Eomesodermin, a target gene of Pou4f2, is required for retinal ganglion cell and optic nerve development in the mouse

Among the six neuronal cell types in the vertebrate retina, retinal ganglion cells (RGCs) are the sole output neurons conveying visual information into the brain. RGCs are also the first retinal cell type to differentiate. These features distinguish RGCs from the other retinal cell types and emphasize the importance of elucidating the mechanisms that regulate RGC development in order to create a comprehensive, integrated model of retinogenesis.

In the mouse, differentiating RGCs can first be visualized with molecular markers as early as embryonic day 11.5 (E11.5) when dividing retinal progenitor cells (RPCs) cease mitosis and migrate basally from the neuroblast side of the retinal epithelium towards the vitreal side to the emerging ganglion cell layer (GCL) (Young,1985a; Young,1985b). We and others have investigated the mechanisms controlling RGC fate specification and differentiation, and by doing so, have identified key regulatory factors that are essential for these events to occur normally(for reviews, see Mu and Klein,2004; Mu and Klein,2008; Cayouette et al.,2006). In particular, we have proposed a gene regulatory network model for RGC development featuring a hierarchical cascade of transcription factors that ultimately controls the expression of terminal downstream effector genes (Mu et al.,2005). Two transcription factors occupy key nodes in the RGC gene regulatory network: the proneural bHLH transcription factor Math5 (Atoh7 -Mouse Genome Informatics) and the class IV Pou domain-containing transcription factor Pou4f2 (also called Brn3b). Math5 is responsible for endowing RPCs with the competence to acquire a RGC fate (Brown et al., 2001; Wang et al.,2001), whereas Pou4f2 is positioned genetically downstream of Math5 and is essential for RGC differentiation, cell survival, neurite outgrowth and axon pathfinding (Erkman et al., 1996; Erkman et al.,2000; Gan et al.,1996; Gan et al.,1999; Wang et al.,2000). Notably, RGCs that lack Pou4f2 are still able to differentiate, albeit abnormally, and the expression of many RGC genes is refractory to the absence of Pou4f2 (Mu et al., 2004). This suggests that pathways operate in parallel along with Pou4f2 to control RGC differentiation(Mu et al., 2004; Mu et al., 2005).

For our RGC gene regulatory network model to be meaningful, connections must be made between individual upstream regulatory genes such as Pou4f2 to terminal downstream genes such as those that control axon projections. Unfortunately, with the exception of Pou4f1(Trieu et al., 1999), little is known about regulatory genes that might be direct targets of Pou4f2. We therefore sought to determine whether any genes encoding transcription factors downstream of Pou4f2 are direct Pou4f2 targets, and whether these transcription factors mediate the roles played by Pou4f2 in regulating genes involved in terminal differentiation events. We focused on the eomesodermin(Eomes) gene for the following reasons. Eomes (also referred to as Tbr2) belongs to the Tbr1/Eomes/Tbet subfamily of genes (Naiche et al., 2005)that were initially found to play essential roles during trophoblast and mesoderm development in mice (Russ et al.,2000; Strumpf et al.,2005). Eomes is also expressed in the developing central nervous system and has been implicated in the development of the human central nervous system; a homozygous breakpoint mutation in a Moroccan family silences Eomes and leads to microcephaly(Baala et al., 2007). Recent studies have also indicated that Eomes is a component of a pathway that regulates glutamatergic neurogenesis in the cerebral cortex and cerebellum in mouse development (Bulfone et al.,1999; Hevner et al.,2006; Quinn et al.,2007). Of particular interest from the standpoint of our studies, Eomes was also found to be expressed in the GCL of the retina(http://www.scripps.edu/cb/friedlander/gene_expression/),prompting us to hypothesize that Eomes plays a role in RGC differentiation.

In this study, we extend the initial findings on Eomes in the developing retina by providing evidence that Eomes is a direct target of Pou4f2, and by demonstrating a role for Eomes in RGC differentiation and optic nerve development. We also discovered a novel role for Eomes: that it regulates myelin ensheathment in the developing optic nerve.

A gene targeting vector was constructed containing a floxed Eomesallele in which exon 3 of Eomes could be deleted by Cre-mediated recombination. We used genomic DNA from R1 embryonic stem (ES) cells to PCR amplify 1.63, 0.45 and 2.79 kb fragments from the Eomes locus containing coding exons 1-5 and associated introns (see Fig. 6A). The amplified products were sequentially subcloned into a targeting vector generated previously in our laboratory (C.A.M., unpublished). The final targeting construct contained two loxP sites inserted into the second and third introns and a PGK-neomycin cassette with flanking Frt sites inserted into intron 3 adjacent to the loxP site (see Fig. 6A). The targeting construct was linearized and electroporated into ES cells, after which G418-resistant ES cells were selected to identify homologous recombination events. Targeted ES cell clones were identified by Southern analysis (see Fig. 6B). Chimeric males were bred to wild-type C57/BL/6J females to generate Eomes^+/flox heterozygotes. To delete the Frt-neomycin cassette, we bred Eomes^+/flox mice with mice containing a Rosa26-FLPeR transgene(Farley et al., 2000). Eomes^flox/flox homozygous mice were obtained by interbreeding against the C57BL/6J background. Eomes^flox/flox mice were viable and fertile.

A CMV-Cre transgenic line was used to generate Eomes^+/Δflox in one-cell zygotes (Su et al.,2002). We then interbred Eomes^+/Δflox mice to recover Eomes^{Δflox/Δflox}embryos, which displayed phenotypes identical to those observed in germline-generated Eomes^-/- embryos(Russ et al., 2000). To delete exon 3 in the developing retina, Eomes^+/flox mice were bred to the Six3-Cre transgenic line, which expresses Cre in neural progenitor cells in the retina(Furuta et al., 2000). The resulting Eomes^+/flox;Six3-Cre males were bred to Eomes^flox/flox females to generate Eomes^flox/flox;Six3-Cre embryos and neonates in which exon 3 was specifically deleted in the retina(Eomes^{Δflox/Δflox};Six3-Cre).

Three PCR primers were used simultaneously to distinguish wild-type, floxed and deleted Eomes alleles (see Fig. 6C): Em15,5′-CATGATTAAGGAAGGCTGGATGCAC-3′; Em16,5′-AAGGAGACAGCCTTCACCCAGGC-3′; and Em18,5′-GACTGAGAAGTAAGAGGGTCAGCAG-3′. PCR primers used to genotype the Six3-Cre transgene were Cre01,5′-AACGAGTGATGAGGTTCGCAAGAAC-3′, and Cre02,5′-CGCTATTTTCCATGAGTGAACGAACC-3′.

To generate embryos carrying the Eomes-enhancer transgene, a 2.8 kb region from base pairs -1182 to -3918 (translational start site, +1) of Eomes was subcloned into a plasmid containing the hsp68-lacZ-pA reporter gene(Kothary et al., 1989). The Eomes-enhancer transgene was used to generate transgenic embryos,which were collected at the desired times.

The US Public Health Service Policy on Humane Care and Use of Laboratory Animals was followed in all experiments using mice, and the M. D. Anderson Institutional Animal Care and Use Committee approved all animal protocols used in these experiments.

Embryos and eyes dissected from embryos or animals were fixed,paraffin-embedded and sectioned into 7 μm or 12 μm slices for immunohistochemistry studies or in situ hybridization, respectively(Mu et al., 2004). After de-waxing and rehydration, the sections were stained with Hematoxylin and Eosin for histology studies. In situ hybridization was performed as described by Mu et al. (Mu et al.,2004).

For the immunohistochemistry studies, sections were placed in a microwave oven at 600 watts in 10 mM sodium citrate for 18 minutes to expose the antigen epitopes. Microwave-treated sections were then incubated with primary antibodies. A tyromide signal amplification kit (TSA biotin system,PerkinElmer) was used in conjunction with a NovaRed kit (Vector Labs), to detect Eomes protein expression by color, and the immunostained sections were counterstained with Methyl Green. To detect Eomes expression using indirect immunofluorescence, we used the TSA plus fluorescence kit to optimize the signal intensity. For double-immunofluorescence labeling, Eomes expression was detected as described above and the expression of all other proteins was detected using Alexa-conjugated secondary antibodies (Invitrogen). The primary antibodies were anti-BrdU (Upstate, 1/4), anti-Brn3b/Pou4f2 (Santa Cruz,1/200), anti-GFP (Invitrogen, 1/1000), anti-NFL (Invitrogen, 1/250),anti-syntaxin (Chemicon, 1/1000), anti-Tbr1 (Chemicon, 1/1000) and anti-Tbr2/Eomes (Chemicon, 1/1000). HRP-conjugated secondary antibody for tyromide signal amplification was obtained from Jackson ImmunoResearch. Flat-mount retinas were prepared as described by Xiang et al.(Xiang et al., 1995).

TUNEL assays on embryonic retinas were performed using an in situ cell death detection kit (Roche Applied Science) following the manufacturer's instructions. Three paraffin-embedded sections were used for data analysis. For pulse labeling with BrdU, 100 μg of BrdU (Upstate) per gram of body weight was injected intraperitoneally into pregnant females 1 hour before euthanization. Embryos were then processed as described previously by Fu et al. (Fu et al., 2006).

E15.5 retinas were isolated for chromatin immunoprecipitation (ChIP) assays following the procedures described by Wells and Farnham(Wells and Farnham, 2002) and Kiyama and Klein (Kiyama and Klein,2007). Two microliters of the immunoprecipitated chromatin DNA and input genomic DNA were used for PCR amplification with the following primer sets: Eo3b007, 5′-GACCAACTTGCCACAAAAAACCC-3′ and Eo3b008,5′-CTGAACAGGCTTGCTGCATGCTC-3′; or TTN E4F,5′-TTGCAGCAACCACTCTTGTC-3′ and TTN E4R,5′-GCATGATGGGAGAGGACCTA-3′. PCR amplifications were performed for 30 cycles. The amplified products were resolved by 6% polyacrylamide gel electrophoresis and visualized by staining with GelStar nucleic acid gel stain(CAMBREX).

Electrophoretic mobility shift assays (EMSAs) were performed as described by Mu et al. (Mu et al.,2004). Oligonucleotides for the wild-type and mutated site A (see Fig. 4A) were: wild-type,5′-GGGAGTGCTGGTGACCCGTTAATGTTGGAATGTTTTCCTA-3′ and mutant,5′-GGGAGTGCTGGTGACCCGAAAAAGTTGGAATGTTTTCCTA-3′. Pou4f2 protein was synthesized in vitro using a transcription/translation system (Promega), and 2μl of the transcription/translation reaction mixture was used for each EMSA reaction.

U2OS cells were cultured in 10% Dulbecco's modified Eagle's medium and 10%fetal bovine serum at 37°C in a 5% CO₂ atmosphere. For the Eo2.8k (A^mB^m)-HSP68p-LacZ-pA plasmid, the core Pou4f2 binding site A was changed from 5′-GTTAATGTT-3′ to 5′-GAAAAAGTT-3′ and site B was changed from 5′-ATTAATGAG-3′ to 5′-AAAAATGTG-3′ using the Quikchange site-directed mutagenesis kit (Stratagene). Full-length Pou4f2 cDNA was subcloned into pIRES-hrGFP-1a (Stratagene) for transactivation experiments. Transfections were carried out on cover slides placed in six-well culture plates using FuGene HD (Roche) following the manufacturer's instructions. Cells were co-transfected with 0.5 μg of the Pou4f2expression construct or the empty expression vector and 0.5 μg of the reporter plasmid. X-gal staining was performed 36 hours after transfection. Transfection efficiency was determined by visually inspecting the number of GFP-expressing cells for each experiment. All experiments were performed in triplicate.

We determined the spatial and temporal pattern of Eomestranscripts in the retina at E14.5 and E16.5 using sectioned in situ hybridization. At E14.5, the first signs of Eomes expression in the retina were mainly in the GCL, although cells within the neuroblast layer(NBL) were sparely labeled (Fig. 1A). At E16.5, Eomes continued to be expressed in the GCL and NBL (Fig. 1B). We also found Eomes to be strongly expressed in the forebrain, a previously known site of Eomes expression(Bulfone et al., 1999)(Fig. 1B). To determine whether Eomes is expressed at stages earlier than E14.5, we performed immunohistochemistry with anti-Eomes antibody in sections obtained at E13.5. Although Eomes protein was readily detected in the forebrain at this stage, it was not detected in the retina (Fig. 1C).

Because the GCL consists of displaced amacrine cells as well as RGCs, it was possible that Eomes was expressed in either both or one of these neuronal cells types. To test these possibilities, we performed immunohistochemistry in sections of wild-type and Math5^-/-retinas obtained at E16.5. Retinas of Math5^-/- embryos lack RGCs and in their place are increased numbers of displaced amacrine cells(Wang et al., 2001). Although wild-type retinas strongly expressed Eomes in the GCL, Eomesexpression in Math5^-/- retinas was virtually undetectable(Fig. 1D,E). These results demonstrated that RGCs in the GCL are the major site of Eomesexpression from E14.5 to E16.5, a time when newly forming RGCs are populating the GCL.

Pou4f2 is one of the earliest markers of RGC differentiation(Gan et al., 1999) and is expressed in newly forming RGCs within the NBL and in differentiated RGCs that have migrated to the GCL (Gan et al.,1999). By comparing Pou4f2 and Eomes protein expression, we determined how much the expression of the two transcription factors overlapped and their potential relationship to each other. Although Eomes expression was not detected in retinas at E13.5 (Fig. 1C), Pou4f2 was already expressed at high levels at this time(Gan et al., 1999). By E14.5,Eomes expression was apparent and coincided with the existence of a subset of Pou4f2-expressing cells within the GCL(Fig. 1F-H). However, many Pou4f2-expressing cells, particularly those still migrating to the GCL, did not express Eomes. In addition, although Pou4f2 was expressed in RGCs throughout the GCL, Eomes was not expressed in the peripheral-most region(Fig. 1H). This would be in keeping with the fact that neuronal differentiation begins at the center and advances to the periphery of the retina. At E16.5, expression of Pou4f2 was largely confined to the GCL and many, but not all, of the Pou4f2-expressing cells also expressed Eomes (Fig. 1I-K). These results suggested that RGCs increasingly express Eomes as they migrate to the GCL and undergo terminal differentiation.

Weaker expression of Eomes within the NBL at E14.5 and E16.5 did not overlap with the expression of Pou4f2 and persisted throughout retinogenesis. As cells within the NBL are a mixture of mitotically active RPCs and newly formed postmitotic cells that have left the cell cycle and committed to a retinal cell fate (Le et al.,2006), we determined whether Eomes was expressed in mitotic cells by pulse-labeling S-phase cells of the E18.0 retinas with BrdU and comparing the relative distributions of the S-phase cells and Eomes-expressing cells. We detected little, if any, overlap between these two cell populations(Fig. 2A-C), suggesting that Eomes-expressing cells are mainly postmitotic.

We also observed that retinal Eomes expression persisted into postnatal life. In particular, Eomes was observed to be strongly expressed in the GCL and in the innermost region of the emerging inner nuclear layer (INL)of newborn mice (P0) (Fig. 2D). Weaker expression was observed throughout the NBL(Fig. 2D). However, Eomes expression was largely absent in the GCL but persisted in the INL and NBL of retinas of newly born Math5^-/- mice(Fig. 2E). Because amacrine cells are the major cell type in the innermost region of the INL, it is possible that the Eomes-expressing cells were amacrine cells and that the Eomes-expressing cells in the NBL were amacrine cell progenitors.

At P12, Eomes expression was detected in a subpopulation of cells in the GCL, where its expression co-localized with that of the RGC marker neurofilament light subunit (NFL) (Fig. 2F). At P30, Eomes expression in the INL co-localized with that of the pan-amacrine cell marker syntaxin (Fig. 2G) but not of other amacrine cell markers calbindin and parvalbumin (data not shown). A few Eomes-expressing cells in the INL were found to be colocalized with the amacrine marker calretinin (data not shown). These results suggested that Eomes-expressing cells in the INL constituted a distinct subpopulation of amacrine cells.

Because Pou4f2 is expressed in newly arising RGCs and is required for RGC differentiation, Eomes might be positioned genetically downstream of Pou4f2, making it a potential target gene of Pou4f2. If this were the case, Eomes expression would necessarily depend on the presence of functional Pou4f2. To test this possibility, we determined the level of Eomes protein expression in retinas isolated from Pou4f2^+/-(wild type) and Pou4f2^-/- embryos. We performed the analysis at E14.5 because Pou4f2^-/- retinas have the same number of RGCs as wild-type retinas at that stage. Eomes expression in the GCL strictly depended on Pou4f2; Eomes protein in the GCL was undetectable in Pou4f2^-/- retinas but clearly detectable in Pou4f2^+/- control littermates(Fig. 3A,B).

These results indicated that Eomes is genetically downstream of Pou4f2 but they did not show whether Pou4f2 is directly involved in regulating Eomes transcription. We therefore determined whether conserved noncoding sequences containing Pou4f2 consensus DNA-binding sites[5′-(A/G)TTAATGAG(C/T)-3′; Xiang et al.(Xiang et al., 1995)] could be identified within the Eomes locus. We focused our search on a 20 kb stretch of genomic DNA that encompasses the Eomes gene along with 9.5 kb of DNA upstream of the Eomes translational initiation codon and 4.1 kb of DNA downstream of the Eomes translational termination codon. By comparing human, mouse, rat and dog genomes, we identified several blocks of conserved noncoding sequences that were located within the upstream and downstream DNA sequences flanking Eomes(Fig. 3C). Furthermore, two putative Pou4f2-binding sites were identified in two closely separated regions. One site (site B, Fig. 3C,D) began at base pair -2937 (5′-ATTAATGAG-3′) and its sequence precisely matched that of the consensus Pou4f2-binding site. The other site (site A, Fig. 3C,D)began at base pair -3854 (5′-GTTAATGTT-3′) and differed from another Pou4f2-binding site by one base pair. However, Pou4f2-binding site B was not conserved in the corresponding position in human, rat and dog genomes(Fig. 3D). By contrast,Pou4f2-binding site A was located within a highly conserved 192 bp region(Fig. 3D), and this site was present in all four vertebrate genomes with the exception of one mismatched nucleotide in the dog genome.

To determine whether the conserved upstream regions containing the Pou4f2-binding sites were capable of directing expression in the developing retina, we constructed a transgene containing a 2.8 kb fragment that included both Pou4f2-binding sites fused to an hsp68-promoter-lacZ-pA reporter gene (Fig. 3E). The resultant transgenic E14.5 embryos expressed lacZ in retinas and limbs, a previously known region of Eomes expression(Russ et al., 2000)(Fig. 3E). In control experiments, a transgene construct with disabling mutations in sites A and B did not express in either the limb or retina (data not shown). In the wild-type construct, lacZ was expressed in a narrow strip within the inner part of the retina just above the GCL(Fig. 3F).β-Galactosidase-stained retinal sections co-stained with anti-Eomes antibody, showed that most of the lacZ-expressing cells did not co-express the endogenous Eomes protein (compare Fig. 3A with 3F). A possible explanation for the lack of overlap is that the lacZ-expressing cells represent newly forming RGCs in the process of migrating to the GCL. To determine whether this was the case, we performed immunohistochemistry analysis ofβ-galactosidase-stained retinal sections from E14.5 transgenic embryos using anti-Brn3b (Pou4f2) antibody. From these experiments, it was clear that most of the lacZ-expressing cells also expressed Pou4f2, although Pou4f2 expression was present in many cells that were not expressing the transgene (Fig. 3G). These results demonstrated that the 2.8 kb fragment is sufficient to direct transgene expression in a subpopulation of RGCs and that this subpopulation also expresses Pou4f2. However, the transgene appeared to be expressed at earlier stages in RGC differentiation than the stages when endogenous Eomes is expressed. The reason why transgene and Eomes expression do not correlate is uncertain, but it suggests that crucial cis-regulatory elements are missing in the 2.8 kb region. Consistent with this, a 225 kb Bac transgenic mouse line containing Eomes fused to an eGFP reporter (Bac Eomes::GFP) (Kwon and Hadjantonakis,2007) expresses GFP in the GCL of E14.5 and E16.5 retinas in a manner similar to that of endogenous Eomes expression(Fig. 3H-K).

To determine whether either Pou4f2-binding sites A or B within the 2.8 kb fragment bind to Pou4f2, we used EMSA to detect Pou4f2 binding in vitro and ChIP to detect Pou4f2 binding in vivo. EMSA was performed using an in vitro-synthesized full-length Pou4f2 protein and an oligonucleotide probe containing the putative Pou4f2-binding site A within the conserved 192 bp region. We detected a specific Pou4f2-DNA complex but no complex was formed with transcription-translation lysates that lacked the Pou4f2 template(Fig. 4A). The formation of the Pou4f2-DNA complex was inhibited when excess homologous unlabeled oligonucleotide competitor but not when excess oligonucleotide competitor containing a mutated Pou4f2 site was added. Addition of the anti-Pou4f2/Brn3b antibody also blocked complex formation(Fig. 4A). Notably, an oligonucleotide competitor containing the Pou4f2-binding site B did not interfere with complex formation, suggesting that this nonconserved sequence was not a bona fide Pou4f2-binding site.

ChIP analysis was performed using chromatin prepared from E15.5 retinas and primer pairs that would reveal Pou4f2 bound to the Pou4f2-binding site within the conserved 192 bp region. This showed that a 122 bp fragment was amplified after immunoprecipitation with anti-Brn3b antibody(Fig. 4B), but not with preimmune IgG, which served as a negative control (data not shown). In addition, sequences derived from the gene encoding the muscle-specific protein Titin were not immunoprecipitated with anti-Brn3b antibody(Fig. 4B). We also performed ChIP experiments using anti-acetylated-histone H3 antibody to determine whether the Pou4f2-binding site was within a region of open (acetylated)chromatin, a finding that would indicate active expression. The 122 bp fragment was also amplified after immunoprecipitation with anti-acetylated-histone H3 antibody, whereas sequences from the gene encoding Titin were not (Fig. 4B). Both the Pou4f2-binding site and Titin sequences were amplified after immunoprecipitation with anti-histone H3 antibody, which ubiquitously marks the presence of nucleosomes (Fig. 4B). These results implicate Pou4f2 as a direct transcription regulator of Eomes expression acting at the Pou4f2-biding site A within the conserved 192 bp region. A similar ChIP analysis was performed using the nonconserved Pou4f2-binding site B. A very weak PCR product was observed when primer pairs were used that amplified a 157 bp fragment containing the putative Pou4f2 site B (data not shown). This result was consistent with the view that the nonconserved Pou4f2 site B is likely to have only a minor function, if any, in vivo.

To provide further evidence that Pou4f2 is a transcriptional regulator of Eomes, we performed transient transactivation assays in U2OS cells using the 2.8 kb fragment-lacZ reporter gene construct containing Pou4f2-binding sites A and B (see Fig. 3E). A control construct, Eo2.8K(A^mB^m)-HSP68p-LacZ-pA, was generated in which binding sites A and B were mutated. The number of cells expressing lacZ was sharply increased when the wild-type construct was co-transfected with the Pou4f2 expression plasmid (Fig. 5A,B), with little increase in the number of cells with the mutated construct (Fig. 5C,D). These results demonstrate that Pou4f2 is able to activate transcription from the 2.8 kb fragment and suggest that Pou4f2-binding to site A, and to a lesser extent site B, enhances Eomes transcription in RGCs.

To assess the role of Eomes in the developing retina, we generated mice with a floxed allele of Eomes and bred them with a Six3-Cre transgenic mouse line to delete exon 3 of Eomes in the developing retina (Fig. 6). Histological retinal sections of Eomes^{Δflox/Δflox}mice at P24 showed a slight reduction in the thickness of the GCL compared with Eomes^+/flox controls, whereas the INL and ONL were normal (Fig. 7A,B). This suggests that in the absence of Eomes, the RGCs are reduced in number and that they possess other defects as well.

To visualize the RGC axons, we prepared flat mounts of retinas that were immunohistochemically stained with anti-NFL antibody. In Eomes^{Δflox/Δflox}P30 retinas, we observed ∼30% reduction in the number of RGC axons in both the central and peripheral regions of the retina when compared with Eomes^+/flox retinas(Fig. 7C-F). The density of the axon bundles was also notably reduced, and their orientation towards the optic disk appeared abnormal (arrowheads, Fig. 7F). The reduction in the number of RGCs and the defects associated with RGC axons associated with Eomes^{Δflox/Δflox}retinas were similar to those identified in Pou4f2^-/-retinas, although the overall effects appeared less severe in Eomes^{Δflox/Δflox}retinas.

As RGCs undergo enhanced apoptosis between E15.5 and E18.5 in Pou4f2^-/- retinas (Gan et al., 1999), we performed a TUNEL analysis of sectioned retinas from Eomes^+/flox and Eomes^{Δflox/Δflox}embryos and postnatal mice to look for differences in apoptosis patterns between the two. Although there was no significant enhancement of apoptosis through E16.5 (data not shown), at E18.5 we detected a dramatic increase in apoptotic cells in Eomes^{Δflox/Δflox}retinas in both the GCL and NBL with respect to Eomes^+/flox retinas (arrowheads, Fig. 7G-J). At P1, P6 and P12,the number of apoptotic cells was consistently higher in Eomes^{Δflox/Δflox}retinas than in Eomes^+/flox controls, but the differences were less dramatic than they were at E18.(Fig. 7I,J). Apoptotic cells were also observed in the GCL, INL and ONL. We did not observe significant differences between Eomes^{Δflox/Δflox}retinas and Eomes^+/flox controls after P12. These results suggest that the reduced number of RGCs that we observed in the retinas of P30 Eomes^{Δflox/Δflox}mice resulted from the enhanced apoptosis of RGCs within the GCL between E18 and P12. However, the reason for increased cell death in the retinal NBL, INL and ONL of embryos and postnatal Eomes^{Δflox/Δflox}mice is uncertain. Some of the apoptotic cells might represent the Eomes-positive amacrine cells and their progenitors that weakly expressed Eomes in the NBL at earlier stages.

As the expression of Eomes in RGCs correlated with the time when RGC axons reach the optic chiasm, this raised the possibility that Eomes might play a role in regulating some aspects of RGC axon formation. To assess this possibility, we isolated optic nerves from adult Eomes^{Δflox/Δflox}and Eomes^+/flox mice, and examined their ultrastructure(Wang et al., 2000). In a section about 0.5 cm from the optic disk, a region where mature RGC axons are well myelinated and ensheathed, low-magnification TEM showed that the area of Eomes^{Δflox/Δflox}optic nerves was 30% smaller than that of Eomes^+/floxoptic nerves (Fig. 8A,B),consistent with the finding that RGC numbers were reduced by ∼30%. At higher magnifications, we observed striking differences between Eomes^{Δflox/Δflox}and Eomes^+/flox optic nerves(Fig. 8C,D). In particular, the axons in the optic nerves of Eomes^+/flox mice were ensheathed with a well-organized, compact myelinated layer. However, only a few axons in the optic nerves of Eomes^{Δflox/Δflox}mice were myelinated and where present, myelin ensheathment was thinner,disorganized and loosely packed (Fig. 8C, parts iii-vi, Fig. 8D, parts iii-vi). Moreover, in the optic nerves of Eomes^{Δflox/Δflox}mice, we observed neurites that were smaller in diameter than those of normal axons and that contained large numbers of microtubules rather than the neurofilaments normally associated with RGC axons in the optic nerve(arrowheads, Fig. 8D, part vi). This finding was highly reminiscent of the abnormalities we observed in the optic nerves of Pou4f2^-/- mice reported earlier(Wang et al., 2000). Because we had not in that study determined whether myelination was abnormal in Pou4f2^-/- optic nerves, we performed a similar TEM analysis in Pou4f2^-/- mice. We found, indeed, that there were similar defects in myelination and ensheathment in Pou4f2^-/- mice (data not shown), providing additional evidence that Eomes mediates the functions of Pou4f2 in RGC and optic nerve development.

In this study, we found that Eomes is expressed in differentiated RGCs within the GCL but not in committed progenitors or newly forming RGCs within the NBL. We further observed that Eomes expression begins when Pou4f2 expression is maximal and depends on Pou4f2. Our results therefore provide evidence that Eomes is a direct target gene of Pou4f2. We also found that a 2.8 kb fragment upstream of Eomes, which encompasses a highly conserved 192 bp region that contains a Pou4f2-binding site, directs RGC expression in transgenic embryos. Furthermore, Pou4f2 binds efficiently to and activates transcription from this site. Analysis of Eomes^{Δflox/Δflox}retinas indicated that Eomes plays roles similar to but more restricted than that of Pou4f2 in RGC differentiation, cell survival and optic nerve development. Overall, our results argue strongly for our hypothesis that Eomes is positioned immediately downstream of Pou4f2 and is a mediator, at least in part, of Pou4f2 function. Eomes, therefore, appears to be an important transcription factor that connects Pou4f2 to the expression of genes associated with terminal differentiation, particularly those involved in axon outgrowth and optic nerve development.

When fused to a minimal promoter and a lacZ reporter transgene, a 2.8 kb fragment upstream of Eomes that contains several conserved noncoding sequence blocks directs reporter gene expression to a subpopulation of cells within the NBL that co-express Pou4f2. The fact that the cells expressing the transgene are in the NBL and that they also express Pou4f2 strongly suggests that the cells are newly forming RGCs in the process of migrating from the NBL to the GCL. As transgene expression in the retina is highly restricted to these Pou4f2-expressing cells, this implies that Eomes expression is dependent on the presence of Pou4f2. However, endogenously expressed Eomes is clearly found in RGCs that also express Pou4f2 but have already migrated to the GCL. Thus, the 2.8 kb fragment does not faithfully reproduce the endogenous expression pattern of Eomes. Nevertheless, retinal expression of the transgene is highly restricted to RGCs. A possible explanation for this is that additional cis-regulatory elements lying outside the 2.8 kb fragment modulate the spatiotemporal activity of Eomes. Indeed, a 225 kb Bac Eomes-eGFP transgene appears to recapitulate endogenous Eomes expression, suggesting that Eomes is regulated by multiple enhancer inputs, one of which is within the 2.8 kb region. In addition, non-deleted Eomes transcripts are downregulated in Eomes^{Δflox/Δflox}retinas, suggesting that Eomes regulates its own expression (C.-A.M. and W.H.K., unpublished).

Major defects were observed in the retinas of Eomes^{Δflox/Δflox}mice. This included reduced numbers of RGC, abnormal RGC axon outgrowth and an aberrant and disorganized myelin sheath surrounding RGC axons. The massive cell death in the GCL associated with the loss of Eomes indicates that Eomes is required for the survival of at least some RGCs. Presumably, the absence of Eomes results in abnormal differentiation, which in turn triggers apoptosis. The increased number of apoptotic cells in other retinal layers might be associated with the weak expression of Eomes observed in non-RGCs or might be an indirect consequence of abnormal levels of apoptotic cells within the GCL.

One of the striking retinal defects in Eomes^{Δflox/Δflox}embryos is the disorganized and loosely packed myelin sheaths surrounding RGC axons in the optic nerve. Axon ensheathment is essential for the normal propagation of membrane potential, and when defective, results in slow and continuous depolarization along the axon(Waxman et al., 1995). It is therefore likely that the propagation of visual signals into the brain is severely compromised in Eomes^{Δflox/Δflox}mice. Because axon outgrowth and myelination require changes in the cell adhesion properties of axons (Bartsch,2003), the abnormal ensheathment possibly has its source in defective cell adhesion. Many members of the T-box transcription factor family, including Eomes, are involved in regulating genes whose products are important for cell adhesion (Strumpf et al., 2005; Inman and Downs,2006). Eomes acting downstream of Pou4f2 might therefore regulate the genes whose products are required for substrate adhesion.

Several members of the T-box transcription factor family besides Eomes are also expressed in retinal development. The vertebrate orthologs of the Drosophila gene optomotor-blind, Tbx2, Tbx3 and Tbx5, are expressed early in the dorsal part of the developing retina and are thought to function in dorsoventral axial patterning(Pflugfelder et al., 1992; Sowden et al., 2001). In addition, Tbx12 and Tbx20 are expressed in the developing retina and Tbx12 is required for the formation of retinal cells and the organization of retinal layers (Carson et al., 2000; Carson et al.,2004; Kraus et al.,2001).

A conserved DNA-binding T-domain is found in all T-box transcription factors (Kispert and Herrmann,1993; Bruneau et al.,2001; Conlon et al.,2001; Paxton et al.,2002; Wang et al.,2004) and a consensus T-box element is believed to be the binding site for all members of the T-box transcription factor family. Thus, it is possible that other T-box factors partially compensate for the loss of Eomes. To examine whether the absence of Eomes resulted in an upregulation of other T-box genes, which would be indicative of compensation, we determined the expression patterns of eight T-box genes in Eomes^+/floxand Eomes^{Δflox/Δflox}retinas. We were able to detect retinal expression of only Trb1 and Tbx5, and no significant differences in expression were observed between Eomes^+/flox and Eomes^{Δflox/Δflox}retinas (see Fig. S1 in the supplementary material). It therefore seems likely that Eomes has evolved a specialized function in RGC development and is positioned in the RGC gene regulatory hierarchy downstream of Math5 and Pou4f2.

We thank Jan Parker-Thornburg, Chad Smith, the Genetically Engineered Mouse Facility, Kenneth Dunner, Jr, members of the High Resolution Electron Microscopy Facility and Jody Polan for advice and guidance on the confocal microscopy; Mathew Chempanal for technical support; and Xiuqian Mu for valuable input on the manuscript. These facilities, the DNA Analysis facility and the Research Animal Support Facility are supported in part by a National Institutes of Health-National Cancer Institute Comprehensive Cancer Center Support grant (CA16672). The work was supported by grants from the National Institutes of Health-National Eye Institute to W.H.K. (EY11930 and EY010608-139005) and Y.F. (EY013128), and by the National Institute of Child Health and Human Development to K.H. (HD052115). We also acknowledge support from the Robert A. Welch Foundation (G-0010) to W.H.K.