The chicken RaxL gene plays a role in the initiation of photoreceptor differentiation

The vertebrate retina is an exquisitely sensitive organ for light detection and information processing. Vision begins with the reception of various wavelengths of light by the photoreceptor cells that line the back of the retina. Photoreceptor cells are highly specialized for the capture of light and for phototransduction, the conversion of light into a biochemical signal. Phototransduction takes place in a series of membranous discs comprising a unique structure: the photoreceptor outer segment. The details of this biochemical process have been well studied and provided us with the first appreciation of a complex signal transduction process. However, the details,and even some of the fundamental aspects of the development of these cells are much less well understood. From lineage analyses, it is known that photoreceptors derive from multipotent retinal progenitor cells(Turner and Cepko, 1987;Holt et al., 1988;Wetts and Fraser, 1988). These progenitor cells generate all six retinal neuronal cell types, and a retinal glial cell type, in a sequential fashion (reviewed byAltshuler et al., 1991). How the retina generates the appropriate cell types at each developmental time to form a functional retina with the correct ratio of each cell type remains an interesting issue. We have begun to understand the mechanisms of how each cell type is generated, including the photoreceptor cells(Cepko et al., 1996;Morrow et al., 1998;Levine et al., 2000).

Rods, which comprise the majority of photoreceptors in rodents and in humans, are the most light-sensitive photoreceptor cell types. They are susceptible to degeneration, in some cases, because of mutations that effect the development of rods. Studies of human diseases (reviewed byClarke et al., 2000) and a comprehensive analysis of genes expressed in photoreceptor cells(Blackshaw et al., 2001) have provided us with a source of candidate genes for the study of photoreceptor development. Cone photoreceptor development remains somewhat more mysterious. Cone photoreceptors are much less sensitive to light, but are active in the light intensities typical of daylight and of our brightly lit night. Cones provide us with high acuity vision, because of their high density in several regions of the retina and the high density of cells that compute the information from the photoreceptors and report it to the brain. Cone photoreceptors are also susceptible to degeneration, particularly in the prevalent human disease of the elderly, age related macular degeneration(reviewed by Weber, 1998). It is thus of great interest to learn how both rods and cones develop, not only to provide us with a basic understanding of retinal development, but also to allow for replacement of these cells in retinal degenerations, and/or to provide us with other points at which to intervene in disease processes.

Many types of birds have a need for high-acuity vision during the day. They typically have cone photoreceptors as the major photoreceptor cell type. Such is the case with chickens, which have a rod-free, cone-rich central zone,similar to that of humans (Morris,1982; Bruhn and Cepko,1996). We have been investigating the development of rod and cone photoreceptors in chickens and in mammals. In the chick, we and others have found two Rax genes, Rax/Rx and RaxL(Ohuchi et al., 1999). TheRax gene is expressed in all retinal progenitor cells, which is similar to the expression of mouse Rax(Furukawa et al., 1997a). By contrast, we found that RaxL is expressed in both retinal progenitor cells and early developing photoreceptors. RaxL homologs, includingrx1 and rx2, have been previously reported in zebrafish and medaka. In zebrafish, rx1 and rx2 are expressed in cone, but not rod, photoreceptors (Chuang et al.,1999). We provide evidence that chick RaxL is required for the earliest stage(s) of photoreceptor development, most probably by acting at the stage of commitment to the photoreceptor fate, and/or at the earliest stages of photoreceptor differentiation. We also report the presence of a second Rax gene (RAX2) in humans that may be the human homolog of RaxL. Wang, Zack and their colleagues have also identified this human gene and have found mutations in this gene in several individuals with retinal degenerations (D. Zack, personal communication). However, the significance of these mutations has not yet been established.

The cDNA encoding the homeodomain of chick RaxL was isolated by RT-PCR from E3 chick eyes using degenerate primers based on the mouseRax sequence. The full-length chick Rax and RaxLcDNAs were further isolated by screening a random-primed chick E6-E8 retinal library using the cDNA fragment encoding the RaxL homeodomain. The full-length chick Rax cDNA (pKScRax) was constructed by ligating 5′ and 3′ partial cDNA clones of pSKcRax into the EcoRI site of pBluescriptKS. The plasmid pKScRax contains the Rax open reading frame, 175 bp of the 5′ untranslated region (UTR) and 101 bp of the 3′ UTR. The full-length RaxL cDNA clone, pSKcRaxL, containsRaxL ORF, 214 bp of the 5′ and 256 bp of the 3′ UTR.

The BbsI/EcoRI and BsaI/EcoRI DNA fragments encoding oar/paired-tail motif deletion of Rax andRaxL were PCR amplified using TGAGAAGACCCCATGCACCCTCCCGGC and GAGAATTCCATGGCTCCCAGGGGCTG, and AAGGTCTCAGATGTTCCTCAATAAGTG and GAGAATTCCATGGGCTGCATGCCCTG as prime pairs, and were subcloned intoNcoI/EcoRI site of pSlaxEn vector to generate pSlaxEnRaxΔC and pSlaxEnRaxLΔC, respectively. TheNcoI/EcoRI DNA fragment encoding the homeodomain of RaxL was PCR amplified using AACCATGGCTGCTGCTGAGGAGGAACAGCCC and GAGAATTCGGACAGCATGGGGGTGTCGTG as primers and subcloned into pSlaxEn vector to generate pSlaxEnRaxLHD. The ClaI DNA fragments of pSlaxEnRaxΔC,pSlaxEnRaxLΔC and pSlaxEnRaxLHD were subsequently cloned into pRCAS(B)retroviral vector (Hughes et al.,1987) to generate pRCASEnRaxΔC, pRCASEnRaxLΔC and pRCASEnRaxLHD, respectively. The same fragments were also subcloned into pCS2 expression vectors (Rupp et al.,1994; Turner and Weintraub,1994) to generate pCSEnRaxΔC, pCSEnRaxLΔC and pCSEnRaxLHD, respectively. The RaxL expression vectors were generated as follows: the DNA fragment encoding the N-terminal region of RaxL was PCR amplified using CGACCATGGAGATGTTCCTCAATAAGTGT and GTGCCCGCCATAGGGGGG as primers; the NcoI/AflIII fragment of this PCR product together with AflIII/SacII(blunted) DNA fragment encoding the C-terminal region of RaxL were ligated into theNcoI/EcoRV site of pSlax21(Chen et al., 1999) to product pSlaxcRaxL vector; the ClaI fragment and ClaI/SpeI fragment of pSlaxcRaxL were further subcloned into the ClaI site of pRCAS(A) and the ClaI/XbaI region of pCS2 to generate pRCAS(A) cRaxL retroviral and pCScRaxL expression vectors, respectively. The Rax expression vector was generated as follows: the BbsI/EcoRI DNA fragment encoding Rax ORF was PCR amplified from pKScRax using TGAGAAGACCCCATGCACCCTCCCGGC and M13 reverse primers and subcloned intoNcoI/EcoRI locus of pSlax21 to generate pSlaxcRax; theClaI DNA fragment of pSlaxcRax was further subcloned intoClaI site of pCS to generate pCScRax expression vector. TheRcaI fragment of mouse Crx was cloned into the NcoI locus of pSlax21 to generate pSlaxmCrx. The ClaI fragment of pSlaxmCrx was further subcloned into pRCAS(A) to construct pRCAS(A)mCrx retroviral vector. In situ hybridization probes specific to Rax andRaxL were transcribed from pKScRaxspl and pKScRaxLspl, respectively,which includes 5′ UTR, 3′ UTR and the homeodomain deleted coding regions of Rax and RaxL, respectively. The pKScRaxspl was constructed by ligating two EcoRI/BamHI PCR fragments intoEcoRI site of pBluescriptKS. These two PCR fragments were amplified from pSKcRax using T7 primer and GCGGATCCCTCCTCGTCCGACGGCTTCCC primer pair, T3 primer and AAGGATCCAGCCGCTCCCCGCAGGCG primer pair. The pKScRaxLspl was constructed in a similar way in that T3 (GCGGATCCTTCCTCCTCAGCAGCAGCTGG) and T7(AAGGATCCAACCGGCCGCCCATGACG) were used as two PCR primer sets.

Plasmid DNA containing 0.05% of Fast Green was injected into the right optic vesicle of Hamburger-Hamilton stage 9 to stage 11 chick embryos in ovo. Immediately after injection, the embryo was subjected to electroporation using the Tokiwa CUY-21 square electroporator with 10 mV for three cycles of 50 mseconds pulse and 950 mseconds chase.

Whole-mount and section in situ hybridization were performed as described(Chen and Cepko, 2000). Flat-mount in situ hybridization was performed as described(Bruhn and Cepko, 1996) with the following modification. The flat-mounted retinal tissues were hybridized overnight at 70°C with digoxigenin-labeled RNA probes of specific cell markers together with the fluorescein-labeled RNA probe of engrailed repressor domain. After hybridization, the retinas were washed and blocked as described and incubated overnight at 4°C with 1:2000 dilution of AP-conjugated anti-digoxigenin antibody (Roche Diagnostics Coporation) in TBST and 1%heat-inactivated sheep serum. Retinas were washed several times in TBST and further detected with NBT and BCIP until the desired purple signal developed. The developing reaction was stopped by washing three times with TBST (pH 5.5)and heating at 70°C for 2 hours in the same buffer to dissociate anti-digoxigenin Ab. The pictures of the retinas with the first in situ signal were taken before detecting the second signal. To detect the second signal,the heat-inactivated retinas were blocked in TBST and 10% sheep serum for 2 hours and incubated overnight at 4°C with 1:2000 dilution of AP-conjugated antifluorescein antibody (Roche Diagnostics Corporation) in 1% sheep serum/TBST. After washing in TBST, the second in situ signal was detected with BCIP alone until the desired blue color developed. The developing reaction was stopped in TBST (pH 5.5) and pictures of the retinas with both the first and second in situ signals were taken. To further detect the total viral infection, the retinas were further subjected to 3C2 mAb staining based on the protocol described in the Immunostaining section, after treatment at 70°C for 2 hours to dissociate anti-Fluorescein Ab.

COS cells were grown in DMEM with 10% fetal calf serum. Five micrograms of pRET1-CAT reporter (Furukawa et al.,1997b), 1 μg of pSVβ (Clontech), 5 μg of pCScRaxL with increasing amounts of pCSEnRaxΔC, pCSEnRaxLΔC or pCSEnRaxLHD, and decreasing amounts of pCS2 to make a total 26 μg of plasmid DNA were transfected onto 10 cm dishes using Superfect as the transfection reagent according to the manufacture's protocol (Qiagen). Cells were harvested for a CAT assay 48 hour post-transfection as described(Chen et al., 1996). Five to 10 μl of cell extract without heat treatment were used for measuring theβ-galactosidase activity at room temperature in 1 ml of Z buffer (60 mM Na₂HPO₄/40 mM NaH₂PO₄/10 mM KCl/1mM MgSO₄) containing 1 μl of β-mercaptoethanol and 0.5 mg/ml of ONPG. The reactions were stop with 0.5 ml of 1 M Na₂CO₃ and OD420 were measured. The OD420 value, which reflects the transfection efficiency of each extract, was used to normalize the CAT value from each transfection.

Monoclonal antibodies to chick visinin were generated by Maine Biotechnology Service Incorporation (S. Bruhn and C. Cepko, unpublished) using purified chick visinin protein as an antigen (gift from Dr A. Polans)(Polans et al., 1993). One of the visinin mAbs (7G4) was deposited into The Developmental Studies Hybridoma Bank at the University of Iowa. For western blot analysis, chick retinas were harvested and sonicated in whole cell extract buffer (20 mM HEPES pH7.6/150 mM NaCl/0.5 mM DTT/0.2 mM EGTA/0.2 mM EDTA/25% glycerol) with proteinase inhibitor cocktail (Roche Diagnostics Corporation). The cell lysates were collected after centrifugation for 15 minutes at 4°C and the protein concentration was determined by Bradford analysis (BioRad protein assay) using bovine serum albumin as a standard. The retinal extracts containing 25 μg protein were run on a 10% precast SDS-PAGE gels and transferred to a nitrocellulose membrane according to the manufacture's protocol (Invitrogen). The transferred nitrocellulose membrane was stained with Ponceau S to confirm that an equal amount of protein was loaded and transferred in each lane before blocking with 5% nonfat milk in PBST (0.1% Tween-20 in PBS). Two-thousand-fold dilution of visinin mAb ascites fluid was used as a primary antibody and peroxidase-conjugated goat anti-mouse IgG (1:4000 dilution) (Jackson Immunoresearch Laboratory) was used as a secondary antibody. The western blot signal was further detected with ECL reagent (Amersham).

Retinal cryosections (20 μm) were blocked with 5% sheep serum/0.02%TritonX-100/PBS for 30 minutes at room temperature. The sections were subsequently incubated with visinin mAb (1:100 dilution of hybridoma culture supernant) for 1 hour. After several washes in PBS, the sections were incubated in biotinylated anti-mouse IgG (1:500 dilution) (Vector) for another hour. The Vectastain ABC kit (Vector) and DAB peroxidase substrate kit(Vector) were further used for amplifying and detecting the signal according to manufacture's protocol.

Papain (100 units/ml) (Worthington Biochenical Corporation) was first activated in Hank's balanced salt solution (HBSS) containing 10 mM HEPES pH 7.6, 2.5 mM cysteine and 0.5 mM EDTA for 15 minutes at 37°C. Dissected chick retinas were incubated in activated papain solution for 40 minutes at 37°C. Retinal pellets were gently triturated and incubated in 0.1 mg/ml of DNaseI/HBSS for 10 minutes. The dissociated retinal cells were further washed twice with HBSS and fixed in 4% paraformaldehyde for 5 minutes at room temperature. The protocol for antibody staining on fixed cells is the same as staining tissue sections. The Cyt2-conjugated anti-rabbit IgG and Cyt3-conjugated anti-mouse IgG (1:500 dilution) (Jackson Immunoresearch Laboratories) were used for secondary antibodies. After two washes with PBS,the cells were suspended in 1% formaldehyde/PBS for FACS analysis.

Viral-infected retinas were fixed in 4% paraformaldehyde/PBS and embedded in OCT compound (Tissue-Tek) after cryoprotection in 30% sucrose solution. Cryosections (20 μm) were subjected to the TUNEL assay using the in situ cell death detection fluorescein kit (Roche) according to the manufacture's protocol. Retinal sections were then further stained with 3C2 mAb and Cyt3-conjugated goat anti-mouse IgG (1:400 dilution) (Jackson Immunoresearch Laboratories) to visualize the viral infected areas.

Full-length chick Rax cDNAs were isolated from a chick E6-E8 retinal library using a RT-PCR fragment encoding the homeodomain region of chick Rax. Two distinct Rax cDNAs, Rax and RaxL, were isolated. The open reading frame (ORF) of Rax encodes a 316 amino acid protein. The RaxL cDNA encodes an ORF containing 228 amino acid residues. Sequence comparison of Rax and RaxL showed that they share 100%amino acid identity in their paired-type homeodomains. Scattered sequence similarity was also found in the region C-terminal to the homeodomain,including the conserved oar/paired-tail motif, which has been found in several paired-type homeobox genes (Furukawa et al., 1997a; Mathers et al.,1997). Unlike mouse Rax and chick Rax, the chick RaxL contains very little sequence that is N-terminal to the homeodomain. Moreover, the highly conserved octapeptide, identified in some of the paired-domain and homeodomain proteins including mouse Rax and chicken Rax, is missing in RaxL. During the course of our studies, the Rax and RaxL cDNAs were reported by Ohuchi et al. (Ohuchi et al., 1999). The cDNA of Rax that we isolated has roughly 0.15 kb more 5′ UTR sequence than the published Rax and some base pair differences throughout the cDNA. Our RaxL isolate contains roughly 0.2 kb more sequence information in both the 5′ and 3′UTRs. The sequences of both Rax and RaxL have been submitted to GenBank with the Accession Numbers AF420600 and AF420601, respectively.

Through searches of the human EST and genome databases, a second human Rax(RAX2) gene was found and located on human chromosome 19. The correspoinding EST was also isolated from a human retinoblastoma cell line(I.M.A.G.E. clone ID3344166). An amino acid sequence comparison showed 93%identity in the homeodomain region between RaxL and the human protein (RAX2). Scattered similarity was also found outside of the homeodomain. Interestingly,like RaxL, the RAX2 exhibited a very short sequence N-terminal to the homeodomain, and was lacking the octapeptide(Fig. 1A). Based on these sequence similarities, this gene is probably the human homolog ofRaxL.

The expression patterns of Rax and RaxL in early chick embryos (Hamburger-Hamilton stage 8 to stage 20) were analyzed by whole-mount in situ hybridization. To avoid cross-hybridization between Rax andRaxL through their conserved homeodomain regions, specific RNA probes with the homeodomain regions deleted were used. Rax RNA was detected in the anterior neural folds at stage 8 (data not shown). By stage 11,Rax was expressed in the entire forebrain region(Fig. 1B) and highly concentrated in the optic vesicles and the ventral midline structure, the infundibulum (Fig. 1C,arrowhead and arrow, respectively). By stage 12, when the optic vesicles have formed, the Rax signal remained strong in the optic vesicles and in the infundibulum, but became weak in the anterior and dorsal forebrain(Fig. 1D,E). By stage 14, the expression of Rax was confined to the retina and ventral diencephalon(Fig. 1F,G arrowhead and arrow,respectively). The retinal and ventral diencephalon expression of Raxpersisted to stage 20, the oldest stage we have analyzed by whole-mount in situ hybridization (Fig. 1H,I,and data not shown).

RaxL was expressed in an overlapping but not identical pattern to that of Rax. The transcript of RaxL was first found at stage 9 in the ventral anterior neural tube (data not shown). By stage 11,RaxL was highly expressed in the ventral optic vesicles, while the signals in the anterior forebrain and infundibulum were barely detectable(Fig. 1J,K). In contrast toRax, by stage 12, RaxL was expressed strongly in the optic vesicles, very weakly in the infundibulum, and at undetectable levels in the dorsal and anterior forebrain (Fig. 1L,M). By stage 14, RaxL was expressed only in the retina and no infundibulum expression was observed(Fig. 1N,O compare withFig. 1G arrow). Interestingly,the early retinal expression of both Rax and RaxL was not uniform throughout the optic vesicle. RaxL was expressed in a high ventral to low dorsal gradient transiently from stage 13 to stage 17(Fig. 1N,P, and data not shown). By contrast, Rax was expressed at a high level in both dorsal and ventral domains and at a low level in the middle region of the retina at similar stages (Fig. 1F,H, and data not shown).

As both Rax and RaxL expression was observed in the eyes of early chick embryos, a further detailed analysis of retinal expression was carried out on retinal sections from embryonic day 5 (E5) to E19(Fig. 2). The Raxtranscript was detected in the majority of retinoblasts at a high level in the E5 retina (Fig. 2A). By E6, two domains with undetectable Rax expression were observed. One was adjacent to the pigment epithelium, presumably comprising differentiating photoreceptor cells, and the other was adjacent to the vitreous, presumably comprising ganglion cells (Fig. 2B, arrows). This pattern is consistent with the observation that mouse Rax is highly expressed in proliferating retinal progenitors and is downregulated in differentiated retinal cells(Furukawa et al., 1997a). At E7, the Rax transcript was found in a small population of cells residing in the future inner nuclear layer (INL) throughout the retina, which are likely to be the remaining retinal progenitors(Fig. 2C,D). Raxexpression was further restricted into a narrow domain in the INL at E9(Fig. 2I). At E11, when almost all retinal progenitor cells have become postmitotic, we observed a low level of Rax signal in the middle of the INL(Fig. 2J). Because a low level expression of Rax was found in Muller glial cells of the mouse retina(Furukawa et al., 2000), it is likely that this small population comprises the remaining retinal progenitors and/or differentiating Muller glial cells. However, the identity of theseRax-expressing cells needs further characterization. A faintRax signal remained at E14 (Fig. 2K) and became undetectable at E19, immediately before hatching(Fig. 2L). We found noRax expression in the retina of post-hatched chicks at one month of age (P30) (data not shown). Based on the in situ hybridization analysis ofRax, which shows a similar expression pattern to that of the mouseRax gene, and given the amino acid sequence similarity of Rax and mouse Rax, chick Rax is likely to be the homolog of the mouseRax gene.

The RaxL transcript was detected at a lower level thanRax throughout the retina at E5(Fig. 2E). By E6, theRaxL non-expressing domain was seen in the ganglion cell region(Fig. 2F, arrow). However,unlike Rax, the RaxL signal remained in the developing photoreceptor layer at this stage (compareFig. 2B,F). As development proceeded, some retinal cells expressing higher levels of RaxLappeared near the pigment epithelium at E7(Fig. 2G), presumably the progenitors fated to be photoreceptors. This pattern of expression in the future outer nuclear layer (ONL) persisted(Fig. 2H). The increase in staining of the future ONL progressed from the center to the periphery(compare Fig. 2H,G) in a pattern that coincides with the overall temporal developmental of the retina. As development progressed, two expression domains of RaxL resulted;one weak expression zone overlapping that of Rax in the INL,representing the remaining retinal progenitor cells(Fig. 2H, arrowhead), and one strong expression zone located in the future ONL, most likely representing the developing photoreceptor cells (Fig. 2H arrow). This two-domain expression pattern of RaxLpersisted until E11 with decreasing signal in the INL and increasing signal in the ONL (Fig. 2M,N). By E14,RaxL was detected only in photoreceptors, and this expression was downregulated to an undetectable level by E19(Fig. 2O,P). No RaxLexpression was found in the P30 chick retina (data not shown). The expression of RaxL in retinal progenitor cells, and later at a high level in developing photoreceptors, suggests a key role for RaxL in the early stage of photoreceptor development.

To determine whether RaxL plays a role in photoreceptor development, we ectopically expressed full-length RaxL protein in optic vesicles using a retroviral expression vector. The optic vesicles of chick embryos were infected with a RaxL retrovirus at Hamburger-Hamilton stage 10 and infected retinas were harvested between E6 and E7. The development of photoreceptor cells was analyzed using the photoreceptor marker, visinin, by flat-mount in situ hybridization (Yamagata et al., 1990). We detected no ectopic expression ofvisinin-expressing cells in the RaxL infected retina (data not shown), suggesting the RaxL is not sufficient to promote photoreceptor cell fate choice. We then examined if RaxL is necessary for photoreceptor cell development by introducing a putative dominant-negative allele of RaxL. As RaxL shares the identical amino acid sequence in the homeodomain region with Rax, dominant-negative RaxL could potentially interfere with both RaxL and Rax functions. To minimize this possibility and maintain as much RaxL specificity as possible, we made a fusion construct containing the engrailed repressor domain and RaxL with deletion only of the oar/paired-tail motif(EnRaxLΔC). The similar fusion construct has been shown as a dominant negative allele of Xrx1 (Xenopus homolog of Rax) inXenopus embryos (Andreazzoli et al., 1999). The EnRaxLΔC retroviral vector was electroporated into optic vesicles of Hamburger-Hamilton stage 10 chick embryos. After electroporation, the EnRaxLΔC transfected retinal cells should produce EnRaxLΔC virus, which subsequently infects neighboring retinal cells to create some viral infected patches in the retina. As we have found that electroporation with high concentrations of DNA can lead to a nonspecific small eye phenotype, we electroporated the viral construct at the low concentration of 0.1-0.2 μg/μl. The infected retinas were harvested at E7.5-E8, and the expression of endogenous visinin and exogenous EnRaxLΔC was analyzed by double in situ hybridization using visinin and engrailed probes, respectively. The infected retinas showed the same overallvisinin pattern as control non-electroporated retinas. However, 73%(11 out of 15) of the retinas electroporated with the EnRaxLΔC had several patches with low visinin expression(Fig. 3A). These patches were within the infected areas, detected with an engrailed probe(Fig. 3B in greenish blue). In some cases, there was no reduction in visinin expression in infected areas. This could be due to a low level expression of the EnRaxLΔC protein, which could occur because of the interference of some viral integration sites. We found no correlation between the size and location of infected patches and the reduction of visinin. To examine if viral infection grossly altered retinal morphology by affecting progenitor cell proliferation, a concern because RaxL is expressed in retinal progenitors, the retina was sectioned after double in situ hybridization.Fig. 3C shows that the viral infected patches spanned the entire thickness of the retina and that the retinal thickness remained normal within those patches. However, there were fewer cells expressing visinin transcript within the EnRaxLΔC virus-infected domains. To further investigate whether EnRaxLΔC interfered with the proper development of photoreceptor cells, or just visinin marker gene expression, we examined the expression of another photoreceptor-specific gene, RXRγ (Rxrg)(Hoover et al., 1998).Rxrg exhibited a uniform pattern of expression in the control,uninfected E7.5 retina (data not shown). However, when the retina was electroporated with EnRaxLΔC viral construct, the virus infected patches showed significant reduction of Rxrg expression in 82% (14 out of 17)of the infected retinas (Fig. 3D-F). The fact that two independent photoreceptor specific genes were reduced by EnRaxLΔC, makes it likely that the development of photoreceptor cells is affected by EnRaxLΔC. These results suggest thatRaxL is required for the development of photoreceptors, and that introducing EnRaxLΔC did not alter retinal cell proliferation.

To determine whether EnRaxLΔC interfered specifically withRaxL, we electroporated a similar viral construct, EnRaxΔC,containing the engrailed repressor domain and Rax, with deletion of the oar/paired-tail motif. Seven and 16 infected retinas were tested forvisinin and Rxrg expression, respectively. We found all the retinas tested exhibited normal visinin and Rxrg expression within the EnRaxΔC infected patches(Fig. 3G-I and 3J-L,respectively). These results strongly suggest that RaxL, but notRax, is required for photoreceptor cell development. Interestingly,we observed a similar photoreceptor phenotype with 100% penetrance when we introduced a dominant-negative allele comprising the engrailed repressor domain fused with the RaxL homeodomain (EnRaxLHD)(Fig. 3M,N). However, when a control viral construct, EnIrx, which carries the homeodomain of Irxfused to the engrailed repressor domain(Bao et al., 1999), was introduced into chick retina, the normal visinin and Rxrgexpression was observed within the EnIrx infected patches(Fig. 3O,P, and data not shown).

In order to examine whether the effects on retina cell differentiation by EnRaxLΔC virus were specific to photoreceptors, the expression of markers of other cell types were analyzed. The expression of Brn3a(now known as Pou4f1) the ganglion cell marker(Liu et al., 2000),Chx10, the bipolar cell marker(Belecky-Adams et al., 1997;Chen and Cepko, 2000), andPax6, which marks horizontal, amacrine and ganglion cells(Belecky-Adams et al., 1997),were analyzed on EnRaxLΔC infected retina. Virus infected E9 retinas,including seven retinas for Brn3a, 6 retinas for Chx10 and 8 retinas for Pax6, were analyzed. None of these markers were affected by expression of EnRaxLΔC (Fig. 4). We also found no changes on the expression of these markers when EnRaxLHD was introduced into chick retina (data not shown). These results strongly suggest that the RaxL gene is required for proper development of photoreceptor cells, but not other retinal neurons.

We have shown that EnRaxLΔC, which theoretically acts as a dominant negative allele of RaxL, blocks normal photoreceptor differentiation. To determine whether EnRaxLΔC indeed functions as a dominant negative allele of RaxL, the transactivation activities of RaxL and EnRaxLΔC were analyzed using a reporter construct encoding the chloramphenicol acetyltransferase (CAT) gene driven by five copies of the Ret1 enhancer element (RET1-CAT) (Fig. 5A). The Ret1/PCE1 site, an enhancer element present in many photoreceptor specific genes, is required for photoreceptor specific expression of these genes (Kikuchi et al.,1993). Fig. 5Bshows that RaxL transactivated the RET1-CAT reporter construct 53-fold above the control expression vector when transiently transfected into COS cells,suggesting that RaxL is a strong transcriptional activator which can transactive photoreceptor specific genes through the Ret1 enhancer element. By contrast, Rax transactived the same reporter construct more weakly (ninefold above the control vector) (Fig. 5B), suggesting that Rax might recognize different enhancer elements that perhaps function in progenitor cells. To test the dominant-negative activity of EnRaxLΔC, the RET1-CAT construct was transiently co-transfected with vectors expressing RaxL and/or various engrailed-fusion constructs into COS cells. The CAT activity of cell extracts was assayed 48 hours after transfection. In the presence of an increasing amount of EnRaxLΔC, the activation activity of RaxL was reduced in a dose-responsive manner. An equal amount of EnRaxLΔC repressed the RaxL activation activity to 25.3%, and three times more EnRaxLΔC further repressed the activity of RaxL to 3.9%(Fig. 5C). This decrease of CAT activity is not due to the overexpression of a homeodomain protein, as cells transfected with four doses of RaxL showed similar CAT activity to those transactivated by a single dose of RaxL (data not shown). EnRaxLΔC repressed the transactivation of RaxL specifically, as EnRaxLΔC showed no effect on the activity of a CAT reporter driven by the SV40 enhancer elements (data not shown). These data suggest that EnRaxLΔC functions as a dominant negative form of RaxL, and that expression of EnRaxLΔC inhibits the endogenous RaxL activity. Interestingly,we also found dominant negative activities of EnRaxΔC and EnRaxLHD on RaxL transactivation activity in a dose-related manner(Fig. 5C). The dominant negative effects of EnRaxΔC and EnRaxLHD were similar but weaker than EnRaxLΔC.

We have demonstrated that EnRaxLΔC can inhibit the transcription activity of RaxL in tissue culture cells, suggesting that the phenotype observed by expressing EnRaxLΔC in the chick retina is due to a block of endogenous RaxL activity. If this assumption is correct, the dominant negative phenotype created by EnRaxLΔC should be rescued by coexpression of RaxL in ovo. As the phenotype created by EnRaxLΔC did not show full penetrance, we decided to use the dominant-negative EnRaxLHD, which is more effective, and thus facilitate the interpretation of a rescue experiment. We electroporated EnRaxLHD alone or together with RaxL into chick optic vesicles. The EnRaxLHD and RaxL retroviral constructs carry the type B and type A envelope proteins, respectively, which allows co-infection of both viruses into the same cells. After the detection of EnRaxLHD virus, we stained the infected retinas with the 3C2 mAb, which recognizes a matrix core protein of Rous Sarcoma virus (Potts et al.,1987). The 3C2 mAb recognizes both viruses. When EnRaxLHD alone was electroporated, we observed strong inhibition of the visinin signal, which correlated with EnRaxLHD infected patches, detected by engrailed expression. The staining pattern of 3C2 mAb perfectly matched the engrailed staining pattern (Fig. 6A-C). This observation allows us to assume that 3C2 stained areas with no engrailed signal represents the RaxL-only infected region. However, the patches with engrailed signal may express EnRaxLHD virus alone, or express both EnRaxLHD and RaxL viruses. To ensure that most of the EnRaxLHD infected cells were also infected with RaxL virus, we electroporated three times as much RaxL viral construct as EnRaxLHD construct when co-electroporation was performed.Fig. 6D-G show that when both EnRaxLHD and RaxL were introduced, the inhibitory effect on visininexpression by EnRaxLHD was dramatically reduced(Fig. 6D,E, arrows). These data demonstrated that ectopic expression of RaxL rescued the dominant-negative phenotype generated by EnRaxLHD. This rescue was specific to RaxL. Chick Rax can not rescue the photoreceptor phenotype(Fig. 6H-K). Similarly, when the mouse Crx viral construct, encoding another paired-type homeodomain protein, which is required for photoreceptor maturation but not for initial photoreceptor cell generation (Furukawa et al., 1997b), was introduced with EnRaxLHD, no rescue ofvisinin expression was found (Fig. 6L-N). These in vivo rescue results strongly suggest that the inhibition of photoreceptor gene expression was due to a block of endogenousRaxL activity. The fact that Crx could not rescue or bypassRaxL function suggests that RaxL is required beforeCrx function during photoreceptor cell development. This is consistent with the idea that RaxL is required in the early stage of photoreceptor cell generation.

The reduction in the expression of photoreceptor specific genes could be due to EnRaxLΔC interfering with proper photoreceptor differentiation and/or photoreceptor survival. To further address these issues, we took advantage of the consistent penetrance of EnRaxLHD to quantify the number of photoreceptors following expression of EnRaxLHD. Two monoclonal antibodies(mAb) against chick visinin were generated, 6H9 and 7G4, which exhibited specificity for chick visinin (S. Bruhn and C. Cepko, unpublished). They behaved similarly in both western blots and immunohistochemical assays, with the results of 7G4 shown in Fig. 7A-C. 7G4 recognized a single band of 24 kDa from chick retinal extracts by western blot analysis, which corresponds to the predicted size of chick visinin (Fig. 7A). A low level of visinin protein was evident at E5.5 and the level increased gradually as more photoreceptors differentiated between E6.5 and E8.5(Fig. 7A). This time course is consistent with the expression profile of visinin transcripts(Bruhn and Cepko, 1996). The immunostaining on retinal sections further demonstrated that the visinin mAb recognized differentiating and mature photoreceptors, which are localized to the developing ONL in differentiating (E7) and mature (E18) retinas(Fig. 7B,C). These analyses demonstrate that the visinin mAb is a reliable early marker of photoreceptors.

To quantify the number of photoreceptor cells, optic vesicles were electroporated with EnRaxLHD, RaxL or control RCAS retroviral constructs at Hamburger-Hamilton stage 10 and infected retinas were harvested and dissociated at E8. FACS analysis was performed on dissociated retinal cells after staining with 7G4 mAb against visinin and antiserum against p27, an Avian Leukemia Viral protein (SPAFAS) (Fig. 7D,E). As viral infections only occurred in some patches of the retina and we expected the action of EnRaxLHD to be cell autonomous, only infected cells were scored for visinin expression. The percentage of visinin and p27 double-positive cells among the viral infected population (p27 positive) was calculated after a total of 250,000 cells were counted from each retina. In two independent experiments, 10.8% and 15.7% (on average) of control virus infected retinal cells were visinin-positive photoreceptors. However, when the retina was infected with the EnRaxLHD virus, the percentage of photoreceptors was significantly decreased to average 7.5% and 10.4%,respectively (Fig. 7F). We found no significant change in the VC1.1-positive population, which comprises amacrine and ganglion cells, when the retina was infected with EnRaxLHD virus(data not shown). Interestingly, we found a slight increase in visinin-expressing photoreceptors in retinas infected with RaxL virus (from average 10.8% and 13.5% to 12.3% and 15.8%, respectively)(Fig. 7F). Overexpression of RaxL thus slightly increased the number of photoreceptors, and interfering with the endogenous RaxL by overexpression of the dominant negative EnRaxLHD led to a significant reduction of differentiating photoreceptor cells in the retina.

Results from both the FACS analysis and whole-mount in situ hybridization showed that decreasing the number of differentiating photoreceptors did not lead to an increase of the other retinal cell types scored following introduction of dominant negative RaxL. These data suggested that interfering with the normal function of RaxL did not induce a change in retinal cell fates. The reduction in differentiating photoreceptors could then be a block in photoreceptor cell differentiation and/or induction of photoreceptor cell death, or an effect on proliferation that affects only photoreceptor cells. The latter case is very unlikely, as photoreceptors are made by a multipotent progenitor (Fekete et al., 1994) and thus other cells would be affected, as would general proliferation, if this were the case. Nevertheless, to test the effect of dominant negative RaxL on cell proliferation, the anti-phospho-Histone H3 antibody was used to detect mitotic cells on E7.5 retinal sections electroporated with the EnRaxLHD viral construct. The virus infected patches were visualized with 3C2 mAb(Fig. 8, red). At E7.5 very few M-phase cells were found in or near the ventricular surface (green nuclei inFig. 8A), and there was no significant difference between virus-infected patches and adjacent non-infected areas (Fig. 8A,B). We also found that EnRaxLHD had no significant effect when scored for phospho-Histone H3 staining on E5.5 and E6.5 retinas when there were more mitotic cells (data not shown). These data suggest that the decrease in differentiating photoreceptor cells by EnRaxLHD was not due to interference with progenitor cell proliferation. We then examined the possibility that reduction was due to apoptosis. The TUNEL assay was performed on E7.5 retinal sections electroporated with EnRaxLHD or control EnIrx viral constructs. TUNEL-positive cells were found only occasionally in normal E7.5 retinal sections. Very few TUNEL-positive cells were found in control EnIrx infected retina (Fig. 8C,D) and non-infected patches in EnRaxLHD infected retina(Fig. 8E,F). However, many TUNEL-positive cells were observed in the EnRaxLHD infected patches(Fig. 8E, green and yellow dots). The same viral construct induced no apoptosis when electroporated into chick brain (Fig. 8G,H),suggesting that overexpression of EnRaxLHD does not lead to non-specific apoptosis. The specific increase of apoptosis in EnRaxLHD infected retina provides an explanation for the decreased number of photoreceptor cells. Interestingly, we found that the TUNEL-positive cells were not concentrated in the photoreceptor layer, but spanned the radial thickness of the retina.

We have investigated the expression patterns of Rax andRaxL and performed functional analyses of RaxL. Our data indicate the RaxL is required for the early steps in the development of photoreceptor cells.

We have isolated cDNAs encoding two members of chick Rax family,Rax and RaxL. Rax is highly expressed in the optic vesicles,retinal progenitor cells, and the ventral diencephalon, in a pattern similar to that of the mouse Rax/Rx(Furukawa et al., 1997a;Mathers et al., 1997) and the published chicken Rax gene(Ohuchi et al., 1999). However, contrary to the previous report that RaxL is highly expressed in the developing retina and ventral diencephalon(Ohuchi et al., 1999), we found that RaxL is expressed in the optic vesicles and retinal progenitor cells, but is absent from the ventral diencephalon. We reason that this difference is due to the specificity of the RaxL probe used in each study. The RaxL probe used previously contains the homeodomain,which shares 96% nucleotide (173 out of 180 nucleotides) identity to theRax homeodomain region. The RaxL probe containing the homeodomain region can recognize both Rax and RaxLtranscripts, and therefore can cross-hybridize with Rax in the ventral diencephalon. The fact that we do not observe RaxL in the ventral diencephalon allows us to conclude that the RaxL expression pattern resembles that of the zebrafish homologs rx1 and rx2. rx1 and rx2 are expressed in the optic primordium and are absent from the ventral midline of the diencephalon. More interestingly, similar toRaxL, rx1 and rx2 are also downregulated as the retina differentiates, except in the ONL where they continue to be expressed at high levels in photoreceptors. The photoreceptor cells where rx1 andrx2 expressed are cones, but not rods(Chuang et al., 1999).RaxL is also expressed in cones as cones comprise 80% of chick photoreceptors and the RaxL-expressing population comprises the majority, if not all, of the photoreceptors. However, we cannot determine ifRaxL is also expressed in rods. We speculate that RaxLhomologs are expressed in cone, but not rod, photoreceptor cells in vertebrates. Such conserved expression pattern and gene sequences suggest an important function for RaxL in photoreceptor development. In mammals,the expression of a human RaxL homolog (RAX2) in a retinoblastoma cell line further suggests a role of RaxL in retina development. In addition, mutations in the human RAX2 gene have been found in individuals with photoreceptor degeneration, which, if shown to be causal, would further establish the importance of RaxL homologs in photoreceptor cell development and/or function (D. Zack, personal communication). A mouse RaxL homolog has not been isolated. Surprisingly, the human RAX2 syntenic region is missing in the mouse genome (T. Matsuda and C. Cepko, unpublished). It is possible that the mouseRaxL homolog is located in a different location in the mouse genome,and is expressed at very low abundance because it is expected to be in cone photoreceptors, which comprise only 2.2% of retina cells in the mouse(Young, 1985). It is also possible that the mouse has no RaxL homolog. The function ofRaxL may be carried out by the mouse Rax/Rx gene, as mouseRax/Rx has been reported to be expressed in photoreceptor cells and can transactivate photoreceptor specific genes(Kimura et al., 2000).

Photoreceptor cells develop in a temporal gradient from the central to the peripheral retina. In the peripheral chick E7 retina, a subset of retinal cells that express a high level of RaxL spans the retinal epithelium,except in the differentiating ganglion cell layer. As development proceeds,centrally located RaxL-expressing cells become concentrated in the photoreceptor layer. This pattern is consistent with RaxL being expressed in mitotic progenitors that are in the process of producing photoreceptors, and/or in newly postmitotic photoreceptors. The high level expression of RaxL in such populations places RaxL at an important point in early photoreceptor development.

Chick photoreceptor genesis is reported to begin sometime between E3 and E5 in different studies (Kahn,1974; Spence and Robson,1989), with the bulk of photoreceptor genesis occurring between E5 and E6 (Prada et al., 1991;Belecky-Adams et al., 1996). Photoreceptors do not differentiate morphologically until E9.5, when the inner segments appear (Meller and Tetzlaff,1976). The outer segments appear on E13(Meller and Tetzlaff, 1976),and the synapses from photoreceptors to bipolar cells are evident on about E18(Hughes and LaVelle, 1974). As discussed above, we found that RaxL is expressed in developing photoreceptors, but not in mature photoreceptors on E19, suggesting thatRaxL is not required for the maintenance or survival of mature photoreceptors. Furthermore, apoptosis was observed as early as E7.5 when proper RaxL function was blocked, also indicating that RaxLis required for an early step in photoreceptor development.

There are two populations of retinal progenitor cells expressing theRaxL transcript. One is the majority of retinal progenitors, which expresses a low level of RaxL and a high level of Rax. The other population is a small subset of cells that expresses a high level ofRaxL. Our data show that overexpression of a fusion construct,EnRaxLHD, interferes with survival of a subset of cells located predominantly in the middle retinal layer. This is the area where mitotic progenitor cells reside and thus it is possible that EnRaxLHD interferes with survival of a subset of mitotic cells. These may be the same cells that express high levels of RaxL in this area, and we would propose that these are the cells that are in the process of producing photoreceptor cells. We believe that it would be this subset of cells, rather than all progenitor cells, based upon the observations that EnRaxLHD does not interfere with general progenitor proliferation, as the number of mitotic cells and the overall thickness of infected areas, as well as differentiation of other retinal cell types, were not significantly affected. Alternatively, the dying cells located in the middle of the retina following transduction with EnRaxLHD are newly produced,postmitotic cells that are fated to be photoreceptor cells. It is not known if cells in this state would be located in this area as there are no markers for cells that are newly postmitotic and fated to be photoreceptors. Although photoreceptor cells are usually located in the outer nuclear layer, it is possible that they briefly reside in the middle of the retina prior to migrating to the future outer nuclear layer. It is curious that murine cones do display an inward migration prior to undergoing full differentiation in the mouse (Rich et al., 1997).

Despite the identical amino acid sequence in the homeodomain regions ofRax and RaxL, dominant-negative EnRaxLΔC, which included most of the RaxL sequence outside of homeodomain region,seemed to maintain it's specificity and interfere mainly with the function ofRaxL, and not Rax. It is possible that the enhancer sequence recognized by Rax in progenitor cells is different from that recognized by RaxL in early photoreceptors. The finding that RaxL transactives the photoreceptor specific Ret1 enhancer element more efficiently than Rax supports this idea. However, this idea remains to be confirmed after the identification of the authentic binding elements of Rax andRaxL. It is also possible that the expression level of Raxis higher than that of RaxL in the retinal progenitor cells and that the expression level of EnRaxLΔC was not high enough to interfere withRax function. Alternatively, the function of Rax in progenitor cells may be dispensable since other paired-type homeodomain genes,e.g. Pax6 and Chx10 are highly expressed in retinal progenitor cells. The similar dominant-negative construct, EnRaxΔC,which contained most of Rax, had no effect on photoreceptor cell differentiation, further supporting the notion that the sequence outside of the homeodomain region provides significant specificity in ovo. Although EnRaxΔC can interfere with the transactivation activity of RaxL when assayed on a simplified reporter construct (RET1-CAT) in tissue culture cells,it appears not to function as a dominant-negative allele of RaxL on complex photoreceptor promoters in ovo. Our finding that interference with the endogenous RaxL activity by overexpression of EnRaxLΔC disturbs an early step in photoreceptor development, but not the general progenitor pool, suggests that only the progenitor population in the process of producing photoreceptor cells, or newborn photoreceptor cells, is affected. Without the proper activity of RaxL, photoreceptor cells cannot differentiate properly and, as a result, undergo apoptosis.

Several photoreceptor-specific transcription factors have been identified over the last several years. Among them, neuroD (now known asNeurod1) a basic helix-loop-helix gene, is expressed in retinal photoreceptors transiently in chick and is sufficient to generate more photoreceptors when overexpressed in chick retina(Yan and Wang, 1998). In mice,Neurod1 is expressed in retinal progenitor cells as well as in developing photoreceptor and amacrine cells, and is maintained in a subset of mature photoreceptors. Analysis of a Neurod1 knockout mouse and overexpression of Neurod1 in rats shows that it is not required for the initial formation of photoreceptor cells(Morrow et al., 1999). Thus the role of Neurod1 in photoreceptor cell development is not the same in chick and mouse, or perhaps it is not required in chick or mouse photoreceptor development. Further studies are needed to clarify its role.

Crx, an otx-like homeodomain gene, is expressed in newly generated photoreceptors, including both cones and rods, as well as at a low level in bipolar cells in mice and a high level in bipolar cells in zebrafish(Furukawa et al., 1997b;Chen et al., 1997;Liu et al., 2001). Interestingly, in zebrafish, Crx is expressed in mitotic cells presumably fated to produce photoreceptor cells, while in murine retinal cells, the expression of Crx appears to be initiated in cells that are fated to be photoreceptors, just after exit from the cell cycle. The timing of chick Crx expression appears to be the same at it is in mouse (T. Furukawa, personal communication). Functional studies in rodents have shown that Crx is required for a high level of expression of many photoreceptor specific genes. It is required for maturation, but not for the initial generation, of photoreceptors(Furukawa et al., 1997b;Livesey et al., 2000). Another important transcription factor in photoreceptor development is Nrl, a basic motif- leucine zipper transcription factor. Nrl is expressed in rod, but not cone, photoreceptors (Swain et al., 2001). It physically interacts with Crx and synergistically transactivates the rhodopsin promoter in vitro(Mitton et al., 2000). Analysis of Nrl mutant mice has revealed that it is a critical determinant of early rod photoreceptor cell development(Mears et al., 2001). A similar function is ascribed to Nr2e3 (also known as PNR),which encodes a ligand-dependent retinal nuclear receptor. Nr2e3 is expressed in photoreceptor cells(Kobayashi et al., 1999), and mutations in Nr2e3 lead to an increased number of cone cells in mice and the enhanced S cone syndrome, a disorder of photoreceptor cells, in humans(Haider et al., 2000;Haider et al., 2001). We provide evidence that a Rax family member, RaxL, is required for the initial generation of photoreceptors in chick. RaxL is expressed in cone photoreceptors. We hypothesize that RaxL andNrl are required for the early stages of cone and rod cell fate determination, respectively. Later in development, as cones and rods take up their final stages of differentiation, Crx plays the major role in supporting photoreceptor-specific gene expression. Overexpression ofCrx failed to rescue the photoreceptor phenotype induced by a dominant-negative allele of RaxL, further supporting the idea of early role of RaxL in photoreceptor development.

We are grateful to Dr A. Polans for the gift of visinin protein. We thank members of the Cepko and Tabin laboratories for helpful discussion and support. This work was supported by NIH grant EY09676.