|Home | About | Journals | Submit | Contact Us | Français|
The localization of mRNAs to subcellular compartments provides a mechanism for regulating gene expression with exquisite temporal and spatial control and recent studies suggest that a large fraction of mRNAs localize to distinct cytoplasmic domains. In this review, we focus on cis-acting RNA localization elements, RNA-binding proteins, and the assembly of mRNAs into granules that are transported by molecular motors along cytoskeletal elements to their final destination in the cell.
The process of mRNA localization and regulated translation has classically been considered to be a mechanism used by a handful of transcripts to spatially and temporally restrict gene expression to discrete sites within highly polarized, asymmetric cells. To date, the best-studied examples of mRNA localization all involve transcripts whose protein products play specialized roles within well-defined subcellular compartments. These include the mRNA encoding the transcriptional repressor ASH1 in budding yeast, which inhibits mating type switching. ASH1 mRNA is transported to the bud tip of a dividing cell such that it is delivered only to the nucleus of the daughter cell, thereby ensuring that the mother and daughter cells have distinct mating types (Paquin and Chartrand, 2008). In fruit fly Drosophila, the localization of mRNAs, such as bicoid, oskar, and nanos to anterior and posterior poles of the oocyte, helps establish morphogen gradients that underlie the proper spatial patterning of the developing embryo (Johnstone and Lasko, 2001). Similar processes occur in oocytes of the frog Xenopus, where localization of the mRNA encoding the T-box transcription factor VegT localizes to the vegetal pole and induces endodermal and mesodermal cell fates in the embryo (King et al., 2005). In fibroblasts, β-actin mRNA localizes to the lamellipodia, where its translation is required for cytoskeletal-mediated motility(Condeelis and Singer, 2005). In oligodendrocytes, the mRNA encoding myelin basic protein (MBP) is transported into the distal processes where myelination occurs (Smith, 2004). During brain development, local translation of mRNAs in axonal growth cones allows neurons to respond to local environmental cues as the distal axonal processes navigate towards their synaptic partners (Lin and Holt, 2007). In the mature brain, the regulated translation of synaptically localized mRNAs allows each of the thousands of synapses made by a given neuron to autonomously alter its structure and function during synaptic plasticity, thereby greatly enhancing the computational capacity of the brain (Martin and Zukin, 2006).
Although these examples are of RNAs encoding proteins with specialized local functions, more recent studies indicate that the localization of mRNAs to particular subcellular compartments may be much more prevalent than previously thought. In a recent study involving high-throughput, high resolution in situ hybridizations of over 3,000 transcripts in Drosophila embryos, 71% were found to be expressed in spatially distinct patterns (Lecuyer et al., 2007). Similarly, in mammalian neurons, it was once thought that only a handful of mRNAs localized at synapses. However, more recent studies indicate that hundreds of mRNAs are present in neuronal processes, where they encode diverse functionalities (Eberwine et al., 2002; Martin and Zukin, 2006). Further, the analysis of RNA localization in migrating fibroblasts (Mili et al., 2008), Xenopus oocytes (Blower et al., 2008) and Drosophila embryos (Lecuyer et al., 2007) may reveal subcellular compartments that had previously been unappreciated, and thus these findings may lead to a more detailed and nuanced understanding of cellular architecture.
What are the advantages of regulating gene expression by mRNA localization? The most obvious is that it allows gene expression to be spatially restricted within the cytoplasm. A second advantage is it that this spatially restricted gene expression can be achieved with high temporal resolution given that local stimuli can regulate translation on-site instead of requiring a signal to be delivered to the nucleus to initiate transcription, followed by mRNA export, cytoplasmic translation, and subsequent targeting of the protein to the site of stimulation. A third advantage is one of economy—localized mRNAs can be translated multiple times to generate many copies of a protein, which is much more efficient than translating mRNAs elsewhere in the cell, then transporting each protein individually to a distinct site. A fourth advantage, exemplified by the localization of MBP mRNA in oligodendrocytes, is that the local translation of proteins can protect the rest of the cell from proteins that might be toxic or deleterious in other cellular compartments.
The targeting of mRNAs to specific subcellular sites involves multiple steps. The cellular “address” of transcripts is encoded by cis-acting elements in the RNA. As detailed below, these cis-acting elements, called “localization elements” or “zipcodes” are most frequently found in the 3′ untranslated region (UTR), although in some cases they are present in the 5′UTR or in the coding sequence. Localization elements are recognized by specific RNA-binding proteins that often function both in transcript localization and translational regulation. Several studies indicate that the processing of pre-mRNAs in the nucleus is required for the recruitment of RNA binding proteins that determine the RNA’s eventual localization in the cytoplasm—that is, the nuclear history of an mRNA is critical to its cytoplasmic localization (Giorgi and Moore, 2007). The complex of RNAs and RNA binding proteins, called ribonucleoproteins (RNPs), in many cases forms part of a larger structure called an RNA transport granule, which is transported by motor proteins along cytoskeletal elements to its final destination in the cell, where additional mechanisms anchor the RNA in place. Finally, there are mechanisms to maintain the RNA in a translationally-repressed state during delivery and to regulate its translation at the right place and right time after delivery [for review, see (Besse and Ephrussi, 2008)].
In this review, we focus on the following steps of mRNA localization: the identification of cis-acting elements; the identification and function of trans-acting RNA binding proteins; the assembly of RNA transport granules; and the role of molecular motors and cytoskeletal elements in mRNA localization. Insight into each of these steps has emerged from studies that make use of a variety of approaches, including genetic analysis of mRNA localization in model organisms such as yeast and Drosophila, cell biological imaging of individual mRNAs in fixed and live cells, and bioinformatic analysis of localized transcripts. The field is at a stage where many individual pieces of data have been collected, and from these pieces, patterns and general principles are beginning to emerge. Our aim in this review is to highlight these general principles while providing specific and illustrative examples.
The initial indication that mRNAs could localize within cells came from in situ hybridization studies in which distinct mRNAs showed very specific patterns of localization within a cell. Many of these studies were done in large egg cells or in asymmetric cells such as fibroblasts, oligodendrocytes and neurons. The first studies indicating that cis-acting RNA elements are required for localization involved genetic and microinjection studies in which elements of the localized mRNAs were fused to hybrid genes in order to identify sequences that were required for localization. These and many subsequent studies showed that localization elements are most often found in the 3′UTR, and can range in length from five or six to several hundred nucleotides. Although no clear consensus has emerged as a “localization” sequence, localization elements have been shown in some instances to function in multiple cell types suggesting that they are recognized by common, shared RNA binding proteins (Bullock and Ish-Horowicz, 2001).
The following additional principles have emerged from studies aimed at identifying cis-acting localization elements: localization elements are often repeated and redundant; distinct localization elements mediate distinct steps in localization; and localization elements can form secondary structures, usually stem loops, that are critical for localization. Cis-acting elements can promote localization in three different ways: 1) by active and directed transport of the transcript to a subcellular site (the most common mechanism described to date); 2) by mediating the local stabilization and regulated degradation of mRNAs; and 3) by locally trapping an mRNA that diffuses throughout the cytoplasm.
The systematic identification of cis-acting elements that mediate bicoid mRNA localization to the anterior pole of Drosophila oocytes illustrates many of the principles of directed transport. Macdonald and Struhl (Macdonald and Struhl, 1988) first showed that 625 base pairs (bp) of the 3′UTR of bicoid was necessary for localization of a hybrid mRNA to the anterior pole of oocytes. By expressing transgenes containing smaller deletions, they subsequently identified several elements (called bicoid localization elements, or BLEs) within this 625 bp region that were required for localization (Macdonald et al., 1993). These studies further showed that one of these elements, BLE1, consisting of 50 nucleotides, formed a stem loop structure that was specifically required for bicoid mRNA transport from the nurse cells into the oocyte. Additional stem loop structures were required for later steps in localization and still an additional stem loop was required for RNA anchoring at the anterior pole (Ferrandon et al., 1997; Macdonald and Kerr, 1997). Mutations in BLEs that altered the primary sequence of the localization element while conserving the stem-loop structure were shown to permit mRNA localization (Ferrandon et al., 1997), demonstrating the critical role that secondary structure play in mRNA localization. Finally, bicoid RNA was shown to dimerize in vitro through interactions between specific hairpin loop structures, and in embryo injection assays this dimerization was shown to be essential for binding of the RNA-binding protein Staufen (Ferrandon et al., 1997). Staufen, in turn, is necessary for bicoid mRNA localization at the anterior during the late stages of oogenesis (St Johnston et al., 1991; Weil et al., 2006), suggesting that dimerization is an important step in bicoid localization. These elements are conserved in the 3′UTR of other Drosophila species (Luk et al., 1994), which not only demonstrated their functional importance, but also provided an initial basis for using conservation algorithms to search for localization elements in UTRs [see for example (Doyle et al., 2008)].
In Xenopus oocytes, the transcript encoding Vg1, a TGF-β family member with mesoderm-inducing capacity, is distributed uniformly in early oocytes, localizes to the vegetal hemisphere during mid-oogenesis, and anchors at the vegetal pole of the oocyte during late oogenesis (Melton, 1987). Microinjection of synthetic transcripts containing elements of the Vg1 RNA showed that 340 nucleotides of the 3′UTR were required to localize the mRNA to the vegetal pole (Mowry and Melton, 1992). Comparison of this region between two frog species revealed two 5–6 nucleotide-long sequences, called VM1 and E2 elements or motifs (Lewis et al., 2004). The Vg1 3′UTR contains multiple, clustered copies of these localization elements that act synergistically to localize Vg1 mRNA (Deshler et al., 1997; Lewis et al., 2004). The clustering of repeated, occasionally redundant localization elements has been observed in other localized transcripts, and is thought to facilitate binding of trans-acting factors to form a transport-competent RNP.
The cis-acting sequences that mediate ASH1 mRNA targeting to the bud tip provide another example of repetitive and synergistic clustering of localization elements. ASH1 mRNA contains four localization elements, three of which (E1, E2A, E2B) are in the coding sequence of ASH1, whereas the fourth (E3) overlaps the coding sequence and 3′UTR (Chartrand et al., 1999; Gonzalez et al., 1999). These localization elements were all predicted to form stem-loop structures, and mutations that disrupt secondary structure formation are localization-incompetent. Each element on its own is capable of localizing a reporter RNA, although the presence of four elements increases the efficiency of localization (Chartrand et al., 2002). One idea that has been put forth is that multiple, clustered repeats of localization elements in transcripts function to create a local concentration of RNA binding proteins. These in turn may promote the binding of additional proteins that bind not to the RNA but to the RNA binding proteins and are necessary for RNA localization (Arn et al., 2003). Furthermore, it appears that many of the RNA binding proteins involved in RNA localization bind to individual sites with low affinity and specificity, leading to the notion that RNA transport complexes may involve a large number of low-affinity interactions of proteins with the RNA, none of which is absolutely essential for localization (Arn et al., 2003).
One of the earliest identified localization elements was from the chick β–actin mRNA (Kislauskis et al., 1994). β–actin mRNA was found to localize to the lamellipodia of motile fibroblasts (Lawrence and Singer, 1986), and experiments expressing reporter plasmids containing elements of the β–actin mRNA revealed that a 54 nucleotide-long sequence in the 3′UTR was essential and sufficient for mRNA localization. This element was termed the “zipcode,” given that it contained the cytoplasmic delivery address for transport. This sequence, and in particular the hexanucleotide sequence ACACCC, was found to be conserved in β–actin transcripts from many other species. Chick β–actin mRNA contains tandem repeats of this hexanucleotide motif, and mutations in this region inhibits localization (Ross et al., 1997). Secondary structure analysis again predicted that the β-actin zipcode forms a stem-loop structure.
Analysis of the targeting of MBP mRNA to the myelinating compartments of oligodendrocytes, done by microinjection of synthetic reporter RNAs into oligodendrocytes, led to the identification of two distinct localization elements (Ainger et al., 1997). One of these, called the RNA trafficking signal (RTE) or A2RE (because it binds heterogeneous ribonucleoprotein A2), is only 11 nucleotides long (Munro et al., 1999), and is the shortest known localization element sufficient for RNA targeting. Homologous sequences are present in transported mRNAs in other species and cell types (Ainger et al., 1997), and the MBP A2RE has been shown to be sufficient to transport mRNAs into neuronal dendrites (Shan et al., 2003). The A2RE does not, however, act on its own to fully localize MBP mRNA. Rather, it directs transport out of the soma and towards oligodendrocyte processes, but is not sufficient to target the mRNA into the myelinating compartment. For this, a second much larger localization element, called the RNA localization region or signal (RLR or RLS), is necessary. The RLR extends over much of the 3′UTR and is predicted to form a complex secondary structure with multiple stem-loops. Intriguingly, the RLR is only necessary when the protein-coding region of the RNA is included in the reporter transcript; in the absence of a protein coding region, the A2RE is sufficient to fully localize the reporter RNA (Ainger et al., 1997).
In the examples described above, the identification of localization elements has been relatively straightforward and uncontroversial. This, however, is not always the case. As one example, studies aimed at identifying sequences that mediate the dendritic localization of the mRNA encoding CamKIIα in neurons have generated conflicting results. Thus, Mori et al. (Mori et al., 2000) defined a 94 nucleotide-long element in the 3′UTR of CamKIIα that was sufficient for dendritic localization of the transcript whereas Miller et al. (Miller et al., 2002) showed that CamKIIα mRNA containing the 94-nucleotide element but lacking the rest of the 3′UTR was not dendritically localized in a genetically modified mouse expressing the CamKIIα transgene. A third distinct, large localization element in the middle of the 3′UTR of CamKIIα was identified by Kindler and colleagues (Blichenberg et al., 2001). The inconsistency in these results may be due to differences between the localization of reporter constructs that are expressed as cDNAs and reporter constructs that are expressed as genomic constructs and undergo RNA processing in the nucleus, and could therefore depend on a different and physiologically more relevant set of cis-acting elements and RNA binding proteins for localization.
mRNA localization can also occur by selective stabilization of the transcript. The best-characterized example of this is the mRNA encoding heat shock protein 83 (hsp83), which localizes to the posterior pole plasm of Drosophila embryos. In this “protection/degradation” mechanism of localization (Bashirullah et al., 2001), hsp83 mRNA is degraded everywhere in the cytoplasm except in the pole plasm. Both degradation and local protection at the posterior pole have been mapped to elements in the 3′UTR of hsp83 mRNA. Hsp83 transcript degradation involves the binding of the multifunctional RNA-binding protein Smaug, which recruits the CCR4/Not deadenylase and thus triggers deadenylation and subsequent degradation of multiple transcripts in Drosophila (Tadros et al., 2007).
nanos is another mRNA whose localization is in part the result of protection in one location, and degradation elsewhere; only 4% of nanos mRNA is localized at the posterior pole of the Drosophila embryo (Bergsten and Gavis, 1999), but this fraction of the mRNA pool is stable whereas nanos mRNA elsewhere in embryo is targeted by 3′UTR-bound Smaug for deadenylation and degradation (Zaessinger et al., 2006). As a consequence, the concentration of nanos is more than a hundred times greater in the posterior pole cytoplasm than elsewhere in the embryo.
A second mechanism, “diffusion/entrapment,” also contributes to nanos localization, which occurs during the late stages of oogenesis, when strong cytoplasmic flows move nanos RNA swiftly throughout the oocyte such that it can readily encounter a localized actin-based anchor at the posterior pole (Forrest and Gavis, 2003). Such facilitated diffusion and entrapment, while promoting the enrichment of nanos RNA at the posterior pole, is not very efficient, and the added engagement of a protection/degradation mechanism, as well as local translation of nanos, ensures that nanos activity is properly restricted to the posterior pole of the embryo.
Regulated nuclear events, such as splicing and alternative polyadenylation site selection, can generate different RNA isoforms with different targeting specificities. Thus, in rat hippocampal neurons, two mRNA isoforms for brain-derived neurotrophic factor (BDNF) are produced by the differential use of polyadenylation sites, resulting in the selective inclusion of localization elements in the long isoform, which target the message to dendrites (An et al., 2008). In Drosophila, the Crumbs/Stardust (Std)/PATJ complex is a determinant of epithelial polarity that is targeted to the apical membrane of young embryos by the cytoplasmic scaffolding protein Std. An elegant study has shown that std mRNA is developmentally regulated. During early stages of development, std is apically localized as a result of alternative splicing that results in differential inclusion of the 3rd exon, which contains an apical targeting element (Horne-Badovinac and Bilder, 2008).
The lack of conserved primary sequence motifs and the relatively small number of localization elements that have been defined experimentally have hindered the development of bioinformatic tools for predicting localization elements in other mRNAs. However, approaches based on the conservation of primary sequence have achieved some limited success. As one example, the CAC motif found in the E2 element of Xenopus Vg1 mRNA was used to search for other mRNAs containing this motif, and led to the identification of similar motifs in most of the transcripts known to localize to the vegetal pole of the Xenopus oocyte (Betley et al., 2002). Other efforts have concentrated on the identification of conserved secondary structures, and in particular on stem loop structures. Using alignment algorithms to look for common, shared secondary structures, however, is significantly more complex than it is for shared primary sequences. Nonetheless, the development of new tools for motif discovery (Doyle et al., 2008) and the increasing availability of information on the localization of transcripts promise to dramatically improve prediction methods in the future. As one example, bioinformatic analysis of mRNAs found to show similar patterns of localization in Drosophila embryos led to the identification of nine predicted “localization” motifs, all of which contained stem-loop structures (Rabani et al., 2008). Predictive approaches and tools will likely continue to improve as more localized mRNAs are discovered and characterized, as the structures of specific RNAs and RNPs are solved, and as more powerful computational algorithms based on RNA secondary structures are developed. Importantly, the development and utility of these algorithms will depend on careful experimental validation.
Localization elements do not function independently to target mRNAs for delivery; rather, they are recognized by trans-acting proteins that bind the RNA to form an RNP. As such, these play a critical role in directing and regulating mRNA localization. The identity of the trans-acting factors involved in mRNA trafficking has emerged primarily from two types of studies: genetic screens for genes involved in mRNA localization and affinity purification of proteins that bind the identified localization elements.
One of the best-characterized trans-acting factors involved in mRNA localization is Staufen. Staufen was first identified in genetic screens because of its role in pattern formation, reflecting its function in localizing oskar and bicoid mRNAs in Drosophila oocytes (St Johnston et al., 1991). Staufen and oskar are interdependent for their localization to the posterior pole during oogenesis (Ferrandon et al., 1994), most likely through the interaction of Staufen with the oskar 3′UTR (Jenny et al., 2006). Staufen, which binds stem-loop structures within the bicoid 3′UTR forms RNPs that can travel along microtubules when injected into early embryos (Ferrandon et al., 1994). More recent work demonstrating that mammalian Staufen is involved in the targeting of mRNAs, including CamKIIα mRNA, to neuronal dendrites (Kiebler et al., 1999; Tang et al., 2001) indicates that its function in localizing transcripts within the cytoplasm is evolutionarily conserved (Roegiers and Jan, 2000). Staufen has five distinct RNA binding domains (four in mammalian Staufen), each of which binds double-stranded RNA (dsRNA) (St Johnston et al., 1992). Interestingly, on their own these domains bind dsRNA indiscriminately, indicating that additional proteins may be recruited to the RNP to achieve specificity (Ferrandon et al., 1994; St Johnston et al., 1992).
To identify additional factors mediating the localization of bicoid mRNA in Drosophila oocytes, Irion and St. Johnston (2007) used a fluorescent protein tag to visualize the localization of Staufen. Surprisingly genetic screens using this construct revealed that a subunit of the ESCRT-II complex, VSP22 (homolog of yeast vesicular sorting protein 22) is involved in bicoid localization. The ESCRT complex is best characterized for its role in endosomal trafficking, where it is required for the sorting of ubiquitinated membrane proteins from endosomes to multivesicular bodies. All components of the ESCRT-II complex were subsequently shown to be required for the last step of bicoid localization to the anterior pole of oocytes, and one of the components, VSP36, was demonstrated to bind directly to the proximal part of stem-loop V in the bicoid 3′UTR (part of the BLE). Binding involved the GLUE domain, the same domain of VSP36 previously shown to bind ubiquitinated membrane proteins. Xenopus VSP36 also bound the bicoid 3′UTR, indicating that this mechanism is conserved in vertebrates. These findings raise the intriguing possibility that endosomal sorting and mRNA localization are linked, or that they use similar molecules to direct trafficking (Rusten and Stenmark, 2007).
Affinity purification methods in chick embryo fibroblasts led to the identification of a 68 kD protein that binds to the β-actin mRNA zipcode (Ross et al., 1997). This protein was called zipcode-binding protein, or ZBP1. ZBP1 contains two RNA recognition motif (RRM) RNA binding domains, and four hnRNP K homology (KH) RNA binding domains (Farina et al., 2003). Distinct functions have been identified for each of these domains. Specifically, the KH domains mediate binding to the zipcode, formation of an RNP and association with actin microfilaments, whereas the RRM domains are required for the localization of the β-actin RNP. Homologs of ZBP1 have been identified in Xenopus, Drosophila, human, and mouse, and in each case have been implicated in mRNA localization. Thus, in Xenopus oocytes, the ZBP1 homolog Vera binds to localization elements in Vg1 3′UTR and is required for localization to the vegetal pole (Deshler et al., 1997). ZBP1 is also present in mammalian neurons, where it binds β-actin mRNA (Zhang et al., 2001). In developing neurons, ZBP1 localizes to growth cones, where stimulus-induced local translation of β-actin is required for growth cone navigation (Lin and Holt, 2007). In mature neurons, ZBP1 undergoes activity-dependent trafficking and dynamic localization in dendrites and spines (Tiruchinapalli et al., 2003). Of note, ZBP1 has been shown to be phosphorylated by Src kinase, resulting in reduced binding of ZBP1 to RNA and to increased translation of β-actin in neuroblastoma cells (Huttelmaier et al., 2005). This finding suggests that ZBP1 functions both in mRNA localization and translational repression.
Another zipcode binding protein, ZBP2, was recently identified by affinity purification (through binding to the zipcode) and is a predominantly nuclear protein that also affects β-actin localization in the cytoplasm (Gu et al., 2002). ZBP2 is a homolog of human KH domain-containing splicing regulatory protein (KSRP). Like ZBP1, ZBP2 orthologs identified in other species have also been shown to be involved in mRNA localization. Thus, in rat, the homolog of ZBP2 is MARTA1 (for MAP2-RNA trans-acting protein), which binds the 3′UTR of MAP2, a dendritically localized mRNA in neurons (Rehbein et al., 2000), and in Xenopus, the homolog of Zbp2, VgRBP71, also binds the Vg1 mRNA, which is localized to the vegetal pole of the egg (Kroll et al., 2002). Recent studies of chick ZBP2 have indicated that ZBP2 binds the nascent β-actin zipcode cotranscriptionally, and facilitates the binding of ZBP1 to the zipcode (Pan et al., 2007). The role for ZBP2 in the nucleus thus provides one example of the principle that the association of proteins with the RNA in the nucleus is required for ultimate localization in the cytoplasm. It also provides an example of how interactions between RNA binding proteins and RNA serve to recruit and stabilize additional proteins to form a large RNP. One consequence of this cooperative binding is that it becomes critical to analyze the function of individual RNA binding proteins in cells that are null for the endogenous protein. Thus, although all the data suggest a requirement for ZBP1 in β-actin mRNA transport in mammalian cells, this formally remains to be shown, by analysis of genetic knockout or knockdown strategies.
A striking example of the role of nuclear processing in transcript localization within the cytoplasm involves the exon junction complex (EJC), a set of proteins that bind to mRNAs during splicing. The core proteins of the EJC, eIF4AIII, Barentsz, Mago Nashi and Tsunagi (fly orthologs of MLN51, Magoh and Y14), are deposited upstream of exon-exon junctions on mRNAs concomitant with splicing, and are thought to remain bound to the mRNA in the cytosol until they are removed during first round of translation (Tange et al., 2005). Several of these proteins were first identified as genetic mutants affecting oskar localization in Drosophila oocytes, and subsequently all four members of the EJC core were shown to be essential for this process. Consistent with the requirement for splicing in EJC deposition, the correct localization of oskar mRNA to the oocyte posterior pole was shown to depend on the presence of the first intron (Hachet and Ephrussi, 2004), in addition to the 3′UTR. However, the identity of the intron was irrelevant and only its position mattered, suggesting that placement of the EJC at a precise position on the RNA might be important for proper architecture of the RNP for oskar transport.
The notion that EJC components may be required for mRNA localization gained further support from studies of mammalian neurons, which showed first that Magoh, Y14 and MLN14 are all present in dendrites (Glanzer et al., 2005; Macchi et al., 2003) and further that eIF4AIII is associated with neuronal mRNA granules and dendritic mRNAs (Giorgi et al., 2007)., Knockdown of eIF4AIII in neurons increased synaptic strength and enhanced expression of the activity-regulated cytoskeletal (arc) protein, which is encoded by an mRNA that is localized to dendrites. Antibodies to eIF4AIII immunoprecipitated a number of transcripts localized to dendrites, including mRNAs encoding dendrin, arc, MAP2, CamKIIα, GluR1 and NR1.
Studies of the RNA binding proteins that bind to ASH1 mRNA in yeast have provided perhaps the greatest mechanistic insights into the role of trans-acting factors in mRNA localization [for review see (Paquin and Chartrand, 2008)]. Genetic and biochemical analyses identified a number of RNA binding proteins required for ASH1 localization to the bud tip, including She2p, a protein that shuttles between nucleus and cytoplasm and that recruits She3p, which binds the type V myosin protein Myo4p (also called She1p). Together, She2p, She3p and Myo4p form what has been called the “locasome,” which mediates trafficking of the ASH1 mRNA along the actin cytoskeleton. She2 binding to ASH1 mRNA is also required for translational repression, an example of an RNA binding protein that functions both in the directed trafficking and the translational regulation of the localized mRNA. Loc1p, a nuclear protein that is involved in ribosome assembly and export, has also been shown to be required for the binding of She2p to the ASH1 mRNA, providing another example of a nuclear RNA binding protein affecting cytoplasmic localization of the RNA binding protein (Long et al., 2001).
In mammalian oligodendrocytes, the identification of a short sequence in the MBP mRNA that mediated localization to myelinating processes allowed the use of synthetic oligonucleotides containing this sequence to affinity purify the transacting RNA binding protein from rat brain (Hoek et al., 1998), which turned out to be hnRNP A2. hnRNP A2 has subsequently been found to transport mRNAs in the dendrites of mammalian neurons (Shan et al., 2003).
Mutations in Fragile X Mental Retardation Protein (FMRP) are the cause of the most common form of X-linked mental retardation (for review, see (Bassell and Warren, 2008). FMRP is an RNA binding protein that contains two KH domains and one Arg-Gly-Gly (RGG) RNA binding domain, and binds to a number of localized transcripts, including MAP1b, PSD95, and its own mRNA. Binding is thought to occur in part through recognition by the RGG domain of a stem “G-quartet” loop in the target mRNA. FMRP is transported into dendrites in a neuronal activity-dependent manner, associates with polyribosomes and represses translation. As such, it provides another example of a trans-acting factor with dual functions in mRNA localization and translational repression. FMRP has also been shown to associate directly with kinesin, and to travel along microtubules, thereby linking bound mRNAs with the kinesin-motor pathway for transport (Dictenberg et al., 2008).
Localized mRNAs are transported in large structures containing many RNAs and proteins, often termed RNA transport granules or RNA granules. In a seminal study, Carson and colleagues visualized MBP mRNA transport in live oligodendrocytes and concluded that a population of large RNA granules, containing multiple localized transcripts, served as the vehicle for mRNA transport (Aigner et al., 1993). Later studies in neurons also indicated that mRNAs are transported by a heterogeneous population of motile RNA granules (for review see (Kiebler et al., 1999). Purification of RNA granules from neurons has revealed some of the components of these large RNPs. Krichevsky and Kosik (Krichevsky and Kosik, 2001) isolated a sucrose gradient fraction reflecting complexes larger than polysomes in cultured neurons, and showed that it contained ribosomes and the RNA binding protein Staufen. They further showed that depolarization of neurons disrupted the structure of the RNA granule, presumably freeing the mRNAs to now be translated.
Hirokawa and colleagues (Kanai et al., 2004) purified large RNA granules from mouse brain that associate with the tail of the kinesin motor protein KIF5. These RNA granules contained CamKIIα and arc mRNAs. Proteomic analysis led to the identification of proteins previously known to be involved in mRNA localization in neurons, including Staufen and FMRP, as well as new trans-acting factors involved in mRNA localization, including Pur-α, hnRNP U and polypyrimidine tract binding protein-associated splicing factor (PSF). They further showed using RNAi knockdown, that Pur-α, hnRNP U, PSF and Staufen were all required for the dendritic localization of CamKIIα mRNA.
In a separate study, Elvira et al (Elvira et al., 2006)isolated RNA granules from developing rodent cortex and used proteomic analysis to identify their components. The composition of these RNA granules differed somewhat from those identified by Kanai and colleagues (Kanai et al, 2004). They were enriched for β-actin mRNA, and not for CamKIIα mRNA and contained ribosomes, RNA binding proteins such as Staufen, hnRNP A2, as well as the DEAD box 3 helicase, which had previously been implicated in RNP assembly. Taken together, these two studies suggest that there are multiple species of RNA granules, each containing distinct populations of mRNAs and RNA binding proteins.
An early step in the formation of RNA granules may be the assembly of oligomeric RNAs and RNPs. Recent studies of oskar mRNA have been informative in this regard. Although splicing is required for oskar mRNA localization, the observation that intronless reporter RNAs bearing only the oskar 3′UTR can localize by association or “hitch-hiking” with endogenous oskar mRNA provides an in vivo demonstration of the capacity of multiple RNAs to associate in transport complexes (Hachet and Ephrussi, 2004). Polypyrimidine tract binding protein (PTB) mediates this 3′UTR-dependent oligomerization and is required for translational repression of unlocalized oskar mRNA (Besse et al., 2009). Consistent with this, in vitro analysis of oskar translational repression in oocyte extracts has shown that the mRNA can occur in a “oligomeric” state in which protein-protein interactions can lead to the assembly of large “silencing particles” (Chekulaeva et al., 2006).
RNA transport particles can also contain several different RNAs. In budding yeast, the same Myo4p/She2p/She3p complex localizes more than 20 different mRNAs to the bud tip. By labeling pairs of these RNAs in vivo with distinct tags and tag-specific fluorescent RNA binding proteins, and tracking single particles in live cells, Lange et al. (Lange et al., 2008) have shown that multiple RNAs are co-assembled into complexes and coordinately transported to their target site, conceivably allowing their co-translation and local assembly of specialized protein complexes.
Recent attention has focused on the relationship between maternal and neuronal RNA transport granules, stress granules, and RNA processing bodies (P-bodies) (Anderson and Kedersha, 2006; Kiebler and Bassell, 2006). Stress granules form in the cytoplasm of plant and mammalian cells following environmental stress. They consist of stalled ribosomal initiation complexes, mRNAs that encode most cellular proteins other than heat-shock proteins, translation initiation factors as well as a number of RNA binding proteins involved in mRNA localization, such as Staufen, FMRP, and cytoplasmic polyadenylation element binding protein (CPEB). Stress granules have been postulated to serve as a “triage” site for mRNA degradation, storage or reinitiation (Anderson and Kedersha, 2006). P-bodies, present in eukaryotic cells from yeast to mammals, are uniformly sized structures containing components of the 5′-3′ mRNA decay machinery, nonsense-mediated decay pathway and RNA-induced silencing complex. Like stress granules, P-bodies have been shown to contain RNA binding proteins that are components of RNA transport granules, including Staufen and FMRP. P-body specific markers have been detected in dendrites of Drosophila and mammalian neurons (Kiebler and Bassell, 2006). This finding, together with recent indications that the translation of transcripts localized to dendrites may be regulated by miRNAs, raises the possibility that mRNAs may undergo dynamic trafficking between RNA transport granules, P-bodies, and stress granules (Kiebler and Bassell, 2006).
The cytoplasm of most asymmetric cells can be very large—for example, the length of neuronal axons and dendrites often exceeds the diameter of the nucleus by orders of magnitude. It is thus likely that dedicated transport pathways exist to deliver mRNAs to their specific subcellular destinations within the cytoplasm. The results of genetic, biochemical and cell biological studies have indicated that microtubules and actin filament networks provide a railway for trafficking of mRNAs within the cytosol, with the microtubule motor proteins kinesin, dynein, and myosin providing the vehicle for transport along these pathways.
The localization of ASH1 mRNA to the yeast bud provides a model example of myosin-mediated mRNA transport along actin microfilaments. As described above, one of the RNA binding proteins involved in the localization of ASH1 mRNA to the bud of yeast, She3p, serves as an adaptor that links the ASH1 mRNA to the motor protein Myo4p (also called She1p). Myo4p is a type V barbed-end-directed myosin motor and has been shown to direct transport of substrates along actin microfilaments in living yeast (Reck-Peterson et al., 2001). It is a nonprocessive motor and the presence of multiple RNA elements in ASH1 for binding of the She2p-She3p-Myo4p complex has been postulated to allow for continuous transport of the ASH1 mRNP to the bud tip. Yeast mutants that prevent bundling of actin cables have been shown to result in mislocalization of ASH1 mRNA (Long et al., 1997; Takizawa et al., 1997). In addition to a role for actin in the targeting of ASH1 mRNA, actin plays a role in anchoring ASH1 mRNA at the tip (Beach et al., 1999). Thus, disruption of cortical actin at the bud tip alters ASH1 mRNA localization (Beach et al., 1999; Takizawa et al., 1997). Of note, the actin cytoskeleton is important for anchoring of a number of other messages, including, among others, β-actin mRNA in fibroblasts (Farina et al., 2003), bicoid mRNAs in Drosophila oocytes and embryos (Weil et al., 2008), Vg1 mRNA in Xenopus oocytes (Yisraeli et al., 1990) and arc mRNA in vertebrate neurons (Huang et al., 2007). An intriguing set of recent studies have shown that ASH1 mRNP also associates with the endoplasmic reticulum in a manner that does not depend on ribosome association, indicating that mRNA transport and ER trafficking may be coupled (Aronov et al., 2007; Schmid et al., 2006).
mRNA localization in other cell types has been shown to depend on a polarized microtubule cytoskeleton. In the Drosophila oocyte at mid-oogenesis, microtubules grow from the anterior and lateral cortex of the cell and show some degree of overall polarization, as indicated by the enrichment of kinesin microtubule plus-end motor protein at the posterior pole. The transport of gurken mRNA (a transforming growth factor alpha member) to the anterior-dorsal corner of the Drosophila oocyte depends on specific subsets of microtubules and the motor protein dynein, whose function switches to that of a static anchor when the RNA reaches its target site (Delanoue and Davis, 2005; Delanoue et al., 2007; MacDougall et al., 2003). Experiments in which dynein is inactivated indicate that at the same time, bicoid mRNA is transported in a dynein-dependent manner to the microtubule minus ends at the anterior of the oocyte (Duncan and Warrior, 2002; Januschke et al., 2002). However, live imaging of fluorescently tagged bicoid mRNA has revealed that, although a portion of the mRNA is localized to the anterior during mid-oogenesis, the bulk of the RNA is localized and maintained at the anterior later on, by a dynamic process involving continuous active transport of bicoid by dynein on anterior microtubules. Similarly, RNA tagging and live imaging indicates that kinesin-1 motor protein transports oskar mRNA to the posterior pole of the oocyte in a random walk on a weakly polarized cytoskeleton (Zimyanin et al., 2008). During early embryogenesis, pair-rule transcripts such as hairy are localized apically, by bidirectional transport on microtubules. Cis-acting localization elements on the mRNA dictate the number of motors associating with the mRNAs, and thereby determine the speed, frequency and duration of movement, and ultimately the localization of the mRNPs. Two proteins, Eqalitarian and Bicaudal D, appear to function as adapters that mediate the association of dynein with the localization elements (Bullock et al., 2006).
In neurons, where the lengths that mRNAs are transported are especially great, microtubules have also been demonstrated to play a critical role. As one example, studies of staufen-dependent dendritic mRNA transport have highlighted a fundamental role for microtubules (Kiebler et al., 1999; Tang et al., 2001). Hirokawa and colleagues (Kanai et al., 2004) demonstrated a role for the microtubule anterograde motor KIF5 in transporting many dendritically localized transcripts and further showed that alterations in the concentrations of KIF5 modulated the dendritic localization of RNA granules in neurons. Genetic, pharmacological and siRNA-mediated inhibition of kinesins have been shown to inhibit FMRP transport into dendrites, and have further indicated that FMRP interacts with at least two distinct kinesin isoforms, KLC (the light chain component of KIF5, Dictenberg et al., 2008) and KIF3C (Davidovic et al., 2007). The finding that FMRP can use two kinesin motors indicates that molecular motors may play redundant roles in mRNA transport. Consistent with such redundancy, a recent study from Mowry and colleagues has shown that multiple kinesins coordinate the transport of mRNAs in Xenopus oocytes (Messitt et al., 2008)
Several studies have indicated that neuronal activity modulates the transport of mRNAs into dendrites (Sossin and DesGroseillers, 2006), and it will be interesting to determine whether this modulation occurs as a result of post-translational changes in the RNA binding proteins, in the composition of RNA granules, or perhaps as modifications of microtubules or motor proteins. Kandel and colleagues have recently shown in Aplysia neurons that kinesin heavy chain isoforms are upregulated during learning-related synaptic plasticity, and show that this upregulation results in an increase in transport of essential components from the soma to the synapse (Puthanveettil et al., 2008). Alternatively, stimulus-induced post-translational modifications of tubulin may alter the association of molecular motors with microtubules, and thereby regulate the transport of mRNAs to dendrites and synapses.
mRNA localization coupled with regulated translation is emerging as a common and fundamental mechanism regulating gene expression in many cell types. Initially discovered as a means of restricting the synthesis to specific, specialized compartments of highly polarized and asymmetric cells, more recent systems-level analyses indicate that mRNA localization and regulated translation may be used more widely as a means of regulating gene expression not just within the temporal but also within the spatial dimension.
The power of this type of gene regulation is readily appreciated in neurons. Neurons are among the more dramatically polarized of all eukaryotic cells, elaborating processes that extend great distances and that form up to 10,000 synaptic connections with target neurons. During development of the brain and during plasticity of the adult nervous system, each of these synaptic compartments can undergo experience-dependent changes in structure and function. The persistence of these changes in synaptic efficacy often depends on changes in gene expression and mRNA localization and regulated translation provides a mechanism for spatially restricting this new gene expression to the synapse. This type of regulation is not, however, simply “synapse-specific,” but rather is integrated with the regulation of gene expression within the nucleus. The fact that the cytoplasmic localization of mRNAs is influenced by processing of the pre-mRNA in the nucleus provides a clear example of how nuclear events can be integrated with synaptic events. It further illustrates the elegant integration of spatial and temporal information processing that can occur within an individual cell: synaptic stimuli can trigger changes in transcription in the nucleus, the addition of RNA binding proteins during pre-mRNA processing can dictate the subsequent localization of the mRNA within the cytosol, and the transcript, after reaching its destination, can undergo translational regulation in response to local stimulation.
As a field, the study of mRNA localization is at an exciting crossroads. Studies over the past two decades have uncovered a wealth of details regarding the specific ways in which individual mRNAs localize within different cell types. From all this information, common patterns and principles are beginning to emerge. Our aim in this review was to highlight these principles while describing the details of specific findings. To briefly summarize: the localization of mRNAs is determined by cis-acting localization elements that are frequently, but not always found in the 3′UTR of the mRNA. In most cases, the primary nucleotide sequence of the localization element appears to be less important than the secondary structure, which commonly consists of stem-loop structures. Localization elements are often found in multiple copies within an mRNA, either as repetitions of similar elements, or as combinations of unique elements that often mediate distinct processes in mRNA localization. Common families of RNA binding proteins have been shown to bind to the localization elements, although the specific motifs recognized by each have not yet been fully determined. Another general emerging principle is that the nuclear history of the mRNP is critical to its ultimate localization within the cytoplasm. Thus, the RNA binding proteins that are loaded on the mRNA during transcription and nuclear mRNA processing (including during splicing) have been found to be required for the ultimate localization of the mRNA within the cytoplasm. Further, the composition of the RNP can be remodeled during trafficking. Studies in diverse systems have indicated that mRNAs are transported in large RNA transport particles or granules. An important concept is that the mRNAs are translationally repressed within these granules, and recent studies have suggested that noncoding RNAs and microRNAs may be components of these RNA granules. The mechanisms underlying the translational repression and depression also appear to involve conserved mechanisms (Besse and Ephrussi, 2008). RNA granules are transported by motor proteins (kinesins, dyneins and myosins) along microtubule or actin microfilament networks. In many cases, the mRNA is anchored at its final destination in an actin-dependent manner.
The development of new methodologies that permit high-resolution imaging of mRNA localization in fixed and in living cells promises to further elucidate the process of mRNA localization. These include high-throughput, sensitive in situ hybridization techniques, as well as new methods for tagging RNAs with fluorescently-tagged RNA binding proteins or oligonucleotides that allow imaging of mRNAs as they move within living cells. As an increasing number of localized mRNAs are identified, bioinformatic approaches aimed at identifying common cis-acting elements are likely to be much more successful. In particular, experimentally validated algorithms that incorporate secondary structure into their predictions promise to enable systems-level identification of potential new localized transcripts. New methods for detecting and perturbing translation within subcellular compartments will elucidate the regulation and the function of local translation of specific transcripts. Together, these advances will likely provide enough pieces of data to solve the puzzle of how and why mRNAs localize within cells.
We thank D. Black, D. O. Wang and L. Zipursky for critical reading of the manuscript, and E. Lecuyer and H. Krause for the image in Figure 2. Work on mRNA localization in our laboratories is supported by the NIH (KCM) and the DFG and the EMBL (AE).