Six3 inactivation reveals its essential role for the formation and patterning of the vertebrate eye

Vertebrate eye development is initiated at the end of gastrulation with the determination of the retina anlage in the anterior neuroectoderm. Subsequently, the initially single eye field splits into two symmetric retinal primordia that evaginate laterally from the forebrain and become proximodistally patterned into optic stalk and retina (Chow and Lang, 2001; Jean et al., 1998).

Members of the evolutionarily conserved Six and Pax families of transcription factors (Bernier et al., 2000; Cheyette et al., 1994; Loosli et al., 1998; Oliver et al., 1995; Quiring et al., 1994; Toy et al., 1998; Zuber et al., 1999) share a temporally and spatially overlapping expression in the developing retina. Loss- and gain-of-function data indicate a key role for Pax6 in eye development (reviewed by Gehring and Ikeo, 1999). Both Six3 and Pax6, when overexpressed, mutually activate each others expression resulting in the formation of ectopic eye structures, suggesting also for Six3 a role in the establishment of retinal identity (Chow et al., 1999; Kobayashi et al., 1998; Loosli et al., 1999).

The transcription factor Six3 contains a homeodomain and a Six domain (Oliver et al., 1995; Toy et al., 1998), and is specifically expressed in the anterior neuroectoderm including the eye field and specific parts of the abutting surface ectoderm in all vertebrates analyzed so far (Bovolenta et al., 1998; Granadino et al., 1999; Loosli et al., 1998; Oliver et al., 1995; Seo et al., 1998; Zhou et al., 2000). In addition, at subsequent stages of development Six3 is expressed in the developing retina. Overexpression of Six3 results in expanded and ectopic retinal primordia in medaka (Loosli et al., 1999) and Xenopus (Bernier et al., 2000), and in an enlarged forebrain in zebrafish (Kobayashi et al., 1998). However, it remains elusive how and to what extend Six3 contributes to this process as experimental loss-of-function data for Six3 have not been reported so far.

We used a morpholino oligo-based gene knock-down approach (Nasevicius and Ekker, 2000) and gain-of-function analysis in the teleost medaka (Oryzias latipes) to determine the role of Six3 in the initial establishment of the retinal primordia. Inactivation of Six3 leads to the complete absence of forebrain and eyes, indicating the key role of Six3 in patterning of the anterior neural plate and establishment of retinal identity.

We further established hypomorphic conditions, under which retinal precursor cells still form, to address the later function of Six3, identifying a role of Six3 in proximodistal patterning of the optic vesicle. Our epistasis analysis establishes a genetic network and shows that Six3 function in proximodistal patterning is mediated by the regional specification gene Vax1. Strikingly, the hypomorphic phenotypes closely resemble humans heterozygous for mutations in Six3, who suffer from holoprosencephaly, microcephaly and microphthalmia.

Wild-type Oryzias latipes from a closed stock at EMBL-Heidelberg were kept as described (Köster et al., 1997).

Morpholinos (Gene Tools, LLC, OR) targeted against Six3, Pax6 and GFP were dissolved in 1×Yamamoto ringer (Yamamoto, 1975). Morpholino dilutions were injected into one cell of medaka embryos at the two-cell stage resulting in an even distribution. Morpholinos targeted against GFP (MoGFP: 5′-CAGCTCCTCGCCCTTGCTCACCATG-3′), which did not affect embryonic development, were injected at 0.06 mM. Binding specificity of MoSix3 was shown using an RNA containing the respective morpholino target sequence fused to the GFP-coding region (MoSix3t-GFP: 5′-GTCCGCCGCCTGCCTTCCCACACGC-GFP-3′). Co-injection of MoSix3t-GFP RNA (130 ng/μl at one-cell stage) and MoSix3 (0.05 mM at the two cell stage together with the red fluorescent lineage tracer dye rhodamine dextran (RD) (to follow its distribution) completely blocks GFP expression. However, GFP expression was not affected when five base pairs of the target sequence were exchanged (MoSix3ct-GFP: 5′-GaCCGCCGaCTGCCTaCCCtCACcC-GFP-3′). To test the Pax6 morpholino (0.06 mM) specificity, MoPax6t-GFP (5′-ATGATGCAGAACAGTCACAGTGGCG-GFP-3′) and MoPax6ct-GFP (5′-ATaATGCAaAACAGTaACAaTGGaG-GFP-3′) constructs were used. Co-injection of MoPax6 with MoPax6t-GFP RNA blocks GFP expression, whereas co-injection with MoPax6ct-GFP does not affect GFP expression.

In vitro transcription of mouse Six3 RNA and injection at 100 ng/μl into medaka embryos was performed as described (Loosli et al., 1999).

Whole-mount in situ hybridization was performed using digoxigenin and fluorescein labeled RNA riboprobes as described (Loosli et al., 1998). Vibratome sections were cut following standard procedures (Loosli et al., 1999).

To determine the role of Six3 in patterning of anterior neuroectoderm, we interfered with Six3 function by injecting a morpholino oligonucleotide (Nasevicius and Ekker, 2000) directed against Six3 mRNA (MoSix3) that specifically blocks translation. Control experiments show that MoSix3 specifically and stably inhibits its target in a concentration-dependent manner (Fig. 1 and Materials and Methods). Sequence specificity of MoSix3 was shown using RNA containing the respective morpholino target sequence fused to the GFP-coding region (MoSix3t-GFP). Co-injection of MoSix3t-GFP RNA and MoSix3 completely blocks GFP expression (n=22/24; Fig. 1A-C). Conversely, GFP expression was not affected when five base pairs of the target sequence were exchanged (MoSix3ct-GFP) (n=36/36; data not shown) indicating the specificity of MoSix3. The morpholino induced phenotype can be specifically titrated by co-injection of RNA containing the morpholino target site: co-injection of MoSix3 and its target MoSix3t-GFP did not affect the embryonic morphology, whereas co-injection of MoSix3 and MoSix3ct-GFP resulted in the Six3 morphant phenotype (data not shown). Finally, the Six3 morphant phenotype can be rescued by co-injection of mouse Six3 RNA (data not shown).

Analogous control experiments have been performed for the Pax6 morpholinos. Co-injection of MoPax6 with MoPax6t-GFP RNA blocks GFP expression (n=19/23; Fig. 1D-F), whereas co-injection with MoPax6ct-GFP does not affect GFP expression (n=25/25).

These control experiments show that both the Six3 and Pax6 morpholinos interfere specifically with their targets in a concentration-dependent manner and thus allow us to study the role of these genes in eye development in the loss-of-function situation, using either hypomorphic or amorphic conditions.

Inactivation of Six3 by injection of high amounts of MoSix3 (Six3 morphants) causes an eyeless phenotype and the deletion of the forebrain. In these embryos already at late neurula/early somitogenesis stages morphologically abnormal cells were detected in the presumptive forebrain, indicating ectopic cell death (Fig. 2A,B). TUNEL labeling (Winkler et al., 2000) in Six3 morphants at the late gastrula stage revealed a large number of apoptotic cells in a region overlapping with the Six3 expression domain in the anterior neuroectoderm (Fig. 2C-E). Thus, in the absence of Six3 function, these cells undergo ectopic cell death resulting in the absence of forebrain and eye structures. This shows that Six3 function is essential for the establishment and maintenance of the anterior neuroectodermal identity including the presumptive forebrain and the retina anlage, consistent with the early expression pattern at the late gastrula/early neurula stage.

Six3 continues to be expressed also in the developing retina and forebrain, suggesting additional later functions. In line with this, Six3 overexpression induces the formation of patterned optic cups at ectopic locations in the brain (Loosli et al., 1999). Using molecular markers specific for the proximal region [optic stalk, Vax1 (Winkler et al., 2000)] and distal region [neuroretina, Rx2 (Loosli et al., 1999)] of the developing optic cup, we found that Six3 overexpression by mRNA injection results in stalk cells positive for Vax1 in the midbrain abutting the Rx2 expressing ectopic neuroretina (Fig. 2F-H). The formation of a complete proximodistally patterned ectopic optic primordium hints at an important role of Six3 in later aspects of eye patterning. In severe Six3 morphants, however, this later role cannot be examined, due to the complete absence of these structures.

We circumvented this problem by using lower MoSix3 concentrations that generate hypomorphic phenotypes in a concentration-dependent manner (Table 1 and Fig. 2I-K). Mildly affected Six3 morphants exhibit mild cyclopia, small eyes and a reduction of proximal regions of the pigmented retinal epithelium (PRE). The intermediate phenotype is characterized by severe cyclopia and the loss of forebrain structures. TUNEL analysis of embryos at the late gastrula stage revealed apoptotic cells scattered over the entire anterior neuroectoderm. When compared with embryos injected with higher amounts of MoSix3, the abundance of apoptotic cells was lower. We observed a close correlation between the abundance of ectopic cell death at the late gastrula stage (intermediate, 32%; mild, 41%; n=79) and the phenotypic severity at subsequent stages of development (intermediate, 33%; mild, 40%; n=159; compare with Table 1). Mutations in the human SIX3 gene result in holoprosencephaly (Granadino et al., 1999; Wallis et al., 1999). In these heterozygous individuals, this dominant disorder causes craniofacial malformations that include cyclopia, microphthalmia and midfacial defects. Thus, the medaka Six3 morphant faithfully phenocopies reduced Six3 activity in humans, indicating that this human holoprosencephaly disorder can be studied using the morpholino-based medaka model.

We analyzed mildly affected morphants at early somitogenesis stages using specific molecular markers. The homeobox gene Vax1 is expressed in the developing ventral forebrain and optic stalk (Hallonet et al., 1998; Winkler et al., 2000) and loss of function analysis in mouse show its essential role for the normal development of these proximal eye structures (Hallonet et al., 1999). In mildly affected Six3 morphants, Vax1 expression was completely lost. This highlights a severe reduction of proximal eye structures (n=14/16; Fig. 3A,B) and an essential regulatory input of Six3 on Vax1 expression. On the other hand, the two expression domains of the distal retina marker Rx2 were fused anteriorly and reduced in size (n=15/25; Fig. 3C,D) consistent with the observed microphthalmia (Fig. 2J). The finding that the proximal region of the retinal primordium is more sensitive to reduced Six3 levels than the distal region indicates a role of Six3 in proximodistal patterning, in part mediated by the transcriptional regulation of Vax1.

In these Six3 morphants the telencephalic Emx1 expression domain is significantly smaller and non-overlapping with the Rx2 expression domain (Fig. 3C,D), consistent with the loss of forebrain structures at later stages of development (Fig. 2K) and the microcephaly observed in individuals with SIX3 mutations (Wallis et al., 1999).

As retinal precursor cells are still present in mildly affected morphants, as visualized by Rx2 expression, we were able to study the role of Six3 in the genetic network that underlies early eye development. Consistent with the demonstrated post-transcriptional effect of morpholinos on gene expression (Nasevicius and Ekker, 2000), initial Six3 transcript levels and distribution are not affected in Six3 morphants (Fig. 3G,H). It has been suggested that Six3 acts in a regulatory feedback loop (Loosli et al., 1999). We demonstrate this at later stages (six-somite stage), when expression of Six3 was not detectable in the retinal precursor cells of Six3 morphants, whereas weak expression was observed in the region of the developing hypothalamus (Fig. 3I,J). This shows that Six3 expression in retinal precursor cells crucially depends on Six3 activity, confirming the suggested role of Six3 in maintenance of its own expression. However, for its hypothalamic expression Six3 receives additional independent input at this stage.

Proliferation and morphogenesis of the retinal primordium requires the activity of the homeobox gene Rx3 (Loosli et al., 2001). Genetic epistasis analysis in medaka indicated a regulatory input of Six3 to Rx3 expression. To evaluate the contribution of Six3, we analyzed Rx3 expression in Six3 morphants. Notably, while retinal Rx3 expression initially appears unaffected, reduced Six3 function results in the loss of Rx3 expression at the four-somite stage (Fig. 3E,F). Thus, Rx3 receives an initial input that is at least partially independent of Six3 and in a second phase depends entirely on Six3 function in the developing optic vesicle. Rx3 remained expressed only in the region of the forming hypothalamus, where Six3 and Rx3 expression domains do not overlap. The observed microphthalmia can in part be explained by the loss of Rx3, resulting in a reduced proliferation of retinal progenitor cells.

Overexpression of the evolutionarily conserved transcription factor Pax6 results in ectopic Six3 expression and eye formation (Chow et al., 1999). On the other hand, ectopic retina formation in response to Six3 overexpression in fish is preceded by ectopic Pax6 expression (Loosli et al., 1999), indicating that Six3 and Pax6 can crossregulate each other during eye formation. Similar to Six3, the expression of Pax6 is not eye specific, but their respective domains overlap in the retina anlage at the late gastrula stage (Loosli et al., 1998; Oliver et al., 1995). It has been argued that the region of overlapping expression defines the retina anlage and its formation requires their crossregulatory interaction (Loosli et al., 2001). In Six3 morphants, up to the onset of somitogenesis, Pax6 transcript levels were unaltered (Fig. 3K,L and data not shown). At the six-somite stage, however, the Pax6 expression level in the retinal precursor cells was significantly reduced, but unaffected in the presumptive dorsal forebrain where Six3 is not expressed (Fig. 3M,N). The finding that Pax6 expression in retinal precursor cells is sensitive to Six3 activity corroborates their regulatory interaction (Bernier et al., 2000; Chow et al., 1999; Loosli et al., 1999). Although at early stages both genes appear to be independent of Six3 function, the strength of this regulatory interaction increases during the development of the retinal primordium.

We used the morpholino-based gene knockdown approach to interfere with Pax6 function. The specific and stable inhibitory effect of MoPax6 on its target was verified in control experiments (see above, Fig. 1 and data not shown). Pax6 morphants exhibit a ‘small eye’ phenotype in a concentration-dependent manner (Table 2 and Fig. 4A,B). In situ hybridization analysis of Pax6 morphants injected with either 0.03 mM (Fig. 5D,G,J) or 0.06 mM MoPax6 (data not shown) revealed that, at the early stages of development, Pax6 interference does not affect the expression of all molecular markers, for example, Pax2, Six3, Pax6 or Rx genes (Fig. 5D,G,J and data not shown). This is in good agreement with the analysis of the respective murine homologs in small eye Pax6^–/– mutant mice (Bernier et al., 2001).

To address the cooperation of Six3 and Pax6 in optic vesicle formation, we co-injected MoSix3 and MoPax6. A MoPax6 concentration was used that, when injected alone, results in a very low frequency of the small eye phenotype (Fig. 4E). This was combined with MoSix3 at a concentration that causes predominantly a weak hypomorphic phenotype when injected alone (Fig. 4E). The co-injection results in a significant change of the ratios of the two observed phenotypes. In MoSix3-injected embryos, less than one third of the affected embryos exhibit the intermediate phenotype, whereas co-injection results in a more than twofold increase (Fig. 4E). The co-injection of MoSix3 and MoPax6 leads to phenotypes that are morphologically indistinguishable from morphant phenotypes caused by higher concentrations of MoSix3 alone (compare Fig. 2J with Fig. 4C and Fig. 2K with Fig. 4D and data not shown). Although MoPax6 alone only mildly affects eye formation, it potentiates the MoSix3 interference. This indicates that the genetic network underlying early retina formation and the subsequent patterning of the eye depends on the combined regulatory input of Six3 and Pax6.

To examine this interaction in more detail, we analyzed the expression of molecular markers upon co-interference. Whereas the expression of Rx2, Rx3, Vax1 and Emx1 were similarly reduced (data not shown), we observed a differential response of the transcription factor Pax2 in morphologically indistinguishable single and double morphants. Pax2 expression in the retina was not detectable in single injected Six3 morphants (Fig. 5A,B). In MoSix3/MoPax6 co-injected morphants however, Pax2 is still expressed (Fig. 5C,E). Even in the more severely affected double morphants, Pax2 expression levels were not affected in the remaining retinal tissue. In addition, in Pax6 morphants Pax2 expression remained unchanged (Fig. 5C,D). Thus Pax2 expression is sensitive only to Six3 interference but not to Pax6 interference at early developmental stages. This differential molecular response indicates that Six3 and Pax6 have partially non-overlapping sets of target genes, consistent with the different phenotypes observed in the single Six3 and Pax6 morphants.

In addition, the expression of Six3 and Pax6 was differentially affected comparing single Six3 and Six3/Pax6 double morphants. Although Six3 expression at the six-somite stage is strongly reduced in single MoSix3-injected embryos (Fig. 3I,J), its expression level remains unaffected in Pax6 or double morphants (Fig. 5F-H). However, retinal Pax6 expression at this stage was more severely reduced in double morphants than in Six3 morphants (Fig. 3M,N, Fig. 5I,K). Pax6 knockdown alone had no effect on its own expression (Fig. 5I,J). Thus, Pax6 expression in retinal precursor cells depends on the combined regulatory input of Pax6 and Six3. Conversely, Six3 expression is less sensitive to this combined input indicating a mainly Six3-dependent regulation. Consistent with our results, it has been reported that Six3 is still expressed in the retinal primordium of Pax6^–/– mouse embryos (Bernier et al., 2001).

Our experiments demonstrate that Six3 is essential for the formation of a discrete domain within the anterior neuroectoderm. In the absence of Six3 function, cells in the Six3 expression domain undergo apoptosis resulting in the absence of forebrain and eye. Conversely, overexpression of Six3 results in retinal hyperplasia, indicating that one function of Six3 is the control of proliferation in the presumptive retinal cells. In addition, Six3 functions in the determination of the naive anterior neuroectoderm as loss of function results in the absence of the respective structures, while ectopic Six3 expression leads to their ectopic formation.

Interference with gene function on the translational level using the gene knockdown technique offers advantages compared with other loss-of-function approaches. They allow to generate hypomorphic conditions, under which feedback loops can be uncovered. Our analysis provides direct evidence that Six3 acts in a regulatory feedback loop on its own expression in vertebrates. Initially, Six3 expression in the retina is unaffected but subsequently becomes crucially dependent on Six3 function. However, Pax6 has no influence on Six3 expression at the early stages of eye development. This finding is consistent with Pax6^–/– mutant mice where Six3 is still expressed. Interestingly, the genetic network underlying gene regulation in the forming hypothalamus appears to be different from what we observed in the retina. Here, Six3 and Rx3 expression patterns do not overlap at early somitogenesis stages. The expression of both genes is neither affected in Six3 single nor Six3/Pax6 double morphants. It will be interesting to determine how the early expression of these genes is regulated and to what extend these genes are involved in the development of the hypothalamus.

Our gene knockdown data show that Six3 and Pax6 interact genetically at early stages of eye development. However, the morphological and molecular consequences of the loss of Six3 function are more severe. Pax6 acts on gene expression including its own only in cooperation with Six3. As seen for Pax6 morphants, small eye Pax6^–/– mouse embryos initially form optic vesicles (Hill et al., 1991; Matsuo et al., 1993; Quinn et al., 1996; Ton et al., 1991), indicating that also in mouse Pax6 is not required for the formation of the retina anlage in the neuroectoderm.

TUNEL-labeling experiments at the late gastrula stage revealed that injection of lower amounts of either MoSix3 or MoSix3/MoPax6 results in a reduced and variable degree of apoptosis as compared with high amounts of MoSix3 (data not shown). In both, single- and double-injected morphants, cells that undergo ectopic cell death were distributed over the entire anterior neural plate under all different hypomorphic conditions analyzed, suggesting a similar mechanism acting in both, single and double morphants.

In both, single- and double-injected morphants, the distribution of apoptotic cells at the late gastrula stage shows no regionalization, suggesting that proximodistal patterning is a later function of Six3. The close correlation of the abundance of cell death at the late gastrula stage and phenotypic severity at subsequent stages suggests that the early role of Six3 (determination) and its later function (patterning) are not completely independent. Future experiments will address to what extent the mechanisms for determination and patterning are coupled.

Following its role in the determination and formation of the retina anlage, Six3 functions in proximodistal patterning of the optic vesicle. Six3 activity regulates the expression of the regional specification gene Vax1, which is required for the formation of ventral forebrain and proximal eye structures. Graded loss of Six3 function shows that the distally expressed genes Rx2 and Pax6 are less sensitive than Vax1 underscoring the proximodistal patterning activity of Six3. In addition, at this stage Six3 controls proliferation and morphogenesis by regulating the expression of Rx3, which is essential for these processes in the developing optic vesicle (Loosli et al., 2001). Future experiments will aim to identify factors mediating the differential activity of Six3 along the proximodistal axis.

We thank A. Ephrussi, P. Rørth, M. Treier, D. Arendt, L. Poggi, C. Grabher and all members of the Wittbrodt laboratory for critically reading the manuscript and helpful suggestions; and A. Krone for technical assistance. This work was supported through grants from the Deutsche Forschungsgemeinschaft and the HFSPO to J. W.