XSEB4R, a novel RNA-binding protein involved in retinal cell differentiation downstream of bHLH proneural genes

Dissecting genetic cascades responsible for the development of the nervous system has helped to understand some of the molecular mechanisms involved in the genesis of distinct neuronal subtypes. The best-known molecular network implicated in neural cell fate decisions involves proneural and differentiation genes, which encode bHLH and other transcription factors, as well as neurogenic genes belonging to the Notch/Delta signalling pathway(Bertrand et al., 2002). The retina has been used as a model to study the role of these genes during retinal histogenesis because of its simple organisation and the limited number of neuronal types that it contains (Cepko,1999; Perron and Harris,2000; Vetter and Brown,2001). The order of expression of several proneural and neurogenic genes has been studied in the retina, focusing on the proliferative ciliary marginal zone (CMZ). The CMZ is a region at the peripheral edge of the retina where cells are spatially ordered with respect to their development, with stem cells closest to the periphery, retinoblasts in the middle and differentiating precursors at the central edge (Wetts et al., 1989; Dorsky et al.,1995). In the CMZ, neurogenic and proneural genes are activated first in retinoblasts, followed by differentiation genes in differentiating precursors, reflecting a genetic hierarchy among these genes(Perron et al., 1998; Perron et al., 1999b). Overexpression or loss-of-function experiments of differentiation genes in the retina affects retinal cell type distribution(Perron and Harris, 2000; Cepko, 1999). For example,loss of ath5 (atoh7 - Zebrafish Information Network; Atoh7 - Mouse Genome Informatics) function in zebrafish and mouse prevents the differentiation of ganglion cells. Conversely, overexpression experiments show that ath5 promotes ganglion cell production at the expense of other cell types in Xenopus. It has therefore been proposed that ath5 is essential for retinal ganglion cell differentiation (Kanekar et al.,1997; Kay et al.,2001; Morrow et al.,1999; Brown et al.,2001). Recently, several lines of evidence have converged to propose that a combinatorial code of bHLH and homeobox proteins is responsible for the specification of the correct neuronal subtypes. For example,co-expression of the mouse bHLH genes Math3 (Neurod4 - Mouse Genome Informatics) and NeuroD (Neurod1 - Mouse Genome Informatics) with the homeobox genes Pax6 or Six3significantly increases their ability to promote amacrine cell genesis(Inoue et al., 2002). As the various retinal cell types are born in a sequential order, it has also been proposed that the Notch/Delta pathway could generate cell diversity by controlling when a cell is released from lateral inhibition(Dorsky et al., 1995; Dorsky et al., 1997; Perron and Harris, 2000). If a neuroblast is released during early retinogenesis, it gives rise to an early born retinal cell type, whereas if it is released during late retinogenesis it gives rise to a late born retinal cell type.

Interactions among neurogenic, proneural and differentiation genes have been extensively studied during primary neurogenesis in Xenopusallowing the establishment of a genetic cascade(Ferreiro et al., 1993; Turner and Weintraub, 1994; Bellefroid et al., 1996; Ma et al., 1996; Chitnis and Kintner, 1996; Perron et al., 1999b). These interactions encompass mainly transcriptional regulation. However, this genetic network probably requires other levels of gene regulation. For instance, it has recently been found that XNeuroD function during primary neurogenesis and retinogenesis can be inhibited by glycogen synthase kinase 3β (Marcus et al., 1998; Moore et al., 2002). This post-translational phosphorylation regulation is crucial for the proper function of XNeuroD (Moore et al.,2002). Post-transcriptional regulation at the mRNA level,involving RNA binding proteins, is also known to play a key role in gene regulation (Burd and Dreyfuss,1994; Perrone-Bizzozero and Bolognani, 2002). Once mRNAs are transcribed, RNA-binding proteins can control all subsequent maturation steps from splicing and translation, to mRNA transport and stability (Harford and Morris, 1997). According to the motif contained in RNA-binding proteins, one can distinguish several families. The largest family of RNA-binding proteins is characterised by the presence of RNA recognition motifs (RRM), domains composed of 90-100 amino acids that are only moderately conserved with two consensus sequences (an octamer and a hexamer sequence called RNP1 and RNP2, respectively). The number of RRMs per protein varies from one to four (Burd and Dreyfuss,1994). Recent advances in the analysis of several RNA-binding proteins during development have increased the perspectives in this developmental biology field.

In the nervous system, a large number of genes are regulated post-transcriptionally via the interaction of their mRNAs with specific RNA-binding proteins. At present, we know several RNA-binding proteins involved in the development and plasticity of the central nervous system(CNS). However, little is known about their precise role and their RNA targets. During neurogenesis for example, the Staufen protein mediates prospero mRNA localisation, which is important for neuroblast asymmetric division in early Drosophila embryogenesis(Matsuzaki et al., 1998). In mammals, the two homologues of Staufen (Stau1 and Stau2) are involved in mRNA transport in dendrites, and they also interact with ribosomes, suggesting an additional role in translation regulation. However, vertebrate RNA targets of Stau1 and Stau2 have not yet been identified(Duchaine et al., 2002; Kiebler and DesGroseillers,2000). Another conserved RNA-binding protein family is the Musashi family. In mammals, two members, Musashi1 (Msi1) and Musashi2 (Msi2), are expressed in neural precursor cells. Antisense ablation experiments suggest that Msi1 and Msi2 are cooperatively involved in the proliferation and maintenance of CNS stem cell population(Sakakibara et al., 2002). Concerning the targets of these genes, Msi1 represses the translation of Numb,an antagonist of Notch (Okabe et al.,2001), and it has recently been suggested that Msi1 also mediates the post-transcriptional regulation of the microtubule-associated protein Tau(Cuadrado et al., 2002). The ELAV/Hu proteins belong to a RNA-binding protein family largely conserved across species. In vertebrates, most members are neuron specific, and have been shown to be essential for nervous system development and function through the regulation of the stability of their mRNA targets, including GAP43, Tau or MYCN (Beckel-Mitchener et al.,2002; Aranda-Abreu et al.,1999; Manohar et al.,2002; Perrone-Bizzozero and Bolognani, 2002). These examples emphasise the important role of post-transcriptional factors during neurogenesis.

With the aim of advancing our knowledge of the genetic network involved in retinal cell fate determination, we have characterised a novel RNA-binding protein and we have studied its function during retinogenesis. We present the cloning and spatio-temporal expression of Xseb4R, which encodes a putative RNA-binding protein containing a single RRM. A related gene, Xseb4, has been previously isolated in Xenopus(Fetka et al., 2000). While Xseb4 is mainly expressed in muscles, Xseb4R is strongly expressed in neural tissues. We show here that overexpression of Xseb4R during primary neurogenesis or in the retina has a proneural effect. Blocking Xseb4R function using morpholino oligonucleotides leads to the opposite effect. Using classical overexpression experiments in the early Xenopus embryo, we demonstrate that Xseb4R is responsive to neurogenin, NeuroD and the Notch/Delta signal transcription cascade. In the Xenopus nervous system, several RNA-binding proteins have been identified previously but their functions remain elusive (Good et al.,1993; Gerber et al.,1999; Perron et al.,1999a). Our present data suggest that the RNA-binding protein XSEB4R has a proneural function during Xenopus neurogenesis.

Large-scale whole-mount in situ hybridisation was performed for screening a tadpole head (ZAP express phage) cDNA library, as described by Souopgui et al.(Souopgui et al., 2002). Briefly, single recombinant phages were eluted in 96-well microplates. Fluorescein-labelled antisense RNA probes were transcribed from templates obtained by PCR amplification of cDNA inserts for these single phages. Four sets of flat-bottom 24-well devices for simultaneous whole-mount in situ hybridisation were used per round of screening. cDNA clones with an interesting expression pattern were sequenced and matched with the DDBJ/EMBL/GenBank sequence information. GenBank Accession Number:AY289193.

Two antisense morpholino oligonucleotides (Mo) against Xseb4R were designed (sequences complementary to AUG are underlined), Mo1(GTGCATGGTCACAGGCAAATTCACC) and Mo2 (starting 2 nucleotides after AUG; AAAGTTGTGTCTTTTTGCACGGTGT), as well as a Mo against GFP cloned into the pCS2 plasmid (TCCTTTACTCATGGTGGATCCTGCA). The standard control morpholino (cMo: CCTCTTACCTCAGTTACAATTTATA) was used as a control (Genetools). Two kinds of Mo have been used: crude Mo (Mo1 and cMo) for blastomere injections and Special Delivery Mo (Mo1, Mo2, MoGFP and cMo), where the non-ionic crude morpholinos are paired to a complementary `carrier' DNA in order to be transfected (Morcos,2001; Ohnuma et al.,2002).

The full-length Xseb4R cDNA was cloned into the pCMV vector, and the open reading frame (ORF) subcloned into the pCS2 vector. The flag-tagged version was engineered by subcloning the ORF into the pCS2-Flag vector. Another construct, called Xseb4R-5′UTR, has been subcloned into pCS2. This construct contains the ORF as well as the region of the 5′UTR complementary to Xseb4R Mo1. Xseb4R-GR was generated by subcloning the ORF into pCS2-GR, a vector initially generated by inserting the GR coding sequence into the XhoI and XbaI sites within the multiple cloning sites of the pCS2 vector.

Capped Xseb4R, Xngn1 (a gift from C. Kintner), XneuroD (a gift from E. Bellefroid), XNotch ICD (a gift from E. Bellefroid), Xseb4 (a gift from R. Rupp) and NLS lacZ RNAs were prepared from CS2 plasmids after Not1 linearisation using mMessage mMachine kit (Ambion). RNAs were injected in a volume of 5 nl, at a concentration of 50 pg/nl (when not notified), into a single blastomere of embryos at the two-cell stage. 5-20 ng of Mo1 or control Mo were injected into a single blastomere of embryos at the two-cell stage. lacZ mRNA was co-injected as a marker. Histochemical staining for β-galactosidase activity was performed to visualise the distribution of the co-injected lacZ mRNA. Embryos were collected at the neurula stage and subjected to in situ hybridisation as described below. Xseb4R-GR-injected embryos were continuously treated, or not, from the stages indicated in the results with 10 μM, final concentration, of dexamethasone.

The Qiagen Rneasy mini kit was used for RNA isolation from oocytes or embryos of different developmental stages. All RNA preparations were treated with DNase I (Qiagen) and checked with 32 cycles of histone H4-specific PCR(Niehrs et al., 1994) for DNA contamination. RT-PCR was carried out using the Gene Amp RNA PCR kit(Perkin-Elmer). The following primers, annealing temperatures and cycle numbers were used:

histone H4, forward (F) 5′-CGGGATAACATTCAGGGTATCACT-3′, reverse(R) 5′-ATCCATGGCGGTAACTGTCTTCCT-3′, 58°C, 25 cycles; and

Xseb4R, (F) 5′-GGAACCTGCAGAGCGCATTTACTA-3′, (R)5′-GTCAGGCTGGAGCTGTTGAGGCTG-3′, 60°C, 33 cycles.

PCR products were separated on 2% agarose gels.

Digoxigenin (DIG)-labeled antisense RNA probes were generated for Xseb4R according to the protocol of the manufacturer (Roche). For analysis of expression in the whole embryo during development, in situ hybridisation was performed as previously reported(Souopgui et al., 2002). For analysis of expression in the retina, whole-mount in situ hybridisation was performed as described previously(Shimamura et al., 1994),apart from that embryos were bleached(Broadbent and Read, 1999) just before the proteinase K step. After NBT/BCIP (Roche) staining, embryos were then vibratome sectioned (50 μm).

BrdU was injected intra-abdominally, and the animals were allowed to recover for 2-8 hours post-injection. BrdU was detected using the BrdU labeling kit (Roche) after a 45-minute treatment in 2N HCl. For double staining, the mRNA was first detected by whole-mount in situ hybridisation (as described above). Embryos were then cryostat sectioned and BrdU immuno-stained.

Immunohistochemistry was performed on 4% paraformaldehyde-fixed tissues. Cryostat sections (12 μm thick) were incubated with primary antibodies,anti-Islet1 (a gift from S. Thor), anti-Flag (Stratagene), anti-CD2 (Serotec)or anti-BrdU (Roche), and visualised using anti-mouse fluorescent secondary antibodies (Alexa, Molecular Probes). To visualise the nuclei, sections were incubated for 5 minutes in Hoechst solution (10 μg/ml) and washed three times in PBS.

DNA was transfected into the presumptive region of the retina of stage 18 embryos, as previously described (Holt et al., 1990; Ohnuma et al.,2002). Mo (10 ng) were similarly transfected. Embryos were fixed at stage 41 and cryostat sectioned (12 μm). GFP-positive cells were counted and cell types were identified based upon their laminar position and morphology, as previously described(Dorsky et al., 1995).

Polyclonal antibodies against XSEB4R have been raised by Eurogentec. This antibody is directed against the N-terminal peptide of XSEB4R: HTVQKDTTFT.

Capped synthetic Xseb4R mRNA containing the part of the 5′UTR against which Mo1 is directed was prepared from pCS2 plasmids after Not1 linearisation using mMessage mMachine kit (Ambion). 500 pg of this RNA and 5 ng of Mo (Mo1 or Control Mo) were injected into both blastomeres of embryos at the two-cell stage. Embryos were harvested at stage 10, frozen in liquid nitrogen and stored at -80°C until further analysis. For preparation of extracts, frozen embryos were homogenised in 10 μl of extraction buffer (50 mM β-glycerophosphate, 20 mM EGTA, 15 mM MgCl₂, 1 mM DTT, pH 7.3) with proteases inhibitor cocktail(complete Mini, Roche), centrifuged for 15 minutes, and the supernatants collected. Proteins were then separated on 4% stacking and 10% resolving SDS-polyacrylamide gels (PAGE), as described by Laemmli(Laemmli, 1970). The separated polypeptides were electrophoretically transferred from gels to nitrocellulose membranes and processed for immunoblotting. Blots were then incubated for two hours with the primary antibody against XSEB4R, diluted 1:100 in 10% dried skimmed milk in Tris-buffered saline-Tween buffer [TBST: 1.37 M NaCl, 0.2 M Tris (pH 7.5), 1% Tween-20]. The peroxidase conjugated anti-rabbit (Vector)was used as secondary antibody at a dilution of 1:5000 in 5% dried skimmed milk in TBST buffer, for a two hour incubation. Blots were developed using the chemoluminescence kit (Amersham) and the reactivity was visualised on hyperfilm ECL (Amersham).

Apoptotic cells were detected by TUNEL methods using the `In situ cell death detection kit, TMR red' (Roche) on 12 μm cryostat sections of stage 34-41 lipofected embryos.

During embryogenesis in Xenopus, XDelta1, XNotch1 and its downstream targets define a synexpression group(Niehrs and Pollet, 1999). Their specific expression pattern in the form of stripes in the open neural plate (Coffman et al., 1990; Chitnis et al., 1995; Koyano-Nakagawa et al., 1999; Lamar et al., 2001) offers a reliable criterion to search for novel candidate genes belonging to the Notch signalling pathway. The clone JS124 was identified using this strategy in a large random expression pattern screen of a tadpole head cDNA library by whole-mount in situ hybridisation(Souopgui et al., 2002). Nucleotide sequence analysis revealed that it encodes a Xenopushomologue of the vertebrate and invertebrate RNA-binding protein SEB4. It is clearly distinct from the previously known XSEB4(Fetka et al., 2000), and closely related to murine SEB4, human SEB4B, and to a Caenorhabditis elegans hypothetical protein (Fig. 1). To distinguish the two SEB4 genes identified in Xenopus laevis, we designate the new gene as XSEB4R ('R' for related). XSEB4R is one of the few members of the RRM protein family, containing only one RRM located at the N terminus. The vertebrate SEB4 protein also shows two other conserved domains in the C-terminal portion(Fig. 1).

The role of a given gene during development is reflected in its specific spatio-temporal pattern of expression. To investigate what role Xseb4R plays during Xenopus development, we examined the tissue distribution of its transcripts by RT-PCR and whole-mount in situ hybridisation techniques at different stages of development. The first method revealed the presence of Xseb4R transcripts at all stages from the oocyte (indicating a maternal expression of this gene) to late tadpole stage(not shown). However, by whole-mount in situ hybridisation, Xseb4Rtranscripts are first detected at stage 10.5, broadly around the blastopore(Fig. 2A). This discrepancy probably comes from a difference in sensitivity of the techniques. The expression of Xseb4R before stage 10.5 must be too low to be detectable by in situ hybridisation. At this stage, Xseb4R expression is exclusively localised in the mesoderm(Fig. 2B). As development proceeds, Xseb4R-expressing cells arise in three bilateral longitudinal stripes lateral to the dorsal midline within the open neural plate (Fig. 2C). Transverse sections of embryos from developmental stage 14 show a double layer of Xseb4R-positive cells, one corresponding to the mesoderm and the other one to the sensorial layer of the ectoderm where primary neurons are born (Fig. 2D). As neurogenesis proceeds, the same pattern of expression is maintained, as the lateral stripes of Xseb4R-expressing cells converge towards the dorsal midline during the process of neural tube folding (Fig. 2E). At embryonic stage 20, Xseb4R expression clearly follows the formation of the central nervous system (CNS), including the expression in the area designated to form the eye, olfactory placodes,forebrain, midbrain, hindbrain and spinal cord(Fig. 2F). This correlation of Xseb4R expression and CNS formation becomes more evident from tail bud stage 24 onwards. At this stage, Xseb4R, like many members of Delta/Notch pathway, is further expressed in the tail bud(Fig. 2G). In swimming tadpole stage 32 embryos, Xseb4R is strongly expressed in the CNS, and particularly in the subventricular zone of the neural tube, an area containing specified neuroblasts (Fig. 2I).

The expression of Xseb4R in the other germ layers is quite dynamic. As mesodermal derivatives, Xseb4R-expressing cells are found in a group of cells associated with the pronephros and the blood islands(Fig. 2J,L). Xseb4Rexpression is also detected in endodermally derived structures, i.e. in the liver progenitors (Fig. 2K). As development proceeds, Xseb4R expression fades out from the developing liver but is observed in an area associated with the duodenum(Fig. 2M,O), and in a more posterior group of cells that may contribute to rectum/anus formation(Fig. 2N). This expression is lost at around stage 44.5 (Fig. 2O). Beyond embryonic stage 46, no Xseb4R signal is detected in the endodermally derived structures (data not shown).

To better study Xseb4R expression during retinal development, we examined sectioned embryos after whole-mount in situ hybridisation and analysed more carefully its expression in the developing eye. At stages 28 and 32, when most cells in the optic vesicle are proliferating, Xseb4Rexpression is distributed throughout the neural retina(Fig. 3A,B). From stage 34 onwards, Xseb4R expression is no longer observed in the central retina, where neurons start to differentiate. It rather becomes restricted to the margins, where retinoblasts continue to proliferate, and to the lens(Fig. 3C). At stage 40, when all cells in the central retina are postmitotic, Xseb4R expression is observed in the ciliary marginal zone (CMZ), the only region of the retina where retinogenesis is still occurring(Fig. 3D). Xseb4Rexpression is, however, not detected in the most peripheral region of the CMZ(Fig. 3D), where stem cells are present (Dorsky et al.,1995).

In order to determine whether Xseb4R is only expressed in proliferating precursors, we performed in situ hybridisation experiments combined with anti-BrdU immunohistochemistry. In the CMZ, Xseb4R is not expressed in stem cells but is expressed in proliferating cells in the middle region of the CMZ (Fig. 3E,G). In addition, some expression of Xseb4R is detected in a few postmitotic cells in the most central region of the CMZ(Fig. 3G). This expression pattern in the CMZ is very similar to that of neurogenic and proneural genes(Perron et al., 1998; Perron et al., 1999b).

As a RNA-binding protein can play a role in the nucleus, in the cytoplasm,or in both, we wanted to determine the subcellular localisation of XSEB4R protein. With this aim, we constructed a Flag epitope-tagged form of XSEB4R(XSEB4R-FLAG). We co-transfected this construct into retinoblasts at stage 18 using the in vivo lipofection technique(Holt et al., 1990). We then analysed its subcellular localisation by immunostaining with an anti-Flag antibody in the retina. We found that XSEB4R-FLAG is concentrated in the cytoplasm of retinoblasts (cells where the endogenous gene is expressed, see above), whereas only a faint staining is detected in the nucleus(n=177 examined cells; Fig. 3H,J). The subcellular localisation of XSEB4R-FLAG thus suggests that XSEB4R is involved in RNA metabolism regulation at a cytoplasmic level.

During retinal neurogenesis, the different cell types of the retina are born in a sequence that is conserved across species. Retinal ganglion cells are born first, bipolar cells and Müller glial cells last(Holt et al., 1988). As Xseb4R is expressed coincidentally with bHLH genes involved in retinoblast determination and differentiation, it may also play an important role in regulating the determination/differentiation of these cells. To test this hypothesis, we misexpressed Xseb4R in the developing retina by in vivo lipofection of Xseb4R DNA into the optic vesicles of stage 18 embryos. GFP DNA was co-transfected, allowing identification of transfected cells in stage 41 retina, when most cells in the central retina are postmitotic and fully differentiated (Holt et al., 1988). The analysis of retinal sections transfected with Xseb4R and GFP shows that GFP-positive cells are present mainly in the ganglion cell layer and in the photoreceptor layer, while very few positive cells are formed in the inner nuclear layer(Fig. 4B). This is very different from a control retina transfected only with GFP, where inner nuclear layer cells are the most represented cells(Fig. 4A).

We wondered whether the decrease of cells in the inner nuclear layer(amacrine, bipolar, horizontal and Müller cells) after overexpression of Xseb4R was due to massive apoptotic cellular death. To test this hypothesis, we performed a TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling) assay at various developmental stages (34, 38 and 41) and counted the number of apoptotic cells. At stage 34,we counted apoptotic cells in the whole neuroepithelium and found an average of 3.8 apoptotic cells per control retina (n=32 retinas) and 3.7 apoptotic cells per retina transfected with Xseb4R (n=34 retinas). At stage 38 and 41, where cells are organised in layers, we counted apoptotic cells in each of the three cell layers and calculated the percentage of apoptotic cells in the inner nuclear layer. We found that 56%(n=308 cells in 15 retinas) or 49% (n=65 cells in 4 retinas)of apoptotic cells in control retinas reside in the inner nuclear layer compared with 55% (n=578 cells in 18 retinas) or 48% (n=60 cells in 4 retinas) in retinas transfected with Xseb4R at stage 38 or 41, respectively. Therefore this suggests that there is no significant increase in apoptosis following Xseb4R transfection in cells in the inner nuclear layer.

Another hypothesis is that cells supposed to be in this layer have changed their cellular fate in favor of ganglion or photoreceptor cells. To analyse this hypothesis quantitatively, we counted the different types of cells transfected with Xseb4R. We indeed found that overexpression of Xseb4R leads to a significant increase of ganglion cells and photoreceptors at the expense of amacrine, bipolar and Müller cells(Fig. 4C). We confirmed that Xseb4R transfected cells observed in the ganglion cell layer are indeed differentiated ganglion cells by staining with an anti-islet1 antibody(a ganglion cell marker, data not shown). We found that the number of horizontal cells also had a tendency to decrease but this effect was rarely significant (probably due to the low number of horizontal cells) and highly variable from one experiment to another. Nevertheless, our results suggest that overexpression of Xseb4R leads to a proneural-like effect,pushing progenitor cells to differentiate prematurely as ganglion or photoreceptor cells at the expense of late born cells (bipolar and Müller cells). We found exactly the same phenotype when we overexpressed Xseb4R-Flag (Fig. 4C),which demonstrates that the FLAG epitope does not alter the function of the XSEB4R protein, which strengthens the subcellular localisation of XSEB4R-FLAG(see above). Xath3 is a bHLH gene that leads to a very similar effect when lipofected in the retina (Perron et al., 1999b). To compare the strength of their effects, we lipofected side-by-side Xath3 and Xseb4R in the same batch of embryos. Xath3 and Xseb4R both increase photoreceptors and ganglion cells by the same magnitude (data not shown). The XSEB4R homologue, XSEB4, shows a high degree of similarity to XSEB4R(Fig. 1). We therefore wondered whether the specificity of XSEB4R arises from its expression in neural tissue(Xseb4 being mostly expressed in muscle) or from protein-functional differences. We therefore lipofected side-by-side Xseb4R and Xseb4. We found that these two genes cause the same effects in the retina (data not shown), suggesting that XSEB4, at least when overexpressed,can interact with the same targets than XSEB4R.

To further characterise XSEB4R function during retinogenesis, we tried to block its function during retinogenesis. We therefore set up a new protocol to transfect morpholinos (Mo) into retinoblasts in vivo. Mo are antisense oligonucleotides that block the translation of a target gene with a high specificity (Ekker and Larson,2001; Heasman,2002). Recently we have shown that Mo can be efficiently lipofected in retinoblasts and that they do not interfere with retinogenesis under a certain threshold (Ohnuma et al.,2002). In order to determine whether lipofected Mo could indeed block the translation of a target gene, we first tested this technique with Mo directed against the mRNA encoding GFP (GFP Mo). We therefore colipofected GFP Mo, or a standard control Mo, plus a GFP plasmid and then analysed the intensity of the GFP fluorescence in the retina. To prevent any subjectivity in the analysis, we added a filter that decreased the fluorescence light of the microscope, implying that low fluorescent cells would be below the detectable threshold. As a positive control, we also co-transfected a plasmid encoding the CD2 protein. CD2 is a membrane protein(Brown et al., 1987) for which a very good antibody is available. We then analysed GFP fluorescence among lipofected retinas (CD2 positive). If the GFP intensity is low (below the detectable threshold), then cells would be only CD2 positive, whereas if the GFP intensity is normal or high, cells would be both CD2 and CFP positive. We found that GFP intensity in retinas transfected with GFP plus GFP Mo is strongly diminished compared with control retinas transfected with GFP plus a control Mo (Fig. 5A). This experiment thus shows that lipofected Mo can specifically and effectively reduce translation of their target genes in such lipofection experiments, and can therefore be used to block the function of a given gene in the retina.

To block XSEB4R function, we decided to lipofect in vivo Xseb4R Mo(Mo1) into retinoblasts. We first tested the specificity of Xseb4RMo1. For this purpose, we raised a polyclonal antibody against the N-terminal region of XSEB4R (see Materials and methods). On western blots, this antibody indeed recognises the XSEB4R protein, which migrates at 22 kD(Fig. 5B). In order to test the specificity of Xseb4R Mo1, we co-injected into two-cell stage embryos Xseb4R mRNA and Xseb4R Mo1, or a control Mo. We then analysed at stage 10, by western blot, the presence of XSEB4R protein. We found that Xseb4R Mo1 inhibits XSEB4R protein expression but not that of a control protein, α tubulin (Fig. 5B). This result shows that Xseb4R Mo1 indeed specifically and efficiently blocks the translation of Xseb4RmRNA.

In order to obtain a knock down of Xseb4R during retinogenesis, we used two different Mo directed against Xseb4R, namely Xseb4RMo1 and Xseb4R Mo2. Some authors have reported that some Mo can have unspecific effects at a certain threshold concentration that is Mo dependent(Heasman, 2002). These effects may complicate the study of a not yet characterised gene because it is impossible to distinguish between an unknown loss-of-function phenotype and an unspecific effect. We therefore used two Mo directed against two different regions of Xseb4R mRNA sequences (see Materials and methods) and compared their effects. We have previously studied the optimum concentration of a control Mo that does not lead to any toxic effect in the retina(Ohnuma et al., 2002). We therefore targeted a subcritical concentration of Xseb4R Mo1 or Xseb4R Mo2 into the developing retina by in vivo lipofection of optic vesicles in stage 18 embryos together with GFP DNA as a tracer(Fig. 5C). Control embryos were co-lipofected with GFP DNA and a standard control Mo. Transfected retinal cells were counted on stage 41 embryos. Consistent with our overexpression data, the two Xseb4R Mo lead to the same phenotype,characterised by a significant increase of Müller cells at the expense of ganglion cells (Fig. 5C),indicating a change of cell fate specification in which the latest born cells of the retina are promoted at the expense of the earliest born cell type. To rule out the possibility that the decrease of ganglion cells was due to apoptosis, we performed a TUNEL assay at stages 34 and 38, and counted the number of apoptotic cells. At stage 34, we found an average of 3.2 apoptotic cells per control retina (n=37 retinas) and 2.9 apoptotic cells per retina transfected with Xseb4R Mo1 (n=40 retinas). At stage 38, we found that 25% (n=539 cells in 19 retinas) of apoptotic cells in the retina reside in the ganglion cell layer compared with 24%(n=461 cells in 17 retinas) in retinas transfected with Xseb4R Mo1. Therefore, this suggests that there is no significant increase in apoptosis following Xseb4R Mo1 transfection in cells in the ganglion cell layer.

We then wondered whether Xseb4R could also have a proneural-like effect during primary neurogenesis. We therefore injected transcripts encoding XSEB4R unilaterally into two-cell stage embryos, using lacZ mRNA as a tracer. The embryos were collected at stage 15 and the expression of N-tubulin, a neuronal-specific marker, was analysed by whole-mount in situ hybridisation. Surprisingly, overexpression of high concentrations of Xseb4R RNA (>100 pg) inhibits N-tubulin expression(Fig. 6B; 100%, n=85 embryos). However, to the contrary, low dose of Xseb4R RNA (50 pg)promoted the formation of ectopic N-tubulin-positive neuroepithelial cells in the lateral ectoderm (74%, n=35 embryos; Fig. 6C). The lateral band of N-tubulin is indeed expanded laterally. Ectopic N-tubulin-positive cells were never observed in control lacZ-injected embryos (Fig. 6A).

As high doses of Xseb4R RNA were associated with gastrulation defects, we attributed the resulting suppression of neuronal differentiation to an early function of this gene and this hypothesis is supported by the expression of Xseb4R in the involuting mesoderm during gastrulation(Fig. 2A,B). To circumvent this handicap, we generated a Xseb4R-GR construct by fusing the ligand-binding domain of the glucocorticoid receptor (GR) to the C-terminal end of Xseb4R cDNA. RNAs prepared from this plasmid were microinjected at different concentrations and activation of the fusion protein was investigated at a range of developmental stages from blastula stage 9.5 to neurula stage 13.

When 100 pg of Xseb4R-GR transcripts were injected into one or two animal blastomers of 4- to 8-cell stage embryos, and then treated with dexamethasone at developmental stages between 9.5 and 10, a high number (80%, n=34/42 embryos) of the injected embryos still showed a significant reduction of N-tubulin expression(Fig. 6D). However, a high proportion (53%, n=29/55 embryos) of the injected embryos induced at a developmental stage of between 10.5 and 11 showed a significant increase of N-tubulin-positive cells within the territory of primary neurogenesis(Fig. 6E), whereas no increase was observed in the control (n=52)(Fig. 6G). This result strongly suggests that when activated between stage 10.5 and 11, Xseb4R is able to upregulate the process of neuronal differentiation. However, a low proportion of the injected embryos showed no phenotype (10/55, 18%) or showed a reduction (16/55, 29%) of N-tubulin expression (data not shown),indicating that some of the dexamethasone-treated embryos were not at the right competence to signal the proneural activity of XSEB4R.

To gain more insight into the role of Xseb4R during primary neurogenesis, we decided to analyse the effects of its knock down and to compare them with our overexpression data. Various amounts of Xseb4RMo1 were injected into one cell of a two-cell stage embryo. The embryos were collected at stage 15 and the expression of N-tubulin was analysed by whole-mount in situ hybridisation. Overexpression of the lowest dose, 5 ng of Xseb4R Mo1, results in a phenotype that is not significantly different from that observed in embryos injected with a control Mo(Fig. 6J). However, 10 or 20 ng of Xseb4R Mo1 significantly inhibits N-tubulin expression(Fig. 6I,J). Taken together,these results suggest that Xseb4R plays an important role in primary neurogenesis

As we reported above, Xseb4R is expressed in the CMZ of the retina, as well as during primary neurogenesis in the region where many regulators of retinogenesis and primary neurogenesis (such as proneural,neurogenic and differentiation genes) are also expressed. In addition, our functional studies during retinogenesis and primary neurogenesis suggest that Xseb4R is involved in neuronal differentiation. Xseb4Rexpression is therefore likely to be responsive to proneural and neurogenic signalling pathways. We addressed this question directly in vivo by analysing the transcriptional regulation of Xseb4R in response to the activated proneural and neurogenic pathways. mRNAs encoding XNgnr1 or its downstream target, XNeuroD, which also encodes a transcriptional regulator, were injected into one of the two blastomeres of two-cell stage embryos. lacZ mRNA was co-injected as a tracer. Embryos were fixed at neurula stage and stained by X-gal treatment for probe distribution control. Results obtained by whole-mount in situ hybridisation show that Xseb4Rexpression is ectopically activated by XNgnr1 (100%, n=135),suggesting that this gene functions downstream of the neuronal determination bHLH factor (Fig. 7A,B). Similar results were obtained with XNeuroD (78%, n=86; Fig. 7C,D). We then injected 100 pg of Xseb4R-GR transcripts into one of the blastomeres of 4-cell stage embryos, treated the embryos with dexamethasone between stage 10.5 and 11, and analysed the expression of XNgnr1 and XNeuroD. Under these conditions, although N-tubulin expression is upregulated (see above), no ectopic expression of XNgnr1 or XNeuroD was observed (data not shown). Altogether these results indicate that Xseb4R functions downstream of these bHLH factors.

Xseb4R expression, though broadly detected in the open neural plate, exhibits a `salt and pepper'-like pattern, indicating that Xseb4R-expressing cells might be subject to lateral inhibition. To address this question, we analysed Xseb4R expression in ICD-Notch (a constitutively active form of XNotch1)-injected embryos. Our results reveal a reduction of Xseb4R transcription (89%, n=110; Fig. 7E,F),indicating that Xseb4R is regulated, directly or indirectly, by the Delta/Notch signalling pathway. Hence, Xseb4R is an additional component of the proneural and neurogenic signalling pathways involved in neuronal cell differentiation during Xenopus development.

We have identified a new putative RNA-binding protein, XSEB4R, which contains a single RRM domain. Xseb4R is strongly expressed in the nervous system during development. Focusing our study on the retina, we showed that Xseb4R expression is restricted to proliferating retinoblasts and postmitotic differentiating precursors. Overexpression experiments, as well as loss-of-function analysis, converge to suggest that XSEB4R is the first known RNA-binding protein displaying a proneural effect during retinogenesis. We also showed a similar function during primary neurogenesis. Finally, we have positioned Xseb4R in a genetic cascade involved in neurogenesis.

Xseb4R encodes a putative RNA-binding protein containing a single RRM. Some RRM-containing proteins also contain other RNA-binding domains. For example, Vg1-RBP/vera has two RRMs as well as four KH domains (reviewed by Yaniv and Yisraeli, 2002). However, in the XSEB4R protein sequence, besides the RRM, we have not detected any other consensus sequences revealing the presence of another known domain. The C-terminal two-thirds of XSEB4R is somewhat enriched with proline residues, known to be involved in protein-protein interactions (reviewed by Williamson, 1994). A single RRM is known to be sufficient to bind RNA(Scherly et al., 1989), but better target specificity has been demonstrated when several RRMs are present(Kuhn and Pieler, 1996). These data suggest that XSEB4R could act through a multi-protein complex, conferring a higher specificity. The two additional conserved domains, observed in the C-terminal half of the vertebrate SEB4 protein sequence, could be involved in the formation of this protein complex.

RNA-binding proteins can regulate the expression of their mRNA targets at different steps of mRNA processing and translation(Harford and Morris, 1997). We have shown that an epitope-tagged XSEB4R protein is mainly cytoplasmic in retinal progenitors. We therefore propose that this protein could regulate the transport, the stability or the translation of its mRNA targets in the cytoplasm.

We found that Xseb4R is strongly expressed in the developing nervous system. As we have not detected any expression in cranial ganglia, Xseb4R expression in the nervous system seems to be restricted to the CNS, at least during early neurogenesis. However, its expression is not restricted to the nervous system. Xseb4R expression is multiphasic and arises in (1) the mesoderm during gastrulation, (2) the neuroectoderm during neurulation, and (3) different organs of the endoderm during organogenesis. Expression of the related Xenopus gene Xseb4,as Xseb4R, is not restricted to a single tissue. Xseb4 is indeed mainly expressed in muscle but also in the lens(Fetka et al., 2000). These genes seem therefore to be involved in different developmental processes.

In the retina, Xseb4R is expressed in retinoblasts of the developing optic vesicle. To compare its expression with that of other genes involved in retinogenesis, we took advantage of the specific properties of the CMZ. According to the expression of various genes and to cell division activity, the CMZ has been divided into four different regions from the peripheral to the central retina (Perron et al., 1998). Proneural and differentiation genes are not expressed in the first zone of the CMZ, which contains retinal stem cells, but in proliferating retinoblasts and some postmitotic neurons(Perron et al., 1998). We showed that Xseb4R is also expressed in the CMZ, in both proliferating retinoblasts and postmitotic neurons, and is excluded from the most peripheral region. This expression correlates nicely with the expression of Xseb4R in the subventricular zone of the neural tube during primary neurogenesis, where neurons are in the transition step between proliferation and differentiation. Some differentiation genes, such as XNeuroD or Xath3, are also expressed in a subset of neurons in the central retina (Perron et al.,1998). Xseb4R expression, however, is restricted to the CMZ in the mature retina, like some proneural genes such as Xash1,Xash3 and XNgnr1, or the atonal-like gene Xath5. The expression of neurogenic genes belonging to the Notch/Delta signalling cascade is also similarly restricted to the CMZ. The expression of Xseb4R in this region of the CMZ together with several proneural and neurogenic genes suggests that it is involved in crucial steps of retinogenesis, where precursor cells become determined and differentiate into a particular type of retinal neuron or glial cell.

In Xenopus, several strategies to reveal a loss-of-function effect have been used. For example, fusion constructs between the DNA-binding domain of a transcription factor and an activator domain (VP16), or a repressor domain (Engrailed), have been used extensively (e.g. Mariani and Harland, 1998). Ectopic expression of dominant-negative variants is another alternative. This is often the case for transmembrane receptors (e.g. McFarlane et al., 1996). Researchers working on RNA-binding proteins belonging to the ELAV family have constructed dominant-negative versions of these proteins. This was possible because ELAV-type proteins contain three RRMs(Robinow et al., 1988). It has indeed been proposed that truncated constructs, missing one or two RRMs,behave as dominant-negative constructs(Akamatsu et al., 1999; Kasashima et al., 1999; Anderson et al., 2000). However, because XSEB4R only contains one RRM we could not use such a strategy.

Mo have recently been used in developmental studies in a wide range of model organisms to block the translation of a target mRNA(Ekker and Larson, 2001; Heasman, 2002). So far, in Xenopus, Mo have been used in blastomere injection experiments. In order to target a single tissue where we expect a loss-of-function phenotype,we set up a protocol to lipofect Mo in vivo into specific regions(Ohnuma et al., 2002). In this paper, we demonstrate the efficiency of this Mo lipofection strategy in the retina. Our results with Xseb4R Mo suggest that this gene is required for ganglion cell production. Whether Xseb4R specifically promotes ganglion cells or simply promotes neurogenesis remains to be determined. Indeed, because ganglion cells are the first cells to be born in the retina,one cannot distinguish between these two hypotheses. Nevertheless, it is important to note that the Xseb4R knockdown phenotype reflects the opposite phenotype, as obtained in the gain-of-function experiments. Indeed when we overexpress Xseb4R in retinoblasts, they tend to differentiate precociously. It has indeed been shown previously that when precursors are forced to adopt an early fate, an increase in ganglion and cone photoreceptor is observed (Dorsky et al.,1997). When they are forced to differentiate slightly later an increase in the number of cone and rod photoreceptors is observed(Dorsky et al., 1997). Although overexpression of Xseb4R induces a severe phenotype,significantly affecting almost all cell types, the Xseb4R knockdown phenotype is less severe, as photoreceptor cells, amacrine and bipolar cells are not affected. This suggests that the Mo strategy does not lead to a complete loss of function, but rather reduces the amount of XSEB4R protein in vivo. Alternatively, this could be due to a redundant function of another RNA-binding protein partly compensating for the absence of XSEB4R. Nevertheless, our knockdown experiment, together with our gain-of-function experiment, strongly suggest that Xseb4R promotes neurogenesis during retinal development.

So far, mainly transcription factors have been found to have such an effect in retinogenesis. Indeed, overexpression of Xath3 in the retina leads to exactly the same phenotype as Xseb4R overexpression (this study)(Perron et al., 1999b). This paper thus provides the first example of an RNA-binding protein displaying proneural properties in the retina.

During primary neurogenesis, we found that Xseb4R could also have a proneural-like effect, i.e. activation of N-tubulin expression when overexpressed, only at a particular dose or at a particular time during development. Such discrepancy in Xseb4R expression pattern(Notch/Delta-like pattern) and resulting proneural-like function has already been observed with the Hes6 gene(Koyano-Nakagawa et al.,2000). Furthermore, functional characteristics of XBF1, an anterior neural plate-specific winged helix transcription factor, just like the data that we report on Xseb4R overexpression, revealed ectopic N-tubulin expression at low doses and inhibition of this same marker at high doses (Hardcastle and Papalopulu,2000). These differential effects of XBF1 were found to correlate with its role in cell proliferation. In the context of our studies, we did not know which early functions could be mediated by Xseb4R. However,because we saw gastrulation defects when a high dose of Xseb4R was injected, one hypothesis is that it interferes with early development,preventing us from analysing its specific effect during neurogenesis. By using an inducible construct we managed to overcome this problem and to analyse the effect of Xseb4R on early neurogenesis. We indeed found that when activated towards the end of gastrulation, i.e. between stage 10.5 and 11,just like with low doses of Xseb4R, Xseb4R-GR promotes ectopic neurogenesis as well. This effect is in accordance with the Xseb4Rknockdown effect, which inhibits neuronal differentiation. Therefore, these results suggest that during primary neurogenesis Xseb4R has, as in the retina, proneural properties.

As a result of Xseb4R expression in the CMZ and of our functional analysis, we wanted to identify genes belonging to the genetic cascade involved in neurogenesis that could regulate Xseb4R expression. We found that the proneural gene XNgnr1 is able to induce strong ectopic expression of Xseb4R in the whole ectoderm. We have also shown that the atonal-like gene XNeuroD (a differentiation gene) also induces ectopic expression of Xseb4R. In addition, our analysis reveals that Notch/Delta signalling negatively regulates Xseb4Rexpression. This is similar to the inhibition found for several bHLH proneural genes, such as XNgnr1 (Chitnis and Kintner, 1996). Altogether, these results suggest that Xseb4R is a component of the genetic cascade involved in neurogenesis. It would be interesting now to characterise genes that function downstream of Xseb4R during neurogenesis and that may be post-transcriptionally regulated by this cytoplasmic RNA-binding protein.

It has recently been reported that Xenopus NeuroD is regulated post-translationally by the kinase GSK3β(Moore et al., 2002). This observation, together with ours, illustrates the fact that post-transcriptional and post-translational regulators may have a crucial role in neurogenesis. These regulatory mechanisms, in contrast to transcriptional gene regulation, have been poorly studied so far during vertebrate neurogenesis.

We would like to thank E. Bellefroid, R. Rupp and C. Kintner for plasmids,S. Thor for the anti-islet antibody. We are also grateful to Karl Johnson for comments on the manuscript. We would also like to thank J. Hamdache and L. Sinzelle for technical assistance. This work was supported by a grant from the DFG (CMPB Göttingen) to T.P. and an EC Biotechnology grant(QLG3-CT-200101460), an ARC grant (5749) and a Retina France Grant to M.P. The EC Biotechnology grant supports M.A.A.