|Home | About | Journals | Submit | Contact Us | Français|
BMPs play multiple roles in development and BMP signaling is essential for lens formation. However, the mechanisms by which BMP receptors function in vertebrate development are incompletely understood. To determine the downstream effectors of BMP signaling and their functions in the ectoderm that will form the lens, we deleted the genes encoding the type I BMP receptors, Bmpr1a and Acvr1, and the canonical transducers of BMP signaling, Smad4, Smad1 and Smad5. Bmpr1a and Acvr1 regulated cell survival and proliferation, respectively. Absence of both receptors interfered with the expression of proteins involved in normal lens development and prevented lens formation, demonstrating that BMPs induce lens formation by acting directly on the prospective lens ectoderm. Remarkably, the canonical Smad signaling pathway was not needed for most of these processes. Lens formation, placode cell proliferation, the expression of FoxE3, a lens-specific transcription factor, and the lens protein, αA-crystallin were regulated by BMP receptors in a Smad-independent manner. Placode cell survival was promoted by R-Smad signaling, but in a manner that did not involve Smad4. Of the responses tested, only maintaining a high level of Sox2 protein, a transcription factor expressed early in placode formation, required the canonical Smad pathway. A key function of Smad-independent BMP receptor signaling may be reorganization of actin cytoskeleton to drive lens invagination.
Lens formation is a classical example of embryonic induction, depending on tissue interactions that begin during gastrulation (Grainger, 1992). The morphogenesis of the lens commences on embryonic day 9 (E9) in mice after the optic vesicle, an outpocketing of the ventral diencephalon, comes in close contact with the surface ectoderm. The apposed tissues become tightly adherent (Fig 1A), followed by thickening of the ectoderm in contact with the optic vesicle to form the lens placode (Fig. 1B). On E10, the placode invaginates, together with the optic vesicle, giving rise to the lens pit and optic cup (Fig. 1C). As the lens pit separates from the surface ectoderm, it forms the lens vesicle (Fig. 1D). Cells in the posterior of the lens vesicle withdraw from the cell cycle and elongate to form the primary fiber cells on E11, occluding the lumen of the vesicle (Fig. 1E). A fully formed lens consists of an anterior sheet of proliferating epithelial cells covering a posterior mass of post-mitotic fiber cells (Fig. 1F). In most species, failure of the optic vesicle to contact the surface ectoderm prevents lens formation.
Lens formation involves signaling by two members of the bone morphogenetic protein family of morphogens, BMP4 and BMP7 (Dudley et al., 1995; Furuta and Hogan, 1998; Jena et al., 1997; Luo et al., 1995; Wawersik et al., 1999). As members of the TGFs superfamily, BMPs activate a heteromeric complex of type I and type II receptors. The receptors signal by phosphorylating cytoplasmic Smad proteins (R-Smads), which then associate with the co-Smad, Smad4. The R-Smad/Smad4 complex accumulates in the nucleus to modulate transcription (Heldin et al., 1997). There is increasing evidence that TGFs superfamily receptors also activate pathways that do not depend on the canonical Smad pathway (Derynck and Zhang, 2003; Heldin and Moustakas, 2006; Moustakas and Heldin, 2005).
BMP7 and the type I BMP receptors Acvr1 (originally called Alk2) and Bmpr1a (Alk3) are expressed in the mouse lens placode (Dudley and Robertson, 1997; Furuta and Hogan, 1998; Wawersik et al., 1999; Yoshikawa et al., 2000). In contrast, expression of the third type I BMP receptor, Bmpr1b (Alk6), appears to be limited to a part of the future retina and the head mesenchyme in the developing eye (Furuta and Hogan, 1998). Although BMP4 expression is initially seen in both optic vesicle and the overlying ectoderm, it becomes restricted to the optic vesicle during placode formation (Furuta and Hogan, 1998). Germline deletion of Bmp4 or Bmp7 prevented lens formation in most (Bmp7) or all cases (Bmp4) (Dudley et al., 1995; Furuta and Hogan, 1998; Jena et al., 1997; Luo et al., 1995; Wawersik et al., 1999). Expression pattern, tissue recombination and other rescue experiments in the Bmp4 null background showed that BMP4 is required for the opticvesicle to manifest its lens-inducing activity (Furuta and Hogan, 1998). However, BMP4 was not able to induce lens formation from the ectoderm unless the optic vesicle was also present. This result raised the possibility that BMP4 functions in lens formation by promoting the production of an inducer by the optic cup, rather than by acting directly on the ectoderm. The expression pattern of BMP7 suggests that it functions predominantly in the lens placode to regulate lens induction (Wawersik et al., 1999). However, because BMP4 and BMP7 null mice lack the ligands in the lens placode, the optic vesicle and in the periocular mesenchyme, conclusively determining the functions of BMP signaling in these mutually interacting tissues is difficult. A few of the genes downstream of BMP signaling are known, but the signaling pathways and cellular processes that orchestrate lens formation by BMP signaling are not understood. (Furuta and Hogan, 1998; Wawersik et al., 1999). The early embryonic lethality of the Bmp4 knockout mice and the variability in the phenotype of the Bmp7 null animals complicate attempts to address these issues.
To better understand the cellular and molecular mechanisms underlying lens induction, we used a Pax6-Cre transgene (LeCre) that is expressed in the early lens placode (Ashery-Padan et al., 2000) to inactivate one or both of the two type I BMP receptors that are expressed in the lens-forming ectoderm. We found that Bmpr1a contributed to the survival of placode cells, while Acvr1 promoted their proliferation. Although neither receptor was required for lens formation, conditional deletion of both from the surface ectoderm reduced placode thickening and prevented lens invagination, leading to eyes that lacked a lens. These results demonstrate that BMPs signal directly to the lens forming ectoderm to promote lens formation.
To investigate the downstream pathways that mediate BMP signaling, we also deleted the R-Smads, Smad1 and Smad5, or the co-Smad, Smad4 from the prospective lens placode. Among the several aspects of lens induction that were regulated by BMP receptor signaling, a few were mediated by the canonical Smad signaling pathway, but most were not. Lens formation, placode cell proliferation, the expression of FoxE3, a transcription factor required for later lens development, and expression of the abundant lens protein, αA-crystallin were regulated by BMP receptors in a Smad-independent manner. Placode cell survival depended on R-Smad signaling, but was independent of Smad4. Of the aspects of lens formation studied, only full expression of Sox2, a transcription factor expressed early in placode formation, was mediated by BMP signaling through the canonical Smad pathway.
The Bmpr1a; Acvr1DCKO (double conditional knockout) lens ectoderm cells failed to reorganize the actin cytoskeleton to their apical ends at the onset of invagination. Based on this observation, we propose that the reorganization of the actin cytoskeleton, which drives the invagination of the lens placode, is an essential function of BMP signaling leading to lens formation.
Mice expressing Cre recombinase under the control of Pax6 P0 enhancer/promoter (Le-Cre) were described previously (Ashery-Padan et al., 2000). Primers for genotyping mice carrying the Cre transgene or the floxed alleles used in this study, (Acvr1fx(exon7) (Dudas et al., 2004) and Bmpr1afx(exon2 )(Gaussin et al., 2002), Smad4fx(exon8) (Yang et al., 2002), Smad1fx(exon2) (Huang et al., 2002) and Smad5fx(exon2) (Umans et al., 2003)) were genotyped by PCR. Genomic DNA from embryonic tail tissue was extracted using the HotSHOT method (Truett et al., 2000). PCR conditions were selected according to the Universal PCR protocol (Stratman et al., 2003). Mice that were homozygous floxed for BMP receptor or Smad genes, one of which was Cre-positive, were mated to generate 50% Cre-positive (conditional knockout, CKO) and 50% Cre-negative offspring (WT). Cre-positive animals were always mated to Cre-negative animals, assuring that Cre-positive offspring inherited only one copy of the Cre transgene.
Embryos or post-natal heads were fixed in 10 % formalin overnight at room temperature, embedded in 5 % agarose, processed and embedded in paraffin and sectioned 4 μm. For morphological studies, sections were stained with hematoxylin and eosin (Surgipath, Richmond, IL). For antibody staining, the slices were deparaffinized and rehydrated. Endogenous peroxidase activity was inactivated with 3% H2O2 in methanol for 30 min at room temperature for those samples that would be treated for horseradish peroxidase (HRP). Epitope retrieval was performed in 0.01 M citrate buffer (pH 6.0) either at 100° C for 20 min using a water bath or by placing slides in a Decloaking Chamber (Biocare Medical, Walnut Creek, CA) for 3 min Slides were then incubated in blocking solution containing 20% inactivated normal donkey serum for 30 min at room temperature followed by incubation in primary antibodies overnight at 4° C. The primary antibodies used were anti-αA crystallin at 1:1000 dilution (a gift from Dr. Usha Andley), anti-Smad4 (Epitomics, Burlingame, CA) at 1:100 dilution, anti-Pax6 at 1:500 dilution (Developmental Studies Hybridoma Bank, Iowa City, IA) and anti-Sox2 at 1:1000 dilution (Chemicon, Temecula, CA). Slides were then incubated for 1 hr at room temperature either with Alexa-Fluor-labeled secondary antibodies (Molecular Probes, Eugene, OR) or biotinylated secondary antibodies (Vector Laboratories, Burlingame, CA). Slides incubated with biotinylated secondary antibodies were treated with the ABC-peroxidase reagent from Vectastain Elite ABC Kit (Vector Laboratories, Burlingame, CA) followed by treatment with diaminobenzidine (DAB) (Sigma, St. Louis, MO) and H2O2. The slides were washed with PBS, and counterstained with hematoxylin (Surgipath, Richmond, IL).
To compare the levels of nuclear Pax6 in the knockouts, 12 nuclei were picked at random from the placode and 12 from the optic vesicle. After outlining the nuclei, the staining intensities were measured using ImageJ software, version 1.36b (National Institutes of Health, Bethesda, MD). To compare the levels of nuclear Sox2 in the knockouts, 12 nuclei were picked at random, using TOPRO as the nuclear marker, from the placode and optic vesicle. Nuclei were outlined and intensities of Sox2 immunostaining were measured using ImageJ.
Terminal deoxynucleotidyl transferase (TdT)-mediated deoxyuridine triphosphate nick end-labeling (TUNEL) was done with an Apoptag kit (Chemicon, Temecula, CA). The deparaffinized slides were treated with 3% H2O2 in methanol for 30 min, followed by proteinase K treatment (20 μg/ml) for 15 min. Slides were incubated with TdT enzyme in equilibration buffer for 1 hour at 37° C. The reaction was terminated with wash buffer provided by the manufacturer for 10 min at room temperature. Anti-digoxigenin- peroxidase conjugate was added for 30 min at room temperature, followed by DAB + H202 treatment. Slides were counterstained with hematoxylin.
For BrdU staining, pregnant females or post-natal day 3 (P3) mice were injected with 50 mg/kg of body weight of a mixture of 10 mM BrdU (Roche, Indianapolis, IN) and 1 mM 5-fluoro-5′-deoxyuridine (Sigma, St. Louis, MO) and sacrificed after 1 hour. A monoclonal anti-BrdU antibody (1:250) (Dako, Carpinteria, CA) was used with a Vectastain Elite Mouse IgG ABC kit as described above. Sections were counterstained with hematoxylin.
Some embryos were fixed in 10% formalin for an hour. The heads were cut in half, embedded in 5% agarose and sectioned into 100 μm thick sections using a vibrating tissue slicer (EM Sciences, Hatfield, PA). The sections were permeabilized and blocked in PBS supplemented with 0.5% Triton X-100 and 5% goat serum and labeled overnight with antibodies specific for pSmad1/5/8 (Cell Signaling Technology, Danvers, MA), FoxE3 (a gift from Dr. Peter Carlsson), αA crystallin (from Dr. Usha Andley), E-cadherin (Cell Signaling Technology, Danvers, MA), ZO-1 (Invitrogen, Carlsbad, CA), laminin α1 (generated by Dr. Dale Abrahamson), laminin α5 (generated by Dr. Jeffery Miner), laminin γ1 (Chemicon, Temecula, CA) and entactin (Chemicon, Temecula, CA). All these antibodies were used at a dilution of 1:250. The sections were then washed using PBS with 0.5% Tween 20, incubated in the second secondary antibody for 2 hours, washed again in PBS with Tween 20 and mounted using a 1:1 dilution of VectaShield (Vector Laboratories, Burlingame, CA) in PBS. Some of the sections were incubated with fluorescent labeled phalloidin, TOPRO or TOTO-1 along with the secondary antibodies. Alexa-Fluor labeled secondary antibodies (Molecular Probes, Eugene, OR) and Alexa-Fluor labeled phalloidin (Molecular Probes, Eugene, OR) were used at 1:1000 dilution. TOPRO or TOTO-1 (Molecular Probes, Eugene, OR) were used at a dilution of 1:10,000.
All the brightfield images were taken using the Olympus BX60 microscope (Olympus, Melville, NY) and Spot camera (Spot Diagnostic Instruments, Sterling Heights, MI). The fluorescent images were taken either using the Olympus BX51 with Spot camera or the Zeiss 510 confocal microscope (Carl Zeiss, Thornburgh, NY).
To analyze the thickness of the placode, 5 equidistant points were marked across the length of the placode on the images of H&E stained embryo heads sections and the height of the tissue was measured at those points using Spot camera software. The extent of contact between the surface ectoderm and the optic vesicle was also measured using the Spot camera software.
For statistical analysis of two groups of samples, an unpaired t-test was performed using GraphPad InStat, Version 3.05 (GraphPad Software, San Diego, CA). Error bars are ± S.E.M.
Previous studies have shown that the BMP ligands BMP4 and BMP7 are required for lens formation (Dudley et al., 1995; Furuta and Hogan, 1998; Jena et al., 1997; Luo et al., 1995; Wawersik et al., 1999). To determine the roles of different BMP receptors in the formation of the lens, we inactivated the genes for the type I BMP receptors Bmpr1a or Acvr1 in the surface ectoderm using Cre recombinase driven by the Pax6 P0 promoter/enhancer (Le-Cre). In these transgenic mice, Cre is expressed in the prospective lens placode by E9.0 (Ashery-Padan et al., 2000).
Loss of Bmpr1a did not prevent lens formation (Fig. 2B). Bmpr1aCKO lenses were reduced in size with defects in fiber cell differentiation (Beebe et al., 2004). In the lens placode, loss of Bmpr1a resulted in a more than two-fold increase in the TUNEL labeling index (p < 0.0001) (Fig. 2E), with no significant alteration of the BrdU labeling index (Fig. 2J).
Deletion of Acvr1 also resulted in the formation of lenses that were reduced in size (Fig. 2C), with several defects that appeared later in lens formation (Rajagopal et al., 2008). Acvr1CKO placodes showed a significant decrease in BrdU labeling compared to placodes from wild-type littermate (p < 0.01) (Fig. 2K), but with no difference in the TUNEL-labeling index (Fig. 2F). Although the two type I BMP receptors separately maintained normal levels of cell proliferation and survival during the formation of the lens placode, neither was required for lens formation.
To test whether the type1 BMP receptors act redundantly to regulate lens formation, we generated Bmpr1a; Acvr1DCKO mice. The double receptor knockout ectoderm displayed a fully penetrant phenotype of failure of lens formation (Fig. 2D). At E10.5, the wild-type lens vesicle expressed the lens protein, αA-crystallin (Fig. 2H). In the Bmpr1a; Acvr1DCKO eye, αA-crystallin-positive cells were not detected (Fig. 2I). Consistent with the expression of BMP ligands and receptors in the lens placode (Dudley and Robertson, 1997; Furuta and Hogan, 1998; Wawersik et al., 1999; Yoshikawa et al., 2000), wild-type placode cells stained with an antibody against the phosphorylated form of the BMP-activated Smads, Smad1/5/8 (pSmad1/5/8) (Fig. 2M). In contrast, the Bmpr1a; Acvr1DCKO ectoderm showed greatly reduced staining for pSmad1/5/8 (Fig. 2N). In the knockout embryos, pSmad1/5/8 levels were maintained in the optic vesicle, where Cre is not expressed. The double receptor knockout ectoderm displayed increased cell death (p < 0.05) (Fig. 2G) and decreased proliferation (p < 0.05) (Fig. 2L), similar to the single receptor knockouts.
Since lenses fail to form upon inactivation of Bmpr1a and Acvr1, we determined whether these receptors signal through the downstream co-Smad, Smad4, to promote lens formation. Conditional deletion of Smad4 did not inhibit lens formation (Fig. 3A). Smad4 antibodies stained the cytoplasm and nuclei of cells in wild type lens placodes and optic vesicles (Fig. 3B). However, Smad4 staining was undetectable in all but a few of the placode cell nuclei in Smad4CKO embryos (Fig. 3C). The increased cell death and decreased proliferation observed in the Bmpr1a; Acvr1DCKO surface ectoderm was also not seen in Smad4CKO lens placodes (Fig. 3D and E). Therefore, Smad4 is not required for lens formation or promotion of cell survival or proliferation in the lens placode.
In the canonical Smad signaling pathway, R-Smads interact with Smad4 to regulate gene expression in response to ligands of the TGFs superfamily (Moustakas et al., 2001). Since lens formation did not require Smad4, we determined whether it required the R-Smads that are downstream of BMP receptors, Smad1 and 5. The third BMP-specific R-Smad, Smad8, is not expressed in the surface ectoderm during lens formation (Arnold et al., 2006). Inactivation of Smad1 and Smad5 did not prevent lens formation (Fig 3F). Greatly reduced staining for pSmad1/5/8 in the Smad1; Smad5DCKO placode (Fig. 3H) demonstrated that the R-Smads had been efficiently deleted. The antibody to pSmad1/5/8 binds to a phosphopeptide that is nearly identical in the three proteins. Therefore, our data suggest that Smad8 expression did not increase to compensate for the absence of Smad1 and 5. Microarray analysis of transcripts from wild type and Bmpr1a; Acvr1DCKO lens placodes confirmed that Smad8 mRNA was not detectable in either genotype (not shown). Thus, signaling by BMP4 and BMP7 through Bmpr1a and Acvr1 activates Smad-independent pathways to promote lens formation.
Although the R-Smads, Smad1 and Smad5, were not required for lens formation, we determined whether they might promote cell survival or proliferation in response to BMP signaling. Smad1; Smad5DCKO placodes showed a significant increase in TUNEL-positive cells compared to the wild type littermates (p < 0.05) (Fig. 3I). Lens placodes lacking either Smad1 or Smad5 also had significantly more TUNEL labeling (not shown). The BrdU labeling index was unaffected in the R-Smad double knockouts (Fig. 3J) and in each of the single knockouts (not shown). Therefore, Bmpr1a signals through Smads 1 and 5 to promote cell survival, while Acvr1 promotes lens cell proliferation by a Smad-independent mechanism.
Pax6, a transcription factor with paired and homeobox domains, is required in the surface ectoderm for lens formation (Ashery-Padan et al., 2000). Pax6 expression is gradually lost in the Bmp7 null placodal ectoderm, which fails to form a lens (Wawersik et al., 1999). Since Bmpr1a; Acvr1DCKO ectoderm does not form a lens, we tested the possibility that Pax6 levels might be altered in the knock-out ectoderm. Pax6 levels were indistinguishable in wild-type placodes and in double receptor knockout surface ectoderm (Fig. 4A and B). As expected, the levels of Pax6 were also unaltered in Smad4CKO and Smad1; Smad5DCKO placodes (Fig. 4C and D). Quantification confirmed the similar levels of Pax6 protein in the wild type, the double receptor knockout ectoderm and in Smad1; Smad5DCKO and Smad4CKO placodes (Fig. 4A′-D′). Expression of Sox2, a HMG box-containing transcription factor, is reduced in the lens-forming ectoderm of Bmp4 null mice and lost in the ectoderm of Bmp7 knockout mice (Furuta and Hogan, 1998; Wawersik et al., 1999). Quantification confirmed that Sox2 protein was present at a lower level than wild type in the double receptor knockout ectoderm and in Smad1; Smad5DCKO and Smad4CKO placodes (Fig. 4E–H and 4E′-H′). Therefore, maximal levels of Sox2 depend on BMP receptor signaling through the canonical R-Smad-Smad4 pathway.
FoxE3 is a lens-specific member of the forkhead family of transcription factors. It is initially detected late in placode formation (Blixt et al., 2000). Although lenses form in mice mutant for Foxe3, severe defects in lens cell proliferation and survival appear soon after invagination (Blixt et al., 2000; Medina-Martinez et al., 2005). Wild-type placode cells at E10 (33 somites) showed nuclear FoxE3 staining (Fig. 4I), but no specific staining was detected in the surface ectoderm of Bmpr1a; Acvr1DCKO littermates (Fig. 4J). However, FoxE3 expression was unaltered in Smad4CKO and Smad1; Smad5DCKO embryos at a similar stage (Fig. 4K and L). Expression of αA-crystallin commenced in a few cells at the 30-somite stage (early lens invagination) (Fig. 4M). In 36-somite embryos (the lens vesicle stage), αA-crystallin was present in all lens cells (Fig. 4M′). Immunostaining for αA-crystallin was not detected in Bmpr1a; Acvr1DCKO prospective lens ectoderm in 33-somite embryos (Fig. 4N), but was present in Smad4CKO and Smad1; Smad5DCKO lens pits (Fig. 4O and P). Therefore, Bmpr1a and Acvr1 signaling initiates the expression of FoxE3 and αA-crystallin in the lens placode in a Smad-independent fashion.
During the formation of the wild-type lens placode, the surface ectoderm and the underlying optic vesicle remain in close contact, only separating during the invagination of the lens and optic vesicle (Fig. 5A and A′). The surface ectoderm and the underlying optic vesicle were in close contact in wild type and Bmpr1a; Acvr1DCKO, 24 somite embryos (Fig. 5B, C). However, at the late placode stage (28 somites) the contact area between these tissues decreased (p < 0.0001; Fig. 5B′, C′). Although wild-type and Bmpr1a; Acvr1DCKO lens-forming ectoderm were of similar thickness early in placode formation (Fig. 5D), the Bmpr1a; Acvr1DCKO ectoderm failed to thicken as much as wild type during placode formation (Fig. 5D′). The ventral region of the placode, where separation from the optic vesicle first occurred, was thinner than the dorsal region. Thus, signaling through Bmpr1a and Acvr1 may contribute to placode formation by maintaining contact between the surface ectoderm and the optic vesicle (Hendrix and Zwaan, 1975).
Since the polarized distribution of plasma membrane proteins is important for cell organization and the function of the cytoskeleton (Knust, 2000), defects in cell polarity could hinder cell elongation or prevent the basal secretion of basement membrane components. Because Bmpr1a; Acvr1DCKO ectoderm cells appeared less organized than wild type and did not fully elongate or remain adherent to the optic vesicle, we determined whether these cells showed altered expression or distribution of markers of cell polarity. The expression and distribution of the adherens junction marker, E-cadherin, and the tight junction marker, ZO-1, appeared similar in wild-type and double knockout ectoderm cells (Fig. S1). This suggested that the cell polarity was not grossly affected in the Bmpr1a; Acvr1DCKO cells.
The lens placode and the optic vesicle are held in contact by a zone of extracellular matrix (ECM), which is secreted by both tissues (Hendrix and Zwaan, 1975; Silver and Wakely, 1974). Since Bmpr1a; Acvr1DCKO ectoderm had reduced contact with the underlying optic vesicle, we determined whether any defects were apparent in the basal laminae of the two epithelia or in the ECM between them. The overall composition and individual components of the basal lamina and ECM were not obviously affected in the double BMP receptor knockouts (Fig. S2).
One proposed mechanisms of lens invagination involves the contraction of actin microfilaments at the apical ends of placode cells (Wrenn and Wessells, 1969). Prior to lens invagination, staining with fluorescently-labeled phalloidin showed that filamentous actin (F-actin) was distributed around the apical, basal and lateral surfaces of wild type lens placode cells (Fig. 6A). As the placode began to invaginate to form the lens pit at E10.0, phalloidin staining decreased along the lateral surfaces of the cells and increased at their apical ends (Fig. 6C). In contrast to the discontinuous distribution seen at the placode stage (Fig. S1E), the apical distribution of ZO-1 appeared continuous as wild-type placode cells began to invaginate (Fig. 6E). This suggested that contraction of apical actin filaments draws the apical ends of the placode cells together to cause bending of the placode and formation of the lens pit. In Bmpr1a; Acvr1DCKO ectoderm cells, F-actin did not accumulate at the apical ends of the cells, remaining uniformly distributed around the cell periphery (Fig. 6B, D). At the same time, the ZO-1 distribution remained discontinuous at the apical ends of the cells (Fig. 6F), implying failure of apical contraction. The apical redistribution of F-actin occurred normally in Smad4CKO and Smad1; Smad5DCKO embryos (Fig. 6G and H). We conclude that BMP signaling through Bmpr1a and Acvr1 promotes reorganization of the actin cytoskeleton prior to lens invagination in a Smad-independent manner, thereby facilitating lens morphogenesis.
Germline knockout of Bmp4 or Bmp7 in mice demonstrated that BMP signaling is essential for lens induction, although these studies did not identify the cellular mechanisms underlying this process (Dudley and Robertson, 1997; Furuta and Hogan, 1998; Jena et al., 1997; Luo et al., 1995; Wawersik et al., 1999). Here, we provide evidence that the type I BMP receptors, Bmpr1a and Acvr1, act redundantly and in a Smad-independent manner to mediate lens formation. Although their most important functions do not require Smads, some of their actions depend on receptor-activated Smads, independent of Smad4, while others use the canonical R-Smad-Smad4 pathway. Each receptor activates one or more Smad-independent mechanisms that redistribute the actin cytoskeleton to the apical ends of lens placode cells, a process that is likely to drive lens invagination.
Our results provide genetic evidence that Bmpr1a and Acvr1 play redundant roles; either receptor is sufficient for lens formation. Another example of redundant BMP receptor function is seen during retinal development. Developing retinae lacking Bmpr1a and Bmpr1b (Alk6) exhibit severe eye defects resulting from reduced growthand failure of retinal neurogenesis, defects not seen in the single receptor knockouts (Murali et al., 2005). In addition to demonstrating their redundancy, the present study provides a unique example of dissimilar and shared functions performed by two BMP receptors during the formation of a single tissue.
Deletion of individual type I BMP receptors showed they have unique functions, which are not required for lens formation. Bmpr1a promotes the survival of placode cells, whereas Acvr1 promotes their proliferation. Acvr1 signaling is mitogenic for other cell types during their differentiation, specifically, the neural crest-derived cells of Meckel’s cartilage (Dudas et al., 2004). As in the lens placode, deletion of Bmpr1a in the lung epithelium leads to increased apoptosis (Eblaghie et al., 2006). However, unlike its function in the placode, loss of Bmpr1a in the lung epithelium also results in decreased proliferation (Eblaghie et al., 2006). The distinct functions of Bmpr1a and Acvr1 in the lens appears to be mediated by different downstream pathways. While promotion of cell survival appears to require the R-Smads, Smad1 and Smad5, cell proliferation is maintained by one or more Smad-independent mechanisms. Analyzing Bmp4 and Bmp7 null surface ectoderm for defects in cell survival and proliferation may determine whether one or both ligands elicit these distinct responses from the type I BMP receptors.
Our studies revealed that, among the several facets of lens formation regulated by BMP signaling, some are mediated by the canonical Smads, but most are not. Several instances of Smad-independent signaling downstream of BMP receptors have been reported [reviewed in (Moustakas and Heldin, 2005)]. Most of these involve the activation of specific MAP kinase modules, particularly the p38 pathway. Type II BMP receptors may also signal through the cytoplasmic kinase, LIMK1 (Foletta et al., 2003; Lee-Hoeflich et al., 2004; Wen et al., 2007). Further studies are needed to determine whether Bmpr1a and Acvr1 promote lens formation by activating one of these pathways, or an as yet unknown Smad-independent signaling cascade.
BMP signaling promoted placode cell survival and maintained full expression of Sox2 protein in a Smad-dependent manner. Lens placodes lacking Smad1, Smad5, or both R-Smads showed increased cell death, with no change in cell proliferation. However, maintaining cell survival did not require the Co-Smad, Smad4, providing evidence for an R-Smad-dependent, Smad4-independent signaling mechanism. This is reminiscent of the effect of deleting Smad4 from the mouse epiblast, where only a subset of BMP-dependent responses were affected (Chu et al., 2004). However, the requirement for the R-Smads in these Smad4-independent responses has not been examined. Although Smad4 is usually considered to play a central role in Smad-dependent signaling, emerging evidence suggests that R-Smads can bind to factors other than Smad4, such as TIF-1γ (He et al., 2006). Although Smad4 is expressed during early stages of lens development and localizes to the nuclei of placode cells, it is possible that surrogates of Smad4, similar to TIF-1γ, mediate some of the effects of Bmpr1a and Acvr1 during early lens development.
Bmpr1aCKO or Acvr1CKO lens placodes thicken normally and maintain contact with the underlying optic vesicle. However, the double knockout ectoderm had diminished ability to sustain contact with the optic vesicle late in placode formation, especially in the ventral region of the placode. Cells in this region were also not as thick as cells in the dorsal region of the placode. Cells of Bmpr1a; Acvr1DCKO placodes had normal apical-basal polarity and contributed to a basement membrane that appeared normal in extent and composition. Therefore, it seems less likely that precocious separation was caused by defects in the secretion of components of the extracellular matrix. Contact between the optic vesicle and the ectoderm was initially established, but was not maintained as the optic vesicle began to invaginate. The invagination of the ventral portion of the optic vesicle differs from invagination of the dorsal and lateral regions, due to the formation of the ventral optic fissure (Morcillo et al., 2006). This may account for the separation of the optic vesicle from the ectoderm in this region. Separation of the surface ectoderm and the optic vesicle was also observed in Bmp4 null embryos (Furuta and Hogan, 1998). It seems most probable that the inability of the knock out surface ectoderm to invaginate synchronously with the optic vesicle caused these tissues to separate.
Previous studies showed that neural crest cells could inhibit the formation of a lens from head ectoderm (Sullivan et al., 2004). Mesenchyme cells were present in the space between the ventral placode and the optic vesicle, raising the possibility that that these cells inhibited lens formation. However, it is not clear what mechanism would have induced the mesenchyme cells to invade the space between the placode and the optic vesicle, since no mesenchyme cells were present in this space at earlier stages of placode formation. Therefore, it seems unlikely that neural crest cell invasion was responsible for inhibiting lens formation.
The underlying cellular mechanisms that lead to failure of lens formation in Bmp4 or 7 knockouts were unclear (Furuta and Hogan, 1998; Jena et al., 1997; Wawersik et al., 1999). In Bmp4 null mice, mRNA encoding Sox2, a transcription factor that is normally expressed in the lens placode, fails to accumulate (Furuta and Hogan, 1998). In mice lacking BMP7, expression of Pax6, Sox2 and secreted Frizzled-Related Protein-2 (sFRP2), an antagonist of the Wnt signaling pathway, is reduced in the surface ectoderm (Wawersik et al., 1999). Conditional deletion of Pax6 in the prospective lens-forming ectoderm prevents lens formation (Ashery-Padan et al., 2000). Mice lacking Sfrp1 and 2 do not appear to have obvious lens defects (Satoh et al., 2006) and mice lacking Sox2 in the lens forming-ectoderm have not been reported. Thus, it seems possible that decreased expression of Pax6 and Sox2 contributes to the failure of lens formation in Bmp4 or 7 null mice. However, deletion of Bmpr1a and Acvr1 did not decrease the levels of Pax6 protein. Sox2 levels were lower in the double knockouts, but were also lower in ectoderm cells lacking Smad4 or the R-Smads, which formed lenses. The results of the present study differ in some ways from those obtained from germline deletion of Bmp7 (Furuta and Hogan, 1998; Wawersik et al., 1999), where absence of Bmp7 resulted in a gradual loss of Pax6 mRNA and a marked reduction in Sox2 transcripts in the lens-forming ectoderm. This difference may be because proteins would be expected to persist longer than their mRNAs. Germline deletion of Bmp7 could also impair the earlier development of tissues that contribute to Pax6 and Sox2 expression in the placode.
It is not clear to what extent thickening of the lens placode is necessary for the subsequent invagination of the surface ectoderm. Hendrix and Zwaan proposed that adhesion between the placode cells and underlying optic vesicle promoted lens invagination by preventing spreading of the surface ectoderm, resulting in placode thickening and subsequent invagination (Hendrix and Zwaan, 1975). Enhancement of placode thickening, either by ethanol treatment or ligature of the ectoderm, resulted in more rapid invagination (Wakely, 1984). Pax6CKO surface ectoderm does not thicken at all (unpublished data) and Bmpr1a; Acvr1DCKO ectoderm forms thinner placodes. In neither case was a lens formed. These observations support the idea that lens placode formation may be required for lens invagination. However, because the cellular mechanisms responsible for lens placode formation are not known, this hypothesis is not easily tested.
Constriction of the apical ends of placode cells has long been suggested as the driving force for lens invagination (McKeehan, 1951; Wrenn and Wessells, 1969) and for the bending of other epithelial sheets. For example, Xenopus embryos lacking Enabled, a member of the Ena/Vasp family, showed defects in apical actin organization and constriction, which impeded the bending of the neural plate (Roffers-Agarwal et al., 2008). Similarly, FGF-induced apical actin redistribution contributes to the initiation of otic placode invagination (Sai and Ladher, 2008). Apical constriction during lens invagination has been compared to a ‘drawstring’ mechanism (Fig. 7) in which the assembly of a network of microfilaments at the apical ends of placode cells is essential for apical constriction and invagination. Our data support this hypothesis, as we observed extensive actin redistribution to the apical ends of the cells during lens placode invagination and constriction of the apical ends of the placode cells, processes that failed to occur in Bmpr1a; Acvr1DCKO ectoderm. Placode cells lacking Smad4 or Smad1 and 5 showed normal apical re-localization of F-actin and went on to form lenses. If this view is correct, future studies may focus on the link between Smad-independent BMP signaling and redistribution of the actin cytoskeleton. Such a link is suggested by the effects of BMP signaling on actin redistribution during the formation of dendritic spines and in growth cone guidance (Lee-Hoeflich et al., 2004; Wen et al., 2007). The relative importance of Smad-independent BMP signaling in development has not been thoroughly evaluated. The results of this and other recent studies (Chu et al., 2004) suggest that much remains to be learned about the downstream pathways used by BMP receptors to mediate their effects.
Lens induction is defined by two hallmark features. At the molecular level, it is characterized by the expression of an array of transcription factors and lens-specific proteins. The cellular aspects of lens formation involve the generation of a placode, followed by its invagination. Our study demonstrates that BMP receptors mediate both the molecular and cellular aspects of lens formation. BMP receptor signaling influences the expression of three molecular markers, Sox2, FoxE3 and αA-crystallin. Both cellular mechanisms contributing to lens formation, lens placode formation and lens invagination, are impaired when BMP receptors are deleted from the lens forming ectoderm. Defects in cell proliferation and survival and/or loss of contact with the underlying optic vesicle may reduce the extent of placode formation, but a placode does form. Failure of cytoskeletal rearrangement appears to be the principal cause of failure of lens invagination in the BMP receptor knockouts.
We thank Drs. Ruth Ashery-Padan and Judy West-Mays for providing the Le-Cre mice and are grateful for the expert technical assistance of Chenghua Wu and Mary Feldmeier with genotyping and Belinda McMahan and Jean Jones with histology and immunohistochemistry. We are indebted to Drs. Jeffrey Miner and Peter Carlsson for the laminin and FoxE3 antibodies, respectively. This work was supported by an unrestricted grant from Research to Prevent Blindness and NIH Core Grant EY02687 to the Dept. of Ophthalmology and Visual Sciences, NIH grant EY04853 to DCB and 5R01HL074862 and 5R01DE013085 to VK. Drs. Lieve Umans and An Zwijsen are indebted to Dr. Danny Huylebroeck for his support of their work.