We compared Pax6/PAX6 (mouse + human) expression in wild-type and PAX77^+/+ eyes by immunohistochemistry on sections of E12.5 and E14.5 embryos. The pattern of expression of Pax6/PAX6 in the eye of PAX77^+/+ embryos reproduces that in the wild-types (Fig 1A, B). As expected, the levels of Pax6/PAX6 in PAX77^+/+ eyes appear increased compared to those in wild-types (Fig 1A, B). We also measured the levels of specifically mouse Pax6 mRNA and both mouse and human Pax6/PAX6 mRNA in the retinae of E14.5 wild-type and PAX77^+/+ embryos by quantitative RT-PCR (qRT-PCR). We found that the total levels of Pax6/PAX6 mRNA are increased by about 2.5 fold in PAX77^+/+ compared to wild-type retinae (Fig 1C). This increase is lower than might be expected from the increase in gene copy number (5–7 fold), suggesting that Pax6 negatively autoregulates in the retina, as it does in the brain 13. In agreement with this, we found that the levels of endogenous mouse Pax6 mRNA are decreased in PAX77^+/+ retinae compared to wild-type retinae (Fig 1C), indicating that overproduction of PAX6 from the YAC transgenes represses production from the endogenous Pax6 loci. For convenience, we refer elsewhere in this paper to the overexpression in PAX77 mice as being an overproduction of Pax6, rather than Pax6/PAX6, since the human and mouse protein sequences are identical.

At early stages of embryogenesis, up to E12.5, the eyes of PAX77^+/+ embryos appear similar to those of the wild-types in both size and morphology (Fig 2A, G). They do not show the lens defects reported in Pax6^+/- embryos at this age 89. We tested for more subtle defects of cell proliferation and death at this age but found no evidence for either. We used the technique of sequential labeling of cells in S-phase with BrdU and IdU 1415 to estimate the cell-cycle length (TC) and the length of the S-phase (TS) of retinal progenitors (Fig 3). We did not find any significant 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 161718192021222324252627282930. 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.

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 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).

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 3435 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 336, are present in the mutant retina but they are not as regularly aligned as in the wild-type retina (Fig 4C, D and arrowheads in Fig 4I). Pax6, Islet1 and Syntaxin are expressed by amacrine cells 336 (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 33638 (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 3839 (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 4041, is implicated in retinal lamination and RGC axon guidance 4041. 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 1142434445. 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 464748. 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 5152. The expression of Six3 and Ngn2 was not significantly altered in the PAX77^+/+ compared to the wild-type retina.

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 wild-type 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 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.

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 Brn3a-expressing 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 3435 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.

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.

Small-eye heterozygote (Pax6^Sey/+) embryos have only one functional copy of Pax6 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 424449], 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.

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 575859, 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.

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 536263646566). 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. 62666869). 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 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 6470, 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.

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 4041, 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.

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 al12 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.

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.

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.

Pregnant females were injected intra-peritoneally with 200 ml of 10 mg/ml (in 0.9% NaCl) IdU (Sigma), then 1.5 h later with the same dose of BrdU (Sigma) and they were sacrificed after 30 min. 10 μm wax sections were immunostained with mouse anti-BrdU/IdU (Becton Dickinson, clone B44, 1:100) and rat anti-BrdU (Abcam, clone BU1/75, 1:100). Cells were counted in 100 μm wide sampling boxes in the distal and proximal retina of 3 wild-type and 3 Pax77^+/+ E12.5 embryos. Each cell count was repeated on at least 3 non-adjacent sections from each embryo.

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

Immunostaining for Pax6, glial fibrillary acidic protein (GFAP), Brn3a, Islet1, bromodeoxyuridine (BrdU), Syntaxin and N-Cadherin (Ncad) was performed on 10 μm paraffin sections as described in Martynoga et al. (2005) 15. For studies with BrdU, pregnant females were injected intra-peritoneally with 200 μl of 10 mg/ml BrdU (in 0.9% NaCl; Sigma) and were sacrificed after 1 hour. Immunostaining for Chx10 and Pax2 was performed on 10 μm frozen sections as described in Martynoga et al. (2005) 15. R-Cadherin (Rcad) immunostaining was performed on frozen sections as described in 81. Sections of whole embryos were cut horizontally and sections of postnatal eyes were cut parallel to the optic nerve.

For whole-mount immunofluorescence, dissected retinae from E16.5 embryos were incubated in blocking buffer (20% goat serum, 0.2% Triton-X100 in phosphate buffered saline [PBS]) for 20 min, followed by anti-neurofilament or anti-L1 antibody overnight at 4°C. For neurofilament detection, sections were rinsed in 0.2% Triton-X100 in PBS, and incubated with an Alexa488-conjugated anti-rabbit secondary (Molecular Probes, Inc). For L1 detection, a biotinylated secondary antibody was used with Alexa488-conjugated streptavidin (Molecular Probes, Inc).

Primary antibodies were: Pax6 (1:400, DSHB), GFAP (1:50, Dako), Brn3a (1:300, Chemicon; MAB1585), Islet1 (1:50, DSHB), Ncad (1:500, BD), BrdU (1:200, Becton Dickinson), Syntaxin (1:100, Santa-Cruz sc-12736), Chx10 (1:1000, a gift from C. Cepko), Pax2 (1:200, Cambridge Bioscience), Rcad (1:100, a gift from M. Takeichi), Neurofilament (1:100, BIOMOL, USA; NA1297), L1 (1:50, Chemicon; MAB5272).

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 was performed as described in Pratt et al. (2006) 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.

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 (n = 3 litters): Ncam (5'-GACCATCAGGAATGTGGA-3' and 5'-AGGCTTCACAGGTCAGAGT-3'; 179 bp product); R-cadherin (5'-CAGTGAAACAGGGGACATC-3' and 5'-ATACGGTTCTCAGGAACCTC-3'; 216 bp product); Ngn2 (5'-CAAACTTTCCCTCTCTGATG-3' and 5'-CATTCAACCCTTACAAAAGC-3' ; 197 bp product);L1 (5'-ACCCTGAGGCATTACACCTG-3' and 5'-CAACTGCTCTTTGCTTTCCC-3'; 140 bp product); Six3 (5'-CTGGAGAACCACAAGTTCAC-3' and 5'-GATCCTGCAGGTACCACTC-3'; 232 bp product); mouse Pax6 (5'-AACACCAACTCCATCAGTTC-3' and 5'-ATCTGGATAATGGGTCCTCT-3'; 153 bp product); human + mouse Pax6 (5'-TAGCGAAAAGCAACAGATG-3' and 5'-TCTATTTCTTTGCAGCTTCC-3'; 250 bp product); Shh (5'-CCCTTTAGCCTACAAGCAGT-3' and 5'-CCACTGGTTCATCACAGAG-3'; 232 bp product) and GAPDH (5'-GGGTGTGAACCACGAGAAAT-3' and 5'-CCTTCCACAATGCCAAAGTT-3'; 121 bp product). Quantitative RT-PCR was performed using Qiagen Quantitect SYBR Green PCR Kit (Qiagen, USA) and a DNA Engine Opticon Continuous Fluorescence Detector (GRI, UK). The abundance of each transcript in the original RNA sample was extrapolated from PCR reaction kinetics using Opticon software and normalised to the level of GAPDH transcript.