Mechanisms and consequences of subcellular RNA localization across diverse cell types

Abstract

Essentially all cells contain a variety of spatially restricted regions that are important for carrying out specialized functions. Often, these regions contain specialized transcriptomes that facilitate these functions by providing transcripts for localized translation. These transcripts play a functional role in maintaining cell physiology by enabling quick response to changes in cellular environment. Here, we review how RNA molecules are trafficked within cells, with a focus on the subcellular locations to which they are trafficked, mechanisms that regulate their transport, and clinical disorders associated with misregulation of the process.

Introduction

Gene expression is regulated through a complex coordination of events that occur at distinct steps during the lifetime of a mRNA. Whereas control of RNA synthesis by RNA polymerases, transcription factors, and chromatin states is widely studied, less well-characterized post-transcriptional mechanisms are also essential in defining gene expression landscapes in individual cell types. These cell-type specific gene expression profiles define the protein composition of cells and ultimately cell-type specific functions. One poorly understood aspect of post-transcriptional regulation is RNA localization.

Much of the work on RNA localization has been performed in neuronal systems because early studies recognized the importance of RNA localization in these polarized cell types, and they are amenable to study due to their size and morphology. Neurons are highly polarized, with long distances between the soma, where RNAs are synthesized in the nucleus, and projections (axons/dendrites/neurites), where cells must quickly respond to stimuli (Figure 1A). These cell types have therefore been a critical model system to understand how RNA localization controls gene expression using imaging-based approaches as reviewed previously^1–5. A number of neurological diseases, including amyotrophic lateral sclerosis (ALS) and fragile X syndrome (FXS), have been associated with misregulated RNA localization in neurons. Recent advances in high-throughput sequencing and microscopy techniques have also revealed that RNAs are localized to subcellular compartments (e.g. endoplasmic reticulum and mitochondria) in many nonpolarized cell types^6–9. This observation has led to some key questions. Do neurons and other cell types use a common set of mechanisms to achieve RNA localization? In other words, does an ‘RNA localization code’ exist that transcends cell types? What are the biological roles of creating localized transcriptomes that are different across subcellular compartments and cell types? What are the consequences of misregulated RNA localization? Despite the realization that local transcriptomes contribute to cell function across diverse cell types, many gaps in our understanding of how these processes are regulated still exist.

Figure 1: Mechanisms of RNA localization.

A) Local RNA transcriptomes are established in neurons to facilitate local protein production in response to stimuli. RNA binding proteins are essential for transport of these RNAs. B) RNAs are transported along cytoskeletal networks in various cell types. Short range transport is typically mediated by actin filaments and myosins, which can facilitate retrograde and anterograde movement. Long range transport is achieved with microtubules in both retrograde and anterograde movement using dynein and kinesin, respectively. C) Concentration gradients of RNAs are established using anchors composed of proteins that sequester RNA.

How does RNA get there?

Several RNA localization mechanisms have been observed in various cell types. Active transport of RNA as components of ribonucleoprotein (RNP) granules along cytoskeleton scaffolds is the most widely observed mechanism¹⁰. Long-range localization of RNA is often accomplished via active transport in neurons (Figure 1B), but cytoskeletal-associated RNA localization has also been observed in spherical cells such as yeast and Drosophila embryos where sometimes shorter distances are traversed^10–13. Transcripts can also achieve localization with a combination of active transport followed by anchoring once they have arrived at a specific location. This mechanism has been characterized in multiple cell types including Aplysia sensory neurons¹⁴ and yeast^11,15. Additionally, in Drosophila oocytes passive diffusion followed by anchoring of nanos mRNA helps establish a concentration gradient across the cell^16,17 (Figure 1C). Lastly, local transcriptomes can be established by selective RNA degradation or local protection. This occurs when an RNA is unstable or susceptible to decay unless it is stabilized or protected within a specific location¹⁸. Even though this mechanism seems energetically wasteful, there is evidence that this process is utilized by Drosophila embryos¹⁹. These mechanisms of RNA localization are not mutually exclusive, and a single RNA transcript may be influenced by a combination of mechanisms to achieve finely tuned regulation of its localization²⁰. Additionally, all of these localization mechanisms require additional protein factors to interact with the RNA. These RNA binding proteins (RBPs) are critical for understanding how RNAs are localized. Below we will discuss in detail how RNAs are localized through different mechanisms in different cell types.

Active transport of RNAs within cells

There is abundant evidence for active RNA transport along cytoskeleton scaffolds. This mechanism was characterized by inhibition of RNA transport following treatment with cytoskeletal depolymerizing drugs including nocodazole and cytochalasin^21,22. Long range and short range cytoskeleton-mediated transport are thought to use different machinery. Long range transport is mostly mediated by dyneins and kinesins on microtubules whereas shorter transport is mediated by myosins on actin filaments, as reviewed elsewhere^1,10 (Figure 1B).

The cytoskeleton is composed of two major structures, microtubules and microfilaments. RNA transport mechanisms have been described using both of these structures. Microtubules are polymers of the protein tubulin. In neurons, transport of many mRNAs [e.g. beta-actin (ActB)²³, Activity-regulated cytoskeleton-associated protein (Arc)²⁴, and calcium/calmodulin-dependent protein kinase II (CaMKIIɑ)²⁵] to axons and dendrites often occurs along microtubules²⁶. Microtubules are also implicated in RNA transport within Drosophila embryos (bicoid (bcd) mRNA²⁶ and oskar (osk) mRNA¹⁸) as well as myelin basic protein (MBP) mRNA in oligodendrocytes²⁷. Cardiomyocytes also require microtubules for transport of transcripts to sarcomeres²⁸.

Transport along actin microfilaments by myosin motors is responsible for the bulk of RNA transport in yeast¹⁰, where ASH1 mRNA is localized into dividing daughter cells and accumulates at the bud tip to regulate mating type switching¹¹. ASH1 mRNA is exclusively localized along actin filaments yet there is also evidence for local sequestration activity after reaching the bud tips¹². In this case, transport along actin is responsible for both getting transcripts to the correct region and refining their localization at the bud tip.

Both microtubules and actin filaments are used to actively localize RNAs in many cell types. RNA movements are facilitated by larger RNP complexes which contain both protein and RNA. Interactions between these components and adapter and motor proteins associated with the cytoskeleton facilitate active transport of sometimes multiple RNAs at once²⁹. For example, CaMKIIα, Neurogranin, and Arc mRNAs are targeted to dendrites within the same RNP granule³⁰. The exact composition of cytoskeleton-associated RNPs can also regulate the speed and direction of transport^31,32. However, combinatorial effects of each factor make it difficult to discern exactly how an RNA achieves its specific destination. Further research into what factors are included in RNP complexes and how they impact trafficking is necessary to understand how RNAs are actively transported along the cytoskeleton, and RNP composition and dynamics are extensively reviewed here³³.

Anchoring of RNAs creates local RNA enrichment

RNAs that arrive near their destination via either active or passive mechanisms can be anchored into position through interactions with proteins and the cytoskeleton. In addition to facilitating active transport, actin also plays a role in refining and stabilizing transcripts at a destination. Both active transport as well as anchoring to a specific region sometimes requires the cytoskeleton to achieve stable localization^17,21.

Active transport followed by specific anchoring of RNA is a common mechanism in Drosophila and Xenopus; however, both active and passive mechanisms prior to anchoring are used. For vg1 mRNA, microtubules are responsible for transport to the vegetal hemisphere, whereas microfilaments are important for anchoring at the cortex²². However, gurken mRNA requires dynein for both transport to and retention at the anterodorsal corner, indicating that anchoring can occur via microtubules^18,34. Nos mRNA does not associate with microtubules in Drosophila and is thought to localize solely through diffusion and cytoplasmic streaming¹⁶. It is trapped by an unknown protein complex containing Rumpelstiltskin at the posterior end of the embryo^17,35. Even in a large Drosophila embryo, RNA anchoring alone is sufficient for maintenance of RNA localization.

Similar mechanisms are used in budding yeast to generate local concentrations of Ash1 mRNA to control mate type switching. Active transport of ASH1 mRNA into the daughter cells is dependent on interaction with myosin motor protein She1. Without She1, ASH1 mRNA does not localize correctly^11,15. However, if She1 is available but Bni1¹¹ or Bud6³⁶ is mutated, specific localization to the bud tip is unstable. This strongly indicates that RNA is generally localized by active transport but specifically retained by anchoring factors like Bni1 and Bud6.

RNA anchoring is observed in other eukaryotes as well. Beta-actin mRNA is anchored to protrusions in chicken embryo fibroblasts as well as the leading edge of rat adenocarcinoma cells by EF1ɑ³⁷. This protein is capable of simultaneously binding mRNA and F-actin. Loss of its actin binding domain results in diffuse beta-actin mRNA that is no longer localized to the leading edge. Other anchoring factors include the adenomatous polyposis coli complex (APC) which is also important for RNA localization in fibroblast protrusions. It associates with RNP granules, and without APC expression, RNAs no longer localize to protrusions³⁸. Therefore, anchoring of RNAs to the cytoskeleton is critical for defining the precise localization of transcripts in several cell types.

Local RNA transcriptomes are regulated by selective degradation

Control of local transcriptomes via selective degradation or protection in subcellular regions is less studied and has been observed in only a few cell types (Figure 1C). For example, Drosophila embryos localize Heat Shock Protein 83 (Hsp83) mRNAs to the posterior pole by protecting transcripts from degradation in that region^39,40. These transcripts are widely expressed throughout the embryo until they are selectively degraded following fertilization of the embryo. This is largely regulated by sequence in the 3’UTR of Hsp83 that can be deleted or exchanged for other UTRs to disrupt its spatial stability phenotype³⁹. This sequence in the 3’ UTR is bound by Smaug, resulting in degradation of the transcript¹⁹.

In addition to regulation of RNA abundance through canonical RNA degradation pathways (e.g. via the TRAMP complex), non-canonical roles of nonsense-mediated decay (NMD) have also been characterized. NMD was initially described to degrade mRNAs containing premature termination codons (PTCs); however, this mechanism also regulates the abundance of non-PTC containing RNAs in dendrites⁴¹ and axons⁴², which in turn influences synaptic plasticity and growth cone formation, respectively. While the degradation strategies that control RNA localization come at an energetic cost, it is effective at creating specific localized transcriptomes in a regulated way.

An RNA has arrived. Now what?

Local RNA transcriptomes may serve a variety of roles in establishing cell type specific functions by creating temporally and spatially regulated proteomes to guide differentiation and responses to environmental stimuli. The predominant theory on how RNA localization accomplishes this is through enhancing local protein production efficiency by having the mRNA translated where the protein ultimately exerts its effect. Transporting a single mRNA is energetically efficient since a single mRNA molecule can be used to synthesize multiple copies of proteins⁴³, reviewed more extensively here^2,18,44. Similarly, colocalization of mRNAs encoding members of a protein complex could produce high concentrations of multi-subunit complex components in a specific location, thus increasing complex formation efficiency and decreasing the formation of off-target, nonfunctional complexes^45–48. mRNAs can also be localized in a translationaly silenced form. Local signals are therefore needed before initiating protein synthesis⁴⁹. Finally, mRNAs can be localized to prevent protein toxicity in specific subcellular compartments, as is the case for MBP in oligodendrocytes²⁷. However, there could be many additional reasons for RNA localization that are not yet clear. For example, the RNAs themselves might have a function as is the case for many non-coding RNAs (e.g. miRNAs and lncRNAs) and at least one mRNA⁵⁰.

Translation of localized RNAs

In many cell types, mRNA co-localizes with its encoded protein. This has been observed in neurons⁵¹, Drosophila embryos⁵², and HEK293 cells⁷. This phenomenon is suggestive of local translation which has been shown across species and cell types. Due to the difficult nature of quantifying local transcriptomes and proteomes, most studies focus on a single mRNA and its encoded protein. Below we discuss multiple specific examples that support the importance of mRNA localization in producing higher local protein concentrations (Figure 1A). In these examples, local mRNA translation can be important to the function of the synthesized protein and overall cell function.

In yeast, ASH1 mRNA is locally translated in the daughter bud to control mate type switching. Normally, ASH1 mRNA is not translated until it arrives in the bud tip⁵³. Depletion of Puf6p, a translational repressor of ASH1 mRNA, causes ASH1 mRNA translation before it arrives at the bud tip, resulting in both cells having the same mating type⁵³. Thus, ASH1 mRNA must be properly localized, and only translated in this location to function correctly.

Local translation has been demonstrated to occur at neuronal dendritic spines, which are substantial distances from the cell nucleus. It has long been suggested that ribosomes locally translate mRNAs, but we now know that ribosomes are localized to dendritic spines along with many translation factors⁵⁴ and engage in active translation in many neuronal cell types (reviewed here³). One of the best examples is beta-actin mRNA, which is localized to dendritic spines in neurons⁵⁵. Reducing beta-actin mRNA translation with morpholinos provided evidence that local translation of this transcript in retinal neurons is required for axonal branching and stabilization⁵⁶.

The majority of protrusion-enriched transcripts in fibroblasts are locally translated to produce localized proteins⁵⁷. This is especially true for actin-associated proteins which help structure the protrusions. Furthermore, mRNAs encoding ribosome subunits have been shown to localize to protrusions of migrating cells in a LARP6 dependent manner. Local translation of ribosomal proteins at these protrusions enhances ribosome biogenesis and local protein production⁵⁸. In each of these cases, the localized mRNA helps determine localization of the resulting proteins which aid critical cell functions.

Assembly of multi-subunit protein complexes

Formation of multi-subunit complexes can depend on chaperones, subunit stoichiometry, and local concentrations of each subunit. Local translation of the protein subunits can increase their local concentrations, thereby facilitating complex assembly by reducing the physical distance between components⁵⁹. Additionally, spatially isolating complex members could help reduce unwanted protein interactions that might result in non-canonical complexes or other unfavorable interactions.

Local transcriptome mapping using proximity labeling has revealed that RNAs encoding complex members co-localize to the ER and mitochondria in yeast⁸. These results support previous observations that mRNAs encoding several mitochondrial associated complex members are found near the mitochondria (reviewed here⁶⁰). Presumably, the mRNAs that encode ER and mitochondrial protein complexes localize to the organelle to facilitate local translation and assembly.

This phenomenon also occurs in less defined subcellular locations. mRNAs encoding the seven subunits of the Actin stabilizing complex Arp2/3 localize to fibroblast protrusions where they are required for stable growth⁴⁶. Localization of these seven separate transcripts to protrusions aid in Arp2/3 assembly through local translation and assembly. In some cases, local translation of protein complex members can result in co-translational assembly as observed for heteromeric ion channels⁴⁷ and histone acetyltransferase⁴⁸.

RNA localization in response to environmental stimuli

mRNAs can also be localized in response to a specific signal, such as following glutamate stimulation in neurons, as reviewed here⁴⁹. It is suggested that mRNAs are localized to produce local proteins that help remodel or stabilize activated synapses.

Intestinal absorptive enterocytes also localize RNAs across their apicobasal axis in response to refeeding to initiate recovery after starvation⁶¹. This reorganization changes the translational status of many mRNAs and may contribute to the cellular response to the refeeding stimulus, although exactly how this happens is unknown.

Control of cell size and projection length might also use active mRNA transport followed by local protein translation⁶². This model suggests that localization of mRNA to projections, local translation to produce protein, followed by retrograde protein diffusion back to the nucleus allows the cell to sense projection length. This mechanism has been suggested to regulate cilia, flagella^63,64, and neurite length^65,66. However, it could be broadly applicable to other subcellular components such as organelle size.

Role of cis and trans factors in RNA localization

The RNA localization mechanisms discussed above rely on multiple components, specifically RNA and their protein binding partners. First, cis-elements within the RNAs themselves, often referred to as zipcodes, ‘mark’ the RNA for a destination. Information can be stored in the RNA in the form of linear sequences or in RNA structure. These RNA features facilitate interactions between the RNA and specific trans-acting RBPs, which typically associate with RNAs through an RNA binding domain (RBD)^31,67. Further interactions between RBPs and other proteins, such as those associated with the cytoskeleton⁶⁸, create a network of interactions that regulate subcellular localization patterns of RNA. RNP granules can vary in their composition. This produces a heterogenous pool of RNPs inside of cells, further complicating study of their protein and RNA components. For example, Arc and CaMKIIα mRNAs associate selectively with Barentsz (Btz)- and Staufen 2 (Stau2)- containing RNPs, respectively whereas RNAs such as Lysophospholipase 1 (Lypla1) and Cap-Binding Protein (CBP80) are found in both of them⁶⁹. A handful of RBPs and their cognate zipcodes are known (Table 1). However, these complex multivalent interactions have been difficult to dissect. For example, a single RNA can interact with multiple RBPs at once, and likewise a single RBP can have the capacity to recognize different zipcodes and therefore multiple RNAs. Below, we will focus on the role and identification of these cis- and trans-acting factors.

Table 1: Known RNA/RBP pairs that drive RNA localization.

Many additional RBPs and mRNAs are known to facilitate localization and be localized; however, we did not include those in this table since either the RBP component or the mRNA zipcode is unknown. abbreviations used: cytoplasmic polyadenlyation element (CPE), spliced oskar localization element (SOLE), exon junction complex (EJC), hnRNP A2 response element (A2RE), untranslated region (UTR), open reading frame (ORF), pumilio homology domain (PHD), RNA recognition motif (RRM), K homology (KH), double stranded RNA binding domain (dsRBD), arginine-glycine-glycine (RGG).

RBP; RBDs (if known)	mRNA (species)	zipcode	transport mechanism	cellular function	references
STAU; 4 conserved dsRBDs	oskar (Drosophila) bicoid (Drosophila) vg1 (Xenopus) CamKlla (mammalian)	3′ UTR (SOLE; EJC deposition) 3′ UTR and 5′ stem loops 3′ UTR (366nt) 3′ UTR (30nt)	active transport (kinesin) with biased random walk active transport (dynein) and anchoring active transport (kinesin) and anchoring active transport	germ line differentiation embryonic patterning embryonic development memory formation; dendrites	183–186 88–90, 187 22, 188, 189 190–192
ZBP1, VERA, IMP1; 2 RRM and 4 KH	ActB (mammalian) vg1 and vegT (Xenopus) Tau (rat)	3′ UTR, 5′-GGACU-3′ (4-8) and 5′-ACA-3′ (22-24) 3′ UTR (element X4) U-rich sequence	active transport, diffusion, and anchoring active transport (kinesin) and anchoring active transport (kinesin)	fibroblast movement and axon guidance embryonic endoderm differentiation axonal polarity maintenance	23, 193–196 77, 197–200 201–203
hnRNPA/B, Squid, Hrp48; RRM, KH, and RGG	MBP (mammalian) gurken (Drosophila) oskar (Drosophila)	3′ UTR (21nt A2RE) 5′ UTR 3′ UTR (SOLE; EJC deposition)	active transport active transport (dynein) and anchoring active transport (kinesin)	myelin sheath formation embryonic development germ line differentiation	204, 205 34, 206, 207 208, 209
CPEB	MAP2 (rat) ZO-1 (mouse) cyclinB1 and Xbub3 (Xenopus)	3′ UTR CPE 3′ UTR (5 conserved CPEs) 3′ UTR CPE	active transport unknown unknown	microtubule assembly epithelial tight junction assembly and polarity mitotic spindle and division	84 85 210
Rump	nanos (Drosophila)	3′ UTR (4 partly redundant regions)	diffusion, streaming, and anchoring	embryonic polarity	16, 35
ARC1	Arc1 (mammalian)	3′ UTR	active transport and exovesicles	synaptic plasticity	179, 215–215
She2/She3	ASH1 (yeast)	ORFs and 3′ UTR	active transport (myosin) and anchoring	mate type switching	11, 12, 211, 212
Sec27	OXA1 (yeast)	3′ UTR and ORF	unknown	mitochondrial inner membrane biogenesis	216, 217
Puf3p; PHD	COX17 (yeast)	3′ UTR (UGUR motif)	unknown	mitochondrial biogenesis and motility	218–220
RBP-L/RBP-P; 3 RRM	glutelin (rice) prolamine (rice)	ORF (repeated motifs) and 3′ UTR (U-rich) ORF (repeated motifs) and 3′ UTR (U-rich)	active transport active transport	grain development grain development	221, 222 221, 222
FMRP; KH and RGG	MAP1b and CamKlla (mammalian)	G-quadruplex	active transport	neurogenesis and memory formation	140, 141, 223, 224
SMAUG	hsp83 (Drosophila)	3′ UTR	degradation and local protection	development; maternal transcript elimination	19, 40
LARP6	ribosomal protein mRNAs (mammalian)	5′ TOP	unknown	fibroblast protrusion formation	58
TDP-43; RRM and Gly-rich C-term	NEFL and RAC1 (mammalian)	3′ UTR	active transport	neuronal development and plasticity	109, 225
NOVA	girk2 (mouse)	intronic and 3′ UTR YCAY	unknown	spinal motor neuron dendrite activity	226

Cis-elements or zipcodes

Zipcode elements serve as the signal to target an RNA for delivery to a specific subcellular location. They can vary from a few bases to a kilobase in their length. Although hundreds to thousands of RNAs are known to be trafficked to a variety of subcellular locations, the sequence elements within those transcripts that regulate transport have been identified for only a few dozen. Most of the known localization regulatory elements are located in 3′ untranslated regions (UTRs) of mRNA, although some localization elements have been identified in 5′ UTRs^14,70, coding regions^{11,12,71–73}, and intronic sequences⁷⁴. Additionally, localization elements are often repetitive and redundant. This repetition might allow many low affinity sequences to act in union to achieve specificity. Most of the zipcodes identified are thought to function independently and be modular in nature. Mutation and truncation experiments using reporter transcripts have led to the discovery of a small number of necessary and sufficient zipcode sequences. Combined with additional biochemical methods, these approaches have uncovered zipcodes that depend on primary sequence and/or RNA structures for their function and have been reviewed here^75,76 (Table 1).

An RNA zipcode alone is not always enough to facilitate localization. Some elements are necessary but not sufficient, suggesting that neighboring RNA sequences can be an important context for the zipcode. For example, xcat2 mRNA in Xenopus oocytes localizes to the vegetal pole using six repeats of a short motif, UGCAC, present within the 3′ UTR of xcat2. However, insertion of this motif into the 3′ UTR of another mRNA, vg1, did not result in its localization to the vegetal pole⁷⁷. Additionally, nuclear processing also plays an important role in mRNA localization. For example, oskar localization at the posterior pole of the Drosophila oocyte requires both spliced oskar localization element (SOLE, which comprises nucleotides of both exon 1 and exon 2) and the exon junction complex (EJC) deposited during splicing for efficient transport along the microtubules by kinesin^71,78.

Many bacterial RNAs are also differentially distributed at locations such as in the cytoplasm in a helical like pattern, at membranes, and at polar or septal sites^79,80. In E. coli, the bglG–bglF operon, which codes for proteins necessary for aryl-β-glucoside metabolism, localize to the cell membrane. The bglG–bglF zipcode is located within the sequence encoding the first two transmembrane helices of bglF and is uracil-rich, a common feature observed in transmembrane proteins across species^81,82.

Another example of an element that hints at conservation across species and cell types is the cytoplasmic polyadenylation element (CPE). CPE is a nucleotide sequence (UUUUUAU) within the 3′ UTR of most mRNAs which binds to the CPE-binding protein (CPEB). Although this interaction has been most studied for its effects on translation upon synaptic activation⁸³ studies have shown that CPEB can also act as a transport protein for dendritic mRNAs such as MAP2 and intestinal apical RNA such as ZO-1 responsible for maintaining tight junction assembly and cell polarity^84,85.

Lastly, many zipcode elements depend on RNA secondary and/or tertiary structures, e.g. stem loops, for recognition by trans-factors^31,86. For example, bicoid mRNA 3’UTR stem loops^87,88 are recognized by the RBP Staufen (STAU)^89,90. The complex roles RNA structure plays in defining zipcodes contributes to the difficulty of finding them since structural elements are much more difficult to define and identify computationally than linear sequence motifs.

Cis-elements in non-coding RNAs

Non-coding RNA such as miRNAs and lncRNAs also use sequence elements to direct localization patterns (Table 2). For example, miR-29b contains a hexanucleotide 3′motif (AGUGUU) responsible for nuclear localization⁹¹. This is in contrast to the cytoplasmic enrichment observed commonly for miRNAs. Additionally, the lncRNA MALAT1 uses sequences present at both the 5′ and 3′ ends to target its localization to paraspeckles within the nucleus⁹². Recent studies using massively parallel reporter assays (MPRAs) have also discovered C-rich motifs present in several lncRNAs that lead to their nuclear enrichment^93,94. lncRNA BC1 and it’s human analog BC200 are enriched in dendrites and synapses, where it represses translation locally in nerve cells by inhibiting eIF4a-mediated unwinding of RNA duplexes^95–97. The lack of evidence for the role of lncRNAs is primarily due to the fact that techniques developed to study complex RNA-RNA interactions are still in their infancy. Further, there has been some evidence that RNA modifications might also play a role in localization. For example, mitochondrial localization of tRNA^Glu in the RNA virus Leishmania is driven by a specific modification of the tRNA⁹⁸.

Table 2: Localized noncoding RNA and their RBP partners.

abbreviations used: K homology (KH), arginine-glycine-glycine (RGG).

RBP; RBDs (if known)	ncRNA (species)	zipcode	transport mechanism	cellular function	references
FMRP; KH and RGG	BC1 (mouse) BC200 (human)	unknown unknown	unknown unknown	dendritic mRNA translation regulation dendritic mRNA translation regulation	95, 227 95, 227
ANT2	mir29-b (human)	unknown	unknown	cell division	91, 228
RNPS1, SRm160, and IBP160	NEAT2/MALAT1 (human)	region E and region M	unknown	nuclear speckle function	92

Identification of potential cis-ating elements using computational methods (e.g. MEME⁹⁹, HOMER¹⁰⁰, SeAMotE¹⁰¹, and others dependent on structure^102–105 have had limited success. Despite our increasing understanding of principles governing cis-localization elements using specific examples, we still lack a systematic charaterization of general rules. One of the challenges to overcome is the implementation of strategies that combine sequence motifs and structural features while keeping in mind the redundancy observed in many known zipcode examples. In the future, focusing on transcriptome-wide studies coupled with high-throughput functional assays along with advanced in silico analyses will further efforts to uncover novel zipcodes.

Trans-factors that regulate localization

RBPs regulate all aspects of RNA metabolism, including transcription, splicing, translation, and decay. RNA localization is also subject to regulation by RBPs, which bind and recognize RNA elements to form RNP complexes. RBPs generally contain a variety of RBDs that impart specificity. Some of the common RBDs are the RNA recognition motif (RRM), K homology (KH) domain, DEAD-box motif, and zinc finger domains¹⁰⁶. These domains allow for sequence specific binding to short single stranded RNA (ssRNA) motifs. For example, RBPs including heterogeneous nuclear ribonucleoproteins (hnRNPs) recognize specific elements such as a 21-nt sequence known as hnRNP A2 response element (A2RE) present in many dendritic mRNAs¹⁰⁷. Additionally, some RBPs recognize RNA secondary structure, sequence elements present within structural elements, or a combination of the two. Many RBPs also contain intrinsically disordered regions (IDRs) that could regulate the assembly and dynamics of RNP complexes¹⁰⁸. Most of the RBPs involved in localization have been identified through genetic screens and/or affinity purification of proteins that bind the localized transcript. Using these approaches, conserved roles in localization have been identified for a few RBPs across various species⁶⁷ (Table 1).

The complex combinatorial regulation of RNA localization by multiple RBPs was demonstrated in a recent study¹⁰⁹. The use of genetic manipulations and high-resolution imaging showed that TDP-43 binds to Rac1 mRNA. Additional interactions occur between the mRNA and either FMRP or Staufen. These interactions further regulate Rac1 mRNA transport to mouse dendrites by facilitating anterograde or retrograde movement, respectively. This illustrates that a number of different RBPs in cooperation with each other might dictate the appropriate location of the target mRNA.¹⁰⁹

Protein capture using oligo-dT to purify polyA mRNA followed by mass spectrometry has found that over 1500 proteins directly bind polyA containing RNA¹¹⁰. Further experiments then identified the sequence binding preferences of many of these RBPs. Two techniques, RNAcompete¹¹¹ and RNA Bind-n-Seq (RBNS)¹¹², were developed to provide an inventory of RBP motif preferences and their specificity. To complement these in vitro approaches and identify RNA sequences bound by RBPs in vivo, a number of techniques involving crosslinking combined with RBP immunoprecipitation such as RIP-seq¹¹³, CLIP-seq¹¹⁴, HITS-CLIP¹¹⁵, eCLIP¹¹⁶, and PAR-CLIP¹¹⁷ were developed. These methods, in conjunction with further variations (reviewed in¹¹⁸), have enabled high resolution profiling of transcriptome wide RBP-RNA binding interactions. It still remains to be seen if these techniques and databases can be routinely successfully applied to the study of RNA localization on a transcriptomic scale to derive regulatory principles that govern the process. However, given their utility in the study of other RNA-based processes, the possibility is intriguing.

Methods that allow the detection of RNA-RNA interactions may also provide insight into RNA localization mechanisms. These include techniques which combine the use of cross-linking agents, proximity ligation, and RNA-seq to profile RNA interactomes (CLASH, SPLASH, PARIS, LIGR-seq and MARIO)¹¹⁹. Combining these techniques with RBP-based approaches may have the potential to yield rich datasets that detail the RBP-RNA and RNA-RNA interactions that regulate RNA localization.

Clinical Relevance

RNA localization is critical for proper development from yeast to humans^11,120–122. These initial observations, coupled with recent technological advances, have linked RNA localization to clinical disease (Figure 2). To date, RNA localization defects contribute to neuromuscular/skeletal diseases, neuronal disorders such as Alzheimer’s, depression, embryonal disorders, and cancer initiation¹²³. But which RNAs are mislocalized, what binding partners regulate their localization, and what causes their mislocalization remain unknown. Furthermore, the lack of functional studies have stalled our understanding of the direct significance of RNA mislocalization during the establishment of disease phenotypes.

Figure 2: Misregulation of RNA localization is relevant to disease phenotypes.

A) RNA localization patterns to various subcellular compartments are disrupted in various diseases. B) Establishment of EMT is associated with RNA localization changes. C) Gradients of RNA concentrations are observed between the apical and basal axis of epithelial cells; however, it is not clear if disruption of these patterns contributes to disease phenotypes. D) Asymmetric distribution of RNAs helps establish developmental states and disruption of these patterns contributes to establishment of cancer.

Neuronal and Neuromuscular Diseases

A handful of neurological disorders have been attributed to defects in RNA localization, and have previously been reviewed extensively here^123–126. Below we highlight a handful of these key findings for amyotrophic lateral sclerosis (ALS), spinal muscular atrophy (SMA), fragile X syndrome (FXS), and myotonic dystrophy (DM). RNA localization has also been implicated in other neurological disorders such as depression and Alzheimer’s, though many more studies are needed to find a definite link^127,128. Taken together, these associations suggest that small perturbations in RNA localization can manifest in a variety of neuronal diseases, though few have been thoroughly explored.

ALS is a neurodegenerative disease characterized by the loss of motor neurons¹²⁹. Mutations in a number of genes found in ALS patients have been implicated to cause defects in RNA metabolism, including alternative splicing, stabilization, and localization. These include SOD1, TARDBP (TDP-43), FUS, and C9orf72, all of them covered extensively in other reviews¹³⁰. Mutations in the RBP TDP-43 are strongly linked to ALS¹³¹. TDP-43 has also been shown to regulate RNA localization¹³², but how the potential mislocalization of RNA in ALS neurons relates to disease phenotype is still unknown.

SMA is another neurodegenerative disease associated with a loss of motor neurons¹³³. SMA results from insufficient levels of the protein SMN. SMN, like TDP-43, facilitates RNA localization to neurites¹³³. Reduced levels of SMN results in mislocalization of various transcripts, such as ActB, growth-associated protein 43 (GAP43), and neuritin/cpg15^123,134. Overexpression of RNP granule assembly and transport regulating RBPs, HuD and IMP1/ZBP1, can restore localization of transcripts and partially rescue the phenotype. These findings also implicate SMN as a regulator of RNA transport^135,136. However, as with ALS, it is unknown whether and how RNA mislocalization in SMA neurons contributes to phenotypes.

FXS is a neurological disorder that results from the loss of FMRP expression¹³⁷. Loss of FMRP results in dysregulated RNA localization in neurons, resulting in slower RNA granule localization as well as mislocalized mRNAs, such as Kif26a in dendrites^138,139. The role of FMRP in regulating neuronal RNA localization is well established and covered in other reviews^139–141. Again, whether mislocalization of RNAs within FXS neurons is related to disease phenotypes is unknown.

Misregulated sequestration of RNA can have dramatic impacts on cellular function. In DM, mutant Dmpk mRNA contains expanded CUG repeats that trap transcripts in the nucleus. The retention of these transcripts in the nucleus is sufficient to generate a DM phenotype in mice¹⁴². This example illustrates the importance of exploring mechanisms behind transcript localization and mislocalization to better understand its impact on human disease.

Stem Cells and Development

Development relies on strict temporal and spatial control of gene expression^{17,39,143,144}. Control of RNA localization is one method cells use to regulate cell fate and function. For example, spermatid maturation and non-genomic inheritance of traits is influenced by localization of specific RNA transcripts (mainly small RNAs) as the sperm develops to various compartments, such as the head, mid-piece, or tail. It is becoming increasingly clear that RNAs are transferred to sperm as they mature. Irregularities in the amount of RNA in spermatid have been associated with sperm maturation, affecting human male infertility, and others have implicated these RNAs in seeding non-genetic inheritance in zygotes^145,146. While future studies are needed to investigate the implication of mislocalized transcripts, it is clear local RNAs play a key role during sperm maturation and inheritance^147–149.

Similarly, defined RNA localization patterns also govern temporal control of translation in oocytes and cell polarity in Drosophila embryos¹²⁰. These observations suggest that RNA localization might be a conserved phenomenon necessary for cell differentiation during development. Indeed, localization of RNA is important in the developing mammalian embryo as well. Discovered more than 100 years ago, the balbiano body is an organelle consisting of endoplasmic reticulum, golgi, proteins, and RNA¹⁵⁰. It is known to be asymmetrically positioned across species and is involved in fate determination through RNA sequestration, though its importance in mammals remains unknown.

Localization of individual RNA transcripts affects mammalian cell differentiation. For example, NEAT2/MALAT1 displays differential localization patterns during development. Regulation of its localization is important for RNA shuttling, splicing, and differential gene expression^92,151. NEAT2/MALAT1 is localized to the nucleus in the oocyte, but becomes cytoplasmic in the developing embryo, only to return to its nuclear localization in differentiated cell states¹⁵². CDX2 mRNA localization is also polarized in blastomeres to establish cell polarity in early mammalian embryonic development¹⁴⁴. The consequence of mislocalizing these RNAs has yet to be reported.

RNA localization patterns are also particularly important during asymmetric cell division. Certain cell types, e.g. B and T cells, hematopoietic cells, and ganglions in the nervous system^153–155, utilize asymmetric cell division for fate determination. For example, localization of CYCLIN D2 mRNA in neural progenitor cells of the developing neocortex allows them to maintain their pluripotency whereas the daughter cell differentiates^156,157. These observations suggest that RNA localization might be a conserved phenomenon that regulates cell differentiation during development.

Developmental disorders attributed to disruption of RNA localization patterns are continuing to be discovered. Loss of the RBP Gle1 disrupts centrosomal RNA localization and centrosome function in human cells, leading to lethal motor neuron disease and fetal hydrops^158–160. Dyskeratosis Congenita is a stem cell disorder caused by mislocalization of TERT transcripts from Cajal bodies to nucleoli, resulting in telomerase shortening in patient stem cells¹⁶¹ and predisposing affected individuals to aplastic anemia.

Mature cell function also relies on proper RNA localization. As discussed above, intestinal epithelial cell RNA localization helps regulate local protein production along the apicobasal axis⁶¹ (Figure 2C). Consequences of disrupting RNA localization in these cell types may underlie disease states such as inflammatory bowel disease, maldigestion, or local immune responses^162,163.

Cancer

Asymmetric RNA localization to daughter cells may contribute to cancer initiation and progression^164,165 (Figure 2D). Recently, hundreds of RBPs involved in post-transcriptional RNA modifications, alternative splicing, and localization have been implicated in numerous cancers^166,167. These studies have led to speculations that RNA mislocalization may play a role in cancer progression. Whether these observations are a cause, consequence, or correlation is unclear. In Melanoma, cytoplasmic and mitochondrial localization of long noncoding RNA SAMMSON has been shown to increase clonogenic potential, and knock-down of SAMMSON decreases melanoma viability¹⁶⁸. Additionally, the noncoding RNA GINIR associates with centrosomal proteins, with over-expression resulting in disruption of Brca1 protein activity and more aggressive tumors¹⁶⁹.

A link between RNA localization and epithelial mesenchymal transition (EMT) is also emerging (Figure 2B), supported by the observation that altered localization of certain transcripts promotes cell adhesion and migration¹⁷⁰. The long noncoding RNA LINC00460 promotes EMT in head and neck squamous cell carcinoma by localizing to the nucleus¹⁷¹. SAM68 transcripts are normally nuclear, but become cytoplasmic during cell adhesion processes, a key factor in promoting cancer metastasis^172,173. Local protein translation at the tips of filopodia also increases Rho and Rho Kinase (ROCK) activation in the pseudopodia of tumor cells^86,170. The local transcriptome found there includes mRNAs encoding for proteins such as M-Ras, whose overexpression is known to be sufficient for EMT transition¹⁷⁴.

The observations described above implicate misregulation of RNA localization during disease establishment, but very few cases have established a direct link. This is due to the technical challenges associated with functional studies of RNA and RBP localization, particularly in disease contexts. These challenges are reflected by the paucity of data we have describing direct links between individual RNAs, their RBP regulators, mechanisms that drive localization, and ultimately the dependence of cellular functions on establishment of local transcriptomes. Fundamental insights have been gained from manipulatable model systems, but the relevance of mechanisms observed in these systems to disease phenotypes is unclear. Further studies utilizing advanced cell culture models such as iPSCs and organoids, advancing RNA visualization tools, and developing more sophisticated approaches to analyze large data sets may help to address these challenges.

Discussion

Many transcriptome-scale sequencing^{51,57,175–177} and imaging¹⁷⁸ experiments have illuminated how pervasive the phenomenon of RNA localization is. However, we are still left without a mechanistic understanding of how the vast majority of these RNAs end up at their destination and why their transport is regulated. Specific examples of mislocalized RNAs leading to specific phenotypes are rare, even though there are many connections between RNA localization and cellular function or human diseases. For thousands of localized RNAs, further work is needed to understand the cis-elements and trans-factors that regulate their transport as well as the effect that their localization has on cell function and physiology. These studies are required before we can begin to formulate a more thorough understanding of general mechanisms that govern RNA localization, some of which might transcend cell types and species.

In addition to the RNA localization mechanisms described above, it is likely that additional RNA localization patterns exist, possibly even some that are trans-cellular. Recent studies have shown transport of Arc mRNA across synapses between motor neurons and muscle cells^179,180. This process is mediated through interactions between the retroviral-like Arc1 protein and retrotransposon-like sequences within the 3′ UTR of its own mRNA. Disruption of trans-synaptic Arc1 mRNA transfer leads to defective synaptic plasticity. Furthermore, genetic mutations in the human Arc protein are linked to autism¹⁸¹ and schizophrenia¹⁸². These findings are intriguing, and suggest that we have much yet to learn about modes of RNA localization, cell types that use them, and consequences of its disruption.

Synopsis.

Here we review what is currently known about mechanisms of RNA localization in a variety of cell types. Many reviews have previously focused on mechanisms, consequences, and diseases associated with defective RNA localization in the central nervous system. However, here we have compiled what is known about mechanisms and consequences of RNA localization across other cell types and organisms, including yeast, oocytes, epithelial cells, and in diseases such as cancer and those affecting development.