Patterning and development of the mid- and hindbrain (MHB) depends on the midbrain–hindbrain organizer (MHO), located at the boundary between these two regions. A variety of transcription factors (e.g., Otx2, Gbx2, Pax2, Pax5, En1, En2) and secreted molecules (e.g., Fgf8, Wnt1, Shh) is expressed at this MHB boundary, and their mutual interactions are responsible for the correct development of the MHB region (for review, see Wurst and Bally-Cuif, 2001; Raible and Brand, 2004).
One of these molecules is the secreted fibroblast growth factor 8 (FGF8), which has organizer activity on its own. Gain-of-function studies have shown that FGF8 is sufficient to induce expression of midbrain/rhombomere 1 (r1) genes at ectopic positions (Liu et al., 1999) and is also sufficient to induce the formation of ectopic midbrain and cerebellar structures (Martinez et al., 1999). Fgf8 conditional knockouts, in which region-specific inactivation of Fgf8 is achieved by the expression of the Cre recombinase in the En1 locus, lack the midbrain and the anterior hindbrain, including ventral structures (Chi et al., 2003). This finding clearly substantiates that FGF8 is not only sufficient but also necessary for normal development of the MHB region. Two other FGFs highly related to FGF8, FGF17 and FGF18, also have been shown to be involved in development of the MHB region. Both FGF17 and FGF18 are involved in the regulation of progenitor cell proliferation (Xu et al., 2000; Liu et al., 2003).
FGFs exert their function by means of high-affinity receptors. These FGF receptors (FGFRs) belong to the tyrosine kinase family of receptors and comprise four proteins (FGFR1–4). FGF receptors are single transmembrane glycoprotein receptors containing three Ig-like loops in the extracellular domain and a split tyrosine kinase intracellular domain (for a review, see Powers et al., 2000; Reuss and von Bohlen und Halbach, 2003). It has been shown by in vitro studies that FGF8 can bind to the four FGF receptors with different affinities (FGFR4 = FGFR3 > FGFR2 > FGFR1), with FGFR1 showing almost no binding (MacArthur et al., 1995; Blunt et al., 1997; Xu et al., 2000). However, the binding of FGFs to their receptors has been shown to be modified by cell type-specific heparan sulfates in vivo (Kan et al., 1999). Thus, the affinity of FGF8 to the four FGF receptors might differ between the in vitro and the in vivo situation.
Several studies report on the expression patterns of Fgf receptors in different species and at different time points of development. There have been studies in the developing Xenopus (Golub et al., 2000), zebrafish (Thisse et al., 1995; Carl and Wittbrodt, 1999; Tonou-Fujimori et al., 2002), chicken (Heuer et al., 1990; Orr-Urtreger et al., 1993; Wilke et al., 1997; Walshe and Mason, 2000), and rat (Wanaka et al., 1991) central nervous system (CNS).
In mouse embryos, expression studies concentrated on certain developmental stages or regions of the neural tube (Orr-Urtreger et al., 1991; Yamaguchi et al., 1992; Peters et al., 1993; Ozawa et al., 1996; Bansal et al., 2003), some of them reporting on the expression of Fgfrs in the MHB region (Ishibashi and McMahon, 2002; Liu et al., 2003; Trokovic et al., 2003).
Taken together, it was shown in zebrafish, chicken, and mouse embryos that Fgfr1 is expressed throughout the MHB region, whereas Fgfr2 is sparing the MHO and surrounding tissues. Fgfr3 was found in the neuroepithelium of the diencephalon and caudal hindbrain, whereas the midbrain and anterior hindbrain appeared to be completely devoid of Fgfr3 expression. The gap in expression around the MHO seemed even larger for Fgfr3 compared with Fgfr2 expression. Fgfr4 transcripts were found in the developing mouse CNS by reverse transcriptase-polymerase chain reaction (RT-PCR; Cool et al., 2002), but no detailed expression analysis is available. Based on these in vivo expression patterns, despite the contradictory results of the in vitro binding assays, it was believed that the receptor transmitting the FGF8 signal in the MHB region is the FGFR1.
To test this hypothesis, a conditional knockout of Fgfr1 was created, leading to an inactivation of Fgfr1 in the En1 domain (i.e., in the caudal mid- and rostral hindbrain; Trokovic et al., 2003). Mutant mice lack the inferior colliculi and the vermis of the cerebellum, structures of the dorsal mid- and hindbrain, respectively. However, no loss of ventral tissues was detected. This phenotype is in sharp contrast to the conditional Fgf8 knockout where Fgf8 is inactivated in the same region. In this Fgf8 mutant, the dorsal as well as most of the ventral midbrain and anterior hindbrain are missing, strongly suggesting that the other FGF receptors must be involved in FGF signal transduction at the ventral MHO. Therefore, this study was performed to revisit the expression patterns of the Fgf receptors Fgfr1–4 in early mouse embryonic stages, with special emphasis on the ventral MHB.
The expression of Fgfr1, Fgfr2, and Fgfr3 was studied in the MHB region and was analyzed from embryonic day (E) 8.5 until E12.5 in comparison to the expression of Fgf8 and the MHB marker En1. Fgf8 is known to be expressed in the anterior r1 from E8.0 on, whereas from E9.5 on, it becomes restricted to a sharp band in the anteriormost part of r1 (Crossley and Martin, 1995). En1 is expressed across the MHB boundary in the caudal midbrain and anterior r1 (Davis and Joyner, 1988). Using both markers on consecutive sections hybridized with the Fgf receptors allowed us to map precisely the extent of expression of the Fgfrs in this region. Fgfr4 is not expressed in the developing mouse MHB region (our own data) and, therefore, was not analyzed further.
Fgfr1 Is Expressed Throughout the Neural Tube
At E8.5, Fgfr1 is weakly expressed all over the CNS, including the floor plate and basal plate of the MHB. The Fgfr1 expression domain is overlapping with the expression of En1 (Fig. 1A–C). At all stages examined, Fgfr1 also overlaps with the expression of Fgf8 (Fig. 1D–F and data not shown). This widespread expression of Fgfr1 in the mouse CNS is maintained throughout later stages.
Fgfr2 Is Expressed in the Ventral Midbrain and in the Floor Plate
Fgfr2 expression at E8.5 covers most of the embryonic neuroectoderm (Fig. 2C,G). Using En1 as a marker for the midbrain and r1, we found that Fgfr2 overlaps with the En1 expression domain in the midbrain but not in the anterior hindbrain. Instead, there is a small gap between the En1 expression in the anterior hindbrain and the Fgfr2 expression in the caudal hindbrain (Fig. 2B–D). In relation to Fgf8 expression, we found that the Fgfr2 anteriorly abuts Fgf8 expression. Posteriorly, it does not reach the Fgf8 expression domain (Fig. 2F–H). Therefore, Fgfr2 at E8.5 is expressed in the midbrain and in the posterior part of the hindbrain, exhibiting a gap in r1.
By E9.5, Fgfr2 expression still shows a gap. In the basal plate of the caudal midbrain and of the anterior r1, this gap now exactly abuts the Fgf8 expression domain. In the alar and roof plate of the midbrain, Fgfr2 expression retracts from the Fgf8 expression domain. Therefore, dorsally, Fgfr2 is not in contact with Fgf8, which itself has a broader expression domain in the dorsal hindbrain (Fig. 2I–N). In contrast to this dorsal gap in expression, Fgfr2 is expressed continuously in the floor plate throughout the mid- and hindbrain. Of interest, Fgfr2 expression in the ventral midline complements the gap in expression of Fgf8 in the floor plate (Figs. 2L–N, 3). At E9.5, this very specific Fgfr2 expression is weak and is seen in cells lining the lumen of the neural tube (Fig. 3D,H,L).
At E11.5, expression of Fgfr2 in the floor plate of the MHO is not restricted to ventricular cells anymore but extends throughout the thickness of the neuroepithelium (Fig. 2S,T). Fgf8 expression at the MHB boundary is restricted to a narrow ring at E11.5, and expression of Fgfr2 in the basal plate follows this contraction of the Fgf8 domain, leaving only a small gap at the MHB boundary, which corresponds to the Fgf8-positive tissue (Fig. 2O–Q). Also, the extent of the dorsal gap changes. By E12.5, this gap has narrowed, so that only the most caudal part of the midbrain and the rostral part of the cerebellar anlage are free of Fgfr2 (data not shown).
Fgfr3 Is Expressed in the Ventral Hindbrain
In contrast to Fgfr2, Fgfr3 expression does not overlap with the En1 expression domain at E8.5 (Fig. 4B–D). The gap formed by the Fgfr3 expression corresponds to the En1 expression domain, indicating that Fgfr3 expression completely spares the midbrain and the anterior hindbrain (Fig. 4D). This finding is supported by the fact that Fgfr3 does abut the posterior border of Fgf8 expression in the hindbrain (Fig. 4H).
At E9.5 and E10.5, ventral Fgfr3 expression closes in and slightly overlaps with the borders of En1 expression in the midbrain and hindbrain. Caudally, it abuts the Fgf8 domain in r1, which is contracting at the MHB boundary (Fig. 4M). From this stage on (E10.5–E12.5), expression of Fgfr3 advances toward the MHB boundary in the ventricular layer of the neuroepithelium, thereby narrowing the gap in ventral expression (Fig. 4K,P). This advancement of Fgfr3 expression correlates with a progressive exclusion of En1 from the ventricular zone of the ventral neural tube in the anterior midbrain and in the hindbrain. Thus, between E8.5 and E12.5, Fgfr3 expression advances from the diencephalon into the midbrain.
By E12.5, only the caudal-most part of the midbrain is devoid of Fgfr3 in the basal plate and floor plate, leaving a gap of only a few cell diameters anterior to the Fgf8 signal (Fig. 4N–R). In the hindbrain, expression of Fgfr3 now overlaps with the Fgf8 expression domain (Fig. 4O,Q,R). Furthermore, only the ventricular zone of the midbrain and anterior hindbrain expresses Fgfr3 at E12.5, which corresponds to weaker expression of En1 in these cells compared with the rest of the neuroepithelium (Fig. 4N–P).
In contrast, in the dorsal MHB region (alar- and roof plate) from E9.5 on, Fgfr3 displays a gap extending from the caudal half of the midbrain until the caudal end of the hindbrain. The anterior to posterior (A/P) extent of this gap in the dorsal midbrain and hindbrain remains unchanged until E12.5. In the dorsal to ventral (D/V) axis, Fgfr3 expression gradually expands further dorsal into the alar plate of the caudal midbrain between E9.5 and E12.5 (Fig. 5 and data not shown).
Taken together, we show that Fgfr1 is weakly expressed throughout the MHB region during all stages examined. In contrast, Fgfr2 and Fgfr3 show dynamic and distinct expression patterns in this region (Fig. 5).
FGF8 is the key signal mediating the activity of the MHO located at the MHB boundary (reviewed in Wurst and Bally-Cuif, 2001). Two other related Fgfs, Fgf17 and Fgf18, are also expressed at the MHB boundary but are probably not involved in patterning of the MHB rather in proliferation (Xu et al., 2000; Liu et al., 2003).
It has been reported previously that, in mouse, the Fgf receptors 2 and 3 are not expressed in tissues adjacent to the MHO (Ishibashi and McMahon, 2002; Liu et al., 2003; Trokovic et al., 2003). In contrast, Fgfr1 has been shown to be expressed along the entire neural tube at E8.5, and this expression pattern of Fgfr1 persists throughout embryonic development.
Therefore, FGFR1 was believed to transmit the FGF8 signal in the embryonic midbrain and hindbrain, leading to the proper development of the structures emanating from this region. However, this hypothesis could not be verified by a conditional knockout for this receptor, in which the Fgfr1 gene becomes inactivated under the control of the En1 promoter. Although, dorsally, there is a loss of the cerebellum and the inferior colliculi, there is no loss of tissue arising from the ventral part of the midbrain and hindbrain. Furthermore, there is no loss of ventral neuronal populations residing in this region (Trokovic et al., 2003). In contrast, conditional mutagenesis of Fgf8 in the En1 expression domain results in a severe loss of midbrain and hindbrain tissue, including loss of markers for ventral midbrain dopaminergic neurons (Chi et al., 2003). Therefore, it has to be assumed that other FGFRs are involved in transmitting the FGF8 signal in the ventral midbrain and hindbrain.
We show here that Fgfr1 is continuously expressed along the neural tube from E8.5 onward, whereas Fgfr2 and Fgfr3 exhibit differences in expression and have distinct dynamics of expression at the MHB boundary. In contrast to earlier studies in mouse and chicken, we can show the presence of Fgf receptors 2 and 3 in the region around the ventral MHB boundary, as for example in the floor plate of the neural tube (Fgfr2), using a very sensitive radioactive in situ hybridization method, which enabled us to detect very low expression levels.
At E8.5, expression of Fgfr2 does not reach the Fgf8 expression domain in the hindbrain, but abuts it in the midbrain. From E9.5 on, there is only a small gap in expression of Fgfr2 in the basal plate of the neural tube, which corresponds to the area of Fgf8 expression at the MHO. Fgf8 is not expressed in the floor plate of the MHB boundary (Crossley and Martin, 1995). One remarkable finding is that Fgfr2 is continuously expressed along the A/P axis of the floor plate from E9.5 onward, even at the level of the MHO, exactly where the Fgf8 expression displays a gap. Dorsal expression of Fgfr2 retracts from the Fgf8 expression at the MHB boundary at E9.5. This dorsal gap gradually closes down until E12.5. Fgfr3 initially, at E8.5, displays a gap in expression in the midbrain and anterior hindbrain. From then on, it gradually progresses from the diencephalic/mesencephalic border toward the Fgf8 expression domain at the MHO but does not abut it in the ventral midbrain until E12.5, the latest stage examined. At E12.5, the posterior Fgfr3 expression even overlaps with the Fgf8 domain (summarized in Fig. 5). Fgfr3 exhibits a broader gap in the dorsal MHB region. The A/P extent of this gap does not change until E12.5, but Fgfr3 expression expands further dorsal, shifting from the basal plate into the alar plate from E9.5 until E12.5. Although a gap in expression of Fgfr2 and Fgfr3 in the MHB region has been described before (Ishibashi and McMahon, 2002; Liu et al., 2003; Trokovic et al., 2003), our results clearly show that the extent of this gap is much smaller than reported, especially in the ventral part of the region.
The differential expression of the three Fgf receptors examined may be the result of a repressor activity triggered by FGF8. This activity may change in a spatiotemporal manner, probably mediated by other downstream factors. Indeed, it has been shown in vitro that FGF8 represses both Fgfr2 and Fgfr3 (Liu et al., 2003). The gap in expression of Fgfr2 at the MHO, exactly where Fgf8 is expressed, and its continuous expression in the floor plate, where Fgf8 is not expressed, substantiate a repression of Fgfr2 by FGF8 in vivo. Fgfr1, which is expressed at the MHO, could mediate the repressive effect of FGF8 on Fgfr2 and Fgfr3.
However, in contrast to Fgfr2, which obviously follows the contraction of the Fgf8 expression domain at the MHB boundary, Fgfr3 in the caudal midbrain never does contact the Fgf8 expression domain at the MHO. Instead, in the hindbrain, expression of Fgfr3 shows an overlap with the Fgf8 expression at later stages. Furthermore, Fgfr3 expression is not continuous in the floor plate at the MHO. This finding hints toward another factor regulating Fgfr3 in vivo during early MHB development. We are well aware that this notion seems to be in contrast to a study in the zebrafish ace mutant, a hypomorphic mutant of Fgf8, in which the Fgfr3 expression domain is expanded into the caudal midbrain upon loss of Fgf8 (Sleptsova-Friedrich et al., 2001). However, this expansion of the Fgfr3 domain might be a secondary effect due to a reduction of engrailed expression in the MHB region of the ace mutant (Reifers et al., 1998). Of interest, we show that Fgfr3 expression at E8.5 complements that of En1. Later, the expression of En1 is gradually excluded from the ventricular zone in the posterior midbrain and in the anterior hindbrain, whereas Fgfr3 expression expands into the ventricular zone in these regions. These mutually exclusive expression patterns of Fgfr3 and En1 indicate a potential interdependence in the regulation of En1 and Fgfr3.
The differences in expression of the three Fgf receptors at the MHB boundary might reflect specific roles for each of them in the development and maintenance of the MHB territory. Fgfr1 is the only Fgf receptor expressed throughout the neural tube at early stages of development. Therefore, FGFR1 could be involved early in general patterning of the neural tube. Expression of Fgfr2 and Fgfr3 in the ventral but not in the dorsal part of the MHB could indicate a role for FGF signaling in D/V patterning. Indeed, FGF8 has been shown to be involved in D/V patterning in medaka (Carl and Wittbrodt, 1999). In zebrafish, on the other hand, Fgfr1 has been shown to be necessary for FGF8 signaling in dorsal as well as ventral aspects. Despite the adjacent expression of Fgfr2 as well as Fgfr3 at the MHB boundary (Tonou-Fujimori et al., 2002), knockdown morphants for Fgfr1 in zebrafish phenocopy the ace mutant, which is a FGF8 loss-of-function mutant (Scholpp et al., 2004). However, species differences between zebrafish and mouse have been discussed before (Scholpp et al., 2004). Therefore, in mouse, the differences between the phenotypes of the conditional Fgfr1 and Fgf8 knockouts (Liu et al., 2003; Trokovic et al., 2003) could be explained by the fact that, dorsally, there is only expression of Fgfr1, which may be necessary for patterning of the dorsal MHB. In addition to Fgfr1, Fgfr2 is expressed in the ventral midbrain from E8.5 on and Fgfr3 in the ventral hindbrain. Both receptors may compensate for the loss of Fgfr1 in the ventral MHB in the conditional Fgfr1 mutant.
Specific combinations of FGFR1 with FGFR2 and FGFR3 in the ventral midbrain and hindbrain could also be necessary for patterning along the A/P axis, triggering different steps in the specification, differentiation, and maintenance of ventral neural cell populations. Indeed, FGF8, together with the secreted factor Sonic Hedgehog, has been shown to be required for the specification of midbrain dopaminergic (DA) and hindbrain serotonergic (5-HT) neurons, which arise in ventral positions anterior and posterior to the MHB boundary, respectively (Ye et al., 1998).
Several mechanisms could mediate the different functions of the FGF receptors in patterning along the A/P and D/V axes in the MHB region. Differences in expression levels and binding activities of the receptors (MacArthur et al., 1995; Blunt et al., 1997; Xu et al., 1999; Walshe and Mason, 2000; Ishibashi and McMahon, 2002), as well as the fact that FGFRs can build heterodimers (Bellot et al., 1991) may lead to activation of different downstream signaling cascades, such as the MAP kinase pathway or the PI3/Akt pathway (reviewed in Pawson and Nash, 2000). Thereby, different sets of target genes may become activated, depending on the combination of the Fgf receptors expressed. In combination, FGFR1 and FGFR2 in the ventral midbrain could be involved in patterning, specification, and differentiation of ventral mesencephalic cell populations, such as the midbrain dopaminergic neurons. In the hindbrain, FGFR1, FGFR2, and FGFR3 together could be responsible for patterning and specification of ventral hindbrain cell populations, including the serotonergic neurons of the ventral hindbrain.
Our results on the expression of FgfR1–3 in the MHB illustrate a possible mechanism to convert one signal, here FGF8, into different outcomes in patterning and specification of neuronal subtypes of the midbrain and hindbrain. To address the precise role of each FGF receptor in the functions of FGF signaling at the MHB boundary, the next step would be to create conditional mouse mutants lacking different combinations of the Fgf receptors.
CD1 mice were purchased from Charles River. Noon of the day of vaginal plug detection was designated E0.5. Embryos were dissected and staged according to Theiler (1989). Embryos were immersion-fixed in 4% paraformaldehyde and embedded in paraffin.
In Situ Hybridization
Embryos were sectioned at 5–8 μm in two planes (coronal and sagittal). In situ hybridization was performed according to a modified version of Dagerlind et al. (1992). A detailed version of the protocol is given at http://www.eumorphia.org/servlet/ECFLP.Frameset. Plasmids used for transcription of antisense probes contained fragments of En1 (Davis and Joyner, 1988) and Fgf8 (Martinez et al., 1999). Plasmids containing fragments of Fgf receptors (specific for both isoforms) were gifts from Roland Lauster (Fgfr1; nt1966-2369 in NM_010206), Clive Dickson (Fgfr2; De Moerlooze et al., 2000), and David Ornitz (Fgfr3; Peters et al., 1993). The probe against Fgfr4 was transcribed from a plasmid cloned by A.A.B. (nt550-1375 in BC033313).
We thank Susanne Laaβ, Katja Mayer, and Miriam Homburg for excellent technical support; Roland Lauster for the Fgfr1 probe, Clive Dickson for Fgfr2 probe, David Ornitz for the Fgfr3 probe. A.A.B. and T.N. were funded by the DFG and the BMBF.