Medaka eyeless is the key factor linking retinal determination and eye growth

The vertebrate eyes form from a single eye field, an anterior ectodermal territory containing a retina anlage and an abutting lens competence field. Under the influence of midline signals, this field splits into two bilateral regions each containing a neuroectodermal retina anlage and an ectodermal lens anlage. During neurulation, the retina anlage evaginates laterally from the forebrain, giving rise to the optic vesicles. These in turn invaginate distally to form the optic cups with neural retina (NR) and retinal pigmented epithelium (RPE) (Grainger, 1996; Jean et al., 1998). Several transcription factors that are required for optic vesicle development have been identified so far (Chow et al., 1999; Jean et al., 1998; Loosli et al., 1999; Porter et al., 1997). However, it is yet unknown in what manner these factors interact to initiate and control optic vesicle formation. What is the molecular link between the determination of the retina anlage and the formation of an optic vesicle?

The transcription factor Pax6 is an essential regulator for eye development conserved in evolution (Quiring et al., 1994; Walther and Gruss, 1991). Pax6 is expressed in the anterior neuroectoderm before optic vesicle formation in fish (Krauss et al., 1991; Loosli et al., 1998), frog (Hirsch and Harris, 1997), and chick (Li et al., 1994). Pax6^–/–Small eye mutant mice develop small optic vesicles, which degenerate during subsequent development (Hill et al., 1991; Hogan et al., 1986). Overexpression of Pax6 in Drosophila and Xenopus can lead to ectopic eye formation (Chow et al., 1999; Halder et al., 1995).

Six3 plays a role in specifying early retinal identity (Bernier et al., 2000; Loosli et al., 1999). This evolutionarily conserved transcription factor contains a homedomain and a Six domain (Cheyette et al., 1994; Seimiya and Gehring, 2000; Toy et al., 1998). It is specifically expressed in the anterior neuroectoderm including the eye field and specific parts of the abutting surface ectoderm in all vertebrates examined (Bovolenta et al., 1998; Granadino et al., 1999; Loosli et al., 1998; Oliver et al., 1995; Seo et al., 1998; Zhou et al., 2000). Overexpression of Six3 leads to expanded and ectopic retinal primordia in medaka (Loosli et al., 1999) and Xenopus (Bernier et al., 2000), and to an enlarged forebrain in zebrafish (Kobayashi et al., 1998). In a feedback loop, Six3 thereby induces its own expression, and that of Pax6 (Loosli et al., 1999). As Pax6 also induces Six3 (Chow et al., 1999), both factors activate each other (Loosli et al., 1999).

Vertebrate members of the Rx gene family are expressed in the early eye field, and, with the exception of medaka Rx2, which is exclusively expressed in the developing retina, also in medial regions of the prosencephalon (Casarosa et al., 1997; Chuang et al., 1999; Deschet et al., 1999; Furukawa et al., 1997; Loosli et al., 1999; Mathers et al., 1997; Ohuchi et al., 1999). Targeted inactivation of Rx in mouse leads to optic vesicle and forebrain deletions, indicating a function of Rx in the development of these structures (Mathers et al., 1997). Similar results have been reported for Xenopus embryos injected with a XRx1 dominant repressor construct (Andreazzoli et al., 1999). Conversely ectopic expression of XRx1 in Xenopus embryos triggers overproliferation of retinal tissue (Andreazzoli et al., 1999; Mathers et al., 1997, Zhang et al., 2000). Thus, the specific involvement of Rx genes in eye development with regard to the determination of the retina anlage and subsequent specification and morphogenesis of the retinal progenitor cells remains to be defined.

In the temperature-sensitive, early larval lethal medaka mutation eyeless (el) optic vesicles do not evaginate at the restrictive temperature (Winkler et al., 2000). This defect in morphogenesis leads to a complete lack of eyes. Temperature shift experiments indicate that gene activity is required before optic vesicle evagination.

We show that the gene affected by the eyeless mutation is Rx3, a member of the Rx gene family. We combine mutant analysis with gain-of-function approaches to place Rx3 function into the genetic network of early eye development. Rx3 is required to initiate and maintain optic vesicle morphogenesis and proliferation, but is dispensable for the formation of the retina anlage. These results provide a link between retinal determination, and the subsequent control optic vesicle morphogenesis and differentiation.

The el mutation was crossed into and kept in two different inbred genetic backgrounds (Cab and Kaga) and kept as closed stocks at EMBL (Winkler et al., 2000). Embryos were staged according to Iwamatsu (Iwamatsu, 1994). Sex averaged linkage analysis showed a close linkage between el and the pigmentation locus b (1.3 cM).

DNA of homozygous el hatchlings (three to five individuals pooled) or individual wild-type fish derived from brother-sister crosses of el/Kaga carriers was isolated (QIAamp kit, Qiagen). A 333 bp fragment was amplified using primers specific for the third exon of Rx3 (up, ACACTCCCTCATCTTTGCTCTCCTTC; low, TGTTCCTTGGCCTTCATCCTCA). PCR conditions were 35 cycles of 94°C for 30 seconds, 61°C for 30seconds, 72°C for 1 minute. PCR products were HaeIII digested and separated on 3% agarose gels, showing cleavage of a Cab-specific 190 bp fragment into 100 bp and 90 bp fragments in Kaga.

A neurula stage λZAP cDNA library was screened with a 440 bp Rx3 PCR fragment (Deschet et al., 1999). Positive clones were plaque purified, converted to pBluescript and sequenced (EMBL Accession Number, AJ298300)

Total RNA was isolated from two-somite stage homozygous mutant embryos and wild-type siblings (Sambrook et al., 1989) and reverse transcribed using an oligo-dT and a Rx3-specific primer (ATGTTCCATTCTGGGCGTCTCAG). A Rx3-specific primer pair (up, TTAGACAAATGTGGCTCCTGGGATCAGCTTCA; low, TGTTCCTTGGCCTTCATCCTCA) was used as follows: 35 cycles of 94°C for 45 seconds, 60°C for 1 minute, 72°C for 2 minutes. PCR products were analysed by gel electrophoresis.

A 440 bp medaka Tbx2 fragment and a 600bp fragment of Tbx3 were RT-PCR amplified from stage 18 RNA using degenerate primers (up, GAGGTIGAG GAYGAYCCIAARGT; low, CCIRTGTCYCTRAAICCYTTIGCRAA). PCR conditions were 5 cycles of 94°C for 1 minute, 53°C for 2 minutes, 72°C of 4 minutes, followed by 30 cycles with annealing at 58°C. PCR products were cloned (TopoTA, Invitrogen) and sequenced (EMBL Accession Numbers: OlTbx2, AJ298301; OlTbx3, AJ298302).

Genomic DNA of adult homozygous mutant el, heterozygous el carrier and wild-type fish was prepared. The DNA was digested over night, separated on a 0.7% agarose gel and transferred to HybondN⁺ (Amersham) by capillary transfer. Filters were hybridised with ³²P-labelled probes under high stringency conditions (Sambrook et al., 1989). Total RNA was isolated from wild-type embryos. Poly(A)⁺ RNA was enriched (dynabeads, Dynal). 10 μg of A⁺ RNA were separated on a formaldehyde agarose gel, blotted and hybridised (Sambrook et al., 1989).

A lambda FixII (Stratagene) genomic library was generated from genomic DNA of homozygous el-mutant fish. A phage covering the 3′ part of the mutation was isolated using the Rx3 cDNA as a probe under high stringency conditions. The remaining part of the mutant locus was cloned as PCR fragments, and the 5′ breakpoint of the insertion was isolated by splinkerette (Devon et al., 1995).

Bac and cosmid clones containing the Rx3 locus were isolated from arrayed libraries (library 756, 73 and 74, RZPD Berlin) probed with Rx3 cDNA. BAC and cosmid DNA was prepared following standard protocols. For injection the DNA was further purified on a Caesium chloride gradient (Sambrook et al., 1989). For the Rx3 rescue plasmid, an 11 kb SnaBI fragment of the Rx3 cosmid was cloned into pBluescript SK+ (Stratagene). The remaining cosmid was re-ligated resulting in the ΔRx3 plasmid. A frameshift was introduced into the second helix of the homeodomain by BspEI digestion of the Rx3 rescue plasmid and re-ligation after filling the ends with Klenow polymerase (Rx3ΔHB plasmid).

Plasmid DNA was injected into the blastomere of one-cell stage embryos from el carrier crosses. Injected embryos were kept under restrictive conditions. The following DNA concentrations were used for injection experiments: Rx3 BAC, 50 ng/μl; control BACs, 50 ng/μl; Rx3 cosmid, 40 ng/μl; Rx3 plasmid, 5 and 20 ng/μl; ΔRx3 plasmid, 45 ng/μl; Rx3ΔHB plasmid, 50 ng/μl and pBluescript, 50 ng/μl.

A 1.2 kb BamHI-BglII fragment of the medaka Rx3 cDNA cloned into pCS2+ was verified by sequencing and TNT transcription/translation (Promega). Linearized pCSmouseSix3, pCSmedakaSix3 (Loosli et al., 1999) and pCSmedakaRx3 plasmid DNA were in vitro transcribed and the purified RNA injected into one blastomere at the two-cell stage (Loosli et al., 1999). RNA concentrations in the injection solution were 200 ng/μl pCSmouseSix3, 200 ng/μl pCSmedakaSix3, when injected into embryos of el carrier crosses; 100 ng/μl, when injected into wild-type embryos; and 50 ng/μl pCSmedakaRx3. For controls, the respective concentration of hGFP RNA was injected. Injected embryos from el carrier crosses were kept at the restrictive temperature (18°C).

Whole-mount in situ hybridisation using antisense riboprobes and vibratome sections of embryos were performed as described (Loosli et al., 1998).

The el mutation, initially isolated in a strain from the Southern population of medaka, was crossed into and kept in the ‘Cab’ inbred strain (Winkler et al., 2000). Crossing of el carrier fish into the Kaga strain revealed a close linkage of el and the pigmentation locus b. The Kaga strain is homozygous for the B allele that results in darkly pigmented melanophores. The Cab background carries the recessive b allele that leads to unpigmented melanophores. Analysis of the offspring of el/Kaga intercrosses resulted in 32 recombinant el embryos (el/el,b/B) in 2422 meiosis. This corresponds to a genetic distance of 1.3 cM and thus places el on linkage group 12 (LG 12) (Naruse et al., 2000) in close proximity to the b locus.

el is a temperature-sensitive mutation. Optic vesicles of homozygous mutant embryos do not evaginate at the restrictive temperature (18°C) and subsequent differentiation is perturbed. At higher temperatures (28°C) in 52% of the el-mutant embryos optic vesicles of variable size form (Winkler et al., 2000). Temperature shift experiments showed that el activity is required at late gastrula/early neurula stages before optic vesicle evagination. We therefore examined genes that are expressed in the retina anlage before optic vesicle evagination. The expression of the retina determination genes Six3 and Pax6 is not affected in the retinal primordia of el-mutant embryos (Winkler et al., 2000). These genes have been mapped to different linkage groups (i.e. Pax6, LG 3; Six3, LG 19 (Naruse et al., 2000)), excluding Pax6 and Six3, as candidates for el.

In the anterior neuroectoderm, the homeobox gene Rx3 is expressed from late gastrula stages onwards (Deschet et al., 1999). The initially narrow, single expression domain widens and is subsequently split into two domains lateral to the forming axis (Fig. 1A). During neurulation and early somitogenesis stages, Rx3 is expressed in the anterior ventral prosencephalon, the optic vesicles and the optic stalk, which connects these tissues (Fig. 1C). Expression in the retina and optic stalk is downregulated at later somitogenesis stages, such that Rx3 expression finally is restricted to the hypothalamus. In the differentiating retina, Rx3 is expressed in the inner nuclear layer (INL). In situ hybridisation showed that in el-mutant embryos, Rx3 expression in all its wild-type expression domains is completely absent at all stages at the restrictive temperature (Fig. 1A-D). This indicates that either Rx3 or an upstream regulator is affected by the el mutation.

To examine whether Rx3 itself or a regulator of Rx3 expression is affected in the el mutation we examined whether Rx3 maps to linkage group 12 and whether the mutation co-segregates with an Rx3 specific polymorphism. We took advantage of the highly polymorphic genetic backgrounds of the inbred Cab and the Kaga strains. A PCR amplified 333 bp genomic Rx3 DNA fragment contains a polymorphic HaeIII restriction site (Fig. 2A). This cleaved amplified polymorphic sequence (CAPS) was used to determine the chromosomal location of the Rx3 gene and to examine a potential linkage between Rx3 and the el mutation.

Cab/Kaga back-cross analysis locates the Rx3 gene to linkage group 12 in close proximity to the b-locus (Fig. 2D). el carrier fish of the Cab strain, in which the mutation has originally been identified, were crossed to Kaga fish and the wild-type, and mutant progeny of the resulting el/Kaga carrier intercrosses were examined for co-segregation by CAPS analysis. The Cab-specific HaeIII polymorphism always co-segregated with the el phenotype in all 532 meiosis analysed (Fig. 2). Thus, the genetic distance of the Rx3 gene and the el mutation is less than 0.19 cM, strongly suggesting that the mutation affects the Rx3 gene proper.

To further analyse the nature of the el mutation we cloned a full-length Rx3 cDNA. Genomic Southern blots probed with a 600 bp 5′ fragment of the Rx3 cDNA suggested an insertion in the Rx3 locus of el-mutant fish (Fig. 2B). Rx3 containing BAC and Cosmid clones were isolated from respective wild-type libraries. The genomic organisation of the Rx3 locus was examined by sequence analysis and Southern blot hybridisation. The Rx3 open reading frame of 876 bp is split into three exons and spans a genomic region of about 2.9 kb (Fig. 2D). The predicted length of the RX3 protein is 292 amino acids. Sequence comparison with other homeobox genes of the Rx subclass shows that the medaka RX3 protein shares highest homology with zebrafish RX3. The RX3-specific octapeptide and the homeodomain are 100% identical and the conserved region in the C-terminal region (OAR) (Furukawa et al., 1997) differ in only one amino acid between medaka and zebrafish RX3. However, the medaka paralogue RX2 differs by two amino acids in the octapeptide, one in the homeodomain and one in the OAR, compared with medaka RX3. In contrast to zebrafish, where three Rx genes have been cloned, we detected two paralogues in medaka.

The homeodomain is encoded by the 3′ part of exon 2 and the 5′ part of exon 3 (Fig. 2D,E). The mutant locus harbours an insertion in the intron intervening exon 2 and 3. The inserted DNA is middle repetitive as indicated by Southern blot hybridisation (Fig. 2C). The insertion of at least 13 kb has no significant homology to any known sequence. Apart from the insertion in intron 2, no further changes were detectable in the mutant Rx3 locus. The open reading frame is unaffected and thus the locus has the potential to express a wild-type RX3 protein.

The tight linkage of el and the pigmentation locus b was used to genetically mark the el mutation. The homozygous el-mutant offspring of an el/+,b/B carrier cross lacks the darkly pigmented melanophores with the exception of less than 1.3% recombinants (Fig. 3A,B). The pigmentation becomes first visible at early somitogenesis stages. At the restrictive temperature, all mutant embryos lack eyes and do not form retinal pigmented epithelium (RPE), thus the mutant phenotype is fully penetrant and the expressivity not variable. Control injections into fertilised eggs affect neither penetrance nor expressivity (Fig. 3E). Mutant embryos injected with either the Rx3 BAC or the Rx3 cosmid form eyes of wild-type morphology in 25% and 39% of the cases, respectively (Fig. 3E).

The Rx3 plasmid contains the Rx3 locus (Fig. 2D) and rescues the el phenotype in 37% of the cases, while the remaining 8.2kb of the Rx3 cosmid (ΔRx3 cosmid) have no rescuing activity (Fig. 3E). Northern blot analysis using the Rx3 plasmid as a probe detects a single transcript of 1.8kb that corresponds to the Rx3 mRNA (not shown) indicating that only the Rx3 gene is expressed in the relevant time window from the Rx3 plasmid. 50% of the rescued homozygous mutant embryos (n=8) identified by the lack of darkly pigmented melanophores show wild type morphology and behaviour (Fig. 3). This rescue is coupled with restored Rx3 expression in mutant embryos (Fig. 4A,B). The finding that the rescued Rx3 expression is confined specifically to the wild type expression domains indicates that all regulatory elements required for wild type expression are contained within the Rx3 plasmid. To examine whether the rescue activity depends on a functional Rx3 protein we introduced a frameshift in exon 2 of the Rx3 plasmid. This results in a deletion of helices 2 and 3 of the homeodomain and the entire C-terminal portion (Rx3ΔHB plasmid, Fig. 2D). The frameshift completely abolishes the rescue activity (Fig. 3E) without affecting its expression (Fig. 4C,D) demonstrating that the rescue depends on the expression of a functional RX3 protein. On the other hand, expression from the rescue construct does not require Rx3 activity (compare Fig. 4C with Fig. 4D), indicating that in contrast to Pax6 and Six3 functional RX3 protein is not required for its own expression.

The T-box-containing transcription factors Tbx2 and Tbx3 are expressed in the developing optic vesicles, ventral forebrain and the otic vesicles starting at early somitogenesis stages in wild-type medaka (Fig. 4H,K). In el-mutant embryos, however, cells of the retinal primordia do not express Tbx2 and Tbx3, whereas the other expression domains are not affected (Fig. 4I,L). Rescued presumptive retinae of mutant embryos express both Tbx2 and Tbx3 (Fig. 4G,J). This indicates that retinal Tbx2 and Tbx3 expression depend on Rx3 activity and that the Rx3 rescue plasmid can provide also this aspect of Rx3 function. The paired like homeobox genes Vsx1 and Vsx2 are expressed in the differentiating retina (Winkler et al., 2000). In the retinae of rescued mutant embryos, expression of Vsx1 and Vsx2 that is lost in el-mutant embryos, is specifically restored (Fig. 4M-P).

Thus, the molecular and morphological analysis of rescued embryos shows that the Rx3 plasmid fully rescues evagination of the optic vesicle and subsequent differentiation of the retina, substantiating that the el-mutant phenotype is caused by the intronic insertion in the Rx3 gene.

At the restrictive temperature (18°C), Rx3 is not expressed in el-mutant embryos (Fig. 1B,D). At the permissive temperature (28°C), however, we found weak Rx3 expression at late neurula to early somitogenesis stages in 56% of the mutant embryos (Fig. 1F). Under these conditions, the mutant phenotype is variable, such that 52% of el-mutant embryos raised at the permissive temperature form small eyes. Thus, Rx3 expression in el-mutant embryos kept at the permissive temperature closely correlates with the variable phenotype. Furthermore, RT-PCR analysis shows that at the permissive temperature a correctly spliced transcript is present in el-mutant embryos (Fig. 1G). Thus, the mutation affects the transcription of the el locus in a temperature-sensitive way, such that the el mutation is amorphic at the restrictive temperature and hypomorphic at the permissive temperature.

Taken together, three lines of evidence indicate that the locus affected in the el mutation is the Rx3 gene: (1) genetic mapping and the detection of an intronic insertion in the Rx3 gene correlate; (2) the expressivity of the phenotype correlates with Rx3 expression levels; and 3) the wild-type Rx3 gene has rescuing activity.

We had shown that overexpression of medaka or mouse Six3 leads to retinal hyperplasia and the formation of ectopic retinal primordia in the mid- and hindbrain (Loosli et al., 1999). Ectopic retina formation in response to Six3 or Pax6 overexpression in Xenopus and medaka fish is preceded by a cross-regulatory interaction of these genes (Bernier et al., 2000; Chow et al., 1999; Loosli et al., 1999). It has been suggested that this interaction is a prerequisite for retinal determination. We examined whether Rx3 functions in this regulatory interaction of Six3 and Pax6.

To address this, we first investigated whether ectopic Six3 expression in el-mutant embryos can relieve the transcriptional repression of the mutant locus at the restrictive temperature. At the two-somite stage, no Rx3 expression was detected in the el-mutant embryos injected with Six3 RNA (n=17, Fig. 5A,B). On the other hand, in all injected wild type siblings (n=67) Rx3 expression was detected (not shown). Therefore, Six3 overexpression does not relieve the transcriptional repression of the mutant Rx3 locus under restrictive conditions. This allows examining the function of Six3 in the absence of Rx3. The finding that optic vesicle evagination is not rescued in the Six3 injected mutant embryos suggests a role of Six3 upstream to or in parallel of Rx3 (Fig. 5A,B).

Using the el mutation, we analysed whether Rx3 activity is required for the crossregulatory interaction of Six3 and Pax6. Ectopic medaka Six3 expression in embryos that lack Rx3 results in an enlarged Pax6 expression domain (n=21/53, 39%) and in ectopic Pax6 expression in the presumptive midbrain (n=9/53, 17%) at early somitogenesis stages (Fig. 5C,D). Similarly, in injected wild-type siblings used as internal control, expanded as well as ectopic Pax6 expression was observed. Thus, ectopic Pax6 expression in response to Six3 overexpression does not require Rx3 activity.

Mouse Six3 overexpression results in the ectopic activation of the endogenous Six3 gene, indicating that Six3 functions in a regulatory feedback loop (Bernier et al., 2000; Loosli et al., 1999). To address whether this regulatory feedback loop of Six3 will also function in the absence of Rx3, we injected mouse Six3 mRNA at the restrictive temperature. We observed ectopic endogenous Six3 expression at the early somitogenesis stages in 22% of injected el-mutant embryos (n=6/27, Fig. 5E,F). Consistent with expanded Pax6 expression, endogenous Six3 expression was also expanded in 40% (n=11) of the injected mutant embryos (not shown).

Thus, the crossregulatory interactions of Six3 and Pax6, as well as the regulatory feedback loop of Six3 are independent of Rx3. The finding that Rx3 is not required for these regulatory interactions preceding the formation of ectopic retinal primordia indicates that Rx3 is not essential for retina determination.

To investigate whether Rx3 is required for Six3 mediated ectopic retina formation, we further analysed the effect of Six3 overexpression in the el-mutant background. As a specific retina marker we used Rx2 that in wild type embryos is exclusively expressed in presumptive neural retina.

In el-mutant embryos at early somitogenesis stages Rx2 is expressed indicating that retinal determination is not affected by loss of Rx3 activity (Fig. 5H). Furthermore, in the Six3 injected mutant embryos we found strongly enlarged Rx2 expression domains (n=12/43, 28%) at early somitogenesis stages (Fig. 5G). Importantly, in injected mutant embryos (n=4/45, 9%) ectopic Rx2 expression was detectable in the presumptive midbrain, as in the wild type siblings (n=10/104, 10%; Fig. 5G arrowhead). This shows that also in the absence of Rx3 activity, Six3 can mediate respecification of presumptive midbrain to a retinal fate. This is in good agreement with a role of Six3 in the determination of retinal primordia independent of Rx3, which acts subsequently in optic vesicle morphogenesis and proliferation.

At late gastrula stages (stage 16), Rx3 is expressed in a narrow domain across the anterior neuroectoderm. Using the forming axis as a morphological landmark, we found that Rx3 expression overlaps with both the Six3 and Pax6 expression domains at this stage (Fig. 6A-C). The crossregulatory interactions of Six3 and Pax6 in early retina determination are not affected by the absence of Rx3 activity in el-mutant embryos at the restrictive temperature, suggesting that Six3 and Pax6 expression do not depend on Rx3. To address whether Rx3 acts downstream or in parallel to Six3, we examined the effects of Six3 overexpression on Rx3 expression and vice versa.

Ectopic expression of mouse Six3 results in an expansion of the Rx3 expression domain (n=19/29, 65%) and in ectopic Rx3 expression (n=8/29, 27%) already at late gastrula/early neurula stages (Fig. 6D). At late neurula/early somitogenesis stages an expansion of the Rx3 expression domain (n=33/49, 67%) and ectopic Rx3 expression in the presumptive midbrain of medaka Six3 injected embryos (n=8/49, 16%) is observed (Fig. 6E). This strongly suggests that Six3 acts as an upstream regulator of Rx3 expression at these stages.

To test whether, in reverse, Rx3 can activate Six3 we overexpressed Rx3. Ectopic Rx3 expression does not affect Six3 expression at the time point when they are first co-expressed in wild-type embryos (late gastrula/early neurula stage) and at midneurula stages (not shown). Neither size nor morphology of the forebrain or optic vesicles is altered by Rx3 overexpression at these stages (not shown). After optic vesicle evagination, however, the optic vesicles of injected embryos are often enlarged as visualised by Six3 (n=13/36, 36%) and Pax6 (n=15/41, 37%) expression (Fig. 6G-I), without affecting their expression domains in the forebrain. The finding that Rx3 overexpression at somitogenesis specifically results in enlarged optic vesicles containing significantly more cells was confirmed by the expanded expression domain of Rx2 (n=34/101, 34%; Fig. 6I) that in wild-type embryos is expressed exclusively in the presumptive neural retina (Fig. 6L).

Thus, Six3 overexpression results in ectopic Rx3 expression. Rx3 overexpression, however, does not alter Six3 expression at stages before optic vesicle evagination, consistent with a role of Rx3 downstream of Six3. The enlargement of optic vesicles caused by higher numbers of retinal cell in response to Rx3 overexpression indicates a role of this gene in the control of proliferation in the optic vesicle.

In this study, we show that the eyeless mutant phenotype in medaka fish is caused by an intronic insertion in the homeobox gene Rx3, resulting in a transcriptional repression of the locus. We show that determination of the retina anlage occurs normally in the absence of Rx3 function, whereas subsequent morphogenesis and differentiation of the retinal primordia do not occur. Our studies indicate a role of Rx3 downstream of the retina determination gene Six3 in morphogenesis and growth control of the optic vesicle. Thus, Rx3 plays a pivotal role in the manifestation of a fate that has been laid down in the organ anlage by upstream players.

Furthermore, we show that the regulatory interactions of Six3 and Pax6 that precede the formation of ectopic retinal primordia do not depend on Rx3 activity. Our results provide novel insights into the regulatory interactions of key players involved in retina determination and subsequent steps of organ formation.

The detailed phenotypic analysis of genetically marked el-mutant embryos shows that the penetrance is complete and the expressivity not variable at the restrictive temperature. At the permissive temperature, however, the expressivity is variable, while the penetrance is not affected. This temperature-sensitive expressivity of the mutant phenotype is tightly correlated with the expression levels of Rx3 in the presumptive retina (Fig. 1). Furthermore, in our rescue experiments the degree of rescue correlates well with restored expression in the evaginating optic vesicles of el-mutant embryos (Fig. 4A,B). Thus, in both situations, optic vesicle evagination and Rx3 expression correlate. In line with this, at the permissive temperature a correctly spliced transcript is detectable by RT-PCR. This strongly suggests that under permissive conditions a functional protein is made, which accounts for the formation of small eyes in the mutant embryos.

The nature of the transcriptional repression remains unclear. Database searches did not reveal any significant homology of the insertion to known sequences. It is conceivable that the insertion influences intronic enhancer elements near the insertion site, thereby resulting in a repression of the locus. However, preliminary data do not suggest the presence of regulatory elements in the second intron (not shown). This does not rule out the possibility that the insertion represses the Rx3 locus by affecting its general accessibility for the transcriptional machinery. Temperature-sensitive splicing appears unlikely, as we cannot detect 5′ parts of the Rx3 transcript using RT-PCR or whole-mount in situ analysis at the restrictive temperature. Further analysis of the regulatory elements of the locus will address the nature of the repression and its temperature sensitivity.

The Rx3 plasmid rescues all aspects of the mutant phenotype, namely optic vesicle evagination and retinal differentiation in embryos raised under restrictive conditions (Fig. 3, Fig. 4). The regulatory elements required for correct expression are contained in 11 kb of genomic DNA present in the Rx3 plasmid as evidenced by specifically restored Rx3 expression in the wild-type expression domains. A modified Rx3 plasmid encoding a non-functional protein also results in specific Rx3 expression in mutant and wild-type embryos. Therefore, a functional Rx3 protein is not required for wild-type Rx3 expression, in contrast to Pax6 and Six3, which act in regulatory feedback loops (Chow et al., 1999; Loosli et al., 1999).

Our study shows that the el mutation genetically separates the determination of the vertebrate retina anlage from its subsequent morphogenesis, organ size control and neuronal differentiation.

Rx3 is not involved in the determination of the retina anlage, as the expression of the retina determination genes Six3 and Pax6 are not affected by either gain or loss of Rx3 activity. Furthermore, Rx2 expression indicative of retinal fate is still present in el-mutant embryos. However, also in absence of Rx3 activity, overexpression of Six3 results in the formation of ectopic retinal primordia.

The targeted inactivation of a mouse Rx homologue results in anophthalmic embryos (Mathers et al., 1997). Similar to the situation in the medaka eyeless mutant, at the neural plate stage, early expression of Six3, Pax6 and Otx2 is not affected in Rx mutant mice (Zhang et al., 2000). This indicates that in Rx^–/– mouse embryos, analogous to medaka, patterning of the anterior neural plate and thus the determination of the retina anlage is not affected.

However, at later stages (E9.0-10.5), upregulation of retina specific expression of Six3 and Pax6 is not detected in Rx^–/– embryos, indicating that retinal progenitor cells are not present (Zhang et al., 2000). At the corresponding stage, the expression of the respective homologues is still detectable in the medaka mutant embryo. The difference is most probably due to a split of function of the medaka Rx3 and Rx2 genes, respectively (see below).

The spatial and temporal expression patterns of the two medaka paralogues Rx2 and Rx3 differ significantly. Medaka Rx3 is expressed already at late gastrula stages in the anterior neuroectoderm, and its expression domain comprises the ventral forebrain as well as the optic vesicle. Rx2, on the other hand, is first expressed several hours later at the late neurula stage in the evaginated optic vesicle, and is thus exclusively expressed in retinal progenitor cells. Thus, Rx2 and Rx3 in medaka are not co-expressed except for a short period in the retinal progenitor cells of the evaginated optic vesicle, suggesting mainly non-overlapping functions of these genes, while Rx3 function is required in the evaginating optic vesicle Rx2 is expressed independently and may play a role in later aspects of retinogenesis.

In zebrafish, three Rx homologues have been isolated. Rx1 and Rx2, which are most similar to medaka Rx2, share the early expression domain with zebrafish Rx3, but show a slightly later onset of expression (Chuang et al., 1999). In the differentiating retina at late organogenesis stages, medaka and zebrafish Rx3 are expressed in the inner nuclear layer overlapping with Six3 expression. Expression of medaka Rx2 is confined to the outer nuclear layer (photoreceptor layer) and the ciliary margin as is Rx1 and Rx2 in zebrafish (not shown).

Based on our study, we propose the following model for early vertebrate retina development (Fig. 7). Patterning of the anterior neural plate culminates in defined expression patterns of Six3 and Pax6. This anterior neural plate patterning relies on the repression of wnt and BMP signalling (Niehrs, 1999), and requires the activity of the Otx transcription factors (Simeone, 1998). In the region where Six3 and Pax6 expression overlap, retinal fate is specified. An Rx3-independent regulatory feedback loop of these genes then ensures the maintenance of the retinal fate.

Six3 overexpression in el-mutant embryos results in dramatically enlarged retinal primordia (Fig. 5G). This expansion does not occur on the expense of forebrain tissue, suggesting that Six3 also affects cell proliferation independently of Rx3 and thereby regulates the size of the retina anlage. Consistent with the suggested role of Six3 in cell proliferation, the closely related Xenopus Optx2 gene controls the size of the optic vesicles by regulating proliferation (Zhou et al., 2000; Zuber et al., 1999).

Under the influence of midline signalling, the retinal anlage is split into two retinal primordia (Varga et al., 1999). Mutations in Six3 cause holoprosencephaly in humans, indicating a requirement for Six3 in this process (Wallis and Muenke, 1999). The two retinal primordia then become localised to the lateral wall of the prosencephalon during neurulation.

Subsequent evagination of the primordia results in the formation of the optic vesicles. For this process, Rx3 function is essential. Functional studies consistently argue for a regulatory role of vertebrate Rx genes in proliferation of retinal progenitor cells in the optic vesicle (Andreazzoli et al., 1999; Mathers and Jamrich, 2000), thus regulating its growth. In the absence of Rx3 function, there is no sign of morphogenesis and the specified retinal precursors do not proliferate and eventually die. Rx3 acts downstream of Six3 and Pax6 that determine the retina anlage. However, it is possible that Rx3 initially also receives input from neural plate patterning genes.

Subsequent development divides the optic vesicle into specific regions that then give rise to neural retina (NR), retinal pigmented epithelium (RPE) and optic stalk. Several genes that are expressed during these later steps of retinal development require Rx3 function directly or indirectly. Interestingly, the expression of Tbx2 and Tbx3 is specifically affected in the retinal primordium, but not in the hypothalamus, where they are also co-expressed. This indicates a differential regulation of Tbx2 and Tbx3 in these tissues.

The authors thank Milan Jamrich and Peter Gruss for discussion of unpublished results, J.-S. Joly for materials, and S. Cohen and M. Treier for comments on the manuscript. This work was supported by an EC-Biotech grant (BIO4-CT98-0399) to J. W.