1998; Seipel et al. 2004; Funayama et al. 2010). Four PIWI proteins have been identified in humans: PIWIL1 (also called HIWI), PIWIL2 (HILI), PIWIL3 (HIWI3), PIWIL4 (HIWI2); three in murine species: PIWIL1 (MIWI), PIWIL2 (MILI), and PIWIL4 (MIWI2); and three in Drosophila: Piwi, Aubergine (Aub) and Argonaute 3 (Ago3) (Table 1). C. elegans has one functional PIWI, PRG-1. Although PRG-1 shares 91% identity at the amino acid level with a second PIWI, PRG-2, a mutation in prg-2 alone does not manifest in any changes of piRNA levels nor does it exacerbate the germline phenotype of a prg-1 mutant (Sharma et al. 2001; Deng et al. 2002; Kuramochi-Miyagawa et al. 2004; Brennecke et al. 2007; Carmell et al. 2007; Nishida et al. 2007; Batista et al. 2008; Das et al. 2008; Wang and Reinke 2008). The piRNA-PIWI partnership relies on piRNAs recognizing their RNA targets by base-pair complementarity while PIWI assumes the effector role (Shen et al. 2018; Zhang et al. 2018). piRNA-PIWI complexes silence their targets at the transcriptional level by recruiting chromatin modifiers to genomic loci and establishing heterochromatic marks or de novo DNA methylation and at the post-transcriptional level through cleavage of target mRNAs by PIWI’s slicer activity (Pal-Bhadra et al. 2002; Aravin et al. 2008; Kuramochi-Miyagawa et al. 2008; Wang and Elgin 2011; Rajasethupathy et al. 2012; Sienski et al. 2012; Le Thomas et al. 2013). In Drosophila, Piwi protein/piRNA complexes engage in the silencing of transposable element expression at the post-transcriptional level while also recruiting proteins, such as heterochromatin protein 1 (HP1a) and Maelstrom (Mael), to establish heterochromatic regions to silence the expression of transposon loci (Hyun 2017; Sato and Siomi 2018).

Table 1.

PIWI proteins in animal models.

Human	Murine	Drosophila	C. elegans
PIWIL1 (HIWI)	PIWIL1 (MIWI)	Piwi	PRG-1
PIWIL2 (HILI)	PIWIL2 (MILI)	Aubergine (Aub)	PRG-2*
PIWIL3 (HIWI3)	PIWIL4 (MIWI2)	Argonaute 3 (Ago3)
PIWIL4 (HIWI2)

^*See Section 1.1

Although PIWIs and their piRNA binding partners have historically been fixtures of the germ line (Lin and Spradling 1997; Cox et al. 1998; Batista et al., 2008), the distribution and function of piRNAs and PIWI proteins in the soma is well documented, including Drosophila salivary glands (Brower-Toland et al. 2007), ovarian and testes somatic cells (Cox et al. 1998, 2000; Malone et al. 2009), embryos (Rouget et al. 2010), and fat body (Jones et al. 2016); human hematopoietic stem cells (Sharma et al. 2001); rat liver (Rizzo et al. 2014); mouse lung and pancreas (Yan et al. 2011; Wasserman et al. 2017); macaque testes somatic cells (Yan et al. 2011); jellyfish muscle (Seipel et al. 2004); sea squirts (Rinkevich et al. 2010; Rabinowitz et al. 2013); and planarian neoblasts (Reddien et al. 2005).

1.2. Endogenous small interfering RNAs (endo-siRNAs)

Another category of sncRNA are endogenous small interfering RNAs (endo-siRNAs). Endo-siRNAs are 20-24 nt long sncRNAs that typically silence their RNA targets and are thought to have evolved to protect against genomic invaders such as viruses or transposons (Ding and Voinnet 2007; Grishok 2013). Endo-siRNAs are derived from long double-stranded RNA (dsRNA) precursors which are then processed by the RNase III endonuclease DICER to generate ~22 nt RNAs (Czech et al. 2008; Ghildiyal et al. 2008; Kawamura et al. 2008). The sense strand from the cleavage product is degraded while the anti-sense, or guide strand, is incorporated into RISC (RNA induced silencing complex) which includes an Argonaute and other co-factors to effect the degradation of a target RNA (Billi et al. 2014; Almeida et al. 2019). In most model organisms, endo-siRNAs, unlike piRNAs, are not a prominent class of sncRNAs. However, endo-siRNAs do represent a major category of sncRNAs in C. elegans, where they are predominantly generated by RdRPs and are modified so as to contain a 5’ triphosphate group, a 3’ OH group, and are biased for a guanosine (G) at their 5’ end (Ambros et al. 2003; Ruby et al. 2006; Pak and Fire 2007; Sijen et al. 2007; Gu et al. 2009). Contrary to piRNAs, whose role in neuronal processes in a broad spectrum of organisms is unambiguous, the existence of endo-siRNAs in the mammalian nervous system is uncertain and largely unexplored despite the presence of the RNAi machinery in mammalian neurons, which shares some of the same components as the endo-siRNA pathway. This may be in part due to the pervasive belief that siRNAs are primarily silencers of transposons and foreign genetic invaders and not regulators of endogenous genes per se. In addition, assigning a small RNA as a bona fide siRNA, which is a DICER-dependent product derived from dsRNA, has proven difficult (Smalheiser et al. 2012; Leighton and Bredy 2018).

Gene regulation by endo-siRNAs can occur through two distinct mechanisms: post-transcriptional gene silencing (PTGS) and transcriptional gene silencing (TGS). PTGS is a cytoplasmic event and occurs when endo-siRNAs associate with their Ago partners to inhibit translation or to effect the destabilization of their cognate target mRNAs. PTGS is well studied and is associated with the RNAi silencing mechanism that is triggered by an exogenous source of dsRNA. TGS is a strictly nuclear mechanism whereby an Ago protein guided by an endo-siRNA silences a target RNA by inhibiting transcription and promoting heterochromatin formation at the target locus (Castel and Martienssen 2013; Schraivogel and Meister 2014; Weinberg and Morris 2016). In C. elegans, the somatic nuclear Ago NRDE-3 shuttles from the cytoplasm to the nucleus when associated with an endo-siRNA, and together with other nuclear factors, silence somatic genes through inhibition of RNA polymerase II elongation and recruitment of histone methyltransferases and other chromatin modifiers to deposit heterochromatic marks (Guang et al. 2008, 2010; Burkhart et al. 2011; Burton et al. 2011).

2. Evidence of piRNA function in the nervous system

2.1. piRNAs in the Aplysia central nervous system

Since their discovery, the function of piRNAs was thought to be primarily to silence transposons and regulate reproduction-related loci in germline tissues. However, in 2012, a remarkable discovery in the sea slug Aplysia defied the restrictive role of the piRNA pathway as a custodian of the reproductive system. While analyzing small RNA libraries to identify miRNAs in the central nervous system (CNS) that might regulate long-term memory, Rajasethupathy and colleagues noticed a second class of small RNAs between 27-30 nt in length that exhibited a 5’ U bias and 2’-O methylation at their 3’ ends, both trademark features of piRNAs. Two of these CNS piRNAs, piR-1 and piR-2, as well as Piwi protein, were confirmed to be present in the brain by quantitative Northern blot analysis and Western blotting. To further investigate the function of the neuronal piRNA/Piwi complexes in memory-related synaptic plasticity, the authors used an elegant coculturing system whereby each of two sensory neurons synapse on a single target motor neuron. By injecting an antisense 2’-O-methyl oligoribonucleotide against Piwi in one sensory neuron of the coculture (and leaving the second neuron unperturbed as the internal control), the level of serotonin (5HT)-induced electrical activity across synaptic connections between the affected sensory neuron and target motor neuron was recorded, effectively measuring the changes in synaptic transmission and memory-related long-term facilitation (LTF) at the synapse. Knockdown of Piwi diminished LTF following exposure of 5HT, indicating that the activity of Piwi and its associated piRNAs affect memory-associated LTF in response to a neuromodulator. In particular, it was found that Piwi, complexed with the aca-piR-F piRNA, regulates the transcription of CREB2, a transcriptional repressor and principal inhibitor of LTF, by increasing the methylation state of a CpG island at its promoter in a serotonin-dependent manner, resulting in the down-regulation of CREB2 expression and the persistence of memory-related synaptic plasticity. These findings show that Piwi/piRNA complexes control long-lasting changes in neurons for the persistence of memory in Aplysia (Rajasethupathy et al. 2012), and introduced an undisputable role for a somatic piRNA pathway in the nervous system (Table 2).

Table 2.

Neuronal functions for piRNAs or PIWI proteins in animal models.

Model organism	Neuronal roles	References
Aplysia	Long term memory	Rajasethupathy et al. 2012
C. elegans	Dauer nictation behavior	Lee et al. 2017
	Axon regeneration	Kim et al. 2018
	Transgenerational inheritance of pathogen avoidance	Moore et al. 2019
Drosophila	Brain tumorigenesis	Janic et al. 2010
	Transposon silencing	Perrat et al. 2013
	ALS model	Wakisaka et al. 2019
Murine	Dendrite spine development	Lee et al., 2011
	Rett Syndrome model	Saxena et al. 2012
	Cerebral cortex development	Zhao et al. 2015
	Locomotor and exploratory behavior	Nandi et al. 2016
	Nerve regeneration	Phay et al. 2016; Sohn et al. 2019
	Contextual fear memory	Leighton et al. 2019

2.2. piRNA neuronal expression and function in C. elegans

Recently, intriguing somatic functions for piRNAs (also known to as 21Us in C. elegans) regarding development and behavior have been discovered in nematodes. First, the piRNA pathway was found to repress axon regeneration following axonal injury. In C. elegans, neuronal regeneration has been characterized for the GABA motor neurons and mechanosensory neurons but remains largely unknown for most neurons (Byrne and Hammarlund 2017). Mechanosensory neurons in nematodes extend axons that transverse from the anterior to the posterior axis of the animal (~1mm for a mature adult) and are responsible for an animal’s response to touch (Chalfie and Sulston 1981). Proteins involved in the transcription (PRDE-1, TOFU-3, and TOFU-5), maturation (PRG-1 and TOFU-7), and secondary amplification (ELK-1, DRH-3, and EGO-1) of piRNAs were found to inhibit the regrowth of the posterior lateral microtubule (PLM) mechanosensory neuron following laser axotomy. Unexpectedly, the axonal regeneration function for PRDE-1 and PRG-1, two essential constituents for piRNA expression in C. elegans, was found to be cell-autonomous and independent of the gonad (Kim et al. 2018). Future work will determine whether the piRNA pathway represents a breakthrough in mechanisms of neuronal regeneration research.

piRNA pathways have also been shown to regulate complex behaviors in C. elegans animals. Exposure to unfavorable growth conditions during early larval stages promotes entry into dauer diapause, which is thought to act as a dispersal mechanism for larvae to locate more favorable environments for growth (Frézal and Félix 2015). To facilitate dispersal, dauer larvae perform a peculiar behavior called nictation in which a worm balances upright on its tail while swinging its body in the air in order to attach to a passing insect or slug. This dauer-specific phoretic behavior is governed by the cholinergic IL2 neurons (Lee et al. 2011b). Using quantitative behavioral assays and linkage mapping, nictation was mapped to a quantitative trait locus (QTL) referred to as nict-1 that harbors four protein coding genes, three pseudogenes, two noncoding RNA genes, and nearly 300 piRNA genes. Further analysis revealed that the piRNA genes were the likely contributors to nictation behavior. Consistent with this result, prg-1/PIWI mutants were defective for nictation, revealing that the piRNA pathway is responsible for regulating a stereotypical behavior in larvae (Lee et al. 2017). Furthermore, evidence suggests that piRNAs can also facilitate transgenerational inheritance of learned behaviors in C. elegans. In the soil, nematodes encounter a myriad of potentially pathogenic microorganisms and have adapted means to mitigate them. The common laboratory strain of C. elegans, N2 Bristol, exhibits attractive behavior towards the pathogenic bacteria Pseudomonas aeruginosa PA14, but avoids the bacteria after four hours (Zhang et al. 2005). Recent work has shown that the learned PA14 avoidance behavior is transgenerationally inherited for up four generations. By comparing the transcriptional changes between mothers and progeny fed with either its control Escherichia coli OP50 diet or with the PA14 pathogen, it was found that the multigenerational transmission of PA14 memory is dependent upon an induction of the transforming growth factor β (TGF-β) signaling ligand, DAF-7, in a single pair of ASI sensory neurons in the progeny (Moore et al. 2019). Because small RNAs have been implicated in transgenerational epigenetic inheritance (See Section 3.1; Rechavi et al. 2014), Moore et al. also examined small RNA populations and found that the expression of over 700 piRNAs changed in response to PA14 exposure. The authors demonstrated that PRG-1/PIWI is required for the upregulation of daf-7/TGF-β in ASI neurons of progeny of PA14 exposed mothers, and that PRG-1/PIWI and proteins downstream, such as histone H3K9 methyltransferase, SET-25, and the H3K9me3 reader, HPL-2, are necessary for the inheritance of PA14 avoidance (Moore et al. 2019). While these studies do not conclusively show that the piRNA pathway is acting in neurons to regulate these behaviors, together they provide intriguing evidence that piRNA functions extend beyond the germ line to regulate specific behaviors in C. elegans.

2.3. piRNA neuronal expression and function in Drosophila

Similar to C. elegans, studies suggest that piRNAs also function to regulate neuronal development and behavior in Drosophila. Since the discovery of Piwi and its PIWI class members, Aub and Ago3, in the germ line, the presence of piRNA pathway members has also been documented in Drosophila somatic tissues (Brower-Toland et al. 2007; Yin et al. 2007; Li et al. 2009; Malone et al. 2009; Rouget et al. 2010; Jones et al. 2016). Mutations in the Drosophila tumor suppressor gene lethal (3) malignant brain tumor, l(3)mbt, cause larval brain tumors due to the malignant transformation of adult optic neuroblasts and ganglion mother cells (Gateff et al. 1993). Transcriptome profiling of l(3)mbt larval brains resulted in the categorization of 102 genes with aberrant expression as l(3)mbt signature tumor genes. Surprisingly, these signature l(3)mbt tumor genes were enriched for genes whose expression was required in the germ line, including piwi and aubergine, as well as 19 piRNAs. Mutations in piwi and aubergine resulted in the attenuation of malignant l(3)mbt tumor growth in larval brains, suggesting that the mis-expression of piRNAs and Piwi pathway members can drive cancer development (Janic et al. 2010). Although the direct targets of these small RNAs have not been identified, this study offers valuable insights on a possible molecular launchpad for the biogenesis of certain types of cancers.

Mobile or transposable elements can account for 45% to a staggering 90% of the human and certain plant genomes, respectively, and are thought to be a significant thrust in genome evolution. In Drosophila, where transposons constitute nearly 20% of the fruit fly genome, transposition events provide neural diversity, behavioral trait variability, and induce neurodegeneration (Kaminker et al. 2002; Kazazian et al. 2004). The mushroom bodies (MBs) in the Drosophila brain are crucial for olfactory memory and can be subdivided into the α’β’, γ, and αβ neurons. The αβ neurons exhibit significantly higher transposon expression and reduced levels of the Aub and Ago3 proteins compared to other types of MB neurons, indicating that transposon suppression is curtailed in the MB αβ neurons. Additionally, more than 60% of transposable elements exhibited significantly higher expression in the brain of mutants defective for aub, ago3 and the piRNA biogenesis factor armitage (Perrat et al. 2013). Thus, the piRNA pathway promotes neural mosaicism and genome heterogeneity in the Drosophila brain by controlling transposon silencing in a region important for learning and memory.

2.4. piRNA neuronal expression and function in murine model systems

In the past decade, studies have debated whether piRNAs or PIWI proteins are present in the mammalian nervous system (Dharap et al. 2011; Lee et al. 2011a; Saxena et al. 2012; Zhao et al. 2015; Ghosheh et al. 2016; Nandi et al. 2016; Roy et al. 2017; Qiu et al. 2017; Leighton et al. 2019; Perera et al. 2019). Using microarray analyses, the cerebral cortex of adult rats was found to express ~40,000 piRNAs (Dharap et al. 2011). In addition, a combination of small RNA sequencing and CAGE (Cap Analysis of Gene Expression) revealed that small RNAs with piRNA features (24-31 nt and 5’ U bias) were present in the adult mouse brain (Ghosheh et al. 2016). However, a reanalysis of these results determined that most piRNA-like sequences expressed in mammalian brain are likely to be non-coding RNA fragments rather than genuine piRNAs (Tosar et al. 2018). Nevertheless, a subsequent study using improved approaches for piRNA selectivity, small RNA next-generation sequencing, and novel bioinformatics techniques found that piRNAs were indeed expressed in brain, albeit at much lower levels than in the testes, and are shorter in length (20-24 nt) than their germ line counterparts (24-32 nt). MIWI (PIWIL1), MILI (PIWIL2), and MIWI2 (PIWIL4) transcripts were also detected in the cerebral cortex and hippocampus providing strong evidence for the activity of functional piRNA pathways in the mammalian nervous system (Perera et al. 2019).

As in other model organisms, the function of piRNAs appear to also regulate brain development in mammals. In mice, PIWI proteins associated with piRNAs were localized in neuronal dendrites of the hippocampus, and suppression of specific piRNAs using antisense inhibition resulted in a reduction of the dendrite spine area (Lee et al. 2011a). MIWI (PIWIL1) was also found to play a role in the development of the cerebral cortex in rodents by affecting the radial migration of cortical neurons by partly modulating the expression of microtubule-associated proteins (Zhao et al. 2015). Similarly, endo-siRNA pathways in C. elegans have been found to regulate migration of the HSN neurons (Kennedy and Grishok 2014). Mammalian piRNAs also seem to respond to nerve injury as described in C. elegans. A study found that a subset of mammalian piRNA-like small ncRNAs (piLRNAs) in rat sciatic nerve axoplasm exhibited significant changes in expression during nerve regeneration. In addition, depletion of MIWI (PIWIL1) by RNAi in cultured rat sensory neurons resulted in increased axonal regrowth following injury (Phay et al. 2018). Recently, it was shown that mice that experienced sciatic nerve transection exhibited differential expression of over 7,500 piRNAs. Furthermore, a reduction of MIWI (PIWIL1) was observed in the Schwann cells of injured sciatic nerve. Cell line transfection with an up-regulated piRNA reduced the expression of the myelin basic protein and enhanced cell migration of Schwann cells compared to the control (Sohn et al. 2019). Together, these results indicate that the piRNA pathway plays a role in the regeneration of injured sciatic nerves by regulating the migration of Schwann cells.

Evidence suggests that piRNA pathway function may also affect mammalian behavior as well. Homozygous (−/−) and, to a lesser degree, heterozygous (+/−) Mili (Piwil1) mutants exhibited increased locomotion and exploratory behavior and reduced anxiety-like behavior compared to wild-type (+/+) mice (Nandi et al. 2016). However, whether MILI was acting in the nervous system to regulate anxiety levels was not determined. In a subsequent study, the question of whether abrogating MIWI (PIWIL1) and MILI (PIWIL2), the two PIWI members involved in the primary biogenesis of piRNAs, would affect cognitive functions was explored using more selective knockdown experiments in mice. Viral knockdown of only Miwi in the dorsal hippocampus had no effect on the acquisition of conditioned fear, locomotion or on their anxiety level, while Mili knockdown mice exhibited hyperactivity. However, simultaneous knockdown of both Miwi and Mili resulted in augmented contextual fear memory, suggesting some functional redundancy between MIWI and MILI functions in the adult mouse brain (Leighton et al. 2019).

3. Evidence of endo-siRNA functions in the nervous system

3.1. Endo-siRNA neuronal expression and function in C. elegans

Since the discovery of RNAi (Fire et al. 1998), numerous studies of endo-siRNAs in C. elegans have been published, particularly regarding their function in the germ line (reviewed in Billi et al. 2014; Almeida et al. 2019). However, studies in the past several years have unveiled surprising roles for endo-siRNAs in the regulation of nematode behavior (Table 3). One of the first studies linking endo-siRNAs to the nervous system was examining the genetic determinants of long-term olfactory adaptation to the odor butanone. C. elegans can sense many volatile odorants and respond accordingly through attraction or avoidance behaviors. The sensation and adaptation to certain odorants, such as the ketone butanone, is mediated by the AWC olfactory sensory neuron and requires the ODR-1 guanylyl cyclase and the EGL-4 cGMP-dependent protein kinase (L’Etoile et al. 2002). By performing a limited reverse genetic screen with mutants that are defective for regulation of gene expression pathways, Juang et al. discovered that a mutation in the somatic Ago gene nrde-3 rendered animals defective for olfactory adaptation. NRDE-3 has been shown to target nascent RNAs in the nucleus and recruit heterochromatin formation complexes to silence gene loci (Guang et al. 2008; Burkhart et al. 2011; Burton et al. 2011). They also found an increase in 22G-siRNAs antisense to the odr-1 coding sequence correlated with olfactory adaptation, suggesting the model that NRDE-3-mediated decrease in odr-1 expression in AWC neurons mediates the olfactory adaptation behavior to butanone (Juang et al. 2013). Similarly, the osm-9 TRPV channel gene that is expressed in numerous sensory neurons and is required for olfactory behaviors is also regulated by endogenous RNAi (Colbert et al. 1997; Sims et al. 2016). C. elegans early larvae that experience environmental conditions unfavorable for growth will enter the dauer diapause stage and resume reproductive development once environmental conditions improve (Cassada and Russell 1975). Adults that transiently passed through the dauer stage no longer express osm-9 specifically in their ADL neurons, resulting in defective olfactory behaviors that are mediated by ADL. For example, C. elegans hermaphrodites that have continuously developed typically avoid high concentrations of the pheromone component ascr#3 (Jang et al. 2012), while adults that have passed through the dauer stage (postdauers) fail to respond (Sims et al. 2016). Postdauer animals with mutations in the Agos ERGO-1 and NRDE-3 fail to downregulate osm-9 in ADL neurons and exhibit ascr#3 avoidance behaviors comparable to continuously developed adults. In addition, mutations in genes encoding components of a protein complex called the Mutator focus, which localizes near nuclear pores and contributes to the amplification of siRNAs, also fail to downregulate osm-9 in postdauer ADL neurons (Sims et al. 2016). Interestingly, members of the Mutator complex, including MUT-16, are also required for dauer formation as mutations in these genes result in a dauer deficient phenotype in multiple adverse environmental conditions. Expression of a mut-16 rescue transgene either pan-neuronally or in a subset of sensory neurons abolishes the dauer formation defects. The authors demonstrated that MUT-16 and the nuclear Argonaute CSR-1 were required for the expression of G proteins that promote sensory signaling; thus, animals defective for endogenous RNAi in their neurons were unable to detect unfavorable environmental conditions (Bharadwaj and Hall 2017). Together, these studies indicate endogenous RNAi pathways as critical arbitrators to how an organism responds to its environment throughout development.

Table 3.

Neuronal functions for endo-siRNAs in animal models.

Model organism	Neuronal roles	References
C. elegans	Olfactory behavior and adaptation	Juang et al. 2013; Sims et al. 2016
	Dauer formation	Bharadwaj and Hall 2017
	Transgenerational inheritance of chemotaxis behavior	Posner et al. 2019
Drosophila	Axon growth	Pepper et al. 2009
	Long term memory	Li et al. 2013
	Transposon activity and ALS model	Krug et al. 2017; Sun et al. 2018
Murine	Cognitive function	Smalheiser et al. 2001

Interestingly, a more recent study has found that changes in attractive olfactory behavior towards benzaldehyde can be inherited transgenerationally via neuronal endo-siRNAs being transmitted through the germ line. Mutations in rde-4, which encodes a double stranded RNA-binding protein, result in a RNAi defective phenotype and are defective in their response to the odorant benzaldehyde (Tabara et al. 2002; Tonkin and Bass, 2003). Posner and colleagues identified neuronal small RNAs from animals whose rde-4 expression was restricted to the nervous system that affected the expression of over 1,000 germline mRNAs. Nearly 15% of these gene expression changes were observed for at least three generations and showed a strong dependency on HRDE-1, an Ago required for transgenerational inheritance (Buckley et al. 2012). The authors demonstrated that the neuronal endo-siRNAs facilitate silencing of a conserved germline gene, saeg-2, encoding a member of the SAEG-1/SAEG-2 histone deacetylase complex (Hao et al. 2011; Posner et al. 2019). SAEG-2 acts downstream of RDE-4 to regulate gene expression, as mutations in saeg-2 partially rescue the chemotaxis defects of rde-4 mutants. Based on their observations, the authors posit an elegant model whereby RDE-4 and neuronal small RNAs mediate chemotaxis for at least three generations via the germline-specific Argonaute HRDE-1 by silencing saeg-2 in the germ line (Posner et al. 2019). Furthermore, this study demonstrates how the nervous system can regulate the reproductive system for multiple generations, providing further evidence against the “Weismann Barrier” dogma stating that genetic information from somatic tissues is not heritable (Sabour and Scholer 2012).

3.2. Endo-siRNA neuronal expression and function in Drosophila

Some of the most compelling evidence for a role of endo-siRNAs in the nervous system of Drosophila is their regulation of transposable elements in the brain. As flies age, the expression of LINE-like elements, gypsy and R1, and the LTR element R2 is upregulated in the brain and the number of gypsy de novo integration events in the mushroom bodies is significantly increased. Ago2 negatively regulates the increased expression and activation of these retroelements during aging. In Ago2 null mutants, young flies (2-4 days old) have the retrotransposon expression levels and activity found in wild-type aged flies (28 days-old). These changes in expression in Ago2 mutants correlate with deficiencies in long-term memory (LTM) compared to wild-type flies that worsen as the flies age. Importantly, these LTM deficiencies were rescued by the introduction of a neuronally expressed Ago2 transgene (Li et al. 2013). This study indicates that the regulation of transposable elements by siRNA pathways is a contributing factor in the neuronal decline characteristic of normal aging, and that disruption of neuronal RNAi leads to LTM defects or the progression of neuronal disease (Li et al. 2013).

3.3. Endo-siRNA neuronal expression and function in murine model systems

As described in the introduction, reports of endo-siRNAs in mammalian model systems remain limited and their function controversial. One study examined whether endo-siRNAs are expressed in the brain and whether their levels are modified during the learning process (Smalheiser et al. 2001). Adult mice were trained to discriminate between two simultaneously presented odors for a reward (Larson and Sieprawska 2002), and hippocampal samples were taken from trained mice, pseudo-trained mice (mice exposed to the same two odors as trained mice but were not rewarded), and naive mice for deep sequencing. Using a siRNA-specific selection procedure and modified bioinformatics pipeline, over 65,500 candidate endo-siRNAs with length distributions between 21-22 nt were identified that mapped with perfect complementarity and uniquely to over 14,500 genes. A subset of these putative endo-siRNAs were derived from predicted hairpin inverted repeats in eight genetic loci (Abca2, Arhgef17, Camk2a, Gap43, Rab40b, Slc17a7, Syn1, and SynGAP1), most of which correspond to genes that are known regulators of synaptic plasticity, signaling, or are synaptic components. Notably, a subset of the siRNAs associated with these loci were significantly increased in trained mice compared to controls. In particular, the gene that harbored the most endo-siRNAs, SynGAP1, is a Ras GTPase-activating protein that complexes with postsynaptic density 95 (PSD-95) and an NMDA receptor to potentiate synaptic plasticity and learning (Komiyama et al. 2002). Immunoprecipitation experiments showed that the SynGAP1 endo-siRNAs associated with an antibody against mouse Ago proteins (Smalheiser et al. 2011). A later study also found an abundant small RNA species in the hippocampus between 20-24 nt that mapped to retrotransposons, genes, and pseudogenes (Nandi et al. 2016). Thus, evidence suggests that endo-siRNAs exist in the brain and may correlate with cognitive functions in mammals, although a causal function of RNAi with learning and memory has yet to be determined.

4. Association of piRNAs and siRNAs with human neuropathologies and their animal models

4.1. Autism spectrum disorder and intellectual disability

The piRNA studies described above deliver a strong argument for the role of the piRNA pathway in the nervous system and raise the question of whether piRNAs and PIWI proteins may be performing evolutionarily conserved roles in humans. In addition, we have described the evidence of endo-siRNA functions in the nervous system, although their role in mammalian brains is less clear than in other animal models. If small RNA pathway components are crucial for neuronal function and behavior in animals, what are the phenotypic consequences of their disruption? Some studies have found links between small RNAs and syndromes associated with intellectual disability and autism.

4.1.1. Murine Rett syndrome model

Rett syndrome is a severe neurological disorder that results in intellectual disability and loss of voluntary movement resulting from mutations in the methyl CpG binding protein 2 gene, MECP2 (Amir et al. 1999). Mice with mutations in Mecp2 exhibit genome-wide aberrant transcription of repeat regions, increased transposition of retrotransposons, and increased abundance of piRNAs in the cerebellum (Yu et al. 2001; Skene et al. 2010, Saxena et al. 2012). A model for Rett syndrome was thus proposed whereby the increased level of retrotransposons in the absence of MECP2 results in the global increase of piRNAs, the consequence of which is anomalous gene expression that can lead to profound neurodevelopmental consequences (Saxena et al. 2012). Indeed, a report examining a set of whole genome sequencing data from over 2,500 families with one child on the autism spectrum disorder revealed a strong association between mutations in piwi genes and autism (Iossifov et al. 2014), suggesting that the piRNA pathway has wide-ranging effects on cognitive functions.

4.1.2. Drosophila Fragile X syndrome model

Endo-siRNAs have also been implicated in intellectual disability and autism spectrum disorder. In Drosophila, Ago2 regulates the expression of dfmr1 (Drosophila fragile X mental retardation gene), the fly model of Fragile X syndrome (FXS) (Pepper et al. 2009). FXS is the most commonly inherited form of intellectual disability and autism and is caused by the absence of a functional FMR1 (fragile X mental retardation 1) protein, which regulates the expression of numerous mRNAs in post-synaptic neurons (Hagerman et al. 2017). Loss of Ago2 results in defects in larval dFMR1-dependent neuromuscular junction synaptic structures and an increased level of dFMR1 in adult brains. While the physical association between Ago2 and dFMR1 has been reported (Caudy et al. 2002; Ishikuza et al. 2002), dfmr1 mRNA was not found to be a direct target of Ago2 nor was the cleavage activity by Ago2 required for dfmr1 regulation, suggesting that Ago2 represses dfmr1 through an unknown mechanism that is independent of its canonical endonucleolytic activity (Pepper et al. 2009).

4.2. Alzheimer’s disease (AD)

The piRNA pathway has also been associated with Alzheimer’s disease (AD). One study detected over 500 piRNAs in the human brain, with more than 450 that were considered AD-associated, and 149 that were differentially expressed in AD-afflicted brain tissue. piRNA target prediction in combination with pathway enrichment analysis uncovered four piRNAs (piR-38240, piR-34393, piR-40666, and piR-51810) that are likely to target four genes (CYCS, KPNA6, RAB11A, and LIN7C) involved in the most significant AD-associated neuronal pathways (Roy et al. 2017). Another study detected ~9,400 piRNAs in the prefrontal cortex tissue, with fourteen showing significant changes in expression in AD cases versus controls (Qiu et al. 2017). Taken together, these studies support the association between a dysregulated piRNA pathway and AD and suggest that certain piRNAs may serve as risk biomarkers for early disease detection.

4.2.1. Drosophila taupathy model

The Drosophila model system has been integral to the investigation of piRNAs as a potential mechanism contributing to some human neural pathologies. Tauopathies, such as Alzheimer’s disease, are a class of neurodegenerative disorders characterized by the aggregation of the microtubule-associated tau protein as neurofibrillary tangles in the brain (Lee et al. 2001). Drosophila that pan-neuronally expressed a transgene carrying an autosomal dominant human tau mutation linked to the familial neurodegenerative disorder, frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17), exhibited some of the trademark features associated with tau-induced neurodegeneration in humans, including the adult onset of progressive neurodegeneration, the accumulation of abnormal tau, and premature death (Wittmann et al. 2001). An overabundance of tau was also associated with extensive loss of heterochromatin in Drosophila, mice, and human genomes, resulting in anomalous expression of genes that are normally silenced. One of the genes with significantly increased expression in the tau transgenic flies was ago3. Repression of Ago3 suppressed tau neurotoxicity in the Drosophila head, suggesting that piRNAs and transposable elements may be deregulated in tauopathies. These findings indicate that tauopathies are detrimental consequences of a conserved mechanism whereby atypical gene expression, including that of ago3/piwil1, follow the relaxation of heterochromatic regions (Frost et al. 2014). In a follow up study, Sun and colleagues demonstrated that heterochromatin relaxation and the reduction of PIWI and piRNAs activate transposable elements as key drivers of tau-mediated neurodegeneration. The authors also identify pathogenic tau as inducers of neuronal transposable element mobilization leading to the aberrant activation of cell cycle and premature neuronal cell death. Moreover, analysis of transposable element levels in postmortem human brains with taupathies also showed significant misregulated expression. Taken together, these studies show that tau-induced heterochromatin decondensation and the consequential deregulation of transposable elements are principal drivers of neuropathies (Sun et al. 2018).

4.2.2. Murine taupathy model

Consistent with the observations in the Drosophila transgenic tau model, heterochromatin loss and increased level of PIWIL1 in motor neurons were also found in mice expressing human tau with a mutated FTDP-17. Additionally, examination of Alzheimer’s disease brains also revealed heterochromatin loss and altered gene expression, including PIWIL1 overexpression in hippocampal neurons. This suggests that heterochromatin loss and dysregulated gene expression are conserved features of taupathies from Drosophila to humans (Frost et al. 2014).

4.3. Amyotrophic lateral sclerosis (ALS)

ALS is a progressive neurodegenerative disease that affects motor neurons in the brainstem, spinal cord, and motor cortex resulting in muscle atrophy, involuntary muscle contraction and stiffness, and dementia. Inspection of postmortem ALS patients revealed neuronal and axonal degeneration in the brainstem and spinal cord, evidence of neuroinflammatory response, and neuronal inclusions due to protein aggregation (Robberecht and Philips 2013). Mutations in a number of genes contribute to the pathogenesis of ALS, including FUS (fused in sarcoma), which is an RNA binding protein with several roles in RNA metabolism such as modulating RNA polymerase II transcription, pre-mRNA splicing and nuclear export, and biogenesis of non-coding RNA (Kwiatkowski et al. 2009; Vance et al. 2009).

4.3.1. Drosophila ALS model

Neuron-specific knockdown of the FUS orthologue in Drosophila, Caz (Cabeza), partially recapitulates certain traits of human ALS pathology such as movement deficiency and motor neuron structural anomalies (Sasayama et al. 2012). Given previously reported roles of piRNAs and Piwi proteins in mammalian neuronal function (Perrat et al. 2013; Nandi et al. 2016), a recent study screened for genetic interactions between caz and components of the piRNA biogenesis pathway using the “rough eye” visible marker. The authors reported a genetic interaction between caz and aub, ago3, wde, and rhino, indicating that Caz function is connected to the piRNA pathway. Overexpression of Aub resulted the mislocalization of Caz from the nucleus to the cytoplasm and exacerbated the defects in locomotion and the synapse of neuromuscular junctions that are features of neuron-specific Caz knockdown. Caz knockdown also reduced the level of mature piRNAs originating from the large somatic piRNA cluster, Flamenco (flam). Interestingly, reduction of Caz increased flam pre-piRNA levels, while Caz overexpression decreased their levels, indicating that Caz suppresses piRNA cluster transcription and modulates additional downstream maturation steps such as piRNA splicing or export. Thus, Aub governs the cellular localization of Caz, which in turn, regulates the levels of flam transcription and mature piRNAs. This study highlights the interconnected functions of Aub, Caz and pre-piRNAs as contributing factors in the neuropathogenesis of ALS (Wakisaka et al. 2019).

Another study investigated the consequences of unchecked retrotransposon activity on neurodegeneration using the Drosophila TAR DNA-binding protein 43 (TDP-43) model (Krug et al. 2017), which is an RNA and DNA binding protein that is mainly expressed in the nucleus of neuronal cells where it regulates RNA processing and is associated with progression of ALS and frontotemporal lobar degeneration (FTLD) (Vanden Broeck et al. 2014). Pan-neuronal or pan-glial transgenic expression of human TDP-43 (hTDP-43) in Drosophila resulted in locomotory impairment and elevated levels of retrotransposable element (RTE) expression, at the genome-wide level. The authors found that the brain of flies expressing pan-glial hTDP-43, but not pan-neuronal hTDP-43, resulted in widespread neuronal cell death and shortened lifespan due to gypsy activation. Using a genetically encoded sensor for Dicer-2 activity co-expressed with hTDP-43, it was found that expression of hTDP-43 in glial cells or mushroom bodies significantly decreased siRNA silencing in those respective tissues (Krug et al. 2017).

5. Concluding remarks

Since small non-coding RNAs were discovered in the early 1990’s, great headways have been made in unravelling how these miniature RNAs govern nearly aspect of animal development. The sheer diversity of non-coding RNAs and their collected functions hark back to the “RNA world” hypothesis first proposed over half a century ago whereby ancient RNAs were capable of transmitting genetic information and of propagating themselves (Woese 1965; Crick 1968; Orgel 1968; Gilbert 1986). It is clear today that small RNAs, amidst their multitude of functions, fulfill that prophecy. The expansive regulatory requirements of the nervous system require neuromodulatory agents that can be generated and can act in an efficacious manner, and small ncRNAs are well suited for those roles. Their diminutive size and potent inhibitory abilities enable sncRNAs to orchestrate the complexities of neuronal circuitries. As such, RNA has long been proposed as the material that confers the transfer of genetic memory from one animal to another and even from one body part to a regenerated body part, as is the case in planarians (McConnell 1966; Golub et al. 1970; Malin et al. 1971). As such, the ramifications of misregulated small RNA pathways on human neurological health are undeniable. Our knowledge of small ncRNAs and their protein partners has rapidly accelerated in the past decades; however, many areas remain largely unexplored (e.g. the contribution of endo-siRNAs to the mammalian nervous system). Nevertheless, we look forward to even more discoveries that can further enlighten the “RNA world” and that could provide valuable insights into formulating therapies or preventative measures to combat neurological disorders in us humans.

Highlights.

• PIWI interacting RNAs play significant roles in animal neuronal development

• Endogenous small interfering RNAs regulate neural processes in animals

• Dysregulation of small non-coding RNA pathways can result in neurodegenerative disorders in mammals

Abbreviations

5HT

seratonin

AD

Alzheimer’s disease

Ago

Argonaute

Ago3

Argonaute 3

ALS

Amyotrophic lateral sclerosis

Aub

Aubergine

CAGE

Cap Analysis of Gene Expression

Caz

Cabeza

circRNA

circular RNA

CNS

central nervous system

dmfr1

Drosophila fragile X mental retardation gene

dsRNA

double stranded RNA

endo-siRNA

endogenous small interfering RNA

flam

Flamenco

FTLD

frontotemporal lobar degeneration

FUS

fused in sarcoma

FXS

Fragile X syndrome

G

Guanosine

HP1a

heterochromatin protein 1a

l(3)mbt

lethal (3) malignant brain tumor

lncRNA

long non-coding RNA

LTF

long-term facilitation

LTM

long-term memory

Mael

Maelstrom

MB

mushroom body

MECP2

methyl CpG binding protein 2 gene

miRNA

microRNA

nt

nucleotides

piLRNA

piRNA-like small noncoding RNA

piRNA

PIWI-interacting RNA

PIWI

p-element induced wimpy testis

PSD-95

postsynaptic density 95

PTGS

post-transcriptional gene silencing

QTL

quantitative trait locus

RdRP

RNA-dependent RNA polymerase

RISC

RNA induced silencing complex

RNAi

RNA interference

RTE

retrotransposable element

sncRNA

small non-coding RNA

snoRNA

small nucleolar RNA

TDP-43

TAR DNA-binding protein 43

TGF-β

transforming growth factor β

TGS

transcriptional gene silencing

U

uridine