Fgf receptor signaling plays a role in lens induction

The lens is a relatively simple structure with which to study the signaling and gene regulation required for normal development. Key experiments have included those of Spemann (Spemann, 1901) that led to the idea of embryonic induction and raised the disputed (Mencl, 1903) possibility that the optic vesicle was required for lens development. In more recent times, molecular genetic approaches have identified a series of signaling molecules and transcription factors that are critical for the optic vesicle-presumptive lens interaction.

Of central interest is Pax6, a homeodomain transcription factor that is necessary for eye (Hogan et al., 1986b; Hill et al., 1991) and lens development (Ashery-Padan et al., 2000; Fujiwara et al., 1994) and, in the context of both invertebrate (Halder et al., 1995) and vertebrate (Altmann et al., 1997; Chow et al., 1999) embryos, is sufficient. The expression of Pax6 in the lens lineage in vertebrates (Grindley et al., 1995) is controlled by at least one highly conserved transcriptional control element (Williams et al., 1998). This 341 bp enhancer has activity in directing gene expression to the lens placode, lens epithelium, immature primary lens fiber cells, corneal epithelium and the epithelium of the lacrimal gland (Kammandel et al., 1999; Makarenkova et al., 2000; Williams et al., 1998). In the mouse, Pax6 expression in the lens lineage is maintained from the placode stage onwards through the action of bone morphogenetic protein (Bmp) 7 (Wawersik et al., 1999), a member of the transforming growth factor β superfamily (Massague, 1998). Bmp7 is normally expressed in the lens placode and its deletion through gene targeting results in a failure of Pax6 expression maintenance (Wawersik et al., 1999) and a variably penetrant phenotype that manifests as anophthalimia in its most severe form (Dudley et al., 1995).

Fibroblast growth factors (Fgfs) have been implicated in late lens development with the demonstration that Fgf-containing retina-conditioned medium and recombinant Fgf 1 and Fgf 2 could stimulate fiber cell differentiation in lens epithelial explants (Chamberlain and McAvoy, 1987; Chamberlain and McAvoy, 1989). In addition, a series of in vivo gain-of-function studies have shown that a number of Fgf ligands can modulate lens development. The placement of Fgf8-containing beads adjacent to the developing chick eye has suggested a role in early lens development (Vogel-Hopker et al., 2000), while overexpression of a variety of ligands in transgenic mice can stimulate premature differentiation of lens fiber cells (Lovicu and Overbeek, 1998; Robinson et al., 1995b). Some members of the Fgf receptor family are also expressed in the lens lineage (de Iongh et al., 1997; de Iongh et al., 1996). Fgfr1 is expressed in the presumptive ectoderm and lens pit. After lens vesicle separation, Fgfr1 and Fgf2IIIc are both expressed in the lens vesicle and presumptive corneal epithelium. As lens development proceeds, Fgfr2IIIc (bek) is expressed throughout the lens epithelium and the transitional zone, but expression declines in maturing lens fiber cells. By contrast, Fgfr2IIIb (KGFR) shows strong expression in the early fibers of the transitional zone with weaker expression in the lens epithelium (de Iongh et al., 1997; de Iongh et al., 1996). Consistent with these observations, it has been shown that when dominant-negative or dimerized soluble forms of fibroblast growth factor receptors (Fgfrs) are expressed in the developing fiber cells, their differentiation is suppressed (Chow et al., 1995; Govindarajan and Overbeek, 2001; Robinson et al., 1995a), indicating that Fgfr activity is necessary. To date, the question of whether Fgfr signaling might be required for earlier phases of lens development, including lens induction, has not been addressed.

In this report, we describe such experiments. We use three distinct strategies to show that at both the morphological and molecular levels, lens induction does not proceed normally if Fgfr activity is perturbed. First, we demonstrate that reduced levels of Pax6 reporter construct expression and Pax6 immunoreactivity result when presumptive lens ectoderm is cultured in the presence of small molecule inhibitors of the Fgfr kinases. Second, we show that when a dominant-negative Fgfr1IIIc is expressed in the presumptive lens ectoderm, early lens development is morphologically abnormal and, in addition, that molecular markers of lens induction, including Pax6 and Sox2, are expressed at reduced levels. Finally, we perform analysis to show that there is a genetic interaction between Bmp7, an established lens inducer (Dudley et al., 1995; Wawersik et al., 1999), and Fgfr signaling at the stage of lens induction. This manifests as more severe lens defects and lower expression levels for Pax6 and Foxe3 (a lens lineage marker) in compound genotype animals. Combined, these analyses indicate that Fgfr signaling has an important role in lens induction and that this pathway converges with the activity of Bmp7 in upregulating the expression of Pax6 (Ashery-Padan et al., 2000; Fujiwara et al., 1994) and Foxe3 (Blixt et al., 2000; Brownell et al., 2000) – transcription factors necessary for lens development.

Explant cultures were prepared from embryonic day (E) 8.0-9.5 wild-type or P6 5.0-lacz reporter (Williams et al., 1998) mice. For E8.0-8.5 embryos, the anterior embryo from head-folds to just posterior to the developing heart was excised and placed within a 15 μl drop of collagen gel according to established procedures (Shannon et al., 1999; Wawersik et al., 1999). For E9.0-9.5 embryos, heads were bisected at the midline and half-heads placed in collagen gel. Explants were cultured in DMEM (Gibco BRL) supplemented with heat inactivated 10% fetal calf serum, glutamine, non-essential amino acids, cholera toxin and an antibiotic-antimycotic (Gibco BRL). For each explantation experiment, a minimum of five optic primordia were cultured for each condition tested and each experiment was repeated a minimum of three times. After culture for 8-48 hours, tissues were fixed for 5 minutes with 4% paraformaldehyde (for labeling with anti-Pax6 antibodies) or with 0.2% glutaradehyde, 1% formaldehyde [for X-gal staining according to established protocols (Song et al., 1996)].

SU9597 is a compound from a new family of inhibitors for the Fgf receptor tyrosine kinases (Mohammadi et al., 1997; Sun et al., 1999) and was used to block Fgf receptor signaling in vitro. The chemical name for SU9597 is 3-{2-[6-(3-methoxy-phenyl)-2-oxo-1,2-dihydro-indol-3-ylidenemethyl]-4-methyl-1H-pyrrol-3-yl}-propionic acid (Makarenkova et al., 2000). For explant cultures, SU9597 was dissolved in dimethylsulfoxide as a 100 mM stock solution, aliquoted and stored at –20°C.

A restriction fragment encoding a C-terminally truncated mutant form of the Fgfr1IIIc (Reid et al., 1990) was subcloned into the P6 EE1.0 plasmid that consists of a PGEM4z backbone, the 341 bp Pax6 ectoderm enhancer [designated EE (Williams et al., 1998)] and a 1 kb fragment containing a minimal Pax6 P0 promoter. This minimal promoter fragment extends from an XhoI site located approximately 1 kb upstream of the P0 transcription start to an SpeI located at the transcription start. The SV40 small t splice and polyadenylation signals were downstream of the truncated Fgfr1 cDNA.

RT-PCR was carried out on separated presumptive lens ectoderm and optic vesicle from E9.0-E9.5 Tfr7/Tfr7 transgenic mice. Separated tissues were obtained by harvesting mouse embryos at E9.0-9.5, removing and bisecting heads and then separating, using fine forceps, the head ectoderm from the epithelium of the neural tube. As the lens placode is very firmly adhered to the presumptive retina of the optic vesicle at E9.5, small volumes of 2 mg/ml dispase (Boehringer Mannheim, #1284908) were pipetted onto the lens placode-optic vesicle junction while the tissue was immersed in room temperature phosphate-buffered saline (PBS). After one or two dispase treatments, the lens placode could be separated from the optic vesicle with ease. The separated tissues were then trimmed to ensure that only presumptive lens ectoderm and optic vesicle was harvested. The Ultaspec reagent (Biotecx, TX) used to isolate total RNA and this reverse transcribed with the Reverse Transcription System (Promega) and oligo-dT primer. The resulting cDNA was then used as template in PCR amplifications with SV40 and GAPDH primers. The amplification conditions were 94°C for 45 seconds, 54°C for 30 seconds and 72°C for 30 seconds for 35 cycles. The SV40 sequence primers are specific for the transgene while the GAPDH primers were used as a positive control. Primer sequences are as follows: SV40-250, 5′TTTGCTCAGAAGAAATGCCA; SV40-450, 3′GCAGTGCAGCTTTTTCCTTT; GAPDH, 5′CTACATGGTCTACATGTTCCAGTA; and GAPDH 3′GTGATGGCATGGACTGTGGTCAT.

Tissues for histological analysis included staged mouse embryos or whole eyes from postnatal animals. Tissue samples were prepared and stained either with Hematoxylin and Eosin or only Hematoxylin using conventional methods (Culling et al., 1985). An assessment of cellular proliferation through S-phase labeling of cells with 5-bromo-2′-deoxyuridine (BrdU) was performed according to standard procedures (Takahashi et al., 1993).

Staged embryos were fixed with 4% paraformaldehyde, cryoprotected in 30% sucrose and 20 μm frozen sections prepared. These were immunofluorescently labeled according to conventional methods (Harlow and Lane, 1988). The anti-Pax6 antibody is an affinity-purified rabbit polyclonal available from Covance (cat#PRB-278P) that was used at a 1:500 dilution in an overnight incubation at 4°C.

All embryos were washed in PBS and fixed in 4% paraformaldehyde at 4°C. Antisense RNA probes were labeled with digoxigenin during in vitro transcription and whole-mount in situ hybridization performed as described previously (Nieto et al., 1996). The probe regions comprised, for murine Foxe3, the entire intronless genomic clone on a ApaI-KpnI fragment of 1064 bp (Brownell et al., 2000) and for murine Sox2, a 1047 bp Xho-AccI fragment that encompasses most of the coding region (Wood and Episkopou, 1999).

The pattern of cell proliferation in normal and transgenic lenses at E13.5 was assessed by mapping BrdU-labeled cells onto a lens coordinate system adapted from previous analyses (McAvoy, 1978; Mikulicich and Young, 1963). Data were expressed as the proportion of total cells that were BrdU positive (the BrdU labeling index) in a given sector. Pregnant female mice were injected with BrdU at the appropriate stage of pregnancy, the mother euthanized 1 hour later and the embryos removed from the uterus by dissection. Heads from E13.5 embryos were fixed overnight in PBS-formalin (4%) and then processed for paraffin embedding. This quantitation was performed using sections that were within three sections either side of the lens center (defined by identifying the largest section) and counts from a minimum of four mice of each genotype pooled. Analysis on E10.5 embryos was performed in a similar way except that fixation was in 4% paraformaldehyde, counting was counting was restricted the lens pit and we used a secondary antibody that was conjugated with the Alexa594 fluorochrome (Molecular Probes) and Hoechst 33258 as a nuclear counterstain. The Student’s t-test was used to assess the significance of the data derived.

Expression of Pax6 in the presumptive lens ectoderm begins as early as embryonic day (E) 7.5 as it is expressed throughout the head ectoderm of the embryo at the head-fold stage (Grindley et al., 1995). However, at E8.75, Pax6 expression was upregulated in the presumptive lens ectoderm that overlies the optic vesicle (Grindley et al., 1995). This second phase of Pax6 expression has been defined by previous investigation as a marker of lens induction (Furuta and Hogan, 1998; Wawersik et al., 1999). As the first in a series of experiments designed to determine whether Fgfr signaling played a role in lens induction, we explanted primordial eye tissues under conditions where Fgfr kinase activity was inhibited and determined whether expression of Pax6 in the presumptive lens region was affected. We used two different methods to assess Pax6 expression. First, we took advantage of a reporter transgenic mouse line (designated P6 5.0-lacZ) that expresses lacZ from a Pax6-derived lens lineage enhancer (Williams et al., 1998). The P6 5.0-lacZ reporter reproduces Pax6 expression in the presumptive lens ectoderm from E8.75 and thus is a marker of lens induction. As a second method, we used antibodies to Pax6 to identify the translation product.

In explant experiments, we used the inhibitor SU9597 to examine the requirement for Fgfr signaling. This is one member of a family of compounds that bind Fgfr kinases at the ATP-binding site and suppress their activity. SU9597 (Sun et al., 1999) is a more cell-permeant variant of the more commonly used SU5402 (Mohammadi et al., 1997). Others have previously shown that these compounds are effective in inhibiting Fgf receptor kinase activity in the 5-20 μM concentration range and that in this range, they have no effect on the epidermal growth factor receptor, insulin-like growth factor receptor and platelet-derived growth factor receptor kinases (Mohammadi et al., 1997). Both SU5402 and SU9597 have also been used previously to demonstrate Fgfr signaling requirements in developmental systems (Makarenkova et al., 2000; McCabe et al., 1999; Schneider et al., 1999).

In one set of experiments, E8.5 embryos were harvested from homozygous P6 5.0-lacz reporter mice, the heads bisected at the midline and explanted into collagen gel either with or without SU9597 at 10 μM. Explants were allowed to incubate for 8 hours until they reached the equivalent of approximately E9.0 and then they were fixed and stained with X-gal in whole mount so that lacZ reporter activity could be detected. Compared with control explants, SU9597-treated 8 hour explants had a consistently observed but only slightly reduced level of X-gal staining (compare Fig. 1A with 1B).

Explants allowed to incubate for a period of 24 hours showed a significant increase in size and this indicated at a general level that the culture conditions permitted development to proceed (compare Fig. 1A,B with 1C,E). Interestingly, after 24 hours of incubation, the Fgfr inhibitor had a dramatic effect on expression of the Pax6 reporter. Control explants showed the anticipated reporter-positive surface ectoderm overlying the optic vesicle (Fig. 1C). SU9597-treated explants showed a greatly reduced level of reporter expression (Fig. 1E). Histological sections prepared from 24 hour explants emphasized the lens placode-specific expression of the reporter in the control (Fig. 1D) and the reduced level of expression with SU9597 treatment (Fig. 1F). This observation implies that Fgf receptors lie upstream of the ectoderm enhancer that normally drives Pax6 expression in the lens placode.

As evidence is emerging that at least two transcriptional enhancers regulate Pax6 expression in the lens placode (Dimanlig et al., 2001; Kleinjan et al., 2001) it was advantageous to assess Pax6 levels in the eye primordium using indirect immunofluorescence. So that we could objectively compare levels of immunoreactivity in control and inhibitor-treated explants, sections were processed and labeled in the same experiment, images acquired digitally under identical lighting conditions, and all figure panels adjusted en masse in the same digital image file. In the Figure panels shown (Fig. 1G-L) both green and yellow represent immunoreactivity for Pax6. The peak signal intensity was replaced with the color yellow and, as such, indicates the highest levels of Pax6.

While Pax6 immunoreactivity was readily detectable in the lens placode of control E8.5 explants cultured for 24 hours (Fig. 1G,H) explants treated with SU9597 showed a reduced level (Fig. 1I,J). Pax6 immunoreactivity was also consistently reduced in the optic vesicle (Fig. 1G-J, ov). In explants harvested at E9.5 and cultured for 24 hours, control explants generated a lens pit with thickened epithelium and strong Pax6 immunoreactivity (Fig. 1K). The presence of SU9597 consistently reduced the width of the lens pit, the thickness of the pit epithelium and suppressed the level of Pax6 immunoreactivity (Fig. 1L). Combined, these data provide evidence that Fgfr kinase activity is required for a normal level of Pax6 expression in the presumptive lens ectoderm, lens placode and lens pit. As Pax6 is known to be critical for lens development (Ashery-Padan et al., 2000; Fujiwara et al., 1994; Wawersik et al., 1999) we can suggest that Fgfr signaling activity is required for normal lens induction.

The described inhibitor experiments have the disadvantage that the inhibitor is active in all cells in the explant and as a result, we cannot determine whether Fgfr activity in the presumptive lens, presumptive retina or both, is required for Pax6 expression. To overcome this limitation, we adopted a second experimental strategy and generated transgenic mice in which a truncated, dominant-negative Fgfr1IIIc was expressed only in cells of the lens placode and lens pit.

To this end, we took advantage of the Pax6-derived lens enhancer (used in the P6 5.0-lacZ reporter mouse described above) that directs transgene expression to the presumptive lens ectoderm. Thus, we generated a transgene construct (Fig. 2A, designated Tfr) that combines the lens enhancer (Williams et al., 1998), the P0 promoter of the Pax6 gene (Xu et al., 1999) and the coding region of a truncated mouse Fgfr1IIIc (Bernard et al., 1991; Reid et al., 1990). SV40 sequences provide splicing and polyadenylation signals. Based on previous experimentation, this construct would be expected to be expressed in the presumptive lens ectoderm beginning at E8.75 (Williams et al., 1998) and to inhibit Fgfr-mediated signaling (Chow et al., 1995; Li et al., 1994; Peters et al., 1994; Robinson et al., 1995a).

The construct was used to generate transgenic mice on the FVB/N background (Taketo et al., 1991) using standard techniques of pronuclear injection (Hogan et al., 1986). Three potential founder animals were generated but only two produced transgenic offspring (Table 1). To confirm the pattern of expression, we performed both whole-mount in situ hybridization and RT-PCR analyses (Fig. 2B-F). An antisense probe to αA-crystallin gave the anticipated strong lens signal in a wild-type E10.5 embryo (Fig. 2B). An antisense probe to SV40 sequences present in the transgene mRNA gave a similar, albeit weaker, pattern of signal in a Tfr7 embryo at E10.5 (Fig. 2C). Importantly, there was no indication of hybridization signal in the retina. A lens-specific hybridization signal was also observed in the Tfr7 line at E11.5 (Fig. 2D) and in the Tfr3 line at E12.0 (Fig. 2E).

Given the limited sensitivity of whole-mount in situ hybridization, we also performed an RT-PCR analysis of transgene expression in the Tfr7 transgenic line. Thus, we harvested eye primordia from hemizygous Tfr7 transgenic mice at E9.5 and separated the presumptive lens ectoderm from the optic vesicle using dispase treatment and dissection. Total RNA was isolated and used to synthesize cDNA. The anticipated 420 bp GAPDH product (Stewart et al., 1997) was amplified from both lens placode and optic vesicle cDNA thus confirming quality of the template (Fig. 2F). PCR primers specific for transgene SV40 sequences amplified a product of the anticipated 200 bp size from lens placode but not optic vesicle (Fig. 2F). Finally, samples prepared without reverse transcriptase did not result in the amplification of any products and this indicated that there was no genomic DNA contamination of the cDNA (Fig. 2F). These data confirm that, as would be expected (Kammandel et al., 1999; Williams et al., 1998), transgene expression in the Tfr7 mice is restricted to the presumptive lens ectoderm. Thus, any phenotypic changes in lens development in the Tfr7 mice could be ascribed to inhibition of Fgfr signaling in lens lineage cells.

As initial analysis of hemizygous Tfr transgenic mice indicated only a mild phenotype, both lines were bred to homozygosity to increase the level of transgene expression. This resulted in a more severe and informative range of phenotypes in both line 3 and line 7 (see summary, Table 1). Because Tfr7 mice showed the most pronounced phenotype, this line has been emphasized. The observation that both lines showed similar changes (including reduced proliferation rates; see below) all but eliminated the possibility of transgene-mediated insertional mutagenesis. Clearly, lens defects can also be rationalized given the expression pattern of the transgene.

Transgenes driven by the Pax6 ectoderm enhancer and P0 promoter are first expressed in the surface ectoderm overlying the optic vesicle at E8.75 but show higher levels at E9.5 as the lens placode forms (Williams et al., 1998). Histological examination of the lens placode in E9.75 Tfr7/Tfr7 transgenic mice revealed a subtle but reproducible decrease in placode thickness and a delay in the initial stages of lens pit invagination (compare Fig. 3A, wild-type, with 3B, Tfr7/Tfr7). Similarly, at E10.5, Tfr7/Tfr7 mice show a lens pit that is reproducibly narrower than in wild-type mice (compare Fig. 3C with 3D). The narrow lens pit observed is a very similar change to that observed in wild-type eye explants exposed to the Fgfr inhibitor (Fig. 1I and J) and is consistent with a role for Fgfr activity in progression of the lens placode to the lens pit.

Subsequent stages of lens development in the Tfr7/Tfr7 mice all showed distinctive defects. At E11.5, when compared with wild-type mice (Fig. 3E) the Tfr7/Tfr7 lens vesicle is small and remains attached to the surface ectoderm (Fig. 3F). At E12.5 in wild-type mice the lens vesicle has closed and the primary fiber cells have differentiated from the posterior wall of the lens vesicle (Fig. 3G, pfc). By contrast, in Tfr7/Tfr7 mice, the lens vesicle has not separated; this is emphasized by a focal invagination of the surface ectoderm (Fig. 3H, red arrowhead) and a failure of mesenchymal cells to migrate between surface ectoderm and lens vesicle (Fig. 3H, black arrows). Normally, the lens vesicle interior is filled by primary lens fibers at this time, but in Tfr7/Tfr7 mice, the vesicle remains small and empty (Fig. 3H). It is also apparent that at E11.5-E13.5, the entire eye in the homozygous transgenics is significantly smaller than in wild-type mice (compare Fig. 3E,F with 3G and H) although at the morphological level, retinal lamination appears unaffected. At the day of birth, wild-type lenses (Fig. 3I) are larger than those in Tfr7/Tfr7 mice and the persistent lens stalk is evident (Fig. 3J). In whole-mount preparations of adult wild-type eyes, the separated cornea and lens are clearly visible (Fig. 3K) while in Tfr7/Tfr7 eyes, the connection between lens and cornea is observed as a column of translucent material. This is very similar to changes observed in Peters’ anomaly in humans and in some heterozygous small eye mice (Hanson et al., 1994).

There was a degree of variability in the eye phenotypes observed both between and within the two transgenic lines. While the lens was clearly smaller in the Tfr3/Tfr3 mice, we did not observe a persistent lens stalk (Table 1). Tfr7/Tfr7 mice also showed phenotypic variability. Besides the typical changes described, in a few transgenic mice, the phenotype was more severe with an extremely small lens and ultimately, a degeneration of the lens and retina (data not shown). Despite this late degeneration response in a few mice, there was no indication of an increased level of programmed cell death in the early stages of lens development. An analysis of the expression pattern of the differentiation markers α-crystallin, β-crystallin (McAvoy, 1978) and MIP26 (Bok et al., 1982) showed that the development of lens fiber cells is delayed in the Tfr7/Tfr7 mice (data not shown). In summary, the morphological analysis of Tfr7/Tfr7 mice indicates that the dominant-negative Fgfr has an effect on all stages of lens development. Important for this analysis however, is the observation that there is an effect on the earliest discernible stage of lens morphogenesis.

In seeking an explanation for the small size of lenses in Tfr7/Tfr7 mice, we quantified the relative cell proliferation rate in controls and transgenics. We performed this analysis on the lens pit at E10.5 and on maturing lenses at E13.5 to determine whether an effect might be observed at multiple stages of lens development. For the assessment at E10.5 we labeled a series of sections (using indirect immunofluorescence; for example, Fig. 4A) from the central lens pit of wild-type and Tfr7/Tfr7 mice and counted total cells and BrdU-positive cells within defined boundaries (Fig. 4A, white lines). The proportion of labeled cells was calculated, a statistical analysis performed and the results presented graphically (Fig. 4B). This indicated that there was a significant reduction of the proliferative index in Tfr7/Tfr7 mice.

For analysis at E13.5, we examined mildly affected Tfr7/Tfr7 eyes where lens-corneal separation was complete. In this instance, BrdU-labeling was performed using alkaline-phosphatase based immunohistochemistry (for example, Fig. 4C). The quantification (Fig. 4D) was performed using a method (McAvoy, 1978; Mikulicich and Young, 1963) in which the number of BrdU-positive cells is determined for a series of angular sectors (Fig. 4C). The data resulting gives a graphical comparison of the proliferation rate at any point along an axis that represents the lens epithelium. Wild-type mice (Fig. 4D, blue line) show the typical pattern of proliferation with peaks centered on 20° and 160° in domains adjacent to immature anterior chamber. Posteriorly, the labeling index decreases to zero at –20° and indicates the point within the lens that fiber cells differentiate. Anteriorly, proliferation becomes progressively lower from 40° to 60° and remains low through the anterior pole (approx. 60° to 120°). Both the Tfr3/Tfr3 and Tfr7/Tfr7 transgenics had reduced levels of proliferation. Line 7 was clearly more severely affected with the BrdU-labeling index lower in all sectors (Fig. 4D, red line). Line 3 showed significantly reduced levels of proliferation in the peak regions only but this was apparent on both the nasal and temporal sides of the lens (Fig. 4D, orange line). These results are consistent with a previous analysis (Robinson et al., 1995a) showing that expression of a truncated Fgf receptor reduced the proliferation rate in the lens lineage. These data are valuable for three reasons. First, they show that the activity of Fgfrs is required for proliferation in the lens lineage from an early stage of development. Second, they identify an additional phenotypic feature that the Tfr3/Tfr3 and Tfr7/Tfr7 transgenics share and emphasize that the changes observed are not a consequence of transgene insertional mutagenesis. Finally, these data provide a link to the phenotype of the Foxe3 mutant mice (dysgenetic lens) where proliferation in the lens lineage is also reduced.

Two models describing signaling events in early lens development suggest that Bmps participate (Dudley et al., 1995; Furuta and Hogan, 1998; Wawersik et al., 1999). In particular, analysis of null mice has indicated that the Bmp7 gene is required for lens induction at the placode stage and that Bmp7 signaling lies upstream of Pax6 expression in a genetic pathway (Wawersik et al., 1999). A role for Bmp7 in lens induction is consistent with the observation that it is expressed in the presumptive lens ectoderm at E9.0 and persists in the lens placode and lens pit until it is downregulated at E11.0 (Wawersik et al., 1999). With this, we could use a third strategy for determining whether Fgfr activity was involved in lens induction by determining whether there was a genetic interaction between Fgfr signaling (as defined by the Tfr7 transgene) and the established lens induction gene Bmp7 [defined by the null allele (Dudley et al., 1995)].

To this end, we crossed Bmp7 heterozygous null and Tfr7/Tfr7 mice through two generations to produce mice with informative genotypes. Mice were initially assessed at the day of birth. A comparison of wild-type (Fig. 5A) and Tfr7/Tfr7 lenses (Fig. 5B) indicated the expected small lens size in the transgenics. Bmp7^–/– animals have a full range of phenotypes ranging from microophthalmia and associated small lenses (Fig. 5C) to anophthalmia (Fig. 5D). By contrast, Bmp7 heterozygous mice have no discernible eye defects (Wawersik et al., 1999) (data not shown). Interestingly, mice that are Tfr7/Tfr7 and heterozygous for the Bmp7-null allele show a more severe phenotype (the two examples represent the typical (Fig. 5E) and mild (Fig. 5F) forms of the phenotype). This was apparent from several features of the Tfr7/Tfr7, Bmp7^+/– eye, including small lens size and the lack of lens vesicle separation and closure (compare Fig. 5B with 5E,F). The lack of lens vesicle closure is indicated by the extrusion of α-crystallin immunoreactive material (data not shown) into the conjunctival sac (Fig. 5E,F, black arrows). In some Tfr7/Tfr7, Bmp7^+/– lenses, there were also extrusions of lens material into the vitreous (Fig. 5F, black arrowheads). Also interesting was the observation that in those Tfr7/Tfr7, homozygous Bmp7 animals with eyes, lens formation was disproportionately affected when compared with other tissues. (Fig. 5G,H). This is illustrated by some lenses that were small, had ruptured capsules and were attached to the cornea (Fig. 5G), and others which were displaced and minute (Fig. 5G,H, arrowheads). By contrast, the retinae of Tfr7/Tfr7, Bmp7^–/– animals were not dramatically reduced in size and showed pseudostratification similar to that observed in wild-type mice. To assess the frequency with which different features arose in wild-type, Tfr7/Tfr7 and Tfr7/Tfr7, Bmp7^+/– mice, we performed histological examination of groups of 10 mice of each genotype. This indicated that regardless of whether we examined lens size, lens vesicle separation, lens vesicle closure or capsule failure, the frequency of occurrence was higher in Tfr7/Tfr7, Bmp7^+/– mice than in Tfr7/Tfr7 or wild type (Table 2). Together with histological observations, this quantitation provides evidence for a genetic interaction between the Bmp7 gene and the Tfr7 transgene. In turn, this implies that Fgfr and Bmp7 signaling cooperate at some stage of lens development.

To extend this analysis and determine whether there might be a Bmp7-Tfr7 genetic interaction during the induction stages of lens development, we examined the expression of lens lineage marker genes. Expression of Pax6 was assessed by immunofluorescence at E9.5 and E10.5. Objective comparisons of Pax6 levels were performed as described for Fig. 1. In this case, the panels included in any one experiment (see brackets, Fig. 6) can be directly compared as the sections shown were labeled at the same time, digital images combined and then adjusted together. As for Fig. 1, we have replaced the peak signal intensity with the color yellow to the same threshold level within an experiment.

In a first experiment, we detected the expected high level of Pax6 immunoreactivity (Fig. 6A, yellow signal) in the lens placode (Fig. 6A, arrowheads) of wild-type mice. By contrast, in Tfr7/Tfr7 mice, the level of Pax6 immunoreactivity was greatly reduced (Fig. 6B, green signal) in the placode (Fig. 6B). Like small molecule inhibitor experiments, this suggested that Pax6 expression was in part dependent on Fgfr activity. In a second experiment (Fig. 6C-E), we compared Pax6 immunoreactivity levels in wild-type (Fig. 6C) Tfr7/Tfr7 (Fig. 6D) and Tfr7/Tfr7, Bmp7^+/– (Fig. 6E) mice. This clearly showed a progressive reduction in Pax6 levels in the lens placode (Fig. 6C-E, arrowheads). In a third experiment, we assessed Pax6 levels at E10.5 in wild-type (Fig. 6F) Tfr7/Tfr7 (Fig. 6G) and Tfr7/Tfr7, Bmp7^+/– (Fig. 6H) mice. As observed at E9.5, the level of Pax6 was progressively diminished as first Fgfr and then Bmp7 activity levels were reduced. Interestingly, it was also apparent that the level of Pax6 immunoreactivity in the presumptive retina was reduced in Tfr7/Tfr7 and Tfr7/Tfr7, Bmp7^+/– at both E9.5 and E10.5. This may reflect the existence of reciprocal induction signals from presumptive lens to presumptive retina that are perturbed when lens induction is abnormal. Regardless, the progressive reduction in Pax6 levels in the lens placode of compound genotype mice further supports the existence of genetic interaction between the Bmp7 gene and the Tfr7 transgene and implies that Fgfr and Bmp7 signaling pathways cooperate as early as E9.5.

In order to understand further how Fgfr and Bmp7 signaling might function in a genetic pathway controlling lens induction, we chose to examine expression of additional lens lineage marker genes in the eyes of Tfr7/Tfr7 and Tfr7/Tfr7, Bmp7^+/– embryos.

The forkhead family transcription factor Foxe3 has been implicated in lens development by its expression pattern and by the phenotype of the mouse mutant dysgenetic lens (dyl) (Blixt et al., 2000; Brownell et al., 2000). In this mouse, there is a failure of lens vesicle separation, inhibition and proliferation, and development of a dysplastic, cataractous lens (Blixt et al., 2000; Brahma and Sanyal, 1984; Brownell et al., 2000; Sanyal and Hawkins, 1979). As this phenotype was very similar to that observed in the Tfr7/Tfr7 mice, we decided to determine whether the expression of Foxe3 was affected. This analysis was performed on E10.5 embryos when the lens pit has just formed. As expected, Foxe3 mRNA was detected in the lens pit of wild-type animals (Fig. 6I). In Tfr7/Tfr7, the Foxe3 mRNA appeared at slightly reduced levels (Fig. 6J) while, in contrast, the Tfr7/Tfr7, Bmp7^+/– mice reproducibly showed a distinct down-regulation (Fig. 6K). This indicated that both Fgfr and Bmp7 signaling are required for a full level of Foxe3 expression that Foxe3 can be placed downstream of both in a genetic pathway.

The transcription factor Sox2 has also been implicated in lens development through regulation of crystallin genes (Kamachi et al., 1995; Kamachi et al., 2001). The Sox2 gene has been used as a marker of lens induction in experiments examining the requirement for Bmp4 and Pax6 (Ashery-Padan et al., 2000; Fujiwara et al., 1994). To determine whether cooperative Bmp7 and Fgfr signaling might function upstream of Sox2 expression, we performed whole-mount in situ hybridization with an antisense Sox2 probe on E10.5 wild-type, Tfr7/Tfr7, and Tfr7/Tfr7, Bmp7^+/– embryos; thick sections permitted the examination of Sox2 expression in the lens pit. Wild-type mice showed the expected pattern of Sox2 signal in the lens pit and the presumptive retina (Fig. 6L, arrowheads and pr, respectively). Tfr7/Tfr7 mice did not show any detectable Sox2 signal in the lens pit (Fig. 6M, arrowheads) and also appeared to have a modified pattern or level of Sox2 expression in the presumptive retina (Fig. 6N, pr). Similarly, Tfr7/Tfr7, Bmp7^+/– mice showed no detectable Sox2 expression in the lens pit and a modified level in the presumptive retina. These analyses indicate that combined Fgfr and Bmp7 signaling function upstream of Pax6, Foxe3 and Sox2 in a genetic pathway for lens development.

We have used three distinct experimental strategies to determine whether Fgf receptor signaling pathways play a role in lens induction. First, we have employed small molecule inhibitors of Fgfr kinases to show that expression of the crucial lens development gene Pax6 is partially dependent on Fgfr activity. Second, we have generated transgenic mice that express a dominant-negative Fgf receptor in the presumptive lens ectoderm and, as a consequence, have abnormalities in lens induction and development. Finally, we have shown that there is a genetic interaction between the Fgfr and Bmp7 signaling pathways during the induction phases of lens development. Because Bmp7 is established as a lens inducer (Wawersik et al., 1999), this indicates that Fgfr signaling is also involved. This analysis shows for the first time that Fgfr signaling is involved in lens induction and defines a genetic pathway that regulates the process.

Previous analysis has defined lens induction at multiple levels. Morphologically, the first indication of lens development (at E9.0-9.5 in the mouse) is the formation of the thickened lens placode in surface ectoderm that overlies the optic vesicle. At E9.5-10.5, the lens placode invaginates in coordination with the lens vesicle and forms the lens pit. At the molecular level, expression of a number of marker genes has defined lens induction. Among these, Pax6 is probably best characterized; it is upregulated in the lens placode beginning at E8.75 and based on both tissue recombination (Fujiwara et al., 1994) and conditional mutation experiments (Ashery-Padan et al., 2000) is necessary for lens development. Misexpression of Pax6 in Xenopus laevis embryos has shown that in this context, Pax6 is also sufficient for lens development (Altmann et al., 1997; Chow et al., 1999). Finally, Bmp7-null mice show a variably penetrant absence of lens development that correlates with the loss of Pax6 expression in the lens placode (Wawersik et al., 1999). When combined, these data provide a strong argument that Pax6 expression in the lens placode defines lens induction. Similarly, it has been argued that increased expression of Sox2 in the early eye is an excellent marker of lens induction; this is based on analysis of both the Bmp4 (Furuta and Hogan, 1998) and Bmp7 (Wawersik et al., 1999) null mice where lens development does not occur and Sox2 expression is not upregulated.

With this, the analysis we have performed makes a strong case for involvement of Fgfr signaling in lens induction. Treatment of eye primordium explants with the Fgfr kinase inhibitor SU9597 downregulates Pax6 expression in the lens placode according to both a Pax6 gene-based reporter construct and immunofluorescent detection of the gene product. This response can be observed at both E9.5 and E10.5 and in the latter case, morphological changes to lens development (formation of a narrow lens pit) are also observed. While the use of small molecule Fgfr inhibitors has proven valuable for characterization of many developmental systems (Makarenkova et al., 2000; McCabe et al., 1999; Schneider et al., 1999) the strategy suffers from the limitation that all cells in an explant are subject to inhibitor effects and as a result, it cannot be determined whether the downregulation of Pax6 expression in the lens placode is an autonomous response of placodal cells or an indirect consequence of the lack of Fgfr activity in the adjacent optic vesicle.

To overcome this limitation, we adopted a second experimental strategy and expressed a truncated, dominant-negative Fgfr1IIIc in the presumptive lens ectoderm. This technique has been employed widely to examine the function of Fgfr signaling in development and has the advantage that, depending upon the spectrum of Fgf ligands represented, a dominant negative mutant receptor may have inhibition activity against all the Fgf receptor isoforms. Thus, with this approach, we can effectively answer a question about the requirement for Fgf receptor signaling even in a system that may involve multiple ligands and multiple receptors and as a consequence, may be refractory to analysis using induced mutations. In these experiments, a truncated mutant of Fgfr1IIIc (Li et al., 1994; Reid et al., 1990) was expressed in the lens lineage beginning at E8.75 using an enhancer from the Pax6 gene (Williams et al., 1998). Two lines of transgenic mice were generated, both of which were used as homozygotes in order to increase transgene expression levels and enhance the phenotypic consequences. The demonstration that both lines of transgenic mice had the same defect in suppression of lens size and lens epithelial cell proliferation indicated that the phenotype observed was not a consequence of insertional mutagenesis. Finally, the confirmation that Tfr7 transgene expression is restricted to lens lineage cells also indicates that cells of the lens placode and lens pit normally respond directly to Fgfr signaling during early lens development.

The morphological and molecular phenotype of the Tfr7/Trf7 transgenic mice provides a strong argument for Fgfr signaling involvement in lens induction. First, we observe distinct morphological defects in lens development from the earliest stages. In particular, there is a delay in formation of the lens placode and its invagination to form the lens pit. At E9.75, The lens placode is thinner than normal, forms a narrow lens pit at E10.5, and throughout embryogenesis the lens of Tfr7/Tfr7 transgenics is smaller than in wild-type mice. Second and more compelling is the observation that the Tfr7/Tfr7 transgenic mice have reduced levels of expression of several genes that are either markers of, or are functionally involved in, lens development. Pax6 expression is distinctly downregulated in the presumptive lens ectoderm of Tfr7/Tfr7 transgenic mice at E9.5 and, importantly, this change is observed before any sign of a morphological defect. A downregulation of Pax6 levels is also observed at E10.5 when the lens pit has formed. The observation that the Tfr7/Tfr7 mice develop Peters’ anomaly is consistent with downregulation of Pax6 in the lens lineage as this phenotype is often associated with Pax6 heterozygosity (Hanson et al., 1994). So too, expression of Sox2 is diminished in the lens pit of Tfr7/Tfr7 transgenic mice. As previous definitions of lens induction have used Pax6 and Sox2 expression in the lens lineage as key markers (Ashery-Padan et al., 2000; Furuta and Hogan, 1998; Wawersik et al., 1999) the current analysis, showing that Sox2 and Pax6 expression are reduced when Fgfr signaling is inhibited, indicates a role for Fgfr signaling in lens induction.

As a third strategy for determining whether Fgfr signaling was involved in lens induction, we determined whether there was a genetic interaction between the Tfr7 transgene and the Bmp7 null allele. This analysis showed that Tfr7/Tfr7, Bmp7^+/– mice have an exacerbated lens phenotype compared with either Bmp7^+/– mice or Tfr7/Tfr7 mice. At the morphological level, Tfr7/Tfr7, Bmp7^+/– mice have a smaller lens and more frequent lens vesicle closure and separation defects than either Tfr7/Tfr7 or wild-type mice. This observation is particularly striking given previous analysis showing that Bmp7 heterozygous mice have no detectable defects (Dudley et al., 1995; Wawersik et al., 1999). Furthermore, we show that the most severe lens defects arise in animals that are homozygous for both Bmp7 and Tfr7. This assessment is complicated by the fact that some Bmp7 homozygous null mice have no eyes, but in those Bmp7^–/–, Tfr7/Tfr7 that do, the consequences for lens development is disproportionate; the retina is near normal in size and morphology while lenses are highly disrupted and very small. From a genetic standpoint, these observations argue that Bmp7 and Fgfr signaling cooperate in an early lens development pathway.

To enhance our understanding of how Fgfr and Bmp7 signaling cooperate, we examined the expression of lens lineage marker genes during induction phases of lens development. This was performed on embryos of wild-type, Tfr7/Tfr7 and Tfr7/Tfr7, Bmp7^+/– genotypes using the markers Pax6, Foxe3 and Sox2 (Furuta and Hogan, 1998; Wawersik et al., 1999). We examined Pax6 immunoreactivity in the lens placode at E9.5 and the lens pit at E10.5 and showed that the lowest levels could be observed in embryos with the Bmp7^–/–, Tfr7/Tfr7 genotype. Similarly, at E10.5, Foxe3 expression levels were reduced progressively in the lens pits of Tfr7/Tfr7 and Tfr7/Tfr7, Bmp7^+/– embryos. Combined, these data indicate that the placodal phases of Pax6 and Foxe3 expression require both Bmp7 and Fgfr signaling. Interestingly, Sox2 expression was undetectable in Tfr7/Tfr7 embryos and remained so in Tfr7/Tfr7, Bmp7^+/– embryos. As it has also been shown that Sox2 expression is lost in the Bmp7 null lens pit (Wawersik et al., 1999) this might imply that either signaling pathway is necessary for Sox2 expression. This interpretation must be applied cautiously, however, as in both this and previous analysis (Wawersik et al., 1999) there may be a limit on Sox2 transcript detection sensitivity given the techniques employed. Regardless of this, this analysis indicates that Sox2 expression in the lens lineage is dependent on both Bmp7 and Fgfr signaling. Because Bmp7 is an established lens inducer, the interaction between the Bmp7 null allele and the Tfr7 transgene indicates that Fgfr signaling is also involved in lens induction. A future challenge will be to identify the ligands for Fgfrs that regulate early lens development. Good candidates are Fgf8 (Lovicu and Overbeek, 1998; Vogel-Hopker et al., 2000) and Fgf15 (McWhirter et al., 1997) as they are both expressed with a timing that might imply an early role.

These experiments define the relationship between several elements of a genetic pathway for lens induction (Fig. 7). Two phases of Pax6 expression in the lens lineage have been defined previously (Grindley et al., 1995). Pax6 is expressed in the pre-placodal head ectoderm (Wawersik et al., 1999) (defined here as Pax6^{pre–placode}) but also later in the lens placode (defined as Pax6^placode). The placodal phase of Pax6 expression is dependent on the activity of Pax6 in the earlier phase (Grindley et al., 1995). The first novel feature of the pathway described is the requirement for Fgfr signaling for Pax6^placode (this relationship could help explain why there is a mild phenotype in the Tfr7/Tfr7 transgenic mice as expression of the dominant-negative Fgfr1 will impose a feedback suppression on the transgene). Furthermore, the model suggests that Pax6^{pre–placode}, Bmp7 and Fgfr signaling act in concert to permit the full level of expression of Pax6^placode (Fig. 7). This model is also consistent with previous analysis of the Bmp7-null mice that indicates Bmp7 is upstream of Pax6^placode (Wawersik et al., 1999).

Similar phenotypes in the dysgenetic lens and Tfr7/Tfr7, Bmp7^+/– mice suggested that Foxe3, the gene mutated in dyl, might lie in the same developmental pathway as Fgfr and Bmp7 signaling. The observation that the level of Foxe3 expression is lowest in Tfr7/Tfr7, Bmp7^+/– embryos and appears to be at an intermediate level in Tfr7/Tfr7 mice argues that Fgfr and Bmp7 signaling also lie upstream of Foxe3 and in combination stimulate a full level of expression (Fig. 7). Consistent with this proposal is the demonstration that Foxe3 expression in the lens placode and pit is absent in the small eye mice (Brownell et al., 2000), thus indicating that its expression is dependent on one of the phases of Pax6 expression. Our accompanying work analyzing a targeted deletion of the Pax6 ectoderm enhancer indicates that Foxe3 expression is lost (Dimanlig et al., 2001). As the ectoderm enhancer is responsible, at least in part, for the placodal phase of Pax6 expression, the combined evidence indicates that Foxe3 lies downstream of Pax6^placode (Fig. 7).

Sox2 is an sry-related HMG family transcription factor that has been implicated in the regulation of lens crystallins (Kamachi et al., 1995; Kamachi et al., 2001) and has been used as a marker to define the process of lens induction (Furuta and Hogan, 1998; Kamachi et al., 1998; Wawersik et al., 1999). It is normally upregulated in both the presumptive lens ectoderm and the optic vesicle at E9.5 when the lens placode thickens (Furuta and Hogan, 1998). Sox2 expression is not upregulated in small eye or Bmp7-null mice and this has suggested that Sox2 lies downstream of both Pax6 and Bmp7 (Furuta and Hogan, 1998; Wawersik et al., 1999). In the current analysis, we show that in both Tfr7/Tfr7, and Tfr7/Tfr7, Bmp7^+/– embryos, Sox2 expression in the lens pit is undectectable. This suggests that Sox2 also lies downstream of Fgfr signaling in a lens induction pathway (Fig. 7). A weakness in the current model is our lack of understanding about the relationship between Sox2 and remaining pathway elements. This uncertainty could be resolved by determining whether Sox2 expression is perturbed in the dyl mouse and, as Bmp4 lies between Pax6 and Sox2 in a lens development pathway (Furuta and Hogan, 1998), whether Foxe3 expression is affected in the Bmp4 null. Inclusion of the sFRP2 marker (Wawersik et al., 1999) in our analysis will also help define distinct stages of lens development.

A future challenge will be to understand which cell-cell interactions are mediated by Fgfr and Bmp7 signaling. It is possible, for example, that placodal Fgf or Bmp7 could initiate an exchange of signals between presumptive lens and retina that is required for placodal Pax6 expression. Some evidence for this comes the observation that when Fgfr or Bmp7 activity is inhibited in the lens placode, Pax6 and Sox2 expression is reduced in the presumptive retina (Fig. 6) (Wawersik et al., 1999). Alternatively, Bmp7 may act in a paracrine or autocrine manner within the placode. The genetic pathway we have described here will form a basis for defining the process of lens induction in molecular detail.

We are indebted to Elizabeth Robertson and Rob Godin for a gift of the Bmp7-null mice. We also thank Ora Bernard for the cDNA to the Fgfr1IIIc, Heather Wood and Vasso Episkopou for the cDNA to murine Sox2, Milan Jamrich for the Foxe3 genomic clone, Sam Zigler for anti-crystallin antibodies, and J. Horvitz for anti-MIP26 antibodies. We are also grateful to members of the NYU Medical Center transgene core facility for generation of the transgenic mice. The Lang laboratory is supported by grants from the National Institutes of Health (EY11234, EY10559 and EY12370).