|Home | About | Journals | Submit | Contact Us | Français|
bicoid mRNA localises to the anterior of the Drosophila egg, where it is translated to form a morphogen gradient of Bicoid protein that patterns the head and thorax of the embryo. Although bicoid was the first identified localised cytoplasmic determinant1-4, little is known about how the mRNA is coupled to the microtubule-dependent transport pathway that targets it to the anterior, and it has been proposed that it is recognised by a complex of many redundant proteins, each of which binds to the localisation element in its 3'UTR with little or no specificity5. Indeed, the only known RNA-binding protein that co-localises with bicoid mRNA is Staufen, which binds non-specifically to dsRNA in vitro6, 7. Here we show that mutants in all subunits of the ESCRT-II complex (Vps22, Vps25 and Vps36) abolish the final Staufen-dependent step in bcd RNA localisation. ESCRT-II is a highly conserved component of the pathway that sorts ubiquitinated endosomal proteins into internal vesicles8, 9, and functions as a tumour-suppressor by removing activated receptors from the cytoplasm10, 11. However, the role of ESCRT-II in bicoid localisation appears to be independent of endosomal sorting, because mutations in ESCRT-I and III components have no effect of the targeting of bicoid mRNA. Instead, Vps36 functions by binding directly and specifically to stem-loop V of the bicoid 3'UTR through its N-terminal GLUE domain12, making it the first example of a sequence specific RNA-binding protein that recognises the bicoid localisation signal. Furthermore, Vps36 localises to the anterior of the oocyte in a bicoid mRNA-dependent manner, and is required for the subsequent recruitment of Staufen to the bicoid complex. This novel function of ESCRT-II as an RNA-binding complex is conserved in vertebrates, and may explain some of its roles that are independent of endosomal sorting.
Genetic screens for mutations that disrupt anterior-posterior patterning of the Drosophila embryo have identified a few genes that are required at different stages for the anterior localisation of bicoid mRNA, but most of these appear to play an indirect role in the process. Mutations in exuperantia (exu) abolish all stages of bicoid mRNA localisation13, 14, but its function is unclear, since Exu protein is a component of a translational repression complex that co-purifies with oskar mRNA, but not with bicoid itself15. Swallow, γ-Tubulin37C, Dgrip75, Dgrip128, and Minispindles are necessary for the localisation of bicoid mRNA from stage 10B of oogenesis onwards, and function to nucleate anterior microtubules that direct localisation at this stage16-18. Finally, Staufen is required for bicoid mRNA localisation at the end of oogenesis and in the early embryo19, 20. Unlike the other trans-acting factors, Staufen is a dsRNA-binding protein, and associates with bicoid mRNA at the oocyte anterior from stage 10B onwards6.
bicoid mRNA is localised by distinct and partially redundant mechanisms at different stages of oogenesis, which may explain why genetic screens have missed many of the essential trans-acting factors. In mutants that only disrupt early localisation, the localisation of the mRNA in late oocytes can rescue anterior development, whereas mutants that only disrupt late localisation result in a gradient of mRNA that induces some anterior patterning, unless bicoid translation is also impaired21-23. To circumvent this problem, we performed a direct visual screen in germline clones for mutants that alter the localisation of bicoid mRNA in living oocytes using GFP-Staufen as a marker24. This screen identified one complementation group, called larsen, that is required for the anterior localisation of bicoid mRNA. GFP-Staufen fails to localise to the anterior cortex of the oocyte in homozygous germline clones of both larsen alleles (Fig.1 a, b, d, e), whereas its posterior localisation with oskar mRNA is unaffected. The stronger allele, lsn5F3-8, is homozygous lethal, and females with a homozygous mutant germline do not lay eggs. However, the weaker allele, lsn2B63, is only semi-lethal, and mutant germline clones result in eggs in which bicoid mRNA forms a gradient across the anterior half of the embryo (Fig.1 f), a phenotype very similar to that seen staufen mutants6.
Both larsen alleles correspond to mutations in the highly conserved protein, Vps22 (supplementary Fig.1). Vps22p, together with Vps25p and Vps36p, forms the ESCRT-II (for endosomal sorting complex required for transport), one of four complexes (Vps27/Hrs complex, ESCRT-I, -II and -III) that act in a linear pathway to sort mono-ubiquitinated transmembrane proteins within the endosomal compartment into internal vesicles, leading to the formation of multivesicular bodies (MVBs)9, 25. Mutants in Drosophila ESCRT-I and ESCRT-II components have been shown to cause a tumorous phenotype, because activated receptors accumulate on endosomes, rather than being removed from the cytoplasm and degraded10, 11, 26. Mammalian ESCRT-II was independently identified as a complex that binds the RNA polymerase II elongation factor, ELL27, 28, while the fission yeast Vps 22/Larsen homologue regulates the expression of centriolar proteins during meiosis29. However, the relationship between these functions and endosomal protein sorting remains unclear.
To test whether the role in bcd mRNA localisation is specific for Larsen/Vps22 or whether the whole ESCRT-II is involved in the process, we examined mutants in the other two components of this complex, vps25 and vps36. Homozygous germline clones of vps25 and vps36 show an identical phenotype to lsn5F3-8: GFP-Stau does not localise to the anterior of the oocyte (Fig.1g-j), and the flies do not lay eggs. We therefore tested whether other MVB sorting mutants also disrupt bcd mRNA localisation. However, bicoid mRNA localises normally in germ line clones mutant for hrs (vps27), the ESCRT-I components, vps28 and ept, and the ESCRT-III component, vps32 (Fig.2c-f). Thus, the ESCRT-pathway appears to be dispensable for bcd mRNA localisation, indicating that ESCRT-II has another function in addition to the sorting of ubiquitinated proteins into MVBs.
Given the role of ESCRT-II in bicoid mRNA localisation, we tested whether any of the proteins of the complex interact with the bicoid localisation signal, using a yeast 3-hybrid assay30. Both Staufen and Vps36, but not Vps22 or Vps25, interact with the bicoid 3′UTR in this assay, resulting in growth of the yeast cells on plates lacking histidine and in the expression of α-galactosidase (Fig.3a). To determine whether Vps36 protein binds directly to bcd RNA, we performed a UV-cross linking assay with purified recombinant Vps36 and in vitro transcribed bicoid 3'UTR. Vps36 cross-links efficiently to the bcd 3′UTR in this assay, indicating that it contacts single-stranded regions of the RNA directly (Fig.3b). Furthermore, this interaction is specific, since the binding is competed by excess unlabelled bcd 3′UTR, but not by control RNAs (Fig.3b, d, e and data not shown). Given the high degree of conservation of ESCRT-II, we tested whether Xenopus laevis Vps36 also interacts with RNA, and found that its N-terminal GLUE domain12 binds specifically to the bcd 3'UTR, indicating that the RNA-binding activity of Vps36 is conserved in vertebrates (Fig.3c).
To map the binding site within the RNA more precisely, we tested the ability of single stem-loops of the bicoid 3'UTR to compete for binding to Vps36. Stem-loop V competed as well as the full-length 3'UTR, whereas stem-loops III, IVb and the distal part of stem-loop V did not compete (Fig.3b, d). We then performed a random mutagenesis to map the Vps36 binding site within stem-loop V, and found that none of the distal mutations affect the interaction, whereas three base changes in the proximal stem and the central loop-region strongly reduce Vps36 binding (Fig.3e and supplementary Fig.2). Thus, Vps36 binds specifically to the proximal part of stem-loop V, and recognises specific bases, most probably in the context of the RNA structure of this region.
We next examined whether Vps36 associates with bicoid mRNA in vivo by generating transgenic flies expressing Vps36-YFP under the control of its endogenous promoter. Vps36-YFP is ubiquitously expressed in the ovary, and shows a general cytoplasmic localisation with some clouds in the nurse cells (Fig.4a and a′). From stage 10B/11 of oogenesis, the protein localises to the anterior cortex of the oocyte in the same region as bicoid mRNA (Fig.4b, b' and c). This anterior accumulation is abolished in exu mutants, which block bicoid mRNA localisation at an earlier stage (Fig.4d), and in sry δ mutants, in which bcd RNA is not transcribed (Fig.4e). Thus, Vps36 co-localises with bicoid mRNA at the anterior, and this localisation is bicoid mRNA-dependent, indicating that the protein binds directly to the RNA in vivo, as it does in vitro.
Both ESCRT-II and Staufen are recruited to bicoid mRNA at stage 10B, and are then required for its localisation during the final stages of oogenesis. staufen mutants have no effect on the anterior recruitment of Vps36 (Fig.4f), however, whereas mutants in larsen/vps22 and vps36 abolish the anterior recruitment of Staufen protein at this stage (Figs.(Figs.11 and and2).2). ESCRT-II therefore binds to bicoid mRNA independently of Staufen, and is required for the subsequent recruitment of the latter to form a functional bicoid mRNA localisation complex.
Our results demonstrate that ESCRT-II is required for the anterior localisation of bicoid mRNA, and is the first identified sequence-specific RNA binding complex that recognises the bicoid localisation signal. This novel activity of ESCRT-II seems to be independent of its well-characterised role in endosomal protein sorting, and is conserved, since Xenopus Vps36 also interacts specifically with the bcd 3'UTR, suggesting that ESCRT-II may play a role in mRNA localisation in vertebrates. In future, it will be important to determine whether this conserved RNA-binding activity of ESCRT-II plays any role in endosomal trafficking, or in the other proposed functions of the complex in transcription elongation and the regulation of centriole assembly.
A more detailed description of the materials and methods used can be found under supplementary information.
Flies were grown under standard conditions on corn meal medium. Germline clones were induced using the autosomal DFS technique. For detailed information and the genotypes of the flies used see supplementary information.
For the yeast 3-Hybrid assay we used the system based on the transcription factor GAL 4 and the RNA binding protein RevM10 from HIV-1, as well as its binding element (Rev responsive element, RRE)30.
The purified proteins were pre-incubated with the competitor RNAs for 20 min at room temperature in X-link buffer (PBS + 5% Glycerol, 2 mM DTT, 0.2% NP-40, 100 μg/ml heparin and 100 μ/ml tRNA). After addition of the probe and a further 20 min of incubation the cross-linking was performed using a ‘Stratalinker 1800’ at maximum power for 5 min. The samples were treated with RNaseA (0.5 μg/ml) for 1 h at 37°C and after SDS-PAGE the radioactivity was visualised using a phosphoimager.
Stainings were performed as described. Pictures were taken using a BioRad 1024 Confocal Microscope and processed with Adobe Photoshop.
Supplementary Figure 1
Comparison of ESCRT-II proteins
a shows a schematic comparison of Vps25, Vps22/Lsn and Vps36 from yeast (Sc, Saccharomyces cerevisiae), humans (Hs, Homo sapiens) and flies (Dm, Drosophila melanogaster). The winged-helix domains A and B are represented as grey boxes (WHA and WHB), the GLUE domain of Vps36 as a red box. Amino acid similarities between the proteins are given as percentages. b sequence alignment between Vps22/Lsn from flies, humans and yeast. The mutations in lsn2B6-3 (blue triangle, R64->C) and lsn5F3-8 (red triangle, R2->Stop) are indicated.
Supplementary Figure 2
Random mutagenesis of stem-loopV
a shows the four pairs of oligonucleotides that were used for the mutagenesis. For each experiment, we replaced one or two of the oligonucleotides with “degenerate” oligonucleotides in which each base had a 10% probability of being mutant. The oligonucleotides were annealed, the internal pairs were phosphorylated, and they were then ligated together and cloned into pSP73, which had been modified to contain MluI and EagI sites. In vitro transcription then results in stem-loopV RNA. In b the mutations in stem-loop V that decrease its affinity for Vps36 are indicated. They correspond to positions 419, 427 and 428 of the full-length 3'UTR.
We are very grateful to Roger Williams, Olga Perisec and Hsiangling Teo for giving us the purified recombinant Xenopus laevis Vps36 GLUE domain, and for their helpful advice on ESCRT-II. We would like to thank John Oberdick and Zulung Zhang for sharing their unpublished results on rat Vps36. This work was supported by a Wellcome Trust Principal Research Fellowship to D. St J. and by the Max-Planck-Gesellschaft (U.I.).