RNA localization mechanisms in oocytes

The early stages of embryogenesis are dependent on the proper spatial and temporal distribution within the oocyte (and ultimately within the egg) of the vast stockpile of maternal molecules accrued during the elaborate and lengthy process of oogenesis. Among these molecules, various localized RNAs are crucial players involved in the generation of oocyte polarity and/or patterning of the embryo. In addition, certain RNAs localized in germinal granules (also know as P-granules or polar granules) of the oocyte are involved in the specification of germ cell fate. Some of these localized RNAs are noncoding structural or regulatory RNAs, whereas others become translated into localized proteins. Here, we discuss recent work that has provided insight into the multifaceted phenomenon of RNA localization in oocytes.

In many invertebrates and vertebrates, the position of the anterior-posterior and dorsal-ventral axes of the embryo and the formation of the germ line are controlled by the asymmetric localization of various RNAs (and/or their translation products) in the oocyte (Hashimoto et al., 2004; Jansen, 2001; King et al., 1999; Kloc et al., 2001; Kloc et al., 2002a; Palacios and Johnston, 2001; Riechmann and Ephrussi, 2001; Zhou and King, 2004). In Drosophila, the localization of nanos and oskar mRNAs to the posterior pole of the oocyte, and bicoid RNA to the anterior pole, plays a role in the formation of the anterior-posterior body axis, pole plasm assembly and germ cell function. Mislocalization of oskar RNA in females that have a fourfold increase in osk gene dosage or in females bearing the osk bcd 3′UTR transgene (which contains the bicoid anterior localization signal) results in misexpression of Oskar protein, which in turn causes mislocalization of Nanos protein and repression of bicoid. This results in a bicaudal embryo, its anterior replaced by a mirror image of the posterior (Ephrussi and Lehmann, 1992; Kim-Ha et al., 1991; Smith et al., 1992; Wharton and Struhl, 1991). Another localized mRNA, gurken, signals to the epithelial cells surrounding the oocyte for the proper formation of the dorsal-ventral axis (Gonzales-Reyes et al., 1995; Neuman-Silberberg and Shüpbach, 1993; Nilson and Shüpbach, 1999).

In ascidians, the localized mRNAs macho-1 and pem determine muscle fate and anterior and dorsal patterning, respectively (Nishida and Sawada, 2001; Yoshida et al., 1996). In sea urchin, the frog Xenopus laevis and zebrafish, subsets of RNAs are also localized to the animal and vegetal poles of the oocyte (Braat et al., 1999; Bruce et al., 2003; Di Carlo et al., 2004; Howley and Ho, 2000; King et al., 1999; Kloc et al., 2001; Kloc et al., 2002a). In Xenopus oocytes, the vegetally localized mRNAs Vg1, VegT and Xwnt 11 are mesodermal and endodermal determinants and determinants of the left-right axis in the embryo (Hyatt and Yost, 1998; Joseph and Melton, 1998; Ku and Melton, 1993; Rebagliati et al., 1985; Thomsen and Melton, 1993; Zhang et al., 1998; Zhang and King, 1996). Other vegetally localized mRNAs in Xenopus, such as Xcat2 and Xdazl, are believed to play a role in germ cell determination or migration (Houston et al., 1998; Mosquera et al., 1993). In zebrafish, several mRNAs are localized either to the animal or vegetal pole of the oocyte (Howley and Ho, 2000). Two of them: the mRNA encoding the DEAD-box RNA helicase Vasa and the mRNA encoding the T-box gene product eomesodermin are cortically localized in vitellogenic oocytes, and might play a role in germ cell determination and the specification of the organizer, respectively (Bruce et al., 2003; Knaut et al., 2000; Knaut et al., 2002).

RNA is exported from the oocyte nucleus through nuclear pores. Once in the cytoplasm, RNA can be trapped by or targeted to specialized transporting organelles or bind to cytoskeletal elements that assist in transport of RNA to its ultimate destination. Alternatively, asymmetrical distribution can be achieved through a combination of RNA destabilization/degradation and local protection (Tadros et al., 2003). In certain animals, the oocyte nucleus is transcriptionally quiescent and most RNAs are synthesized in the accessory cells (called nurse cells in insects), from which they are exported to the oocyte. In this case, independent mechanisms are often responsible for RNA transport to and within the oocyte.

Two pathways of RNA localization to the oocyte vegetal cortex have been identified in Xenopus - the early pathway and the late pathway (Fig. 1). The early pathway, also known as the messenger transport organizer (METRO) pathway, functions in early oogenesis (stages I and II) and transports germinal granules, various RNAs (some of them involved in germ cell specification), and germline-specific mitochondria in a specialized organelle that in Xenopus is called the mitochondrial cloud (also known as the Balbiani body) (Heasman et al., 1984; Forristall et al., 1995; Kloc and Etkin, 1995) (Figs 1 and 2). The Balbiani body was discovered by von Wittich in 1845 in oocytes of spiders. Between 1864-1893, Balbiani conducted comprehensive studies of this organelle in oocytes of myriapods and spiders; later, this organelle was discovered in oocytes of various invertebrates and vertebrates, including humans (de Smedt et al., 2000; Guraya, 1979; Heasman et al., 1984; Kloc et al., 2004b). Interestingly, the existence of the Balbiani body in the oocytes of marsupials and eutherian mammals (including nonhuman primates and humans) is not widely known. This situation arises primarily from the fact that the mouse, the dominant model organism in mammalian studies, does not have a Balbiani body in its oocytes. In this respect, the mouse is probably an exception among mammals. However, as a consequence of its absence in the mouse, there is not a single study that experimentally addresses the role of the Balbiani body in mammalian oocytes.

The Balbiani body is a spherical membraneless structure that forms in contact with the oocyte nucleus in early oogenesis, and fragments and disperses in late oogenesis (Figs 1 and 2). Its constant components are mitochondria, endoplasmic reticulum (ER), membranous vesicles and lipid droplets. In Xenopus stage I oocytes, the Balbiani body is ∼40 μm in diameter, contains half a million mitochondria, which differ in morphology and metabolism from those of cytoplasmic mitochondria, and is rich in membranous vesicles, and ER cysternae. The vegetal apex of the Xenopus Balbiani body (called the METRO region) contains germinal granules and various localized RNAs (Fig. 2) (Forristall et al., 1995; Heasman et al., 1984; Kloc et al., 2004b). For nearly 150 years, the role of the Balbiani body remained a complete mystery, although on the basis of its biochemical and ultrastrucural composition, it was commonly believed that this organelle participates either in the formation of lipids or in the multiplication of mitochondria (Guraya, 1979). In 1993, Kloc et al. showed that the Balbiani body/mitochondrial cloud in Xenopus is a vehicle that transports localized RNA to the vegetal pole of the oocyte.

Cis-acting elements directing RNA to the mitochondrial cloud (mitochondrial cloud localization elements, MCLEs) have been identified in the 3′UTR of Xcat2 mRNA and in the noncoding Xlsirts RNA in Xenopus oocytes (Allen et al., 2003; Kloc et al., 1993; Zhou and King, 1996). Xcat2 mRNA also has a germinal granule localization element (GGLE) in its 3′UTR (Kloc et al., 2000). Recently, Claussen et al. identified a 300-nucleotide element in the 5′UTR of XNIF mRNA that directs it to the mitochondrial cloud (Claussen et al., 2004). The localization of RNAs into the mitochondrial cloud is independent of microfilaments and MTs and probably involves a diffusion and entrapment mechanism (Kloc et al., 1996; Chang et al., 2004). The movement of the mitochondrial cloud to the vegetal cortex of the oocyte is also independent of MTs and microfilaments, and the mechanism underlying this movement remains a mystery; one possibility is passive translocation along the cytoplasm streaming towards the vegetal cortex (Kloc et al., 2004b).

The Balbiani body has recently been discovered in the oocytes of several insects (Fig. 2) (Bradley et al., 2001; Cox and Spradling, 2003; Jaglarz et al., 2003; Kloc et al., 2004b). In Drosophila, it contains germline-specific mitochondria, forms in the oocyte in early oogenesis, and might facilitate localization of oskar mRNA to the posterior pole of the oocyte (Cox and Spradling, 2003). The Balbiani body differs from Drosophila sponge bodies, which were previously thought to be equivalent to the Xenopus Balbiani body. The sponge bodies form at the surface of nurse cell nuclei later in oogenesis, associate with Exuperantia (Exu), a novel protein that is a core component of a protein complex involved in the localization of mRNAs within the nurse cells and oocyte, and mediate transport of bicoid RNA (Wilsch-Brauninger et al., 1997).

Another ooplasmic organelle involved in the localization of RNAs in oocytes is the ER. In ascidians, the maternal mRNAs pem and macho-1 are localized in the egg on the network of cortical rough ER, which compacts after fertilization to form the centrosome-attracting body (CAB), which is responsible for the unequal cleavages (Sardet et al., 2003). Recently, Chang et al. showed that, in Xenopus, the ER vesicles are involved in the entrapment of Xcat2 and Xdazl mRNA within the mitochondrial cloud (Chang et al., 2004). Since these experiments used injected synthetic fluorescent RNA synthesized from the Xcat2 3′UTR and Xdazl localization element, it remains to be seen whether the same mechanism is involved in the localization of endogenous Xcat2 and Xdazl mRNAs to the mitochondrial cloud.

Oocytes have a highly polarized system of MTs and molecular motors that are responsible for the proper localization of most localized RNAs in Drosophila and of late-pathway (also known as the Vg1 pathway) RNAs in Xenopus (Fig. 1).

In typical somatic cells, MT minus-ends are arranged around a MT-organizing center (MTOC) close to the nucleus, and MT plus-ends occupy the cell periphery. This polarity is recognized by molecular motors, which are typically unidirectional - either minus-end directed or plus-end directed (Cohen, 2002; Lopez de Heredia and Jansen, 2004; Welte, 2004). However, traffic along the MTs in oocytes is much more complex than previously thought. As a consequence, different experiments on the same subject often yield contradictory results and conclusions, depending on the method used by the researchers (Stephenson, 2004). Thus, some of the original papers and even some of the review articles on MT trafficking are extremely confusing.

Injection of exogenous RNAs and analysis of mutants indicate that, in Drosophila, oskar mRNA is localized to the posterior pole by plus-end-directed motors (kinesins), whereas bicoid mRNA is localized to the anterior pole and gurken mRNA is localized to the anterior-dorsal region by minus-end-directed motors (dyneins) (Brendza et al., 2000; Cohen, 2002; Lopez de Heredia and Jansen, 2004). These findings raise a fascinating question: how can the same molecular motor direct bicoid and gurken RNAs to different cellular locations? Recently, MacDougall et al. used time-lapse photography to track the localization of fluorescent gurken RNA in living Drosophila oocytes (MacDougall et al., 2003). They showed that the movement of gurken RNA is dependent on dynein and MTs but occurs in two distinct steps using distinct arrays of MTs. First, gurken particles move to the oocyte anterior; later, they turn dorsally towards the nucleus. The authors suggested that, in this instance, as well as in others, the use of distinct MT networks might be responsible for transport by the same motor of different cargos to different destinations. Recent experiments showing that kinesin heavy chain (Khc) mutants that exhibit impaired oskar localization also, unexpectedly, exhibit impaired localization of gurken and bicoid mRNA indicate that anterior and anterior-dorsal transport also require kinesin (Duncan and Warrior, 2002; Welte, 2004). Several models for cooperation between minus-end-directed and plus-end-directed motors in the oocyte have been proposed. For example, kinesin might indirectly activate dynein (through an adaptor protein) or these two motors might cooperate by recycling one another - i.e. kinesin could transport dynein to MT plus-ends, where dynein would bind RNA cargo and transport it to the minus-ends, and vice versa (Cohen, 2002; Duncan and Warrior, 2002; Januschke et al., 2002; Welte, 2004).

Further complications arise because the polarity of the Drosophila oocyte cytoskeleton changes during oogenesis (which is divided into 14 stages) (Theurkauf et al., 1992). Just after formation of the oocyte (stage 1), the MTOC assembles at the anterior pole of the oocyte and MTs extend from it to the nurse cells. The oocyte then grows by acquiring molecules and organelles (such as centrosomes and mitochondria) from the nurse cells by actin-based and minus-end-directed, MT-based transport (Swan et al., 1999). In early oogenesis (stages 2-6), the MTOC shifts from the anterior to the posterior pole, resulting in localization of oskar mRNA and Gurken protein as a crescent at the oocyte posterior. Subsequently, Gurken signals to overlying follicle cells and they adopt a posterior fate. The genes required for the MTOC shift from the anterior to the posterior pole in early oogenesis include the armitage gene, which encodes a homolog of SDE3, an RNA helicase involved in post-transcriptional gene silencing, and probably other genes whose products are involved in RNA silencing (see below), such as spindle-E, aubergine and maelstrom. In armitage mutants, the MTOC does not switch to the posterior pole, which impairs posterior localization of oskar mRNA and causes its premature translation (Chekulaeva and Ephrussi, 2004; Cook et al., 2004; Tomari et al., 2004). It is possible that this defect in MOTC switching is caused not by the mutation in armitage but, directly or indirectly, by the premature synthesis of Oskar protein. However, the fact that oocytes overexpressing Oskar from the transgene have normal MT organization (Riechmann et al., 2002) makes it very unlikely that MT polarization and oskar silencing are co-dependent.

At the beginning (stage 7) of mid oogenesis (stages 7-10), the posterior follicular cells send an unknown signal back to the oocyte, causing a re-polarization of the oocyte MTs; the MTOC at the oocyte posterior disassembles, and new MTs are nucleated from the anterior and lateral oocyte cortex (Schnorrer et al., 2002). At this stage, localized oskar mRNA becomes translated at the posterior of the oocyte, bicoid mRNA is localized at the oocyte anterior, and gurken mRNA and protein become located at the anterior-dorsal corner of the oocyte, which sets up the final polarity of the oocyte.

The polarization of the Drosophila cytoskeleton seems to require more than just a simple switch in the position of the MTOC; it also depends on the partitioning-defective (PAR) proteins (Huynh and St Johnston, 2004). Evolutionary conserved PAR proteins are involved in the establishment and maintenance of cell polarity during development and in somatic cells in various species. These proteins include PAR-1, a serine/threonine kinase; PAR-2, a RING finger protein; PAR-3, which has a PDZ domain [PDZ domains are homologous in diverse signaling proteins including postsynaptic density protein-95 (PSD-95), a protein involved in signaling at the post-synaptic structures; DLG, the Drosophila Discs Large protein; and ZO-1, the zonula occludens 1 protein]; PAR-4, a homolog of mammalian kinase LKB1; and PAR-5, a homolog of 14-3-3 (which are highly conserved proteins involved in the regulation of protein phosphorylation and mitogen-activated protein kinase pathways). In Drosophila oocytes, PAR-1 regulates posterior patterning through phosphorylation and thus stabilization of Oskar (Riechmann et al., 2002) and, together with other PAR proteins, PAR-1 might control the dynamics of MT assembly. In par mutants, the MT minus-ends do not switch from the anterior to the posterior oocyte cortex in early oogenesis (Vaccari and Ephrussi, 2002), and there is a disorganization of the MT cytoskeleton in mid oogenesis in which MT minus-ends are located around the oocyte cortex and MT plus-ends are directed towards the oocyte center (Benton et al., 2002; Shulman et al., 2000). Although the molecular targets and mode of action of PAR proteins within the oocyte cytoskeleton are unknown, it is possible that PAR kinases, like the mammalian PAR-1 homolog, MARK, regulate MT stability by phosphorylation of MT-associated proteins (Benton et al., 2002; Drewes et al., 1997).

In Xenopus oocytes, several mRNAs, including Vg1 mRNA, localize to the vegetal cortex during mid and late oogenesis (oocyte stages III-V) via the MT-dependent, late (or Vg1) pathway (Fig. 1) (Kloc and Etkin, 1995; Melton, 1987; Weeks and Melton, 1987; Yisraeli et al., 1990; Zhou and King, 1996; Zhou and King, 2004). The MT-dependent localization of Vg1 mRNA starts with the `streaming' of RNA towards the vegetal cortex around the so-called wedge region of the oocyte, which contains a subdomain of ER organized on top of the mitochondrial cloud. Either the ER serves as a matrix for the formation of polarized MT tracks that transport Vg1 mRNA, or Vg1 binds to the ER vesicles (possibly through the Vg1RBP/Vera protein), and the ER-Vg1 mRNA complexes are transported on the MTs to the oocyte vegetal cortex (Deshler et al., 1997; Kloc and Etkin, 1998).

There is some evidence that the movement of Vg1 mRNA to the Xenopus oocyte vegetal cortex is mediated by the plus-end-directed motor kinesin II (Betley et al., 2004) and possibly by kinesin I (Yoon and Mowry, 2004). Paradoxically, the majority (around 96%) of MTs in Xenopus stage III and VI oocytes have their minus-ends directed towards the cortex, which is the reverse of what is typical for eukaryotic cells (Pfeiffer and Gard, 1999). The majority of the molecules transported to the cortex should therefore use minus-end-directed motors, such as cytoplasmic dyneins or minus-end-directed kinesins. Possible explanations for these apparently contradictory findings (i.e. that movement of Vg1 to the vegetal cortex is mediated by plus-end-directed motors but that most MTs have minus-ends directed towards the cortex) are that, in Xenopus, as in Drosophila, the polarity of the MTs shifts during oogenesis and/or that the RNA-binding factors decide the directionality of the RNA transport. Rat neurons provide an example of this: the ribonucleoprotein granules of the RNA-binding protein Staufen fused to green fluorescent protein (GFP) are able to associate with both kinesin and dynein, and occasionally reverse the direction of their movement on MTs (Köhrmann et al., 1999; Villace et al., 2004). Although we do not understand the molecular events that lead to the changes in the direction of movement along MTs, the novel Halo-like proteins discovered in mutants that exhibit defective transport of lipid droplets might be involved. Halo proteins might affect the duration of the particular run of the cargo by altering the binding strength between a particular motor and MTs, which would change the frequency of dissociation of motor proteins from MTs, or by controlling the number of motors that bind to MTs. Another possibility is that Halo proteins act in trans, changing the net direction of the cargo translocation: in the presence of Halo, the net movement of the cargo is plus-end directed whereas, in the absence of Halo, the net movement of cargo is minus-end directed (Fig. 3) (Gross et al., 2003; Cohen, 2003; Welte, 2004).

RNA localization via MTs depends on the presence of cis-acting localization elements (LEs) located in the 3′UTR (common), the 5′UTR (rare), or both (Table 1) (Betley et al., 2002; Bubunenko et al., 2002; Claussen et al., 2004; Kloc et al., 2002a; Sasakura and Makabe, 2002; Zhou and King, 1996). Proper secondary structure is often important for the functioning of LEs, and several LEs can act redundantly. The best characterized are those of: Vg1, VegT, fatvg, Xvelo1 and XNIF mRNAs and noncoding Xlsirts RNA in Xenopus (Chan et al., 1999; Claussen and Pieler, 2004; Claussen et al., 2004; Kloc et al., 1993; Kwon et al., 2002; Mowry and Cote, 1999; Yaniv and Yisraeli, 2001); oskar, nanos, gurken and bicoid mRNAs in Drosophila (Bergsten and Gavis, 1999; Brunel and Ehresmann, 2004; Bullock and Ish-Horowicz, 2001; Crucs et al., 2000; Kim-Ha et al., 1993; Macdonald and Kerr, 1997; Macdonald and Kerr, 1998; Macdonald and Struhl, 1988; Thio et al., 2000; Wagner et al., 2004); and a group of maternal mRNAs in zebrafish and ascidians (Knaut et al., 2002; Sasakura and Makabe, 2002).

Multiple RNA-binding proteins that interact with these elements, and thus link RNA, MTs and molecular motors, have been identified (Table 1), primarily in Drosophila and Xenopus. For example, Staufen, Mago nashi, Y14 (also known as Tsunagi), Barentsz (Btz) and Hrp48 are components of the oskar mRNA transport complex (Hachet and Ephrussi, 2001; Huynh et al., 2004; Mohr et al., 2001; St Johnston et al., 1991; van Eeden et al., 2001), and Egalitarian (Egl) and Bicaudal D (BicD) are components of the dynein motor complex in Drosophila oocytes (Navarro et al., 2004). Staufen, a double-stranded-RNA-binding protein is essential for the localization of oskar and bicoid mRNA in oocytes and prospero mRNA in neuroblasts (Li et al., 1997; Zhou and King, 2004). Mago nashi and Y14, a putative RNA-binding protein, are highly conserved from Schizosaccharomyces pombe to humans, and in Drosophila they are required for the correct localization of oskar mRNA at the posterior pole. Mago nashi and Y14 interact both in vitro and in vivo. They also interact with the mRNA export factor TAP, and are associated with mRNA during its export from the nucleus to the cytoplasm (Hachet and Ephrussi, 2001; Le Hir et al., 2001a; Le Hir et al., 2001b). Btz, which shares some sequence similarity with the human nucleo-cytoplasmic shuttling protein malignant lymph node 51 (MLN51), is essential for the localization of oskar mRNA to the posterior pole of the Drosophila oocyte. When expressed in Drosophila oocytes, the mammalian homolog of Btz interacts with mammalian Staufen in an RNA-dependent manner (Macchi et al., 2003; van Eeden et al., 2001).

Hrp48 is a member of the heterogeneous nuclear ribonucleoprotein A/B (hnRNPA/B) family of RNA-binding proteins that bind 5′ and 3′ regulatory regions of oskar mRNA. Because the Hrp48 mutant exhibits defective formation of GFP-Staufen particles, it is believed that HRP48 plays a role in the assembly of Staufen-oskar mRNA transport particles (Huynch et al., 2004). BicD is essential for the organization and maintenance of the polarized MT network and the structural integrity of Drosophila oocytes and nurse cells. BicD protein contains four heptad-repeat domains typical of intermediate filament proteins and might be an integral part of the cytoskeleton (Oh and Steward, 2001). BicD forms a complex with Egl, which interacts directly with dynein light chain (Dlc) through an Egl domain distinct from that responsible for the binding to BicD (Navarro et al., 2004).

In Xenopus, Vg1RBP (also known as Vera) binds to various subelements of LEs in Vg1, VegT and Xvelo1 mRNAs (Claussen and Pieler, 2004; Git and Standart, 2002; Kwon et al., 2002). The Xenopus homolog of Drosophila Staufen associates with kinesin I and the vegetally localized RNAs Vg1 and VegT in oocytes. The function of the Vg1 localization element (VLE) in VglRNA is blocked by the expression in oocytes of dominant-negative XStau234 (a mutant form of XStau containing only a double-stranded-RNA-binding domain) (Yoon and Mowry, 2004). The presence of Staufen in RNA-transporting complexes in Drosophila and Xenopus indicates common elements in localization mechanisms in invertebrates and vertebrates.

Another example of the commonality of RNA localization mechanisms between different organisms comes from the study of zebrafish Vasa, whose RNA is a component of the germ plasm in the embryo. In zebrafish stage I oocytes, vasa mRNA is uniformly distributed within the cytoplasm. In stage II oocytes, it is localized to the cortical cup; in stage III oocytes, it is enriched in the animal cortex. In late-stage oocytes (starting from stage IV), vasa mRNA remains cortical and is localized to germ plasm granules. In cleaving embryos, vasa mRNA segregates asymmetrically to the germ plasm and subsequently to the founder population of primordial germ cell (PGCs). Although nothing is known about the mechanisms responsible for the localization of vasa mRNA in zebrafish oocytes, the asymmetrical segregation of vasa mRNA to the germ plasm in the embryos depends on intact MTs and also on the presence of the product of the maternal-effect gene nebel (Howley and Ho, 2000; Knaut et al., 2000). Recently, Knaut et al. showed that when vasa mRNA 3′UTR (which directs its localization to the germ plasm in zebrafish embryos) is fused to a GFP reporter and injected into Xenopus oocytes, it directs reporter to Xenopus germ plasm (Knaut et al., 2002). Although there is no obvious similarity between the 3′UTR of zebrafish vasa mRNA and the 3′UTR of the Xenopus vasa-like DEADSouth mRNA, both are able to direct transcripts to germ plasm in either species (Knaut et al., 2002). The experiments discussed above show not only the universality of the localization mechanisms between oocytes and embryos and between different species, but also the intrinsic interspecies similarity of the germ plasm.

Asymmetrically localized RNAs in the oocyte must be anchored at their final destinations. RNA can be transported directly to its anchor, it can be captured as it diffuses by, or a combination of these two mechanisms can be involved. Drosophila oocytes contain an actin-filament-based anchor at the posterior pole that requires the F-actin-binding protein Bifocal, tropomyosin II, Dmoesin and Cap for proper organization (Babu et al., 2004; Baum, 2002). Dmoesin is required for the crosslinking of F-actin to the membrane of the posterior pole. In Dmoesin-deficient flies, which have defects in localization of oskar mRNA, the actin filament network is loose and detached from the plasma membrane (Jankovics et al., 2002; Polesello et al., 2002). Recently, Cha et al., studying the role of Khc in oskar mRNA localization in Drosophila, re-evaluated the distribution and polarity of the MTs in the Drosophila oocyte, and the role of the posterior anchor in the localization of oskar mRNA (Cha et al., 2002). They concluded that, during late oogenesis, the MT minus-ends are uniformly distributed over the oocyte cortex, the lowest MT density being at the posterior. In stage 8, the plus-end-directed movement mediated by kinesin transports oskar mRNA towards the oocyte interior, preventing its association with unwanted cortical sites. In late stage 8 and early stage 9, the majority of MTs at the posterior pole depolymerize, unmasking (making available) the cortical actin anchor that tethers oskar mRNA (Fig. 4). Entrapment at the posterior pole also mediates localization of nanos mRNA in late oogenesis and possibly also cyclin B, gcl (germ-cell-less) and pgc (noncoding polar granule component) RNAs (Dalby and Glover, 1992; Forrest and Gavis, 2003; Jongens et al., 1992; Nakamura et al., 1996). Similarly, the microfilament-rich oocyte cortex is necessary for the anchoring of vegetally localized RNAs in Xenopus oocytes (Elinson et al., 1993; Forristall et al., 1995; Kloc and Etkin, 1995; Klymkowsky et al., 1991; Pondel and King, 1988; Yisraeli et al., 1990). The disruption of actin and/or cytokeratin filaments releases late-pathway RNAs from the vegetal cortex (Alarcon, and Elinson, 2001). Interestingly, two vegetally localized RNAs - early-pathway noncoding Xlsirts and late-pathway VegT mRNA - are necessary for the anchoring of other late-pathway RNAs to the oocyte cortex (Heasman et al., 2001; Kloc and Etkin, 1994). Indeed, we and others have shown that Xlsirts and VegT function as structural RNAs in the organization of the cytokeratin cytoskeleton at the vegetal cortex of Xenopus oocytes (M.K., K. Wilk, M. Bilinski and L.D.E., unpublished).

The latest theme in the RNA localization saga concerns how the events preceding the export of RNA into the oocyte cytoplasm - either events in the oocyte nucleus or events in the nuclei and cytoplasm of the accessory/nurse cells - determine the fate of RNA, including its stability, translatability, mode of transport and destination. The most striking finding is the discovery that the polarity of RNA transport depends on the source of RNA and its binding partners. Fluorescent bicoid RNA injected directly into Drosophila oocyte cytoplasm moves on MTs to the closest cortical surface. However, if first injected into the nurse cell cytoplasm and then withdrawn and re-injected into the oocyte, the bicoid RNA moves to the anterior. Transport to the proper destination depends on the ability of Exu protein and bicoid mRNA to be assembled into complexes in nurse cell cytoplasm. Moreover, this finding indicates that the direction of transport along the oocyte microtubular system in Drosophila depends on the identity of the proteins bound to the RNA cargo (Cha et al., 2001).

Drosophila oskar provides another example. The hybrid RNA lacZ/oskar-3′UTR is efficiently transported to the posterior pole of the oocyte (Kim-Ha et al., 1993), which implies that the oskar RNA 3′UTR contains all signals necessary and sufficient to direct RNA to the posterior pole. However, the hybrid RNA cannot move to the posterior pole in the absence of endogenous, properly spliced oskar mRNA. This leads to the conclusion that the hybrid molecule piggybacks on the endogenous oskar mRNA and suggests that the oskar RNA 3′UTR, although necessary, is not sufficient for transport to the posterior pole and that the splicing events in the nucleus regulate cytoplasmic localization of mRNA (Hachet and Ephrussi, 2004).

Splicing not only removes introns from RNA, but also deposits a stable protein complex called the exon-exon junction complex (EJC) upstream of mRNA exon-exon junctions. In Drosophila and Xenopus, the EJC, besides containing proteins directly involved in splicing, contains the Y14-Mago heterodimer, which binds to the nucleo-cytoplasmic shuttling complex TAP/p15, which transports spliced RNA through the nuclear pores. In addition, the EJC serves as an anchoring point for the factors responsible for the nonsense-mediated decay (NMD) of mRNA containing premature termination codons (le Hir et al., 2001a; le Hir et al., 2001b; Wagner and Lykke-Andersen, 2002). Hachet and Ephrussi showed that the Drosophila EJC, deposited on oskar mRNA in a splicing-dependent manner in the nucleus, is necessary for the oskar mRNA 3′UTR to assemble into a functional localization complex (Hachet and Ephrussi, 2004). The authors suggested that the first EJC landmark on oskar mRNA mediates subsequent interactions with other factors and specifies the structure of the oskar mRNA localization complex.

In Xenopus oocytes, the fate of localized RNA also probably depends on its nuclear or cytoplasmic history. When synthetic Xcat2 mRNA is injected into the nucleus of stage I or early stage II Xenopus oocytes, it behaves like endogenous Xcat2 mRNA, following the mitochondrial cloud pathway and entering germinal granules (Fig. 5 and Table 1) (Kloc et al., 1996; Kloc et al., 2000). However, when synthetic Xcat2 RNA is injected into the nuclei or the cytoplasm of older oocytes (stage III/IV), it follows the MT-dependent Vg1 pathway and probably uses a kinesin motor (Fig. 5 and Table 1) (Betley et al., 2004; Kloc et al., 1996; Zhou and King, 1996). Similarly, when an early-pathway XNIF mRNA is injected into stage I oocytes it moves to the mitochondrial cloud, but when the same RNA is injected into stage III oocytes it follows the late (Vg1) pathway, and its LE binds to the same proteins that bind the Vg1LE (Table 1) (Claussen et al., 2004).

The findings discussed above suggest that the mode of transport in Xenopus oocytes and Drosophila oocytes depends not only on the cis-acting elements but also on the identity and availability of the factors with which they associate. Indeed, modification of the localized RNA both in the nucleus and in the cytoplasm has recently been observed in Xenopus. Modification of the late-pathway RNAs Vg1 and VegT in stage III/IV oocytes begins in the nucleus, in which hnRNP1 and Vg1RBP/Vera bind to the RNA, forming a core complex. After export into the cytoplasm, this complex undergoes remodeling and recruitment of additional components (the proline-rich, RNA-binding protein Prrp and XStau) that lead to the formation of an RNP-transporting/anchoring complex (Kress et al., 2004; Zhao et al., 2001).

Another process directed by the events in the nucleus is the regulation of translation. To produce proteins at the right time and place within the oocyte, localized mRNAs must be translationally silent not only during their transport but also at the final destination until their protein product is required. In Drosophila oocytes, Oskar protein does not accumulate before its mRNA becomes localized to the posterior pole in mid oogenesis (Markussen et al., 1995; Rongo et al., 1995). The fact that unlocalized oskar mRNA is associated with polysomes indicates that either regulated protein degradation or inhibition of translation is involved (Benton and St Johnston, 2002; Braat et al., 2004; Riechmann et al., 2002). The Bruno protein binds to Bruno-response elements (BREs) in the oskar mRNA 3′UTR, and mutation of these elements causes premature translation of oskar (Kim-Ha et al., 1995). Bruno also binds to the gurken mRNA 3′UTR and plays a role in its translational regulation (Filardo and Ephrussi, 2003).

At least eight distinct transacting factors, including Btz, Y14-Mago (a component of the EJC; see above), Yps and Cup, assemble with oskar mRNA into RNP particles and participate in oskar mRNA localization and its translational repression (Hachet and Ephrussi, 2001; Palacios et al., 2004; van Eeden et al., 2001; Wilhelm et al., 2000). Wilhelm et al. have proposed that an interaction between Cup and eukaryotic initiation factor 4E (eIF4E) blocks the initiation of oskar mRNA translation in early oogenesis (Wilhelm et al., 2003). In late oogenesis, after oskar becomes localized at the posterior, Cup no longer associates with eIF4E, and this de-represses translation (Lasko, 2003; Nakamura et al., 2004; Wilhelm et al., 2003). The translational regulation of oskar and gurken mRNAs also depends on the members of the hnRNP family, such as Hrp48/p50/Hrb27 and Squid (Goodrich et al., 2004; Gunkel et al., 1998; Yano et al., 2004).

Little is known about translational regulation of localized RNA in Xenopus oocytes. Vg1 mRNA undergoes 3′UTR-dependent translational repression before being localized in stage IV oocytes (Dale et al., 1989; King et al., 1999; Tannahill and Melton, 1989). VgRBP71, a KH-domain protein that interacts with the Vg1 mRNA LE, acts as a translational activator of Vg1 mRNA by promoting the removal of its translational repressor element. VgRBP71, which has RNA-strand-separating activity, binds to the 3′ end of Vg1LE, inducing the cleavage at an adjacent polyadenylation signal. As a consequence, the Vg1 mRNA becomes polyadenylated at this site and the downstream translational repressor element is removed (Kolev and Huber, 2003). Xcat2 and Xdazl mRNAs are probably silent throughout oogenesis. However, exogenous Xcat2 mRNA is translated upon injection into stage IV oocytes (Zhou and King, 1996). Since endogenous Xcat2 mRNA is sequestered within germinal granules (Kloc et al., 1998; Kloc et al., 2002a; Kloc et al., 2002b), it might be inaccessible to the components of the translation machinery. By contrast, exogenous Xcat2 RNA injected into stage III/IV oocytes and unable to enter germinal granules is readily available to undergo translation.

The translational silencing in Xenopus oocytes might also depend on Y-box proteins such as mRNP3 and FRGY2/mRNP4, which are present in mRNP storage particles and probably mask or inhibit translation (Bouvet and Wolfe, 1994; Darnbrough and Ford, 1981; Murray et al., 1992). Interestingly, the Y-box protein Yps, related to Xenopus FrgY2 (implicated in translational silencing), was discovered in an RNP complex containing oskar mRNA (see above) (Wilhelm et al., 2000). Recently, Tanaka et al. suggested that, in ascidians, translational repression of the localized mRNA cipem also depends on the formation of an RNP complex containing the Y-box protein CiYB (Tanaka et al., 2004).

RNA silencing regulates a plethora of vital functions in a variety of organisms, acting through small noncoding RNAs [short interfering (si)RNAs and/or micro (mi)RNAs] that are produced by DICER-mediated cleavage of long double-stranded or hairpin RNA precursors, respectively. Single-stranded miRNA or siRNA in RNA-induced silencing complexes (RISCs) recognize complementary target mRNA. Perfect complementarity results in target degradation, and partial complementarity results in translational repression (Bartel, 2004; Dykxhoorn et al., 2003; Finnegan and Matzke, 2003; Nolan and Cogoni, 2004).

Recent observations raise the exciting possibility that components of the RNA-silencing pathway regulate localized RNAs in oocytes. In Drosophila, the armitage gene, which encodes a component of the RNA-silencing pathway that is also involved in MT polarization (Cook et al., 2004; Tomari et al., 2004), is required for oskar mRNA silencing. Armitage is thought to participate in the assembly of the RISC and, by colocalizing with the oskar mRNA, it can spatially restrict RNA-silencing activity to its particular location (Cook et al., 2004).

Another gene required for RNA silencing in Drosophila and efficient translation of oskar mRNA is aubergine, which encodes a member of the RNAi-defective/Argonaute1 (RDE1/AGO1) protein family (Wilson et al., 1996). In addition, aubergine is involved in the localization of Maelstrom, a member of the Drosophila spindle-class family, to the nuage (Findley et al., 2003). The nuage is a specialized, perinuclear, electron-dense structure present in germ cells across the animal kingdom (Kloc et al., 2004b). It is believed to be a precursor of germinal (polar/P) granules, and recent studies indicate that it participates in the exchange of components between the nucleus and the germinal granules (Findley et al., 2003; Kloc et al., 2004a; Kloc et al., 2002b). In animals ranging from nematodes to mammals, the conserved component of nuage is the DEAD-box RNA helicase Vasa (Raz, 2000; Snee and Macdonald, 2004), which is probably involved in translational control of RNAs localized in the nuage and also, indirectly, in RNA silencing (Findley et al., 2003). In Drosophila, Maelstrom shuttles between the nucleus and cytoplasm, possibly via the nuage and sponge bodies (Findley et al., 2003). Maelstrom mutants also exhibit defective processing of Vasa and cause the mislocalization of two proteins involved in RNA silencing, Dicer and Argonaute2, which suggests a possible connection between the nuage and the RNA-silencing pathway in Drosophila oocytes (Findley et al., 2003).

Recently, a Vasa-like DEAD-box helicase and components of the splicing machinery, including Sm proteins, have been found in the nuage in Xenopus oocytes (Bilinski et al., 2004). This suggests that a subset of Sm proteins have a splicing-independent role. The nuage thus might serve as a platform for the shuttling of proteins and RNAs between the nucleus and cytoplasm and for the assembly of RNP particles necessary for the correct localization of various mRNAs into the germinal granules. These results also link the nuage and the RNA-silencing pathway (Barbee et al., 2002; Bilinski et al., 2004; Findley et al., 2003). Another possibility is that, as in Drosophila, there is a connection in Xenopus oocytes between splicing events and the fate of mRNAs localized in the nuage and germinal granules.

The mechanisms governing RNA localization in oocytes are clearly extremely complex. For example, the mechanisms governing the localization of endogenous RNAs are very often different from those governing the localization of exogenously introduced counterparts. The development of new imaging technologies and in vivo techniques allowing monitoring of the transport of endogenous RNA in real time (Bratu et al., 2003; Cha et al., 2001; Famulok, 2004; Forrest and Gavis, 2003; Snee and Macdonald, 2004; Stephenson, 2004) should alleviate the problems arising from the current heterogeneity of research techniques.

The recent findings that the fate of a localized RNA depends on its nuclear and/or cytoplasmic history and its connection to the RNA-silencing pathways, and the discovery that coding and noncoding localized RNAs can play a structural role in the organization of the oocyte cytoskeleton, should open completely new avenues in the study of RNA localization in oocytes. The majority of RNA localization studies in oocytes are limited to model organisms such as Drosophila and Xenopus, which are not necessarily reflective of other dipterans (Bullock et al., 2004) or amphibians. For example, in directly developing frog Eleutherodactylus coqui, in contrast to Xenopus, the mRNAs of the Vg1 and VegT orthologs are localized at the animal pole of the oocyte (Beckham et al., 2003). Findings like this indicate the necessity of widening studies to other, lesser-known species and taxa. Such studies could reveal which phenomena and mechanisms operating in RNA localization are fundamental to a given group of organisms and which evolved to accommodate the specific developmental needs of particular species.

This work was supported by grants from NSF (L.D.E.).