Xrx1 controls proliferation and neurogenesis in Xenopusanterior neural plate

Genetic studies in Drosophila have been instrumental in identifying molecular pathways involved in the control of neurogenesis. This process is initially regulated by positional information provided by prepatterning genes that control the site-specific expression of proneural genes (Gomez-Skarmeta et al.,1996). The role of proneural genes, which code for transcription factors of the basic helix-loop-helix (bHLH) class, is to define clusters of cells competent to originate neuronal precursors(Chitnis, 1999). Within each cluster, a single cell is selected to become a neuroblast by the expression of neurogenic genes. This is accomplished by lateral inhibition, a process mediated by the membrane-bound ligand Delta and the Notch receptor. One cell in the cluster becomes committed to a neuronal fate by expressing higher levels of Delta, which in turn activates the Notch receptor in adjacent cells,forcing them to remain uncommitted. The search for vertebrate homologues of proneural and neurogenic genes has led to the discovery that key regulators of neurogenesis are evolutionary conserved(Chitnis, 1999). In particular, many studies on vertebrate neural induction and neurogenesis have been performed in the amphibian Xenopus laevis because of the experimental accessibility of its embryos and the small number of early differentiating primary neurons. In this species, it was shown that the dorsal mesodermal region called Spemann's organizer(Spemann, 1938) neuralizes the dorsal ectoderm by secreting noggin, chordin and follistatin which antagonize BMP4, an epidermalizing signal (Sasai and De Robertis, 1997). Following neural induction, the spatial distribution of neuronal precursors within the posterior neural plate is controlled by prepattern genes of the Xiro, Gli and Zic families. In particular, Zicr1, Xzic3 and Gli proteins induce neurogenesis, while Zic2,Xiro1, Xiro2 and Xiro3, acting as anti-neurogenic transcription factors,restrict the expression domains of proneural genes(Nakata et al., 1997; Mizuseki et al., 1998; Brewster et al., 1998; de la Calle-Mustienes et al.,2002). Regulators of these transcription factors include the Hedgehog genes, which promote proliferation by repressing Gli3 and activating Zic2, and retinoic acid (RA), a posteriorizing morphogen that induces neuronal differentiation by inhibiting the expression of Hedgehog genes (Franco et al., 1999). As early as the end of gastrulation, the first sites of neurogenesis within the neural plate are marked by the expression of X-ngnr-1, which encodes an atonal type bHLH protein, followed by the activation of X-Delta-1 and finally of N-tubulin, a marker of differentiated neurons. The expression of these genes is restricted to three longitudinal rows on either side of the dorsal midline, where individual cells are selected for differentiation through the action of Delta/Notch(Chitnis, 1999).

Although the molecular mechanisms underlying the control of neurogenesis in the posterior nervous system are beginning to be unraveled, less is known about factors controlling neuronal differentiation in the anterior neural plate. Lineage tracing and pulse-labeling experiments(Hartenstein, 1989; Eagleson et al., 1995), as well as analysis of neuronal differentiation markers(Hartenstein, 1993; Papalopulu and Kintner, 1996),have shown that anterior neural plate cells undergo neuronal differentiation significantly later than cells of the posterior neural plate. An as yet unanswered question is what are the factors controlling this phenomenon and how they are related to regulators of posterior vertebrate neurogenesis. So far, only a small group of transcription factors expressed in the anterior neural plate, including XBF-1, Xanf-1, Xsix3 and Xoptx2,have been shown to play a role in delaying neuronal differentiation and/or promoting proliferation (Bourgouignon et al., 1998; Ermakova et al., 1999; Zuber et al., 1999; Bernier et al., 2000; Hardcastle and Papalopulu,2000). However, because the spatiotemporal expression of these genes does not coincide with the entire proliferative region of the anterior neural plate, additional genes are likely to be involved. In this work, we propose that Xrx1, a homeobox gene required for eye and anterior brain development, is one such factor. We report that Xrx1 is expressed in the entire proliferative anterior neural plate surrounded by cells expressing X-ngnr-1, X-Delta-1 and p27Xic1, a cell cycle inhibitor. Xrx1 microinjection inhibits X-ngnr-1,X-Delta-1 and N-tubulin expression, and counteracts RA- and X-ngnr-1-mediated differentiation, while at the same time activating proliferation. These effects are independent of Notch signaling and are restricted to the most rostral region of the embryo. Xrx1 exerts its function by activating Xhairy2 and Zic2, the expression of which in the anterior neural plate overlaps with that of Xrx1, and by repressing p27Xic1. Accordingly, loss-of-function experiments show that Xrx1 is required for the normal proliferation of the anterior neural plate. These data indicate that Xrx1 possesses the appropriate activities and spatiotemporal expression pattern to be one of the factors responsible for the maintenance of anterior neuronal precursors in a proliferative state.

Xenopus embryos were generated and staged as described(Nieuwkoop and Faber, 1967; Newport and Kirschner, 1982). Whole-mount in situ hybridization on embryos and animal caps was performed essentially as described by Harland(Harland, 1991). Histological examination was performed according to Casarosa et al.(Casarosa et al., 1997). For HUA treatment, stage 10 devitellinised embryos were added to a 20 mM hydroxyurea, 150 μM aphidicolin in 0.1× MMR solution, as described by Hardcastle and Papalopulu (Hardcastle and Papalopulu, 2000), and kept in this solution until fixation.

Capped synthetic RNAs encoding for Xrx1 (20-100 pg), X-chh (1 ng), X-shh (1 ng)(Ekker et al., 1995), X-ngnr-1 (40 pg) (Ma et al.,1996), XRALDH2 (1.5 ng)(Chen et al., 2001), Notch-ICD (30 pg-1.8 ng) (Chitnis et al., 1995), X-Delta-1^stu (500 pg)(Chitnis et al., 1995), XBF-1 (150 pg) (Bourgouignon et al., 1998) and Xhairy2 (125 pg) (Davis et al., 2001) were generated by in vitro transcription and co-injected with lacZRNA(100-500 pg) into one blastomere at the two-cell stage or into a dorsal blastomere at the four-cell stage. The optimal concentration of each batch of RNA was identified through injection of various doses followed by analysis of either the phenotype or the expression of specific markers. For animal cap experiments, capped synthetic chordin (150 pg per blastomere)(Sasai et al., 1995), X-ngnr-1 (40 pg per blastomere) and Xrx1 (360 pg per blastomere) RNAs were injected into both blastomeres at the two-cell stage and animal caps dissected at stage 9. When sibling control embryos reached stage 16 or 17, animal caps were fixed and stored in ethanol at -20°C. For retinoic acid treatment, injected animal caps were incubated in 2×10^-6 M RA in 0.5×MMR where they were cultured until stage 16. For the experiments shown in Fig. 2I-L, Fig. 3E, Fig. 5P,Q, the total amount of RNA injected, either in the experimental or in the respective control samples,is the same. This was achieved by adjusting the amount of lacZ RNA in the control samples. The Xrx1 antisense morpholino used was:5′-TCAGGGAAGGGCTGTGCAGGTGCAT-3′ (Gene Tools LLC). A standard morpholino oligo (Gene Tools LLC) was injected as control. Immunostaining with anti-phosphorylated H3 antibody was performed as described by Saka and Smith(Saka and Smith, 2001). BrdU incorporation was performed essentially as described by Hardcastle and Papalopulu (Hardcastle and Papalopulu,2000).

Xrx1 is a homeobox gene initially expressed in the anterior neural plate in territories fated to give rise to the retina, diencephalon and part of the telencephalon. Its overexpression induces overgrowing of the neural retina, pigmented epithelium and anterior neural tube, while inhibition of its function leads to a strong reduction or absence of the eye and anterior brain(Casarosa et al., 1997; Mathers et al., 1997; Andreazzoli et al., 1999). As cells of the anterior neural plate are characterized by prolonged proliferation and delayed neuronal differentiation, we decided to analyze whether Xrx1 is involved in the control of these activities. To better define Xrx1 expression in the context of early neurogenesis,we compared the expression domain of Xrx1 with that of early neuronal differentiation markers. Double in situ hybridization experiments showed that at early neurula Xrx1 is expressed in a territory that is precisely circumscribed by the expression of the neurogenic gene X-Delta-1 and abuts on the anterior expression domains of the neuronal determination gene X-ngnr-1 (Fig. 1A,B). Histological sections showed that within this area Xrx1 is expressed exclusively in the deep sensorial layer of the neuroectoderm where both primary and secondary neurons will form(Hartenstein, 1989)(Fig. 1C). Regions marked by X-Delta-1 and X-ngnr-1 expression coincide with prospective sites of neuronal differentiation. In particular, the most anterior semicircular stripes of X-Delta-1 and X-ngnr-1 correspond to the presumptive olfactory placodes, part of the telencephalon and laterally to the epiphysis, while the more posterior X-Delta-1 medial stripe coincides with the ventral midbrain(Eagleson and Harris, 1990; Bourguignon et al., 1998). Although expression of several other genes partially overlaps with that of Xrx1, to our knowledge Xrx1 is the only gene described so far whose expression completely fills the anterior gap of X-Delta-1expression, thus corresponding exactly to the proliferative region of the anterior neural plate.

We next investigated which factors localize Xrx1 expression to the proliferative region of the anterior neural plate. Because the Hedgehog genes were shown to induce proliferation and delay differentiation in the early neural plate (Franco et al.,1999), we decided to test if they affect Xrx1 expression. We found that both sonic hedgehog (X-shh) and cephalic hedgehog (X-chh) are able to activate Xrx1 at early neurula stage (X-shh 68%, n=44; X-chh 56%, n=36; Fig. 1D,E). Xrx1 is ectopically activated by Hedgehog signaling only in an area that surrounds the endogenous Xrx1 expression domain, despite of the broader distribution of the injected RNA. Because of the lack of Xrx1expression in X-Delta-1 and X-ngnr-1 positive regions, we looked if proneural gene expression plays a role in restricting Xrx1expression. Overexpression of X-ngnr-1, which efficiently induces ectopic N-tubulin expression (100%, n=31; Fig. 1G), strongly represses Xrx1 expression (100%, n=32; Fig. 1F). These data indicate that Xrx1 expression is not compatible with neuronal differentiation and that the anterior expression of proneural genes like X-ngnr-1defines the perimeter of the Xrx1 expression domain.

The coincidence of Xrx1 expression with the proliferating area of the anterior neural plate, where neuronal markers are not expressed, led us to think that Xrx1 might be part of the system preventing precocious neurogenesis in this area. To test this hypothesis, we analyzed the expression of neuronal differentiation markers in Xrx1-injected embryos during early neurulation. We observed that X-ngnr-1, X-Delta-1 and N-tubulin are all repressed in the anterior region by Xrx1overexpression (X-ngnr-1, 91%, n=58; X-Delta-1,90%, n=31; N-tubulin, 97%, n=92; Fig. 2A-D), while the repressive effects are weak in the posterior expression domains of these markers (Fig. 2A-C,E). Sox2, a general neural marker, was not affected at this stage (0%, n=90; Fig. 2F),indicating that Xrx1 acts on neuronal differentiation but not on neural induction. As a positive control, Xrx1 ectopically activates XBF-1 in the lateral border of the neural plate (58%, n=24; Fig. 2G), as previously described (Andreazzoli et al.,1999). Xrx1 effects on neurogenesis are distinct from those observed upon XBF-1 overexpression(Bourguignon et al., 1998). In fact, injection of XBF-1 at doses that cause suppression of endogenous N-tubulin also leads to ectopic activation of N-tubulin along the boundary of the injected area in the posterior neural plate (94%, n=36; Fig. 2H). In a complementary approach, we tested if Xrx1 has the ability to inhibit ectopic neurogenesis induced by X-ngnr-1overexpression. Injection of X-ngnr-1 induced a massive expression of N-tubulin, the in situ signal of which covered the β-gal staining (100%, n=41; Fig. 2I, also compare the injected versus uninjected side in Fig. 1G). At variance,co-injection of X-ngnr-1 and Xrx1 resulted in a considerable attenuation of N-tubulin activation (95% with reduced ectopic expression, n=45; Fig. 2K). These data were confirmed by animal cap experiments where X-ngnr-1 ability of inducing N-tubulin(Ma et al., 1996) was inhibited by Xrx1 (X-ngnr-1 + lacZ 100% positive, n=58; X-ngnr-1 + Xrx1 96% negative, 4% weakly positive, n=62; Fig. 2J,L).

Retinoic acid has been shown to control the timing of neuronal differentiation being able to accelerate neurogenesis in anterior neural cells(Papalopulu and Kintner, 1996; Sharpe and Goldstone, 2000). Although RA is thought to function mainly in the posterior neural plate and mesoderm during early development (Chen et al., 1994), it has been shown recently that XRALDH2, one of enzymes involved in RA synthesis, is expressed also in an anterior site(Chen et al., 2001). A double in situ hybridization revealed that Xrx1 expression adjoins, but does not overlap, the XRALDH2 anterior expression domain(Fig. 3A). To determine the causes of this spatial relationship between Xrx1 and XRALDH2, we looked at the effect that the overexpression of each of these genes exerts on the other. We found that Xrx1 and XRALDH2 exhibit mutually repressive activities(XRALDH2-injected embryos: 75% with reduced Xrx1 expression, n=24; Fig. 3B; Xrx1-injected embryos: 83% with reduced XRALDH2 expression, n=24; Fig. 3C), which could explain the generation of adjacent, non-overlapping expression domains. To analyze if Xrx1 could also counteract the effects of RA on neuronal differentiation, we took advantage of an animal cap system that recapitulates anterior neurogenesis. Papalopulu and Kintner(Papalopulu and Kintner, 1996)showed that noggin-injected animal caps cultured until neurula stage(stage 16) express NCAM but not N-tubulin. Initiation of N-tubulin expression is observed only when these explants are cultured until tailbud stage (stage 27); addition of RA accelerates this process leading to activation of N-tubulin by stage 16. We used chordin as a BMP antagonist in animal caps, and found that at stage 16 chordin alone does not induce N-tubulin (0% positive, n=26), while RA treatment of chordin-injected animal caps robustly activates a punctate expression of N-tubulin (97% positive, n=33; Fig. 3D). Interestingly, Xrx1 expression, which is strongly induced in chordin-injected animal caps (96% positive, n=25), is completely abolished by RA treatment (0% positive, n=26; Fig. 3D). This observation again inversely correlates Xrx1 expression and neurogenesis. To test whether the repression of Xrx1 was required to activate neurogenesis we treated with RA animal caps that had been co-injected with chordinand Xrx1 RNA. Co-injection of Xrx1, but not of lacZ, effectively inhibited N-tubulin expression in RA-treated caps (Chordin + lacZ 100% positive, n=25; Chordin + Xrx1 80% negative, 20% weakly positive, n=30; Fig. 3E). Thus, Xrx1 appears to counteract RA-mediated neuronal differentiation through a dual action: upstream of RA production by repressing the expression of XRALDH2, and downstream of RA or acting on a parallel pathway, as shown by its ability to impede RA-mediated differentiation in chordin-injected caps.

Cells of the anterior neural plate are not only characterized by delayed differentiation but also display a protracted proliferating state. Therefore,we asked whether Xrx1 plays a role in the control of proliferation at early neurula stage. To achieve this, embryos injected with either Xrx1 or Xrx1-EnR, a dominant-negative construct(Andreazzoli et al., 1999) were tested for BrdU incorporation. In these experiments the number of BrdU-positive cells in the injected side was compared with that of the uninjected control side taking also into account their anteroposterior distribution. We found that the anterior neural plate of Xrx1-injected embryos displayed a 44% increase of BrdU-positive cells in the injected side compared with the control side (an average of 42.2 positive cells per section in the injected side, n=1099, versus 29.2 cells per section in the control side, n=760; P<0.001; Fig. 4A,C). On the contrary, no significant difference was detected in the posterior neural plate of Xrx1-injected embryos (an average of 19.6 positive cells per section in the injected side, n=413, versus 19.3 cells per section in the control side, n=406; Fig. 4B,C). However, the anterior neural plate of Xrx1-EnR-injected embryos shows a 33% decrease of BrdU-positive cells in the injected side compared with the control side (an average of 18 positive cells per section in the injected side, n=519, versus 26.8 cells per section in the control side, n=773; P<0.001; Fig. 4D,F). In addition, no significant difference was observed in the posterior neural plate (an average of 14.3 positive cells per section in the injected side, n=253,versus 13.9 cells per section in the control side, n=235; Fig. 4E,F). Altogether, these results suggest that Xrx1 is involved in controlling cell proliferation specifically in the anterior region of the neural plate.

To gain insight into how Xrx1 may exert its effects, we analyzed the expression of potential Xrx1 target genes in injected embryos. The criteria that we used to select the genes tested are the following: (1)the gene has to be expressed in an area where Xrx1 is also expressed,or in a surrounding region affected by Xrx1 overexpression; (2) it must play a role in controlling cell differentiation and/or proliferation. Zic2, a gene encoding a zinc-finger transcription factor, has been shown to have an antineurogenic function(Brewster et al., 1998). Besides being expressed in the posterior neural plate in stripes that alternate with the primary neurons, Zic2 is also expressed in the anterior neural plate. Comparing the expression profiles of Xrx1 and Zic2, we found that they mostly overlap in the presumptive forebrain,although Zic2 RNA extends slightly more to the anterior and Xrx1 expression more to the posterior(Fig. 5A,B). Overexpression of Xrx1 leads to an expansion of Zic2 expression, which extends along the mediolateral axis (95%, n=62; Fig. 5D,E), while no ectopic expression is observed in the majority of cases in the posterior neural plate. Xhairy1 and Xhairy2 are homologues of the Drosophila hairy gene that are known to act as transcriptional repressors and to inhibit neuronal differentiation (Dawson et al., 1995; Koyano-Nakagawa et al., 2000). Both genes display diffuse expression in the anterior neural plate that overlaps with Xrx1 expression(Fig. 5C,G). Moreover, Xhairy2, but not Xhairy1, shows a stronger stripe of expression that coincides with the anterior and lateral borders of Xrx1 expression. Injection of Xrx1 induces ectopic expression of Xhairy2, which, as in the case of Zic2, does not extend posteriorly (72%, n=50; Fig. 5F). In addition,overexpression of Xrx1 also leads to a repression of the anterior domain of Xhairy1 (90%, n=40; Fig. 5J). As Xhairy1and Xhairy2 are known to repress each other, and because there is evidence that Xrx1 might work as a transcriptional activator(Andreazzoli et al., 1999), it is possible that Xhairy2 is first activated by Xrx1 and that this excess of Xhairy2 could be responsible for the downregulation of Xhairy1. Xoptx2, a homeobox gene of the Six homeobox family, has been shown to play a role in controlling the proliferation of the retina(Zuber et al., 1999). Analysis of Xoptx2 expression in Xrx1-injected embryos showed that,although at stage 14 the expression of this gene does not appear to be affected (89% normal expression, 11% slightly reduced expression, n=67; not shown), a remarkable expansion is observed at stage 18(75%, n=56; Fig. 5L). This response is similar to the one previously described for the related gene Xsix3 (Andreazzoli et al.,1999). p27Xic1, an inhibitor of cyclin/cdk required for primary neurogenesis (Su et al.,1995; Carruthers et al.,2003; Vernon et al.,2003), is expressed anteriorly in two semicircles very similar to those characterized by X-Delta-1 expression. We observed that Xrx1 expression is complementary to that of p27Xic1, being delimited dorsally and ventrally by the two p27Xic1 expression domains (Fig. 5H). Overexpression of Xrx1 suppresses p27Xic1 expression,particularly in its most anterior domains (96%, n=83; Fig. 5K). Finally, we looked at cyclin D1, which has a strong expression site in the eye field(http://www.xenbase.org/)(Vernon and Philpott, 2003)and has been implicated in mouse eye development(Fantl et al., 1995). Although Xrx1-injected embryos did not show any difference in cyclin D1 expression at early neurula stage (stage 13, 100% normal expression, n=46; not shown), a notable expansion was observed at late neurula stage (stage 18, 83%, n=24; Fig. 5I). Thus, Zic2 and Xhairy2 are activated and p27Xic1 is repressed by Xrx1 at early neurula stage, while Xoptx2 and cyclin D1 are activated only at a later stage under the influence of Xrx1.

X-Notch-1 is expressed in the anterior neural plate in the region occupied by Xrx1 (Fig. 5M), and its activity in preventing differentiation of neurons has been described (Chitnis et al.,1995). Moreover, the mouse Xrx1 homologue (Rx1,also called Rax) has been shown to activate Notchtranscription during retinogenesis(Furukawa et al., 2000). For these reasons, we tested whether, during early neurulation, Xrx1 and X-Notch-1 affect one another's expression. Analysis of Xrx1-injected embryos at various stages during early neurulation failed to show any transcriptional activation of X-Notch-1 (stage 13,0%, n=54; stage 15, 0%, n=33; stage 18, 0%, n=38; Fig. 5N). Furthermore,injection of several doses of a constitutively active form of X-Notch-1 (Notch-ICD)(Chitnis et al., 1995) did not show activation of Xrx1 at early neurula (30 pg 86% normal, 10%slightly reduced expression, 4% slightly expanded expression, n=30;500 pg 84% normal, 8% slightly reduced expression, 8% slightly expanded expression, n=37; 1.8 ng 85% normal, 11% slightly reduced expression,4% slightly expanded expression, n=27; Fig. 5O). These results suggest that during early neurulation, Xrx1 is not affected directly by Notch signaling, and that Xrx1 does not affect Notch expression. Finally, to test if the effects of Xrx1 on neurogenesis are mediated by lateral inhibition, we co-injected Xrx1 with X-Delta-1^stu, an antimorphic version of X-Delta-1(Chitnis et al., 1995). Even though under these conditions, lateral inhibition was blocked, as shown by an excess of N-tubulin-expressing cells in the posterior neural plate(Fig. 5P,Q, arrow), Xrx1 was still able to repress the trigeminal ganglia expression of N-tubulin (75% absent expression, 25% reduced expression, n=40; Fig. 5Q).

The ectopic expression of Xhairy2 and Zic2 as well as the repression of X-ngnr-1 and p27Xic1 in Xrx1-injected embryos could be triggered independently of cell proliferation or,alternatively, could result from an expansion of the proliferating neuroectoderm. To distinguish between these two possibilities, we asked whether Xrx1 can affect Xhairy2, Zic2, X-ngnr-1 and p27Xic1 expression in embryos where cell division has been blocked by treatment with hydroxyurea and aphidicolin (HUA)(Harris and Hartenstein,1991). HUA treatment severely affected anti-phoshoH3 staining, a marker of cells in mitotic prophase, as well as Xoptx2 ability of expanding Xrx1 (Zuber et al.,1999) and resulted in smaller embryos with reduced optic vesicles(Fig. 6E-I). Under these conditions, Xrx1 is still able to expand Zic2 (96%, n=78) and Xhairy2 (71%, n=74) expression and to repress the expression of X-ngnr-1 (97%, n=36) and p27Xic1 (84%, n=44), although not to the same extent as in untreated embryos (Fig. 6A-D). These data suggest that the regulation of Xhairy2, Zic2, X-ngnr-1 and p27Xic1 observed in Xrx1-injected embryos does not depend exclusively on proliferation. This observation may be consistent with the finding that Xrx1 is able to convert competent ectoderm to an anterior neural fate (Kenyon et al.,2001).

As Xrx1 microinjection expands Zic2 and Xhairy2expression even in HUA-treated embryos, we tested whether Xrx1 could activate these genes in non-neuralized ectoderm. To achieve this, we analyzed the expression of Zic2 and Xhairy2 in Xrx1-injected animal caps. At the same time, we also looked at the expression of XBF-1, a gene activated by Xrx1 in the lateral border of the anterior neural plate, Sox2, a general neural marker, and XK81, an epidermal marker (Fig. 7). Control uninjected caps showed no expression of Xhairy2,Zic2, XBF-1 and Sox2 (0% in all cases; Xhairy2, n=39; Zic2, n=32; XBF-1, n=33; Sox2, n=41), while they expressed XK81 (100%, n=20). By contrast, Xrx1-injected animal caps expressed Xhairy2 (60% positive,21% weakly positive, n=72) and XK81 (100%, n=32)while showing no activation of Zic2, XBF-1 and a weak activation of Sox2 (Zic2, 0%, n=65; XBF-1, 0%n=37; Sox2, 8% weakly positive, n=41). Thus, among the genes activated by Xrx1 in the early neurula embryos, Xhairy2 appears to be the only one whose induction by Xrx1is independent of ectoderm neuralization.

To further analyze the requirement of Xrx1 function in controlling the expression of genes involved in anterior proliferation and neurogenesis,we used, as an alternative to the Xrx1-EnR construct, an antisense morpholino approach (Heasman et al.,2000). Injection of 10 ng of a morpholino oligo directed against Xrx1 (MoXrx1) into dorsoanimal blastomeres at the eight-cell stage generates embryos displaying severe reduction of eyes and anterior head(97%, n=102; Fig. 8A). This phenotype, which is very similar to the one obtained after Xrx1-EnR injection and to Rx1^-/- mice(Mathers et al., 1997; Andreazzoli et al.,1999), is completely rescued by co-injection of 80 pg of Xrx1 RNA (99%, n=86; Fig. 8B) and is not observed upon injection of control morpholino (96% normal embryos, n=75; Fig. 8C). Analysis of X-ngnr-1 expression in both MoXrx1- and Xrx1-EnR-injected embryos showed an anterior expansion of the expression domains of this gene, which tend to fuse medially. This phenotype is first observed at stage 14 (Xrx1-EnR: 82% n=34; MoXrx1: 83%, n=79; Fig. 8D-F) and is maintained at stage 18 (Xrx1-EnR: 75%, n=28; MoXrx1: 78%, n=38; Fig. 8H-J). The expanded expression domains of X-ngnr-1 appeared to correspond to the telencephalon rather than the trigeminal ganglia. This was confirmed by the observation that MoXrx1 did not affect the expression of FoxD3, a marker of neural crest cells that contribute also to the trigeminal ganglia (100%, n=19; Fig. 8G,K). To analyze simultaneously the effects of Xrx1 knockdown on neurogenesis and eye-field specification we co-hybridized MoXrx1-injected embryos with X-ngnr-1 and Xrx1 (Fig. 8L-N). We found that the medial expansion of X-ngnr-1correlates with a smaller Xrx1 expressing area,which in extreme cases was totally abolished, with no overlap of the two markers (85% reduced Xrx1 expression, Fig. 8M; 15% absent Xrx1 expression, Fig. 8N, n=33). Analogous results were also obtained using Xsix3 as eye field marker (not shown). A very similar phenotype was observed in HUA-treated embryos (100% reduced Xrx1 expression, n=25; Fig. 8O),suggesting that the effects observed in Xrx1 loss of function are mainly due to a reduced proliferation of the anterior neural plate. As gain-of-function experiments indicated a potential role of Xhairy2downstream of Xrx1, we tested whether Xhairy2 could rescue MoXrx1 effects. Bilateral injection of MoXrx1 resulted in reduction, but not complete abolishment, of the expression of Xhairy2(98%, n=50; Fig. 8P,Q), whereas Xhairy2 overexpression led to a severe repression of X-ngnr-1 expression (67%, n=36; Fig. 8R). Co-injection of MoXrx1 and Xhairy2 resulted in embryos displaying the typical anterior expansion of X-ngnr-1 observed in embryos injected with MoXrx1 alone, thus indicating that Xhairy2 is not able to rescue the MoXrx1 phenotype (84%, n=94; Fig. 8S). Among the genes activated by Xrx1 overexpression and repressed by Xrx1inactivation (Andreazzoli et al.,1999), XBF-1 is of particular interest, because, like Xrx1, it acts by controlling p27Xic1 expression(Hardcastle and Papalopulu,2000). Therefore, we tested whether XBF-1 could rescue the effects of MoXrx1 injection. Overexpression of XBF-1alone repressed anterior X-ngnr-1 expression while inducing ectopic X-ngnr-1 in the posterior neural plate (90%, n=30; Fig. 8T,U). When co-injected with MoXrx1, XBF-1 was able to restore a wide X-ngnr-1-free area in the anterior neural plate (82%, n=70; Fig. 8V). These results suggest that XBF-1 might work downstream of Xrx1 and/or that these genes function by controlling common mechanisms.

Comparing the expression domain of Xrx1 with those of genes involved in promoting neuronal differentiation or blocking cell proliferation,we observed that Xrx1 expression is complementary to the expression of X-ngnr-1, X-Delta-1, XRALDH2 and p27Xic1. In particular, Xrx1 completely fills the area surrounded by the anterior domain of X-Delta-1. Other transcription factors known to control neurogenesis in this area include XBF-1, Xoptx2 and Xanf-1 (Bourgouignon et al., 1998; Ermakova et al.,1999; Zuber et al.,1999). XBF-1 is expressed only in the most anterior border of the neural plate, adjacent to the rostral expression of X-Delta-1 and p27Xic1 (Bourgouignon et al., 1998; Hardcastle and Papalopulu,2000), where it plays a crucial role in defining the border between proliferative and non-proliferative areas. Xoptx2 is expressed in the anterior neural plate later than Xrx1(Zuber et al., 1999) and Xanf-1 is expressed in a domain broader than the region occupied by Xrx1 (data not shown). Thus, the early Xrx1 expression appears to be restricted, spatially and temporally, to the proliferative area of the anterior neural plate more specifically than that of other transcription factors. How is Xrx1 expression localized to this area and how are its expression boundaries defined? We found that both positive and negative signals contribute to this positioning, as Xrx1 is activated by chordin and Hedgehog and is repressed by retinoic acid and X-ngnr-1. As BMP antagonists are known to act as general forebrain inducers (Sasai and De Robertis,1997), chordin may be responsible for the initial activation of Xrx1 in the anterior neuroectoderm. Hedgehog signaling has been shown to delay differentiation and promote proliferation at early neurula stage (Franco et al.,1999); thus, the observation that Hedgehog signaling can activate Xrx1 correlates with the expression of Xrx1 in anterior proliferating cells. Different Hedgehog genes share the same activity, with regional specificity being achieved by differential expression of the family members. Thus, it is expected that X-shh and X-chh were equally effective in Xrx1 activation, although it is likely that X-chh, which displays anterior expression(Ekker et al., 1995), assumes this function in normal development. At later stages, when the proximodistal axis of the diencephalon/eye-tract develops, the response of Xrx1 to the Hedgehog gradient may become more complex. In fact, Xrx1 is expressed in the retina and retinal pigmented epithelium, where the level of Hedgehog is low, is absent in the optic stalk, and is expressed again in the diencephalon floor, a region exposed to the highest level of Hedgehog signal. Xrx1 expression in the anterior neural plate is also controlled by negative regulators such as RA and X-ngnr-1, which appear to prevent Xrx1 expansion to surrounding regions. This is reflected in the spatial relationship between Xrx1 on one side, and XRALDH2,a RA-producing enzyme (Chen et al.,2001), and X-ngnr-1 on the other. Xrx1 is not simply regulated by factors controlling neuronal differentiation, but plays an active role in this process. In fact, Xrx1 is able to repress neuronal differentiation markers and to counteract the effects of X-ngnr-1 and RA on neurogenesis. Both in the case of neurogenin and RA, Xrx1 action appears to be twofold. Xrx1 acts upstream by repressing X-ngnr-1 and XRALDH2 expression, and downstream or in parallel, by blocking the differentiation promoting activity of these factors. Although at early neurula stage, XRALDH2 and Xrx1expression domains are adjacent but do not overlap, at later stages XRALDH2 expression extends into the retina(Chen et al., 2001). This indicates that although at later stages RA is required for retina formation,for the initial specification of the eye field it is important to keep RA in the anterior neural plate at a very low level. The RA degrading activity of Cyp26 provides protection from RA coming from the posterior neural plate(Hollemann et al., 1998). However, the lack of expression of Cyp26 in the most rostral region of the embryo leaves this area exposed to RA generated by the anterior domain of XRALDH2. Xrx1, the expression pattern of which is complementary to that of Cyp26 (data not shown), may be one of the elements that counteracts the RA signal in the anterior neural plate.

Besides counteracting neuronal differentiation, Xrx1 promotes proliferation in the anterior neural plate. In fact, BrdU incorporation in gain- and loss-of-function experiments provides a direct evidence that Xrx1 is both necessary and sufficient to regulate proliferation in the anterior neural plate. In particular, loss of Xrx1 activity reduces anterior neural plate proliferation to levels similar to those observed in the posterior neural plate, indicating that Xrx1 is one of the main factors responsible, directly or indirectly, for the increased proliferation of the anterior neural plate. So far, among the Rx genes that promote an enlargement of the retina, a proliferation inducing activity has been suggested for medaka Rx3(Loosli et al., 2001) but not for zebrafish rx1 and rx2(Chuang and Raymond, 2001). Although species-specific differences may exist, the orthology relationship between vertebrate Rx genes has not yet been completely clarified.

We previously noticed that the phenotypic effects of Xrx1overexpression are restricted to the eye-anterior brain region, despite of the wider distribution of the injected RNA(Andreazzoli et al., 1999). In the present work, we find that Xrx1 is able to induce proliferation and repress neuronal differentiation in an anterior-specific manner. As Xrx1 is a transcription factor, it presumably acts by regulating the expression of target genes. Interestingly, previous experiments have shown that Xrx1 microinjection activates ectopic expression of XBF-1 in the lateroanterior border of the neural plate(Andreazzoli et al., 1999)(Fig. 2G). As the effects of Xrx1 on neurogenesis are also observed in regions where XBF-1 cannot be activated by Xrx1, and the Xrx1expression domain is larger than that of XBF-1, additional factors are likely to be regulated by Xrx1 in the anterior neural plate. Although Xrx1 appears to be a transcriptional activator(Andreazzoli et al., 1999; Chuang and Raymond, 2001), its overexpression repressed X-ngnr-1, X-Delta-1, N-tubulin, XRALDH2 and p27Xic1. Consistently, Xrx1 activates Zic2 and Xhairy2, two transcriptional repressors involved in delaying neuronal differentiation. As these two genes have an anterior expression domain that partially overlaps with that of Xrx1, they may mediate the repressive effects of Xrx1 in this system. Interestingly, mouse Rx1 can activate Hes1, a hairy homologue, during retinogenesis(Furukawa et al., 2000),indicating that genes of the Hairy family might be evolutionary conserved Rx1 targets. Moreover, mutations in human ZIC2 induce holoprosencephaly, and the mouse knockout of Hes1 affects eye morphogenesis, phenotypes that are similar to those produced by the loss of function of the Rx1 gene (Tomita et al., 1996; Mathers et al.,1997; Brown et al.,1998; Andreazzoli et al.,1999). However, as Zic2 is not induced by Xrx1in animal caps, additional factors are likely required for Zic2activation. Both the ectopic activation of Zic2 and Xhairy2,and the repression of X-ngnr-1, X-Delta-1, N-tubulin, XRALDH2 and p27Xic1 by Xrx1 overexpression are restricted to the anterior neural plate, suggesting that only this region is competent to respond to Xrx1. Xrx1 loss-of-function experiments resulted in reduction, but not abolishment, of Xhairy2 and Zic2expression (Fig. 8Q; data not shown), indicating that Xrx1 is not the only factor responsible for their anterior activation. Conversely, X-ngnr-1 anterior expression was expanded medially, probably as a consequence of the reduction of the eye field. This phenotype, which is essentially reproduced in HUA-treated embryos,is consistent with a severely reduced anterior proliferation. Accordingly, the functional inactivation of Xrx1 does not appear to be sufficient to induce widespread ectopic X-ngnr-1 across the anterior neural plate,presumably because of the persistence of Zic2 and Xhairy2expression. Co-injection experiments revealed that XBF-1, but not Xhairy2, is able to rescue the anterior expansion of X-ngnr-1 observed in MoXrx1-injected embryos. These data indicate that Xhairy2 cannot maintain a normal level of proliferation in the anterior neural plate in the absence of Xrx1 function. The ability of XBF-1 to re-establish a X-ngnr-1-free region suggests that Xrx1 might work in part through XBF-1 and/or that both genes control anterior neural plate proliferation acting on common regulators, as is the case for p27Xic1.

An important mechanism used during development to prevent neuronal differentiation is lateral inhibition, a process mediated by transduction of the Notch signal. We considered the possibility that Xrx1 might work by increasing lateral inhibition. This hypothesis was supported by the co-expression of Xrx1 and X-Notch-1 at early neurula and by data indicating that mouse Rx1 activates Notch transcription during retinogenesis (Furukawa et al.,2000). By contrast, we did not find activation of X-Notch-1 in Xrx1-injected embryos during early-mid neurulation (stages 13-18). Furthermore, Xrx1 expression could not be stimulated by expression of a constitutively active form of Notch at early neurula stage. Another way in which lateral inhibition could be triggered is by overexpression of Delta, but this possibility could also be ruled out as Xrx1 represses X-Delta-1 expression,probably as a consequence of X-ngnr-1 inhibition. Finally, we checked if Xrx1 repression of neuronal differentiation could be prevented by blocking lateral inhibition. We observed that co-injection of Xrx1and an antimorphic form of Delta, known to block lateral inhibition, does not affect the ability of Xrx1 of repressing neuronal differentiation in the anterior regions of the embryo.

These data suggest that Xrx1 does not work through lateral inhibition involving Delta and Notch, but may bypass this system through the activation of Xhairy2, a target gene of Notch(Davis et al., 2001). In general, lateral inhibition is probably not responsible for preventing precocious neuronal differentiation in the anterior neural plate. In fact, the inability of noggin-injected animal caps, which display an anterior neuroectodermal character, to undergo neuronal differentiation at early neurula stage is not mediated by lateral inhibition(Papalopulu and Kintner,1996). Similarly, the inhibition of neuronal differentiation after injection of high doses of XBF-1 is not due to increased lateral inhibition (Bourguignon et al.,1998).

In Drosophila, prepattern genes that are expressed before the onset of neurogenesis control the region-specific activation of proneural genes. Prepattern genes include hairy and the Iroquois family homeobox genes (Gomez-Skarmeta et al.,1996; Fisher and Caudy,1998). In vertebrates, homologues of the Iroquois genes play a similar role, functioning during early neurulation in the specification of neural precursors in the posterior neural plate(Bellefroid et al., 1998; Gomez-Skarmeta et al., 1998; de la Calle-Mustienes et al.,2002; Itoh et al.,2002). We notice several similarities between the activities of Xenopus Iroquois (Xiro) genes and Xrx1, as these genes: (1) repress neuronal differentiation at early neurula; (2) do not work through lateral inhibition; (3) are repressed by X-ngnr-1 and activated by hedgehog signaling; (4) upregulate Xhairy2 and Zic2; and (5) act after neural induction and before the selection of neuronal precursor cells.

Moreover, the loss of function of Rx genes in vertebrates as well as of the Iroquois complex in Drosophila, results in the absence of the structures where these genes are normally expressed(Cavodeassi et al., 2001; Mathers et al., 1997; Andreazzoli et al., 1999; Loosli et al., 2001). Beside these similarities, it is worth noting that while the Iroqouis genes play a role in positioning domains of proneural gene expression, this function has not been demonstrated for the Rx genes. However, the complementary expression of Xrx1 and Xiro genes together with their similar activities suggest the existence of two gene systems, one acting in the anterior and the other in the posterior neural plate, the function of which is to control the timing and delimit the location of neuronal differentiation.

In conclusion, Xrx1, by counteracting differentiating signals and promoting proliferation in a region-specific manner, plays a crucial role in executing a program that, after neural induction, leads to the correct differentiation and patterning of the anterior neural plate.

We thank Drs A. Chitnis, S. C. Ekker, W. A. Harris, C. Kintner, G. Lupo, J. L. Maller, N. Papalopulu, T. Pieler, A. Ruiz i Altaba, T. D. Sargent, M. Tsang, D. L. Turner and M. Zuber for plasmids. We are grateful to Dr L. Bally-Cuif and Dr A. Chitnis for valuable discussion and critical comments on this manuscript, and to Dr N. Papalopulu and Dr D. L. Turner for helpful suggestions and discussion. We also gratefully acknowledge Dr I. Appolloni for her help with some in situ hybridization experiments, Dr P. Malatesta for assistance with statistics, M. Fabbri and D. De Matienzo for technical assistance, and S. Di Maria for frog care. This work was supported by grants from MURST, FIRB Neuroscienze (RBNE01 WY7P) and EC Quality of life and Management of Living resources program (QLG3-CT-2001-01460).