Transient Expression of Fez Family Zinc Finger 2 Protein Regulates the Brn3b Gene in Developing Retinal Ganglion Cells*

Retinal ganglion cells (RGCs) are projection neurons in the neural retina that relay visual information from the environment to the central nervous system. The early expression of MATH5 endows the post-mitotic precursors with RGC competence and leads to the activation of Brn3b that marks committed RGCs. Nevertheless, this fate commitment process and, specifically, regulation of Brn3b remain elusive. To explore the molecular mechanisms underlying RGC generation in the mouse retina, we analyzed the expression and function of Fez family zinc finger 2 (FEZF2), a transcription factor critical for the development of projection neurons in the cerebral cortex. Fezf2 mRNA and protein were transiently expressed at embryonic day 16.5 in the inner neuroblast layer and the prospective ganglion cell layer of the retina, respectively. Knockout of Fezf2 in the developing retina reduced BRN3B+ cells and increased apoptotic cell markers. Fezf2 knockdown by retinal in utero electroporation diminished BRN3B but not the coexpressed ISLET1 and BRN3A, indicating that the BRN3B decrease was the cause, not the result, of the overall reduction of BRN3B+ RGCs in the Fezf2 knockout retina. Moreover, the mRNA and promoter activity of Brn3b were increased in vitro by FEZF2, which bound to a 5′ regulatory fragment in the Brn3b genomic locus. These results indicate that transient expression of Fezf2 in the retina modulates the transcription of Brn3b and the survival of RGCs. This study improves our understanding of the transcriptional cascade required for the specification of RGCs and provides novel insights into the molecular basis of retinal development.


Retinal ganglion cells (RGCs) are projection neurons in the neural retina that relay visual information from the environ-
Retinal ganglion cells (RGCs) 3 are generated from a pool of retinal progenitor cells. In all vertebrates examined, RGCs are the first born neurons in the retina. In mice, multipotent retinal progenitor cells begin to exit mitosis and differentiate into RGC-competent precursors around embryonic day 11 (E11). RGC competence is fundamentally a cell-intrinsic process that is activated by the early expression of basic helix-loop-helix transcription factor MATH5, targeted deletion of which blocks the formation of most RGCs (1). MATH5 is expressed during or after the terminal mitosis of retinal progenitor cells (2), leading to a regulatory cascade for RGC development. MATH5 induces the expression of class IV POU domain transcription factor Pou4f2 (also known as BRN3B), which is essential for the survival of differentiated ganglion cells (3,4). Remarkably, when Brn3b is selectively deleted, a significant portion of RGCs dies by enhanced apoptosis (5,6). Moreover, BRN3B acts as a safeguard mechanism by repressing cell differentiation of amacrine and horizontal cells during the fate commitment of RGCs (7). Although BRN3B marks committed RGCs, it regulates RGC maturation together with LIM homeodomain transcription factor ISLET1, which co-expresses with BRN3B in post-mitotic differentiating RGCs (8 -10).
MATH5 is critical for the competence and, hence, formation of RGCs (1,2,11). The regulation of fate commitment of RGCs from the competent cells is unclear because MATH5ϩ progenitors have the potential to give rise to all seven major retinal cell types (12). In addition, the temporal expression patterns of Math5 and Brn3b mRNA are not identical. Brn3b shows a progressive increase starting from E12.5 and persists at E16.5, when Math5 starts to decline (2,5,13). A recent lineage tracing study of Math5-expressing cells in the retina also argues that not all RGCs descend from MATH5ϩ progenitors (14). Therefore, the precise transcriptional cascade leading to the generation of RGCs remains to be explored.
RGCs transmit visual information from the environment to the central nervous system and are the sole projection neurons in the retina. The projection neurons in the cerebral cortex have been extensively studied by in vivo genetic manipulations. It was found that transcription factor FEZF2 (also known as FEZL and ZFP312) was expressed by early neural progenitors in the ventricular zone of the dorsal telencephalon. Zinc finger protein FEZF2 represses the expression of Hes5 in proliferating neural stem cells (15,16) and regulates the differentiation of telencephalic precursors from mouse embryonic stem cells (17). Fezf2 is required for the proper specification and differentiation of subcerebral projection neurons in the deep layers of the cerebral cortex (18 -20). Overall, these observations led us to hypothesize that Fezf2 may regulate the specification of retinal projection neuron RGCs.
In this study, we examined the expression and function of Fezf2 in developing retinas. We found that Fezf2 was transiently expressed in retinas during the generation of RGCs. To investigate the role of Fezf2, we inhibited its expression by RNA interference and deleted the gene by conditional Fezf2 knockout in the retina. Our results showed that knockdown and knockout of Fezf2 led to BRN3B inhibition and impaired the formation of RGCs. Moreover, FEZF2 regulated the transcription of Brn3b by binding to a 5Ј regulatory sequence of the Brn3b locus.

Experimental Procedures
Animals and Genotyping-Animal procedures were performed according to the protocol approved by the Animal Care Committee of Wenzhou Medical University. Tg (Fezf2-EGFP) mice were created by the GENSAT project and obtained from the Mutant Mouse Regional Resource Center. Fezf2 flox/flox mice were provided by Nenad Sestan (Yale University). Exon 2, which encodes amino acids 1-279 of 455 in FEZF2, was targeted for removal (21). The Tg(Six3-cre)69Frty mouse was created by Yasuhide Furuta (RIKEN Center for Developmental Biology) (22) and was initially obtained from Lin Gan (University of Rochester Medical Center). Dr. Furuta granted permission for its use by our group. The Fezf2 second exon was deleted in retinas by crossing Fezf2 flox/flox and Six3-Cre mice. The retina-specific knockout of Fezf2 was identified by PCR genotyping using three primers located in exon 1-3 of Fezf2 as described before (21). The loss of Fezf2 expression was further confirmed by quantitative PCR (q-PCR) or Western blotting in one eye, whereas the counterpart eye from the same embryo was taken for the knockout study.
Tissue Preparation-Mouse pups were put on ice to induce hypothermia while pregnant dams were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg). Eyes from prenatal mice were enucleated under a dissecting microscope and fixed by immersion in 4% paraformaldehyde. All tissues were cryoprotected in 30% sucrose at 4°C, frozen by liquid nitrogen vapor, and then embedded in OCT compound (Sakura Tissue Tek, Torrance, CA). Cryosections of 16 m were mounted on Superfrost microscope slides (Fisher Scientific) for in situ hybridization and immunofluorescence (IF) staining.
Plasmid Constructs-For in vivo inhibition of Fezf2 expression, we used Zfp312 siRNA-1 and Zfp312 siRNA-2 with targeting sequences GGAGAACTCGGCCTTGACA at 775-793 and GGTGTTCAATGCTCACTAT at 886 -904 of the coding region of Fezf2, respectively. The scrambled scRNA-1 and scRNA-2 were created by introducing three mutations in the target sequences. The Zfp312 siRNA-1 and its scramble were cloned in the lentiviral vector pLVTH, whereas Zfp312 siRNA-2 and its scramble were created on the basis of pCGLH. All constructs were verified in a previous study of Fezf2 functions in the cerebral cortex (18). For Fezf2 expression, full-length Fezf2 was cloned into pCAGEN (Addgene), which harbors a CAG promoter. For ChIP, a V5 tag (GKPIPNPLLGLDST) was inserted into Fezf2-pCAGEN and pCAGGS-GFP following the startup codon ATG of Fezf2 and GFP, respectively, by PCR in-frame cloning.
To measure the promoter activity in the 5Ј regulatory region of Brn3b, a conservative 1-kb fragment upstream from the starting site of Brn3b was amplified by PCR using primers tcgctttcttccttctctgaa and cggcaataatctaataatagctgtcaa. The PCR product was cloned into pGL4.18, a promoterless firefly luciferase vector co-expressing a neomycin resistance gene (Promega). Mouse neuroblastoma N2a cells were transfected with the pGL4.18 promoter and cultured in the presence of G418 for 2 months to select cells stably expressing the plasmid. All constructs were confirmed by sequencing prior to the in vivo and in vitro experiments.
Retinal in Utero Electroporation-Plasmids were delivered to the developing retina by retinal in utero electroporation (23).
Briefly, E13.5 pregnant mice were anesthetized and attached to a clean table. The abdominal cavity was opened to expose the uterine horns. DNA solution (4 g/l in Tris-EDTA buffer containing 0.1% fast green) were injected through the uterus wall using a polished pulled glass micropipette. The head of each embryo was placed between tweezer-type electrodes, and five square electric pulses (38 V, 50 ms) were passed at 1-s intervals using a BTX electroporator. The wall and skin of the abdominal cavity were aligned and sutured. The embryos were allowed to develop in vivo until E17.5 or postnatal day 7 (P7). Then the mice were sacrificed to remove the brains together with the eyes for fixation in paraformaldehyde.
FACS-To study the effect of Fezf2, we isolated E17.5 retinas from Fezf2-siRNA-transfected mice. Retinas were peeled from the eyeballs and collected under a stereomicroscope (Zeiss). The tissues were minced and incubated with 0.05% trypsin for 8 min at 37°C. Single cell suspensions were prepared by trituration, suspension, and filtration. GFPϩ cells were separated by FACS using a FACSAria II cell sorter (BD Biosciences) and were taken for measuring gene expression by q-PCR.
Real-time RT-PCR-To quantitate the gene expression of Fezf2 and Brn3b, we synthesized cDNA from the RNA isolated from the retina (SuperScript TM first-strand synthesis system, Invitrogen). q-PCR was performed using an ABI 7900 system (Applied Biosystems) and monitored on the basis of SYBR Green I dye detection with the primer sets for Fezf2 (accatggtctccagggaag and tcggaacgcatctccttg) and Brn3b (tcgctttcttccttctctgaa and cggcaataatctaataatagctgtcaa). The primer sets are unique to the targeted genes and do not map to any other genes in the mouse genome according to a BLAST search. Gene expression levels were normalized to the reference gene Actin. The Ct value was defined as the number of PCR cycles required for the fluorescent signal to grow beyond a threshold.
Chromatin immunoprecipitation-ChIP was performed by following the protocol of the EZ ChIP kit (Millipore, 17-371). N2a cells were transfected with pCAG-V5-Fezf2 and pCAG-V5-GFP. The cells were cross-linked with 1% paraformaldehyde and lysed on ice. The lysates were sonicated to shear the chromatin and incubated overnight at 4°C with anti-V5 antibody bound to Dynal protein A/G beads (1:10, Abcam, ChIP grade ab9116). The precipitate from V5-FEZF2 and V5-GFP in Dynabeads was extensively washed and eluted at 65°C for 15 min with elution buffer (1% SDS and 10% 1 M NaHCO 3 in H 2 O). The eluted protein-chromatin complexes were reverse-crosslinked by incubation with 5 M NaCl overnight at 65°C. Both ChIPed and input DNA were treated with RNaseA and proteinase K, followed by DNA purification.
PCR and q-PCR of ChIP DNA-To determine whether FEZF2 binds to Brn3b genomic DNA, we amplified a tentative regulatory region of Brn3b (Ϫ315ϳϩ310) (chr8, 78436342-78436967) by PCR with a primer set (GTGCAGGCTGCCTG-CGTGA and CCGAAGCACCGACCTGGCTAG). The input DNA from the transfected cells was taken as a positive control for gel analysis of ChIP DNA. In a similar manner, the region covering FEZF2 binding motifs (CAGCAACC) (Ϫ230ϳϪ137) was examined by q-PCR with primers (GCAGGCTCCGGT-CCTT and ACAGCAGAGCGCCTGC). A part of the 3Ј UTR of Brn3b (1507-1595) was amplified as a negative control by primers (ACCGGAATTGTTTTCTGATAGC and CTTTCC-CTTCGCCCTCTTT).
Data Analysis-For all q-PCRs and the luciferase assay, data were presented as mean Ϯ S.D. from three independent experiments with duplication of reading. For cell counting from immunostaining, mean Ϯ S.D. were calculated from 12 sections on the basis of three to four samples (eyes). The control and experimental groups were compared by independent samples t test using SPSS 17.0 software (SPSS Inc.) (*, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.0005).

Results
Transient Expression of Fezf2 in the Developing Retina-Retinal cells are generated consecutively from neuroepithelium in the optic cup starting from E11.5 and continuing into the early postnatal period (24). RGCs are the first born retinal cells generated from mid-to late gestation (24,25). We measured the expression of Fezf2 in the retina by RNA in situ hybridization. Fezf2 signals were not detected in the E12.5 retina but appeared by E14.5 and E16.5, when massive RGC production occurred (25). Expression of Fezf2 was diminished by P1 and disappeared at P7 (Fig. 1A), when RGC generation was completed. Fezf2 mRNA was expressed in the inner neuroblast layer (NBL) of the retina but not detected clearly in the outer neuroblast layer, a presumptive ganglion cell layer (GCL) (Fig. 1B). Compared with the neocortex, the retina had a relatively weaker expression of Fezf2 (Fig. 1C). The temporal expression pattern of Fezf2 suggests that this transcription factor may be involved in regulating RGC development.
Although Fezf2 mRNA was hardly detected in the GCL of the developing retina when assayed by in situ hybridization, the FEZF2 protein was found in the GCL at E16.5, as revealed by IF ( Fig. 2A). FEZF2 was co-localized with BRN3B (Fig. 2B), a marker protein of RGCs, indicating FEZF2 expression in the developing RGCs. The discrepancy in the location of Fezf2 mRNA and protein is likely caused by the transient expression Fezf2 Knockout Reduces RGCs and Increases Apoptotic Cells-To assess the role of Fezf2 in RGC development, we analyzed Fezf2-null retinas generated by crossing Fezf2 flox/flox and Six-3-Cre mice. Fezf2 deletion was confirmed by the presence of the inserted loxp and absence of the Fezf2 second exon, as determined by genotyping PCR (Fig. 3, A and B). As expected, expression of Fezf2 decreased in the isolated null retina compared with the control on E17.5 (Fig. 3C). Knocking out Fezf2 attenuated Brn3b mRNA levels (Fig. 3C) and reduced BRN3B protein in the isolated retina (Fig. 3D). The number of BRN3Bϩ cells was reduced in the retinal sections from knockout mice (Fig. 3E). The wild-type retina had 42.6 Ϯ 3.0 BRN3Bϩ cells (per field, ϫ40) in the GCL, which decreased to 23.8 Ϯ 2.1 when Fezf2 was deleted (Fig. 3F). We also noticed an almost complete reduction of BRN3Bϩ cells in the NBL (Fig. 3, E and F). A few BRN3Bϩ cells scattered in the NBL of wild-type mice represent the newly generated ganglion cells en route to the GCL.
Previous studies have established that Brn3b is required for preventing nascent RGCs from apoptotic cell death during retinal development (5,26). Therefore, we investigated apoptotic cells in control and Fezf2-null retinas by the activation of CASPASE III. The cells with activated CASPASE III were found in the GCL of the Fezf2 Ϫ/Ϫ retina (Fig. 3E). The Fezf2-deleted retina had 2.5-fold more CASPASE IIIϩ cells than the wild type at E17.5 (Fig. 3G), when maximum apoptosis occurs in Brn3b knockout mice (5). The number of apoptotic cells in the Fezf2 knockout retina is comparable with that in the Brn3b-deleted retina at E17.5 (5). The increase in apoptosis in the Fezf2 Ϫ/Ϫ retina suggests that Fezf2 is needed for the proper function of Brn3b in the developing retina.
The reduction of RGC marker BRN3B was similarly observed in the P7 retina. In Fezf2-null retinas, BRN3Bϩ cells were reduced to about 47.3% of those in wild-type littermates (Fig. 4,  A and B). As an alternative approach to determining the cell types that reduced in Fezf2 Ϫ/Ϫ retinas, we immunostained retinas with anti-SMI-32, which predominantly labels the axons of RGCs (6). As shown in Fig. 4C, Fezf2 Ϫ/Ϫ retinas had a decreased population of SMI-32 immunoreactive processes, suggesting a loss of RGC axons. Consistently, the optic nerve appeared thinner in the Fezf2-deleted retina compared with the control (Fig.  4D). In addition, we noticed that axons from Fezf2 knockout RGCs occasionally bifurcated within the retina (Fig. 4C), reminiscent of aberrant axonal trajectories as seen in the Brn3b knockout retina (27). All of these results support the hypothesis that the transient expression of Fezf2 is essential for the proper expression of Brn3b and formation of RGCs.

. Targeted disruption of Fezf2 leads to a reduction in BRN3B؉
RGCs and an increase in apoptosis. A, the Fezf2 second exon harbors two loxp sites flanking the second exon. P1, P2, and P3 are the three genotyping primers located in exon 1-3 of the Fezf2 locus. B, gel electrophoresis of the PCR products from genotyping. The presence of the first loxp site and the absence of exon 2 were determined by PCR with the primer sets of P1-P2 and P1-P3, respectively. Expression of Cre was also confirmed in the knockout mice by PCR (data not shown). C, E17.5 retinas were isolated from wild-type and Fezf2-null mice, and the relative expressions of Brn3b and Fezf2 were measured by q-PCR. The knockout mice had diminished expression of Fezf2 and Brn3b. D, expression of BRN3B and FEZF2 was measured in isolated retinas from control and Fezf2-null mice by Western blotting with ␣-tubulin as an internal control. E, immunostaining of BRN3B and activated CASPASE III at E17.5 in control and Fezf2 knockout retinas. Knocking out Fezf2 reduced BRN3Bϩ cells and increased apoptotic cells. Scale bars ϭ 200 and 50 m for low and high magnitude, respectively (BRN3B) and 50 m (CASPASE III). F, the numbers of BRN3Bϩ in the GCL and NBL per field (ϫ40). G, total of CASPASE IIIϩ cells per section in control and Fezf2-null retinas. **, p Ͻ 0.005; ***, p Ͻ 0.0005.

Fezf2 Regulates Brn3b in the Developing Retina
Fezf2 Knockdown Inhibits BRN3B Expression-Fezf2 knockout reduced BRN3Bϩ RGCs in the retina, either as direct effects of inhibition of Brn3b expression or as effects of developmental changes leading to loss of BRN3Bϩ RGCs indirectly. To determine whether the loss of BRN3Bϩ RGCs was caused by a cell-autonomous or a non-cell-autonomous mechanism, we knocked down Fezf2 in a small population of retinal cells within the context of a healthy environment by retinal in utero electroporation. Fezf2 in the developing retina was knocked down with lentiviral shRNA that specifically inhibits Fezf2 expression by RNA interference (18). Plasmid pLVTH-Fezf2-siRNA-1, which harbors a reporter gene, Gfp, was delivered into the eye cups of E13.5 embryos by in utero electroporation. The consequences of Fezf2 inhibition were examined in the transfected retinas at E17.5, 1 day after the maximal expression of endogenous Fezf2 (Figs. 1 and 2). The cells expressing the siRNA moved to the GCL normally, as did those expressing pLVTH-Fezf2-scRNA-1, a scramble RNA that does not target any sequences. This result suggests that inhibition of Fezf2 does not interfere with the normal migration of RGCs. We analyzed the expression of marker genes, MATH5, ISLET1, and BRN3B, in transfected cells by IF. Substantial populations of scramble RNA-electroporated cells expressed MATH5, ISLET1, and BRN3B, as measured by their colocalization with the co-transfected GFP (Fig. 5A). MATH5 and ISLET1 were expressed similarly in Fezf2-siRNA-transfected cells as in the scramble transfectants (MATH5 siRNA, 24.8%; scRNA, 23.2%; ISLET1 siRNA, 23.8%; scRNA, 21.4%). In contrast, compared with the scrambled control, introducing Fezf2-siRNA-1 resulted in a massive reduction of BRN3B-positive cells (siRNA, 1.1%; scRNA17.2%) (Fig. 5B). Likewise, Fezf2-siRNA-2, but not the correspondent scramble control, diminished BRN3Bϩ cells in the developing retina (data not shown). The similar effect introduced by two siRNAs that target different locations in FEZF2 supports the hypothesis that the inhibition of BRN3B was attributed to the knockdown of FEZF2 specifically. Fezf2 inhibition abolished BRN3Bϩ cells, but not ISLET1, which coexpresses with BRN3B, demonstrating that Fezf2 is essential for the expression of Brn3b rather than controlling the survival of BRN3Bϩ RGCs directly.
In mature retinas, BRN3A and BRN3B are each expressed in approximately 70% of RGCs in an overlapping pattern. Brn3a was considered a downstream target of Brn3b (28). To evaluate the expression of BRN3A in the cells with diminished BRN3B, we knocked down Fezf2 with Fezf2-siRNA-2. Fezf2 inhibition did not significantly change the ratio of colocalization of BRN3Aϩ and the transfectants (Fig. 5, D and E), indicating that Fezf2 inhibition had no effect on the expression of BRN3A. Our result is consistent with a previous study with a knockin alkaline phosphatase to visualize individual genetically altered RGCs. It was found that Brn3b is not required for the maintenance of Brn3a expression in RGCs. The loss of Brn3a in Brn3b-null retinas is more likely due to an overall reduction in BRN3Aϩ RGC number rather than inhibition of Brn3a expression (27).
Because FEZF2 is a transcription factor, we envisioned that the transcription and mRNA levels of Brn3b may be subject to regulation by FEZF2. To test this, we inhibited Fezf2 expression by electroporating the pLVTH-Fezf2-siRNA-1 and pLVTH-Fezf2-scRNA-1 plasmids into the eye cups of E13.5 embryos.  The electroporated cells were dissociated from the retinas at E17.5 and isolated by FACS on the basis of expression of the reporter GFP. Assays with q-PCR revealed that Fezf2-siRNAexpressing cells had a decreased level of Brn3b mRNA relative to those expressing the scramble control following the inhibited expression of Fezf2 by the interference RNA (Fig. 5C). This data suggests that Brn3b depends on the transient expression of Fezf2 during retinal development.
To determine whether FEZF2 is a determinant of Brn3b expression in vivo, we also performed a gain-of-function study by in utero electroporation of Fezf2-pCAGEN into the developing retina. Expression of Fezf2 under a CAG promoter resulted in overall inhibition of RGC-related genes, including Math5, Islet1, and Brn3b (Fig. 6). A strong ectopic expression of this transcription factor may dramatically alter the phenotype of host cells. It was found that striatal progenitors change their fate from medium spiny neurons to cortex-like projection neurons after ectopic expression of Fezf2 (29). Intrinsic transient expression, but not a high and sustained expression of Fezf2, is essential for RGC development in the retina.
FEZF2 Binds to and Activates a 5Ј Regulatory Region of Brn3b-Because both knockout and knockdown of Fezf2 decreased Brn3b mRNA in retinas (Figs. 3C and 4C), Fezf2 may regulate Brn3b at the transcription level. FEZF2 acted as a histone deacetylase-associated repressor down-regulating multiple oncogenes through direct binding to their promoters (30). In zebrafish, FEZF2 binds to a core consensus-binding site (CAGCAACC) by which the transcription factor drives a forebrain enhancer activity of reporter genes both in vitro and in vivo (31). Four contiguous FEZF2 binding motifs are located at Ϫ274, Ϫ195, ϩ21, and ϩ211 of the transcriptional starting site of the Brn3b locus (Fig. 7A). We cloned this tentative regulatory region containing the Fezf2-binding motifs into pGL4.18, a luciferase reporter vector, for measuring promoter activity. The pGL4.18-Brn3b promoter was stably expressed in neuroblastoma N2a cells, which were then transiently transfected with Fezf2-pCAGEN and pCAGEN or pLVTH-Fezf2-siRNA-1 and pLVTH-Fezf2-scRNA-1. The relative luciferase activity of Fezf2-transfected cells was 2.1-fold higher than that of the control. Conversely, the activity of the Fezf2-siRNA-transfected group was decreased by 51.6% compared with those of the scrambled control (Fig. 7B). The results indicate that this regulatory sequence near the 5Ј UTR is positively regulated by FEZF2 and constitutes a part of the Brn3b promoter.
To investigate whether the increased transcriptional activity mediated by the Brn3b promoter resulted from FEZF2 binding, we generated V5-tagged GFP and Fezf2 expression constructs and expressed the fusion proteins in N2a cells. The DNA fragments bound to FEZF2 were isolated by ChIP with a V5 antibody and analyzed by q-PCR and gel electrophoresis following PCR. The region by which FEZF2 activated the reporter activity was successfully amplified by PCR using the ChIP DNA as a template (Fig. 7C). Consistently, the predicted binding fragment of FEZF2 was enriched by 3.7-fold compared with the 3Ј UTR, as assayed by q-PCR (Fig. 7D). In contrast, V5 antibody failed to pull down the Brn3b promoter region from cells expressing V5-GFP (Fig. 7C). The data suggest that FEZF2 binds to a promoter region that contains FEZF2 consensus binding sites and helps to drive the transcription of Brn3b.

Discussion
Retinal neurons start to generate from neuroepithelial cells in the optic cup during the middle of gestation. Newly generated neurons exit the cell cycle and migrate toward the inner (vitreal) side of the optic cup. By E13.5, the retina begins to transform from a uniform sheet of neuroepithelial cells into two  ChIP with an anti-V5 antibody precipitated FEZF2-bound DNA from N2a cells expressing V5-tagged FEZF2 but not those expressing V5-GFP. The DNA was analyzed by gel electrophoresis (C) of the PCR products for the presence of the tentative promoter and measured by q-PCR (D) for enrichment of the binding sequence. The input of ChIP and the pulldown without V5 antibody were taken as positive and negative controls of the PCR, respectively. The control sequence for the q-PCR in D is a region in the 3Ј UTR of the Brn3b locus. *, p Ͻ 0.05; **, p Ͻ 0.005.

Fezf2 Regulates Brn3b in the Developing Retina
layers of prospective GCLs and NBLs. This study examined the expression and function of FEZF2, a key transcription factor regulating cortical projection neurons, in the developing retina. We found that a peak of transient expression of Fezf2 mRNA occurs at E16.5 in the NBL, where the retinal progenitors are located (Fig. 1). Interestingly, FEZF2 protein was found in the GCL and colocalized with the RGC marker BRN3B (Fig. 2). The transient expression of Fezf2 in migrating retinal cells may allow its protein to accumulate in post-migratory neurons in the GCL. The pattern of Fezf2 expression is spatially analog to that of Math5. A knockin lacZ at the Math5 locus appeared in the retina, including the GCL (2), despite Math5 mRNA being expressed in the NBL at E15.5 (11). The expression of Fezf2 at E16.5 supports it as a later regulator of Brn3b. Fezf2 may not regulate peak formation (E11.5-E14.5) but help the survival of RGCs around E17.5, when maximal apoptosis occurs (5).
The roles of Fezf2 in the developing retina were examined by the study of loss of function. We found that Fezf2 knockout retinas had reduced BRN3Bϩ RGCs (Figs. 3 and 4). To determine whether this reduction is an effect of inhibition of Brn3b expression or developmental changes leading to the loss of BRN3Bϩ RGCs, we knocked down Fezf2 in a small population of retinal cells with RNA interference by in utero electroporation (18). The results supported the hypothesis that the knockdown of Fezf2 repressed BRN3B cell-autonomously. The inhibition of BRN3B was not accompanied by any loss of ISLET1 and BRN3A, two RGC markers that largely overlap with the expression of BRN3B (Fig. 5). This observation indicates that reduction of BRN3B is the cause, not the result, of the decrease in BRN3Bϩ RGCs in the knockout retina. The BRN3B deficit increased the apoptosis of RGCs and reduced optic nerve fibers (Figs. 3 and 4), as expected for its function in the maintenance and survival of RGCs (5,6).
Our data revealed that Fezf2 was essential for the proper expression of BRN3B, whereas it had no effects on BRN3A (Fig.  5, D and E), a closely related protein in the family of class IV POU domain transcription factors, indicating a specific action of Fezf2 on Brn3b. A recent study found that the high-mobility group domain-containing transcription factors SOX4/SOX11 express in the developing mouse retina and function redundantly to regulate the generation and survival of RGCs (32). SOX4 and SOX11 transactivate Fezf2 by competing with the specific repressor SOX5 in the cerebral cortex (33). It is not clear whether SOX4 and SOX11 participate in the transient activation of Fezf2 in the retina. However, our results are consistent with the observation that targeted disruption of SOX4 or SOX11, positive regulators of Fezf2 (33), caused a reduction of RGCs in retinas (32).
The mechanisms responsible for FEZF2 regulation of Brn3b were explored in this study. A 5Ј regulatory region of Brn3b containing a putative FEZF2 binding motif was enriched in the DNA-protein complex in N2a cells expressing V5-FEZF2 but not in those expressing V5-GFP (Fig. 7). This observation suggests that FEZF2 may bind to a tentative promoter of Brn3b. An in vitro assay found that FEZF2 activated a reporter construct carrying the 5Ј regulatory part of Brn3b. The result indicates that FEZF2 enhances the transcription of Brn3b by binding to the genomic locus of Brn3b. Notably, the Wilms tumor gene Wt1, which encodes a zinc finger transcription factor, also activates the transcription from the 5Ј regulatory sequence of Brn3b (34). The 5Ј sequence, including the 5Ј UTR, may contain positive control elements driving the transcription of Brn3b. Mouse embryos with targeted disruption of Wt1 exhibit thinner retinas and enhanced apoptotic cell death of RGCs (34). This is in parallel to our observation that Fezf2-null retinas had increased apoptosis and reduced optic nerves (Figs. 3 and 4). Future studies are needed to determine whether Fezf2 knockout leads to a visual deficiency in mice.
Although transient expression of Fezf2 is required for Brn3b, sustained expression completely blocked BRN3B (Fig. 6). The underlying mechanisms responsible for the inhibition of Brn3b are not clear. It remains to be excluded that the genomic locus of Brn3b may harbor unknown regions mediating inhibition by FEZF2. On the other hand, sustained expression of Fezf2 blocked MATH5 in addition to BRN3B and ISLET1 (Fig. 6). We speculate that Fezf2 could function as a repressor of Math5 and, thereafter, inhibit its downstream targets. It is known that targeted deletion of Math5 abolishes the retinal expression of Brn3b (2) despite Fezf2 expressing transiently in the developing retina. Therefore, if Math5 were depressed by a sustained Fezf2, then Brn3b would be blocked. With the inhibition of Math5, the phenotype of retinal cells may be altered by ectopic expression of Fezf2, a fate-determining gene of dorsal telencephalic progenitors and cortical projection neurons (17,29).
In conclusion, we determined the expression and function of Fezf2 in the developing retina. Transient expression of Fezf2 is essential for the development and survival of BRN3Bϩ RGCs by regulating the transcription of Brn3b. This study provides novel insights into the molecular mechanisms underlying RGCs and retinal development.