Sox2, Tlx, Gli3, and Her9 converge on Rx2 to define retinal stem cells in vivo

Abstract

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.

Keywords: de‐differentiation; gene regulation; neural stem cells; retinal stem cells; transcriptional network.

Figure 1

Rx2 functions as retinal stem cell (RSC) marker

1. A rx2 mRNA is strongly detected in the peripheral CMZ of juvenile medaka (black box).

2. B Expression analysis of boxed area in (A). Transgenic reporter (rx2::Tub-GFP in green) overlaps with Rx2 immunostaining (in red).

3. C Cross-section of a transgenic rx2::H₂B-mRFP juvenile medaka. rx2⁺ cells are located in the most peripheral domain of the CMZ (bracket), limited centrally by Atoh7::mYFP. rx2⁺ cells in the central retina are Müller glia (arrowhead) and photoreceptors (arrow).

4. D, E Transplantation of rx2::H₂B-mRFP, Wimbledon cells into wild-type medaka demonstrates that rx2⁺ cells represent the most peripheral cells in an ArCoS.

Data information: Scale bars represent 25 μm in (B), 50 μm in (C), and 100 μm in (D).

Figure 2

Individual rx2-positive cells are stem cells for the neural retina (NR) or the retinal pigmented epithelium (RPE)

1. The stochastic expression of fluorescent proteins (FPs) allows single-cell labeling within the rx2⁺ expression domain.

2. ArCoS generation by individual rx2⁺ cells defines rx2 as a marker for RSCs for the NR, and the RPE. rx2 labels individual RSCs for (left) the NR and (right) the RPE.

3. All rx2⁺ cells forming ArCoSs in the NR are multipotent, generating all NR cell types.

Data information: CMZ, ciliary marginal zone; GCL, ganglion cell layer; INL, inner nuclear layer; MG, Müller glia; ONL, outer nuclear layer; PRC, photoreceptor cell. Scale bars represent 50 μm (A) and 100 μm (B, C).

Figure 3

Transcriptional regulators of rx2 are expressed in the post-embryonic CMZ

1. A A luciferase-containing vector (pGL4.1-rx2) was co-transfected with individual cDNA clones (pSport6.1-cDNA) and an internal control (pRL-CMV) to measure the transcriptional response of the rx2 pCRE in BHK cells.

2. B–E Dual luciferase assays show dose-dependent activation and repression of rx2 cis-regulatory activity in BHK cells. All values were normalized against the luminescence recorded in cells transfected with pCS2+ and reporters only. Values indicate averages of four replicates. Error bars indicate standard deviation.

3. F–Q Confocal stacks of two-color WISHs with probes against sox2 (G, red), tlx (J, red), her9 (M red), gli3 (P, red), and rx2 (H, K, N, Q, green) on medaka hatchlings. Vibratome sections show the distinct expression patterns of sox2, tlx, her9, and gli3, which partially overlap with the expression of rx2 in the peripheral CMZ. Dashed line demarcates boundary between RPE and NR. Scale bar represents 25 μm.

Figure 4

Clonal gain of gli3 or her9 restricts rx2 expression and proliferation in the CMZ

1. A Transgenic hormone-inducible lines for overexpression of gli3, her9, and H₂B-EGFP (control).

2. B Addition of RU-486 induces mosaic overexpression of transgenes in the Rx2 domain.

3. C–K The cross-sections of transgenic control embryos (rx2::LexPR LexOP::H₂B-EGFP, C–E) were compared to retinae mis-expressing gli3 (rx2::LexPR LexOP::Gli3 LexOP::H₂B-EGFP, F–H) and her9 (rx2::LexPR LexOP::Her9 LexOP::H₂B-EGFP, I–K) in the rx2 domain 9 days post-fertilization (dpf). Sustained clonal gain of Gli3 (F–H) or Her9 (I–K) inhibited rx2 expression as indicated by loss of Rx2 protein (in red). Scale bars represent 10 μm.

4. L–T Gli3 (O–Q) or Her9 (R–T) gain-of-function clones frequently lacked PCNA protein compared to control cells (L–N). Dashed lines demarcate RSC domain in the CMZ. Dotted outlines highlight affected cells. Scale bars represent 10 μm.

Figure 5

Expression of sox2 or tlx promotes stem cell features in terminally differentiated neurons

1. A Hormone-inducible expression plasmids for co-injection into transgenic rx2::Tub-GFP or wild-type embryos.

2. B Positive clones were traced by the expression of FPs (LexOP::Cherry), encoded by co-injected reporter plasmids. In combination with the ubiquitous cska promoter, this approach facilitated mosaic expression throughout all three nuclear layers in transgenic rx2::Tub-GFP embryos. Expression of the candidate factors was hormonally induced (4 dpf) when the majority of cells in the central retina (CR) had exited the cell cycle and already differentiated into the neuronal and glial cell types.

3. C, D Compared to control (C), combined expression of tlx and sox2 (D) triggered ectopic rx2 pCRE activation.

4. E–N In addition to ectopic rx2 expression, PCNA protein was detected upon sox2 or tlx expression at 7 dpf.

Data information: White arrowheads point to representative co-expressing cells. GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; ONL, outer nuclear layer; OPL, outer plexiform layer; RSCs, retinal stem cells. Scale bar represents 10 μm.

Figure 6

The rx2 pCRE contains conserved Sox- and Gli-binding sites, which interact with Sox2 and Gli3 proteins

1. A Result of BLAT search for the 2.4-kb medaka rx2 pCRE using the UCSC Genome Browser. The region, which contains the predicted Sox- and Gli-binding sites, is conserved. Blue peaks indicate conservation of coding and non-coding DNA.

2. B Sox2 and Gli3 proteins bind to predicted sites within the rx2 pCRE. The DNA–protein interaction is reduced/abolished through mutations introduced in the binding motifs.

3. C, D Reporter vectors carrying mutations (indicated by pink bars) in Sox (rx2 pCRE mtSox) or Gli (rx2 pCRE delGli) motifs in rx2 pCRE followed by firefly luciferase were co-transfected with 20 ng of sox2 or gli3 cDNA. Results of dual luciferase assays show that relative luciferase activity is reduced with rx2 pCRE mtSox compared to rx2 pCRE (C). Repression mediated by Gli3 is attenuated with rx2 pCRE delGli (D). Values were first normalized against the luminescence recorded in cells transfected with pCS2 + and reporters only (averages of four replicates). Normalized averages of modified rx2 pCREs (mtSox or delGli) were then normalized against the respective averages of wild-type rx2 pCRE.

4. E, F Fold enrichments of two independent ChIP-PCR assays for Sox2 (E) and Gli3 (F) on rx2 pCRE. Enrichments for Sox2 and Gli3 TFBSs were normalized against control luciferase fragments (summarized in Supplementary Table S1).

Figure 7

Sox- and Gli-binding sites are crucial for the rx2 cis-regulatory activity in the CMZ

1. A Transgenic lines carrying the rx2 pCRE with mutations (indicated by pink bars) in the predicted Gli- (rx2 pCRE delGli) and Sox-binding site (rx2 pCRE mtSox) followed by a nuclear FP were generated for in vivo validation.

2. B–C‴ Loss of the Gli-binding site in the rx2 pCRE generates shift of rx2 reporter expression into the RPE. Images in (B′–B‴) are magnifications of the area boxed in (B), and images in (C′–C‴) are magnifications of the area boxed in (C).

3. D–E‴ Mutation of Sox-binding site abolishes rx2 cis-regulatory activity in the CMZ. Images in (D′–D‴) are magnifications of the area boxed in (D), and images in (E′–E‴) are magnifications of the area boxed in (E).

Data information: RPE, retinal pigmented epithelium; RSCs, retinal stem cells. Scale bars represent 50 μm in (B, C, D, E) and 25 μm in (B′, C′, D′, E′).

Figure 8

Rx2 balances fate decisions of retinal stem cells

1. Adult wild-type (top) and rx2 homozygous mutant fish (bottom).

2. Transplantation of permanently labeled wild-type cells at the blastula stage into a wild-type host results in the formation of both NR (asterisk, green) and RPE (arrowhead, blue) ArCoSs.

3. Control transplantations (wild-type to wild-type).

4. Transplantation of wild-type cells into rx2-mutant host blastulae results in the preferential formation of NR ArCoSs.

Figure 9

A proposed model summarizing spatial regulation of rx2 expression and RSC-specific features through sox2, tlx, her9, and gli3

1. Schematic illustration outlining the spatial distribution of rx2 (red), sox2 (black dashed line), tlx (green), her9 (green), and gli3 (blue) transcripts in the CMZ and RPE. Purple indicates overlap between rx2 and gli3; yellow indicates overlap between rx2 and tlx/her9.

2. RSC-specific gene regulatory network operating in vivo. sox2 (black) and tlx (green) directly interact with the rx2 pCRE to activate rx2 expression in the CMZ as well as other stem cell features. Her9 (green) restricts rx2 expression centrally, and Gli3 prevents Rx2 from being expressed in the RPE. Proximity to the hedgehog morphogen (gray) secreted from the ganglion cell layer could dictate the activating function (purple) in the peripheral CMZ and repressive action (blue) in the RPE of Gli3 protein isoforms.