Bmp signaling is required for development of primary lens fiber cells

The mature lens is a polarized structure consisting of a mitotic epithelial layer that covers the anterior surface and terminally differentiated lens fiber cells that occupy the interior volume and the posterior surface. In 1963, it was shown that if the lens of a chick embryo was removed and rotated 180° so that the epithelial layer faced the retina, it would completely repolarize over a few days (Coulombre and Coulombre, 1963). This implied that all of the signals necessary for the establishment of lens polarity are present in the ocular media. Since this experiment, the identity of lens polarization signals has been of interest.

One element of lens polarization is the differentiation of the lens fiber cells that express crystallin proteins and impart the features of transparency and refractility (McAvoy, 1978). Evidence that fiber cell differentiation signals have their origin in the retinal cup comes from experiments showing that rotated mouse lenses are only able to repolarize in the presence of retinal tissue (Yamamoto, 1976). Several signaling pathways have been implicated in the regulation of lens polarity. These include the Fgf signaling pathway, which is important for fiber cell differentiation (Chow et al., 1995; McAvoy and Chamberlain, 1989; Robinson et al., 1995a; Robinson et al., 1995b; Stolen and Griep, 2000), and the Igf1 pathway that can regulate fiber cell differentiation in the chick (Beebe et al., 1980; Beebe et al., 1987) and epithelial cell proliferation in the mouse (Shirke et al., 2001). Involvement of TGFβ ligands in lens development has recently been uncovered by the attenuated fiber cell differentiation phenotype observed when a dominant-negative TGFβ receptor is expressed in the lens (de Iongh et al., 2001).

Several bone morphogenetic protein (Bmp) family ligand genes are expressed during early development of the eye. These include Bmp4 and Bmp7 (Furuta and Hogan, 1998; Wawersik et al., 1999), each of which is believed to play an important early role. Bmp7 has a role in lens induction (Wawersik et al., 1999) where it cooperates with Fgf receptor signaling (Faber et al., 2001). Bmp4 has also been implicated in lens development and differentiation as it is genetically upstream of Sox2 (Wawersik et al., 1999), a transcription factor that regulates expression of crystallin genes (Kamachi et al., 2001). The Bmp4 transcript is present in both the presumptive lens and presumptive retina but is expressed predominantly in the dorsal optic cup as primary fiber cell differentiation begins at E11.5 (Furuta and Hogan, 1998). Within the lens lineage, Bmp7 is expressed in the presumptive lens ectoderm, lens pit and lens vesicle but is downregulated thereafter (Wawersik et al., 1999). Based on characterized expression patterns (of Gdf6, for example) (Chang and Hemmati-Brivanlou, 1999), it might be expected that many other Bmp family ligands will have a role in different aspects of eye development. Bmp signaling occurs by means of ligand binding a primary (type 2) receptor, a complex formation with a transducer, a (type 1) receptor and consequent phosphorylation of Smad proteins (Massagué, 1998; Piek et al., 1999). Three known type 2 receptors [Bmpr2, Actr2a (Acvr2 – Mouse Genome Informatics) and Actr2b (Acvr2b – Mouse Genome Informatics)] and four known type 1 receptors [Actr1 (Acvr1 – Mouse Genome Informatics; also known as Alk2), Bmpr1a (also known as Alk3), Bmpr1b (also known as Alk6), Actr1b (Acvr1b – Mouse Genome Informatics)] are expressed in the developing eye (Dewulf et al., 1995; Feijen et al., 1994; Furuta and Hogan, 1998; Obata et al., 1999; Verschueren, 1995; Yoshikawa et al., 2000).

In this study, we have investigated the involvement of Bmp signaling in development of the lens. We show that the Bmp binding and inhibition protein noggin can suppress primary fiber cell elongation and lens size in explant culture. We also expressed a dominant negative Bmp type 1 receptor (Bmpr1b; also known as Alk6) in the developing lens in transgenic mice. Both the Pax6 ectoderm enhancer (Kammandel et al., 1999; Williams et al., 1998; Xu et al., 1999) and the αA crystallin promoter (Chepelinsky et al., 1985) were used to drive transgene expression. These mice show defects in the differentiation of primary lens fiber cells, suggesting that Bmp ligands are important for this aspect of lens development. Importantly, we show using anti-Bmpr2 (Gilboa et al., 2000) and anti-phospho-Smad (Itoh et al., 2001) antibodies that at embryonic day (E) 12.5, when primary fiber cell differentiation is beginning, equatorial lens cells have Bmp signaling machinery and are responding to Bmp signals. Finally, we show that the primary lens fiber cell differentiation defect is radially asymmetrical, perhaps implying that there are distinct differentiation stimuli active in different quadrants of the eye.

A restriction fragment encoding a dominant negative mutant form (a point mutation resulting in K231R) (Zou and Niswander, 1996) of the rat Alk6 cDNA (Bmpr1b) was sub-cloned into the polylinker of the plasmid pαASPLONO (Chow et al., 1995) to generate the transgene αA-Alk6^DN (Fig. 2A). A second transgene, designated EE-1.0-K-Alk6^DN (Fig. 2A) was generated in pSPLONO (Chow et al., 1995) by subcloning a 340 bp, PCR amplified fragment encoding the Pax6 ectoderm enhancer (Kammandel et al., 1999; Williams et al., 1998) adjacent to a 1 kb, SpeI-XhoI fragment containing the P0 promoter of Pax6. Subsequently, the cDNA encoding dominant-negative Alk6 and containing a Kozak consensus (Kozak, 1986) translational start codon was subcloned downstream of the P0 promoter. The oligonucleotide used to add the efficient start codon was of the sequence ACTAGT-GGATCC-TACGTA-CCACCATGG. Transgenic FVB/N (Taketo et al., 1991) mice were generated and screened according to established methods (Hogan et al., 1986).

Whole-mount in situ hybridization was performed as described (Nieto et al., 1996). 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). Section in situ hybridization using radioactively labeled SV40 sequence probes was performed as described previously (Robinson et al., 1995b).

Tissues for histological analysis included staged mouse embryos of whole eyes from postnatal animals. Tissue samples were prepared and stained either with Hematoxylin and Eosin (H/E) using conventional methods (Culling et al., 1985). Figures in this paper were prepared digitally using a Sony DKC1000 digital camera and Adobe Photoshop software.

Staged embryos fixed were fixed in 4% PFA in PBS and immunofluorescently labeled according to conventional methods (Harlow and Lane, 1988). Polyclonal rabbit antisera for MIP26 (Horwitz and Bok, 1987) was used at a dilution of 1:200, Polyclonal rabbit antisera for γ-crystallin (Zigler and Sidbury, 1976) were both used at a dilution of 1:500. The anti-Bmpr2 (Gilboa et al., 2000) and anti-phospho-Smad (Itoh et al., 2001) antibody are affinity-purified rabbit polyclonals from ten Dijke and P. Knaus that were used at a 1:500 and 1:1000, respectively.

Reconstructions were created by tracing digital images of 15-20 H/E stained histological sections of wild-type and transgenic lenses in the program Canvas. Canvas tracings were then converted to simple line drawings and imported into the 3-D program FormZ. Individual section tracings were stacked along a z-axis and carefully aligned in both the x and y dimensions. The ‘skin’ command was then used to mold a skin over the aligned sections. The resulting lenses were then identified as whole objects that could then be rotated in three dimensions so that all sides of the lens could be observed.

Explants of wild-type E10.5 lenses were made following the protocol described elsewhere (Wawersik et al., 1999). Briefly, E10.5 embryos were dissected under a dissecting microscope in PBS at 4°C and the eyes placed in DMEM at 4°C. When eyes from an entire litter had been collected, the presumptive retinal pigmented epithelium was carefully removed with needles. Eyes were then placed in 33% rat-tail collagen (two-thirds DMEM pH 4.0, one third rat-tail collagen, made basic with 0.8M NaCO₃ – 16 μl/200 μl 33% collagen), and left to gel for about 10 minutes. After hardening, the explant collagen gel was immersed in DMEM + 10% FCS with or without human recombinant Noggin (Regeneron Pharmaceuticals) at 300 ng/ml. Cultures were left to grow for 48 hours then fixed with quick-fix (Culling et al., 1995) and processed for paraffin wax sectioning.

As an initial test of the involvement of Bmp signaling in primary lens fiber cell differentiation, we performed a series of explant experiments where Bmp activity was inhibited by the antagonist noggin (Zimmerman et al., 1996). This general strategy has previously been used to assess lens responses (Wawersik et al., 1999). We performed this analysis with E10.5 tissues that have already been through the induction phases of lens development. E10.5 mouse eye primordia were dissected, the retinal pigment epithelium removed (to allow good survival of retinal cells), and the tissue placed in rat-tail collagen gel submerged in defined media. After incubating at 37°C for 48 hours, spherical lenses with elongated and histologically identifiable lens-fiber cells develop.

The addition of the Bmp antagonist noggin to whole eye explants at E10.5 resulted in the development of smaller lenses after 48 hours of culture. We measured the total area (in arbitrary units) of the largest lens section of five noggin-treated and six control explants and showed that there was a statistically significant difference (Fig. 1A). The smaller size of the lenses could also be seen in histological sections of control (Fig. 1B) and noggin-treated (Fig. 1C,D) explants. As primary fiber cells make up most of the area of a lens section at the stage the explant assay was performed, these data indicate that directly or indirectly, Bmps are required for primary fiber cell development. Difficulty in orienting explants of this size precluded a meaningful analysis of primary fiber cell development asymmetries.

To address the question of whether Bmp receptor activity was important for lens development using a second experimental strategy, we generated transgenic mice that expressed a dominant-negative form of the type 1 receptor Alk6. Alk6 is normally expressed only at low levels in the developing eye (Furuta and Hogan, 1998) but would be expected to inhibit the activity of some Bmp ligands (Massagué, 1998). Transgenic mice were generated on the FVB/N background (Taketo et al., 1991) using the constructs designated EE-1.0-K-Alk6^DN and αA-Alk6^DN (Fig. 2A). The EE-1.0-K-Alk6^DN construct takes advantage of the ectoderm enhancer from Pax6 that has activity in the presumptive lens ectoderm, primary lens fiber cells as well as the early and mature lens epithelium (Williams et al., 1998) (Fig. 2B). By contrast, the αA-Alk6^DN construct carries a portion of the αA-crystallin regulatory region that has activity only in primary and secondary lens fiber cells from about E11.5 (Chepelinsky et al., 1985) (Fig. 2B). Thus, these constructs provide distinct patterns of transgene expression and were designed to indicate whether different lens regions were responding to Bmp signals.

To test the consequences for lens development of Alk6^DN expression, we generated transgenic mice with both the EE-1.0-K-Alk6^DN and αA-Alk6^DN constructs. As phenotypes were mild in heterozygous mice, we bred a number of lines to homozygosity. The most severe defects were observed in αA-A6^DN line 11 and of EE-A6^DN line 48 homozygous animals. Homozygous mice from these two lines were used in all subsequent analysis. For brevity, homozygous mice are referred to by the transgene designation. A summary of the phenotypes in transgenic lines is given in Table 1.

In order to examine transgene expression, in situ hybridization analysis was performed (Fig. 2C-K). Whole-mount P6 5.0-lacZ embryos (Williams et al., 1998) provided a positive control for lens hybridization signal at E13.5 using a probe to the SV40 sequences included in the Alk6 transgenes (Fig. 2C). Negative controls included wild-type whole-mount embryos hybridized with follistatin (Fig. 2D) or noggin (Fig. 2E) probes and these provided contrasting patterns of hybridization where the lens was negative. An SV40 hybridization signal was detectable in the lens of whole-mount αA-Alk6^DN transgenics from E11.5 and later [Fig. 2F (E13.5) and Fig. 2G (E11.5)] and in EE-1.0-K-Alk6^DN transgenics from E9.5 (for example, E11.5, Fig. 2H). Section in situ hybridization of E12.5 control and transgenic tissues were also performed using radioactively labeled SV40 sequence antisense probes (Fig. 2I-K). Matched brightfield and darkfield micrographs indicate that, as would be anticipated, EE-1.0-K-Alk6^DN transgenics show hybridization signal throughout the lens but primarily in the developing lens epithelium (Fig. 2I). By contrast, αA-Alk6^DN transgenics show transgene expression in the differentiating primary lens fiber cells (Fig. 2J). Importantly, there is no indication of differences in hybridization signal intensity on the nasal or temporal side of the lens.

To permit a detailed assessment of the phenotype in αA-Alk6^DN and EE-K-Alk6^DN transgenic mice, we performed a histological analysis from E9.5-E18.5. No phenotypic change was observed during the induction phases of lens development, but clear defects were observed in primary fiber cell elongation. Lenses of E13.5 mouse embryos from lines of both EE-1.0-K-Alk6^DN and αA-Alk6^DN transgenics show primary fiber cell elongation defects. Interestingly, however, the defect was not distributed evenly across the lens width. This was most obvious in Hemotoxylin stained sections (Fig. 3A-C) when examining the boundary between equatorial cells and the primary fiber cell mass. In wild-type lenses, this boundary exists only anterior to the equator (Fig. 3A) but in EE-1.0-K-Alk6^DN line 48 and αA-Alk6^DN line 11 transgenics, the boundary extends close to the posterior lens pole but only on the nasal side (Fig. 3B,C). The nature of the defect is emphasized by the distribution of the fiber cell differentiation markers γ-crystallin and MIP26 (Fig. 3D-I). In the E13.0 wild-type lens, both γ-crystallin and MIP26 are restricted to the fiber cells and are found at high levels of immunoreactivity along the posterior aspect of the lens from equator to pole (Fig. 3D,G). By contrast, E13.0 lenses from both αA-Alk6^DN and EE-1.0-K-Alk6^DN transgenics show nasal side domains that have dramatically reduced levels of immunoreactivity for both markers (Fig. 3E-I). This indicates that expression of the dominant-negative Alk6 from either transgene construct inhibits primary fiber cell differentiation.

In the day-of-birth lenses, the lack of extension of the primary lens fiber cells leads to a break along the sutures (small arrowheads Fig. 3J-L). Secondary fiber cells have compensated for the short central cells by elongating around the edges (seen most clearly in Fig. 3K). Transgenic mice also have smaller lenses than wild type. This might be expected because the primary fiber cells make up a good proportion of the lens bulk at this stage. Adult mice of both lines (red arrowheads, side view, Fig. 3M-O) show cataracts. The appearance of this type of cataract seems to result from the growth of the secondary fiber cells elongating around the short primary fiber cells of the lens nucleus.

In histological sections taken for analysis, transgenic lenses often appeared flatter (a shorter pole-pole distance) than their wild-type littermates. To assess this, we measured the dimensions of E13.5 lenses from wild-type, EE-1.0-K-Alk6^DN line 48 homozygous and αA-Alk6^DN line 11 homozygous embryos. Transgenic lenses of both lines were not significantly smaller equator-to-equator, but the graph of lens shape (Fig. 4) shows that there are significant differences in pole-pole distance. This might be expected in lenses where the primary fiber cell development is perturbed as the elongation of these cells expands the lens along the pole-pole axis.

In order to understand the developmental basis of the asymmetrical primary fiber cell elongation defect in αA-Alk6^DN and EE-1.0-K-Alk6^DN transgenics, we decided to assess carefully wild-type lens development. Histological sections of wild-type lenses from E10.5 to E13.5 indicate that asymmetries are apparent. At E10.5 and E11.5, the lens appears symmetric (data not shown) but by E12.5 it is clear that there is an asymmetric pattern of lens fiber cell elongation. Coronal sections taken from ventral to dorsal of wild-type E12.5 embryos (Fig. 5A-C) shows that in the ventral lens, fiber cells elongate first on the temporal side. Sections from the dorsal half of the lens do not show this asymmetrical distribution of primary fiber cell differentiation. In a second example (Fig. 5D-F), the differentiation marker MIP26 emphasizes that primary fiber cell differentiation begins on the temporal side of the lens vesicle.

Because the αA-Alk6^DN, EE-1.0-K-Alk6^DN defect at E13.5 was subtle, we decided to model the defect in three-dimensions (3D). The goal of the modeling was to show the boundary between epithelial cells and fiber cells as a 3D surface. In order to do this, serial lens sections from E13.5 wild-type and homozygote αA-Alk6^DN-11 and EE-1.0-K-Alk6^DN-48 animals were collected, and sections spaced about 20 μm apart were traced. These tracings were stacked in a 3D drawing and rendering program, and one surface was extended over the spherical lens model. A second surface was extended over the arcs making up the epithelial-fiber cell boundary (Fig. 5G,H). The purple three-dimensional surface represents the boundary between the basal surface of the epithelial cells and the tips of the elongating fiber-cells. The glassy sphere represents the lens capsule. Homozygous EE-1.0-K-Alk6^DN (data not shown) and aA-Alk6^DN (Fig. 5G,H) lenses appeared similar. In the context of this analysis, the phenotype of the αA-Alk6^DN and of EE-K-Alk6^DN lenses at E13.5 is more easily understood. The implication is that the late differentiating ventronasal domain of lens vesicle epithelium has not elongated in αA-Alk6^DN and EE-1.0-K-Alk6^DN embryos, and results in the observed nasal side defect at E13.5. The further implication is that Bmp receptor activity is required for primary fiber cell differentiation in this discrete domain.

The Smads are a family of signal transduction molecules active in the Bmp pathways and their phosphorylation indicates that a given cell is responding to Bmp signals (Miyazono et al., 2000). Taking advantage of recently generated antibodies specific for the type 1 receptor phosphorylated forms of Smad1, Smad5 and Smad8 (Korchynskyi et al., 1999), it was possible to directly examine whether Bmp signaling pathways were functional in the developing lens. This strategy allows the spatial pattern of Bmp responses to be examined. We also chose to use antibodies to the cytoplasmic domain of Bmpr2 (Gilboa et al., 2000) to assess its distribution. Though Bmpr2 has been reported to be expressed in the adult rat eye (Obata et al., 1999) the expression pattern in the developing mouse eye has not so far been described.

To further validate the phospho-Smad and Bmpr2 antibodies in the mouse, we performed immunofluorescence labeling of embryonic regions where Bmps were known to be expressed. Specifically, we examined the surface ectoderm in the region of the E9.5 olfactory placode and first branchial arch as in these locations, there is strong expression of Bmp4 (Fig. 6A) (Dudley and Robertson, 1997). With ligand expression, the Bmp signaling pathway might be expected to be active. Anti-phospho-Smad antibody labeling showed strong nuclear immunoreactivity in the epithelium of the olfactory placode and first branchial arch (Fig. 6B,C, red arrowheads) in a pattern that corresponded closely with the domain of Bmp4 expression. The anti-Bmpr2 antibodies also labeled an adjacent section in the olfactory placode and first branchial arch epithelium but the labeling pattern appeared more cell surface than nuclear (Fig. 6D,E, red arrowheads). This indicated a strong spatial correlation between the Bmp4 ligand, the Bmp type 2 receptor and nuclear phospho-Smad labeling suggesting that we are able to detect cells in which this pathway is active.

Interestingly, we could detect distinct patterns of anti-phospho-Smad and anti-Bmpr2 antibody immunoreactivity in the developing eye. Phospho-Smad immunoreactivity was detected in the lens vesicle at E12.5 in the equatorial cells adjacent to the anterior retinal rim (Fig. 6F, red arrowheads). The labeling was nuclear as indicated by the turquoise color obtained when merging phospho-Smad immunoreactivity with blue Hoechst nuclear staining (Fig. 6G, red arrowheads). Strong nuclear phospho-Smad immunoreactivity was also present in presumptive retinal and pigmented epithelial cells in anterior optic cup rim (Fig. 6F,G, broken red line). Importantly, there was also clear cell surface immunoreactivity for anti-Bmpr2 antibodies in cells of the lens vesicle equator (Fig. 6H,I, red arrowheads) and the anterior optic cup rim. Anti-phospho-Smad and anti-Bmpr2 antibodies labeled lens vesicle cells on the nasal side (Fig. 6F-I, white N) but was also present on the temporal side (data not shown). Combined, these data indicated that Bmp signaling pathways are active in the equatorial lens vesicle cells that are the likely precursors of the primary lens fiber cells.

We have used three distinct experimental strategies to address the issue of whether Bmp signaling is involved in development of the mouse lens. We have taken advantage of the Bmp inhibitor noggin (Zimmerman et al., 1996) to suppress Bmp activity in explant cultures of eye primordia. This results in attenuated differentiation of primary fiber cells as indicated by the smaller lenses that develop. In addition, we show that expression of a dominant-negative form of the type 1 Bmp receptor Alk6 from two different lens-expressed transgenes has resulted in defects in primary lens fiber cell development. Interestingly, the primary fiber cell differentiation defects arise only in the ventronasal quadrant of the lens. Our analysis of wild-type lens development shows that the asymmetrical lens development defect in the transgenic animals has its origin in a transient phase of asymmetrical primary fiber cell differentiation in the normal lens. To our knowledge, this is the first time this aspect of mouse lens development has been recognized. Finally, we have taken advantage of antibodies to Bmpr2 (Gilboa et al., 2000) and Bmp-responsive phospho-Smads (Korchynskyi et al., 1999) to determine whether there is evidence for Bmp signaling machinery and responsiveness in cells of the early lens. This has shown that the equatorial cells of the lens vesicle express Bmpr2, but most important, also show nuclear immunoreactivity with antibodies raised to phospho-Smad1. This indicates that at the inception of primary fiber cell differentiation, the precursors are responding to Bmp signals. These data strongly suggest that Bmp signaling is important for primary fiber cell development but also raise a number of interesting issues.

The anti-phospho-Smad1 antibody we have used in this analysis detects phosphorylated Smad1, Smad5 and Smad8 (Korchynskyi et al., 1999). In some circumstances this antibody crossreacts at a low level with phospho-Smad3 (that functions downstream of Tgfβ receptors) (Korchynskyi et al., 1999) potentially complicating interpretation of the current data. Importantly however, there is evidence that at physiological expression levels, Tgfβ receptors do not dimerize with Bmp receptors (Massagué, 1998) and so there is no expectation that dominant-negative receptors from each class would cross-react. Thus, with the current analysis and recent work in which the role of Tgfβ signaling in lens development was investigated (de Iongh et al., 2001) we can conclude that both Bmp and Tgfβ signaling are independently important in the development of the primary lens fiber cells. This suggestion is reinforced by the observation that primary fiber cell development can be inhibited in explant culture by noggin (Fig. 1), an inhibitor of BMP but not Tgfβ activity.

The published evidence indicates that both Bmp7 and Bmp4 are expressed in the eye primordium (see Fig. 7 for summary) and suggests that they have important roles in induction of the lens (Furuta and Hogan, 1998). Bmp7-null mice demonstrate that Bmp7 regulates expression of the transcription factor Pax6 in the lens placode (Wawersik et al., 1999). As Pax6 is both necessary (Ashery-Padan et al., 2000) and sufficient (Altmann et al., 1997) for the earliest stages of lens development, this has argued that Bmp7 has an important role in lens induction (Wawersik et al., 1999). By contrast, Bmp4 does not regulate expression of Pax6, but is required for the upregulation of Sox2 (Furuta and Hogan, 1998), a transcription factor that is involved in regulating the expression of crystallin genes sometimes in a complex with Pax6 (Kamachi et al., 2001). As the ectoderm enhancer used to direct transgene expression in the EE-1.0-K-Alk6^DN transgenic mice is expected to be expressed in the lens lineage from E8.75 (Williams et al., 1998), the question arises of why we do not see a defect in lens induction. There are several reasons. One explanation is that the level of Bmp4 and Bmp7 signaling activity in the induction phases of lens development is high, and that the transgene we describe does not provide a sufficiently high level of expression to inhibit Bmp signaling pathways. Another, more satisfying, explanation is based on the biochemistry of the Bmp receptor signaling system.

It has been established that type 1 and type 2 Tgfβ family receptors can heterodimerize in various combinations (Massagué, 1998). Expression studies have shown that the type 1 Bmp receptors Alk2 and Alk3 are expressed in the early lens (Furuta and Hogan, 1998; Yoshikawa et al., 2000) (Fig. 7). Furthermore, the current analysis shows that Bmpr2 is expressed in the developing lens from E9.5 to E12.5 (Fig. 6 and see Fig. 7 for summary). Thus, there are a limited number of receptor heterodimers that might mediate development. It has also been shown that despite the option to heterodimerize, the different type 1/type 2 receptor combinations have very different abilities to signal in response to a particular ligand (Table 2). Significantly, Bmp7 has a higher affinity for the Alk2/Bmpr2 combination than for any other including Bmpr2/Alk6. This argues that in the EE-1.0-K-Alk6^DN transgenic mice, the dominant-negative Alk6 is unlikely to impede Bmp7 signaling at the stage of lens induction. Additionally, as Bmp7 does not appear to signal efficiently through Alk3 (Macias-Silva, 1998), it is likely that signaling through Bmpr2/Alk2 (Liu et al., 1995) is the major effector of Bmp7 action in lens induction. It is more difficult to explain why dominant-negative Alk6 does not apparently inhibit Bmp4 signaling in lens induction as it can clearly pair with Bmpr2 and efficiently transmit Bmp4 signals (Table 2). It will be interesting to observe the outcome of lens lineage conditional Alk2 and Alk3 targeting experiments that might definitively identify the type 1 receptor(s) crucial for the induction phases of lens development.

The current analysis provides strong evidence that as well as the early inductive action of Bmp7 and Bmp4, lens development involves a later phase of Bmp signaling that stimulates development of the primary lens fibers. The transcriptional control elements used to drive expression in the αA-Alk6^DN and EE-1.0-K-Alk6^DN transgenic mice have been well characterized and the lens expression patterns completely understood (Kammandel et al., 1999; Williams et al., 1998) (Fig. 2B). We have also confirmed lens expression (Fig. 2F-H). The Pax6 ectoderm enhancer and the αA-crystallin promoter have expression patterns that overlap in the equatorial cells of the lens vesicle starting at E11.5. Interestingly, these are precisely the cells that based on immunoreactivity for anti-phospho-Smad and anti-Bmpr2 antibodies have the signaling machinery for Bmp signaling and are responding to Bmps at E12.5. Thus, we can suggest (Fig. 8) that the observed suppression of primary fiber cell differentiation is a consequence of dominant-negative Alk6-mediated inhibition of Bmp signaling in the presumptive primary lens fiber cells at the equator of the E12.5 lens vesicle.

Although there has been little published concerning the developmental distinction between primary and secondary lens fiber cells, it has been shown that there is a clear morphological difference (Shestopalov and Bassnett, 2000). Primary fiber cells, the first cells to elongate after the closure of the lens vesicle, appear to have a different cellular morphology and organization than secondary fiber cells in both the human and chick. Primary lens fiber cells are heterogenous in size and shape, while secondary fiber cells appear to be uniformly smooth. The current analysis suggests that Bmp signaling may be one determinant of the distinctions between primary and secondary lens fiber cells.

The data we describe shows that the lens normally goes through a stage where development is anatomically asymmetrical. Specifically, we show that primary lens fiber cells begin to elongate first in the ventrotemporal quadrant of the lens at E12.5, but by E13.5, the nasal side cells have also elongated and the obvious asymmetry is lost. To our knowledge, this is the first time that this aspect of mouse lens development has been recognized.

In the wild-type lens vesicle, Bmpr2 and phospho-Smad immunoreactivity are present in the entire equator and this suggests that all these cells are responding to Bmp signals. This raises the question of why, in the αA-Alk6^DN and EE-1.0-K-Alk6^DN transgenic mice, where lens vesicle stage expression is not expected to be asymmetrical, the primary fiber cell differentiation defect is. The most obvious explanation is that there may be a Bmp ligand that is important for primary fiber cell differentiation that is restricted in its distribution to the ventronasal quadrant of the eye. A preliminary analysis of the expression patterns of many Bmp family members in the primordial eye shows that many are expressed and that the patterns are complex (S. C. F. and R. A. L., unpublished). Given the possibility that Bmps can act as both homodimers and heterodimers and have distinct signaling activity as a consequence (Suzuki et al., 1997), coupled with the existence of soluble modulators of their activity (Cho and Blitz, 1998), it is not unreasonable to suggest that Bmp responses within the eye may be finely restricted spatially.

Thus, we propose a model in which the normal lens vesicle contains two subpopulations of primary lens fiber cell progenitors that respond to distinct differentiation stimuli. We propose that in the ventrotemporal and dorsal lens vesicle, there is an early differentiating population (Fig. 8, blue) in which fiber cell elongation is unaffected by dominant-negative Alk6. This could mean that this domain of lens vesicle cells differentiates in response to non-Bmp ligands, or that if a Bmp ligand is involved (as the uniform distribution of phospho-Smad might suggest), it does not bind efficiently to Alk6-containing receptor heterodimers. We also suggest that there is a group of lens vesicle cells located on the nasal side that differentiates later (Fig. 8, red) in response to a Bmp ligand that can be inhibited by dominant-negative Alk6. This model explains both the normal asymmetry in primary fiber cell differentiation and the asymmetrical phenotype of the αA-Alk6^DN and EE-1.0-K-Alk6^DN transgenic mice.

Other analyses of eye development have identified mutant phenotypes and gene expression patterns that reflect asymmetries. Gene targeting of the retinoid receptors Rxra and Rxrg has shown that this pathway is important for development of the ventral eye (Kastner et al., 1994). In particular, compromised function of these receptors resulted in a ventral rotation of the entire lens suggesting that retinoid signaling is important for development of ventral eye structures. It may be interesting to determine whether retinoid signaling and Bmp-mediated primary fiber cell differentiation are developmentally related. Another example of interest is the expression patterns of the forkhead family members Foxg1 and Foxd1. These are expressed in the nasal and temporal retina, respectively (Dirksen and Jamrich, 1995; Hatini et al., 1994; Tao and Lai, 1992), and it is possible that this type of gene expression pattern might reflect spatially restricted fiber cell differentiation stimuli that are likely to have their origin in the developing retina.

We thank Peter ten Dijke for supporting this work with suggestions, with a critical reading of the manuscript and by providing antibodies to phospho-Smad1. We also thank Helen Makarenkova for help with in situ hybridization, Petra Knaus for antibodies to Bmpr2, Sam Zigler for anti-crystallin antibodies and J. Horvitz for anti-MIP26 antiserum. The laboratory of M. L. R. is supported by RO1 EY12995 from the NEI/NIH. The Lang laboratory is supported by grants from the NEI/NIH (RO1s EY11234, EY10559, EY12370 and EY14102) and by funds from the Abrahamson Pediatric Eye Institute at Children’s Hospital Medical Center of Cincinnati.