The vertebrate mesencephalic-isthmo-cerebellar region, the so-called midbrain/hindbrain (MH) domain, is characterised by the expression of several genes, such as Otx2, Gbx2, Pax2, En2, and Fgf8. The domains of expression of the transcription factors Otx2 and Gbx2 extend into rostral and caudal portions of the embryo, respectively (see Alvarado-Mallart, 2000). Fate map experiments, using the chick/quail marker, have shown that, at least from HH10 (Hamburger and Hamilton, 1951) onward, the caudal end of Otx2 expression coincides with the caudal limit of the mesencephalic primordium (Millet et al., 1996). At HH20, Otx2 and Gbx2 domains abut, and their common boundary separates mesencephalic from cerebellar domains (Hidalgo-Sánchez et al., 1999a; Garda et al., 2001). Two additional transcription factors, Pax2 and En2 (Joyner, 1996), are expressed on both sides of the Otx2/Gbx2 boundary with a double decreasing gradient extending over the mesencephalic and cerebellar domains. Finally, Fgf8, a member of the fibroblast growth factor family, is expressed just caudal to the Otx2 domain (Crossley and Martin, 1995), overlapping the area that strongly expresses Gbx2 (Hidalgo-Sánchez et al., 1999a). Mutation in all these genes exhibits defects in the development and differentiation of the MH domain (reviewed by Rhinn and Brand, 2001).
The MH boundary represents a secondary organizer (Martínez et al., 1991, 1995; Marín and Puelles, 1994). Crossley et al. (1996) have reported that Fgf8 is the molecular signal capable of inducing MH phenotype. Numerous studies recently have been devoted to understanding how the MH domain develops, focusing on the patterning signal (see Alvarado-Mallart, 2000; Nakamura, 2001; Rhinn and Brand, 2001; Martínez, 2001). However, some aspects of the mutual genetic regulation are still poorly understood. Recently, we have shown that Otx2 positive prosomeres, transplanted to various levels of the MH domain, are induced to express an MH genetic cascade only at the areas where the graft contacts the host Gbx2- and Fgf8-positive cerebellar domain, but never when the transplant contacts Otx2-expressing territories (Hidalgo-Sánchez et al., 1999b). We here further analyse the mutual interactions taking place between the p1/p2 grafted neuroepithelia and the host cerebellar domain, and the temporal sequence of gene expression during the formation of the induced ectopic MH domain.
RESULTS AND DISCUSSION
Graft/Host Interaction Rapidly Affects Host Fgf8 Expression and Induces Pax2 and En2 Expression Within the p1–p2 Graft
Chimeric embryos with p1/p2 grafts within the MH domain (Fig. 1) were fixed at various stages after transplantation, and the temporal sequence of gene expression leading the graft to develop a MH phenotype (Hidalgo-Sánchez et al., 1999b) was analysed in serial sections from the same embryo. The analysis of a significant number of cases showed that Pax2 and En2 are the first genes to be induced within the prosencephalic graft (Fig. 2D,E), forming decreasing gradients from the host/graft interface. These two induced expressions can already be observed at HH13–14 (n = 14) and are maintained at all the developmental stages analysed (see Fig. 3A,E,F, for stage HH14).
Interestingly, at HH13–14, Fgf8 expression is still absent within the grafted territory (Fig. 2A,B). However, graft/host interaction rapidly affects host Fgf8 expression, i.e., on the cerebellar plate of the operated side, the domain of Fgf8 expression extends throughout the host/graft interface (Fig. 2F), much farther caudally than on the nonoperated side where Fgf8 expression is restricted to the characteristic isthmic ring (compare Fig. 2F with G). The Otx2-positive grafted primordium may affect the host Gbx2-positive epithelium by either preventing the retraction of the Fgf8 domain (taking place in normal embryos between HH10 and HH20, see Hidalgo-Sánchez et al., 1999a) and/or inducing Fgf8 expression in areas just underlying the grafted neuroepithelium. This ectopic Fgf8 expression within the host could cause Pax2 and En2 induction within the graft. However, data obtained in a zebrafish Fgf8 mutant, ace, shows that Fgf8 is required to maintain, but not to initiate, Pax2 and En2 expressions (Reifers et al., 1998). Thus, it can be proposed that confrontation of the host Gbx2+ and the graft Otx2+ neuroepithelium could trigger Pax2 and/or En2 induction. It is known that expression of one of these genes is shortly followed by the induction of the other. Indeed, Pax2 is a direct regulator of En2 in the genetic cascade controlling MH development (Song et al., 1996), and mis-expression of Pax2 within the diencephalon induces En2 expression (Okafuji et al., 1999). Also, up-regulation of Pax2 expression is obtained when En2 is mis-expressed within the diencephalic domain (Araki and Nakamura, 1999).
Initially, Otx2-Repressed and Gbx2-Induced Territories Do Not Abut
At early stages (HH13–14), the whole extension of the graft maintains its original Otx2 expression (Fig. 2C) and has not yet been induced to express Gbx2 (not shown). At HH14–15 (n = 20), Otx2 repression occurs within the graft, and Gbx2 is induced in the repressed territory. Interestingly, Gbx2 does not extend as far as Otx2 repression (Fig. 3C,D). The former territory shows a sharp border, whereas the border of the latter territory is diffuse. Thus, now there is no Otx2/Gbx2 interface because a band with the absence or a very low level of Otx2 expression and also negative for Gbx2 appeared within the graft. At HH15–16 (n = 12), the Otx2-negative/Gbx2-negative band within the grafted neuroepithelium persists (Fig. 4C,D). At these stages, Gbx2 expression is stronger than in previous stages and always shows a sharp borderline, whereas the Otx2-negative territory extends much further and still presents a diffuse boundary.
Fgf8 Is the Last Gene to Be Induced Within the Grafted Territory
At HH13–15, Fgf8 transcripts are still absent from the grafted primordium (Figs. 2A,B, 3A,B). At HH15–16 (n = 12), a few grafted cells start to express Fgf8 (Fig. 4A,B). The induced cells are always located in the Otx2-repressed/Gbx2-induced territory (Fig. 4C,D). In stages HH16–17 (n = 6), ever more cells become positive for Fgf8 (Fig. 5A,B). The pattern of gene expression previously described in this type of graft (Hidalgo-Sánchez et al., 1999b) can already be observed at HH17–18 (n = 9), i.e., the area where Otx2 has been down-regulated (Fig. 5D) seems to retract because it becomes restricted to the territory where both Gbx2 (Fig. 5E) and Fgf8 (Fig. 5C) are up-regulated.
Is the Localisation of the Intragraft MH Boundary Dictated by the Gbx2 Induced Domain?
In the normal MH domain, Otx2 and Gbx2 are already expressed at gastrulation (Shamim and Mason, 1998; Niss and Leutz, 1998; Alvarado-Mallart, 2000). At HH7–8, these two domains are not initially complementary, there being a small gap between them that gradually shrinks and disappears (Garda et al., 2001). At HH10, Otx2 and Gbx2 genes show an overlapping expression in a very narrow band that shrinks and ceases to be visible (Garda et al., 2001). At HH20, the Otx2/Gbx2 boundary is established and the Otx2 and Gbx2 domains abut (Hidalgo-Sánchez et al., 1999a). We know that the Otx2 and Gbx2 common boundaries are mutually regulated (Millet et al., 1999; Broccoli et al., 1999). An interesting point in our analysis is that, at early stages, a territory negative for Gbx2 and expressing null or very low levels of Otx2 is found inside the prosencephalic grafts, similar to that observed in HH7–8 normal embryos (Garda et al., 2001). This territory disappears at stage HH17–18, when the induced MH boundary becomes stabilised. In addition, the limit of the intragraft Gbx2-positive domain seems to be fixed as soon as this transcription factor becomes induced, contrary to the limit of the repressed Otx2 territory. Thus, we can hypothesize that the localization of the intragraft MH boundary may be dictated by the Gbx2-induced domain.
Application of beads impregnated with FGF8 secreted proteins in the mesencephalon and the caudal diencephalon represses Otx2 expression around the beads (Martínez et al., 1999). A negative feedback loop has been described between Otx2 and Fgf8 in a dosage-dependent manner (Acampora et al., 1997; Suda et al., 1997). This negative feedback has been hypothesized as being responsible for the short range effect of FGF8-impregnated beads (Martínez et al., 1999). More recently, Garda et al. (2001) have shown that FGF8 is also able to induce Gbx2. It is very possible that, in our grafts, the repression of Otx2 and induction of Gbx2 could be caused by different doses of FGF8 that cause, shortly after transplantation, a different behaviour of Otx2/Gbx2 expression. These findings open the possibility that several mechanisms may be involved in the final stabilisation of the Otx2/Gbx2 MH boundary.
The transplants were performed at the stage of 10–11 somites (HH10), by using quail embryos as the donors and chick embryos as the hosts (Fig. 1), following Alvarado-Mallart and Sotelo (1984). We positioned the p1–p2 alar plate just caudal to the MH boundary (Fig. 1). The resulting chimeric embryos were fixed from 5 to 24 hours after grafting (stages HH13–HH18) by immersion in a 4% paraformaldehyde solution in 0.12 M phosphate buffer (pH 7.6). They were postfixed overnight in the same fixative at 4°C, washed and cryoprotected in 10% sucrose solution in the same buffer, and embedded in a solution of 7% albumin-gelatine, 10% sucrose, in the same buffer. The blocks were frozen and serially sectioned in a cryostat. Sagittal sections, 20 μm thick, were mounted on three sets of parallel slides and treated for in situ hybridization (ISH) with Fgf8, Gbx2, and Otx2 probes (see Hidalgo-Sánchez et al., 1999b). The anti-quail monoclonal antibody QCPN (Developmental Studies Hybridoma Bank) was used to localize the graft in sections hybridized with the Fgf8 probe. In these cases, the diaminobenzidine immunoreaction was intensified with nickel-ammonium-sulfate (dark blue color). In some cases, the section tested for Otx2 and for Gbx2 were immunohistochemically double-stained for En2 expression by using the 4D9 monoclonal antibody (Patel et al., 1989) and for Pax2 expression by using the rabbit anti-Pax2–specific antiserum (Zymed Laboratories, Inc.), respectively. We used either the Sternberger-PAP immunoreaction or a secondary antibody coupled to the fluorescent Cy3. In some cases, we treated one series of sections by ISH with a Pax2 chick probe (see Hidalgo-Sánchez et al., 1999b) instead of testing Gbx2 expression.
We thank C. Sotelo for stimulating discussions and A. Mallart for his kind help with the manuscript. M.H-S. was supported by a fellowship of the “Junta de Extremadura, Consejería de Educación y Juventud-Fondos Europeos,” Spain. Part of this work has been presented as an abstract to the Millennium Meeting of the FENS, Brighton, UK (Eur J Neurosci 2000:12(Suppl 11);325).