FGF17b and FGF18 have different midbrain regulatory properties from FGF8b or activated FGF receptors

Early patterning of the vertebrate presumptive midbrain and rhombomere 1(r1), which dorsally gives rise to the cerebellum, is regulated by a local organizer situated at the mid/hindbrain junction (reviewed by Joyner et al., 2000; Liu and Joyner, 2001a; Wurst and Balley-Cuif, 2001). FGF8, a member of the fibroblast growth factor (FGF) family, is expressed in r1 adjacent to the mid/hindbrain junction and has organizer activity. FGF8 can induce the patterned expression of many midbrain/r1 genes and the formation of ectopic midbrain or cerebellar structures depending on the cellular environment and isoform (a or b) of FGF8 protein used. Furthermore,loss-of-function studies in mouse and zebrafish have shown that Fgf8is required for normal development of the midbrain and cerebellum(Meyers et al., 1998; Reifers et al., 1998; Chi et al., 2003). Fgf17 and Fgf18, which encode proteins more closely related to FGF8 than the other FGF family members, are also expressed in the mid/hindbrain junction region in broader domains than Fgf8, including the posterior midbrain (Maruoka et al.,1998; Xu et al.,1999). In biochemical and cell culture assays FGF17b and FGF18 have similar receptor binding properties and ability to induce proliferation when compared with FGF8b (Xu et al.,1999; Xu et al.,2000). In zebrafish, mRNA injection experiments indicate that Fgf8 and Fgf17 have similar effects on gastrulation(Reifers et al., 2000). Loss of Fgf17 in mouse results in truncation of the posterior midbrain(inferior colliculus) and reduced proliferation of the anterior cerebellum(Xu et al., 2000), whereas Fgf18 does not appear to be required for midbrain or cerebellum development (Liu et al., 2002; Ohbayashi et al., 2002). There is clearly overlap in function between at least Fgf8 and 17,as removal of one copy of Fgf8 on an Fgf17 mutant background leads to an exaggerated cerebellum phenotype(Xu et al., 2000). The exact functions of each FGF protein therefore are not clear.

Fgf8 mRNA is differentially spliced to generate multiple protein isoforms. FGF8a and FGF8b are the primary isoforms expressed in r1(Sato et al., 2001) and they differ by only 11 amino acids that are included in FGF8b. Surprisingly, we have shown that these two FGF8 isoforms produce very different phenotypes when mis-expressed in transgenic mouse embryos(Liu et al., 1999). Ectopic expression of the a isoform of Fgf8 in the midbrain and caudal forebrain results in both expansion of the midbrain and ectopic expression of En2, but not other genes expressed in the midbrain and r1 (Lee et al., 1997; Liu et al., 1999). The EN transcription factors alone cannot mediate the midbrain expansion, as similar ectopic expression of En1 does not induce the same phenotype(Lee et al., 1997), and Fgf8a produces midbrain expansion even in En2 mutants (D. Song and A.L.J., unpublished data). In contrast to FGF8a, the b isoform produces exencephaly and a rapid transformation of the midbrain and diencephalon into an anterior r1 fate (Liu et al., 1999) that includes repression of the midbrain gene Otx2, expansion of the hindbrain gene Gbx2 and an anterior shift in organizer genes (Fgf8/Wnt1). A further study showed that GBX2 and EN1/2 are both required for FGF8b to regulate some midbrain/r1 genes(Liu and Joyner, 2001b).

Recently, the functions of FGF8a and b also were elegantly compared in chick following electroporation of different concentrations of DNA expression constructs. Similar to what was observed in mouse, Fgf8a causes expansion of the midbrain and Fgf8b transforms the midbrain into a cerebellum based on early gene expression changes and later morphology(Sato et al., 2001). Interestingly, the initial effect of FGF8b is to reduce growth of the midbrain. Thus, FGF8a and b have distinct activities, both on growth and regulation of gene expression. Of relevance, 100 times lower levels of Fgf8b induce an expanded midbrain. These results, and other studies(Martinez et al., 1999; Liu et al., 1999), have led to the suggestion that a high level of FGF8 signaling induces cerebellum development and a lower level induces midbrain development. If this is the case, then strongly inducing the FGF pathway using activating mutations in FGFRs should mimic the effects of FGF8b. Furthermore, given the dual functions of FGF8 proteins in midbrain and cerebellum development, it is important to determine whether FGF17 and 18 are similar to FGF8a or b.

FGF signaling is mediated by fibroblast growth factor receptor (FGFR)proteins, which belong to a family of tyrosine kinase-containing transmembrane proteins that bind to FGF molecules and mediate FGF signaling (reviewed by Powers et al., 2000). Loss-of-function studies in mouse have demonstrated that various FGFRs are essential in processes such as gastrulation, limb outgrowth and lung morphogenesis (reviewed by Liu and Joyner,2001a). In vitro studies have indicated that in the presence of heparin, all three FGFs present in the mid/hindbrain region bind to the c isoforms of FGFR2 and FGFR3 with high affinity, but not to FGFR1(Blunt et al., 1997; Xu et al., 1999). Interestingly, in mouse and chick embryos Fgfr2 and Fgfr3are not expressed near the mid/hindbrain organizer and Fgfr1 is expressed at low levels (Ishibashi and McMahon, 2002; Walshe and Mason, 2000), raising the question of whether FGFR2/3 mediate FGF signaling from the organizer. Indeed, a recent study of mice lacking Fgfr1 specifically in the midbrain and r1 showed that Fgfr1is the primary FGF receptor required in midbrain and cerebellum development(Trokovic et al., 2003).

The Sprouty (Spry) family of proteins are antagonists of multiple tyrosine kinase-containing receptors including those for epidermal growth factor and FGF. In Drosophila, spry is expressed in cells receiving Fgf signals,and loss of spry phenocopies gain-of-function mutations in fgf (breathless) or fgfr (branchless)(Hacohen et al., 1998). There are multiple Spry members in the vertebrates, two of which (Spry1 and Spry2) are expressed in the mid/hindbrain region of mouse and chick embryos and induced by FGF (4 or 8b)-soaked beads in chick embryos(Chambers et al., 2000; de Maximy et al., 1999; Minowada et al., 1999). Thus,similar to other signaling pathways, FGF induces a negative feedback loop, and a fine balance between activating and suppressing signaling must be required for proper midbrain and cerebellum development.

In this study, we compared the activity of FGF17b and FGF18 to FGF8 in midbrain/cerebellum development using the chick electroporation assay. Strikingly, mis-expression of Fgf17b or Fgf18 at similar levels to Fgf8 induced expansion of the midbrain and regulation of midbrain genes similar to FGF8a. Of significance, among the four FGF proteins tested, only FGF8b induces Gbx2 and represses Otx2 producing a broad Gbx2+/Otx2– domain that abuts the Otx2positive cells in the remainder of the midbrain. Interestingly, FGF8b induces organizer genes at the new Gbx2/Otx2 border, whereas FGF8a induces Fgf8 in scattered cells in the midbrain. In addition, only FGF8b strongly induces the feedback inhibitors Spry1 and Spry2,and we show that Spry1 is a direct target of FGF8 signaling. Consistent with the idea that FGF8b induces a higher level of signaling,mis-expression of activated FGFRs leads to induction of Gbx2 and Spry1/2 and repression of Otx2 similar to FGF8b, although the induction is in scattered cells and does not produce a late phenotype of cerebellum induction.

Mouse brain explant culture, RNA in situ hybridization of wholemount tissue and sections was carried out as previously described(Liu and Joyner, 2001b; Liu et al., 1999). In ovo electroporation in chicken embryos was performed as described previously(Timmer et al., 2001) with some modifications. Specifically, cDNAs for mouse Fgf8a(Lee et al., 1997), Fgf8b (Liu et al.,1999), Fgf17b (Xu et al., 1999) and Fgf18(Maruoka et al., 1998), and human mutant FGFR genes were cloned into a chicken expression vector pMiwIII (Muramatsu et al.,1997) such that they are under the control of a chickenβ -actin promoter. Two mutant forms of human FGFR1, N546K and K656E (M. Mohammadi, unpublished), as well as one mutant form of human FGFR2,C342Y (Mansukhani et al.,2000), were used in this study. The expression constructs were injected into the midbrain ventricles of stage 9-12 chicken embryos(Hamburger and Hamilton, 1992)and two electrodes were placed on either side of the rostral brain. Five rectangular electric pulses of 10 volts, for 50 mseconds were then delivered.

Human FGF17b protein was kindly provided by Shaun K. Olsen and M. Mohammadi. An in situ probe for chicken Gbx2 was generated by RT-PCR from stage 18 chicken brain RNA according to chicken sequences published in GenBank. A probe for chicken Otx2 was made by Dado Boncinelli. The chick Fgf8 probe was from Brigid Hogan, and Spry1 and Spry2 probes were from Gail Martin. The chick Wnt1 probe was from Marion Wassef.

Previous studies in mouse showed that Fgf8, Fgf17 and Fgf18 are all expressed in the isthmus region of eight- to nine-somite mouse embryos with Fgf8 in the broadest domain(Xu et al., 2000). To determine the temporal sequence of mid/hindbrain expression of the three Fgf genes, we examined gene expression in whole mouse embryos. Fgf8expression was first seen at the four-somite stage(Fig. 1A,C), whereas Fgf18 expression was not detected until slightly later at the five-somite stage (Fig. 1D,Eand data not shown). Fgf17 expression was first detected in the midbrain/r1 region at the six-somite stage(Fig. 1I and data not shown). At E9.5, both Fgf17 and Fgf18 were strongly expressed in domains encompassing the posterior midbrain and anterior r1(Fig. 1F,J), with the Fgf17 expression domain being broadest. Fgf8 expression, by contrast, was restricted to a small domain at the anterior border of r1(Fig. 1B).

As Fgf8 expression precedes that of Fgf17 and Fgf18, we investigated whether FGF8 can regulate the expression of Fgf17 and Fgf18 in the mouse brain. Explants were taken from the anterior midbrain at E9.5 and cultured with FGF8b-soaked beads as described previously (Liu et al.,1999). By 40 hours, Fgf18 was induced by FGF8b-soaked beads (n=4/4, Fig. 1H)but not by BSA-soaked beads (n=0/4, Fig. 1G). By contrast, Fgf17 was not induced by either FGF8b-soaked beads (n=0/4, Fig. 1L) or BSA-soaked beads(n=0/4, Fig. 1K). A time course of Fgf18 induction was then performed. Fgf18mRNA was not detected in midbrain explants after 8 hours of exposure to FGF8b(n=4), but was present by 16 hours (n=3) (inset in Fig. 1H and data not shown). We have previously shown that Fgf8 is not induced by FGF8 in the same assay (Liu et al., 1999). Taken together, these studies demonstrate that Fgf8 is the first Fgf expressed in the mid/hindbrain region and suggest that it, in turn, induces Fgf18 (directly or indirectly) in surrounding cells.

Induction of Spry1 and Spry2 by FGF-soaked beads has been shown to occur more rapidly in chick embryonic brains than other midbrain/r1 genes such as En and Wnt1(Chambers et al., 2000; Minowada et al., 1999). We therefore sought to determine whether the Spry genes are direct targets of FGF8b using our mouse brain explant assay, as protein synthesis inhibitors can be added to the medium. First we examined whether expression of the mouse Spry genes are similarly controlled by FGF signaling, using explants from prosomere 1 (p1), where neither Spry is expressed(Fig. 2A,B). Similar to in the chick, we found that Spry1 and Spry2 were rapidly induced within 4 hours by FGF8b-soaked beads (n=4/4 for each gene, Fig. 2D,H and data not shown),but not BSA-soaked beads (n=0/4 for each gene, Fig. 2C), in p1 explants. Next we added 50 μg/ml cyclohexamide or ethanol to the medium, and found that Spry1 is not induced by this treatment(Fig. 2E,G), whereas Spry2 is induced by cyclohexamide alone (data not shown). Regulation of Spry1 by FGF8b in the presence of the protein synthesis inhibitor was then tested, and indeed Spry1 was found to be induced(n=6/6; Fig. 2F). By contrast, the induction of En1, En2 and Gbx2 by FGF8b-soaked beads was efficiently blocked by cyclohexamide (data not shown), consistent with our observation that it takes at least 8 hours for these genes to be induced by FGF8b-soaked beads (Liu and Joyner, 2001b). These results show that induction of Spry1, but not of En1, En2 and Gbx2 by FGF8b is direct.

As Fgfr2 and Fgfr3 are not expressed in the cells surrounding the Fgf8 domain in r1(Walshe and Mason, 2000; Ishibashi and McMahon, 2002)(Fig. 3A-F), this raises the question of whether FGF8b regulates these receptors as well as Fgf18. We therefore examined the effects of FGF8b on the expression of Fgfr genes in p1 brain explants. Fgfr1 was expressed at low levels in these explants and the expression was not altered by FGF8b-soaked beads after 40 hours (Fig. 3G,H; n=4/4). Fgfr2 expression was maintained in p1 brain explants in the presence of BSA-soaked beads (n=4/4), with the highest level being along the dorsal midline (Fig. 3I). Significantly, FGF8b-soaked beads downregulated Fgfr2 expression in the surrounding cells(Fig. 3J; n=4/4). Fgfr3 expression was also maintained in control p1 brain explants(Fig. 3K; n=4/4), and was repressed by FGF8b-soaked beads (Fig. 3L; n=4/4). The repression of Fgfr3 by FGF8b seemed to be more efficient than repression of Fgfr2, consistent with Fgfr3 expression being more restricted to the rostral midbrain at E9.5 than Fgfr2.

Given that Fgf17 and Fgf18 are expressed in the midbrain,as well as in r1, it is important to determine the activity of these FGFs compared with FGF8a and FGF8b. A previous study using beads soaked in FGF18 and placed in the caudal diencephalon showed that it can induce Fgf8and apparently an ectopic midbrain after 3 days(Ohuchi et al., 2000), but did not address whether it can induce a cerebellum or regulate other genes. The function of FGF17 in the midbrain/r1 has not been explored in such a gain-of-function assay. Only one isoform of FGF18 has been described and it contains a 12 amino acid insert in the same position that FGF8b has an 11 amino acid insert compared to FGF8a (Xu et al., 1999). Three isoforms of FGF17 have been described and one(referred to as FGF17b) has an 11 amino acid insert in a similar position to FGF8b, whereas FGF17a lacks this insert. FGF17c has a stop codon that truncates the protein before the conserved FGF domain. To compare the activity of FGF17b and FGF18 with FGF8a/b in midbrain and cerebellum induction we used the electroporation assay described by Sato et al.(Sato et al., 2001) in which FGF8a induces midbrain development and FGF8b represses midbrain development and later induces a cerebellum. Expression constructs containing mouse cDNAs for Fgf17b or Fgf18 (see Materials and methods) were electroporated into the midbrain and caudal forebrain region of chick embryos(Fig. 4A) at stages 9-12(Hamburger and Hamilton, 1992)at a concentration of 1 μg/μl, and examined for changes in midbrain morphology and midbrain/r1 gene expression.

As a control for our experiments, we repeated the experiments of Sato et al. (Sato et al., 2001) by electroporating mouse Fgf8a and Fgf8b expression vectors at different concentrations (Table 1). The transfection efficiency was monitored by visualizing the expression of co-electroporated GFP (Fig. 4B) and by examining the transgene RNA using section in situ hybridization (see Fig. 5). Similar to recent studies (Sato et al.,2001), when Fgf8b was electroporated into the midbrain and caudal forebrain at a concentration of 0.1 μg/μl to 3 μg/μl,midbrain development was repressed and the region was probably transformed into an ectopic cerebellum (n=22/22; Fig. 4D, compare with the control in 4C, and data not shown). By contrast, electroporation of 0.01μg/μl Fgf8b (n=6/6, data not shown), or 1-2μg/μl Fgf8a (n=5/5; Fig. 4E) led to expansion of the midbrain and transformation of the diencephalon into a midbrain. Analysis of expression of the Fgf8a and Fgf8b mRNA produced by the expression vectors showed that similar levels and patterns of expression were obtained when 1-2 μg/μl of the DNA was used (inset in Fig. 5A,B). As expected, no Fgf8 mRNA was detected when 0.01 μg/μl Fgf8b was used(data not shown).

Strikingly, electroporation of 1 μg/μl Fgf17b(n=15/15; Fig. 4F and Table 1) or Fgf18(n=17/17; Fig. 4G and Table 1) led to expansion of the midbrain, similar to the phenotype seen with Fgf8a. This phenotype was not due to reduced levels of expression of the constructs compared to the Fgf8b (or Fgf8a), as similar levels of mRNA were produced by the Fgf17b and Fgf18 vectors (inset in Fig. 5I,M). To analyze the phenotype in more detail, chick embryos were processed for section RNA in situ analysis 24 hours after electroporation and analyzed for midbrain/r1 gene expression. Sato et al. (Sato et al.,2001) found that FGF8b induces Gbx2, Pax2/5, En1/2 and represses Otx2 and Pax6, whereas FGF8a only induces En1/2. Thus, a key distinction between the activity of FGF8a and FGF8b is that only FGF8b induces Gbx2 and represses Otx2(compare Fig. 5A,B with 5E,F). Consistent with the similar phenotype produced by FGF17b, FGF18 and FGF8a,neither FGF17b nor FGF18 induced Gbx2 or repressed Otx2(Fig. 5I,J,M,N). We found that electroporation of a low level of Fgf8b vector (0.01 μg/μl)also did not alter Gbx2 and Otx2 expression (data not shown). The response of the endogenous Fgf8 gene to the four FGFs was very interesting. FGF8b (1 μg/μl) was found to induce Fgf8 in a sharp band of cells at the new Gbx2/Otx2 boundary in the dorsal and lateral midbrain (Fig. 5G; see Fig. 7A,B). By contrast, FGF8a induced Fgf8 in scattered cells in the midbrain(Fig. 5C), whereas FGF17b and FGF18 or low FGF8a did not induce detectable levels of Fgf8(Fig. 5K,O and data not shown). With all expression vectors, En2 was upregulated broadly (data not shown). Finally, we examined the induction of Spry1 and Spry2, as we found that Spry1 is a direct target of FGF8b signaling. Only FGF8b strongly induced the Spry genes in the midbrain, and FGF17b or FGF18 only weakly induced Spry1 in some experiments (Fig. 5H,L,P and data not shown). Taken together, the gene expression studies and morphological analysis demonstrate that when Fgf17 and Fgf18 are expressed at similar mRNA levels, they enhance midbrain development similar to Fgf8a.

Given that FGF17b and FGF18 have similar FGF receptor-binding affinities to FGF8b, at least for FGFR2 and FGFR3 in a tissue culture assay(Xu et al., 2000), it might be expected that the three proteins activate FGF signaling to the same level. One possible reason why electroporation of Fgf17b and 18expression constructs does not produce similar changes in gene expression and proliferation to Fgf8b is that production or secretion of FGF17b and FGF18 protein is less efficient than FGF8b. To determine whether the same concentrations of FGF17b or FGF8b have identical activities, we used our mouse brain explant assay to compare the changes in gene expression induced by beads soaked in FGF17b compared with FGF8b. Similar to the in vivo results, FGF17b was found to induce En1 (n=7/7) and only weakly induce Spry1 (n=6/6) in midbrain explants(Fig. 6I,L; data not shown). Furthermore, Gbx2 was induced in only three out of 12 midbrain explants and FGF17b did not repress Otx2 (n=8/8) in midbrain explants (Fig. 6C,F; data not shown). In the same experiments, FGF8 strongly induce En1(n=5/5), Spry1 (n=3/3) and Gbx2(n=9/9), and repressed Otx2 (n=3/7)(Fig. 6B,E,H,K; data not shown). In all experiments, beads soaked in BSA had no effect on gene expression (Fig. 6A,D,G,H; n≥2). Thus, the difference in activity of FGF17b compared with FGF8b is most likely to be due to intrinsic differences in their ability to induce the FGF signaling pathway.

If the differential effects of FGF8a, 17b and 18 versus FGF8b mis-expression on midbrain development are because FGF8b can activate the FGF pathway more efficiently than the other FGFs, then ectopic expression of activated Fgfr genes should have the same effect as Fgf8b. To test this, we electroporated human FGFR constructs containing activating mutations into the midbrain and caudal forebrain and examined gene expression after 24-36 hours and brain morphology at later stages. We chose three mutant forms of FGFRs to test: one containing a mutation in the extracellular domain(C342→Y) of FGFR2, which leads to receptor activation possibly by inducing spontaneous dimerization of the receptors(Mangasarian et al., 1997);and two containing mutations in the tyrosine kinase domains of FGFR3[N540→K (Bellus et al.,1995)] or [K650→E(Tavormina et al., 1995)]. As the kinase domains are very well conserved among different FGFRs, and Fgfr1 but not Fgfr3 is expressed in the midbrain/r1 region,activated forms of FGFR1 containing the N546K and K656E mutations were used.

We examined gene expression changes in embryos electroporated with 1μg/μl of the FGFR2^C342Y or the FGFR1^N546K vector. Similar to FGF8b, Gbx2, Fgf8and Spry1 and 2 were induced and Otx2 repressed in the midbrain by FGFR2^C342Y (Fig. 5Q,R,T and data not shown). Different from the homogeneous alterations in gene expression produced by electroporation of Fgf8b,the expression of the activated FGFR induced Gbx2, Fgf8 and Spry and repressed Otx2 in patches of cells mainly in the ventricular zone. This is probably due to the cell autonomous function of the activated FGFR compared with the secreted FGF8b protein, as the level and pattern of expression of the FGFR2^C342Y mRNA was similar to the mouse Fgf8 cDNA (inset in Fig. 5Q). The FGFR1^N546K vector produced similar results, when assayed for Fgf8 (n=2/3) and Gbx2 (n=3/3)expression by whole-mount RNA in situ analysis (data not shown).

We next determined the long-term phenotype of transiently expressing activated FGFRs in the midbrain. Unlike FGF8b, the three activated FGFRs(FGFR1^N546K, FGFR1^K656E and FGFR2^C342Y) produced enlargement of the midbrain and diencephalon (Fig. 7A-C; Table 1 and data not shown). At a DNA concentration of as high as 3 μg/μl, ectopic expression of FGFR2^C342Y or FGFR1^N546K caused a similar phenotype to that obtained with 1 μg/μl DNA(Fig. 7B and data not shown). The FGFR1^K656E mutant at 3 μg/μl led to a non-specific loss of the entire brain region including midbrain and cerebellum, preventing a morphological or marker gene analysis(Table 1 and data not shown). Histological analysis of sections through E8-10 chicken embryos (n=2)electroporated with FGFR2^C342Y confirmed that the long-term phenotype of activated FGFRs is an enlarged midbrain, as the electroporated side showed the same histological features of the midbrain as on the control side (data not shown).

One possible reason why transient mis-expression of the FGFRs leads to expansion of the midbrain is that the upregulation of Gbx2 and Spry1/2, and repression of Otx2 does not happen in a sufficient number of cells to transform the midbrain into a cerebellum. Indeed, Sato et al. (Sato et al.,2001) showed that co-electroporation of an Otx2expression vector (1 μg/μl DNA) with an Fgf8b expression vector(0.1 μg/μl DNA) results in expansion of the midbrain, indicating that Otx2 positive cells that receive an FGF8b signal respond by expanding the midbrain, whereas Gbx2 positive cells form a cerebellum. Alternatively, or in addition, FGF8b may eventually lead to transformation of the midbrain into a cerebellum because the organizer is extended along the new Gbx2/Otx2 border and it maintains the transformation. In order to explore these ideas further, we analyzed whole-mount embryos mis-expressing Fgf8b or FGFR2^C342Y for expression of genes normally expressed near the organizer region. In embryos electroporated with the Fgf8b vector, the normal rings of Fgf8 (n=4), Wnt1 (n=4) and En1 (n=3) at the mid/hindbrain junction were repressed on the electroporated side and both genes were induced along the dorsal midline and in a transverse band in the caudal diencephalon, probably adjacent to the induced Gbx2 domain(Fig. 8A-D; n=4 for Fgf8, n=3 for En1). By contrast, expression of the activated FGFR vector induced Fgf8, En1 and Wnt1 only in patches of cells in the midbrain and caudal diencephalon(Fig. 8E,F and data not shown). The resolution of the experiment is not such that we can determine whether these genes are induced at new Gbx2/Otx2 borders. Taken together,these studies are consistent with the idea that a high level of FGF signaling is required to induce Gbx2 and repress Otx2 and that transformation of a broad region into Otx2-/Gbx2+ cells is required for late cerebellum formation.

We show here that FGF8b can positively regulate FGF signaling by inducing Fgf18 expression in brain explants, and that Fgf18expression is initiated slightly later than Fgf8 in the mouse mid/hindbrain junction region. Fgf17 is not expressed until later and is not induced by FGF8 in mouse brain explants. Thus, FGF8 could normally be required to induce expression of Fgf18. Furthermore, although the three FGFs have overlapping spatial distributions after the six somite stage, Fgf17 and Fgf18 should not be able to compensate for a loss of FGF8 protein because they are expressed too late and Fgf18expression is dependent on FGF8 function. A recent study of mice lacking Fgf8 function specifically in the midbrain/r1 region after the five somite stage showed that both Fgf17 and Fgf18 are actually dependent on Fgf8, as their expression is greatly reduced at the seven- to nine-somite stage and gone by the 12- to 15-somite stage in such mutants (Chi et al., 2003). Thus, Fgf8 mutants are equivalent to Fgf8/17/18 triple mutants, and determination of the normal requirement for Fgf17/18 in the midbrain and cerebellum will await analysis of Fgf17/18 double mutants.

Consistent with the expression patterns of Fgfr1, Fgfr2 and Fgfr3, we found that FGF8b is sufficient to repress expression of Fgfr2 and Fgfr3 in caudal forebrain explants. A study of zebrafish ace mutants that have a mutation in fgf8 showed that FGF8 is also required to restrict fgfr3 from the mid/hindbrain junction, because in ace mutants fgfr3 is mis-expressed in the midbrain and r1 (Sleptsova-Friedricha et al., 2001). Thus, Fgf8 negatively regulates FGF signaling by repressing two FGF receptors. Although Fgfr1 is the key receptor that mediates FGF signaling in r1 and the caudal midbrain(Trokovic et al., 2003), the other two FGF receptors might play a role in mediating a low level of FGF signaling in anterior regions of the midbrain.

Expression of two negative regulators of FGF signaling, Spry1 and Spry2, in a broad domain surrounding the mouse mid/hindbrain junction region has indicated that FGF signaling is attenuated by SPRY proteins in this region. Interestingly, we found that only FGF8b strongly induces expression of Spry1 and Spry2 in the chick anterior midbrain or mouse brain explants. Furthermore, using a brain explant assay we demonstrated that Spry1 (and probably Spry2) is a direct downstream target of FGF8b signaling. This indicates that Spry1 expression can be used as a read-out for FGF signaling. Consistent with this, in mouse embryos lacking Fgf8 in the midbrain/r1 after the six-somite stage, Spry2 is maintained at the 7- to 9-somite stage, but greatly reduced by the 13-16 somite stage (Chi et al.,2003).

Taken together, our studies and others show that in mouse FGF8b regulates at least three components of the FGF signaling pathway. First, FGF8b induces expression of another FGF protein, FGF18. FGF8b also directly induces two negative modulators of the pathway (SPRY 1/2), and thus produces a negative-feedback loop. Furthermore, our finding that FGF8b also represses Fgfr2 and Fgfr3 demonstrates that a second negative feedback loop contributes to fine regulation of the level of FGF signaling in r1 and the midbrain to ensure appropriate patterning and growth.

To gain insight into how three Fgf genes orchestrate midbrain and cerebellum development, we explored the activity of FGF17b and FGF18 in comparison to FGF8a and FGF8b in their ability to regulate cell proliferation and gene expression when mis-expressed in the midbrain. Of the four proteins,only FGF8b has the ability to transform the midbrain into a cerebellum. Associated with this unique activity, only FGF8b can induce Gbx2 and repress Otx2 when expressed in the midbrain. Furthermore, and likely of crucial importance for maintaining the transformation, only FGF8b induces an ectopic organizer region at the new Gbx2/Otx2 border in the midbrain. By contrast, FGF8a, FGF17b and FGF18 induce expansion of the midbrain, and do not alter Gbx2 or Otx2 expression. Spry1/2 is strongly induced by FGF8b and only weakly by FGF17b and FGF18, whereas endogenous Fgf8 is only induced locally by FGF8a. This different activity of FGF8b protein can not be due to a higher level of expression of the Fgf8b construct, as it is only at a 100-fold lower DNA concentration at which the mouse Fgf8b mRNA can not even be detected that FGF8b induces a midbrain. As Fgf17and 18 are expressed in the midbrain, although Fgf8 is restricted to r1, Fgf17 and Fgf18 could be the main FGFs that normally directly regulate growth and patterning of the midbrain.

Mouse mutant analyses have shown that Fgf17 is more important in the midbrain than Fgf18, because only Fgf17 mutants have a truncation of the posterior midbrain (Xu et al., 2000; Liu et al.,2002; Ohbayashi et al.,2002). Our comparison of the activities of FGF17 and FGF18 show that Fgf18 could also function with Fgf17 in regulating midbrain development. Loss-of-function studies have also shown that Fgf17 plays a role, along with Fgf8, in regulating late proliferation of the anterior cerebellum(Xu et al., 2000). Our finding that FGF17b and FGF18 have such distinct activities from FGF8b in the midbrain are in contrast to previous tissue culture studies that indicated the proteins have similar binding affinities to FGFR2c and FGFR3c and similar functions in regulating proliferation (Xu et al.,1999; Xu et al.,2000). One possibility was that FGF17b and FGF18 proteins are not produced or secreted as efficiently as FGF8b in the chick midbrain. We have ruled out this possibility by showing that when similar concentrations of FGF17b and FGF8b protein are compared in mouse brain explant assays, they differentially regulate gene expression similar to the electroporation experiments. Thus, the intrinsic activity of FGF8b is different from that of FGF17b, possibly because the 11 amino acid inserts in the two proteins are distinct. Our study demonstrates the importance of testing the activity of proteins in vivo where they normally function. Finally, although Fgf8alone encodes two proteins sufficient for directing development of both the midbrain and cerebellum, Fgf17 and Fgf18 probably augment the proliferative and midbrain inducing ability of Fgf8a or a low level of Fgf8b.

It is possible that the difference in the phenotypes produced by mis-expression of Fgf8a versus Fgf8b is quantitative,because in vitro studies have shown that FGF8b has a much higher affinity for FGFRs than FGF8a. Consistent with this, electroporation of a low concentration of Fgf8b expression vector has similar effects to high concentrations of Fgf8a (Sato et al.,2001), and some Wnt1-Fgf8a transgenics have phenotypes similar to Wnt1-Fgf8b transgenics(Liu et al., 1999). By contrast, FGF17b and FGF18 have similar binding affinities and proliferation activities to FGF8b in vitro (Xu et al.,1999; Xu et al.,2000), but do not behave like FGF8b when mis-expressed in the midbrain or applied to brain explants. However, the biochemical studies were carried out using FGFR2 and FGFR3, which are not the major receptors that mediate midbrain/r1 patterning (Trokovic et al., 2003). It is possible that there are qualitative differences in the way FGF8b interacts with FGFR1, that allow FGF8b to activate the downstream pathway more efficiently. We addressed this possibility by asking whether high level FGF signaling is sufficient to induce Gbx2 and repress Otx2 using activating mutations in FGFR1 and FGFR2. Indeed, the activated FGFRs regulate key target genes similar to FGF8b. Of significance, the activated FGFRs strongly induce Spry1/2and Gbx2 and repress Otx2.

Given the changes in gene expression induced by activated FGFRs, it was perhaps surprising that the long-term phenotype of transient expression of activated FGFRs is expansion of the midbrain. We suggest that in transient mis-expression studies such as electroporations, Gbx2 must be induced in a homogeneous domain so that a new organizer can form along the extended Gbx2/Otx2 border, and the organizer can then maintain the long-term transformation of the midbrain into a cerebellum. In support of this idea,when Gbx2 is electroporated into the midbrain, Otx2 is only transiently repressed in scattered cells in the anterior midbrain, and although the isthmus is expanded anteriorly, no ectopic cerebellum forms(Katahira et al., 2000). As electroporation produces mosaic gene expression, the secreted protein FGF8b,but not the activated FGF receptor, can induce Gbx2 throughout the electroporated region. In addition, although the activated FGFRs can induce Fgf8 as well as Wnt1 and En1 in the midbrain, it is in patches of cells because of the cell-autonomous nature of the receptors. As the response of Otx2-expressing midbrain cells to FGF8b is proliferation of the midbrain (Sato et al., 2001), and there are Otx2-positive cells present on the side of the midbrain electroporated with the activated FGFRs, this could account for the later expansion of the midbrain.

Based on our studies and those of others, we suggest the following steps in midbrain and cerebellum development in mouse(Fig. 9). At the four-somite stage, Fgf8 is induced in the presumptive r1 territory by an unknown factor. Pax2 is required for this induction(Ye et al., 2001) and OTX2 inhibits Fgf8 from being induced in the midbrain(Li and Joyner, 2001; Martinez-Barbera et al.,2001). FGF8b then induces Fgf18 in the surrounding cells,producing a larger domain and gradient of Fgf mRNA that extends into the midbrain. FGF8b also maintains two negative feedback loops by inducing Spry1 and Spry2 expression and inhibiting Fgfr2 and Fgfr3. Fgf17 is then induced by an unknown mechanism that is dependent on Fgf8 (Chi et al.,2003) in a broader domain than Fgf18, further extending the gradient of Fgf mRNA expression. FGF17 and FGF18 protein, and possibly FGF8a and a low level of FGF8b, then regulate proliferation of the midbrain and cerebellum and En expression. The narrow domain where Fgf8 is expressed becomes the isthmus because of the activity of FGF8b (Li et al., 2002), and the adjacent Otx2-negative r1 cells become the cerebellum. We have recently shown that by the 15-somite stage Gbx2 is not required in r1 for cerebellum development, but is required earlier to specify r1(Li et al., 2002). Thus, once Fgf8 expression in r1 is stabilized, perhaps by a secreted factor from the midbrain (Irving and Mason,1999), a key function of high level signaling by FGF8b is to maintain a cascade of gene expression in the midbrain/r1 that maintains an Otx2-negative domain in r1 in which the cerebellum develops.

We are grateful to Shaun Olsen and M. Mohammadi for providing the FGF17b protein. We thank Drs MacArthur, Mohammadi, and Mansukhani and Basilico for providing us with the cDNAs for Fgf8b, mutant FGFR1, and mutant FGFR2, respectively. We also thank Drs Martin, Ornitz, Hogan and Wassef for providing us with in situ probes for mouse Spry1 and Spry2, mouse Fgfr1, Fgfr2, Fgfr3, Fgf17, Fgf18, chicken Fgf8, Wnt1 and chicken Otx2. We are grateful to Mark Zervas and Sema Sgaier for critically reading the manuscript. L.A.N. and A.L.J. are investigators of the Howard Hughes Medical Institute, and A.L. is a Howard Hughes Medical Institute research fellow. Some of this work was supported by a grant from NINDS (R01-NS35876) to A.L.J. and an MSKCC support grant to L.A.N. J.Y.H.L. is supported by an NIH fellowship.