Six3 regulates optic nerve development via multiple mechanisms

Malformations of the optic nerve lead to reduced vision or even blindness. During optic nerve development, retinal ganglion cell (RGC) axons navigate across the retina, exit the eye to the optic stalk (OS), and cross the diencephalon midline at the optic chiasm en route to their brain targets. Many signalling molecules have been implicated in guiding various steps of optic nerve pathfinding, however much less is known about transcription factors regulating this process. Here we show that in zebrafish, reduced function of transcription factor Six3 results in optic nerve hypoplasia and a wide repertoire of RGC axon pathfinding errors. These abnormalities are caused by multiple mechanisms, including abnormal eye and OS patterning and morphogenesis, abnormal expression of signalling molecules both in RGCs and in their environment and anatomical deficiency in the diencephalic preoptic area, where the optic chiasm normally forms. Our findings reveal new roles for Six3 in eye development and are consistent with known phenotypes of reduced SIX3 function in humans. Hence, the new zebrafish model for Six3 loss of function furthers our understanding of the mechanisms governing optic nerve development and Six3-mediated eye and forebrain malformations.

. In current models of Six3 loss of function, eye formation is severely abrogated from early stages. Mice homozygous for Six3 null mutation lack all forebrain structures including eyes 16 and conditional removal of Six3 from the mouse eye field resulted in arrested neural retina specification 17 . In medaka, reduced Six3 function resulted in reduced eye tissues and cyclopia was also observed 18 . In zebrafish there are three six3-related genes, six3a, six3b and six7 19,20 . The combined loss of zebrafish six3b and six7 function resulted in anophthalmia or microphthalmia 21 . Hence, to date it has been difficult to study the roles of Six3 in later stages of eye development, when Six3 is expected to function based on its expression, which is initially observed throughout the developing eye at optic vesicle stage, becomes largely limited to the OS and ventral eye at optic cup stage 19,20,[22][23][24] , and is limited to the retina inner nuclear layer and RGCs after neurogenesis [25][26][27] .
Here we describe a new zebrafish model for Six3 loss of function that reveals new roles for Six3 affecting later eye development. We show that zebrafish embryos with reduced Six3a and no Six3b functions have malformed eyes with large optic disc colobomas. The optic nerve in these embryos is hypoplastic and severely defasciculated and RGC axons make multiple pathfinding errors, both in eye and in forebrain. These errors appear to result from multiple mechanisms, including abnormal eye patterning, failure of OS differentiation, abnormal differentiation of RGCs and deficiencies in the diencephalic midline. Altogether, our findings show that Six3 is required for normal RGC and optic nerve development and identify six3a;six3b double mutants as a useful model for revealing new roles for Six3 in eye development.

Results
six3a;six3b-deficient embryos. To identify new roles for Six3 in embryonic development we screened for mutations in six3a by TILLING 28,29 . From all identified mutations, only the vu129 allele, in which a T to C transition changes a highly conserved Leucine in position 183 in the homeodomain to Serine (L183S) (Fig. 1A,B), caused a severe reduction in Six3a activity, as tested by six3a misexpression (Supplementary Fig. S1). These results identify six3a vu129 as a strong hypomorphic allele.
As both six3a vu129 and six3b vu87 homozygous mutants appear normal during embryonic and early larval stages 21 , we examined whether there was redundancy between these genes by generating double mutant embryos. We intercrossed double heterozygotes for six3a vu129 and six3b vu87 and examined their progeny, which appeared normal until 2 days post-fertilization (dpf). However, at 2 dpf, and more prominently at 3 dpf, light could be seen passing through the lenses in some embryos, whose eyes also appeared slightly smaller (Fig. 1D). The passage of light suggested a defect in the retinal pigmented epithelium (RPE) layer, which covers the neural retina, and indeed, the medial region of the RPE appeared incomplete (Fig. 1F). These embryos died around 2 weeks post-fertilization. Genotyping revealed that affected embryos were almost strictly double mutants, carrying four mutant six3 alleles (n = 22/25), and few (n = 3/25) carried three mutant alleles. We hereafter refer to double mutants as six3a;six3b double mutants or Six3-deficient embryos.
To better understand the eye phenotype we performed histological sections at 5 dpf, which revealed large colobomas in the optic disc region with ectopic retinal tissue protruding towards the brain (Fig. 1H). Additionally, optic nerve axons, which normally form a single bundle that exits the eye at the optic disc region (Fig. 1G,I), appeared severely defasciculated (Fig. 1H,J) and the optic chiasm could not be detected (Fig. 1I,J).
Together, these results show that reduced Six3 function through partial loss of Six3a and complete loss of Six3b activities causes abnormal eye and optic nerve morphogenesis.
Multiple optic nerve abnormalities in Six3-deficient embryos. We analyzed more closely the effects of Six3 loss of function on development of the optic nerve. In zebrafish, RGCs begin to differentiate around 28 hours post-fertilization (hpf) 30,31 , axons exit the eye around 32 hpf, cross the chiasm at 34-36 hpf and fully innervate the tectum at 72 hpf 32,33 . We began by examining the optic nerve at 40 hpf, when axons should have crossed the diencephalic midline. To visualize optic nerve axons we used antibodies against DM-GRASP (zn-5) 34,35 or against acetylated α -tubulin (Ac-T). Compared to normal ( Fig. 2A), Six3-deficient embryos had hypoplastic optic nerves, axons failed to reach the midline and fewer RGCs were detected, especially in the ventral retina (Fig. 2B). By 48 hpf, optic nerve axons of Six3-deficient embryos reached the midline but the optic nerve remained hypoplastic and there was no clear structure of an optic chiasm (Fig. 2D). Moreover, some axons appeared to project abnormally inside the eye and many axons appeared to stray from the optic nerve before and in the midline region (Fig. 2D).
To better visualize pathfinding of the optic nerve we used lipophilic dyes for anterograde labelling of RGC axons from one eye, at 3 and 5 dpf. Whereas in WT embryos axons from one eye were always traced only to the contralateral tectum (n = 22) (Fig. 2E), in Six3-deficient embryos axons were found both in the contralateral and ipsilateral tecta (n = 31/35) and also in the telencephalon (n = 34/35) (Fig. 2F, S3).
As the ventral retina appeared to be more strongly affected by Six3 loss of function, we asked whether RGC axons from both dorsal and ventral regions exhibited similar pathfinding defects. We labelled dorsonasal (DN) and ventrotemporal (VT) eye quadrants with DiI and DiO, respectively, and examined RGC axons. In WT embryos (n = 15), all axons from VT and DN retina invariably projected contralaterally (Fig. 2G). In Six3-deficient embryos, axons of VT RGCs made many more pathfinding errors than DN axons. In all examined embryos (n = 17), many VT axons projected to the ipsilateral tectum and/or forebrain, whereas DN axons transverse sections through forebrains of normal (G,I) and Six3-deficient (H,J) 5 dpf embryos. "R" in (H) marks ectopic retinal tissue of one eye. Arrowhead and arrows in I point at the optic nerve inside and outside the eye, respectively, which cannot be clearly seen in the double mutant; asterisks mark location of the optic chiasm; bracket in J marks an apparently defasciculated optic nerve. (C,D,G-J) dorsal is up, (C-F) anterior is to the left. Scale bars are 50 μ m.
Scientific RepoRts | 6:20267 | DOI: 10.1038/srep20267 projected mostly to the contralateral tectum. However, some DN axons also showed aberrant pathfinding and axons projecting to the contralateral tectum were defasciculated (Fig. 2H). Despite the multiple pathfinding errors, many axons reached either the contralateral or ipsilateral tectum; however, it was unclear whether they formed functional connections. To address this question we performed the visual motor response (VMR) test, which monitors locomotion of larvae that is normally triggered by transitions between light and dark conditions 36 (see Methods sections for details). We found that although Six3-deficient larvae had a lower baseline activity than WT, they responded to transition from light to dark by a sharp increase of locomotor activity, similarly to WT ( Supplementary Fig. S2). When lights were turned on, both WT and Six3-deficient larvae increased their activity, however the response of Six3-deficient larvae was delayed and somewhat slower compared to WT ( Supplementary Fig. S2). These results suggest that Six3-deficient larvae can see, indicating functional connections are formed in the tectum. The mechanisms responsible for differences in locomotor behavior compared to WT larvae remain to be determined.
Six3 is required for normal eye patterning. Next, we addressed potential mechanisms that could cause optic nerve misrouting in Six3-deficient embryos. Optic nerve pathfinding errors and coloboma have been associated with abnormal eye patterning and reduced function of genes required for ventral retina and OS development. In Medaka, reduced Six3 function resulted in abnormal eye patterning with reduced expression of ventral eye genes 18 . Interestingly, the coloboma phenotype of six3a;six3b double mutant embryos is highly reminiscent of combined loss of function of transcription factors Vax1 and Vax2 37,38 . We therefore examined vax1 and vax2 expression at 30 hpf, after initial eye patterning has been completed. At this stage, both vax1 and vax2 are normally expressed in the POA, weakly expressed in the OS, and in the ventral retina vax2 is prominently expressed whereas vax1 expression is limited to the region of the choroid fissure ( Fig. 3A,C,E). In double mutant embryos, vax1 expression (n = 6) was clearly present in the OS region, and appeared upregulated. Additionally, the OS appeared abnormally large. In the POA, vax1 expression was reduced and retinal expression appeared comparable to normal (Fig. 3B). vax2 expression (n = 3) was reduced in the ventral retina, more prominently in the nasal region, almost absent from the OS and reduced in the POA (Fig. 3D,F). These results show that vax gene expression is differentially affected by Six3 loss of function, with vax2 being generally reduced and vax1 reduced in the POA and upregulated in the OS. Another transcription factor important for eye and optic nerve morphogenesis is Pax2, whose loss of function results in optic fissure coloboma and similar RGC axon pathfinding defects to Six3-deficient embryos 39,40 . At 30 hpf, pax2a was expressed in the optic fissure and OS of normal embryos, whereas in double mutants (n = 6) expression in the optic fissure was reduced and was absent from the OS (Fig. 3G,H) (also see below).
In contrast to the reduced expression of ventral genes in double mutants, expression of dorsal eye genes tbx5 (n = 2) (Fig. 3I,J) and raldh2 (n = 3) (Fig. 3K,L) was expanded. These results suggest that Six3-deficient eyes are partially dorsalized, an abnormality that can contribute to RGC axon pathfinding defects.
Optic stalk is induced but fails to differentiate when Six3 function is reduced. The OS plays a critical role in guiding growing RGC axons and its cells differentiate into astrocytes of the optic nerve. As the OS of Six3-deficient embryos failed to express pax2a and appeared abnormal at 30 hpf, we examined whether it was specified in the mutants by labelling for pax2a expression during optic cup formation. At 18 hpf (18-somite stage), pax2a was expressed in the developing OS of double mutants (n = 3) although in a slightly reduced domain, but this expression disappeared by 24 hpf (n = 6) ( Fig. 4A-D). Hence, the OS was specified and the reduction of Six3 activity resulted in failure to maintain pax2a OS expression. Next we examined OS morphogenesis in live embryos in which all cell membranes were labelled by membrane-tethered EGFP. Normally, the OS narrows significantly from 23 hpf to 30 hpf (n = 3-5 embryos for each time point) (Fig. 4E,G). In double mutant embryos, the OS failed to narrow and remained wider than normal (n = 2) (Fig. 4F,H). Hence, reduced Six3 activity also interfered with OS morphogenesis.
To find whether OS cells of double mutants differentiated into glia of the optic nerve we generated an rx3:-Kaede transgenic line, in which cells derived from the eye field, including OS and eye cells, express the photoconvertible protein Kaede 41 . When OS cells are photoconverted at 24-27 hpf in otherwise WT embryos and are imaged again at 48 hpf or later, photoconverted glial cells are clearly seen at the periphery of the optic nerve (n = 4) (Fig. 4I,K). By contrast, when the same experiment is performed in double mutants, only few or no photoconverted glial cells can be identified in the optic nerve (n = 3) (Fig. 4J,L), suggesting that when Six3 activity is reduced, OS fails to differentiate normally.
Together, the data show that OS morphogenesis and differentiation are abrogated in Six3-deficient embryos. The absence of a normal OS likely contributes to the observed RGC axon pathfinding errors.
Reduced cxcl12a signalling from the optic stalk. OS cells normally express the chemokine ligand cxcl12a, which was shown to function in intra-retinal guidance of RGC axons by attracting them towards the OS 10,42 . The failure of OS differentiation in Six3-deficient embryos prompted us to examine cxcl12a expression at 30 hpf, when axons navigate within the retina. Indeed, cxcl12a expression in the OS was strongly reduced (Fig. 4M,N), providing a possible mechanism underlying intra-retinal pathfinding defects in six3a;six3b double mutants.
Abnormal diencephalic midline formation in Six3-deficient embryos. After exiting the eye and OS, all RGC axons in zebrafish cross the anterior diencephalic midline, where they form the optic chiasm in close proximity to the diencephalic post-optic commissure (POC) and ventral to the telencephalic anterior commissure (AC) [43][44][45] . The region flanked by the AC and POC is the POA, in which expression of semaphorin 3D (sema3d) is required for midline crossing of RGC axons 46 . Additionally, correct expression of Slit family ligands slit1a, slit2 and slit3 in this region is required for normal commissure formation 45 . The failure to form a normal optic chiasm and the close proximity of the chiasm and commissures prompted us to examine expression of these signalling molecules in Six3-deficient embryos at 32 hpf, shortly before RGC axons normally approach the midline. We found that POA expression of sema3d and the three Slit-family genes was strongly reduced or missing (Fig. 5A-H). Moreover, where the AC and POC could be detected, they appeared much closer to each other than normal (Fig. 5B' ,D' ,F').
The abnormally close apposition of commissures raised the possibility that the POA was under-developed, providing an anatomic rather than regulatory explanation for the lack of signalling molecule expression. To test this hypothesis we examined expression of netrin1a (ntn1a), which is normally expressed at 32 hpf in the telencephalon and hypothalamus, leaving a gap in the POA (Fig. 6A). By contrast, in Six3-deficient embryos (n = 4) we observed a continuous expression domain throughout the anterior diencephlic midline and the telencephalic and hypothalamic domains appeared fused (Fig. 6B), supporting the idea that the POA was strongly reduced. Further confirmation for the reduction in POA was obtained by labelling the AC and POC at 28 hpf. In double mutant embryos (n = 3), the distance between the POC and AC was strongly reduced and commissures appeared fused together (Fig. 6C,D). We next labelled embryos for islet1 (isl1), which is normally expressed in the POA at 24 hpf. In Six3-deficient embryos (n = 5), isl1 POA expression could not be detected and importantly, the region ventral to the telencephalon appeared reduced at its dorsoventral dimension, suggesting that the POA was already under developed at 24 hpf (Fig. 6E,F).
These results demonstrate that in Six3-deficient embryos the diencephalic region where optic nerve axons form the chiasm is severely reduced, leading to diminished expression of signalling molecules and likely contributing to abnormalities in midline crossing of RGC axons.
Delayed and abnormal differentiation of Six3-deficient RGCs. The abnormalities described so far in retinal, OS and POA development likely contribute non cell-autonomously to pathfinding errors of RGC axons. However, reduced Six3 levels might also cause abnormalities in RGCs themselves that would lead to abnormal optic nerve development. For example, intra-retinal navigation errors could arise from lack of RGC pioneers 11 or loss of cxcr4b function in RGCs 10 , and navigation errors outside the retina are found when function of the Slit receptor Robo2 in RGCs is abrogated 6,9 . One line of evidence that RGC development is affected by Six3 loss of function comes from our earlier observation that ventroanterior RGCs, which are the first to differentiate, were not detected or appeared reduced in number in double mutants, at a time point when later-differentiating RGCs already sent axons (Figs 2B and 7B). As ventroanterior RGCs are part of the population of RGC pioneers whose axons guide later born RGC axons in the retina 11 , it is possible that their reduction/absence contributes to the phenotype of intra-retinal guidance errors.
We also examined retinal expression of chemokine receptor cxcr4b, which is required in RGCs for navigating towards the OS where its ligand, cxcl12a is expressed 10 . cxcr4b expression begins in a ventral patch at 24 hpf, before RGCs differentiate and continues in differentiated RGCs 47 . At 30 hpf, strong cxcr4b expression was present in RGCs of WT embryos, but was strongly reduced in Six3-deficient retinas (n = 4) (Fig. 7C,D). We considered two possible explanations for the lack of cxcr4b expression, namely specific downregulation in mutant RGCs or delayed differentiation of RGCs. To distinguish between these possibilities we labelled embryos for cxcr4b at 48 hpf and found that at this time point, Six3-deficient RGCs indeed expressed cxcr4b throughout the circumference of the eye, albeit at lower levels than normal (n = 3) (Fig. 7E,F), suggesting that reduced Six3 function caused delayed RGC differentiation. To verify this conclusion we labelled Six3-deficient embryos for isl1, which Scientific RepoRts | 6:20267 | DOI: 10.1038/srep20267 is normally expressed in differentiated RGCs 48 . Consistent with the idea of delayed differentiation, isl1 expression was present in normal eyes and missing from eyes of double mutants at 34 hpf (n = 5) (Fig. 7G,H), but was present in double mutant eyes at 48 hpf (n = 6) (Fig. 7J). As cxcr4b is expressed in Six3-deficient RGCs, it is unlikely to play a significant role as a cell-autonomous factor in intra-retinal pathfinding in the double mutants. However, the results uncover a role for Six3 in regulating timely differentiation of RGCs.
Next, we asked if a cell-autonomous mechanism contributes to abnormal pathfinding errors outside the eyes of Six3-deficient embryos. As the pathfinding errors exhibited by Six3-deficient RGC axons are reminiscent of those seen in astray mutants 6,9 , we focused on robo2 expression. robo2 is expressed in RGCs soon after their differentiation, beginning at 31 hpf, and more robustly by 36 hpf 9 . However, in Six3-deficient embryos robo2 levels were strongly reduced even at 48 hpf, when many RGC axons have already reached the midline (Fig. 7K,L). Hence, Six3-deficient RGCs fail to activate expression of robo2 on time and have reduced ability to respond to specific guidance cues in their environment.
Together, the results suggest that both cell-autonomous and non cell-autonomous mechanisms likely contribute to the pathfinding errors of Six3-deficient RGC axons.

Discussion
In this work we present a new model of Six3 loss of function, in which reduction in total Six3 activity through inactivation of six3a and six3b, two out of three zebrafish six3-related genes (six3a, six3b, six7), results in malformed eyes and optic nerves. Because the eye malformations are relatively mild compared to previously described models of Six3 loss of function 16,18,21 , six3a;six3b double mutants provide an opportunity to study in vivo how Six3 functions during optic cup and neurogenesis stages, thus affording identification of previously unknown functions of Six3 in eye and optic nerve development. The different combinations of double loss-of-function reveal the division of work between zebrafish six3-related genes. Although all three genes are expressed from early stages in similar patterns 20,49 , loss of six3b with downregulation of six7 causes severe microphthalmia or anophthalmia through failure to form optic vesicles 21 (our unpublished observations), whereas strong reduction of six3a function with loss of six3b function results in only mild microphthalmia and in eye abnormalities that are apparent at later stages. Labelling six-3a;six3b double mutant embryos at early segmentation stages for eye field and anterior neural plate markers suggests that early patterning of the anterior neural plate is normal (Supplementary Fig. S4). Interestingly, combined loss of six3a and six7 does not lead to overt eye phenotypes (A.S. and A.I., unpublished observations). Hence, six3a appears not to have a significant requirement during the transition from eye field to optic vesicle stage and to act redundantly with six3b only in specific aspects of eye and optic nerve development. This division of work between six3-related genes makes zebrafish a convenient system in which to dissect the roles of Six3 during different stages of eye formation. Interestingly, a previous study described the combined loss of six3a and six3b through the use of antisense morpholino oligonucleotides (MOs) against both genes 50 . This approach led to more significant microphthalmia than what we found, a difference that might stem from differences between the levels of loss of function achieved by mutation versus the use of MOs. six3a;six3b mutants also uncover a specific role for Six3 in promoting development of the POA. The POA was recently shown to develop as a distinct morphogenetic entity, the optic recess region (ORR), which is organized around the optic recess of the third ventricle 51 . Our results suggest that development of this region is particularly sensitive to Six3 levels and further studies will elucidate the mechanisms through which Six3 promotes POA development.
Importantly, the ventral forebrain deficiencies and optic nerve hypoplasia found in six3a;six3b mutants place Six3 as a potential causative gene in additional congenital diseases involving the forebrain and eyes such as septo-optic dysplasia (SOD) and Kallman syndrome. Indeed, a mutation in SIX3 was reported in a patient with Kallman syndrome 52 and a connection to SOD is supported by the fact that Six3 expression is regulated by Sox2, a causative gene in SOD [53][54][55] .
We find that reduced levels of Six3 lead to a wide repertoire of RGC axon pathfinding errors. Intra-retinal pathfinding defects are relatively mild and are more prominent in the ventral retina. One possible mechanism is the failure of some ventroanterior pioneers to differentiate on time (Figs 2B and 7B) 11 . Another likely cause for the intra-retinal errors is the strongly reduced level of cxcl12a expression in the OS. However, the phenotype in Six3-deficient embryos is surprisingly mild when compared to reports on reduced Cxcl12a levels by morpholino or mutation 10,42 . One possible explanation is that residual cxcl12a, at levels undetected by ISH, is present in the OS, providing attraction for RGC axons.
Innervation of the ipsilateral tectum in the mutants appears to be mediated by several factors. Non-autonomous mechanisms include abnormal development of the OS and diencephalic midline, the latter resulting in reduced levels of sema3d. Interestingly, similar abnormalities in OS morphogenesis and differentiation were found when Pax2 function was lost 39,40 , suggesting Pax2 is a major mediator of Six3 function in OS development. A reduced POA was also reported in Pax2 mouse knockout but not in zebrafish pax2a mutants 39,40 . Hence, this aspect of the phenotype is likely not mediated by reduced Pax2 function.
The possibility of a cell-autonomous mechanism that contributes to pathfinding errors outside of the eye is raised by the reduced level of robo2 in Six3-deficient RGCs. Nevertheless, some phenotypes reported for robo2 loss of function were not observed in six3a;six3b double mutants; this could be explained by low levels of robo2 still present in Six3-deficient RGCs.
six3a;six3b double mutants uncover novel roles for Six3 in RGC differentiation. Firstly, there is delayed differentiation of RGCs and possibly other retinal cell types (A.I., T.A. and A.M.R., unpublished). This is likely mediated by regulation of cell cycle progression and/or exit. Indeed, Six3 has been shown to influence the cell cycle in other contexts, mostly during earlier stages of eye and anterior neural plate development [56][57][58][59] . Future analyses will determine the mechanisms by which Six3 influences timing of retinal differentiation.
A second effect of Six3 on RGC differentiation is abnormal gene expression in RGCs. In this work, for example, we found strongly reduced levels of robo2 in RGCs at 48 hpf, which cannot be attributed solely to delayed differentiation as at this stage many mutant RGCs have axons that have reached the midline (Fig. 2D) and express the differentiation marker isl1 (Fig. 7J). Hence, Six3 appears to influence RGC development through affecting specific gene expression. Additional roles for Six3 in differentiated RGCs are also likely, based on its expression in these cells after neurogenesis 25,27 , and it will be important to identify these functions and their contribution to RGC biology.
Many of the RGC axon pathfinding defects in Six3-deficient embryos are also seen in mutants in which Hh pathway activity is compromised [6][7][8] . A connection between Six3 and Hh signalling is of particular interest given that both have been implicated in HPE. In mice, Six3 was shown to be required for expression of Sonic hedgehog (Shh) in the anterior ventral diencephalic midline 60,61 , suggesting Six3 acts upstream of Hh signalling in patterning the anterior diencephalon. However, work in zebrafish showed that Six3 does not function upstream of Hh but rather in parallel in patterning the telencephalon 62 . In the current study, several observations are consistent with Six3 functioning upstream of Hh signalling, namely reduced levels of known Hh downstream targets such as pax2a, cxcl12a and vax2 38,42,63 . However, other findings argue against reduced Hh activity in six3a;six3b double mutants, and include lack of reduction in ptch2 and gli1 levels in the midline (A.S., A.M.R. and A.I., unpublished observations), no reduction in vax1 levels in the OS (Fig. 3A,B) and no cyclopia, which would be expected if the deficiencies in midline development were mediated by reduced Hh signalling 38,45 . Interestingly, the enlarged optic stalk, coloboma and ectopic retinal tissue in the brain are more reminiscent of patched2 mutants, in which Hh pathway activity is upregulated 64 . However, molecular changes found in ptch2 mutants such as expanded pax2 expression and increased Hh signalling are also inconsistent with our findings in Six3-deficient embryos. Hence, based on current findings we propose that Six3's function in optic nerve development is not mediated by regulation of Hh pathway activity.
Interesting similarities also exist between phenotypes observed in Six3-deficient embryos and lhx2b (belladonna) mutant embryos. For example, in lhx2b mutants RGC axons typically fail to cross the midline and POA development appears reduced 6,12 . These similarities are particularly interesting given that lhx2b has been proposed to function as a mediator of Six3 function during early eye and forebrain growth 50 . However, in lhx2b mutants AC and POC fail to form unlike in Six3-deficient embryos and we did not observe significant reduction in lhx2b levels in Six3-deficient embryos (A.M.R. and A. I., unpublished). Hence, while it is still possible that lhx2b functions downstream of Six3 in forebrain patterning and RGC axon guidance, it does not appear to be a significant mediator of Six3 activity in this context.

Methods
Fish lines and genotyping. six3a vu129 mutation was identified by TILLING (Targeting Induced Local Lesions IN Genomes) as previously described 28,29 . six3b vu87 mutation has been described 21 . six3b vu87 and six-3a vu129 fish were maintained in AB and TL backgrounds. Genotyping of six3a vu129 : a 176 bp fragment was amplified from six3a gene using forward primer 5′ -AGTTTCCCCTGCCTAGAACC-3′ and reverse primer 5′ -AAACCAATTTCCGACCTGTG-3′ . After digestion with AleI, the mutant allele is cut into 80 bp and 96 bp fragments whereas the wild-type (WT) allele remains uncut. six3b vu87 was genotyped by amplifying a 168 bp fragment using forward primer 5′ -TCAACAAGCACGAGTCCATC-3′ and reverse primer 5′ -GCAGCTTCTCTGCTTCTTGG. After digestion with MseI, the mutant allele is cut into 65 bp and 103 bp fragments whereas the WT allele remains uncut. Generation of Tg(rx3:Kaede)huj8 line: Kaede coding sequence from pKaede-S1 (Amalgaam) was cloned into pENTR1A (Invitrogen) to generate pME-Kaede, which was used for inserting Kaede sequence by Gateway LR recombination downstream of rx3 promoter sequence in rx3-destination plasmid 65 . Tg(rx3:Kaede) fish were generated by co-injection of rx3:Kaede DNA and Tol2 transposase synthetic RNA as described (Kawakami et al. 2004). Founder fish were identified by transgene expression in their progeny.
All experiments were approved by the Hebrew University Authority for Biological and Biomedical Models. The methods were carried out in accordance with the approved guidelines.
RNA injection. pCS2-six3a: six3a coding sequence was PCR-amplified and cloned into BamHI and XbaI sites of pCS2+ vector. pCS2-six3a-L183S: L183S mutation was introduced by site directed mutagenesis into pCS2-six3a. Sequence was confirmed for both constructs and synthetic capped RNA was prepared by NotI digestion and transcription with SP6 polymerase (mMESSAGE mMACHINE, Ambion).
Visual motor response (VMR) assay. Automatic quantification of larval movement was performed as previously described 36 . Briefly, at 6 dpf, six3a;six3b double-mutant and WT larvae were placed, individually, in 48-well plates in an observation chamber of DanioVision Tracking System (Noldus Information Technology, Waningen, Netherland). After 3 hours of light adaptation, larvae were subjected to 4 intervals of 30 min light/30 min darkness. In all experiments, larvae were subjected to locomotion analyses between 12:00 to 4:00 PM in a sound-and temperature-controlled (26 °C) behavioral testing room. Locomotor activity was tracked and analyzed using Ethovision 8.0 software (Noldus Information Technology). Described results are combined from two independent experiments.
In situ hybridization, immunohistochemistry and histology. Whole-mount in situ hybridization (WISH) using riboprobes was performed according to standard protocols 66 . BM Purple (Roche) was used as color substrate.
For histology, embryos were fixed in 4% paraformaldehyde (PFA) overnight at 4 °C, washed with PBT, dehydrated in EtOH series and embedded in JB4 resin (Polysciences, Inc.) according to manufacturer's instructions. 4 μ m sections were cut with LKB8800 Ultratome III microtome and stained with methylene blue-azure II.

Photoconversion and imaging.
To block pigmentation when imaging embryos older than 24 hpf, embryos were raised from 22 hpf in the presence of 0.003% N-Phenylthiourea (PTU; Sigma-Aldrich #P7629). Live embryos were anaesthetized using tricaine and mounted in 0.5% low-melting-point agarose (SeaPlaque, Lonza). For photoconversion, images of eye and OS region were acquired by minimal exposure using a 488 nm laser. A selected region of interest (ROI) was exposed to 405 nm laser scanning until all green fluorescence was converted to red, as determined by imaging with 488 nm and 555 nm lasers. The embryos were released from agarose and incubated at 28.5 °C in the dark until imaged again.
Images were acquired using Zeiss LSM700 confocal microscope and Axio Imager.M2 compound microscope or with Discovery.V8 stereoscope and AxioCam MRc digital camera (Zeiss). Microscope objective used were 40 × 1.0 NA water objective or 25 × 0.8 NA or 10 × 0.3 NA. Images were exported as JPEG or TIFF files using ZEN 2009 LE software (Zeiss) and figures were assembled using Adobe Photoshop CS4 software.