Musashi: a translational regulator of cell fate

Cell-fate specification in vivo is achieved by differential gene expression, which can involve regulation at various levels, including transcription, RNA processing, translation and post-translational protein modification. In cell-fate decisions, transcriptional regulation has been thought to play the central role; however, the role of `translational'regulation has remained largely unknown. In contrast to the vast number of studies of transcriptional control of gene expression, only a few studies of translational regulation have been reported. The shortage of in vivo studies of translational regulation is largely due to the lack of simple, established methods for analysing the behaviour of RNA in organisms. To date, the translational regulation of maternal mRNA during early animal development has been characterised mainly biochemically in eggs from sea urchin(Wagenaar and Mazia, 1978), Drosophila (Wreden et al.,1997) and other animals (reviewed byGray and Wickens, 1998). However, such approaches cannot easily identify translational control of zygotic genes involved in cell-type specification and examine its function in multicellular tissues. To overcome this difficulty, powerful genetic approaches using C. elegans or Drosophila have been adopted. These studies have unveiled an important role for translational control,although the detailed molecular mechanisms explaining how RNA-binding proteins are involved in the translational control of their target mRNAs remain to be elucidated. Here, we outline the current understanding of translational control in development and discuss the functions of the evolutionarily conserved, neural RNA-binding protein Musashi.

The C. elegans germ line provides an excellent model system for genetic studies of cell fate decision. C. elegans has two sexes: XO animals, which develop as males; and XX animals, which are hermaphrodites. The hermaphrodites initially produce sperm but switch to producing oocytes. This cell-fate switch is controlled by the translational repression of tra-2 mRNA (Goodwin et al., 1993), which encodes a predicted membrane protein; this protein acts as a negative regulator of downstream genes required for male development, through the interaction of tra-2 mRNA 3′-UTR and an RNA-binding protein, GLD-1 (Jan et al.,1999). In addition, FEM-3, which controls sperm-oocyte fate by promoting spermatogenesis, is translationally regulated(Ahringer and Kimble, 1991) by FBF (fem-3 binding factor), a member of the FBF RNA-binding protein family that includes FBF-1 and FBF-2(Zhang et al., 1997). Furthermore, CPEB (cytoplasmic polyadenylation-element-binding protein)proteins, evolutionarily conserved RNA-binding proteins that bind to and regulate the translation of specific mRNAs, have been shown to control two key steps in C. elegans spermatogenesis. Specifically, the CPEB protein FOG-1 specifies sperm cell fate and another CPEB protein, CPB-1, executes this decision by regulating translation of a specific set of mRNAs(Luitjens at al., 2000;Jin et al., 2001). Future work revealing the mechanisms by which these CPEB proteins act to regulate the cell fate should provide interesting insights into this process.

Sexual development in Drosophila depends on the ratio of X chromosomes to autosomes and involves a cascade of RNA splicing events(Baker, 1989). Recent findings indicate that translational control plays a part in this process. The expression of the neural sex-determination factor Fruitless (FRU) is regulated translationally in a sex-specific manner(Usui-Aoki et al., 2000). FRU protein is produced in motor neurons in a male-specific manner and directs the formation of the male-specific muscle of Lawrence (MOL). FRU is not detected in female neurons, despite the fact that similar levels of fru mRNA are expressed in male and female neurons; expression of the fru gene thus appears to be translationally regulated. This sex-specific translational repression of fru mRNA is mediated by the binding of an RNA-binding protein, Transformer (TRA), one of the sex-determining gene products (e.g. Sex-lethal, Doublesex, TRA), to fru mRNA in female neurons. In Drosophila sex determination, TRA had previously been known to act as a splicing activator of the precursor mRNA of a downstream gene(doublesex) (reviewed by Baker,1989). Thus, the latest findings indicate that TRA is a multifunctional RNA-binding protein that acts as both a splicing activator and a translational repressor.

Post-transcriptional gene regulation is involved in asymmetric cell division during Drosophila development. Previous findings indicated only the involvement of asymmetric sorting of mRNA(Li et al., 1997;Takizawa et al., 1997;Broadus et al., 1998) or protein (Rhyu et al., 1994;Kraut et al., 1996;Ikeshima-Kataoka et al., 1997)between two daughter cells (reviewed byJan and Jan, 1998), which provides no evidence for translational control of this process. However,recent work has shown that translational control does in fact occur.

The Drosophila external sensory organ is an excellent model system for studies of mechanisms that regulate asymmetric cell divisions during development (Jan and Jan,1998). Previously, we identified a neural RNA-binding protein,Musashi (MSI), that is required for the asymmetric cell division of the sensory organ precursor cell (SOP) of the Drosophila adult external sensory organ (Nakamura et al.,1994). In the wild-type fly, the SOP generates two second-order precursor cells: a non-neuronal precursor cell (the IIa cell) and a neuronal precursor cell (the IIb cell). In contrast, in the loss-of-function msi mutants, the SOP fails to undergo asymmetric cell division and instead gives rise to two IIa cells. Consequently, the number of socket and/or shaft cells is increased at the expense of neurons and glia, which results in a `double bristle'phenotype^*(Fig. 1). Molecular analysis of the msi locus demonstrated that it encodes an RNA-binding protein that has two tandem RNA-recognition motifs (RRM-1 and RRM-2), each of which includes two short, highly conserved motifs: RNP-1 (eight residues) and RNP-2(six residues) (reviewed by Dreyfuss et al., 1993). This structure led us to propose that post-transcriptional control plays a role in this asymmetric cell division. However, the precise role of MSI in this process remained obscure, mainly because of the difficulties of determining its in vivo target RNA.

Intensive biochemical and genetic studies demonstrated that the in vivo target of MSI is the tramtrack69 (ttk69) mRNA, which encodes the key determinant of IIa versus IIb fate(Okabe et al., 2001). The TTK69 protein is a zinc-finger-type transcriptional repressor, whose expression is necessary and sufficient to specify a non-neuronal identity(reviewed by Jan and Jan,1998). The regulatory mechanism that ensures that TTK69 protein is present only in the IIa cell was unknown. Surprisingly, we found that ttk69 mRNA is expressed in both IIa and IIb cells at apparently equal levels, which indicated that the synthesis of the TTK69 protein is regulated translationally rather than transcriptionally. The underlying molecular mechanism was revealed by intensive analysis of a gain-of-function allele of ttk (ttk¹)(Xiong and Montell, 1993). The ttk¹ mutants possess a P-element insertion in the 3′UTR of the ttk69 mRNA and ectopically express TTK69 protein in the presumptive IIb cell. This suggested that the cell-type-specific translational repression of TTK69 depends on cis-acting repressor sequences in the ttk69 3′ UTR. Subsequently, in vitro selection experiments(Buckanovich and Darnell, 1997)and biochemical assays showed that MSI protein specifically binds to cis-acting repressor RNA sequences that contain GU_3-6[G/AG] repeats in the 3′ UTR of ttk69 mRNA to execute its translational repression. Indeed, in the msi mutant, the translation of ttk69 mRNA is de-repressed and the TTK69 protein is ectopically produced in the presumptive IIb cell (Fig. 2) in a way similar to that seen in the gain-of-function ttk¹ mutant. Consequently, the presumptive IIb cell could have transformed into a IIa cell, thus producing double-bristle phenotype. Although the molecular mechanism responsible for the absence of MSI function in the IIa precursor cells remains to be elucidated, it is possible that the function of MSI is regulated post-translationally by Notch signalling, which is differentially activated in IIa precursor and IIb precursor cells(Okabe et al., 2001).

The Musashi family is an evolutionarily conserved group of neural RNA-binding proteins (Table 1)that has representatives in Drosophila melanogaster(Nakamura et al., 1994), C. elegans (Yoda et al.,2000), Halocynthia roretzi(Kawashima et al., 2000), Xenopus laevis (Richter et al.,1990; Good et al.,1993), mouse (Sakakibara et al., 1996; Sakakibara et al.,2001) and human (Pincus et al., 1998; Good et al.,1998). In mammals, the Musashi family is important for cell fate determination in the broader sense, playing roles in maintenance of the stem-cell state, differentiation and tumorigenesis. The mammalian homologue of MSI, Musashi1, is selectivley expressed in neural progenitor cells, including neural stem cells (Sakakibara et al.,1996; Kaneko et al.,2000). Outside the nervous system, Musashi1 is a selective marker for intestinal stem or early lineage cells (C. Potten and H. Okano,unpublished).

The selective expression of Musashi1 in stem cells or immature cells of these tissues led us to speculate that it plays a role in keeping these cells in an undifferentiated state during post-transcriptional gene regulation. We sought to identify its target RNA by using a strategy similar to that used in the study of Drosophila Musashi. By in vitro selection, we determined that the consensus ligand RNA sequence for mammalian Musashi1 is G/AU_2-3(AGU). We then explored candidates for the in vivo Musashi1 target gene on the basis of the results of in vitro selection experiments as well as expression patterns and functions. We speculated that mRNAs of genes regulating neural differentiation (either positively or negatively) would be downstream targets of Musashi1 since Musashi1 is preferentially expressed in undifferentiated neuronal progenitor cells.

One of the in vivo targets of Musashi1 is m-Numb mRNA, the 3′ UTR of which has a Musashi1-binding site(Imai et al., 2001). The m-Numb and Musashi1 expression patterns overlap in neuroepithelial cells in the ventricular zone of the neural tube(Sakakibara et al., 1996;Zhong et al., 1996;Zhong et al., 1997). Furthermore, m-Numb is involved in the regulation of neuronal differentiation (Wakamatsu et al.,1999). Studies using both gain-of-function and loss-of-function mutations have demonstrated that Musashi1 translationally represses synthesis of m-Numb. Because Numb is an evolutionarily conserved intracellular Notch antagonist (Uemura et al.,1989; Guo et al.,1996; Zhong et al.,1996; Zhong et al.,1997), we expected Musashi1 to be a positive regulator of Notch1 signaling (Fig. 3). Indeed,overexpression of Musashi1 activates Notch1 signaling through a pathway dependent on the action of RBP-Jκ(Imai et al., 2001), a dormant transcription factor of the CSL family that forms a functional complex with an intracellular domain of Notch1 protein within the nucleus(Schroeder et al., 1998). Notch signaling is known to induce the self-renewal of mammalian neural stem cells (Nakamura et al., 2000;Gaiano and Fishell, 2000). By reducing the activity of mammalian musashi genes [including those encoding Musashi1 and Musashi2 (Sakakibara et al., 2001)] through the antisense ablation of Musashi2 protein production in cultured brain cells derived from musashi1^-/- mice, we found that these genes play essential roles in maintaining the undifferentiated state (or self-renewal) of neural stem cells (S. Sakakibara and H. Okano, unpublished).

Musashi1 is expressed in particular types of brain tumor that are likely to have originated from immature brain cells(Toda et al., 2001;Kanemura et al., 2001). Interestingly, the Musashi1 expression level correlates with the malignancy and proliferative activity of the tumor. Furthermore, glioblastoma cells that express high levels of Musashi1 show a significantly higher rate of nuclear localization (and hence activation) of Notch1 than do cells expressing lower levels of Musashi1 (Kanemura et al.,2001). Cells that stained strongly for Musashi1 showed almost no staining for m-Numb. Thus, a high level of Musashi1 expression could have led to the clonal expansion of the above-mentioned tumor cells by activating Notch signaling, presumably through the translational inhibition of m-Numb(Kanemura et al., 2001). Hyperactivation of Notch signaling could result in tumorigenesis, possibly owing to the inhibition of apoptosis and maintenance of a sustained immature state (Artavanis-Tsakonas et al.,1999). The trigger causing Musashi1 overexpression, which could be the initial step in tumorigenesis, remains to be elucidated. Nevertheless, it is clear that members of the Musashi family are involved in determining cell fate not only during normal neural development but also in a particular pathogenic state.

The molecular mechanism underpinning repression of the translation of m-Numb mRNA by Musashi1 remains to be elucidated. However, in the case of TRA, a single RNA-binding protein could act as a multifunctional regulator that controls its target genes at several different steps of post-transcriptional regulation, including splicing, translation, stability control and localization of RNAs. However, why a single RNA-binding protein might have a multifunctional role in post-transcriptional gene regulation remains an open question. Interestingly, intracellular localization of Musashi1 protein is variable (cytoplasmic and/or nuclear), depending on the cell-type and/or developmental stage. Thus, Musashi1 could be involved at a step other than translational control — something we and others are currently investigating.

The work described above has revealed the importance of translational control of zygotic gene expression in the development of multicellular organisms. In addition, it has shown that members of the Musashi family are involved in cell-type specification in neural development — that is, in the determination of neural versus non-neural fates or the self-renewal of neural stem cells. This is accomplished through the translational repression of target mRNAs. Further studies of the upstream signaling mechanisms regulating the expression and/or function of such translational regulators could reveal the whole gene cascade responsible for cell-fate specification during development.

We thank Makoto Nakamura, Shinichi Sakakibara, Yoshihiro Kanemura,Christopher Potten, Hirotaka James Okano, Yasushi Hiromi and all members of the Okano laboratories for helpful discussions.