Genomic analysis of RNA localization

Abstract

The localization of mRNAs to specific subcellular sites is widespread, allowing cells to spatially restrict and regulate protein production, and playing important roles in development and cellular physiology. This process has been studied in mechanistic detail for several RNAs. However, the generality or specificity of RNA localization systems and mechanisms that impact the many thousands of localized mRNAs has been difficult to assess. In this review, we discuss the current state of the field in determining which RNAs localize, which RNA sequences mediate localization, the protein factors involved, and the biological implications of localization. For each question, we examine prominent systems and techniques that are used to study individual messages, highlight recent genome-wide studies of RNA localization, and discuss the potential for adapting other high-throughput approaches to the study of localization.

Keywords:

• subcellular localization
• genomics
• fractionation
• transcriptome
• RNA transport
• RNA-seq
• translation

Introduction

Eukaryotic gene expression is regulated in both time and space. Studies focused on temporal regulation have benefited from the use of a variety of high-throughput genomic technologies for analysis of changes in transcripts, translation status and chromatin during cell state transitions and development.^1-3 Subcellular localization of a transcript can also have profound effects on a transcript's activity, but with several recent exceptions has usually been studied one transcript at a time. In this review we discuss central questions in the field of RNA localization and highlight recent studies that are starting to use genome-wide strategies to address them.

Importance of RNA localization

RNA localization has been linked to many important phenotypes. Two interesting and well-characterized cases involve the localization of oskar RNA in the Drosophila embryo and β-actin RNA in vertebrate fibroblasts.

In its early stages, the developing Drosophila embryo is a syncytium. Proper patterning of the embryo relies on the presence of distinct gradients of proteins within the embryo. Germ plasm develops at the posterior end of the embryo and its formation is dependent on a high concentration of Oskar protein produced from localized oskar RNA.⁴ Mis-localization of oskar RNA to the anterior pole results in the formation of germ plasm at the anterior end.⁵ Interestingly, proper localization of oskar RNA is dependent on the splicing of its first intron, suggesting a link between splicing and localization,⁶ discussed further below.

In vertebrate fibroblasts, the leading edges of the cell are dense with actin. This is due, in part, to the localization of mRNA coding for β-actin to the leading edges.⁷ This localization is necessary for proper cell motility, and inhibition of recognition of the sequences that regulate localization result in the loss of cell polarity and motility.^8-10

In a general sense, RNA localization provides the foundation for a rapid, robust response to a stimulus (Fig. 1). While protein transport to the subcellular site of interest could, and often does, provide a similar output, localization of the mRNA molecule obviates the need for a drawn-out signal transduction process. Additionally, since several protein molecules can be made from a single mRNA molecule, the response can also be more vigorous. Although the majority of work in the field has focused on the localization of mRNAs, newer work has also demonstrated that long noncoding RNAs (lncRNAs) are often subject to specific subcellular localization as well.^11,12

Figure 1. General scheme of mRNA localization. The formation of a splicing-dependent stem-loop and involvement of the EJC is seen in the localization of the transcript oskar in Drosophila embryos. After splicing, mRNAs are exported from the nucleus. Those containing mRNA localization signals, in this case, the stem-loop, are recognized by trans-factors that interact with the localization signal. The trans-factor is integrated into a transport complex that uses motor proteins to move the RNA to its destination. Localized RNAs are often kept in a translationally-repressed state until they reach their destination, whereupon they are released and may be translated to give a localized pool of protein product.

RNA localization systems

RNA localization is often studied in systems where there is clear functional specialization and the level of localization is likely to be robust. Projections of mammalian neurons, both dendrites and axons, are commonly used.^13,14 The bud tip is a well-studied example in S. cerevisiae as ASH1 mRNA localizes there during mating type switching.^15,16 The high concentration of ASH1 mRNA in the bud tip results in high expression of Ash1p, a protein that represses transcription of the mating type regulator HO. Syncytial fly embryos are often used, as RNA localization to specific places in the embryo is important for proper body patterning and development, and the distance from one pole of the embyro to the other is quite large.^4,5 RNAs may also localize to organelles within the cell. Mitochondria are frequently studied as targets for RNA localization,^17-19 as are the cytoskeleton^20,21 and leading edges of fibroblasts.⁷

This review is organized around four central questions in RNA localization: (1) which RNAs are localized? (2) What specific mRNA sequences mediate localization? (3) What protein factors contribute to RNA localization? (4) How does RNA localization impact translation or other biological outcomes? In each section, we review state of the art molecular genetic and cell biological methods and discuss emerging genomic approaches for answering each question.

Identifying Localized RNAs

A primary question in the field is to determine which RNAs specifically localize within cells. The current approaches to this question can provide robust and accurate answers but generally require that each RNA be interrogated separately.

Localized RNAs: Single-gene approaches

The most common technique for interrogating RNA localization is RNA fluorescent in situ hybridization (RNA FISH) (Fig. 2A).⁷ By hybridizing many labeled probes to a single RNA, this method has been improved to allow single molecule sensitivity.^22,23 Although this technique has been successfully used in a variety of settings, including whole organisms,²² it requires that samples be fixed before probing and is thus not compatible with living cells or organisms.

Figure 2. Classical and genomic approaches for studying RNA localization. (A) RNA FISH images from three systems. RNA encoding β-actin is localized to the leading edges of chicken fibroblasts,¹²⁴ while ASH1 mRNA localizes to the bud tip of S. cerevisiae during mating type switching.¹²⁵ RNA encoding odd and the resulting protein product accumulate in distinct segments in the developing Drosophila embryo and are important for proper development and morphology.³⁵ (B) Fractionation of cells followed by high-throughput sequencing can give a more generalized view of RNA localization.²¹ Here, mouse C2C12 myoblasts were treated first with digitonin, which gently lyses cells but keeps membranes generally intact. Cytoplasmic RNA is collected and the cells are then treated with Triton X-100 to solubilize membranes. Membrane-associated RNA is collected and the resulting sample is treated with high salt to elute RNA associated with insoluble cell components. RNA from each fraction is then analyzed by RNA-seq. The location of activity of the protein encoded by each RNA generally match with the location of the RNA. (C) Method overview of FISSEQ.⁴⁶ Cellular structures are crosslinked followed by in situ reverse transcription of the crosslinked RNA using random primers and modified nucleotides that allow further crosslinking. The resulting cDNA is crosslinked to cellular structures and the circularized and amplified using rolling circle amplification. cDNA amplicons are then crosslinked to cellular structures and each other and sequenced in situ using SOLiD sequencing.

Imaging RNAs in live cells is typically done through the use of an RNA binding protein fused to a fluorescent protein. A commonly used system is fusion of the viral coat protein MS2, which binds to a specific RNA hairpin structure. When this hairpin is appended to an RNA of interest, usually in the 3′ UTR, and coexpressed with an MS2-GFP fusion, GFP fluorescence can be used to track RNA localization in vivo.²⁴ This technique has since been improved and expanded to allow visualization in live mice.²⁵ While this method allows the monitoring of dynamics of RNA localization in vivo, it is still limited to the study of one or a few RNAs at a time.

Other methods include the use of molecular beacons. These probes consist of a fluorescent molecule that is held next to a quencher by basepairing.²⁶ When the loop of sequence outside of the stem hybridizes to an RNA of interest, it forces the fluorescent probe and quencher apart, allowing visualization of the RNA.²⁷ By combining biotin-conjugated fluorescent probes with tetravalent streptavidin, Alonas and colleagues were able to visualize single RNA molecules in developing viruses.²⁸

As an alternative to tracking the localization of a single RNA, others have taken the approach of fractionating cells and cloning the resulting RNA into cDNA libraries. An early study cloned RNAs present in processes of rat hippocampal cells in culture, identifying several RNAs involved in neuronal signaling^29-31 and even budded synpatoneurosomes.³² Giant neurites from the sea slug Aplysia have also been analyzed using this approach.³³ Sequencing of several hundred clones from this library revealed an enrichment for RNAs encoding elements of the translational machinery as well as cytoskeletal components. A similar fractionation of rat dorsal root ganglia (DRG) using a compartmental culture system yielded RNA from axons³⁴ and was analyzed by quantitative RT-PCR to test for the presence of several target transcripts in the axon. Although such approaches have less precision in terms of location than imaging-based methods, their increased gene sampling allows a richer picture of localized RNAs.

Localized RNAs: Genomic approaches

Large scale FISH

Some groups have taken existing techniques and found ways to automate them, approaching genome-wide scale. A classic example was the “Fly-FISH” study by Lécuyer and colleagues.³⁵ In this study, the established RNA FISH technique was scaled up using hybridizations in 96-well plates to examine the localization patterns of over 3000 genes in the developing Drosophila embryo. Surprisingly, over 70% of the analyzed genes showed distinct, nonuniform localization patterns, demonstrating unexpected and prevalence and diversity of RNA localization.

Microarray-based approaches

In a similar extension, several groups used microarrays to study differential enrichment of RNA species between two cellular compartments. Marc and colleagues isolated polysomes from both cytoplasm and mitochondria-enriched fractions.¹⁷ Interestingly, transcripts encoding mitochondrial proteins of bacterial origin were more often associated with the mitochondria than were those of eukaryotic origin. Many of these mitochondrial-associated transcripts encoded proteins with transmembrane domains, suggesting a connection between RNA localization to the surface of the mitochondria and efficient import and structuring of membrane-embedded proteins.³⁶ This microarray-based approach was also applied to other systems. RNA from protruding pseudopodia of mouse fibroblasts was enriched for approximately 50 transcripts,³⁷ some of which were associated with fragile X mental retardation protein (FMRP).

Several studies have fractionated neurons and used microarrays to analyze RNAs present in the distal processes vs. the cell bodies. Poon and colleagues used pore-containing culture membranes to separate distal projections from cell bodies in cultured rodent hippocampal neurons.³⁸ Using microarrays, they identified over 100 projection-localized RNAs, many of which were associated with translation. In mammalian cortical neurons, over 300 transcripts were identified as enriched in projections using microarrays, with many of these transcripts coding for proteins with translational and mitochondrial functions.³⁹ A similar transcriptomic analysis of rat DRG axons revealed the presence of mRNAs from almost 3000 genes in the axons.⁴⁰ The RNA content of neuronal projections can also change in response to injury, stimulation or during development.³⁹

The RNA content of even smaller cellular structures can also be interrogated, although the reduced RNA yield from these structures often necessitates that the RNA be amplified prior to measurement. For example, RNA collection and amplification from laser dissected growth cones at the tips of neurites revealed that they, too, contain many different RNA species, again with an enrichment for RNAs involved in translation.⁴¹

RNA-seq-based approaches

Following developments in high-throughput sequencing, later experiments used RNA-seq to measure expression. Minis and colleagues used similar membrane-based fractionation techniques to separate DRG axons and cell bodies,⁴² and found transcripts encoding mitochondrial proteins as well as specific transcription factors enriched in the axon fraction.

Cajigas and colleagues further probed the transcriptome of neural projections using Roche/454 sequencing of sections of rat hippocampal neuropil,⁴³ which are generally depleted of cell bodies and enriched in dendrites and axons. Over 2500 transcripts were identified as present in dendrites and/or axons and were enriched for functions including neurotransmitter secretion and synaptic vesicles. This study also demonstrates a key issue in observing localization with tissue dissections: the cleanliness and specificity of the “localized” fraction. Although the neuropil is enriched for projections, it also contains other components. Using a titration strategy in which RNA from other components was doped in, Cajigas and coworkers were able to identify transcripts specifically enriched in the projections.

RNA-seq has also been applied to study localization in other systems. In X. laevis extracts, taxol-stabilized microtubules were isolated by sedimentation. Microtubule-associated RNAs were shown to be enriched in transcripts involved in cell division and spindle formation.⁴⁴ Fractionation of mouse C2C12 myoblast cells and Drosophila S2 cells into cytosolic, membrane-associated, and insoluble fractions revealed transcripts enriched in each compartment (Fig. 2B). The localization patterns of transcripts were generally associated with the functions of the proteins they encoded. For example, nucleosomal and other nuclear protein mRNAs were enriched in the cytosolic compartment, while mRNAs encoding membrane and secreted proteins tended to be enriched in insoluble and membrane compartments.²¹ In another study, developing Drosophila embryos were cyrosectioned into 60 μm slices and analyzed by RNA-seq, recapitulating RNA localization patterns previously defined by FISH.⁴⁵ This study puts forward an interesting approach that could pave the way for improved resolution in future studies.

Newer approaches may obviate the need for fractionation altogether. By fixing RNA and then reverse transcribing, amplifying, and sequencing it in situ in the cell, Lee and colleagues have combined the resolution of imaging techniques with the depth and throughput of sequencing-based approaches⁴⁶ (Fig. 2C). This method, named FISSEQ, performs all enzymatic steps necessary for sequencing at the site of the RNA transcript. Combining imaging and sequencing in this way may provide powerful new tools for the future study of RNA localization.

A more complete understanding of RNA sequences involved in localization may have implications in human disease as well. Some factors known to be involved in RNA localization like FMRP and muscleblind already have known disease associations, but the extent to which RNA localization is mis-regulated in these diseases is unknown. The identification of RNAs that localize to particular cell compartments may contribute to explanations of disease phenotypes and provide insights into potential modes of treatment.

Identifying RNA Elements that Direct Localization

RNA localization is mediated by cis-elements, which link the RNA to protein complexes that may 1) associate with cytoskeletal elements for subsequent intracellular transport, 2) preferentially stabilize or degrade the RNA in particular subcellular compartments, or 3) tether or anchor the RNA to a particular locations. These cis-elements may be composed of linear sequence motifs or structures.

Localization elements: Molecular genetic approaches

To date, most cis-elements that direct RNA localization have been identified using a combination of image-based and biochemical assays. Commonly, reporters encoding the putative localization element are introduced into a model system of interest, and deletions/mutations are generated to alter localizing activity. Sequential experiments are often required to pinpoint the localizing element with high precision. For example, initial studies describing -actin mRNA at leading edges of fibroblasts used the entire 3′ UTR to demonstrate localizing activity.⁴⁷ Subsequent studies narrowed the region to 54 nucleotides,⁸ culminating in a definition 5-mer and 3-mer with specific spacing requirements, such that each short motif causes the RNA to loop around the localization factor zipcode binding protein 1 (Zbp1).⁴⁸

While the majority of known localizing elements reside in 3′ UTRs, some occur in 5′ UTRs,⁴⁹ or within mature, spliced coding sequence. In some instances, particular elements required for localization extend beyond a simple cis-element, and require elements acquired during mRNA processing. For example, localization of oskar mRNA to the posterior pole of the Drosophila embryo requires both a 3′ UTR cis-element, as well as an exon junction complex deposited adjacent to a 28-nucleotide motif generated by splicing.^6,50

Finally, a novel approach that linked cell survival to RNA localization was used to identify a 50-nucleotide motif sufficient to localize reporter mRNAs to mitochondria.⁵¹ Localization of the ATP2 mRNA to yeast mitochondria is required for proper respiratory function, and provided an assay for a recombination-based screen in which random fragments derived from the ATP2 3′ UTR were used to rapidly screen for localizing activity.

Localization elements: Genomic approaches

Increasingly, genomic approaches are being taken to identify cis-elements involved in RNA localization. In general, an important challenge is to separate cellular compartments in a manner that maximizes spatial resolution, while yielding material suitable for genomic techniques such as microarrays or deep sequencing. Neuron cell bodies have been separated from their projections in rodent and Aplysia systems, typically using membrane filters with passages large enough to allow only projections to pass through,³⁸ by laser capture,⁴¹ or by microfluidic isolation.^40,52

Following identification of transcripts with biased localization, motif analysis has been performed to identify sequence features enriched in particular regions of those transcripts, or sequences that may be conserved between species. Because cis-elements controlling localization often involve RNA secondary structure,^53-56 some may be missed by approaches that search for enrichment or conservation of linear sequences. However, structure-aware algorithms have been developed to aid in the search,⁵⁷ some of which rely on co-evolution of residues that may be involved in secondary structure formation.⁵⁸

Future work that further increases spatial resolution may be productive in the search for localizing elements. For example, subcellular fractionation by differential centrifugation has been useful in identifying protein complexes,^59,60 and may also be useful for identifying RNAs localized to subcellular compartments, provided that the fractionation procedure does not significantly disrupt interactions between RNA and its associated complexes.

Approaches could also be borrowed from the splicing field where a library of random sequences is selected for a particular function.^61,62 For purposes of RNA localization, a randomized sequence could be introduced into the 3′ UTR of a reporter gene, or endogenous gene, using genome engineering approaches. Given a strategy for isolating localized RNAs, high throughput sequencing could then be used to identify RNA elements associated with a particular subcellular location. Numerous variations on this screening methodology can be performed, for example to study different cell types, or different classes of transcripts, e.g., those with signal sequences, genes of varying expression level, or genes with alternative splicing. Following appropriate biochemical and image-based confirmation of cis-element activity, proteins that can interact with active cis-elements can be identified using pulldown/immunoprecipitation-based approaches.

The successful application of deep sequencing to these questions should accelerate the discovery of RNA localizing cis-elements. We now have improved sensitivity to detect transcripts localized to subcellular compartments and the depth of coverage to survey synthetic libraries of cis-elements in an unbiased manner.

Protein Factors Involved in RNA Localization

Mechanisms of RNA localization

In a general sense, selective enrichment of RNA molecules in specific cellular locations occurs through one of three different mechanisms. In the first, RNA is stabilized in one location and degraded in others. In the Drosophila embryo, hsp83 and nanos mRNA are degraded throughout the embryo except for the posterior pole where they are protected from degradation by elements in their 3′ UTR.^63,64 Second, mRNAs passively diffusing through the cytoplasm can be bound and retained by anchoring proteins that are themselves localized to specific cellular sites. This system has been shown to function in at least two different systems: Xenopus oocytes⁶⁵ with the RNAs Xcat2 and Xdazl and Drosophila embryos with nanos mRNA.⁶⁶ Finally, RNAs can also be actively directed to a particular location. Sequence elements that influence active localization often function through recognition of the element by a specific RNA binding protein (RBP) and subsequent targeting of the RNA to localization RNPs. These RNPs, often containing motor proteins like kinesin and dynein, are then transported along the cytoskeletal infrastructure to their destinations.

Efforts to identify a more complete list of all RNA binding proteins through RNA purification and mass spectrometry identified almost 800 proteins with the capacity to bind RNA.⁶⁷ Further work is needed to determine which are involved in RNA localization, but the pool of potential localization factors is clearly quite diverse.

Localization factors: Biochemical and molecular genetic approaches

A number of proteins have been found to interact with a localized RNA and be important for its efficient transport. We will focus on three of the best known: Zbp1, staufen, and fragile X mental retardation protein (FMRP), and present proteomic strategies for identifying new components of localization RNPs.

Zbp1

The mRNA encoding β-actin contains in its 3′ UTR a 54-nucleotide element, termed a zipcode, that is necessary for its localization to the cellular periphery.^7-9 To identify the factor responsible for binding this element, a biochemical approach was taken in which the factor was affinity-purified from a complex extract using probes of the zipcode sequence as bait.⁶⁸ This study identified Zbp1 as the primary factor bound to the zipcode. Zbp1 was later shown to be functionally relevant as RNAi-mediated knockdown of its mammalian homologs resulted in defects in cell morphology.⁶⁹ Later studies demonstrated the importance of Zbp1-mediated β-actin mRNA localization in other systems, in particular neurons and neuronal growth cones.^14,70-72

Staufen

In contrast to the biochemical identification of Zbp1, staufen was first identified through a genetic screen for Drosophila in which pole cell formation and pattern formation across the embryo was inhibited.⁵⁶ Staufen was later shown to be required for proper transport and localization of bicoid and oskar RNA in the embryo and to require specific sequences in the 3′ UTR of bicoid in order to direct its localization.^53,73 In adult flies, staufen is important in many neuronal processes, including long-term memory formation.^74,75

In mammals, there are two staufen paralogs. They are involved in RNA transport in neurons, particularly to dendrites.^55,76 Expression of a dominant negative form of staufen results in inhibition of RNA transport to dendrites,⁷⁷ and neurons from knockout mice are deficient in dendritic delivery of RNPs while the mice themselves show defects in locomotor activity.⁷⁸ Mammalian staufen has four double-stranded RNA binding domains, although not all of them have strong double-stranded RNA binding activity.⁵⁴ Besides a preference for double-stranded RNA, staufen has long been thought to be relatively nonspecific in its RNA binding, but recent work has demonstrated that staufen may in fact prefer distinct double-stranded structures.⁷⁹

Fragile X mental retardation protein

The fragile X mental retardation protein (FMRP) was identified through its location in a fragile site on the X chromosome associated with mental retardation.⁸⁰ The FMRP transcript contains a CGG trinucleotide repeat in its 5′ UTR. When this repeat is expanded, the gene becomes methylated and transcriptionally silenced.^81,82

In mammalian neurons, FMRP is localized to spines and growth cones, while also localizing with a distinct population of RNAs.^83,84 This localization to dendrites is particularly robust in response to depolarization.⁸⁵ In addition, FMRP colocalizes with ribosomes and is believed to be involved in translational repression⁸³ as neurons from FMRP knockout mice show high basal levels of translation but are deficient in translation in response to a stimulus.⁸⁶ In response to stimulation, FMRP regulates localization of its targets as well as their translation as FMRP targets are increased in dendritic cultures after stimulation, but not in FMRP knockout neurons.⁸⁷

Localization factors: Genome-wide approaches

Each of the proteins discussed above was identified individually. As new approaches to nucleic acid sequencing allow richer and more complete pictures of localized transcriptomes, newer proteomic approaches allow more complete characterization of localization RNPs, their components, and their RNA cargos.

Identification of trans factors

Efforts to identify and characterize localization RNPs have mainly centered around strategies of RNP purification whereby a known component of an RNP is purified from lysate and copurifying proteins are identified by mass spectrometry. One of the first studies to do this with an eye toward RNA localization used this strategy with a GST-tagged fragment of the kinesin Kif5 and lysate from mouse brain.⁸⁸ Over 40 proteins, including FMRP and Staufen, copurified, as well as the known dendritically-localizing RNAs CaMKII-α and Arc. RNAi knockdown of some of these proteins inhibited dendritic transport of reporter RNAs.

A similar study using IMP1, a mammalian ortholog of ZBP1, as the purification bait identified hnRNP proteins and ribosomal proteins as RNP components.⁸⁹ An identification of the interactors of staufen2 also identified ribosomal proteins and cytoskeletal components.⁹⁰ Fritzsche and colleagues analyzed the interacting partners of two proteins, barentsz and staufen2, finding several proteins in common between the two but also a large number of differences, implying that the composition of different localization RNP particles may be more diverse than expected.⁹¹ Several translational repressors were found in both complexes, consistent with the idea that transported RNA molecules are translationally repressed while they are in transit.⁹² Interestingly, RNA-induced silencing complex (RISC) components were also found in these particles, raising the possibility of location-specific small RNA-mediated regulation in these cells.

Identification of trans-factor binding sites

Currently, a determination of the binding sites of RNA binding proteins generally involves in vivo crosslinking of proteins to RNA using UV irradiation, followed by purification of the protein of interest and sequencing of the associated RNAs.^93-95 Such approaches, with variations, have been undertaken for many proteins involved in other aspects of RNA metabolism, including alternative splicing, and have recently been applied to proteins involved in RNA localization.

Staufen and FMRP have been particularly studied using such methods. FMRP has been proposed to bind a G-quartet RNA structure.^96,97 Immunoprecipitation of FMRP from mouse brain lysates followed by microarray analysis of the coimmunoprecipitated RNA revealed over 400 potential RNA binding targets of FMRP.⁸⁴ Many of these messages were translationally misregulated in the absence of FMRP and also contained putative G-quartet structures. A later PAR-CLIP⁹⁸ based study identified two independent preferred motifs for FMRP from sequences immunoprecipitated with FMRP.⁹⁹ Biochemical analyses demonstrated that these motifs corresponded to the sites bound by two independent RNA-binding KH2 domains.

Identifying targets of staufen and the rationale behind its binding has been less straightforward. Staufen binds to double-stranded RNA, but whether it has specificity for particular sequences or structural motifs is unclear. RNA immunoprecpitation followed by microarray (RIP-CHIP) analysis of staufen targets in Drosophila embryos revealed a pattern of staufen binding. 3′ UTRs of genes bound by staufen tended to be long and contain structured regions of particular combinations of basepaired regions and loops,⁷⁹ providing evidence for specificity in binding. A similar analysis in C. elegans demonstrated that the worm ortholog, STAU-1, also possesses an ability to bind double-stranded RNA while retaining a specificity for distinct transcripts in vivo.¹⁰⁰ Analyses of human Stau1 and Stau2 by RIP-CHIP showed that there may be few mRNA targets in common between the two human staufen homologs,¹⁰¹ and a separate study showed a possible role for staufen in neural lineage progression by promoting asymmetric segregation of some target RNAs.¹⁰² RIP-CHIP of staufen from rat brain again demonstrated an RNA-binding specificity for staufen, with an enrichment for long 3′ UTRs and stem structures at least 12 base pairs long and containing no more than 2 unpaired bases.¹⁰³ However, since the studies above identify whole transcripts bound by staufen, they give comparatively little immediate data on the direct binding sites of the protein.

In a variant of RIP called RIPiT in which proteins are immunoprecipitated after formaldehyde crosslinking,¹⁰⁴ binding sites are narrowed from a search space of an entire transcript to a short RNA fragment that is protected from nuclease digestion by the bound proteins.¹⁰⁵ In these experiments, Stau1 is observed associating with translating ribosomes and long, stable secondary structures such as inverted Alu pairs. The authors argue that it is the propensity to form double-stranded regions, both in the coding sequence and 3′ UTR, that is the main driver of Stau1 binding.

CLIP experiments have also been performed on SMN, which is involved in snRNP biogenesis, but also colocalizes with RNAs in granules associated with mRNA transport in neurons.¹⁰⁶ These experiments identified hundreds of targets of SMN, many of which were previously shown to localize to axons. Furthermore, the axonal localization of many transcripts bound by SMN decreased upon depletion of SMN, implicating SMN in the trafficking and localization of these transcripts. Thus, defective RNA transport may contribute to pathogenesis in spinal muscular atrophy, a disease caused by mutations in the SMN1 gene.

These experiments are some of the first to take a genome-wide approach to identification of the RNA binding sites of localization factors. As more high-resolution binding data become available for more factors, it will be interesting to begin to cross-reference binding sites with sequences of localized transcripts to begin to piece together a more mechanistic understanding of the regulatory principles of RNA localization. Some localized transcripts may be bound by multiple factors, either concurrently or sequentially, raising the possibility of combinatorial effects. The identification of a larger number of binding sites for known localization factors will inform us about the accuracy of the canonical examples of their binding are as well as give us better characterizations of the sequences required for localization.

Localized Translation and Localization of Proteins

Of course, localization of the RNA itself is commonly not the end of the story. Translation of messages is often repressed while it is in transit and may only be activated in response to a specific stimulus. Well-established methods have been developed to study localized translation, but these are limited to the interrogation of only a few genes at a time and usually use reporters, rather than native transcripts. In this section we will summarize these current methods and highlight recent techniques that examine localized translation on a genome-wide scale.

Localized translation and protein localization: Imaging and proteomic approaches

Localized translation

In a sense, almost all methods used to analyze localized translation address one basic problem. Namely, how can a molecule of protein produced in situ in a particular cellular locale be distinguished from molecules made elsewhere that diffuse to the same location? Several methods have been devised that simply and elegantly answer this question using fluorescence-based reporters.

Two successful methods using the translation of fluorescent proteins involve destabilized and anchored proteins. GFP variants that are destabilized to reduce their half-life to a few hours^107,108 have been used to identify newly synthesized proteins with the logic that older ones would be degraded before they could diffuse to the site of interest.^109,110 GFP molecules that contain a signal sequence for acylation, usually myristoylation, result in the newly translated GFP being inserted into the cell membrane shortly after its translation.¹¹¹ This setup has been used to measure the translation of reporter constructs in dendrites and axons in response to stimuli¹¹² and as a proxy for identifying the localization of the mRNA as well as the site of translation.¹¹⁰

Other approaches ablate or modify the spectral properties of the molecule. Fluorescence recovery after photobleaching (FRAP) based approaches remove fluorescence from existing GFP molecules and then wait for the appearance of signal from newly translated GFP molecules at the site of interest before molecules translated elsewhere arrive by diffusion. For example, the appearance of signal in axons on a short timescale (∼20 min) is interpreted as evidence that the new signal is due to locally translated GFP.^112,113 Similarly, by converting the fluorescence emitted from photoconvertible GFP molecules from green to red, the appearance of new molecules of protein can be analyzed. This technique has been used to monitor the axonal translation of protein¹¹⁴ and the retrograde movement of protein from growth cones back toward the cell body.¹¹⁵

Proteomics of organelles

Subcellular organelles often have proteomic environments distinct from the rest of the cell. Several groups have undertaken efforts to purify certain organelles using differential sedimentation and, using mass spectrometry, determine the proteins contained within them. Over 1000 proteins were identified as present in the Golgi and secretory pathway in rat liver homogenates.¹¹⁶ Similarly, several hundred proteins were seen in synaptic vesicles prepared from rat brain,¹¹⁷ and many more in mitochondria.¹¹⁸ Yoon and colleagues used non-canonical amino acids that could be labeled with a fluorescent dye using Click chemistry. These amino acids are incorporated into newly synthesized proteins. By separating axons from cell bodies and culturing them on their own, the appearance of new proteins in the axonal proteome in response to stimuli could be measured on 2D-gels and by mass spectrometry.¹¹⁹ Although these approaches have identified a rich diversity of proteins in several cellular structures, they rely on rigorous and clean biochemical methods to purify these organelles and structures. An alternative strategy instead relies on a localized enzyme that has the ability to tag nearby proteins and allow their efficient purification. By targeting an enzyme that has the ability to convert biotin from an inactive to an active form to the mitochondria, Rhee et al. purified several hundred proteins with a high degree of spatial precision.¹²⁰

Localized translation: Genome-wide approaches

Although the above organelle-based methods have been highly successful in identifying local proteomes, they are limited to answering questions about which proteins reside in a particular area, not where they are produced. Recent advances in identifying RNA molecules that are in a state of active translation have allowed a more global understanding into the translational output of a cell at any one time.¹²¹ These methods include ribosome footprint profiling in which RNA fragments protected from RNase digestion by ribosomes are sequenced and polysome-sequencing in which mRNA molecules that co-sediment with polysomes are sequenced. By applying these techniques to ribosomes localized to specific parts of the cell, new insights have been gained into the spatial nature of translational regulation.

Fractionation of HEK293 cells into cytosolic and ER-associated fractions followed by ribosome footprint profiling revealed that although messages encoding cytosolic transcripts can be translated on ER-associated ribosomes, mRNAs that were associated with the ER were more highly translated than those in the cytosolic compartment.¹²² Similar experiments in mouse myoblasts showed that the translational activity of mRNAs can be altered by perturbations that change their subcellular localization without affecting their overall abundance.²¹ The translational status of localized messages is also likely to be influenced by bound protein factors.⁹¹

These findings hint at possible broad differences in translation efficiency between different cellular compartments. Only a handful of cellular locations have been probed using approaches that monitor translation globally. However, previous, more targeted work has demonstrated many instances of translational regulation during mRNA localization,^84,123 making this a promising area for future global analysis. Further work in this area is needed to determine the extent to which translation is spatially regulated.

Conclusion

Although genomic techniques have been around for many years, they have only recently become sensitive enough to be efficiently applied to RNA localization. Studies in transcription and splicing have benefited from genomic insights, and future studies in RNA localization may also benefit from applying these methods. The diversity of systems and biological processes associated with RNA localization may also provide new opportunities for the application of high-throughput methods.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.