Overexpression of Pax6 results in microphthalmia, retinal dysplasia and defective retinal ganglion cell axon guidance

Background The transcription factor Pax6 is expressed by many cell types in the developing eye. Eyes do not form in homozygous loss-of-function mouse mutants (Pax6Sey/Sey) and are abnormally small in Pax6Sey/+ mutants. Eyes are also abnormally small in PAX77 mice expressing multiple copies of human PAX6 in addition to endogenous Pax6; protein sequences are identical in the two species. The developmental events that lead to microphthalmia in PAX77 mice are not well-characterised, so it is not clear whether over- and under-expression of Pax6/PAX6 cause microphthalmia through similar mechanisms. Here, we examined the consequences of over-expression for the eye and its axonal connections. Results Eyes form in PAX77+/+ embryos but subsequently degenerate. At E12.5, we found no abnormalities in ocular morphology, retinal cell cycle parameters and the incidence of retinal cell death. From E14.5 on, we observed malformations of the optic disc. From E16.5 into postnatal life there is progressively more severe retinal dysplasia and microphthalmia. Analyses of patterns of gene expression indicated that PAX77+/+ retinae produce a normal range of cell types, including retinal ganglion cells (RGCs). At E14.5 and E16.5, quantitative RT-PCR with probes for a range of molecules associated with retinal development showed only one significant change: a slight reduction in levels of mRNA encoding the secreted morphogen Shh at E16.5. At E16.5, tract-tracing with carbocyanine dyes in PAX77+/+ embryos revealed errors in intraretinal navigation by RGC axons, a decrease in the number of RGC axons reaching the thalamus and an increase in the proportion of ipsilateral projections among those RGC axons that do reach the thalamus. A survey of embryos with different Pax6/PAX6 gene dosage (Pax6Sey/+, Pax6+/+, PAX77+ and PAX77+/+) showed that (1) the total number of RGC axons projected by the retina and (2) the proportions that are sorted into the ipsilateral and contralateral optic tracts at the optic chiasm vary differently with gene dosage. Increasing dosage increases the proportion projecting ipsilaterally regardless of the size of the total projection. Conclusion Pax6 overexpression does not obviously impair the initial formation of the eye and its major cell-types but prevents normal development of the retina from about E14.5, leading eventually to severe retinal degeneration in postnatal life. This sequence is different to that underlying microphthalmia in Pax6+/- heterozygotes, which is due primarily to defects in the initial stages of lens formation. Before the onset of severe retinal dysplasia, Pax6 overexpression causes defects of retinal axons, preventing their normal growth and navigation through the optic chiasm.


Background
The transcription factor Pax6 is expressed dynamically during the early development of the mouse eye in both the surface ectoderm and the optic vesicle, which integrate to generate the structures of the eye [1 -3]. As morphogenesis proceeds, Pax6 is expressed in the lens, retinal pigment epithelium and retina [1][2][3][4]. Most studies examining the functions of Pax6 in eye development have exploited loss-of-function models. The optic cup fails to form in embryos completely lacking Pax6 [5]. Examination of Pax6 +/+ ↔ Pax6 -/mouse chimeras has shown that Pax6 is required cell autonomously in the optic vesicle for maintenance of contact with the overlying surface ectoderm, a necessary event in eye formation, providing a clue that Pax6 may be involved in defining the adhesive properties of these cells [6,7]. Heterozygous Pax6 +/embryos show microphthalmia that is thought to be caused by delayed development of the lens and anterior eye structures [8]. Data from Pax6 +/-↔ Pax6 +/chimeras has demonstrated that heterozygous cells have early cell autonomous deficiencies at lens placode formation and after closure of the lens pit [9]. Studies using Cre-lox technology to selectively disrupt Pax6 in discrete parts of the developing eye have shown that removing Pax6 from the developing surface ectoderm produces an eye lacking a lens but possessing a retina with retinal ganglion cells (RGCs) able to project axons [10]. Removing Pax6 function after the retina forms results in a retina comprising mainly amacrine cells at the expense of other retinal cell types including RGCs [3].
Gain-of-function models are also useful for assessing gene function, particularly where, as in the case of Pax6, absence of function leads to complete abolition of a structure or cell type. In some parts of the developing brain Pax6 has functions in axon guidance as well as in tissue morphogenesis and regulates genes implicated in growth cone navigation [11]. Since absence of Pax6 results in absence of RGCs, it is unclear whether Pax6 has a role in the navigation of retinal axons. Its expression by projecting RGCs [4] indicates that it is poised to fulfil this function. Here, we investigated the formation of the eye, retina and retinal projections in mice that over-express Pax6 and we complemented our analysis of retinal projections in the gain-of-function model by studying retinal projections in Pax6 +/embryos.
Controlled overexpression of Pax6 in its normal domains of expression has been achieved by the generation of a YAC transgenic mouse expressing several copies of the human PAX6 gene, which generates protein identical to mouse Pax6, under the control of all its regulatory elements [12]. PAX77 hemizygous mice carry 5 to 7 copies of the human PAX6 gene, all integrated at the same locus, and are designated PAX77 + [12]. PAX77 homozygous transgenic mice are designated PAX77 +/+ . Schedl et al. [12] showed that over-expression of PAX6 causes microphthalmia in adults but did not study the development of this defect. Here we show that the eye initially forms apparently normally but that retinal defects and microphthalmia emerge and become increasingly severe during subsequent embryogenesis and postnatal life. We also show that Pax6 overexpression causes defects in RGC axon growth and guidance through the optic chiasm, where RGC axons are sorted into those that project ipsilaterally and those that project contralaterally. Our results indicate that although both over-and under-expression of Pax6 cause microphthalmia, the underlying mechanisms are different.
Pax6 is overexpressed in the eye of PAX77 +/+ embryos Figure 1 Pax6 is overexpressed in the eye of PAX77 +/+ embryos. Horizontal sections through the eye of wild-type and PAX77 +/+ embryos at (A) E12.5 and (B) E14. 5, showing the expression of Pax6/PAX6. (C) Levels of endogenous Pax6 mRNA and total Pax6/PAX6 mRNA in the retina of E14.5 PAX77 +/+ embryos, relative to wild-types, determined by real time quantitative RT-PCR and normalised against GAPDH mRNA levels (n = 3 in each case).Pax6 mRNA levels are significantly decreased in the retinae of PAX77 +/+ embryos compared to those of wild-types (Student's t-test, p = 0.014) while Pax6/PAX6 mRNA levels are significantly increased (Student's t-test, p < 0.001). Scale bars: A, 200 µm; B, 500 µm.  cant difference in the estimated TS ( Fig 3B) or TC (Fig 3C) in the proximal and distal retina of E12.5 PAX77 +/+ embryos compared to wild-types (P > 0.05, Student's t test, n = 3 wild-type and 3 PAX77 +/+ embryos). We performed an analysis of cell death in the retina by examining the pattern of dying cells clearly visible on thin plastic sections as pyknotic nuclei. Pyknosis is the most characteristic feature of apoptosis and counting pyknotic nuclei is a standard way of looking at cell death in the retina [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30]. In a previous study we counted apoptotic cells in the brain by counting pyknotic nuclei as well as by counting Terminal Uridine Deoxynucleotidyl Transferase dUTP Nick End Labeling (TUNEL) positive cells and found that both techniques gave the same result [31].We did not observe any obvious increase in the number of pyknotic figures per 2 µm section through the eye of E12.5 PAX77 +/ + embryos compared to wild-types. In sections from eyes of both genotypes there were between 0 and 2 pyknotic figures per section. Thus, there does not appear to be a dysregulation of progenitor proliferation and death at early stages of retinogenesis in PAX77 +/+ embryos.

Overexpression of Pax6 results in optic disc malformations, progressive retinal dysplasia and microphthalmia from E14.5
As development progresses PAX77 +/+ retinae become progressively smaller than normal while the size and appearance of their lenses appears relatively less affected (Fig 2B-F). As a consequence, from postnatal day (P) 2 their vitreous chambers have almost completely disappeared and parts of their retinae and lenses are in direct contact ( Fig  2E, F).
Morphological defects are first detected at E14.5 with the development of persistent hyperplastic primary vitreous (PHPV), a condition that arises when the hyaloid canal does not retract appropriately towards the optic nerve head leaving a tissue nodule (arrow in Fig 2B-D) [32], and malformations of the optic disc (Fig 2B, H). At this stage in the wild-type, the pigmented epithelium (arrowhead in Fig. 2H) and the retinal neuroepithelium stop sharply at the junction of the optic cup and optic stalk, whereas in the mutant, cells continuous with those of pigmented epithelium and neural retinal cells extend along the optic stalk (arrow in Fig 2H). Taken together, these results indicate that failure of the optic nerve head to develop normally is an early feature in PAX77 +/+ embryos. Defects of RGC axons at around this time are analysed in greater detail below.
By E18.5 the retinae of PAX77 +/+ eyes have become significantly smaller than those of the wild-type eyes ( Fig 2D) and the iris appears folded and pushed against the lens epithelium (arrow in Fig 2I). At E18.5, the wild-type retina comprises two well-defined layers. The inner layer is the Pax6 overexpression does not significantly affect retinal pro-genitor proliferation at E125 (A) Horizontal section of the eye of an E12.5 wild-type embryo labeled with anti-BrdU (red), anti-IdU (green) and counterstained with TOPRO3 (blue) Figure 3 Pax6 overexpression does not significantly affect retinal progenitor proliferation at E12.5 (A) Horizontal section of the eye of an E12.5 wild-type embryo labeled with anti-BrdU (red), anti-IdU (green) and counterstained with TOPRO3 (blue differentiated RGC layer and the outer layer is the region of the retina that is still primarily neuroblastic. While both layers are present in the mutant retina, the RGC layer appears to be less structured than in the wild-type retina (Fig. 2J). Four days after birth, when normal neuronal production is declining rapidly and the retina is initiating the differentiation of the inner and outer nuclear layers, rosette-like structures are starting to form in the outer layer of the retina of PAX77 +/+ mice (Fig 2E, K). In addition, the peripheral retina is folded back towards what remains of the vitreous chamber and retinal folds appear more centrally (Fig 2E). At P6, PAX77 +/+ retinae display degrees of rosetting that vary between eyes and also between different regions of the same eye. Figures 2F and  2L show a PAX77 +/+ eye at P6 with very severe rosetting. In all P14 PAX77 +/+ eyes examined the entire outer nuclear layer consisted of dysplastic rosettes ( Fig 4J) and the retina displayed extensive gliosis (shown by glial fibrillary acidic protein [GFAP] staining in Fig 4J).
The abnormality of the optic disc observed in the mutant embryos from E14.5 is similar to that previously described in mice heterozygous for the Kidney and retinal defects (krd) deletion, which display haploinsufficiency of the transcription factor Pax2 [33]. We could not, however, detect by immunofluorescence any obvious difference in the pattern or the level of expression of Pax2 in the eyes of PAX77 +/+ embryos compared to wild-types ( Fig 5).

Analysis of patterns of gene expression in the retinae of PAX77 +/+ and wild-type mice
In the retina of wild-types aged E18.5, proliferative cells are found in the outer layer while the inner layer contains the recently differentiated RGCs. We analyzed the organization of the retina at E18.5 using Brn3a ( Fig 4A) as a marker of RGCs [34,35] and BrdU to label proliferative cells in S-phase ( Fig 4B). There are no obvious differences in the positions and densities of cells labelled with BrdU or Brn3a between PAX77 +/+ and wild-type retina (Fig 4A,  B). A further quantitative analysis of Brn3a labelling in wild-type and mutant embryos, presented below in the context of RGC axonal development, supported these findings. Thus, at E18.5, although the PAX77 +/+ retina is smaller than the wild-type retina, its organization appears to be normal.
We analyzed the expression of a broader range of markers in the postnatal retina, at P6. The RGCs, which express Pax6, Islet1 and Brn3a [3,36], are present in the mutant retina but they are not as regularly aligned as in the wildtype retina (Fig 4C, D and arrowheads in Fig 4I). Pax6, Islet1 and Syntaxin are expressed by amacrine cells [3,36] ( Fig 4C, D, F; arrowheads), Chx10 is a marker of bipolar and retinal progenitor cells [37] (Fig 4E; arrows) and Pax6 and R-cadherin are expressed by horizontal cells [3,36,38] ( Fig 4C, G; arrows). Analysis of the expression of these markers in PAX77 +/+ retina provided evidence that all the major cell types of the inner nuclear layer are produced in the mutant, but their appearances and distributions are not normal at P6. Cells expressing markers associated with cells from the inner nuclear layer accumulate at the outer surface of the retina (Fig 4C, D, F; stars); their appearance suggests that they might have migrated through the outer nuclear layer. In addition, there seem to be fewer horizontal cells in the PAX77 +/+ retina than in the wild-type retina and they are not as regularly spaced ( Fig  4C, G; arrows). In the wild-type retina, R-cadherin is strongly expressed in the outer plexiform layer, which contains the horizontal cells' dendrites ( Fig 4G; arrowheads). The expression of R-cadherin in the outer plexiform layer of the mutant retina is much weaker and does not appear as a continuous line as in the wild-type ( Fig  4G). This suggests that the horizontal cells in the mutant retina do not form normal dendritic connections with cells in the outer nuclear layer. In the wild-type retina, the outer limiting membrane is strongly stained by antibodies against N-cadherin [38,39] (Fig 4H; arrow). The outer limiting membrane in the PAX77 +/+ eye is not as clearly defined as in the wild-type and it appears discontinuous (Fig 4H; arrow). Overall, these results indicate that the major cell types of the normal retina are produced in the mutants but they are disorganized at P6.
We used real time quantitative RT-PCR to analyze the mRNA levels of several genes which are implicated in retina formation and RGC axon navigation and some of which are known targets of regulation by Pax6. We compared the mRNA levels of our candidates in the retinae of wild-type and PAX77 +/+ embryos at E14.5 and E16.5, stages at which retinal dysmorphology first emerges in the mutants (Table 1). Shh, which is expressed specifically by the RGCs [40,41], is implicated in retinal lamination and RGC axon guidance [40,41]. We found a small, significant decrease in the levels of Shh mRNA in the PAX77 +/+ retinae compared to the wild-types at E16.5 (Table 1). We also analyzed the expression of the adhesion molecules Ncam, R-cadherin and L1 which are implicated in axon growth and guidance and are transcriptional targets of Pax6 [11,[42][43][44][45]. L1, in particular, is essential for topographic mapping of retinal axons [43]. We did not detect any difference in expression of these genes in the PAX77 +/ + retina compared to the wild-type. The homeodomain transcription factor Six3 is an important regulator of retinal development and may regulate retinal progenitor proliferation and cell fate specification [46][47][48]. There is evidence that the expression of Six3 is regulated by Pax6 [49]. The bHLH transcription factor Ngn2 has been implicated in the regulation of the specification of RGCs [50] and is a transcriptional target of Pax6 in the telencephalon and spinal chord [51,52]. The expression of Six3 and The different retinal cell types are produced normally in the PAX77 +/+ eye -Ngn2 was not significantly altered in the PAX77 +/+ compared to the wild-type retina.

Overexpression of Pax6 results in intraretinal and retinothalamic errors in RGC axon navigation
In order to test the hypothesis that normal levels of Pax6 are important for the proper navigation of RGC axons we first examined the ability of RGC axons to navigate within the PAX77 +/+ retina. We looked at E16.5 retinae as, by this age in wild-types, large numbers of RGC axons have grown to the optic nerve head and any defects in intraretinal navigation in mutants would be apparent. In wildtype embryos, RGC axons grow radially towards the optic nerve head as revealed by immunostaining RGC axons for L1 (Fig. 6A, B) or neurofilament (Fig. 6C), or labelling cohorts of RGC axons by DiI injection into the peripheral retina ( Fig. 6D-F). In contrast, in PAX77 +/+ embryos RGC axons take a more erratic path. L1 (Fig. 6G, H) and neurofilament ( Fig. 6I) immunostaining showed that, as in the wild-type, RGC axons form bundles that converge on the optic nerve head. However, whereas in the wild-type these bundles run approximately linearly and cover the retinal surface evenly, in the PAX77 +/+ embryo they take a more serpentine course and are not evenly distributed across the retinal surface. Focal DiI injections labelling cohorts of RGC axons ( Fig. 6J-L) showed that although many axons do converge on the optic nerve head their course is erratic.
Despite exhibiting abnormal routing within the retina, a significant number of RGC axons exit the retina at the optic nerve head and grow into the optic stalk in PAX77 +/ + embryos. During normal development the optic nerves Pax2 appears to be expressed normally in the optic stalk of the PAX77 +/+ eye converge on the ventral surface of the hypothalamus to form the optic chiasm. Axons are then sorted into the contralateral or ipsilateral optic tracts and navigate to their targets in the thalamus and superior colliculus. Immunostaining for Pax6 in wild-type (Fig. 7A, B) and PAX77 +/+ (Fig. 7D) embryos shows that Pax6 is not expressed at detectable levels at the optic chiasm at E14.5, the age at which RGC axons are navigating this region. Horizontal sections of wild-type (Fig. 7C) and PAX77 +/+ (Fig. 7E) optic chiasm following DiI injection into the retina reveal that in both cases significant numbers of axons reach the chiasm and continue into the optic tracts. We noticed that our injections consistently labelled more axons in the wild-type than in the PAX77 +/+ embryos suggesting that fewer axons reached this region in the PAX77 +/+ embryos. In order to quantify this we counted the numbers of RGC cell bodies retrogradely labelled by unilateral DiI injection into the thalamus at E16.5 (Fig. 7F). These counts confirmed our initial observation that the PAX77 +/+ retina projects fewer axons to the thalamus: the total number of back-labelled RGCs in the PAX77 +/+ retina was reduced compared to the wild-type (p < 0.05; Mann-Whitney Rank Sum Test). Interestingly, however, there was a disproportionate reduction in the contralateral projection which was reduced by 90% in the PAX77 +/+ embryos (p < 0.05; Mann-Whitney Rank Sum Test) compared to the ipsilateral projection in which there was no significant decrease (p > 0.05; Mann-Whitney Rank Sum Test). This is examined in more detail below.

The number of RGCs in the retina is sensitive to Pax6 gene dosage
We studied whether the reduction in the number of RGCs labelled by DiI from the thalamus in PAX77 +/+ mutants could be accounted for by a reduction in the total number of RGCs in the mutant retina. In view of the interesting observation that there was a disproportionate reduction in the contralateral projection in PAX77 +/+ mutants, we extended the study to include embryos with a range of Pax6 gene dosage. We counted the total number of RGCs, marked by expression of Brn3a, in whole E16.5 retinae from Pax6 Sey/+ , Pax6 +/+ , PAX77 + , and PAX77 +/+ embryos. This analysis showed that all three types of mutant retina produced about a third fewer Brn3a-expressing RGCs than the wild-type ( Fig. 8A; this was significant p < 0.05; Mann-Whitney Rank Sum Test) although there was no significant difference between the mutant genotypes. This reduction in RGC number was accounted for by the reduction in retinal area in Pax6 Sey/+ , PAX77 + , and PAX77 +/+ embryos as the density of Brn3a-expressing cells in the RGC layer and the depth of the RGC layer were similar in Pax6 Sey/+ , Pax6 +/+ , PAX77 + , and PAX77 +/+ embryos ( Table 2). There was no significant difference in the density of Brn3aexpressing cells between ventro-temporal and dorso-nasal retina for each genotype ( Table 2), indicating that the relatively even distribution of RGCs over the retinal surface seen in wild-types is maintained in all three types of mutant.
The reduced number of RGCs at E16.5 resulting from Pax6 overexpression raises the possibility that the onset of ganglion cell differentiation may be delayed. We therefore examined the number and distribution of Brn3a expressing cells at the onset of RGC differentiation [34,35] in E12.5 retinae of Pax6 +/+ , PAX77 + , and PAX77 +/+ embryos (Fig 9). We found no differences in the distribution (Fig 9A-F) or number (Fig 9G) of Brn3a positive cells between the three genotypes. We conclude that the retinal axon guidance defects caused by Pax6 overexpression do not arise from delayed onset of RGC differentiation.

The midline crossing behaviour of RGC axons at the optic chiasm is sensitive to Pax6 gene dosage
RGC axons are sorted into the optic tracts at the optic chiasm. In wild-type mice about 3% do not cross the midline and grow into the ipsilateral optic tract. The remaining 97% cross the midline and grow into the contralateral optic tract. In PAX77 +/+ embryos the increased levels of Pax6 correlate with an increase in the proportion of RGCs projecting ipsilaterally. In order to investigate the relationship between Pax6 and the ipsilateral projection in more detail we examined other lines of mice with different numbers of Pax6/PAX6 gene copies. Alterations to RGC axon trajectories within the PAX77 +/+ retina at E16.5  [5]. Loss of one functional copy of Pax6 in Pax6 Neu/+ embryos produces a hypoplastic optic nerve [53]. Quantification of RGCs retrogradely labelled by unilateral DiI injection into the thalamus confirm that the total projection in Pax6 Sey/+ embryos is reduced to 12% (p < 0.05; Mann-Whitney Rank Sum Test) of that seen in wild-types (Fig 8B). In contrast to the PAX77 +/+ embryos, however, the reduction affects ipsilateral and contralateral projections equally and the proportion of RGCs projecting ipsilaterally is 3%, not significantly different to the 2.5% in the wild-type (Fig 8C; p > 0.05; Mann-Whitney Rank Sum Test). Reducing the number of functional Pax6 gene copies reduces the number of RGC axons projected by the retina to the thalamus but does not affect the proportion that cross the midline at the optic chiasm. Thus, reduction in the number of RGCs, which occurs in both Pax6 Sey/+ and PAX77 +/+ embryos, is not sufficient to explain the increased proportion of ipsilateral projections in PAX77 +/+ mutants, which must be regulated separately by gene dosage.
PAX77 + hemizygotes [12] contain 5 to 7 copies of human PAX6 in addition to the two endogenous mouse Pax6 genes. The Pax6/PAX6 gene dosage in these embryos is therefore intermediate between wild-type embryos and PAX77 +/+ embryos [12]. Quantification of RGCs retrogradely labelled by unilateral DiI injection into the thalamus showed no significant difference in the overall numbers of RGCs projecting into the brain compared to wild-types (p > 0.05; Mann-Whitney Rank Sum Test; Fig  8B), even though the total number of Brn3a-expressing RGCs was less than in wild-types (Fig. 8A). The proportion of RGCs projecting ipsilaterally was about 10%, significantly more than the 2.5% seen in wild-types (p < 0.05; Mann-Whitney Rank Sum Test; Fig 8C). As described above, in PAX77 +/+ embryos there is a significant reduction of the total projection to 10% of its normal size (p < 0.05; Mann-Whitney Rank Sum Test; Figs 7F and 8B) of which 11% is ipsilateral, a significant increase compared to the wild type ipsilateral projection (p < 0.05; Mann-Whitney Rank Sum Test; Fig 8C). The tendency of ipsilaterally projecting RGCs to be most concentrated in the ventro-temporal retina, which has been well established in Pax6 +/+ embryos [eg [42,44,49]], was maintained in Pax6 Sey/+ , PAX77 + , and PAX77 +/+ embryos (Table 3) irrespective of the changes in the total number or proportion of ipsilaterally projecting RGCs. These results, summarised in Fig 8D, show that increasing the number of functional PAX6 gene copies increases the proportion of RGC axons projecting ipsilaterally independently of its effect on the total numbers of RGCs whose axons reach the thalamus.

Discussion
We analyzed the effects of Pax6 overexpression on the development of the neural retina and on its ability to project axons to the brain. Our results indicate that, in the eye of PAX77 +/+ embryos, (1) development proceeds normally until about E12.5, (2) the first defects, which are of the optic nerve and the growth and navigation of RGC axons, emerge at E14.5-E16.5, (3) major retinal cell types are generated and (4) there is progressive retinal dysplasia and microphthalmia from E18.5 into postnatal life. The globe and retina fail to grow and by P4 the vitreous has almost disappeared leaving the retina in contact with the lens. Retinal lamination is also affected with the appearance of many rosettes and abnormal horizontal cells. One clear conclusion from these findings is that while both over-and under-expression of Pax6 cause microphthalmia, the underlying mechanisms are different. Microphthalmia in Pax6 Sey/+ mice is thought to be caused by primary defects of the lens and anterior structures occurring at the time of initial eye formation rather than with later retinal defects [8,9]. Pax6/PAX6 gene dosage affects both the total number of RGCs, the number of RGC axons projected to the thalamus and the proportion that project ipsilaterally

Possible causes of intraretinal RGC axonal defects
During normal development RGC axons initially grow straight towards the optic nerve head. Overexpression of Pax6 does not prevent RGC axons converging on the optic nerve head altogether, but does cause a substantial proportion to take a less direct route than normal. The intraretinal navigation defects occur before E16.5, when the morphology of the PAX77 +/+ retina is not grossly abnormal, making it less likely that the distorted RGC axons are a secondary consequence of retinal dysplasia. Intraretinal axon navigation defects have been reported in mice lacking the secreted axon guidance molecules Slit1 and/or Slit2 [54] or the cell surface receptors EphB2 and EphB3 [55], or netrin1 and its receptor DCC (deleted in colorectal cancer) [56]. Although these loss-of-function phenotypes do not precisely phenocopy the PAX77 +/+ phenotype reported here, it is possible that Pax6 normally regulates the expression of one or more of these molecules and that its overexpression alters their normal balance resulting in the intraretinal axon guidance defects we observed. As Pax6 is expressed by RGCs themselves, and by other retinal cell types, it is possible that the intraretinal pathfinding errors reflect sensitivity of the RGC growth cone and/ or the environment through which it navigates to levels of Pax6. It has recently been postulated that the lens directs RGC axons towards the optic nerve head by secreting repulsive Slit molecules [54]. As Pax6 plays a role in the formation of the lens [10], and overexpressing Pax6 causes defects in lens development [57][58][59], it is quite possible that overexpression of Pax6 in the lens is directly responsible for a component of the intraretinal misrouting. The proper formation of the optic nerve head is also important for intraretinal axon navigation [56] and its altered structure in PAX77 +/+ embryos could influence axon navigation.

Possible causes of RGC axonal defects at the optic chiasm
Retinal axon navigation at the optic chiasm is controlled by genes ranging from transcription factors [60] to cell surface and secreted proteins [61]. Manipulating a small subset of these genes produces specific alterations to chiasm organisation (for example see [53,[62][63][64][65][66]). This complex genetic control may underlie variations in the details of chiasm construction between vertebrate species. In mice the vast majority of retinal axons cross the ventral midline at the optic chiasm and join the contralateral optic tract growing towards their targets in the thalamus and superior colliculus whereas a minority do not cross and join the ipsilateral tract [67]. To date only a small number of transcription factors have been identified which control the balance between ipsilateral and contralateral projections (i.e. [62,66,68,69]). A study of chick retina identified a gradient of Pax6 expression levels consistent with a role in topographic mapping of RGC axons via regulation of the expression of the axon guidance receptor EphB2 [70]. As signalling between EphB receptors and their ephrinB ligands have been implicated in axon crossing behaviour at the chiasm [64], it is possible that the regulation of EphB by Pax6 is conserved between chick and mouse and that overexpressing Pax6 alters the expression of EphB receptors on RGC growth cones and influences their midline crossing behaviour at the chiasm.
To investigate the relationship between Pax6 levels and RGC axon navigation at the optic chiasm we examined an allelic series of genotypes with different Pax6/PAX6 gene dosage. Overexpression of Pax6 resulted in an increase in the proportion of RGCs projecting ipsilaterally combined with either no change in the total number projecting (as in the PAX77 + phenotype) or with a reduction in the total numbers projecting (as in the PAX77 +/+ phenotype). Conversely a reduction in the total numbers projecting occurred without changing the proportion of the ipsilateral projection in the Pax6 Sey/+ phenotype. This provides strong evidence that the increased proportion of ipsilaterally projecting axons in PAX77 +/+ embryos is not simply a consequence of a reduced total projection. As Pax6 is expressed by RGCs at the time their axons navigate the midline but not by chiasm cells at the midline it is likely Distributions of RGCs projecting ipsialterally to the thalamus in Pax6 Sey/+ , Pax6 +/+ , PAX77 + , and PAX77 +/+ embryos in the four retinal quadrants. Note that in all genotypes the ipsilaterally projecting RGCs are enriched in the ventrotemporal (vt) quadrant (shown in bold) compared to the dorsonasal (dn), ventronasal (vn), and dorsotemporal (dt) quadrants regardless of the overall size of the ipsilateral projection (shown in Fig 8 B,C). All values mean ± s.e.m. Sample sizes: Pax6 Sey/+ n = 4; Pax6 +/+ n = 9; PAX77 + n = 5; PAX77 +/+ n = 4.
that Pax6 levels in the RGCs themselves are important in defining the navigational properties of their growth cones. Pax6 overexpression might increase RGC axon growth cones' sensitivity to repulsive guidance cues at the chiasm [64], therefore reducing the proportion that will cross the midline and project contralaterally. This might occur either directly by regulation of axon guidance molecules, such as EphB1 [64,70], or indirectly by regulation of transcription factors, such as Zic2 [62], which program an RGC to project ipsilaterally. In either case this is the first evidence that Pax6 acts as an 'ipsilateral determinant' in its own right.
Brn3a is expressed by a large population of RGCs innervating the principal retinothalamic and retinocollicular pathway [71] so is a useful marker for determining whether the size of the population of RGCs that would normally project to the thalamus is affected when Pax6/ PAX6 dosage is altered. RGCs projecting to ipsilateral targets in the thalamus and superior colliculus do not express Brn3a whereas contralateral projecting RGCs do, indicating that Brn3a expressing RGCs tend to project contralaterally [71]. Embryos either underexpressing or overexpressing Pax6/PAX6 have similar numbers of Brn3a expressing RGCs but different numbers projecting to the ipsilateral and contalateral thalamus indicating that in our models alterations of the size of the ipsilateral or contralateral projection cannot be explained by variations in the numbers of cells expressing Brn3a. The total numbers of Brn3a-expressing RGCs produced by Pax6 Sey/+ , PAX77 + , and PAX77 +/+ embryos at E16.5 were very similar to one another, although in all cases about 30% less than in Pax6 +/+ embryos. The earliest RGC projection is a transient ipsilateral projection from the dorsocentral retina at E12.5 [72]. As Pax6 overexpression does not affect the number or dorsocentral location of Brn3a expressing RGCs at E12.5, it is extremely unlikely that the increased ipsilateral projection following Pax6 overexpression is simply a consequence of delayed retinal development. The 90% reduction in the total numbers of RGCs projecting to the thalamus in Pax6 Sey/+ and PAX77 +/+ embryos cannot be accounted for by a failure to generate RGCs but must be caused by axons failing to get to the thalamus, for example by wandering around in the retina, as seen in PAX77 +/+ embryos. It is possible that some RGCs fail to project axons out of the retina because they have a defect in axonogenesis like that seen when RGCs lack the transcription factor Brn3b [73].

Possible causes of early defects of optic nerve head
Pax6 is initially co-expressed with Pax2 in the early optic stalk before becoming restricted to the optic cup by E12.5. Baumer et al. [74] have shown that maintaining the expression of Pax6 in the optic stalk after E12.5, by expressing Pax6 under the control of a Pax2-upstream promoter fragment, results in the appearance of retinal pigmented epithelium (RPE) in the optic disc, which is one of the earliest abnormalities detected in PAX77 +/+ E14.5 embryos. Although we did not detect Pax6 in the optic stalk of PAX77 +/+ mice after E12.5, it is possible that its overexpression causes a slight delay in its downregulation in the optic stalk. Such a delay could be responsible for the ectopic presence of RPE and retinal cells in the optic disc.
The malformation of the optic disc in PAX77 +/+ embryos could arise as a consequence of the abnormal development of RGC axons. Alternatively, the defects of RGC axon guidance might be secondary to primary defects at the optic nerve head. In this study we show that, at E16.5, RGCs display axon guidance defects and the levels of Shh mRNA, specifically expressed by the RGCs [40,41], are slightly lower than normal. Previous studies have shown that Shh, expressed by the RGCs, is implicated in retinal lamination and RGC axon guidance. Conditional inactivation of Shh in RGCs causes a complete loss of optic disc astrocyte precursor cells, resulting in defective axon guidance in the retina, as well as conversion of the neuroepithelial cells in the optic stalk to pigmented cells [40]. Loss of RGC-derived Shh also causes extensive disorganization of the retina with the appearance of many rosettes in the outer nuclear layer [41]. Thus, the slight decrease in Shh expression found in the mutant retina at E16.5 could reflect abnormalities in RGC differentiation and could contribute to the defects in RGC axon guidance and retinal lamination observed in the PAX77 retina.
In addition, another study has shown that specific ablation of newly formed RGCs results in a decrease of retinal progenitor cell proliferation and abnormally thin retinal layers in the postnatal eye [75]. Thus, a defect in RGC differentiation could also account for the reduced size of the retina in PAX77 +/+ mice. The fact that major retinal defects appear after RGC defects in axon guidance and Shh expression supports the hypothesis that microphthalmia and retinal dysplasia in PAX77 +/+ eyes could be secondary to earlier abnormalities in RGC differentiation caused by Pax6 overexpression.

Conclusion
Higher than normal levels of Pax6 produce complex phenotypes in which several features of eye morphogenesis are affected. Both increased (present study and [12]) and decreased [76][77][78][79][80]Pax6 gene dosage result in defective eye development, but the effects are quite different. Whereas defects of Pax6 +/mice occur mainly in the anterior segment of the eye [76][77][78][79][80], we show here that overexpression of Pax6 causes major defects of retinal development. Increased Pax6 levels also affect the projection of retinal axons. Since retinal axon navigation generally follows eye morphogenesis it is possible that a component of the intraretinal guidance errors we detect is a secondary consequence of alterations in retinal, lens, and/or optic disc anatomy or molecular composition. The navigation errors we detect at the chiasm are likely to reflect a requirement for normal Pax6 levels in programming the responses of RGCs to guidance cues at the chiasm.

Methods
Mice PAX77 hemizygous mice [12], designated PAX77 + , carry 5 to 7 copies of a 420 Kb human PAX6 YAC (Y593) with all copies integrated at the same locus. We refer to the array of integrated YAC Y593 copies as the PAX77 transgene. PAX77 homozygous mice, designated PAX77 +/+ , carry 10 to 14 copies of human PAX6 and were genotyped by fluorescent in situ hybridization using the Fat5 probe as described in Schedl et al. (1996) [12]. As no abnormalities are observed in animals carrying the transgene on a Pax6 Sey/+ heterozygote background Schedl et al [12] concluded that the ocular abnormalities in PAX77 mice are due to overexpression of PAX6 and not to overexpression of a second gene encoded by the YAC. All mice were maintained on an albino CD1 background. The morning of the vaginal plug was deemed E0.5. The first 24 h after birth was deemed P0. Animal care followed institutional guidelines and UK Home Office regulations.

Histology
For each stage and each genotype, at least 2 samples were fixed in 4% paraformaldehyde : 2% glutaraldehyde, dehydrated to 100% ethanol and embedded in cold-polymerizing resin (Technovit 7100, Kulzer Histo-Technik). Sections were cut (5 µm) and stained with cresyl violet. Sections of whole embryos were cut horizontally and sections of postnatal eyes were cut parallel to the optic nerve.

BrdU/IdU analysis of cell-cycle parameters
The BrdU/IdU double labeling protocol is detailed in Martynoga et al.(2005) [15]. In brief, at T = 0 h the pregnant mouse receives an injection of IdU which labels cells going through S-phase from the beginning of the experiment. At T = 1.5 h the female receives an injection of BrdU and the embryos are fixed shortly after (T = 2 h). Thus, the BrdU labels cells which are in S-phase at the end of the experiment; this is the S fraction (S cells ). The cells labeled only by the IdU but not by the BrdU are cells which have left S phase in the interval of 1.5 h between the two injections. This is the leaving fraction (L cells ). The cell cycle length (Tc) and the length of S-phase (Ts) can be calculated using the equations of Shibui et al.(1989) [14]: Ts = Ti/(L cells /S cells ) and Tc = (S cells /P cells ), where Ti is the interval between the two injections and P cells is the total number of proliferating cells. We made the assumption that all progenitor cells of the retina are proliferating. The growth fraction was not determined in this experiment.

Immunohistochemistry
Whole embryos or postnatal eyes were fixed in 4% paraformaldehyde and either processed to wax or cryoprotected in 15% and 30% sucrose and frozen.

Quantification of Brn3a-expressing cells in the retina
Retinae were removed from heads previously fixed in 4% paraformaldehyde and an orientating cut made in the ventral retina of E16.5 embryos. Retinae were boiled in 10 mM sodium citrate buffer pH6.0 for 10 minutes and blocked in 5% goat serum, 0.2% BSA, 0.1% Triton-X1000 in PBS at room temperature before incubating overnight at 4°C in primary antibody [Brn3a Chemicon; MAB1585)] diluted 1:300 in blocking buffer. After washing extensively in 0.1% Triton-X1000 in PBS at room temperature retinae were incubated overnight at 4°C with an Alexa488-conjugated anti-mouse secondary (Molecular Probes, Inc), washed extensively, and mounted with the RGC layer uppermost in 90% glycerol containing MOWIOL and DABCO. For E16.5 retinae, a Leica TCS NT confocal microscope was used to acquire (1) a low power image of the entire retina for calculation of retinal area [see Fig 8A], (2) to scan through the entire Brn3a-expressing RGC layer at high power and measure its depth and (3) to collect a series of optical sections for sampling within the RGC layer.
Sections were 160 µm × 160 µm and were collected at 0.16 µm intervals through 8 µm of retinal depth. The boxed area in Fig 8A shows an example of a sampling area in ventrotemporal retina. We used a stereological approach to calculate the density of RGCs within the sampling volume: optical sections comprising the upper and lower 4 µm of the total 8 µm stack were combined separately to generate images like the high power inset shown in Fig 8A. Simply counting the number of Brn3a-expressing nuclei in each image would give an overestimate of nuclear density in each 160 µm × 160 µm × 4 µm volume as nuclei protruding into the sampling volume both from above and from below would be included in the count. To avoid overestimating the nuclear density, we only counted nuclei protruding into the sampling volume from one side while excluding those that protruded from the other. This was done by using Adobe Photoshop to pseudo-colour the nuclei in the upper stack red and those in the lower stack green and then merging the two images. Nuclei appearing yellow [red and green combined] were present in both upper and lower stacks and so were excluded from the count. This gave an accurate number of Brn3a-expressing nuclei present in the sampling volume and was used to calculate their density. For each retina the density was calculated in the dorso-nasal and ventro-temporal retina. The number of Brn3a-expressing cells in the retina was calculated by multiplying the retinal area × depth of RGC layer × density of Brn3a-expressing cells in RGC layer. For E12.5 retinae, where the density of RGCs is much lower, a Leica TCS NT confocal microscope was used to acquire a scan through the whole retina and total numbers of Brn3a positive cells were counted.

Tract-Tracing
Tract-tracing was performed as described in   [65]. Embryonic heads were fixed at 4°C in 4% paraformaldehyde in PBS overnight. DiI crystals (Molecular probes, USA) were either (1) placed in the optic cup of one eye after removal of the lens or (2) placed in a line over the dorsal thalamus on one side to label axons navigating the optic tract. Heads were returned to 4% paraformaldehyde in PBS in the dark at room temperature for about six weeks to allow tracers to diffuse along axons. In some cases the retinae were removed from the head, cleared in 9:1 glycerol:PBS, and imaged as wholemounts using a Zeiss Axiovert confocal LSM 510 microscope (Zeiss, Germany). In other cases heads were sectioned (200 µm) with a vibratome, cleared in 9:1 glycerol:PBS containing the nuclear counterstain TOPRO3 (1.0 µM, Molecular Probes, USA), mounted in Vectashield (Vector Laboratories, USA), and imaged using an epifluorescence microscope and digital camera (Leica Microsystems, Germany) or a TCS NT confocal microscope (Leica Microsystems, Germany). For quantification of RGC projections the total number of DiI labelled RGCs in each retina was counted in serial vibratome sections using an epifluorescence microscope as follows. Each section was viewed under epifluorescence using a x20 objective at which magnification DiI labelled RGC cell bodies are easily resolved. An eyepiece graticule was used to divide the retina into 50 µm wide bins. Within each bin DiI labelled RGC bodies were counted by focussing through the thickness of the section and counting labelled cell bodies as they came into focus. Individual bin counts were then summed to give totals for RGCs projecting ipsilaterally and contralaterally to the thalamic DiI injection site in each embryo and these totals were used to generate data presented in Fig 7, Fig 8, and Table 3. The numbers of E16.5 embryos used for quantification were as follows: Pax6 Sey/+ n = 4; Pax6 +/+ n = 9; PAX77 + n = 5; PAX77 +/+ n = 4. DiI appears orange and TOPRO3 appears red in epifluorescence images and in confocal images DiI appears red and TOPRO3 appears blue.

Quantitative reverse transcription-PCR (qRT-PCR)
For each sample RNA was extracted from a pool of retinae from a wild-type or PAX77 +/+ litter using Qiagen RNeasy kit (Qiagen, USA). cDNA synthesis was performed as described in [82]. qRT-PCR was performed on cDNA from E14.5 and E16.5 retinae with the following primer pairs