|Home | About | Journals | Submit | Contact Us | Français|
It is widely accepted that vitreous humor-derived FGFs are required for the differentiation of anterior lens epithelial cells into crystallin-rich fibers. We show that BMP 2, 4, and 7 can induce the expression of markers of fiber differentiation in primary lens cell cultures to an extent equivalent to FGF or medium conditioned by intact vitreous bodies (VBCM). Abolishing BMP 2/4/7 signaling with noggin inhibited VBCM from upregulating fiber marker expression. Remarkably, noggin and anti-BMP antibodies also prevented purified FGF (but not unrelated stimuli) from upregulating the same fiber-specific proteins. This effect is attributable to inhibition of BMPs produced by the lens cells themselves. Although BMP signaling is required for FGF to enhance fiber differentiation, the converse is not true. Expression of noggin in the lenses of transgenic mice resulted in a postnatal block of epithelial-to-secondary fiber differentiation, with extension of the epithelial monolayer to the posterior pole of the organ. These results reveal the central importance of BMP in secondary fiber formation and show that although FGF may be necessary for this process, it is not sufficient. Differentiation of fiber cells, and thus proper vision, is dependent on cross-talk between the FGF and BMP signaling pathways.
Following invagination of the lens placode early in embryogenesis, the cells in the posterior half of the lens vesicle elongate to form the primary fiber cells whereas the anterior cells become the initial lens epithelium. All subsequent growth of the lens is due to proliferation of epithelial cells located near the anterior/posterior boundary of the organ (referred to as the lens equator) followed by their differentiation into secondary fiber cells (reviewed in Piatigorsky, 1981; Lovicu and McAvoy, 2005). The transformation from epithelial cell to secondary lens fiber is characterized by a large increase in cell length and upregulation of fiber-specific proteins including crystallins, aquaporin 0, and the beaded filament subunits CP49 and filensin. Eventually, all intracellular organelles are degraded, and DNA, RNA, and protein synthesis ceases. The process of epithelial-to-fiber differentiation continues throughout life, creating an organ that consists of a monolayer of epithelial cells at its anterior face and a mass of secondary fiber cells arranged in concentric layers around a central core of primary fibers.
Over 20 years of research have demonstrated that FGF signaling is essential for normal lens development, although the precise role of any one FGF or FGF receptor (FGFR) family member in this process remains unclear (Robinson, 2006). Conditional triple deletion of the genes encoding FGFR 1, 2, and 3 in the lens pit blocks lens formation at the vesicle stage (Zhao et al., 2008). Interfering with FGFR function in the lens at later developmental periods results in inhibition of secondary fiber formation (Chow et al., 1995; Robinson et al., 1995a; Stolen and Griep, 2000; Govindarajan and Overbeek, 2001). Conversely, overexpression of several members of the FGF family in the lens causes premature fiber differentiation (Lovicu and Overbeek, 1998; Robinson et al., 1995b; Robinson et al., 1998). It is believed that secondary fiber formation begins at the lens equator because this is where epithelial cells are first exposed to the high levels of FGF that diffuse out of the vitreous body (Schulz et al., 1993).
It has frequently been stated that FGF is the only factor known to be capable of initiating epithelial-to-fiber differentiation (Lovicu and McAvoy, 2005). Although perhaps necessary, it is not known if FGF is sufficient for the entire secondary fiber formation process. Indeed, a number of growth factors including EGF, TGFα, PDGF-A, insulin, and IGF-1 can enhance the synthesis of one or more fiber-specific proteins in transgenic mouse and/or cultured lens cells (reviewed in Lovicu and McAvoy, 2005). A key question in lens development is the identity of non-FGF factor(s) that play a physiologically important role in secondary fiber differentiation. Members of the BMP (bone morphogenetic protein) family of growth factors have been shown to be involved in the early stages of lens development. Germline knockout mice lacking either BMP4 or BMP7 have severe defects in lens placode induction and/or development (Furuta and Hogan, 1998; Wawersik et al., 1999), and expression of a dominant-negative form of the ALK6 BMP receptor (BMPR) leads to inhibition of primary fiber differentiation (Faber et al., 2002). Deletion of the ALK3 BMPR in the lens placode and head ectoderm using the Le-Cre transgene permits lens development to progress to later stages, but is associated with defects within (abnormally thin epithelium) and outside (lack of a cell-free anterior chamber) of the lens that are evident by E13, the first time point examined (Beebe et al., 2004). Because the lens is already perturbed at a stage that developmentally precedes secondary fiber formation, the aforementioned models do not define whether BMPs play an essential and direct role in the differentiation of normal epithelial cells into secondary fibers.
One strategy to bypass these limitations is to use primary lens epithelial cells in tissue culture. The most common in vitro system to study fiber differentiation is central epithelial explants, prepared by excising from whole lenses the monolayer of epithelial cells that faced the anterior pole of the lens in vivo. In the normal lens in vivo, cells in the central epithelium do not undergo fiber differentiation. Instead, it is the cells from the equatorial and preequatorial (i.e., peripheral epithelial) regions that are exposed to growth factors (including FGF) from the vitreous humor and that are the direct precursors of the lens fibers. To include these populations, we dissociate lenses into single cells with trypsin (a process that kills terminally differentiated fibers) and plate the surviving epithelial cells as monolayers. As first shown by Menko and colleagues (Menko et al., 1984; Menko et al., 1987), cells prepared from dissociated E10 embryonic chick lenses faithfully recapitulate the in vivo process of epithelial-to-fiber-differentiation. Our chick cultures (termed dissociated cell-derived monolayers; DCDMLs) are cultured on laminin (the major component of the lens capsule, the in vivo substratum for lens epithelial cells), in the absence of serum to reproduce the avascular environment of the lens (Le and Musil, 1998; Le and Musil, 2001a; Le and Musil, 2001b).
In either central epithelial explants from rodent (McAvoy and Chamberlain, 1989) or chick (Le and Musil, 2001a), or chick DCDMLs (Le and Musil, 2001a), purified FGF2 and FGF1 upregulate the expression of fiber differentiation markers and cause the cells to acquire the morphological hallmarks of fiber formation. These FGFs are present in the vitreous humor of all species examined (Caruelle et al., 1989; Schulz et al., 1993), and are capable of diffusing out of intact vitreous bodies at levels high enough to induce the aforementioned changes (Le and Musil, 2001a). We have shown that tissue culture medium conditioned by intact vitreous bodies was also able to induce phosphorylation of Smad1 on Ser463/465, indicative of activation of the canonical BMP signaling pathway (Musil and Boswell, 2006). In the present study, we have used DCDMLs as well as transgenic mice to investigate the role of BMPs in secondary fiber differentiation. Our results indicate that FGF cannot induce secondary fiber formation in the absence of BMP signaling, revealing a previously unknown, essential role for BMP in lens development and a novel type of cross-talk between the FGF and BMP signaling pathways.
Recombinant bovine FGF2, human BMP4, mouse noggin/Fc chimera, and anti-BMP2/4 antibody (MAB3552) were from R & D Systems (Minneapolis, MN). R3IGF-1 was from GroPep (Adelaide, Australia), and recombinant human BMP2 from Genetics Institute, Inc. (Cambridge, MA). PD173074 and 8-CPT-cAMP were from Calbiochem (La Jolla, CA). BMP7 and anti-BMP7 antibodies (rabbit polyclonal 5086 and mouse monoclonal 1B12) were from Creative Biomolecules (now Curis, Cambridge, MA) (Vukicevic et al., 1994; Lein et al., 2002).
Cultures were prepared from E10 chick lenses and plated at 1.6 × 105 cells/well onto laminin-coated 96-well tissue culture plates plates as previously described in Le and Musil (1998). Cells were cultured in M199 medium plus BOTS (2.5 mg/ml bovine serum albumin, 25 mg/ml ovotransferrin, 30 nM selenium), penicillin G, and streptomycin, with or without additives at 37°C in a 5% CO2 incubator. Cells were fed every 2 d with fresh medium. Anti-BMP antibodies and rabbit IgG were used at 40 ug/ml; noggin at 0.5 ug/ml.
DCDML cultures or freshly isolated, intact mouse lenses were labeled at 37°C with [35S]methionine for 4 h in methionine- and serum-free Dulbecco’s minimum essential medium (DMEM) (and solubilized as previously described (Le and Musil, 1998) (Le and Musil, 2001a). [35S]methionine incorporation into total cellular protein and into δ-crystallin was quantitated after SDS-PAGE using a PhosphorImager (Molecular Dynamics) and Quantity-One software (Bio-Rad).
For CP49 or γ–crystallin, DCDML cultures or whole mouse lenses were solubilized in lysis buffer as previously described (Le and Musil, 2001a). For filensin, ERK, and pSmad1, cells were solubilized directly in SDS-PAGE sample buffer and boiled. Either the entire cell lysate from one well of a 96-well culture plate (ERK or pSmad1), or a 1.5 μg (CP49, filensin) or 0.4 μg (γ–crystallin) aliquot was analyzed by SDS-PAGE. Proteins were transferred to polyvinylidene fluoride membranes, and the blots probed with: rabbit anti-mouse CP49 polyclonal serum (#899 or 900), rabbit anti-filensin serum (#76PF) (both generous gifts of Paul FitzGerald, University of California, Davis), anti-p44/42 MAP kinase polyclonal rabbit antibody, anti-phospho-p44/42 MAP kinase E10 monoclonal mouse antibody (specific for activated ERK) (both purchased from Cell Signaling Technology, Danvers, MA), anti-phospho-Smad1 (Ser463/465) (#06-702; Millipore, Billerica, MA), or rabbit anti-rat γ–crystallin antibody (David et al., 1987) (kindly provided by Larry David, Oregon Health & Science University). Immunoreactive proteins were detected using secondary antibodies conjugated to either IRDye800 (Rockland Immunochemicals) or Alexa Fluor 680 (Molecular Probes, Eugene, OR) and directly quantified using the LI-COR Biosciences Odyssey infrared imaging system (Lincoln, NE) and associated software.
Total RNA was extracted using Trizol (Invitrogen, Carlsbad, CA) and the manufacturer’s protocol. RNA samples (1 μg) were reverse transcribed using random primers and the SuperScript III first-strand synthesis system (Invitrogen). The resulting cDNA was amplified using Taq DNA polymerase and previously published primer sets for chicken BMP2, 4, and 7 (Li et al., 2003). The PCR protocol included a 95°C denaturation step for 5 min, followed by 35 cycles of 95°C denaturation (20 sec), 47°C annealing (20 sec), and 72°C extension (30 sec), with a final 7 min extension step at 72°C.
DCDMLs grown on glass coverslips were fixed in 2% paraformaldehyde in PBS and processed for immunocytochemical detection of lens proteins as previously described (Le and Musil, 2001a). Antibodies used included mouse anti-chicken MP28 monoclonal MC15 (kind gift of Ross Johnson, University of Minnesota), rabbit anti-mouse CP49 polyclonal serum (#900), rabbit anti-filensin serum (#76PF), rabbit anti-N-cadherin (C 3678; Sigma), mouse anti-human vinculin Mab (V9131; Sigma), and rat anti-dog ZO-1 monoclonal antibody R40.76 (from Daniel Goodenough, Harvard Medical School). Immunofluorescence images were captured using a Leica DM LD photomicrography system and Scion Image 1.60 software.
Freshly isolated, intact mouse eyes were fixed in 10% neutral buffered formalin, dehydrated through a graded ethanol series followed by xylene, and then embedded in paraffin. Five micrometer-thick sections were cut, deparaffinized, and stained with hematoxylin and eosin.
BMP2 and BMP4 share 92% amino acid identity and common receptors (Massague, 1998). BMP7 is less closely related, and is functionally distinct from either BMP2 or 4 (e.g., Faber et al., 2002; Shou et al., 2000). We used our previously established assays to assess the effects of these BMPs on secondary fiber marker expression in primary embryonic chick lens DCDMLs. Although δ–crystallin is detectable in embryonic chick lens central epithelial cells, its levels are greatly increased during the early stages of epithelial-to-fiber differentiation (Piatigorsky et al., 1972). CP49, an essential component of lens-specific beaded filaments, is absent from undifferentiated central epithelial cells and is considered to be a late marker of fiber differentiation (reviewed in Quinlan et al., 1996). Purified preparations of recombinant BMP2, 4, and 7 (at concentrations of ≥ 5 ng/ml) increased the expression of δ–crystallin and CP49 in DCDMLs over a six day period to an extent similar to that obtained with three known fiber differentiation-promoting factors, namely FGF2, IGF-1, and vitreous body conditioned medium (VBCM) (Le and Musil, 2001a) (Fig. 1A, B). The latter is produced by incubating intact E10 chick vitreous bodies in a Transwell filter unit in serum-free medium overnight, and contains the factors in vitreous humor (including FGF) that cells at the lens equator would have access to in vivo (Le and Musil, 2001a). Similar to what was observed with FGF, IGF-1, and VBCM (Le and Musil, 2001a), upregulation of both fiber markers required a greater than 3 d exposure to BMP. Culturing DCDMLs with BMP affected neither their viability or ability to attain confluency.
A morphological hallmark of differentiation in DCDMLs is the formation of lentoids, phase-bright clusters of elongated cells with the ultrastructural characteristics of newly formed cortical fiber cells (Menko et al., 1987; Le and Musil, 1998; Okada et al., 1971; FitzGerald and Goodenough, 1986). These multilayered clusters, but not the surrounding monolayer of cobblestone-shaped epithelial cells, stain strongly with antibodies directed against fiber-specific proteins including CP49, its obligatory assembly partner in beaded filaments (filensin),and MP28, the chick orthologue of aquaporin 0. Because lentoids vary in diameter, height, and the size of individual cells, it is not possible to rigorously quantitate the expression of fiber markers in lentoids using conventional immunofluorescence microscopy. Qualitatively, however, both the number and average size of lentoids positive for CP49, filensin, and MP28 was markedly increased in BMP-treated cells relative to untreated controls (Fig. 1C). Although we do not know why some cells in DCDMLs form lentoids faster than others, one possibility is that cells that originated from the least differentiated regions of the lens (i.e., the central epithelium) are less likely to undergo lentoid formation within 6 days of culture than cells isolated from more peripheral regions of the epithelium. Western blotting confirmed increased levels of filensin protein in BMP-treated cultures (see Fig. 4C; we do not have reagents suitable for MP28 or δ–crystallin immunoblotting). Thus by both biochemical and morphological criteria, BMP 2,4, and 7 promote epithelial-to-fiber differentiation in DCDMLs, the first demonstration of such an activity in lens cells (see Discussion).
The canonical BMP signaling pathway is via Smad1/5/8 phosphorylation and does not include ERK (Massague, 1998). As assessed by Western blotting using phospho-specific antibodies, DCDMLs respond to purified recombinant BMP2, 4, or 7 (2 – 50 ng/ml) by increasing the phosphorylation of Smad1 in its activation domain (Fig 2A, top), but not by activation of ERK (see Fig. 6A). Phosphorylation of Smad1 by BMP2, 4, or 7 was blocked by a purified recombinant form of noggin, a naturally produced, secreted protein that binds with very high affinity and specificity to these BMP species and prevents them from engaging cell surface BMP receptors (Zimmerman et al., 1996; Groppe et al., 2002) (Fig 2A). VBCM also increased the levels of phosphorylated Smad1 in DCDMLs in a manner blocked by noggin, indicating that vitreous humor contains BMP2, 4, and/or 7 as its main or sole diffusible BMP species. Similar results were obtained with C3H10t1/2 cells, a mouse mesenchymal stem cell line in which BMP signaling is better characterized than in lens epithelium (Bowers et al., 2006; Hoffmann et al., 2002) (Fig. 2A, bottom).
As expected, δ–crystallin levels remained low if DCDMLs were co-cultured with either BMP2, 4, or 7 in the continuous presence of noggin. Noggin also very strongly reduced δ–crystallin synthesis induced by VBCM (Fig. 2B). Importantly, noggin had no significant effect on cell viability, or on the pattern or intensity of the other major [35S]methionine-labeled protein bands (compare first two lanes). Moreover, noggin did not diminish δ–crystallin expression stimulated by the nonphysiological activator fetal calf serum (FCS) (Fig. 2B), further indicating the specificity of the effect. The small but reproducible decrease in the relative amount of δ–crystallin synthesized by cells cultured with noggin alone compared to untreated controls is addressed in subsequent experiments (see Fig. 5).
We have reported that the ability of VBCM and vitreous humor to increase δ–crystallin synthesis in DCDMLs is abolished if FGF is removed using heparin beads, and that FGF eluted from such beads is capable of upregulating δ–crystallin expression to the same extent as unfractionated VBCM (Le and Musil, 2001a). Given this evidence that upregulation of δ–crystallin by VBCM is mediated by FGF, how could a BMP inhibitor such as noggin block this process? The simplest explanation would be if noggin prevented FGF from increasing δ–crystallin levels. This was confirmed by co-culturing DCDMLs with FGF and noggin for 4 – 14 d. At all times and concentrations of FGF2 tested (10 – 50 ng/ml), noggin blocked upregulation of δ–crystallin expression (Fig. 3A). Preincubation of FGF with noggin for 1 – 12 h did not interfere with the ability of the former to stimulate the activation of ERK or p38, confirming that noggin has no direct effect on FGF/FGFR interactions (data not shown). Importantly, noggin did not block upregulation of δ–crystallin by another previously described, BMP-independent inducer of fiber differentiation marker expression, the cAMP analogue 8-CPT (8-(4-chlorphenylthio)-cyclic AMP) (Ireland et al., 1997) (Fig. 3A). Similar results were obtained when BMP 2, 4, and 7 signaling was inhibited using function-blocking antibodies instead of with noggin (Fig. 3A).
The results shown in Fig. 3A strongly implied that upregulation of δ–crystallin by FGF required the activity of noggin-sensitive BMPs endogenously produced by the lens cells themselves. Message for BMP4 and BMP7 was detected in DCDMLs, as well as in whole E10 chick lenses, by RT-PCR (Fig. 3B). No evidence for expression of BMP2 was obtained, despite the ability of the primer set used to amplify the expected fragment from genomic DNA. Similar results were obtained if cells were cultured in the presence or absence of FGF, suggesting that FGF does not affect BMP expression. Embryonic and postnatal mouse lens is also reported to express BMP4 and 7, but not 2 (Furuta and Hogan, 1998; Wawersik et al., 1999; Hung et al., 2002; Thut et al., 2001), consistent with gene microarray data from a human lens epithelial cell line (Dawes et al., 2007).
Stimulation by FGF of expression of δ–crystallin in chick cells (Le and Musil, 2001a), like that of β-crystallin in rodents (Lovicu and McAvoy, 2001), does not require activation of ERK. In contrast, FGF upregulates the beaded filament proteins CP49 (Le and Musil, 2001a) and filensin (Lovicu and McAvoy, 2001) in an ERK-dependent manner, in keeping with numerous studies indicating that expression of crystallin genes and non-crystallin fiber-associated genes are regulated by distinct sets of transcription factors (see DePianto et al., 2003). Despite such differences in regulation, we found that expression of CP49 in response to VBCM or FGF, like that of δ–crystallin, required BMP activity. As shown in Fig. 4A, co-culturing cells with noggin inhibited the ability of VBCM (or of BMP), but not of fetal calf serum, to enhance CP49 expression. Noggin or a mixture of anti BMP-2/4/7 antibodies also reduced CP49 expression in response to FGF, but did not affect upregulation induced by the protein kinase A activator 8-CPT (Fig 4B). Moreover, noggin inhibited the ability of FGF, as well as of BMP, to enhance the levels of full-length filensin (Fig. 4C).
Additional evidence that lens cells synthesize noggin-sensitive BMPs that are important for fiber differentiation came from experiments conducted in the absence of added growth factors. As previously reported (Le and Musil, 1998), DCDMLs have a limited capacity to undergo fiber differentiation-associated changes in morphology (formation of lentoids) when cultured under our standard control conditions (i.e., for 6 days in unsupplemented M199/BOTS medium changed every two days). We noticed that including noggin in the medium appeared to reduce the levels and average size of fiber marker-containing lentoids, but did not affect the typical epithelial staining pattern in undifferentiated epithelial cells of proteins such as ZO-1 and vinculin (Fig. 5A). This decrease in morphological differentiation was accompanied by a small but reproducible reduction in the levels of δ–crystallin, CP49, and filensin protein detected in our standard assays (e.g, see Figs 2B; 4A,C). Extending the culture period from 6 to 14 d resulted in a modest enhancement in the synthesis of δ– crystallin in untreated DCDMLs, and a greater increase in CP49 levels. Enhanced expression of these fiber differentiation markers was blocked if noggin was included in the medium (Fig. 5B). In all cases, noggin-treated cells remained as viable and as capable of synthesizing total cellular proteins as controls. These findings further support the conclusion that lens cell-derived BMPs are not required for lens cell survival, but instead serve as bona fide differentiation factors (see Discussion).
Lens epithelial cells endogenously express low levels of FGF2 and FGF1 both in vivo and in culture (Schulz et al., 1993; Schweigerer et al., 1988; Lovicu et al., 1997). It has been suggested that these factors could influence secondary fiber formation at the lens equator (Lovicu and McAvoy, 1993). To determine whether paracrine/autocrine signaling through FGFs plays a role in BMP-induced upregulation of fiber marker expression, we treated DCDMLs with PD173074, a widely used small molecule inhibitor of FGFR 1–4 tyrosine kinase activity. As expected (Mohammadi et al., 1998), addition of 100 nM PD173074 to DCDMLs effectively blocked activation of ERK (as well as of p38; data not shown) in response to FGF, but not to phorbol esters (Fig. 6A). It did not, however, perturb BMP signaling, in that BMP2, 4 and 7 still activated Smad1 (Fig. 6B). A 6 d incubation of DCDMLs with PD173074 had no major effects on cell viability as indicated by: (1) the level of total protein synthesis (Fig 6C; amount of total protein metabolically labeled with [35S]methionine during a 4 h pulse is 1.03 fold +/− 0.29 of that synthesized in untreated controls; n=4), the ability of the cultures to attain confluency, or by the amount of total protein in PD173074-treated cultures (97.6% +/− 11% of that recovered from controls; n=5). The inhibitor did however, abolish upregulation of δ–crystallin and CP49 levels by exogenously added FGF. In contrast, PD173074 had little effect on expression of δ–crystallin or CP49 induced by BMP (Fig. 6C). We conclude that although lens-derived BMP is required for upregulation of fiber marker expression in DCDMLs by exogenous FGF, upregulation by exogenous BMP is independent of lens-derived FGF/FGFR signaling. Thus the FGF that is required for normal fiber differentiation in vivo is likely to originate from extralenticular sources, in particular the vitreous.
The major steps in lens development are highly conserved between mammals and birds, as are the effects of FGF and vitreous humor on this process (Piatigorsky, 1981; Le and Musil, 2001a; Robinson, 2006). If the results obtained with DCDMLs are relevant to the mammalian lens in vivo, then it would be expected that overexpression of noggin in the lenses of transgenic mice would inhibit the differentiation of epithelial cells to secondary fibers. This was investigated using a strain of transgenic mice (OVE1196) in which noggin was exogenously expressed under the control of a lens-specific αA-crystallin promoter (CPV2; Zhao et al., 2002). Importantly, this promoter becomes active (on E12.5) after formation of the lens placode and primary fiber cells, the stages in lens development already established to require noggin-sensitive BMP signaling. Zhao et al. (2002) have reported that the gross morphology of the eyes of CPV2-Noggin mice is indistinguishable from that of wild type littermates at E12 -18, indicating that (unlike in mouse models in which BMP signaling is disrupted during embryogenesis; Beebe et al., 2004; Faber et al., 2002) the epithelial compartment had developed normally. Consistent with the late onset of abnormalities in the OVE1196 mice, there were also no major differences in the total proteins synthesized by wild type and transgenic lenses at (or before; not shown) postnatal day 7, with the fiber-specific α, β, and γ crystallins being the predominant species metabolically labeled during a 4 h incubation of freshly isolated lenses with [35S]methionine (Fig. 7A; crystallins identified by their size, immunoreactivity on Western blots, and because they constitute > 90% of the total lens proteins detectable by Coomassie blue staining). By P14, the transgenic lenses were visibly smaller than the controls (see Fig. 8B) and contained 32 % +/− 5% less total protein per lens (n=33). This was accompanied by a striking decrease in de novo synthesis of all three crystallin families in 16/17 transgenic pups in 5 litters examined (Fig. 7A). Immunoprecipitation confirmed the near absence of metabolically labeled γ–crystallins, a late marker of fiber differentiation (Fig. 7B, top). Importantly, [35S]met-labeled bands of proteins larger than crystallins (Mr ≥ 40 kDa) were present at wild type levels, indicative of the continued viability of the transgenic lenses (Fig. 7A, arrowheads). Western blot analysis of fractionated lenses showed that γ-crystallins (presumably mostly synthesized prior to P14) remained soluble in the mutants at this time (Fig. 7C), consistent with the absence of obvious lens opacities. Within the next two weeks, however, the lenses of the CPV2-Noggin homozygous and heterozygous mice developed large cataracts and did not increase in size, and the animals became microphthalmic. As expected (Piatigorsky, 1980), γ–crystallin began to abnormally accumulate in the insoluble fraction (Fig. 7C). Similar results were obtained when the OVE 1196 line was crossed for two generations with wild type C57BL/6 mice, suggesting that the phenotype was independent of the mutations in the CP49 and phosphodiesterase 6B genes that perturb expression of their protein products in the purebred FVB/N strain (Simirskii et al., 2006) (data not shown). Although it is not known why the protein synthesis abnormalities in the lenses of CPV2-Noggin mice are not apparent until the second postnatal week, one possibility is that (secreted) noggin must accumulate to relatively high levels before it can neutralize all of the noggin-sensitive BMPs produced by extralenticular ocular tissues (e.g., ciliary and iris epithelium; retina).
Because mature fibers are incapable of protein synthesis, the [35S]methionine-labeled bands shown in Fig. 7A are the products of epithelial and newly differentiating cortical fiber cells. The loss of the capacity of the transgenic lenses to synthesize fiber-associated proteins during the second postnatal week, combined with their growth defects, suggested that noggin overexpression inhibited the formation of new secondary fiber cells. This was confirmed by examination of paraffin-embedded, hematoxylin and eosin-stained sections of wild type and mutant littermates (Fig. 8). Fig 8A, left illustrates the histological hallmarks of normal epithelial-to-secondary-fiber differentiation, with the nuclei of the newly elongating cortical fiber cells forming a characteristic “bow” (indicated by bracket). This pattern is also present in sections from CPV2-Noggin mice at P7. By P11-P14, however, the histology of the outermost (i.e, youngest) equatorial cells in the noggin-overexpressing mice is very different from that of controls (Fig. 8A). Instead of forming a bow, the monolayer of cuboidal epithelial cells could be shown to extend to the posterior pole of the lens. These cells were morphologically similar to epithelial cells at the anterior of the organ, with little or no evidence of apoptosis, multilayering, or conversion to a mesenchymal phenotype (Fig. 8C). Similar changes were observed in P14 sections prepared from a second line of CPV2-Noggin FVB/N mice (OVE1195; slides kindly provided by Dr Shulei Zhao, Lexicon Genetics, Inc.). As previously reported (Zhao et al., 2002), morphological abnormalities were also detected outside of the lens, including perturbation of the ciliary body and processes and loss of the vitreous body (see Discussion).
It has been appreciated since the 1960s that secondary fiber differentiation is critically dependent on extralenticular factors concentrated within the posterior half of the eye (Coulombre and Coulombre, 1963; Yamamoto, 1976). Although there is an extensive literature that FGF from vitreous humor plays a essential role in this process, this is the first report that provides direct evidence that BMPs are also required. We base this conclusion on three novel and unexpected findings, discussed individually below.
At least two previous studies have examined the effect of purified BMPs on chick (Belecky-Adams et al., 2002) and rat (Vinader et al., 2003) lens epithelial cells in culture. Neither found any evidence that purified BMP 2, 4, and/or 7 enhanced morphological differentiation or expression of fiber marker proteins, although Belecky-Adams et al. (2002) reported that adding noggin to vitreous humor partially reduced its ability to stimulate cell elongation over a 5 h period. Both groups used as their model system central epithelial explants. We also did not detect either morphological or biochemical differentiation of chick E6 central epithelial explants treated with BMPs despite robust fiber formation in response to FGF or VBCM (data not shown). A critical difference between such explants and DCDMLs is that the former preparation contains only the least differentiated cells in the lens, whereas DCDMLs include epithelial cells from the more peripheral regions including the equator. Central epithelial cells are less responsive to the differentiation-inducing effects of FGF and insulin/IGF than peripheral populations (Richardson et al., 1993; Richardson et al., 1992; Lovicu and McAvoy, 1992), possibly because the latter have been influenced by factors released from the ciliary body and/or iris. We speculate, although have not experimentally verified, that in central epithelial cells, BMP signaling is not coupled to fiber differentiation, allowing exogenously added FGF to enhance fiber formation without a requirement for either endogenous or extralenticular BMP. Cells in the central epithelium of intact chick and rodent lenses have been reported to express BMP receptors and activated (i.e., anti-phosphoSer463,465-reactive) Smad1 (Zhao et al., 2002; de Iongh et al., 2004; Rajagopal et al., 2007), suggesting that they possess the components of the canonical BMP signaling cascade but lack the capacity to transduce this signal into enhanced fiber differentiation. Interestingly, Rajagopal et al. (2007) have reported that immunoreactivity for activated Smad1 and its binding partner Smad4 is concentrated in undifferentiated epithelial cells of mouse and chick lenses in endosomes, whereas it is localized to the nucleus in elongating fiber cells. These distinct distributions could be linked to differences in the functional outcome of BMP signaling in the two cell types, with transcriptional events (in which Smad 1 and 4 translocate into the nucleus) predominating in the population of cells transitioning into fibers.
Hung et al. (2002) overexpressed BMP7 in the lenses of transgenic mice. Consistent with the aforementioned culture studies, these animal showed no signs of premature fiber differentiation in central epithelial cells. Cells in more peripheral regions can acquire BMPs and other fiber-promoting factors from the ciliary body and/or vitreous humor, which could explain why they did not respond to an additional supply of transgenic BMP7 with accelerated differentiation. The loss of fiber differentiation observed in BMP7-overexpressing lens at P1 is a downstream consequence of ablation of the neural retina at E15.5, a phenotype not observed in the noggin-overexpressing mice.
We have shown that inhibiting BMP 2/4/7 activity with noggin or anti-BMP antibodies blocks upregulation of the fiber differentiation markers δ–crystallin, CP49, and filensin by FGF in DCDMLs. In principle, one mechanism by which lens-derived BMP could be required for exogenous FGF to increase fiber differentiation would be if endogenous BMP acted as a survival factor. This possibility is, however, ruled out by the finding that the viability of cells cultured with noggin in either the absence or presence of other factors (FGF, BMP, VBCM) is indistinguishable from that of noggin-minus cells as assessed by their ability to exclude Trypan Blue, reach confluency, and carry out protein synthesis. Alternatively, BMP could be required to maintain DCDML cells in an epithelial state such that the aforementioned BMP blockers convert them into mesenchymal cells incapable of developing into fibers. We can also discount this mechanism given that noggin-treated cells retain their epithelial morphology and lack the hallmarks of EMT in lens cells including redistribution of vinculin and ZO-1 from cell-cell interfaces (Fig. 5A) or expression of α-smooth muscle actin-positive stress fibers (Stump et al., 2006) (data not shown). Because lens epithelial cells must exit the cell cycle to differentiate (Menko, 2002), BMP blockers could conceivably prevent fiber formation by forcing them into a hyper-proliferative state. Relative to untreated controls, noggin did not, however, enhance entry into mitosis as assessed by anti-phosphohistone H3 immunocytochemistry, nor did it increase the number of cells present after 6 days of culture (data not shown). The remaining possibility is that endogenous BMP acts to promote epithelial-to-fiber differentiation, a conclusion in keeping with our demonstration that exogenously added BMP behaves as a bona fide differentiation factor. Noggin has previously been reported to block at least three FGF-induced processes, namely potentiation of apoptosis in the developing limb bud (Montero et al., 2001), expression of albumin in ventral foregut endoderm (Rossi et al., 2001), and retina regeneration in the embryonic chick (Haynes et al., 2007). In each case, enhancement of the same process by exogenously applied BMPs (or by expression of activated BMPRs) required FGF signaling. This is unlike what we have observed (Fig. 6), indicating that lens fiber differentiation is regulated by a fundamentally different, nonreciprocal type of cross-talk between FGF and BMP signaling.
What could be the mechanism underlying FGF/BMP cross-talk in secondary fiber formation? The simplest scenario would be if FGF stimulates the synthesis of BMPs to levels high enough to support secondary fiber formation in the absence of ongoing FGF signaling. We have, however, failed to find evidence for increased expression of BMPs in FGF-treated DCDMLs by either RT-PCR (Fig. 3B), or by assaying the conditioned medium of FGF-treated cells for BMP bioactivity (data not shown). An alternative possibility would be that FGF sensitizes DCDMLs to low levels of endogenous BMP by increasing the synthesis of either BMP receptors or Smad1. This also does not appear to be the case, given that: (1) adding exogenous BMP increases the amount of immunologically detectable activated Smad1 in DCDMLs within 40 min (Fig. 2A), indicating that neither BMP-responsive receptors or Smads are limiting under basal conditions, and (2) a 1– 6 d exposure of DCDMLs to FGF does not enhance the amount of Smad1 activated in response to either high or low levels of exogenously added BMP (data not shown). Instead, BMP signaling may be required for lens cells to be maximally responsive to the differentiation-inducing effects of FGF. This possibility will be addressed in future studies.
FGF has been considered to be a survival factor for lens epithelial cells (Robinson, 2006). Evidence for this contention in cultured cells includes the report by Renaud et al. (1994) that antisense oligonucleotides against FGF1 reduced the viability of serum-deprived bovine lens-derived cells by 50%. More recently, Tholozan et al. (2007) demonstrated that apoptosis of bovine lens epithelial cells replated on lens capsules in low serum was greatly increased by an inhibitor of FGF release from the lens capsule (the MMP2 blocker OA-Hy), but could be rescued from death by addition of exogenous FGF. In contrast, we have found that DCDMLs remain viable when FGF-mediated signaling was directly blocked with the specific FGFR inhibitor PD173074 ( Fig. 6); similar results were obtained with another FGFR tyrosine kinase antagonist, SU5402 (data not shown). Possible explanations for this apparent discrepancy include pretreatment of lens cells with high serum in the Renaud et al. (1994) and Tholozan et al. (2007) studies (DCDMLs, like lens cells in vivo, are never exposed to serum), cell source (adult cow, compared to embryonic chick), and/or experimental design. Regardless of the reason, our findings demonstrate that FGF-mediated signaling is not an obligatory prerequisite for lens epithelial cell survival in culture. This conclusion is in keeping with previous studies showing that FGF does not prevent apoptotic cell death in embryonic chick lens central epithelial explants (Huang et al., 2003), and is not the factor in lens conditioned medium that rescues P11 rat lens epithelial cells cultured at low density from apoptosis (Ishizaki et al., 1993). We therefore interpret the continued viability of DCDMLs in the presence of noggin not as an example of an FGF-mediated process that is BMP-independent (e.g., noggin-insensitive), but instead as indicating that survival of DCDMLs requires neither lens-derived FGF or BMP. An increase in apoptosis has been reported in the lenses of transgenic mice expressing mutant forms of FGFR expected to act as dominant negative inhibitors of FGF signaling. In at least one such model (Stolen and Griep, 2000), death was enhanced in anterior epithelial cells that did not detectably express the transgene, in keeping with an indirect role for FGF in lens epithelial cell survival in vivo.
In vivo, epithelial cells at the lens equator differentiate in response to factors that diffuse out of the vitreous body. We found that the BMP signaling inhibitor noggin blocked fiber marker expression not only in DCDMLs after addition of vitreous body conditioned medium, but also in transgenic mice after its exogenous overexpression. Analysis of CPV2- Noggin lenses after P7 indicated that de novo synthesis of all major α, β, and γ crystallin species was severely decreased, demonstrating that the requirement for BMP signaling is widespread among fiber proteins and is not confined to avian lens cells. A caveat of our mouse studies is that noggin is a secreted protein. We therefore do not know whether the transgenic noggin blocks fiber differentiation by inhibiting the function of BMPs originating from the vitreous humor, lens cells, or both. Another confounding issue is that noggin overexpression has effects outside of the lens. This includes the postnatal loss of the vitreous body, which may be related to defective formation and/or function of the ciliary body and processes (Zhao et al., 2002). Given that vitreous humor is the major reservoir of FGF for the lens equator, it could be argued that the absence of epithelial-to-fiber differentiation in the noggin over-expressing mice is a downstream consequence of the lack of the vitreous body. These two phenotypes are not, however, obligatorily linked. For example, germline deletion of nectin 1 causes postnatal loss of the vitreous body without abolishing the lens bow or blocking lens growth (Inagaki et al., 2005). Conversely, mice in which expression of the BMPR ALK3 was deleted in the lens using the Le-Cre transgene retain the vitreous body, but develop a layer of epithelial cells at the posterior surface of the lens between P7 and P14 reminiscent of that observed in CPV2-Noggin mice (Beebe et al., 2004). Interestingly, a postnatal inhibition of lens differentiation resulting in posterior extension of the epithelial monolayer was also observed in mice expressing a secreted, dominant-negative form of FGFR3 (Govindarajan and Overbeek, 2001). In these animals, the levels of active ERK in the lens were reduced to almost undetectable levels, indicating that (unlike in mice expressing a transmembrane dominant negative mutant of FGFR1; Chow et al., 1995) FGF signaling throughout the organ was essentially abolished. Thus epithelial-to-fiber differentiation is inhibited when either FGF or BMP signaling in the lens is severely disrupted. Although incapable of proving a direct cause-and-effect relationship between lenticular BMP signaling and secondary fiber formation, these in vivo results support the major conclusion of our culture studies, namely that BMP signaling plays an essential role in FGF-dependent secondary fiber formation. Faber et al. (2001) have reported that the frequency and severity of defects in lens induction in mice expressing a dominant negative mutant of FGFR1 in the presumptive lens ectoderm cells is increased in animals in which one allele of BMP7 had been deleted. It is therefore possible that FGF/BMP signaling cross-talk also plays a required role in a very early stage of lens development.
This work was supported by Grant R01 EY014622 from the National Eye Institute to L. Musil. We thank Drs P. FitzGerald, R. Johnson, D. Goodenough, and L. David for providing antibodies, and Dr. D. Goldman for advice and assistance in the animal studies. Carolyn Gendron provided invaluable assistance with the lens histology.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.