Brn3b/Brn3c double knockout mice reveal an unsuspected role for Brn3c in retinal ganglion cell axon outgrowth

Projection neurons are typically defined as those that project long axons to different regions of neuronal or nonneuronal tissues. Examples include dorsal root ganglia, trigeminal ganglia and retinal ganglion cells (RGCs). Recent studies implicate the mouse Brn3 POU-domain transcription factors as key regulators for axon outgrowth and pathfinding in projection neurons (Huang et al., 1999; Gan et al., 1999; Wang et al., 2000; Erkman et al., 2000; Eng et al., 2001). Similarly, Caenorhabditis elegans UNC-86, an ortholog of the Brn3 factors, directly regulates the expression of genes involved in axonogenesis of mechanosensory neurons in the nematode (Sze et al., 1997; Duggan et al., 1998; Rohrig et al., 2000). In the mouse, Brn3 genes are expressed in projection neurons immediately after the cells exit mitosis and adopt specific neuronal fates. In addition, Brn3 genes are expressed postnatally (Xiang et al., 1995), suggesting a potential function in neuronal maintenance and survival in adult life. Further characterization of the Brn3 factors in mammalian neuronal development could thus shed light not only on de novo axon formation during embryogenesis but also on the failure of projection neurons to regenerate in adults (Goldberg and Barres, 2000).

In the mouse, the Brn3 family is composed of Brn3a, Brn3b and Brn3c (Pou4f1, Pou4f2 and Pou4f3, respectively – Mouse Genome Informatics). These factors share greater than 95% identity within their POU domains, suggesting that they have closely related functions in neuronal differentiation. The Brn3 genes are expressed in newly formed projection neurons with a high degree of spatial and temporal overlap. In the dorsal root ganglia, Brn3a expression is first observed at E9.5 (McEvilly et al., 1996). Expression of Brn3b and Brn3c follows that of Brn3a. In the developing retina, Brn3b expression is first detected at E11.5, when the first RGCs are formed. Expression of Brn3a begins 1 day later, followed by Brn3c, the last Brn3 gene expressed in the retina (Xiang et al., 1995; Gan et al., 1999). In cochlear and vestibular cells of the inner ear, Brn3c is the first of the Brn3 genes to be expressed, at approximately E14 (Xiang et al., 1997a) (S. W. W., unpublished). Thus, in each sensory organ system, a different Brn3 gene is the first to be expressed.

Targeted deletions of the Brn3 genes affect the sensory systems in a manner that reflects the first gene that is expressed. Mice without Brn3a lose their suckling functions and show impaired somatosensory and motor control in large part because of abnormal trigeminal ganglia and dorsal root ganglia (McEvilly et al., 1996; Xiang et al., 1996a; Xiang et al., 1996b). By contrast, Brn3b knockout mice manifest a loss of a large number of RGCs (Gan et al., 1996; Erkman et al., 1996), and Brn3c knockout mice have hearing and vestibular defects because of cochlear and vestibular hair cell degeneration (Erkman et al., 1996; Xiang et al., 1997a). An interesting feature of the Brn3 knockout mice is that each displays a highly individualized phenotype that is not found in the others. Thus, no retinal or inner ear defects were associated with Brn3a knockout mice, no somatosensory/motor or inner ear defects were associated with Brn3b knockout mice and no somatosensory/motor or retinal defects were associated with Brn3c knockout mice.

The Brn3 factors are not required for the initial specification of sensory neurons but are essential for their normal differentiation and survival (Erkman, 1996; Xiang, 1998; Xiang et al., 1998; Gan et al., 1999; Wang et al., 2000; Eng et al., 2001). In Brn3b knockout mice, RGCs are specified and migrate to the ganglion cell layer, but most are not able to project normal axons, and eventually they undergo apoptosis (Xiang, 1998; Gan et al., 1999; Wang et al., 2000). Indeed, neuritic processes emanating from Brn3b-deficient RGCs assume a dendrite-like character indicating that Brn3b regulates the expression of genes critical for axon formation and may suppress dendrite formation. Whether Brn3a or Brn3c have a role in axonogenesis or other processes in RGCs remains unclear because in Brn3a and Brn3c knockout mice, Brn3b or other factors might compensate. A notable aspect of the Brn3b^–/– phenotype is that 20% to 30% of the RGCs are able to send out axons, albeit abnormal ones, and these residual RGCs survive through adulthood (Wang et al., 2000; Erkman et al., 2001). These results predict that a more severe RGC phenotype might occur if one of the later expressed Brn3 genes was also deleted.

To test this prediction, we have generated Brn3b/Brn3c double knockout mice. Unlike Brn3a^–/– mice, which die at birth, both Brn3b^–/– and Brn3c^–/– mice are viable and fertile. We expected that the Brn3b/Brn3c double knockout mice would also be viable. This expectation turned out to be true, and so we were able to characterize retinas and optic chiasms in postnatal double knockout animals. Our results show that Brn3c partially compensates for the loss of Brn3b because the double knockout mice had a greater loss of RGC axons than did the Brn3b knockout mice. Our experiments indicate that Brn3c plays an essential role in RGC axon formation. In addition, Brn3c appears to have distinct functions from those of Brn3b in axon pathfinding.

We previously described a Brn3b knock-in allele in which Brn3b was replaced with the gene encoding human placental alkaline phosphatase (AP) (Gan et al., 1999). We also created Brn3b-GFP and Brn3c-AP alleles (L. G., unpublished). Mice deficient in both Brn3b and Brn3c were generated by mating Brn3b (AP/AP) or Brn3b (GFP/GFP) males with Brn3c (+/AP) females. All compound heterozygous animals appeared normal and were intercrossed to produce the Brn3b (GFP/GFP):Brn3c (AP/AP) double knockouts that were used for subsequent experiments.

Total RNA was isolated from retina of different embryonic stages by TriReagent (MRC), and first-strand cDNA was synthesized with Superscript reverse transcriptase (Life Technologies) at 42°C. cDNA equivalent equal to 40 ng of total RNA was used for PCR amplification with the Hotstart Taq DNA polymerase (Qiagen) in a volume of 20 μl. The reaction mixture was heated at 95°C for 15 minutes, denatured at 94°C for 30 seconds, annealed at 50°C for 30 seconds and extended at 72°C for 1 minute. Different cycle numbers were tested to determine the optimal cycle number for each reaction. The primers for PCR were as follows: Brn3b, forward, 5′-TCTGGAAGCCTACTTCGCCA-3′ and reverse, 5′-CCGGTTCACAATCTCTCTGA-3′; Brn3a, forward, 5′-AGGCCTATTTTGCCGTACAA-3′ and reverse, 5′-CGTCTCACACCCTCCTCAGT-3′; Brn3c, forward, 5′-TCTTCAACGGCAGTGAGCGT-3′ and reverse, 5′-ACACCCTGGAGTGTCCCGTA-3′; and β-actin, forward, 5′-CAACGGCTCCGGCATGTGC-3′ and reverse, 5′-CTCTTGCTCTGGGCCTCG-3′. The PCR products were separated on a 2% agarose gel, and visualized using ethidium bromide.

Retinal tissues were collected at E13.5 and dissected into four to eight pieces. Fresh retinal pieces were placed on laminin- (40 μg/ml) and poly-D-lysine- (100 μg/ml) coated coverslips in a six-well petri dish containing DMEM and 1% N2 neuronal supplement (Gibco BRL), 1% penicillin-streptomycin (Gibco BRL), and 0.2% glutamine (Gibco BRL). In addition to the DMEM, the final culture medium contained 5 μg/ml insulin, 100 μg/ml transferrin, 20 nM progesterone, 0.1 mM putrescine, 30 nM selenium, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.4 mM glutamine. Retinal explants were cultured for 2-3 days at 37°C with 5% CO₂ and air at 99% humidity.

Procedures for AP histochemical staining were modified from Fields-Berry et al. (Fields-Berry et al., 1992). Briefly, retinal explants were fixed with 3.2% paraformaldehyde (EMS) in 0.1 M phosphate-buffered saline (PBS) at pH 7.4 for 3 minutes. Fixed samples were rinsed four times with PBS for 10 minutes each and incubated at 65°C for 5 minutes to inactivate the endogenous AP activity. Heat-treated tissues were incubated in AP buffer (100 mM Tris-HCl (pH 9.5), 100 mM NaCl, 50 mM MgCl₂, and 2 mM levamisal (Sigma)) at room temperature for 30 minutes. Color reactions were carried out in AP detection solution (100 mg/ml BCIP (Sigma) and 1 mg/ml NBT (Sigma)) for 3 to 15 hours at room temperature. The samples were then post-fixed with 3.2% paraformaldehyde for 10 minutes at room temperature and washed twice with TE and once with PBS. Samples were mounted in Fluromount (EMS) for microscopy. Images were examined with an Olympus IX-70 inverted microscope and collected digitally.

Cryosections or wholemounts of retinal tissue were labeled with primary and secondary antibodies following standard immunohistochemical procedures. Briefly, fixed whole retinas or cryosections were washed three times with PBS containing 0.05% Triton X-100 (PBS-T) and blocked with 2% bovine serum albumin (BSA) in PBS-T for 1 hour. Samples were then incubated sequentially for 1 hour each with the first primary antibody, first secondary antibody, second primary antibody and second secondary antibody. Samples were washed three times with PBS-T between each antibody incubation. Primary antibodies were mouse anti-β3 tubulin (Chemicon), rabbit anti-choline acetyltransferase (ChAT) (Chemicon) and mouse anti-neurofilament light chain (NFL) (Zymed). Secondary antibodies were Cy5-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch Lab) and Alexa488-conjugated goat anti-mouse IgG (Molecular probes). Nuclei were visualized by quick rinsing in propidium iodide solution (1/3000 dilution of saturated solution). Labeled tissues were mounted in Fluoromount and examined using an Olympus FV500 confocal microscope equipped with argon, green HeNe and red HeNe lasers. Images were projected from 5 to 12 optical sections with intervals of 0.5 μm. Projected images were pseudo-colored to represent the colors as seen in an epifluorescence microscope. Cy5 emission was assigned blue.

Anesthetized P1 pups were used for anterograde tracing of RGC axons at the optic chiasm. DiI (Molecular Probes) and DiASP (Molecular Probes) were applied to different eyes using a Nanoject (Drummond). For each labeling, 70 nl of 10% DiI or DiASP dissolved in dimethyl formamide was injected three times at different spots of the ganglion cell layer within the tempo-ventral portion of the retina. Optic chiasms were collected from sacrificed mice 20 to 24 hours after injection. The collected optic chiasms were mounted on coverslips in Fluoromount for confocal microscopic analysis.

The pHSVPrPUC/CMVegfp amplicon used in these experiments expresses the enhanced green fluorescence protein (eGFP) under the control of the CMV IE promoter and the gene of interest under control of the HSV IE4/5 promoter (Bowers et al., 2001). The pHSVBrn3c/CMVegfp amplicon was constructed by inserting a cDNA fragment encoding Brn3c into the XbaI and BamHI sites of the parent amplicon. Helper virus-free amplicon packaging and virus purification were performed as described previously (Bowers et al., 2001). Viral infectivity was determined by assessing both GFP expression and transduction efficiency (Bowers et al., 2000). For infection, retinal tissues were incubated with 50 μl of viral products with the same volume of culturing media (Wang et al., 2000) at 37°C for 1 hour. Infected retinal tissues were then transferred to neural culture media with laminin-coated substrate (Wang et al., 2000).

To test whether Brn3a or Brn3c might partially compensate for the loss of Brn3b in RGC differentiation and survival, we first needed to determine whether Brn3a or Brn3c expression was affected in retinas of Brn3b^–/– mice. In wild-type mice, Brn3a and Brn3b are expressed in almost all RGCs, while Brn3c is expressed in approximately 50% of the RGCs (Xiang et al., 1995; Gan et al., 1999). Brn3a and Brn3c expression is diminished in retinas of Brn3b^–/– adults (Erkman et al., 1996; Gan et al., 1996), but the decreased expression is largely due to the presence of fewer RGCs (Gan et al., 1996).

To determine whether Brn3a and Brn3c were expressed in developing retinas of Brn3b^–/– mice, we performed RT-PCR with RNA isolated from wild-type and Brn3b^–/– E14.5 and E16.5 retinas. In the wild-type retinas, robust expression of Brn3b and Brn3a was observed at E14.5 and E16.5, while Brn3c expression was weak at E14.5 and increased several fold by E16.5 (Fig. 1). In Brn3b^–/– retinas, Brn3a and Brn3c expression was attenuated to a half to a third of wild-type levels in both E14.5 and E16.5 retinas, but was nonetheless detectable (Fig. 1). Between E14.5 and E16.5, Brn3b^–/– RGCs undergo enhanced apoptosis relative to controls (Xiang, 1998; Gan et al., 1999), and thus lowered expression levels for Brn3a and Brn3c were at least partially caused by the presence of fewer RGCs or RGCs undergoing programmed cell death. The RT-PCR experiments clearly indicated that although the absence of Brn3b attenuated Brn3a and Brn3c expression in the developing retina, expression was not abolished. The results are thus consistent with those of Gan et al. (Gan et al., 1996), who showed that expression of Brn3a and Brn3c in adult retinas was not noticeably altered on a per RGC basis. It therefore remained possible that Brn3a and Brn3c partially compensated for the loss of Brn3b in Brn3b^–/– mice.

We chose to test whether Brn3c was required for RGC differentiation and whether it partially compensates for the loss of Brn3b by generating Brn3b/Brn3c double knockout mice. This choice was made because Brn3b/Brn3c double knockout mice, unlike Brn3a-Brn3b double knockout mice, were likely to be viable, which would allow us to examine retinal defects in adults as well as embryos. To facilitate the analysis, we inserted the AP gene into the Brn3c locus to create Brn3c-deficient mice (Fig. 2). The Brn3b/Brn3c double knockouts were deaf and had severe vestibular impairment (data not shown). The extreme vestibular defects were greater than those seen in Brn3c knockout mice, suggesting that Brn3b, in addition to Brn3c, might play a role in the differentiation of vestibular neurons within the inner ear. Despite their behavioral abnormalities, the Brn3b/Brn3c double knockouts survived to adulthood and were fertile.

It is known that Brn3c is expressed in approximately 50% of RGCs (Xiang et al., 1995). To determine whether Brn3c was required for axon outgrowth in this RGC subpopulation, we cultured retinal explants from Brn3b^+/+:Brn3c (AP/+) and Brn3b^+/+:Brn3c (AP/AP) embryos. If Brn3c was essential for axon outgrowth, Brn3c-expressing cells, which are defined as cells that would express Brn3c in wild-type retinas and are indicated by AP expression, should not be able to send out axons. We observed AP expression in axons emanating from Brn3c-expressing cells in both Brn3c heterozygous and homozygous null retinas (Fig. 3A-D). Thus, Brn3c was not required for axon outgrowth in Brn3c-expressing cells, at least not if Brn3b was present. AP expression was stronger in the Brn3c (AP/AP) explants than in Brn3c (AP/+) explants because there were two copies of AP rather than one (Fig. 3A,B).

We next determined whether axon outgrowth occurred in Brn3c-expressing cells in the absence of Brn3b. As expected from previous work (Wang et al., 2000), explants that marked Brn3b-expressing cells from Brn3b (AP/AP):Brn3c^+/+ retinas displayed abnormal neurite outgrowth, suggestive of dendrites rather than axons (Fig. 3E). When Brn3c-expressing cells were identified by AP expression in a Brn3b (GFP/GFP);Brn3c (AP/+) retinal explant, we also observed neurite outgrowth (Fig. 3G, arrows). By contrast, Brn3c-expressing cells from Brn3b/Brn3c double mutant explants were unable to send out any processes (Fig. 3F,H). The area of growing neurites of an explant from a Brn3b (GFP/GFP):Brn3c (AP/AP) retina is shown in Fig. 3F and the tissue itself is shown in Fig. 3H. The cells from which the neurites grew did not express AP and so were not Brn3c-expressing cells (Fig. 3F, arrows). These cells were generating neurites independently of both Brn3b and Brn3c, perhaps through events controlled by Brn3a. Nevertheless, numerous viable cells that normally express Brn3c were present within the explanted tissue (Fig. 3H, arrowheads). We also observed fractured membrane debris marked by AP expression in the Brn3b (GFP/GFP);Brn3c (AP/AP) retinal tissue indicating extensive apoptosis of Brn3c-expressing cells (Fig. 3H). Apoptosis may be the result of Brn3c-expressing cells not being able to project axons. It is also possible that some RGCs that lack both Brn3b and Brn3c factors were undergoing apoptosis within the explants before axonogenesis could occur. However, other RGCs lacking Brn3b and Brn3c did not appear apoptotic and yet were still not able to extend axons (Fig. 3H, arrowheads). It is likely that RGCs lacking Brn3b and Brn3c were abnormal both in their ability to survive and their ability to extend axons.

The results with cultured retinal explants demonstrate that Brn3c is not essential for axon outgrowth as long as Brn3b is present. However, in the absence of both Brn3b and Brn3c, Brn3c-expressing cells (as marked by AP expression from the Brn3c locus), are present but are unable to extend axons and undergo programmed cell death.

The experiments using retinal explants indicated that Brn3c had a role in RGC differentiation as a regulator of axon outgrowth and that it might partially compensate for the absence of Brn3b. If true, fewer optic fibers (RGC axons) should be detected in developing retinas that were deficient in both Brn3b and Brn3c. To examine this possibility, we monitored retinas from Brn3b^–/– and Brn3b/Brn3c double knockout embryos for the expression of markers designed to visualize optic fibers and RGC nuclei. In wild-type E16.5 retinas, numerous axons coalesced counter-radially into the optic disk in a well-fasciculated fashion (Fig. 4A,B). Retinas that lacked only Brn3c appeared identical to wild-type retinas (data not shown). In Brn3b-deficient retinas, far fewer optic fibers were observed (Fig. 4D,E), consistent with earlier reports (Erkman et al., 1996; Gan et al., 1996). Moreover, the axons of Brn3b-deficient RGCs were not fasciculated and were misguided (compare Fig. 4B with 4E). They were entangled and crossed or by-passed the optic disk, growing toward the opposite side of the retina rather than entering the optic disk (Fig. 4E).

In Brn3b/Brn3c double mutant retinas, axons were difficult to detect at low magnification (Fig. 4G). Higher magnification revealed sparsely distributed axons (Fig. 4H). However, the few axons that were observed appeared to be correctly routed, unlike the misguided axons in Brn3b-deficient retinas (Fig. 4E,H). Despite the drastic decrease in axons, the overall cell number within the ganglion cell layer was not significantly different between Brn3b^–/– and Brn3b^–/–:Brn3c^–/– retinas (Fig. 4C,F,I). Thus, as was observed in vitro with explanted retinas, the loss of both Brn3b and Brn3c blocked axon outgrowth in the subpopulation of RGCs that would normally express both Brn3b/Brn3c. These results provided strong evidence for Brn3c functioning in RGC differentiation to promote axon outgrowth.

The greater loss of axons in embryonic retinas of Brn3b/Brn3c double knockout mice suggested that adult retinas would also be affected. As is the case for Brn3b-deficient mice, we would expect RGCs that were unable to extend axons would not survive into postnatal life. Retinas were collected from 3-week-old mice and monitored for the expression of markers for RGCs and their axons, amacrine cells and cell nuclei (Fig. 5).

In the adult retinas, we found that different regions of retinas from double knockout mice were differentially affected. Within the dorsal-nasal region, Brn3b/Brn3c mutant retinas had only slightly fewer RGC axons than retinas from Brn3b^–/– mice, although, as expected, neither genotype reached the number in wild type (Fig. 5A-C). However, within the ventral-temporal region, Brn3b/Brn3c double knockout retinas had far fewer axons than their Brn3b-deficient littermates (Fig. 5D-F, arrows indicate nonspecifically labeled blood vessels).

Cell numbers within the ganglion cell layer of the ventral-temporal region were also altered (Fig. 5G-I). Retinas lacking Brn3b had approximately 85% of the cells present in wild-type controls, whereas retinas of the double mutants had only 70% of the cells. However, the reduction in number was less than expected given the substantial loss of axons and was not reflective of the reported value for Brn3b^–/– RGC loss (Gan et al., 1996). The discrepancy was explained by the replacement of RGCs with displaced amacrine cells, as revealed by ChAT staining of starburst amacrine cells (Fig. 5J-O).

Nuclei stained with propidium iodide showed two populations of cells within the ganglion cell layer, based on their size and intensity of staining (green and blue arrows in Fig. 5G-I). Those with large nuclei and diffuse propidium iodide staining were mostly RGCs, and those with small nuclei and intense staining were mostly starburst amacrine cells. In wild-type retinas, RGCs occupied most of the space in the ganglion cell layer with a few displaced amacrine cells scattered among them (Fig. 5G, Fig. 6). In Brn3b^–/– retinas, there was a 65% reduction in the number of RGCs, while the number of displaced amacrine cells increased threefold (Fig. 5H, Fig. 6). In Brn3b/Brn3c double mutant retinas, the number of RGCs was reduced by 60% compared with that of Brn3b^–/– retinas (Fig. 5I, Fig. 6).

In Brn3b/Brn3c mutant mice, it appeared that the loss of RGCs led to an increased percentage of cells in the ganglion layer that were displaced amacrine cells. To further substantiate this point, we prepared retinal sections at randomly selected orientations across the optic disk of Brn3b^–/– and Brn3b/Brn3c double knockout mice at 3 weeks of age. As expected, the optic fiber (RGC axons) was substantially reduced in the Brn3b^–/– retina when compared with a wild-type control (Fig. 5J,K), and was virtually absent in the Brn3b/Brn3c double mutant retina (Fig. 5L).

Closer examination of the ganglion cell layer revealed an average of 15±3 RGCs (n=6) over a distance of 200 μm in wild-type retina (Fig. 5M; white arrows point to RGCs), with about a third of the ganglion cell layer composed of displaced amacrine cells (Fig. 5M). In Brn3b^–/– retinas, 6±2 RGCs (n=6) were identified over the same distance, and displaced amacrine cells occupied two-thirds of the population in the ganglion cell layer (Fig. 5N). In the double-mutant retinas, the RGC number was reduced to 2±2 (n=6), and although the number of amacrine cells was not substantially increased from that found in Brn3b^–/– retinas, almost 90% of the cells in the ganglion layer were displaced amacrine cells (Fig. 5O). These results further demonstrate that Brn3c has a compensatory function in RGC differentiation and axon formation that is uncovered in Brn3b/Brn3c double knockout mice. Mice that are deficient in Math5 (Atoh7 – Mouse Genome Informatics) or Pax6 genes are unable to specify RGCs, and in retinas of Math5^–/– and Pax6^–/– mice, missing RGCs are also replaced by amacrine cells, probably because of a respecification of RGC progenitors (Wang et al., 2001; Marquardt et al., 2001).

Brn3b/Brn3c double mutant retinas but not Brn3b^–/– retinas, suffered from a greater loss of projections emanating from the ventral-temporal region than from the dorsal-nasal region (Fig. 5C,F). In the adult mouse, ipsilateral projections are restricted to RGCs within the ventral-temporal crescent of the retina (Dräger, 1985). It was therefore possible that Brn3c was associated with controlling the ipsilateral growth of RGC axons, and in Brn3b/Brn3c double mutant adult retinas, a defect in ipsilateral projections would be apparent. To test this possibility, we performed anterograde labeling of RGCs with P2 wild-type, Brn3b^–/– and Brn3b-brn3c double mutant mice.

Right-side retinas were labeled with DiI (Fig. 7, red), and left-side with DiASP (Fig. 7, green). Optic chiasms were examined 20-24 hour after dye injection. In wild-type optic chiasms, the majority of axons projected contralaterally, but a small proportion of ipsilateral axons was clearly seen (Fig. 7A,D). In Brn3b^–/– optic chiasms, the total number of axons was reduced, indicating that the majority of RGC axons were not able to reach the optic chiasm (Fig. 7B,E). Many of the axons that reached the optic chiasm of Brn3b^–/– mice were abnormal. They were misrouted dorsally toward the hypothalamus (12/12 examples) or occasionally went into the optic nerve of the other eye (4/12 examples) (Fig. 7B,E). Misrouting of RGC axons in Brn3b^–/– mice has been reported previously by Erkman et al. (Erkman et al., 2000). Interestingly, a larger proportion of axons was directed ipsilaterally in Brn3b^–/– mice, making the ipsilateral/contralateral ratio almost 1.0 (Fig. 7E).

In optic chiasms of Brn3b/Brn3c double knockout mice, the axon number was further reduced (Fig. 7C,F). Unexpectedly, the misrouted axons observed in Brn3b^–/– optic chiasms were not found in the double mutants (Fig. 7C,F; n=6). In addition, virtually no ipsilateral projections were seen in the double mutant optic chiasms (Fig. 7F). The results suggest that in the absence of Brn3b and Brn3c, RGC axons defaulted to a contralateral pathway, presumably because RGCs from the ventral-temporal region of the retina were missing. Brn3c appears to promote enhanced ipsilateral growth and axon misrouting in Brn3b^–/– mice, as both of these defects were eliminated in the double mutant.

We next asked whether Brn3c was sufficient for promoting axon growth in RGCs that were deficient in this process. Cultured retinal explants from E13.5 Brn3b^–/– embryos extend abnormal processes that are short, tangled and dendritic in character (Fig. 3E) (Wang et al., 2000). Except for neurite and migration defects, Brn3b^–/– RGCs appeared to be well differentiated and should therefore be useful in attempts to restore axon outgrowth. When Brn3b was expressed in Brn3b^–/– retinal explants by HSV-mediated gene transfer (Bowers et al., 2001), effective restoration of axon outgrowth was observed (data not shown). We repeated the experiment using an HSV amplicon containing Brn3c (Fig. 8A). In this amplicon, a GFP sequence was expressed under separate control (Fig. 8A). Fig. 8B,C shows efficient GFP expression in an explant derived from a Brn3b^–/– retina infected with an HSV amplicon lacking Brn3c. Expression was found in most cells within the explanted tissue (Fig. 8B) and in neurites extending outward from the tissue (Fig. 8C). These neurites were the dendrite-like processes reported previously (Wang et al., 2000). Moreover, abnormal migrating RGCs characteristic of Brn3b-deficient retinas were present at the edge of the tissue (Fig. 8C). When Brn3c was included in the amplicon, GFP expression was similar to the control vector (Fig. 8D). The abnormally migrating RGCs were not observed and the number of neurites extending outward increased tenfold (Fig. 8E). In addition, the neurites were substantially longer in Brn3b^–/– explants that expressed Brn3c (Fig. 8E). However, the neurites did not resemble normal axons in that they were highly branched (Fig. 8E, inset) suggesting that Brn3b and Brn3c did not function identically to each other in this gain-of-function analysis.

In this study, we have shown that Brn3c has critical functions in RGC axon outgrowth. These functions were not apparent in mice with a targeted deletion solely at the Brn3c locus but were revealed in Brn3b/Brn3c double knockout mice. Brn3c is expressed in approximately 50% of RGCs, while Brn3b is expressed in virtually all RGCs (Xiang et al., 1995; Gan et al., 1999; Wang et al., 2000). The subpopulation of Brn3b/Brn3c-expressing RGCs represents the cells that are probably affected in the double mutants.

When Brn3c, but not Brn3b, was absent from cells normally expressing both Brn3b/Brn3c, RGC differentiation and axon formation occurred. When Brn3b, but not Brn3c, was missing, the cells sent out abnormal processes. When both Brn3b and Brn3c were missing, cells normally expressing both Brn3b and Brn3c were present, which indicated that the RGC differentiation program had occurred, but the RGCs were unable to extend even abnormal axons. The absence of Brn3b and Brn3c led to a loss of Brn3b/Brn3c-expressing RGCs in adults. Presumably, these cells underwent apoptosis because they could not extend axons. RGCs in the ventral-temporal region of the adult retina were more severely affected by the loss of Brn3b and Brn3c than were cells in the dorsal-nasal region, implying that the ventral-temporal region in wild-type mice is enriched in Brn3b/Brn3c-expressing cells. Indeed, ipsilateral projections are missing at the optic chiasm of Brn3b/Brn3c double knockouts, suggesting that Brn3c may be involved in the control of ipsilateral guidance.

The distinct phenotypes exhibited by Brn3a^–/–, Brn3b^–/– and Brn3c^–/– mice had previously suggested a simple model in which each Brn3 factor was necessary for the terminal differentiation and survival of an individual sensory system (Xiang et al., 1997b). Of particular relevance, Brn3b was shown to be required for the normal differentiation of RGCs (Erkman et al., 1996; Gan et al., 1996) and Brn3c for auditory and vestibular hair cells (Erkman et al., 1996; Xiang et al., 1997a). The work described here demonstrates that a more complex relationship must exist between Brn3b and Brn3c in the regulation of RGC gene expression. The absence of both Brn3b and Brn3c genes resulted in a more severe RGC defect than that observed for either single deletion. In fact, published studies on mice that lack only Brn3c have yet to reveal any RGC defects, indicating that the role of Brn3c role in RGC differentiation can be largely or entirely replaced by other factors, most likely Brn3a and Brn3b. Conversely, the RGC phenotype of Brn3b/Brn3c double mutant mice demonstrates that Brn3c plays a compensatory role in the absence of Brn3b. Thus, Brn3b and Brn3c appear to have partially overlapping roles in RGC axon outgrowth, despite the fact that Brn3b is expressed much earlier in development than Brn3c.

We made use of a gain-of-function analysis using Brn3b^–/– retinal explants to demonstrate that Brn3c was able to promote axon outgrowth, even in the absence of Brn3b. The results indicate that high levels of Brn3c suffice to perform many of the functions normally required by both Brn3b and Brn3c. However, the axons in Brn3c-expressing cells were not normal, and it is therefore possible that Brn3c cannot replace all the functions of Brn3b. We favor the notion that Brn3b and Brn3c have distinct as well as overlapping functions in promoting axon outgrowth in differentiating RGCs. These functions must ultimately relate to the sets of genes that are dependent on Brn3b, Brn3c, or both factors for their expression.

The Brn3b^–/– retinal explants contained other cell types besides RGCs, and our analysis did not distinguish whether HSV-mediated expression of Brn3c (or Brn3b) could promote axon outgrowth in non-RGCs or whether the outgrowth was limited to RGCs. This is an interesting distinction because Brn3b^–/– RGCs are at least partially differentiated, whereas non-RGCs within the Brn3b^–/– explants are likely to include uncommitted neuroblasts and neuroblasts that are committed to non-RGC fates. Axons emanating from these cells as a result of Brn3c misexpression would suggest a much broader function for Brn3c. Thus, Brn3c might be capable of initiating the RGC differentiation program in non-RGCs of explanted retina. Consistent with this notion, Liu et al. (Liu et al., 2000) showed that all three Brn3 genes can promote RGC fate in non-RGC progenitors in the chick retina.

A recent study showed that Brn3b regulated genes associated with axon pathfinding in adult mice (Erkman et al., 2000). The residual axons emanating from Brn3b^–/– RGCs exhibit pathfinding defects at every point where axons make decisions about which way to project. Pathfinding defects were found at the peripheral retina, optic disk, optic nerve, optic chiasm and superior colliculus (Erkman et al., 2000). In addition, residual Brn3b^–/– axons do not fasciculate in early retinogenesis. Several genes involved in axon outgrowth and pathfinding have recently been identified as potential target genes of Brn3b by differential hybridization screens, including those for neuritin and Gap43 (Erkman et al., 2000; Mu et al., 2001).

Defects in Brn3b^–/– residual axons are particularly noticeable at the optic chiasm. In addition to misrouted axons, the number of axons projecting ipsilaterally is substantially increased in Brn3b^–/– mice. There may be a close relationship between misguided axons and ipsilateral axons. In normal axon guidance, RGC axons that eventually project ipsilaterally do so by trial and error, and during development many axon projections appear to be wandering; the projections that do not migrate ipsilaterally eventually undergo apoptosis (Godement et al., 1994; Marcus and Mason, 1995).

In contrast to Brn3b^–/– optic chiasms, those from Brn3b/Brn3c double mutants had virtually no misrouted or ipsilateral RGC axons. These results indicate that the presence of Brn3c and the absence of Brn3b enhanced the production of ipsilateral projections, while the absence of Brn3c and Brn3b enhanced their loss. The most severe loss of RGCs in Brn3b/Brn3c double knockout mice derives from the ventral-temporal region of the adult retina, where ipsilateral projections originate. Thus, Brn3c appeared to be required for ipsilateral projections in the subpopulation of Brn3b/Brn3c-expressing cells in the ventral-temporal region of the retina. Preliminary evidence indicates that ipsilateral projections are also missing at the optic chiasm of Brn3c^–/– mice (S. W. W., unpublished). This suggests that Brn3c has a specialized function in axon pathfinding that cannot be replaced by Brn3b or other factors.

During RGC axonogenesis, each Brn3 factor regulates distinct as well as overlapping sets of genes required for the formation of the final RGC axonal network. Although Brn3a^–/– mice display no observable retinal phenotype, we predict that roles for Brn3a in RGC differentiation will be uncovered in Brn3a/Brn3b and Brn3a/Brn3c double knockout mice, and in Brn3a/Brn3b/Brn3c triple knockout mice. Moreover, it is likely that the distinct and overlapping requirements for the Brn3 factors in axonogenesis will extend to dorsal root and trigeminal ganglia, the other projection neurons expressing the Brn3 genes.

We thank Darlene Howard and Ann Casey for assistance in packaging HSV amplicons. Genotype analysis was performed by Lin Liu. We acknowledge the National Eye Institute (EY11930) and the Robert A. Welch foundation for their support to W. H. K. The HSV amplicon project is supported by an AFAR Grant to W. J. B. and an National Institute of Aging grant (AG18254) to H. J.F.