Using whole embryo culture and bead implantation, our analyses have provided new evidence for the critical importance of the level of BMP4 signaling for D-V patterning and growth of the developing mouse eye. Our data suggest that BMP4 patterns the optic cup by dosage dependent regulation of target genes, Tbx2, Tbx3 and Tbx5.
A conserved genetic pathway: BMP4 regulation of Tbx2 subfamily genes
In whole embryo culture experiments, exogenous BMP4 induced Tbx2, Tbx3 and Tbx5 expression, while Noggin repressed/reduced Tbx2 and Tbx5 expression in the developing retina, providing strong support for the idea that BMP4 signaling regulates T-box gene expression during mouse eye development. This is the first demonstration that all three members of the Tbx2 subfamily, Tbx2, Tbx3 and Tbx5 genes are downstream targets of BMP signaling in the mouse eye. A recent paper reported loss of Tbx5 expression in the optic vesicles of Bmp4-/- mice . Previous reports have demonstrated BMP2/4/Dpp regulation of Tbx2 subfamily genes in the heart, limb and retina in other vertebrates and of omb in Drosophila [41, 49, 52–54, 68]. Considered together, these data indicate that BMP2/4 regulation of Tbx2 subfamily genes is an ancient and evolutionarily conserved pathway, which has been recruited to regulate development and patterning of different tissues in vertebrates and invertebrates.
In the developing eye, Tbx2, Tbx3 and Tbx5show differential responses to BMP4 signaling
Our data suggest that in the eye, each T-box gene responds to BMP4 in a different manner. They are consistent with a model in which Tbx5 responds to BMP4 close to the dorsal signaling source, while Tbx2 and Tbx3 can be induced in a wider area by lower levels of BMP4 secreted from the source. Based on the expression pattern of Bmp4 in the dorsal region of the optic vesicle and the optic cup, it is likely that BMP4 protein concentrations will be high in the dorsal and low in the ventral region of the eye; indeed the dorsal retina shows high levels of BMP signaling activity as determined by immunohistochemistry for Phospho-Smad 1/5/8 proteins.
The different spatial response that each T-box gene showed and their close evolutionary relationship support a mechanism of direct regulation by BMP4 signaling, although we have not excluded the possibility that other intermediate factors regulate these genes. Significantly, a recent study showed that the Tbx3 promoter is directly regulated by BMP Smads, in chromatin immunoprecipitation experiments using extracts from embryonic hindlimbs, as well as cotransfection assays . In Drosophila, molecular dissection of cis-regulatory sequences of omb, the homologue of vertebrate Tbx2/3, revealed a Dpp response element  suggesting that omb may also be a direct target of Dpp signaling. Interestingly, Tbx3 responded more vigorously than Tbx2 to exogenous BMP4 administered to the lens, extending all the way to the ventral retina, whereas Tbx2 did not extend as far and may need other factors or is being inhibited by factors present in the ventral retina. The fact that Tbx3 expression is initially more restricted than the Tbx2 expression domain, yet after exogenous BMP4 treatment it shows the greatest expansion suggests that other factors are regulating the expression of T-box genes along with BMP4.
It has been proposed that Tbx5 represses cVax and vice versa in the chick based on overexpression studies [21, 41]. The cascade in the mouse appears to differ, as with increased BMP4 levels, we observed repression of Vax2 without ventral expression of Tbx5. Also, in Vax2 null mice, the Tbx5 expression domain is unchanged . Moreover the separation of Vax2 and Tbx5 expression in two non-overlapping domains (domains 1 and 4) indicate that these factors are unlikely to control each other's expression.
Tbx2 subfamily genes as regulators of apoptosis and proliferation
Downstream targets regulated by the Tbx2, Tbx3 and Tbx5 transcription factors have not been identified in the eye. Several in vitro studies implicate these genes in the regulation of cell proliferation and apoptosis. Ectopic expression of TBX5 in osteosarcoma cells inhibits cell proliferation and induces apoptosis [72, 73]. TBX2 gene amplification has been detected in primary human breast cancer tumours, pancreatic cancer cell lines, and its overexpression detected in melanoma cell lines [74–78]. Both TBX2 and TBX3 repress the expression of p14ARF, and there is evidence for the repression of other cyclin dependent kinase inhibitors p21, p16INK4a, and p15INK4b by TBX2 in vitro [74, 79–81]. Moreover, both TBX2 and TBX3 show anti-senescence properties [74, 82], while a dominant negative form of Tbx2 can induce senescence in a melanoma cell line .
During embryogenesis, a similar pro-proliferative role has not been identified for these genes. In fact, it has been shown that the expression of p21, p19ARF, p16INK4a, and p15INK4b is normal in mice with a targeted mutation of Tbx2, and there is no evidence for a genetic interaction between Tbx2 and p53 (Trp53) . Furthermore, Tbx2 expression has been associated with a low rate of proliferation in the atrioventricular canal during heart development . Tbx5 has also been associated with the regulation of cell proliferation during embryogenesis. Misexpression of TBX5 in the chick heart leads to inhibition of cell proliferation  and in the limb, ectopic expression of Tbx5 (or Tbx4) leads to a truncated limb phenotype due to cell proliferation arrest .
We observed a dosage sensitive increase in the amount of apoptosis in BMP4-treated eyes, which correlates with Tbx5 expression in the dorsal retina, but not with the ectopic Tbx2 or Tbx3 expression in the ventral retina. Alternatively, the homeobox gene Msx2, which is normally expressed in the dorsal neural retina [40, 65], could be mediating the apoptotic effect of the elevated BMP4 levels in the dorsal neural retina. Msx genes mediate BMP induced apoptosis during development of other organs such as the hindbrain and the limbs [57, 62]. The different response of Msx2 and the T-box genes to BMP4 treatment suggests that these genes do not regulate each other. If Tbx2 acts to reduce proliferation during optic cup formation, as recently found in the developing heart , then Tbx2 expansion into the ventral retina may contribute to the observed reduction in levels of proliferation after BMP4 treatment.
BMP4 dosage in the regulation of eye size and shape
Addition of exogenous BMP4 in the mouse optic cup in embryo cultures resulted in a reduction in retinal volume and alterations in eye shape. In chick and Xenopus, malformation of the ventral retina has previously been described after overexpression of BMP4 by electroporation [41, 52]. Increases in gene dosage, as well as loss of function are known to cause congenital eye malformations in humans and mice, for example in the case of the transcription factor genes Pax6 and Foxc1 [85, 86]. Our findings suggest that increased levels of BMP4 signaling may be detrimental; in the human population, such changes could occur, for instance, via genetic variation in regulatory sequences, gene duplication, or gain of function mutations. The recent report that anophthalmia occurs in embryos from crosses between Bmp4 heterozygous mutants and mice homozygous for a disrupted allele of Twisted gastrulation (Tsg), a BMP binding protein that is thought to enhance BMP4 signaling in the eye , also indicates that excess BMP4 signaling may be as harmful as loss of function.
The adverse consequence of reduced levels of BMP4 signaling for eye development has been explored in different systems. That optic vesicle development is arrested in Bmp4 null mice  first indicated the importance of Bmp4 for normal growth. We observed an expansion of optic stalk and a reduction in the neural retina after addition of the BMP antagonist, Noggin, close to the dorsal BMP4 signaling source in the mouse optic cup. In the chick embryo, misexpression of Noggin induces primarily ventral abnormalities that include optic stalk hyperplasia . Conditional gene inactivation has also been used to analyse the effect of loss of the serine/threonine kinase membrane bound BMP receptors; BmprIa, BmprIb, and BmprII are expressed in the developing mouse eye [26, 88, 89]. Whereas targeted deletion of Bmpr1b alone disrupts the ability of many ventrally located ganglion cells to enter the optic nerve head , conditional Bmpr1a-/fx/Bmpr1b-/-;cre double mutant embryos exhibit more profound structural abnormalities due to excess apoptosis and reduced proliferation at E11.5, and show anophthalmia at birth . In these mutants, the loss of Bmpr1a and Bmpr1b likely interfered with BMP4 signaling, as well as signaling by BMP7, BMP2 and BMP3, from non-neural retinal sources .
The findings of altered eye size and shape with loss and gain of BMP4 function, suggest that a BMP4 signaling gradient mediates control of proliferation across the D-V axis during optic cup morphogenesis. With exogenous BMP4, we observed a significantly decreased level of retinal cell proliferation in both the dorsal and the ventral retina. Regional differences in proliferation have recently been found during optic cup growth [4, 5], and we found that the exogenous BMP4 acted to equalise the differences in levels of proliferation across the D-V axis. Other studies have linked excess BMP4 to reduced proliferation and increased apoptosis. For example, the BMP expressing region of the mouse telencephalon has a lower proliferative activity and higher level of apoptosis than the rest of the forebrain, and BMP4-soaked beads can induce apoptosis in telencephalic explants . In the chick, exogenous BMP4 applied to the dorsal optic cup, or at early stages to the telencephalon also increases apoptosis [40, 91]. However, in the same studies, addition of BMP4 to chick retinal cultures at later stages induced proliferation , while reduced levels of BMP signaling (via Noggin misexpression) in the telencephalon was associated only with a decrease in proliferating cells , highlighting that the cellular response to BMP4 signaling appears stage and context dependent. In our study, only at higher doses of BMP4 was widespread apoptosis induced.
Threshold-specific requirements for BMP4 in optic cup development
Our data show that a major role for dorsal BMP4 signaling in the developing mouse optic cup is to regulate gene expression across the D-V axis. It suggests that BMP4 may be critically important for providing a signaling gradient, which is used to demarcate gene expression domains across the D-V axis, with retinal cells expressing different transcription factor genes in response to different threshold levels of BMP4 activity. Variations in Smad binding site distribution between target genes or levels of other active cofactors are likely to contribute to the differential gene responses.
Exogenous BMP4 treatment prior to optic cup formation affected Tbx2 and Tbx5, but not Vax2 or Tbx3, and T-box gene expression changes were not induced in the ventral retina in contrast to the effect of exogenous BMP4 addition to the lens. These differences may be due to the different developmental stage, and/or the different range of BMP signaling from bead implantation in the lens compared with the extraocular mesenchyme. In the latter case the different intervening tissue between the bead and the retina that is exposed to the BMP signal may be significant; when a bead is implanted in the lens the intervening tissue is the lens, whereas from a bead located in the mesenchyme, the BMP signal has to pass through the presumptive RPE. We were only able to detect increased levels of Phospho-Smad 1/5/8 immuno-labeling after BMP4 bead implantation in the lens, which correlates with our findings that lens bead implantation showed the most dramatic effect on gene expression domains. In this study it was not possible to distinguish if the BMP gradient regulates the expression of the Tbx2 subfamily genes directly or indirectly. Rather than the BMP gradient being directly responsible for the nested expression domains it is possible that there are other secondary molecules induced by BMP that carry the signal. However, our data considered together with recent data showing that the Tbx3 promoter is directly regulated by Smads  strongly support the idea that these genes respond directly to a BMP4 signaling gradient.
Although there are some differences in the specific gene expression patterns between chick and mouse, the broadly conserved pattern of gene expression domains across the D-V axis of the optic cup suggests that compartment-like units of clonally related cells defined by restricted gene expression domains, as previously described in the chick , may also exist in the mouse. If this is the case, then our data suggests that altering the BMP4 concentration in the eye re-specifies the compartments of the retina along the D-V axis; increased BMP causes compartments to express markers normally expressed by their more dorsal neighbours, although the response to BMP4 is not uniform in the different genes examined. Our analysis of proliferation and apoptosis support the conclusion that these changes in gene expression domains reflect changes in the identity of retinal territories rather than tissue loss. Instead of being made up of domains 1, 2, 3 and 4, the post BMP-treatment eye consists only of domains 1, 2 and a modified domain 3 (domain 3* that lacks Tbx2 and Vax2 expression but expresses Tbx3). The importance of correct spatial restriction of gene expression for topographic mapping of retinal ganglion cells to the brain is well established [21, 41, 92]; our data suggests it is also important in regulating regional growth of the developing optic cup.
Our finding that exogenous BMP4 not only extends expression of the dorsally expressed T-box genes in the optic cup, but also reduces the ventral marker Vax2 support the idea that BMP4 signaling has a long-range action across the D-V axis, and that BMP4 may repress Vax2 expression. BMP inhibitors, in particular Ventroptin and Dan, are known to be expressed in a ventral-high gradient in the chick eye [93, 94], although these expression patterns have not been reported in the mouse. The role of such inhibitors may normally be to block long range BMP4 signaling from acting in the ventral-most region (domain 4), allowing Vax2 expression, and preventing induction of dorsally expressed genes. In our study, increased levels of BMP4 at optic cup stage are apparently sufficient to overcome endogenous inhibitors, leading to Vax2 repression and induction of T-box gene expression in the ventral retina.
Also consistent with the proposal that BMP regulates Vax2, the recent informative study by Murali et al , showed that with reduced levels of BMP signaling, Tbx5 expression was lost whereas Vax2 expression was expanded into the dorsal mouse retina in Bmpr1a-/fx/Bmpr1b+/-;Cre double mutants. In chick, retroviral overexpression of Bmp4 was previously shown to repress expression of Ventroptin . Alternative to direct regulation of Vax2 by BMP4, interaction with other ventral signaling pathways could be important, as in zebrafish and chick, hedgehog signaling from the ventral midline contributes to regulation of the Vax gene [95, 96]. Mouse embryos cannot be cultured ex utero for long enough to analyse the effect of the shifted gene expression domains on late stages of eye development. Nevertheless, the effect of loss of Vax2 expression in the ventral eye has already been carefully studied in Vax2 null mice and shown to cause profound defects in ventral eye formation and coloboma [20, 22, 97].