RNA localization coupled to translational repression is a general mechanism for creating structural and functional asymmetry within the cell. While there are many possible ways to target an mRNA to its destination, a large fraction of the studied transcripts undertake active transport mediated by cytoskeletal elements (microtubules and actin filaments) and associated mechanoenzymes. Among the best-studied model systems of RNA localization are the oocyte and the early embryo of Drosophila melanogaster, for which many well-characterized tools have been developed to study this cell biological phenomenon in a dynamic, developing system in its in vivo context. In the present paper, we review the current evidence and models explaining the different modes of RNA localization that depend on active transport within cells.
Asymmetric distribution of protein molecules, and thus the establishment of specific functional domains within cells, is a general phenomenon that occurs in simple unicellular organisms as well as in complex multicellular entities. Local deployment of functional protein molecules is often achieved by localizing their blueprint, the encoding mRNA, to the proper subcellular locus [1–5]. As a single mRNA molecule can give rise to numerous copies of the encoded polypeptide during translation, this process can serve as an efficient means of localizing large amounts at the cost of moving only a few molecules. However, to avoid misexpression of the encoded proteins and deleterious consequences arising from their mislocalized function, the encoding mRNAs are frequently localized in a translationally repressed state [6,7]. Both translational repression and mRNA localization are mediated by proteins associated with the RNA, resulting in the formation of L-RNPs (localizing ribonucleoprotein complexes) [2,3], which often contain multiple copies of the mRNA oligomerized either by self-assembly and/or RNA–protein/protein–protein interactions [2,8,9].
There are multiple ways to localize mRNAs: if the cellular geometry and nuclear geometry are favourable, certain mRNA molecules preferentially leave the nucleus through nuclear pores positioned proximally to the destination of the RNA (vectorial RNA transfer) . However, most of the known localizing RNA molecules are not transported in this manner; rather they are exported into the cytoplasm in a non-directed fashion. Once in the cytoplasm, a combination of RNA degradation, thermal motion, facilitated diffusion, active transport and anchoring leads to the delivery of the message to its location. For instance, the Drosophila heat-shock protein 83-encoding mRNA is localized to the posterior pole of early embryos as a consequence of its local protection against de-adenylation and degradation at this site . nos (nanos) mRNA is enriched at the posterior pole of the developing Drosophila oocyte by flows generated by the synchronous movement of ooplasmic organelles (fast ooplasmic streaming) and a localized anchoring mechanism at the posterior pole [12,13]. Interestingly, the mechanism anchoring nos mRNA depends on the localization of osk (oskar) mRNA and the local expression of Oskar protein at the posterior pole of the developing oocyte. Oskar protein acts as one of the axis determinants of the future embryo: it recruits protein molecules and RNAs (including nos) that form the posterior pole ooplasm and that are necessary for the formation of the germline cells and abdominal structures in the developing embryo [5,12,14].
osk mRNA and other localizing mRNA molecules, e.g. grk (gurken)  and bcd (bicoid)  mRNAs that determine the dorsal and anterior structures of the embryo respectively, are synthesized in the nurse cells, the 15 sibling cells of the oocyte, which remain interconnected with the oocyte by ring canals during oogenesis, allowing the transport of macromolecular complexes from one cell to another (Figures 1A and 1B). It has been shown that, within the nurse cells, grk rapidly converges towards the ring canals in an microtubule-dependent fashion. Translocation of grk mRNA from the nurse cells to the oocyte depends on cytoplasmic dynein: hypomorphic mutations in the heavy chain-encoding gene (dhc) lead to a reduced velocity of grk RNP movement and, consequently, a slower accumulation of the transcript in the oocyte . When injected into developing Drosophila embryos during nuclear cycle 14 (Figure 1C), bcd, grk and osk mRNAs, as well as embryonic pair rule transcripts [e.g. hairy, ftz (fushi tarazu), even-skipped and runt] undergo bidirectional movement with net apical directionality and rapidly localize to the apical surface of the embryos, the site of microtubule nucleation [18–21]. This localization depends on the motor function of cytoplasmic dynein [19,21] and requires certain cis-acting RNA sequences and structures, the so-called LEs (localization elements) [18,22]. These elements usually reside in the 3′-UTR (untranslated region) of the mRNA, and they typically include a (putative) stem–loop structure composed of a proximal dsRNA (double-stranded RNA) stem of variable length and a distal loop . As demonstrated by Dienstbier et al. , mRNAs containing these apical targeting LEs have the capacity to interact with Egl (Egalitarian), a non-canonical RNA-binding protein. Egl has been shown to interact with dynein LC8 (light chain 8)  and with the CTD (C-terminal domain) of BicD (Bicaudal D), a phosphoprotein involved in diverse dynein-dependent processes . Thus Egl serves as an adaptor protein between cytoplasmic dynein and mRNA molecules containing apical targeting LEs. In addition, the CTD of BicD acts as a common interface for the binding of Egl and Rab6, a vesicular cargo adaptor [25,26]. Overexpression or injection of ectopic Egl protein suppresses the apical motion of lipid droplets, indicating that the two types of cargo, namely LE-containing RNAs and lipid droplets, compete for the transporter and that the binding of the adaptors is mutually exclusive .
Overview of the mRNA localization model systems, the developing Drosophila oocyte (A and B) and the syntitial blastoderm embryo in nuclear cycle 14 (C and C')
The interaction between an mRNA and Egl depends on the integrity of the LE. Small changes that affect the proximal stem region of these elements, even a single nucleotide substitution, can abolish the mRNA–Egl interaction and thus localization of the transcripts [18,22]. On the other hand, the LEs are autonomously sufficient to drive localization of heterologous mRNAs, provided that their structure remains unaltered in their new context [18,22,27]. Not surprisingly, all the aforementioned localizing mRNAs contain such (putative) stem–loop structures in their 3′-UTR regions [18,22]. However, several mRNAs that localize to the apical surface of polarized epithelial cells have no such elements in their 3′-UTR. The btsz (bitesize) and sdt (stardust) 3′-UTRs are dispensable for apical localization of the endogenous or any heterologous transcripts [27,28], and their LEs are located within the mRNA coding sequence instead. The apical localization of sdt mRNA in follicular epithelial cells was shown to require dynein as well as the presence of the alternatively spliced exon 3 in the transcript (sdt A). Furthermore, on its own, exon 3 can mediate apical localization of a heterologous, GFP (green fluorescent protein)-encoding transcript .
Although direct evidence is still lacking, it has been postulated that other L-RNPs (bcd, osk, nos, K10, etc.) use the same, dynein-based machinery for their transport from the nurse cells, into the oocyte [17,24] based on their ability to use the cytoplasmic dynein–Egl–BicD machinery to localize apically in the developing embryo [18,19]. Indeed, grk and bcd RNPs localize to the oocyte anterior pole, which in terms of polarity is homologous with the apical membrane of epithelial cells [29,30], considered to be a site of microtubule nucleation . As expected, grk and bcd localization is compromised when dynein-mediated transport is impaired by the use of hypomorphic Dhc alleles, overexpression of p50/dynamitin (which disrupts the integrity of the dynactin complex) or injection of function-inhibitory antibodies targeting Dhc or Egl [32–34]. Weil et al. proposed a ‘continual active transport’ model for bcd mRNA localization during late oogenesis, whereby bcd mRNA is effectively localized by repeated rounds of dynein-mediated transport towards the oocyte anterior, from which it is free to diffuse away. This dynamic localization process remains active  until the very end of oogenesis, when bcd mRNA finally gets anchored at the anterior in an actin-dependent manner [35,36]. In contrast, grk mRNA, which localizes to the dorso-anterior corner of the developing oocyte around the nucleus, is not only transported, but also statically anchored by cytoplasmic dynein. Injection of anti-Dhc antibody releases the localized grk mRNA into the cytoplasm, whereas inhibition of Egl blocks the transport process without affecting mRNA anchoring . Similarly, during the localization of wingless, runt and ftz transcripts in the developing embryo, dynein not only transports but also anchors these transcripts, even when the nucleotide cycle of the motor is suspended by vanadate injection . Interestingly, on impairment of the function of the hnRNP (heterogeneous nuclear ribonucleoprotein) Squid, the grk localization process becomes dynamic, similarly to the ‘continual active transport’ reported for bcd mRNA. Simultaneously, the distribution of grk mRNA broadens along the anterior cortex [37,38]. Squid is a component of the grk mRNP and is also necessary for the integrity of the sponge bodies (homologues of RNA processing P bodies in Drosophila), which anchor grk in the dorso-anterior corner of the oocyte [37,38]. Thus these differences between the diverse anchoring processes might explain the different subcellular localizations of RNPs utilizing the same, dynein-mediated transport machinery.
During oogenesis, other localizing mRNAs head from the nurse cells to the opposite pole of the developing oocyte and accumulate along the posterior cortex, the functional analogue of the basal membrane of epithelial cells . Localization of nos mRNA, as mentioned previously, relies on a selective anchoring mechanism established by Oskar protein localized at the posterior pole  and is facilitated by the fast ooplasmic streaming promoted by KHC (kinesin heavy chain) that occurs late in oogenesis . osk mRNA localization also depends on conventional kinesin [13,39–41]. During early oogenesis osk mRNA is enriched and distributed homogeneously in the oocyte, similarly to bcd and grk mRNAs. However, when bcd and grk move towards the anterior, osk mRNA accumulates first in the oocyte centre and then at the oocyte posterior cortex. Both of these localization steps depend on KHC [41,42]. Zimyanin et al.  showed that the posterior localization of osk is mainly achieved by KHC-mediated active transport along microtubules, in a weakly biased random walk towards the posterior pole of the oocyte. On reaching its destination, osk mRNA is anchored via the actin cytoskeleton, myosin V and the newly translated Oskar protein to the posterior plasma membrane [43–45].
Recent evidence suggests that osk RNPs are bound to the KHC motor via the cargo adaptor KLC (kinesin light chain) and Pat-1, an amyloid precursor protein-binding protein in flies homologous with conventional KLC, in a redundant manner , rather than by the RNA adaptor protein FMRP (fragile X mental retardation protein), which in neurons is a mediator of KHC-dependent anterograde mRNP transport . The finding that KLCs are involved in connecting KHC to an RNP cargo is novel and the underlying mechanism has not yet been established. A tempting hypothesis is that conventional kinesin is bound to the cargo by cytoplasmic dynein, which is considered to transport osk from the nurse cells to the oocyte. Ligon et al.  have demonstrated that rat KLCs interact directly with DIC (dynein intermediate chain), an obligate component of the dynein holoenzyme (Figure 2). osk RNPs, on the other hand, seem to be capable to recruit dynein, similarly to grk, bcd and localizing pair rule transcripts , and indirect evidence suggests that this interaction is preserved within the oocyte. Under specific circumstances, such as in the presence of dominant alleles of BicD or overexpression of the RNA targeting protein, Miranda, part of the osk RNP pool, is targeted to the anterior of the developing oocyte where microtubule minus ends are located , giving rise to ectopic abdominal structures and the so-called bicaudal embryos [5,48]. Although the molecular mechanism through which these mutations act is poorly understood, they suggest the existence of a silenced dynein transport machinery associated with osk RNPs in the oocyte.
mRNA interaction with the transport machinery
Another intriguing feature of osk mRNA localization is that it depends on the splicing position of the first intron in the mRNA. In the absence of this particular event, which takes place in the nurse cell nuclei, the mRNA enters the oocyte, where it localizes ectopically around the anterior and lateral cortex and fails to enrich at the posterior pole . Similar phenotypes are observed in mutant oocytes of the EJC (exon junction complex) core components, eIF4AIII (eukaryotic initiation factor 4AIII), Y14, Mago nashi and Barentsz . The requirement for splicing distinguishes osk mRNA from other localizing transcripts, and although the proper molecular mechanisms are unknown, it might serve as a mark for kinesin recruitment and/or of silencing of the dynein transport machinery. Whether osk is unique indeed or whether it is the first member of an RNA family that reaches its destination via switching motors is yet to be determined. Nevertheless, the specific features of osk mRNPs, especially their presumed ability to recruit opposite polarity motors, make them an excellent model to study how RNPs reorganize to switch and co-ordinate motor functions in time and space and more importantly to learn the intimate relationship and bases of interdependence of opposite polarity mechanoenzymes within a developing system.
Cellular Cytoskeletal Motor Proteins: A Biochemical Society/Wellcome Trust Focused Meeting held at Wellcome Trust Genome Campus, Hinxton, Cambridge, U.K., 30 March–1 April 2011. Organized and Edited by Folma Buss (Cambridge, U.K.) and John Kendrick-Jones (MRC Laboratory of Molecular Biology, Cambridge, U.K.).
dynein: hypomorphic mutations in the heavy chain-encoding gene
dynein intermediate chain
fragile X mental retardation protein
kinesin heavy chain
kinesin light chain
light chain 8
localizing ribonucleoprotein complex
I thank my colleagues Virginie Marchand and Sanjay Ghosh and my supervisor Anne Ephrussi for their help and critical comments on the paper.
I.G. received support from the EMBL Interdisciplinary Fellowship (EIPOD) and the EMBO Long-term Fellowship [grant number EMBO ALTF 773-2009].