In ferret, E27.5 and E34.5 are ages that correspond to E15.5 and E17.5, respectively, in mouse in terms of uncrossed RGC production and axon growth to the chiasm (Chalupa and Snider, 1998). At both of these ages, Zic2-positive cells occupy an area in temporal retina larger than in mouse retina that reflects the proportion and location of uncrossed RGCs. At E27.5, Zic2 expression was more central than at E34.5 (compare Figures 5Bd and 5Be). In ferret retina, Zic2-positive and retrograde-labeled zones overlap but are not in register, suggesting as in mouse that Zic2 is expressed in younger ganglion cells, likely prior to or during the early phases of axonogenesis.

In Xenopus tadpoles, Zic2 mRNA was detected in the CMZ and not VT neural retina. At metamorphosis, in contrast, when the ipsilateral projection develops, Zic2 is expressed in VT retina. In adults, Zic2 mRNA was present in VT retina but at lower levels than at metamorphosis, corresponding to the continued but slowed production of uncrossed ganglion cells (Figure 5C).

Finally, Zic2 mRNA was not detected in chick neural retina at two different stages, St24 and St30, when axons reach the chiasm area (Thanos and Mey, 2001), even though Zic2 was localized in spinal cord, ventral diencephalons, and other structures (data not shown).

Thus, the number of retinal cells expressing Zic2 correlates with the spatiotemporal features of the formation of the uncrossed projection in the albino and pigmented mouse, ferret, Xenopus, and chick, and reflects the degree of binocularity in each of these species and strains.

Loss and Gain of Zic2 Function in Retinal Ganglion Cells

Reduced Levels of Zic2 In Vivo Perturb the Development of the Ipsilateral Projection
To test whether Zic2 is required for the formation of the ipsilateral projection, we analyzed embryos from genetically modified mice that have significantly reduced levels of Zic2 (knockdown mice), both heterozygotes (Zic2+/kd) and homozygotes (Zic2kd/kd) (Nagai et al., 2000).

Zic2kd/kd embryos exhibit varying defects in neural tube closure (Nagai et al., 2000). Some Zic2kd/kd embryos also show defects in the eye, in agreement with the proposed role of Zic2 in early eye development (Nagai et al., 1997), but in most Zic2kd/kd embryos, the eyes appeared normal (Figure 6A , Table 1). Immunohistochemistry using the Islet1/2 antibody showed normal patterns of RGCs in Zic2kd/kd or Zic2+/kd mice with normal eyes, even in those embryos with the most severe phenotype in the spinal cord, suggesting that RGC positioning is unaltered in these animals (Figure 6B). Zic2kd/kd embryos also exhibited anatomical malformations in the ventral diencephalon and concomittantly altered expression of molecules important for axon divergence at the chiasm, such as ephrin-B2 (Nakagawa et al., 2000; Williams et al., 2003) (Figure 6B). Importantly, Zic2+/kd mice have an apparently normal chiasm (including normal expression of ephrin-B2) (Figure 6B), making the heterozygote knockdown a more suitable model for this study.

 
Figure 6. Zic2 Is Required for the Proper Development of the Ipsilateral Projection In Vivo

(A) Lateral views of E17.5 littermate embryos: Zic2+/+ (a) and Zic2kd/kd embryos (b and c). Zic2kd/kd embryos usually have normal-appearing eyes (b) although occasionally display an underdeveloped eye (c).

(B) E16.5 Zic2+/+, Zic2+/kd, and Zic2kd/kd retinal sections (cut frontally) stained with α-Zic2 and Islet1/2. Zic2 is expressed in VT retina in Zic2+/+ embryos (a) but was not detected in Zic2kd/kd embryos (g). Some cells express Zic2 in the Zic2+/kd retina (d). Islet1/2 expression was similar in Zic2+/+, Zic2+/kd, and Zic2kd/kd embryos (b, e, and h). Scale bar: 100 μm. (c, f, and i) Ephrin-B2 expression (green) is similar in Zic2+/+ and Zic2+/kd embryos in the chiasm region but is altered in Zic2kd/kd, along with the conformation of the ventral diencephalons. Scale bar: 200 μm.

(C) In (a), (d), (g), and (j), Zic2 expression in E16.5 retinal whole mounts from Zic2+/+, Zic2+/kd, and Zic2kd/kd embryos. Scale bar: 500 μm. In (b), (e), (h), and (k), higher magnification of the VT region. In (c), (f), (i), and (l), DiI anterograde tracing of the retinal projection in the chiasm in the same embryos as in (a), (d), (g), and (j), respectively. Scale bar: 100 μm. (c) Chiasm of a Zic2+/+ embryo. Abbreviations: ON optic nerve; OT, optic tract. Green arrowhead, ipsilateral projection; red arrowhead, contralateral projection; the asterisk marks the chiasmatic midline. (f) Chiasm of a Zic2+/kd embryo with a dramatic reduction of ipsilateral fibers. (i) Chiasm of a Zic2+/kd embryo having no visible axons in the ipsilateral pathway. (l) Chiasm of a Zic2kd/kd embryo with no visible axons growing into the ipsilateral pathway, but fibers stray from the optic nerve and grow into the diencephalon caudal to the chiasm (arrowhead). Scale bar: 200 μm.

(D) Correlation between Zic2+ cell number and ipsilaterally projecting axons in Zic2+/+ and Zic2+/kd embryos. Black circles, Zic2+/+; Green squares, Zic2+/kd. Zic2kd/kd were not represented because they have no Zic2+ cells. Data on Zic2kd/kd are listed in Table 1.
 

Table 1. Retinal Axon Projection Phenotype at the Optic Chiasm of Zic2 Knockdown Embryos
 
Genotype Number of Embryos Age Overall Phenotype Zic2+ Cells in Retina Cip
 
     
+/+ 8 E16–E17 normal 685 ± 116 * 0.30 ± 0.03 *
     
     
+/kd 14 E16–E17 normal 295 ± 122 * 0.11 ± 0.07 *
     
     
kd/kd 1 E16 NTM, one eye not developed ND 0.16, AP
     
     
kd/kd 1 E16 NTM, one eye not developed 35 0.21, AP
     
     
kd/kd 1 E16 NTM, normal eyes 10 UI, AP
     
     
kd/kd 1 E16 NTM, normal eyes 0 UI, AP
     
     
kd/kd 1 E17 NTM, normal eyes 0 UI, AP
     
     
kd/kd 1 E17 NTM, normal eyes 0 UI, AP
     
     
kd/kd 1 E17 NTM, normal eyes 0 UI, AP
     
     
kd/kd 1 E17 NTM, normal eyes ND UI, AP
 
 
Zic2-positive cells were counted in retinal whole mounts for each individual embryo (Zic2+cells). Cip, coefficient of ipsilaterally projecting axons, an index of the size of the projection (see Experimental Procedures and Figure 6). In Zic2+/+ and Zic2+/kd embryos, Zic2-positive cell number and Cip are expressed as an average, with data pooled from ten different litters. In Zic2kd/kd embryos, data are shown for each individual embryo. Abbreviations: ND, not determined; NTM, neural tube malformation; UI, undetectable ipsilateral projections; and AP, aberrant projections straying from the chiasm.
 
 

To investigate whether the ipsilateral projection was reduced in embryos with decreased levels of Zic2, the retinal projection was labeled with DiI at E16.5–E17.5 in Zic2+/+, Zic2+/kd, and Zic2kd/kd embryos. Zic2+/kd embryos displayed a reduced ipsilateral projection compared to Zic2+/+ littermates (Figures 6C and 6D, and Table I). Most Zic2kd/kd embryos lacked an ipsilateral projection, but there were a few cases in which the uncrossed projection was apparent (Table 1). In both Zic2kd/kd and Zic2+/kd genotypes, a subset of axons consistently projected in aberrant directions distal to the optic nerve but before the chiasmatic midline, conferring a frayed appearance to the caudal aspect of the chiasm (Figure 6Dl).

Zic2+/kd embryos showed a reduction in the ipsilateral projection, but there was variability in the phenotype (Figure 6C). To test whether this variability in the uncrossed projection was a consequence of differing levels of Zic2, we compared expression of Zic2 in VT retina with the size of the ipsilateral projection in each embryo by measuring fluorescence in the ipsilateral optic tract and comparing this value to fluorescence in the optic nerve (coefficient of the ipsilateral projection [Cip]; see Experimental Procedures). We found that the number of cells expressing Zic2 in VT neural retina in Zic2+/kd embryos closely correlates with the proportion of fibers that project ipsilaterally (n = 8 for Zic2+/+, n = 13 for Zic2+/kd, Figure 6D; Table 1). It was not possible to make this correlation in Zic2kd/kd embryos because Zic2 levels were not detectable in VT retina. The phenotype of each Zic2kd/kd embryo is described in Table 1.

Zic2 Modulates the Behavior of RGC Axons in Response to Chiasm Cues
To investigate whether Zic2 is sufficient to switch the response of RGCs to inhibitory cues at the optic chiasm midline, we utilized in vitro assays previously established by our lab that reconstitute the interactions of retinal axons with chiasm cells. In these assays, retinal explants are cultured in contact with cells dissociated from the midline region of the ventral diencephalon surrounding the optic chiasm, the native cue-presenting cells that direct the bilateral retinal projection. Neurites from VT retinal explants are more strongly inhibited by chiasm cells than neurites from other retinal regions (Marcus et al., 1995; Wang et al., 1995).

First, we confirmed that Zic2 is not expressed at E14.5 in explants from DT retina after 18–24 hr in coculture with chiasm cells (data not shown). Next, prior to coculture with dissociated chiasm cells, E14.5 DT retinal explants were infected with a recombinant Sindbis virus expressing EGFP alone (EGFP), Zic2 (Zic2-EGFP), or an inactive Zic2 mutant (Zic2Δ-EGFP). Neurite outgrowth from DT retinal explants was not affected by the introduction of EGFP or Zic2Δ-EGFP, but when Zic2 was ectopically misexpressed, neurites from DT explants grew 30% shorter compared to those explants in which EGFP (p < 0.02) or Zic2Δ-EGFP (p < 0.04) were overexpressed (Figure 7B) . To test the possibility that Sindbis virus or overexpression of Zic2 itself was toxic, we infected DT and VT retinal explants with Zic2-EGFP Sindbis virus and plated them without chiasm cells. Axonal outgrowth under these conditions was similar to or better than uninfected DT or VT explants grown without chiasm cells, indicating that the amount of Zic2-Sindbis virus that we added to the explants did not perturb axonal outgrowth (Figure 7B).
 
Figure 7. Zic2 Is Sufficient to Switch the Response of RGCs to Midline Cues from Extension to Inhibition

(A) Representative DT (a–h) or VT (I–p) retinal explants from E13.5 mouse embryos, uninfected (a, e, i, and m), missexpressing EGFP (b, f, j, and n), Zic2-EGFP (c, g, k, and o), or Zic2Δ-EGFP (d, h, l, and p), cocultured with dissociated chiasm cells. Axons are visualized with α-neurofilament (a–d and i–l). (e–h) and (m–p) represent the same explants as (a)–(d) and (i)–(l), respectively, but show EGFP fluorescence. Scale bar: 500 μm.

(B) Quantification of the area occupied by neurites from DT (filled bars) or VT (open bars) retinal explants at E13.5 (left graph) or E14.5 (right graph); uninfected; expressing EGFP, Zic2-EGFP, or Zic2Δ-EGFP; and cocultured with dissociated chiasm cells. Explants that were uninfected or infected with Zic2-EGFP-Sindbis virus and grown without chiasm cells were included as a control for possible toxicity from the virus. **p < 0.001 relative to EGFP infected explants and *p < 0.02 relative to EGFP infected explants (Student's unpaired t- test). Number above bars indicates number of explants. Data were pooled from nine independent experiments.
 

At E14.5, because Zic2 expression has commenced in many cells in retina, it is possible that RGCs in DT retina have already been specified to project contralaterally by other regulatory genes putatively determining the crossing axonal phenotype and thus are less responsive to the introduction of Zic2. We therefore performed the same experiment on DT retinal explants but taken from younger retinae. The misexpression of Zic2-EGFP in DT explants from E13.5 rather than E14.5 retinae led to a dramatic reduction in the extent of neurite outgrowth (about 67%) compared to controls (EGFP [p < 0.001] or Zic2Δ-EGFP [p < 0.003]) (Figures 7A and 7B). In addition, neurite outgrowth from VT retinal explants in which Zic2 was overexpressed was severely inhibited by chiasm cells (65% reduction in neurite outgrowth compared to EGFP-expressing VT explants) (p < 0.001) (Figure 7B). In sum, gain-of-function assays in vitro indicate that Zic2 activity is sufficient to switch the projection phenotype of RGCs with respect to cues provided by chiasmatic cells, from extension to neurite inhibition, and suggest that this behavior is determined in a narrow time interval.

Using this in vitro assay, we also tested whether Zic2 is required to switch the behavior of RGCs toward chiasm cells. Zic2 function was blocked with decoy oligos bearing the Zic DNA binding sequence (Zic2 decoy). As a negative control, we used the same sequence only with point mutations that abolish Zic2 binding (Zic2M decoy). The outgrowth from VT retina transfected with Zic2-decoy oligos was significantly enhanced compared to explants incubated with the lipofection agent alone or explants transfected with Zic2M decoy). In contrast, outgrowth from DT explants transfected with Zic2 decoy was unchanged and was similar to outgrowth in explants transfected with the lipofection agent alone or with Zic2M decoy (see Supplemental Data online at http://www.cell.com/cgi/content/full/114/5/545/DC1).

Together, the loss- and gain-of function analyses implicate Zic2 as an essential determinant of the ipsilateral retinal projection at the optic chiasm.

Discussion

In this study, we show that in the mouse retina, ganglion cells having an ipsilateral projection can be distinguished from the more numerous contralaterally projecting ganglion cells by expression of the transcription factor Zic2. Zic2 expression in VT neural retina is limited to the period when the ipsilateral projection is established at the optic chiasm. Three additional lines of evidence implicate Zic2 as a determinant of RGC identity with respect to laterality of the retinal projection. (1) The spatiotemporal expression of Zic2 reflects the extent of binocularity in four distantly related species. (2) A reduction in Zic2 levels in vivo leads to a decrease in the number of axons that project ipsilaterally at the optic chiasm. (3) Gain-of-function experiments convert the response of DT retinal ganglion cell neurites to cells of the chiasm midline, from extension to inhibition.

Spatiotemporal Aspects of Zic2 Expression in the Retina
During the development of the RGC projections to the brain, Zic2 expression is restricted temporally, from E14.5 to about E17.5, after neurogenesis and during the period in which the permanent ipsilateral RGC axon projection is established, and spatially, exclusively in the region from which the uncrossed projection arises. The finding that Zic2 is more likely to be expressed in RGCs with axons in the optic nerve than in RGCs that have projected into the optic tract suggest that this gene is downregulated as axons transverse the optic chiasm.

The formation of the optic chiasm occurs in two phases. First, early-generated RGC axons originating from dorsal-central retina reach the optic chiasm at E12.5 and project ipsi- or contralaterally. This early ipsilateral projection from dorsocentral retina is thought to be transient (Guillery et al., 1995). In a second phase, at E14.5, RGCs differentiate in a central to peripheral wave to produce the permanent uncrossed projection from peripheral VT retina. Zic2 is not detected in dorsocentral retina at E11.5–E13.5, suggesting that Zic2 exclusively specifies the permanent RGC projection from VT retina.

Together, these data are consistent with the current idea that postmitotically expressed transcription factors have key functions in activating programs for specific pathway choices and target recognition (for review, Shirasaki and Pfaff, 2002) here, with respect to the midline.

Zic2 As a Candidate Transcriptional Regulator of the Uncrossed RGC Projection
The Zic family was originally identified as a group of genes encoding zinc finger proteins expressed in adult mouse cerebellum (Aruga et al., 1996). Zic2 collaborates with Zic1 in cerebellar folial and areal patterning (Aruga et al., 2002) and marks later stages of granule cell differentiation (Rubin et al., 2002). The three best studied members of the Zic family, Zic1, Zic2, and Zic3, are also expressed in the early embryo, including the eye, playing essential roles in CNS development as “prepatterning” genes (Nagai et al., 1997; Brown et al., 1998, 2001; Aruga et al., 2002). Here, we show that Zic2, but not Zic1 or Zic3, is implicated in an additional developmental process, the specification of a subtype of retinal ganglion cell that have uncrossed axon projections with respect to the CNS midline. Thus, it appears that Zic2 plays multiple roles in neural development, as a neuronal inducer in undifferentiated cells, and as a postmitotic factor, acting during later differentiation in VT retina.

Zic2kd/kd mice show variability in the RGC projection phenotype, from a severe reduction to an approximately normal number of uncrossed axons; in all of these variants, a subset of axons strays from the optic chiasm. Zic2 is also highly expressed in the developing optic chiasm (Brown et al., 2003), and accordingly, Zic2kd/kd embryos show anatomical malformations at the chiasm region. The random unraveling of a subset of axons in Zic2kd/kd embryos could be a consequence of perturbations of the ventral diencephalon on which the chiasm forms. In fact, the disposition of ephrin-B2, a major inhibitory cue for VT axons in the optic chiasm midline and thus key to retinal axon divergence in the optic chiasm (Williams et al., 2003), appears to be altered in Zic2kd/kd mice. Thus, the possibility that Zic2 also regulates axon guidance factors at the optic chiasm cannot be excluded and will be investigated in the future. However, the finding that Zic2+/kd embryos have a normal chiasm, together with our gain-of-function experiments in vitro, argue for a role for Zic2 in designating the ipsilateral optic projection, acting primarily within the retina.

Our functional experiments in vitro demonstrate that postmitotic expression of Zic2 is able to switch axon projection phenotype within 24 hr and more potently when it is ectopically expressed at E13.5 than at E14.5, likely before the expression of transcription factors that might specify the contralateral RGC phenotype. This suggests that Zic2 may operate over a relatively short period, in accordance with transcription factor function.

The LIM code expressed in postmitotic neurons has been proposed to participate in cell fate and axon trajectory of neuronal subtypes throughout the CNS (Shirasaki and Pfaff, 2002). Although Zic2 does not belong to the LIM family of homeodomain transcription factors, it seems to act similarly, inducing postmitotic RGC axons to differentially respond to extrinsic guidance cues. However, the present findings point to a novel transcriptional program for axon pathfinding with respect to the midline of the CNS. It will be interesting to investigate the generality of this program, particularly whether Zic family members are involved in determining axon projections in other bilateral pathways in the CNS. Intriguingly, Zic2 has recently found in a subset of cells in the periotic mesenchyme (Warner et al., 2003), potentially relating to the several uncrossed pathways in the acoustic nuclei-brainstem projection.

The experiments presented here suggest that Zic2 controls downstream events that may direct the avoidance of chiasm midline cells by RGCs with an uncrossed trajectory. Eph receptors and their ephrin ligands mediate axonal divergence at the optic chiasm (Nakagawa et al., 2000; Williams et al., 2003). Notably, the EphB1 receptor is expressed in retina in a spatiotemporal pattern identical to Zic2 and loss-of-function analyses in vivo indicate that EphB1 is necessary for formation of the uncrossed pathway at the optic chiasm (Williams et al., 2003). Thus, EphB1 is a prime candidate for regulation by Zic2, a hypothesis in line with LIM homeodomain protein regulation of ephrinA-EphA interactions in dorsoventral selection of motor axon pathways in the limb (Kania and Jessell, 2003).

Zic2 Expression Reflects the Degree of Binocularity
Zic2 expression matches the extent of binocularity across several species, including mouse, Xenopus, and ferret. Thus, although mammals and amphibians use different strategies to establish binocular vision during their lifetime, the fact that Zic2 expression correlates with degree of binocularity suggests a conserved function for Zic2 in modulating the uncrossed RGC axon projection. Moreover, Zic2 is absent in chick retina, again supporting a role for Zic2 in the development of the ipsilateral retinal pathway, since chicks lack a noticeable uncrossed projection.

The findings on the albino retina underscore the hypothesis that Zic2 determines the uncrossed retinal projection. Albino mice display a reduction in the number of uncrossed fibers and a concomitant reduction in Zic2-expressing cells. This reduction might arise because the tempo of retinal neurogenesis is accelerated in the albino (Rachel et al., 2002). A slight alteration in the numbers of cells born on each day of retinal neurogenesis could bias the production of RGCs in favor of one population or the other by affecting the postmitotic expression of Zic2.

In summary, the spatiotemporal expression pattern of Zic2, across species and strains with varying sizes and developmental onset of the binocular projection, along with gain- and loss-of-function studies, implicate Zic2 in the patterning of binocular vision.

Experimental Procedures

Animals
All mouse experiments were performed using C57BL/6 mice except for experiments involving albino embryos for which albino mice (Tyrc/Tyrc) of the C57BL/6 strain were crossed with heterozygous pigmented mice (Tyr+/Tyrc). The Zic2 knockdown embryos were obtained from mice as described in Nagai et al. (2000). Noon of the day on which a plug was found was considered embryonic day (E)0.5. Ferret embryos were purchased at Marshall Farms USA, Inc. Xenopus tadpoles and frogs were obtained from Xenopus, Inc.

Immunoblotting, Immunohistochemistry, and In Situ Hybridization
HEK293 cells and peripheral DT and VT segments of retinae were pooled, lysed, loaded on a gel, and immunoblotted using a pan-Zic antibody (Brown et al., 2003). Fixed retinal whole mounts or vibratome sections were processed for immunohistochemistry with α-Zic1 (AbCAM), α-Zic2 (Brown et al., 2003), or α-Islet1/2 (K4, provided by Tom Jessell, Columbia University). For BrdU staining, BrdU (100 mg/kg) was injected intraperitoneally into pregnant females, and 1 hr later embryos were anesthetized and removed. Fixed retinas were treated with 2 N HCl, 1% Triton X-100, and 0.1 M borate to denature DNA and processed for inmunohistochemistry with α-BrdU antibodies. In situ hybridization was performed according to reported methods (Schaeren-Wiemers and Gerfin-Moser, 1993). A 603 bp StuI-EcoRI segment from Zic2 mouse cDNA and the Xenopus cDNA clone IMAGE: 3378215 were used as templates for mouse and Xenopus Zic2 riboprobe synthesis, respectively.

Anterograde and Retrograde Dye Tracing and Quantification of the Ipsilateral Projection
Anterograde labeling of normal and mutant mice was accomplished as described (Plump et al., 2002) but with the exception that the projection was labeled by placing dye crystals on the optic nerve head of one eye after removing the retina for immunolabeling with anti-Zic2. After DiI labeling, brains were removed and the optic chiasm viewed en face in whole mount with a fluorescence dissecting microscope. To quantify the uncrossed projection, squares were superimposed on the width of the labeled optic nerve proximal to the chiasm and on the ipsilateral optic tract. The fluorescence within the square covering the optic nerve (ONF) and optic tract ipsilateral to the dye labeling (OTF) was measured using Openlab Software. ONF and OTF were normalized with respect to the area in each case, and the coefficient of the ipsilateral projection (Cip) was determined for each individual embryo by applying the following formula: Cip = OTF/ONF + OTF. Retrograde labeling with DiI was performed according to standard methods. Retrograde tracing combined with immunohistochemistry was performed as described (Rachel et al., 2002).

Culture Assay
Retinal explants and dissociated chiasm cells from E14.5 and E13.5 embryos were cocultured as described previously (Marcus et al., 1996) and immunostained with α-neurofilament (3A10, provided by Tom Jessell, Columbia University). The extent of axon outgrowth was quantified using Openlab software to measure the area covered by RGC axons, expressed as mean area (±SEM) occupied by RGC axons (μm2 × 104) (Aa). The amount of axon growth was normalized in each case with respect to the explant size: (Aa/Ae) × 104, where Ae = area covered by each retinal explant.

Sindbis Virus Synthesis and Infection
Sindbis virus encoding EGFP, Zic2-EGFP, or Zic2-Δ-EGFP was generated according to the Sindbis Expression System, Version E (Invitrogen) but using a modified version of the virus with decreased cytotoxicity (gift of Dr. Osten, Max-Planck, Heidelberg). The Zic2 mutant was generated by deletion of an internal SmaI fragment, leading to a nonfunctional protein (Sin-Zic2Δ-IRES-EGFP) that lacks 100 amino acids from the zinc finger domain (Brown et al., unpublished). Sindbis virus was incubated with retinal explants for 1 hr at 37°C in serum-free medium. Excess virus was removed by washing with serum-free medium, explants were plated, and dissociated chiasm cells added.