Transcriptional networks defining stemness in adult neural stem cells (NSCs) are largely unknown. We used the proximal cis‐regulatory element (pCRE) of the retina‐specific homeobox gene 2 (rx2) to address such a network. Lineage analysis in the fish retina identified rx2 as marker for multipotent NSCs. rx2‐positive cells located in the peripheral ciliary marginal zone behave as stem cells for the neuroretina, or the retinal pigmented epithelium. We identified upstream regulators of rx2 interrogating the rx2 pCRE in a trans‐regulation screen and focused on four TFs (Sox2, Tlx, Gli3, and Her9) activating or repressing rx2 expression. We demonstrated direct interaction of the rx2 pCRE with the four factors in vitro and in vivo. By conditional mosaic gain‐ and loss‐of‐function analyses, we validated the activity of those factors on regulating rx2 transcription and consequently modulating neuroretinal and RPE stem cell features. This becomes obvious by the rx2‐mutant phenotypes that together with the data presented above identify rx2 as a transcriptional hub balancing stemness of neuroretinal and RPE stem cells in the adult fish retina.
This study establishes Rx2 as functional determinant of neuro‐epithelial progenitor fate and uncovers the gene regulatory network that governs Rx2 expression.
Rx2‐positive stem cells can give rise either to neuroretina or to retinal pigmented epithelium.
A transcriptional core network of Sox2, TLX, Her9, and Gli3 confines rx2 expression to the peripheral CMZ.
Repression of Rx2 (by Gli2 or in Rx2 mutant clones) favors formation of retinal pigmented epithelium.
Rx2 balances the fate decision of retinal stem cells towards retinal pigmented epithelium or neural retina.
Post‐embryonic neurogenesis relies on the activity of neural stem cells (NSCs) (Adolf et al, 2006; Zhao et al, 2008). The fish neural retina (NR) constitutes an ideal model to study embryonic and post‐embryonic NSCs in their physiological environment. It consists of seven main cell types distributed in three nuclear layers, and all these cell types are added lifelong from retinal stem cells (RSCs) that reside in the peripheral ciliary marginal zone (CMZ) (Johns, 1977; Lamba et al, 2008; Centanin et al, 2011). The CMZ also contributes to the retinal pigmented epithelium (RPE), a monolayer of pigmented cells surrounding and synchronously growing with the NR (Johns, 1977; Amato et al, 2004a; Moshiri et al, 2004; Centanin et al, 2011). As an attractive model for life‐long neurogenesis as well as growth of the RPE, the CMZ has been extensively studied in fish, frog, and chicken (Amato et al, 2004b; Raymond et al, 2006; Lamba et al, 2008).
The lack of genetic tools to follow lineages in these species, however, was a long‐lasting limitation to validate putative RSC‐specific markers, which in turn prevented understanding the regulatory framework that generates and maintains this stem cell fate. In medaka, retinal multipotent NSCs were identified by their virtue to form arched continuous stripes (ArCoS) via cell transplantation at early embryonic stages (Centanin et al, 2011) and by inducible recombination in late embryonic and post‐embryonic stages (Centanin et al, 2014). These experiments demonstrated that the maintenance of the NR and the RPE is achieved by independent RSCs located in the CMZ.
How the function of NSCs is maintained in the CNS remains largely unknown. Two of the best‐studied factors expressed by post‐embryonic NSCs in other niches are Sox2 and Tlx, which have been shown to be crucial for the self‐renewal and differentiation of NSCs (Monaghan et al, 1995; Graham et al, 2003; Shi et al, 2004). Furthermore, fate‐mapping studies showed that sox2 (Suh et al, 2007) and tlx (Liu et al, 2008) are markers for multipotent stem cells in the mammalian brain. Identifying target genes regulated by NSC‐determining TFs is crucial for the understanding of the molecular and signaling pathways underlying adult neurogenesis. Previously, Sox2 has been demonstrated to bind and regulate the gfap regulatory element (Cavallaro et al, 2008). Target genes regulated by Tlx include cell cycle regulators p21 and pten (Sun et al, 2007). The function of Tlx is context dependent; inhibition of target genes relies on the interaction with HDACs (Sun et al, 2007), while Tlx can also act as activator of transcription (Iwahara et al, 2009).
Characterizing a NSC regulatory network crucially depends on the identification of a reliable molecular marker and the main cis‐acting regulators controlling its proper expression pattern. Here, we followed a long‐term lineage analysis to identify the transcription factor (TF) retina‐specific homeobox gene‐2 (rx2) as a bona fide marker for multipotent adult RSCs in the post‐embryonic medaka CMZ. We show that individual rx2+ RSCs give rise to progeny contributing either to NR or to RPE. Next, we identified TFs acting upstream of rx2 in a trans‐regulation assay using the rx2 proximal cis‐regulatory element (pCRE) and characterize sox2, tlx, gli3, and her9 as transcriptional modulators of stem cell features in the post‐embryonic fish retina. Clonal expression of sox2 or tlx induces RSC‐specific characteristics, including ectopic rx2 expression in differentiated neurons of the central retina. Conversely, conditional clonal gain of Gli3 or Her9 function leads to rx2 repression and inhibits stem cell proliferation in the post‐embryonic CMZ. We also identify and validate the cis‐regulatory motifs within the proximal rx2 pCRE that activate expression in the peripheral most CMZ, and repress expression in the adjacent RPE. Additionally, we show that the rx2 expression levels resulting from this transcriptional network establish the balance between NR and RPE stem cells. Our in vitro and in vivo results provide evidence for the importance of direct TF‐DNA binding for proper spatial rx2 expression in RSCs. Taken together, we present a regulatory framework of TFs that establish, expand, and restrict RSC features in the post‐embryonic retina and demonstrate a crucial function of Rx2 in the definition of retinal stem cell types.
Rx2 labels the most peripheral cells in the ciliary marginal zone of the medaka retina
To specifically target RSCs in the CMZ, we followed a candidate gene approach and systematically searched for genes and their regulatory regions with expression confined to the CMZ. Both in amphibians and fish, the retina‐specific homeobox gene‐1 and retina‐specific homeobox gene‐2 (rx1 and rx2 respectively) are expressed in the peripheral CMZ at embryonic and post‐embryonic stages (Locker et al, 2006; Raymond et al, 2006; Borday et al, 2012). In medaka, rx2 is first expressed in the undifferentiated retinal progenitor cells (RPCs) that form the optic vesicle (Loosli et al, 1999). At later stages, when the fish retina already contains all the main cell types and is functionally active, rx2 expression is confined to photoreceptors (cones and rods in the outer nuclear layer, ONL), to the Müller glia cells, and to the peripheral most part of the CMZ, as revealed by in situ hybridization and immunostaining (Fig 1A and B) (Sinn et al, 2014). Medaka transgenic lines in which 2.4 kb of the proximal rx2 pCRE control the expression of a reporter fluorescent protein (FP) (rx2::H2B‐mRFP and rx2::Tub‐GFP) (Martinez‐Morales et al, 2009; Inoue & Wittbrodt, 2011) (Fig 1B and C; Supplementary Fig S1) exhibit the same expression pattern, indicating that the 2.4‐kb rx2 pCRE contains the regulatory cues driving rx2 expression to those cell types.
The post‐embryonic medaka retina grows outwards, with the RSCs located in the most peripheral domain of the CMZ (Centanin et al, 2011). A close analysis at rx2::H2B‐mRFP transgenic juveniles revealed that rx2+ cells locate to the peripheral most domain of the CMZ (Fig 1C), suggesting that Rx2 constitutes a marker for RSCs. On the central side of the CMZ, differentiating Atoh7+ cells demarcate the boundary between the central most domain of the CMZ and the differentiated, layered retina (Fig 1C) (Del Bene et al, 2007; Cerveny et al, 2010). As expected for stem cells, rx2+ cells in the post‐embryonic CMZ barely incorporate the thymidine analog bromodeoxyuridine (BrdU) when applied in short pulses (Supplementary Fig S2A) and similarly, there is only partial overlap with proliferation markers like phospho‐histone H3 (pHH3) and proliferating cell nuclear antigen (PCNA) (Supplementary Fig S2B and C). Most of the BrdU+, PCNA+, and/or pH3+ cells of the CMZ map to the progenitor domain, between the rx2+ cells and the Atoh7+ cells.
In medaka, transplantation of EGFP+ cells from the Wimbledon+/− line (a transgenic line that expresses EGFP ubiquitously during the entire life of the fish) (Centanin et al, 2011) into an unlabeled blastula results in fish with mosaic retinae. In these chimeric fish, RSCs were addressed by their property to form ArCoSs, which are continuous clonal strings of EGFP+ cells consisting of differentiated neurons and glia at central positions, and undifferentiated cells in the peripheral domain. We reasoned that if rx2+ cells were indeed RSCs, all the ArCoSs should contain rx2+ cells located at the most peripheral position of the clone. We therefore transplanted Wimbledon+/−, rx2::H2B‐mRFP donor cells into wild‐type hosts and consistently found rx2+ cells at the peripheral tip in all resulting ArCoSs (Fig 1D and E). Thus, the expression domain of rx2+ and the relative position of rx2+ cells within an ArCoS match the expected location of RSCs and suggest that rx2+ is a RSC marker in the mature medaka retina.
Rx2 is a molecular marker of adult RSCs
The ultimate validation of rx2+ as a stem cell marker is to follow the lineage of an individual rx2+ cell in the medaka CMZ during several months of post‐embryonic life. The Gaudí toolkit allows single‐cell labeling and lineage analysis in medaka (Centanin et al, 2014), based on Cre/LoxP‐mediated recombination. When fish bearing a ubiquitous Cre driver and a Gaudí reporter are induced for stochastic recombination, single RSCs generate induced ArCoSs (iArCoSs) of the same characteristics than the ArCoSs generated by transplantation (Centanin et al, 2014).
To address if rx2 marks RSCs and if the progeny of an rx2+ cell in the CMZ forms an iArCoS, we generated a transgenic line expressing a tamoxifen‐inducible Cre recombinase under the control of the well‐characterized 2.4‐kb rx2 pCRE (rx2::ERT2Cre). We induced Cre recombinase within the rx2 expression domain to trigger recombination in the ubiquitously expressed four‐color reporter cassette (Gaudí2.1) (Centanin et al, 2014). Tamoxifen induction of rx2::ERT2Cre, Gaudí2.1 at 10 dpf resulted in the specific labeling of individual rx2+ cells, as shown by stochastic expression of FPs labeling single photoreceptors, Müller glia, and peripheral cells in the CMZ (Fig 2A).
Long‐term lineage experiments showed that rx2+ cells formed iArCoSs and thus indeed represent RSCs (n = 162 red‐colored iArCoSs distributed over 7 retinae, ranging from 15 to 31 iArCoSs per retina, average 23.1 iArCoSs/retina) (Fig 2B). Lineage analyses indicate as well that every single rx2+ post‐embryonic NR stem cell analyzed is multipotent, equivalent to ArCoS‐forming RSCs in transplantation experiments. Each iArCoS contains the full repertoire of differentiated retinal cell types of the NR (n= 125 NR iArCoSs) (Fig 2C).
Remarkably, rx2+ RSCs as a population give rise to iArCoSs in both the NR and the RPE (Fig 2B). An individual rx2+ RSC, however, can be a stem cell either for the NR or for the RPE (n = 25 NR iArCoSs and 136 RPE iArCoS distributed over 8 retinae, 98.5% of independent iArCoSs).
Identification of Sox2, Tlx, Gli3, and Her9 as transcriptional regulators in control of retinal stemness
To identify genes controlling RSC features, we followed a high‐throughput trans‐regulation screen (Souren et al, 2009) to systematically detect factors operating on the rx2 pCRE. The trans‐regulation screen is based on two nested screens. In a first step, it employs a high‐throughput luciferase assay based upon the co‐expression of an rx2 pCRE reporter construct driving firefly luciferase together with individual full‐length candidate TFs (Fig 3A). This cell culture‐based assay allows transcriptome scale analyses and has been used reliably to identify so far unknown upstream regulators (Souren et al, 2009). We took advantage of the relatively short 2.4‐kb proximal rx2 CRE sufficient to recapitulate the rx2 expression pattern and assayed more than one thousand individual full‐length cDNA clones, which represented a large complement of all putative medaka TFs. We controlled for transfection efficiencies in a dual luciferase‐based screen in cultured cells through co‐transfection of a control plasmid encoding Renilla luciferase (Fig 3A). To exclude potential false positives, we performed a secondary, nested, whole‐mount in situ screen to analyze the expression pattern of putative candidate TFs relative to rx2 by a semi‐automated whole‐mount in situ hybridization approach (Quiring et al, 2004). We eventually selected activating or repressing candidates based on their co‐expression with or adjacent to rx2 in the juvenile CMZ.
This nested screening pipeline delivered clear candidates from the more than one thousand TFs analyzed: sox2 was the top activator, while gli3 and her9 (a medaka Hes1 ortholog) showed the strongest repressive activities. tlx—not initially present in the full‐length TF library—showed a strong activation of the rx2 pCRE (Fig 3C) and was assayed in a parallel candidate screen because of its role in mouse NSCs (Yu et al, 1994; Monaghan et al, 1995; Shi et al, 2004). To test whether Sox2, Tlx, Her9, and Gli3 regulate rx2 transcription in a concentration‐dependent manner, we performed dual luciferase assays with increasing amounts of the respective TF cDNA. For Sox2 (Fig 3B), we observed the activation of relative luciferase activity in a dose‐dependent manner. Likewise, for Tlx (Fig 3C) activation of transcription peaked with the highest cDNA concentration (160 ng), implicating tlx as an activator of rx2 expression. Conversely, stepwise increase of Her9 resulted in the gradual reduction of reporter expression (Fig 3D). Interestingly, Gli3‐mediated repression of rx2 pCRE activity was strongest at the lowest Gli3 concentration (Fig 3E), while increasing cDNA amounts led to a gradual reduction of its repressive potential.
Next, we addressed the expression patterns of sox2, tlx, gli3, and her9 with respect to their putative target gene rx2 in the juvenile CMZ by two‐color fluorescent whole‐mount in situ hybridization (WISH). All four regulators are expressed in nested domains that partially overlap with the rx2 expression domain in the CMZ. We detected transcripts of the pan‐neural determinant sox2 throughout the CMZ overlapping with the Rx2 expression domain (Fig 3F–H). tlx and her9 were both expressed in the central CMZ where they partially overlapped with the rx2 expression domain (Fig 3I–N). gli3 transcripts were found in the peripheral CMZ overlapping with rx2 expression and were also found in the adjacent RPE (Fig 3O–Q). Out of all the rx2 regulators identified in the trans‐regulation screen, gli3 was the only factor expressed in the peripheral RPE adjacent to the CMZ.
Gli3 and Her9 antagonize stem cell features in vivo
To test whether these candidate factors regulate rx2 expression in vivo, we employed a conditional clonal analysis in the post‐embryonic retina. For this purpose, we adopted a hormone‐inducible binary gene expression system, which consists of a TF (LexPR) that upon hormone induction will dimerize and bind to the corresponding promoter element (LexOP) to activate the expression of following genes of interest (Emelyanov & Parinov, 2008). We established transgenic lines expressing LexPR in the CMZ under the control of the rx2 pCRE (Fig 4A). Upon addition of mifepristone (RU‐486), dimerization and LexOP‐dependent transcription were initiated (Fig 4B). By limited exposure to the hormone, we triggered mosaic expression of gli3 or her9 (and the co‐expression of a fluorescent reporter protein) within the very specific Rx2 domain in the CMZ and Müller glia cells as well as in mature photoreceptor cells. This system allows co‐expressing fluorescent reporter proteins and the gene of interest with high efficiency (up to 90%). Consequently, the analyses based on the expression of the fluorescent reporter are conservative and always underestimate the effects of the gene of interest.
To assess the repressive potential of gli3 and her9, respectively, on the expression of rx2 in vivo, we targeted their clonal expression to the peripheral rx2+ CMZ. Gain‐of‐function clones were highlighted by the co‐expression of nuclear FPs (Fig 4F and I). The expression of Rx2 protein was determined by antibody staining and analyzed automatically by a threshold‐based segmentation algorithm. The repression of the pCRE of rx2 by Her9 and Gli3 in the CMZ in this experiment not only affects the endogenous Rx2 but also the ectopic expression of the repressors. Consequently, we consistently underestimated the repressive potential Her9 and Gli3 on rx2. Six days after the clonal induction of Gli3 in the peripheral CMZ (rx2::LexPR LexOP::Gli3 LexOP::H2B‐EGFP), Rx2 protein was lost in 39.4% of the induced gli3+ cells (n = 13/33) in the CMZ (compare Fig 4C–E to F–H). Similarly, 20% of the induced her9+ cells (rx2::LexPR LexOP::Her9 LexOP::H2B‐EGFP) (n = 7/35) had lost Rx2 expression (compare Fig 4C–E to I–K), whereas the GFP alone had no effect on the rx2 expression (n = 1/34) (Fig 4C–E). These findings are in support of the hypothesis of Her9 and Gli3, confining rx2 expression as repressors in the central and peripheral domain of the CMZ.
Since both Gli3 and Her9 showed the ability to repress rx2 expression in vivo, we next addressed the impact of ectopic gli3/her9 expression on the proliferative capacity of RSCs in the CMZ. PCNA staining, which labels S‐phase cells within the CMZ, was low only in very few cells in the CMZ of GFP control retinae (n = 3/38). Conversely, PCNA expression was affected in the gli3 and her9 gain‐of‐function clones at rates comparable to those observed for the repression of Rx2. PCNA was severely affected in gli3 gain‐of‐function clones (n = 39/106) (compare Fig 4L–N to O–Q) or her9 gain‐of‐function clones (n = 21/64) (compare Fig 4L–N to R–T). Taken together, these results indicate that ectopic clonal gli3 and her9 expression in the CMZ represses rx2 expression and impacts negatively on the proliferation of RSCs.
Sox2 and Tlx promote rx2 expression
Since our in vitro characterization and the overlapping expression pattern of sox2 and tlx with rx2 consistently argued for an activating function of sox2 and tlx, we tested the consequences of acute clonal activation of sox2 (cska::LexPR LexOP::sox2) and tlx (cska::LexPR LexOP::tlx) gain‐of‐function (Fig 5A). Gain‐of‐function clones were marked by the expression of red FPs (LexOP::cherry), encoded by co‐injected reporter plasmids. In combination with the ubiquitous cska promoter (Grabher et al, 2003), this approach allowed clonal, mosaic expression throughout all three nuclear layers of the differentiated retina (Fig 5B).
We first examined the consequences of Sox2 and Tlx co‐expression on rx2 promoter activity in vivo. The transgenic reporter line expressed FP under the rx2 pCRE (rx2::Tub‐GFP). The combined expression of sox2 and tlx resulted in strong rx2 reporter activation (n = 40/48) in all three nuclear layers (Fig 5C and D). Individual clonal mis‐expression of sox2 (n = 32/48) (Supplementary Fig S3A–D) or tlx (n = 142/173) (Supplementary Fig S3F–I) also resulted in ectopic rx2 reporter activation with high efficiency. To corroborate that sox2 and tlx activate endogenous rx2 expression in vivo, we combined WISH and immunohistochemistry in whole‐mount preparations. Clones expressing sox2 (n = 53/62) (Supplementary Fig S3E–E‴) or tlx (n = 34/56) (Supplementary Fig S3J–J‴) efficiently triggered the ectopic expression of endogenous rx2 mRNA, which was never detected in controls.
We next asked whether the clonal activation of sox2 or tlx was sufficient to trigger the ectopic induction of RSC features. While control clones in central retinal cell types never showed proliferating activity, the expression of Sox2 (n = 7/11) (Fig 5E–I) or Tlx (n = 3/23) (Fig 5J–N) resulted in the re‐acquisition of proliferative features as indicated by PCNA staining. PCNA+ clones were observed in the INL and ONL, indicating that de‐differentiation and re‐initiation of proliferation were not restricted to one particular type of retinal neurons. Together, these results revealed that both sox2 and tlx induce the endogenous expression of rx2 in vivo, and re‐activate the proliferative potential of post‐mitotic cells.
Sox2 and Gli3 regulate rx2 expression through direct protein–DNA interaction with the rx2 pCRE
Our analysis showed that sox2 and tlx trigger some RSC features in differentiated retinal cells, while gli3 and her9 constrain rx2 expression and stem cell proliferation in the CMZ. To investigate whether these effects are mediated by direct trans‐regulation on the rx2 pCRE, we carried out a combination of in vitro and in vivo analyses.
We started by performing an evolutionary footprinting analysis on the cis‐regulatory elements present in the rx2 pCRE. We identified evolutionarily conserved binding sites for Gli (Sasaki et al, 1997) and for Sox (Danno et al, 2008), the strongest repressor and activator in the in vitro assay, respectively (Fig 6A). Direct interaction between Sox2 and the Rax promoter has been demonstrated in Xenopus (Danno et al, 2008). To test whether Sox2 binds to the predicted, putative Sox‐binding site in the medaka rx2 pCRE, we next performed electromobility shift assays (EMSA). Sox2 showed sequence‐specific binding to the rx2 pCRE, which was weakened upon mutation of the predicted Sox TFBS (Fig 6B). Similarly, the specific affinity of Gli3 protein to the rx2 pCRE was abolished when we tested the rx2 pCRE lacking the Gli‐binding motif (Fig 6B).
Next, we investigated whether the transcriptional regulation of rx2 expression is impaired upon mutation of the identified Sox or Gli TFBSs. For this, we employed the luciferase trans‐regulation assay and compared the transcriptional activity of rx2 pCREs with mutated Sox or Gli motifs (rx2 pCRE mtSox and rx2 pCRE delGli) to the corresponding wild‐type activity (Fig 6C). The transcriptional activation of the rx2 pCRE through the predicted Sox2‐binding site was severely attenuated by the introduction of mutations into the Sox‐binding site (from 50.99 ± 2.19 to 9.93 ± 1.41) (Fig 6D). Conversely, absence of the Gli TFBS lifted the strong Gli3‐mediated repression of the rx2 pCRE (0.37 ± 0.01 to 0.56 ± 0.11) (Fig 6D). These results of the trans‐regulation assays underscore the significance of both identified motifs for the regulation of rx2 expression through Sox2 and Gli3.
To gain further insight into the molecular mechanism underlying the regulation of rx2, we performed chromatin immunoprecipitation (ChIP)‐PCR assays. To this end, rx2 pCRE luciferase plasmids co‐expressed with GFP‐tagged Sox2 or Gli3 fusion proteins were analyzed by immunoprecipitation of chromatin with antibodies directed against GFP. Subsequent amplification of co‐precipitated DNA by qPCR using oligonucleotides flanking the specific TFBSs and the luciferase coding sequence (control), respectively, revealed a specific binding of both factors analyzed to their in vitro‐validated binding sites. This was indicated by a five‐ to sevenfold enrichment of the rx2 pCRE (1st ChIP: 7.71; 2nd ChIP: 5.18) in comparison to the control (Fig 6E) in the Sox2 ChIP. Upon mutation of the Sox motif, this enrichment was severely reduced (1st ChIP: 3.18; 2nd ChIP: 3.03), indicating a specific binding of Sox2 to the predicted binding site in vivo. Similarly, Gli3 was binding to the predicted site in the rx2 pCRE, though with lower affinity (1st ChIP: 1.38; 2nd ChIP: 1.41). In the absence of the binding site (rx2 pCRE delGli), no input enrichment (1st ChIP: 1.07; 2nd ChIP: 1.10) was detectable, indicating the specificity of the interaction (Fig 6F). Similar results were obtained for the TFs Tlx (Yu et al, 1994) and Her9 (Sasai et al, 1992), where we could show a direct site‐specific binding to the rx2 pCRE in EMSA (Supplementary Fig S4) and ChIP‐PCR (Supplementary Fig S5, Supplementary Table S2), respectively. In summary, our biochemical analysis and transcriptional assays revealed binding sites for the TFs Sox2 and Gli3, which are bound by the corresponding factors in a sequence‐specific manner to regulate the activity of the rx2 pCRE.
Sox2 and Gli3 confine Rx2 expression to adult RSCs through direct protein–DNA interaction
To gain insight into the biological relevance of the predicted and characterized Gli‐ and Sox‐binding sites for rx2 expression in vivo, we introduced the same modifications that impaired physical interaction with the respective TFs. We generated stable transgenic lines (rx2 pCRE delGli and rx2 pCRE mtSox) expressing nuclear FPs (Fig 7A) and assayed the impact of Gli3 and Sox2 on Rx2 reporter expression in vivo.
Gli is required for confining Rx2 expression to the periphery of the CMZ and at the same time necessary to prevent its expression in the peripheral RPE. Upon interference with Gli binding to the rx2 pCRE (rx2::H2B‐mRFP delGli), rx2 reporter expression was shifted to the Gli3‐expression domain in the RPE (Fig 7C), consistent with Gli3 acting as an in vivo rx2 repressor (n = 15/15 hatchlings). Simultaneously, rx2 reporter expression was massively reduced in the peripheral CMZ (Fig 7C–C‴), indicating that the single Gli‐binding site identified and characterized mediates both activating (CMZ) and repressing (RPE) activities. These data corroborate the in vitro findings (above) and highlight their relevance in vivo.
The binding of Sox2 is necessary for the expression of rx2 in the CMZ. Mutation of the Sox‐binding site (rx2::H2B‐mRFP mtSox) resulted in massive reduction of rx2 reporter expression in the CMZ, both in terms of levels and cell numbers (Fig 7E–E‴) (n = 14/14 hatchlings) compared to controls (rx2::H2B‐mRFP) (Fig 7D–D‴). Interestingly, the mutation of the Sox‐binding site also affected rx2 reporter expression in Müller glia cells, the only other retinal cell type that co‐expresses sox2 and rx2. In contrast, rx2 mtSox reporter expression was unaffected in post‐mitotic photoreceptor cells, which express rx2, but not sox2 (compare Supplementary Fig S6A–A‴ to Supplementary Fig S6B–B‴).
These results indicate a critical in vivo role of the Sox‐binding site and its interaction with the TF Sox2 for the onset and maintenance of rx2 expression in RSCs in the CMZ and in Müller glia cells. Our analyses support a dual role for the TF Gli3 as a direct context‐dependent mediator of rx2 activation and repression, respectively.
Rx2 activity is required for the balance between RPE and neuroretinal stem cells
To ultimately address the role of Rx2 in the retinal stem cells in the peripheral CMZ, we established mutant alleles targeting the rx2 locus with transcription activator‐like effector nucleases (TALENs). Customized TALEN pairs were designed to bind the coding sequence in the homeodomain of the rx2 gene. Mutations were introduced by injection of mRNA (Ansai et al, 2013) encoding the Rx2‐specific TALEN pair at the one‐cell stage. Successful targeting of the rx2 locus was validated by PCR and followed by sequence analysis in the injected and subsequent generation. Embryos carrying a homozygous mutation in rx2 (RNA null) are transiently delayed in retinal development but ultimately develop morphologically normal eyes with slightly reduced eye size (Fig 8A). They are viable and can be maintained as homozygous stock (Fig 8A). The spatiotemporal redundancy of rx2 expression in comparison to other highly related paralogous genes (rx3 in RPCs, rx1 in RSCs) presents a likely argument for the lack of a strong phenotype in mutant fish.
To challenge the capacity of rx2‐mutant cells to contribute to retinal stem cells, we generated chimeric retinae by transplanting labeled wild‐type cells into rx2‐mutant embryos. When wild‐type blastula cells are transplanted into a wild‐type blastula host, they contribute to both stem cell populations, either to neuroretinal stem cells or to RPE stem cells as indicated by the formation of NR ArCoS and RPE ArCoS, respectively (34 NR ArCoSs and 13 RPE ArCoSs distributed in eight retinae; NR/RPE ratio = 2.6) (Fig 8B and C). Conversely, in the context of a rx2‐mutant background, wild‐type blastula cells transplanted into a mutant host preferentially give rise to neuroretinal stem cell ArCoSs, and contribute only occasionally to the RPE (24 NR ArCoSs and three RPE ArCoSs distributed in eight retinae; NR/RPE ratio = 8) (Fig 8D). Thus, our mosaic analysis revealed a requirement for Rx2 activity in balancing the contribution to the stem cell pools of neuroretinal stem cells and RPE stem cells. In the absence of Rx2 activity, RPE is preferentially formed, resulting in the exclusion of wild‐type contribution in a mosaic situation. These data corroborate the findings presented above and, altogether, highlight an intricate connection between fate determinations of stem cell populations via the activity of Rx2, which ensures the proper balance of the two stem cell populations.
Here, we identify and characterize the TF rx2 as a proxy for retinal stem cells with a key function in balancing retinal stem cell populations and establish a scaffold network of directly interacting TFs that control and likely confine rx2 expression to RSCs in vivo. Strikingly, an individual rx2+ cell is a stem cell either for the NR or for the RPE. Each individual rx2+ NSC is multipotent and gives always rise to the full complement of retinal cell types. By identifying upstream regulators of rx2, we have gained mechanistic insight into the transcriptional control of retinal stemness. The bifunctional rx2+ stem cell domain within the CMZ is defined by both transcriptional activators (Sox2, Tlx) and transcriptional repressors (Gli3, Her9) (Fig 9A). We hypothesize that the combined activities of those factors confine Rx2 expression and facilitate the balanced establishment of stem cells for the NR and the RPE within the CMZ.
We followed an unbiased approach (Souren et al, 2009) to initially identify proteins interacting with the rx2 CRE in vitro. This approach has clear limitations due to the composition of the initially screened gene set and the fact that we only assayed for the activity of individual genes. Nevertheless, we successfully identified key transcriptional regulators of rx2 that directly interact with the rx2 CRE in vitro and in vivo and exert key functions in confining the rx2+ RSCs to the periphery of the CMZ.
Conditional mosaic expression of individual rx2 activators in the central retina resulted in the de‐differentiation of post‐mitotic retinal cells and the induction of stem cell features therein. These terminally differentiated cells are efficiently reprogrammed by the combination of Tlx and Sox2, consistent with reports in cultured NSCs, where a direct interaction between Sox2 and Tlx alleviates a negative feedback loop acting on the tlx promoter (Shimozaki et al, 2012). Additionally, both factors have been shown to be crucial for the maintenance of mammalian NSCs (Bylund et al, 2003; Graham et al, 2003; Shi et al, 2004; Taranova et al, 2006; Liu et al, 2008). We hypothesize that both factors synergize in the central CMZ to establish the niche for neuroretinal stem cells (Fig 9A).
Our data connect Sox2, Tlx, Gli3, and Her9 forming a core scaffold, which converges on the rx2 pCRE and modulates rx2 expression in adult RSCs (Fig 9B). Because these TFs are known to participate in non‐exclusive protein–protein interactions (Kageyama et al, 2007; Qu & Shi, 2009; Kondoh & Kamachi, 2010), it is possible that a combinatorial code with additional cofactors modulates the spatiotemporal function and activity of the core components of the network. Previous studies in Xenopus reported the interdependence of Sox2 and Otx2 for rx activation (Danno et al, 2008; Martinez‐De Luna et al, 2010). Even though we failed to detect Otx2 transcripts in the CMZ (data not shown), we cannot entirely exclude the presence of Otx2 at basal levels in the trans‐regulation assays. However, the graded response to increasing concentrations of Sox2 strongly argues for a Sox2‐mediated rx2 regulation independent of Otx2. Since fish and amphibia possess the highest regenerative capacity amongst vertebrates and show lifelong, post‐embryonic retinogenesis, the difference in rx regulation is unexpected. Addressing how the regulatory scaffold presented here has evolved from fish over frogs to mammals will present one aspect toward our understanding of the gradual loss of proliferative and regenerative capacity in the retina of higher vertebrates.
One of the striking findings of this work relates to the fact that a single retinal stem cell expressing rx2 can either be a multipotential stem cell for the neuroretina or be a stem cell for the RPE. Here, distally adjacent to the CMZ, Gli3 (Fig 9A, blue) acts as repressor, confining rx2 expression (Fig 9A, red) to the CMZ. The shift of rx2 reporter expression from the CMZ to the RPE upon deletion of the Gli‐binding site underlines a dual function of Gli3: It mediates rx2 repression in the RPE as well as rx2 activation in the directly adjacent CMZ. These data together with the fact that rx2‐mutant cells preferentially contribute to the RPE RSC pool (discussed below) indicate that Rx repression (mediated by Gli 3 in the wild‐type retina) is a prerequisite for commitment toward RPE fate.
Here, Gli TFs (Fig 9A, purple) function as activators in the presence of a Shh signal (gray) emanating from the RGC layer (Borday et al, 2012) and crucially contribute to rx2 expression at this site. This can be modulated by additional Gli TFs expressed in the CMZ (data not shown) as reported in fish and frog, regulating rx2 expression via the same evolutionarily conserved motif. Additionally, it has been reported that Gli TFs act as direct transcriptional repressors or activators dependent on the activity of the Hh signaling cascade (Humke et al, 2010) (Fig 9B), a pathway shaping the CMZ (Borday et al, 2012). While multiple Gli proteins are potentially involved in the regulation of rx2 expression in vivo, no SoxB1 genes other than sox2 are expressed in the post‐embryonic medaka CMZ (not shown). Thus, the activity uncovered for Sox2 and its binding site in the rx2 pCRE can be unambiguously attributed to Sox2. Recent studies provided evidence for synergistic action between SoxB1 and Gli TFs to activate Shh target genes during neural tube development (Oosterveen et al, 2012; Peterson et al, 2012). It has been shown that Sox and Gli TFs occupy the same regulatory elements of common targets (Peterson et al, 2012). Given the close proximity of the tested Sox and Gli motif in the rx2 pCRE. A Gli activator could exert its function in concert with Sox2 in the CMZ to activate Rx2 and possibly modulate RSC fate toward the NR. In the RPE, a Gli repressor in the absence of Hh signal could use the same Gli site to restrict rx2 expression independent of Sox2. Interestingly, non‐coding regions bound by Gli1 included a regulatory region of sox2 in neural progenitors (Peterson et al, 2012); therefore, Hh‐dependent Gli may help to maintain sox2 expression in the CMZ in order to keep stem cells and progenitors in an undifferentiated state. The relevance of Gli‐mediated repression of rx2 is corroborated by the requirement of rx2 for balancing neuroretinal and RPE fates. In the absence of rx2 (by repression or mutation), RPE fate is strongly favored. Intriguingly, rx2‐mutant cells display a striking phenotype when challenged with wild‐type cells within the same niche. In a mosaic retina, the absence of Rx2 activity favors the formation of RPE and consequently prevents wild‐type cells to take RPE stem cell fate. Therefore, our data not only highlight an intricate connection between fate determinations of both stem cell domains via the activity of Rx2. They furthermore also imply a cell non‐autonomous feedback activity ensuring a balance of both stem cell populations. The transcriptional confinement of rx2 to the CMZ by the activity of the regulatory scaffold presented here connects the stem cells of the NR and those of the RPE and thus sheds light on the mechanism specifying this composite stem cell niche.
Materials and Methods
Medaka (Oryzias latipes) stocks were maintained as previously described (Koster et al, 1997). All fish are maintained in the closed stocks of COS at Heidelberg University. Fish husbandry and experiments were performed according to local animal welfare standards (Tierschutzgesetz 111, Abs. 1, Nr. 1, Haltungserlaubnis) and in accordance with European Union animal welfare guidelines. The fish facility is under the supervision of the local representative of the animal welfare agency. Embryos were staged according to Iwamatsu (2004).
Injections were done as previously described (Rembold et al, 2006b). For transient expression, driver and effector plasmids were co‐injected (30 ng/μl final concentration) in two‐cell stage medaka embryos (Rembold et al, 2006a).
LexPR and LexOP cassettes were derived from pDs(krt8:LPR‐LOP:G4) and pDs(cry:C‐LOP:Ch) to generate driver, effector, or driver–effector constructs (Emelyanov & Parinov, 2008). A cassette containing the LexOP operator upstream of the Cherry coding sequence was extracted from pDs(cry:C‐LOP:Ch). The Cherry coding sequence was replaced with H2B‐EGFP and H2A‐Cherry. Effectors: LexOP::H2A‐Cherry; LexOP::H2B‐EGFP; LexOP::Cherry. A cassette containing the coding sequence for the LexPR trans‐activator followed by the LexOP operator was released from pDs(krt8:LPR‐LOP:G4) and inserted downstream of the rx2 pCRE (Inoue & Wittbrodt, 2011). Coding sequences for gli3 and her9 were inserted downstream of the LexOP operator. A second LexOP operator followed by H2B‐EGFP coding sequence was added (released from LexOP::H2B‐EGFP). Driver–Effectors: rx2::LexPR LexOP::Gli3 LexOP::H2B‐EGFP; rx2::LexPR LexOP::Her9 LexOP::H2B‐EGFP; rx2::LexPR LexOP::H2B‐EGFP.
A cassette containing the coding sequence for the LexPR trans‐activator followed by the LexOP operator was introduced downstream of the cska promoter (Grabher et al, 2003). Coding sequences for sox2 and tlx were inserted downstream of the LexOP operator. Drivers: cska::LexPR LexOP::Sox2; cska::LexPR LexOP::Tlx,; cska::LexPR LexOP.
The coding sequence of the nuclear FPs in the rx2::H2B‐mRFP vector (Inoue & Wittbrodt, 2011) was replaced by tubulin‐GFP fusion. The 2.4‐kb rx2 pCRE was released through restriction digest and cloned upstream of the luciferase gene into the pGL4.1 luciferase reporter vector (Promega).
Coding sequences for sox2, gli3, and her9 were derived from a full‐length cDNA library based on the pCMV‐Sport6.1 vector (Souren et al, 2009); tlx cDNA was derived from a Lambda ZAP cloning vector. Full‐length Olrx2 (NP_001098373.1) for antibody generation was cloned by PCR from medaka stage 32 cDNA using the following primers: forward primer: 5′‐GGAATTCCATATGATGCATTTGTCAATGGATAC‐3′; reverse primer: 5′‐ CGGGATCCTCACATGTGCTGCCAGG‐3′. PCR products were digested with restriction enzymes NdeI and BamHI, ligated into the pET15b (Merk Millipore), which was cleaved with the same enzymes. pET15b‐Olrx2 was used to bacterially express OlRx2 protein as the antigen for generation of OlRx2 antibody.
The 2.4‐kb Medaka rx2 pCRE was cloned in a pBS/I‐SceI already containing a tamoxifen‐inducible Cre recombinase (from pIndu Perfect) resulting in pBS/I‐SceI/rx2::ERT2CRE/I‐SceI. All constructs generated for transient expression or transgenesis were based on a pBluescript plasmid containing two I‐SceI sites flanking the insert.
N‐terminal mGFP‐flexible linker (FL) fusion protein constructs (mGFP‐FL‐Her9, mGFP‐FL‐Sox2, and mGFP‐FL‐Tlx) for ChIP analysis were generated by Golden GATEway (GGW) cloning system (Kirchmaier et al, 2013). Routinely, respective genes and FL were cloned into appropriate GGW entry vectors and assembled into GGW destination vectors. These GGW destination vectors were used to clone respective mGFP‐FL fusion constructs into pSport6.1 vector. FL was generated by annealing with the sense oligo: GATCCTCCCTGAGCGGTGGAGGCGGTTCAGGCGGAGGTGGCTCTGGCGGTGGCGGATCGGGAGGCGGTGGAAGTGCAGCCGCGGGTG and the antisense oligo: GTACCACCCGCGGCTGCACTTCCACCGCCTCCCGATCCGCCACCGCCAGAGCCACCTCCGCCTGAACCGCCTCCACCGCTCAGGGA. For mGFP‐FL‐Gli3, mGFP‐FL was released from the assembled pSport6.1 vectors via restriction digest and inserted upstream of pSport6.1‐Gli3.
Cell culture, transfection, and luciferase readout were carried out with 1,151 cDNA clones representing the majority of the annotated medaka TFs as previously described (Souren et al, 2009). For each well, pSport6.1‐cDNA (100 ng), rx2::luc2 (40 ng), and pRL‐CMV (5 ng) were co‐transfected.
Amino acid substitutions or deletions were introduced into the rx2 gene regulatory region of rx2::H2B‐mRFP by site‐directed mutagenesis (Quikchange, Stratagene) using the following primers: putative Sox‐binding site, forward primer: 5′‐CCACACAAGCCATTATCTTTCAGACGCTAGATTTGTTGAAAGGAAGTTTTGT‐3′; reverse primer: 5′‐ACAAAACTTCCTTTCAACAAATCTAGCGTCTGAAAGATAATGGCTTGTGTGG‐3′; Gli‐binding site, forward primer: 5′‐GAAGTTTTGTTGAGGCTTCATTAGCAATGTGGTCTGAAAGCAG‐3′; and reverse primer: 5′‐CTGCTTTCAGACCACATTGCTAATGAAGCCTCAACAAAACTTC‐3′.
In vitro translation and EMSA
Medaka Sox2, Tlx, Gli3, and Her9 proteins were generated from cDNA clones by in vitro transcription and translation using the TNT Quick Coupled Transcription/Translation System (Promega). Complementary oligonucleotides were annealed in annealing buffer (10 mM Tris–HCl, pH 8.0, 50 mM NaCl, 1 mM EDTA). Biotin‐labeled oligonucleotides (25 fmol) were incubated with 5 μl protein translation reaction for 1 h at 4°C in binding buffer (for Gli3 and Tlx: 10 mM Tris, 1 mM DTT, 5 mM MgCl2, 5% glycerol, 1 mM EDTA, 0.05% NP‐40, and 50 ng/μl poly(dI‐dC); for Sox2 and Her9: 10 mM Tris, 1 mM DTT, 2.5 mM MgCl2, 100 mM KCl, 5% glycerol, 1 mM EDTA, 0.05% NP‐40, and 50 ng/μl poly(dI‐dC)) in 20 μl total volume.
The binding reactions were subjected to electrophoresis on a pre‐run native 6% polyacrylamide gel in 0.5× TBE. Signal was detected using the LightShift Chemiluminescent EMSA kit (Thermo Scientific) according to the manufacturer's instructions. Oligonucleotide sequences used are summarized in Supplementary Table S3.
pGL4.1 vector containing the rx2 pCRE with different mutations (40 ng) and pRL‐CMV vector (5 ng) were co‐transfected with 20, 40, 80 or 160 ng of medaka cDNA (pSport6.1‐cDNA) per well. A total amount of 205 ng DNA was transfected in each well through addition of empty pCS2+ vector. The assays were carried out in quadruplicate.
ChIPs were performed with the SimpleChIP Enzymatic Chromatin IP kit (New England Biolabs) according to the manufacturer's instructions. BHK cells were co‐transfected with 6 mg pGL4.1‐rx2 and 3 mg pSport6.1‐mGFP‐FL‐cDNA. Chromatin was incubated with 20 μl of agarose beads attached to anti‐GFP antibodies (Chromotek) at 4°C overnight. Quantification of DNA by PCR was carried out in a thermocycler (Bio‐Rad) using ABsolute qPCR SYBR Green Mix (Thermoscientific). Primers used to amplify regions containing putative binding sites within the rx2 pCRE and control primers for firefly luciferase coding sequence are summarized in Supplementary Table S4. Two independent biological replicates were carried out for each genomic region of interest. Each qPCR was performed as duplicates.
Whole‐mount in situ hybridization
For anti‐sense riboprobe synthesis, linear templates were produced from full‐length cDNA clones either through standard PCR (sox2, gli3 and her9) or restriction enzyme digestion 5′ of the start codon (rx2 and tlx). T7 RNA polymerase‐based transcription was carried out as previously described (Loosli et al, 1998). Fluorescent whole‐mount in situ hybridizations were performed as previously described (Schuhmacher et al, 2011). Signals were detected using TSA‐Plus Fluorescein and Cyanine 5 Systems (Perkin Elmer). For combined single‐color fluorescent whole‐mount in situ hybridization and immunostaining, embryos were incubated for 2 days with anti‐fluorescein antibody conjugated to horseradish peroxidase (Roche) and anti‐GFP antibody (Invitrogen) at 4°C. After riboprobe detection using TSA‐Plus Cyanine 3 System (Perkin Elmer), the embryos were incubated with fluorescent‐conjugated secondary antibody and 4′,6‐diamidino‐2‐phenylindole (DAPI, 1:500, Sigma) for 2 days at 4°C. Whole‐mount in situ hybridizations using NBT/BCIP detection were carried out as previously described (Loosli et al, 1998). 25‐μm‐thick sections were obtained using a VT1000S vibratome (Leica) after mounting stained embryos in 4% agarose (Sigma).
16‐μm cryosections and immunostainings were preformed as previously described (Inoue & Wittbrodt, 2011). When necessary, cryosections were subjected to 3% H2O2 in 1% KOH for 30 min prior to immunostainings. Anti‐OlRx2 antibody was raised against the full‐length OlRx2 (NP_001098373.1) recombinant protein in rabbits (Charles River), and affinity‐purified as described previously (Barenz et al, 2013). Primary antibodies used were rabbit anti‐phospho‐histone H3 (1:500; Upstate), mouse anti‐PCNA (1:100; Santa Cruz), rabbit anti‐OlRx2 (1:500), chicken anti‐GFP (1:500, Invitrogen), rabbit anti‐DsRed (1:250, Clontech), mouse anti‐BrdU (1:50, Becton Dickinson), mouse anti‐Islet (1:250, DSHB), mouse anti‐GS antibody (1:50, Chemicon), and anti‐chicken, anti‐mouse, or anti‐rabbit fluorescent secondary antibodies (1:1,000, DyLight488, DyLight549 and DyLight647, Jackson). Cell nuclei were counterstained with DAPI. For BrdU detection, cryosections were dried overnight, rehydrated with 1× PTW (1× PBS pH 7.3, 0.1% Tween) and treated with 2 N HCl for 90 min, and then blocked with 10% sheep serum (Sigma).
Embryos were incubated in a solution of 1 g/l BrdU for 3 days and fixed immediately afterward in 4% paraformaldehyde.
Mifepristone (Sigma, Tocris, Cayman) was dissolved in dimethylsulfoxide (DMSO) to a final concentration of 25 mM and stored as stock solution at −20°C. The stock solution was added to the medium and used at final concentrations up to 20 μM.
Tamoxifen was used at a concentration of 2.5 mM (T5648, Sigma) and applied for 12–24 h. Embryos were washed several times after the treatment. A 50 mM stock solution in DMSO was stored at −20°C.
Establishment of rx2‐mutant alleles
The TALENs designed to bind the Rx2 coding sequence (Jean‐Paul Concordet) were provided in pCS2+ backbones with the following DNA‐binding domains: N106: TCGAGAAGTCCCACTA; N107: TCGTTGCCAGTTCCTC.
In vitro mRNA transcription:
Supercoiled DNA of the pCS2+ plasmid was subjected to restriction digest with NotI and the linearized DNA template was purified using the innuPREP PCRpure kit (Analytik Jena) according to the manufacturer's protocol. TALEN N106 and N107 mRNA in vitro transcription was performed using the mMessage machine (SP6, Ambion) according to the manufacturer's protocol. The mRNA was purified using the RNeasy RNA purification kit (Qiagen) according to the manufacturer's protocol and stored at −80°C.
mRNAs encoding TALENs directed against the rx2 coding sequence were injected at the one‐cell stage as described for medaka (Ansai et al, 2013). To identify mutations in the locus, fin clip DNA of F1 fish was analyzed. The regions of interest for the rx2 TALEN pair were amplified from the isolated genomic DNA with the following primers: rx2 forward primer: 5′‐AACAGTGAGTAGCGGGTCGT‐3′; reverse primer: 5′‐TCTGAGGGATGGAATTCTGG‐3′. The PCR amplification was performed with a proofreading polymerase: 30 s 98°C; (20 s 98°C, 45 s 67°C, 45 s 72°C) repeated for 29 cycles, 5 min followed by 72°C; and followed by 15 min at 72°C with Taq polymerase. The resulting PCR product (930 bp) was enzymatically digested with HpaII. Uncleaved fragments resulting from TALEN‐induced mutations in the HpaII recognition site were purified, cloned, and validated by sequencing. RNA null alleles were crossed to homozygosity and maintained as homozygous stock. Stocks are validated each generation by allele‐specific PCR on fin clip DNA. Wild‐type rx2 alleles were identified with the following primers: forward primer: 5′‐GGGGATTGATGGAGATGGAGT‐3′; reverse primer: 5′‐CGGCTGTAGACGTCTGGA‐3′.
Mutant rx2 alleles were identified with the following primers: forward primer: 5′‐GGGGATTGATGGAGATGGAGT‐3′; reverse primer: 5′‐CCTCCCGTCTGGATAGTGG‐3′.
Transplantations were performed as previously described (Ho & Kane, 1990; Rembold et al, 2006a). For experiments in Fig 1, homozygous Wimbledon males (Centanin et al, 2011) were crossed to heterozygous rx2::H2B‐mRFP females; 10–15 blastula cells from the progeny (100% Wimbledon+/−, 50% rx2::H2B‐mRFP+/) were transplanted to the central part of wild‐type (Cab strain) blastulae. Transplanted embryos were kept in agar‐coated dishes in 1xERM supplemented with penicillin–streptomycin (1:200, Sigma) and selected for GFP+, RFP+ cells in the retina.
For experiments shown in Fig 8, heterozygous Wimbledon males were crossed to wild‐type females; 20–30 blastula cells from the progeny were transplanted to the central part of either rx2‐mutant or a wild‐type blastulae. Transplanted embryos were raised in agar‐coated dishes in 1× ERM supplemented with penicillin–streptomycin (1:200, Sigma), selected for GFP+ clones in the retina 5 days post‐transplantation and grown to the desired age for ArCoS analysis.
Samples were imaged using an Olympus MVX10 binocular coupled to a Leica DFC500 camera, a Nikon AZ100 coupled to a Nikon C1 (entire retinae), or a Leica TCS SPE (sections). Images of NBT/BCIP stainings were taken using a Leica DM5000 scope equipped with a Leica DFC500 camera. For cell counting, an automated segmentation tool (maximum entropy threshold) for ImageJ (Version 1.41o; http://rsbweb.nih.gov/ij/) was used.
RR (in vivo, in vitro and cell culture assays, generation of rx2 mutants) and LC (lineage‐tracing analysis, transplantation assays) performed the experiments with contributions from TT (transplantation assays), DI (Rx2 antibody and GFP fusion proteins), J‐PC (design of rx2 talens) BW (rx2 mutants), and JRM‐M (rx2::Tub‐GFP transgenic line). RR, LC, TT and JW designed the experiments and analyzed the data. RR, LC and JW wrote the manuscript with contributions from DI, TT and JRM‐M.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Dataset S1
Supplementary Dataset S2
We thank N. Foulkes, S. Lemke, F. Loosli, and J.L. Mateo for critical discussion and comments on the manuscript; J.L. Mateo and M. Ramialison for help with the DNA motif analysis; M. Gebert for advice regarding the EMSA; E. Leist and A. Saraceno for fish husbandry; and T. Kellner and L. Schertel for technical assistance. DI received fellowships from the Human Frontier Science Program (HFSP) and the Japan Society for the Promotion of Science (JSPS); TT received a fellowship of the Hartmut Hoffmann Berling International Graduate School (HBIGS) in Heidelberg. The project was supported by the Collaborative Research Center SFB 873 (J.W.) of the German Research Foundation (DFG).
FundingHuman Frontier Science Program (HFSP)
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs 4.0 License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
- © 2015 The Authors. Published under the terms of the CC BY NC ND 4.0 license