A regulatory loop involving PAX6, MITF, and WNT signaling controls retinal pigment epithelium development

Abstract

The separation of the optic neuroepithelium into future retina and retinal pigment epithelium (RPE) is a critical event in early eye development in vertebrates. Here we show in mice that the transcription factor PAX6, well-known for its retina-promoting activity, also plays a crucial role in early pigment epithelium development. This role is seen, however, only in a background genetically sensitized by mutations in the pigment cell transcription factor MITF. In fact, a reduction in Pax6 gene dose exacerbates the RPE-to-retina transdifferentiation seen in embryos homozygous for an Mitf null allele, and it induces such a transdifferentiation in embryos that are either heterozygous for the Mitf null allele or homozygous for an RPE-specific hypomorphic Mitf allele generated by targeted mutation. Conversely, an increase in Pax6 gene dose interferes with transdifferentiation even in homozygous Mitf null embryos. Gene expression analyses show that, together with MITF or its paralog TFEC, PAX6 suppresses the expression of Fgf15 and Dkk3. Explant culture experiments indicate that a combination of FGF and DKK3 promote retina formation by inhibiting canonical WNT signaling and stimulating the expression of retinogenic genes, including Six6 and Vsx2. Our results demonstrate that in conjunction with Mitf/Tfec Pax6 acts as an anti-retinogenic factor, whereas in conjunction with retinogenic genes it acts as a pro-retinogenic factor. The results suggest that careful manipulation of the Pax6 regulatory circuit may facilitate the generation of retinal and pigment epithelium cells from embryonic or induced pluripotent stem cells.

Figure 1. Gene dose of Pax6 regulates dorsal RPE development in an Mitf mutant–lacking MITF protein.

Cryostat sections of eyes of the indicated genotypes and developmental time points were stained with H&E (A–C,G–I,M–O) or labeled for the indicated markers (D–F,J–L,P–R). Note relatively mild RPE thickening in Mitf^mi-vga9/Mitf^mi-vga9 (C,F). In contrast, in the presence of one copy of the Pax6^Sey-Neu allele, there is a massive RPE thickening (I,L) as well as staining for the neuronal marker TUJ1 (L). In the presence of the Pax6 YAC transgene (O,R), however, no such RPE abnormalities are seen. Arrows point to the thickened RPE (C,F,H,K,I,L) or to the corresponding monolayer RPE in (M–R). Scale bar (A–C,G–I,M–O): 115 µm; (D–F, J–L): 90 µm; (P–R): 60 µm.

Figure 2. Generation and analysis of mice lacking the RPE–specific D-isoform of Mitf (Mitf^mi-ΔD/Mitf^mi-ΔD).

(A) Top: Schematic diagrams showing a region of the Mitf gene containing exons 1H, 1D, and 1B; Middle: the targeting construct with a novel BamHI restriction site and a floxed Neomycin cassette in place of 5.8 kbp of the D-Mitf promoter/exon 1D region; Bottom: Mitf gene portion after targeting. Probe ‘a’ recognizes a 5.5 kbp and probe ‘b’ a 6.5 kbp BamHI restriction fragment after targeting while both probes recognize the same 18.5 kbp fragment before targeting (see Materials and Methods for the details on construct design). (B) Southern hybridization of BamHI-restricted genomic DNA from wild-type and homozygous mutants shows the expected bands. (C) Mitf isoforms are upregulated in the RPE of Mitf^mi-ΔD/Mitf^mi-ΔD mice. Quantitative RT-PCR analysis for Mitf-isoforms from wild-type and Mitf^mi-ΔD/Mitf^mi-ΔD RPE fractions. Primers specific for Mitf isoforms A, J, H, D, and M were used to measure the respective RNAs. All values are normalized using the housekeeping gene Usf1. Results (mean values, S.D. and statistical significance [see Experimental Procedures] based on 3 biologically independent samples) are shown as fold change in RNA expression level in Mitf^mi-ΔD/Mitf^mi-ΔD compared to wild type. (D–G) Mice lacking the D-isoform of Mitf have reduced/delayed RPE pigmentation. Eye sections (D,E) from E11.5 wild type and Mitf^mi-ΔD/Mitf^mi-ΔD embryos show reduced MITF (green) and TYROSINASE (red) staining in the RPE (arrows in D,E). (F,G) Whole eye pictures show a mild reduction in pigmentation (F,G). (H,I) Adult Mitf^mi-ΔD/Mitf^mi-ΔD mice are indistinguishable on visual inspection from wild-type mice.

Figure 3. Gene dose of Pax6 regulates dorsal RPE development in a hypomorphic Mitf mutant.

Sectioning and labeling was performed as for Figure 1. In contrast to single Mitf^mi-ΔD/Mitf^mi-ΔD mutants, there is dorsal RPE thickening in Pax6^Sey-Neu/Pax6⁺; Mitf^mi-ΔD/Mitf^mi-ΔD mutants (arrows in B,E,H,K). Some cells in the thickened RPE are positive for CD138, a retinal progenitor marker (arrow in E), or TUJ1, a neuronal marker (arrow in K). Scale bar (A–C,G–I): 115 µm; (D–F,J–L): 90 µm. (M,N) Expression levels of the indicated RNAs in isolated RPE fractions based on quantitative RT-PCR (fold change relative to wild type). Results represent means and S.D. obtained from 3 biologically independent samples, each representing a pool of approximately 40 RPEs. Statistical significance of the results (see Experimental Procedures) is given for multiple pairwise comparisons.

Figure 4. Development of a differentiated laminated retina in Pax6^Sey-Neu/Pax6⁺;Mitf^mi-ΔD/Mitf ^mi-ΔD but not Pax6^YAC/YAC;Mitf ^mi-ΔD/Mitf ^mi-ΔD mice.

(A–L) Sections of eyes from P0 mice of the indicated genotypes were subjected to in situ hybridization for Crx, a photoreceptor marker (A–F) or Math3, an amacrine cell marker (G–L). Note that the ectopic staining is not present in the RPE of Pax6^YAC/YAC;Mitf ^mi-ΔD/Mitf ^mi-ΔD mutants (compare arrows in D,J with E,K for ectopic staining; arrowheads mark normal retinal staining). (M–L′) Immunofluorescent labeling for the indicated markers on P0 eye sections of the indicated genotypes. ISL1 is a ganglion cell marker (M–R), as is PAX6 at this time point (S–X, G′–L′, green). NF160 marks horizontal cells (S–X, red); VC1.1 marks amacrine cells (A′–F′, red); and SYNTAXIN marks synapses (G′–L′, red). Arrows mark the transdifferentiating portions of the RPE in Pax6^Sey-Neu/Pax6⁺;Mitf ^mi-ΔD/Mitf ^mi-ΔD mice (P,V,D′,J′) or the corresponding non-transdifferentiating portions in Pax6^YAC/YAC;Mitf ^mi-ΔD/Mitf ^mi-ΔD mice. The normal retinas continue to express each of these markers (arrowheads in the corresponding figures). Scale bar (A–L): 115 µm; (M–X, A′–L′): 90 µm.

Figure 5. Tfec is regulated by Pax6 and compensates an Mitf mutation in the RPE.

(A–D) In situ hybridization for Tfec in embryonic eyes of the indicated genotypes and developmental time points. Arrows mark areas with altered Tfec expression compared to wild type. (I) Quantitative RT-PCR analysis of Tfec mRNA levels in E11.5 RPE fractions obtained from the indicated mutants. Means and S.D. based on 3 biologically independent samples. Statistical significance shown as for Figure 3. (J) Schematic diagram of the Tyrosinase enhancer-hsp70 minimal promoter-Tfec-V5 expression cassette used for generating transgenic mice. (K) Flat mount of adult RPE from Tyr-Tfec transgenic line stained with Phalloidin (green) and anti-V5 antibody (red). (L,M) Transgenic TFEC rescues eye defects in Mitf^mi-rw/Mitf^mi-rw mice. The eyes of Tg (Tyr-Tfec);Mitf^mi-rw/Mitf^mi-rw mice (n = 18) are bigger than those of Mitf^mi-rw/Mitf^mi-rw mice (n = 16). (N–Q) Transgenic TFEC expression suppresses RPE-retina transdifferentiation in Mitf^mi-rw/Mitf^mi-rw mice. P0 mouse eye sections stained as shown, with nuclei stained with Topo3. (N,O) TFEC-V5 staining is seen only in Tg (Tyr-Tfec); Mitf^mi-rw/Mitf^mi-rw mice (arrows) and MITF staining is below threshold in this area of the RPE. (P,Q) Note absence of ectopic PAX6 and TUJ1 expression in the RPE of Tg (Tyr-Tfec); Mitf^mi-rw/Mitf^mi-rw mice. Scale bar (A–H): 110 µm; (N–Q): 90 µm. (R) Schematic diagram of the Tfec enhancer/promoter region. The positions of conserved potential binding sites for MITF (▪) and PAX6 (•) are given relative to the translation start site of Tfec isoform A . (S) ChIP assays of wild-type RPE fractions dissected from E11.5 embryonic eyes, using PAX6 and MITF-specific antibodies. (T) A 700 bp Tfec enhancer/promoter region containing amplicons I and II (see R) was used for reporter assays in ARPE19 cells co-transfected with expression plasmids for the indicated transcription factors. Results represent normalized mean luciferase activity units obtained from 9 independent transfections. S.D. and statistical significance are indicated.

Figure 6. PAX6 and MITF/TFEC together suppress Fgf15 and Dkk3 in the developing RPE.

(A,B) Quantitative RT-PCR of Fgf15 and Dkk3 RNA of E11.5 RPE fractions of wild type and indicated mutants. Mean, S.D. and statistical significance based on 3 biological replicates. (C–N) In situ hybridization for Fgf15 (C–H) and Dkk3 (I–N) on eye sections of E11.5 embryos of the indicated genotypes. Arrows in E–G,K–M point to the RPE (note ectopic Fgf15 expression in the RPE of Pax6^Sey-Neu/Mitf^mi-vga9 double mutant in F and ectopic Dkk3 expression in the RPE of Pax6^Sey-Neu/Mitf^mi-vga9 double mutant in L). Scale bar (C–N):115 µm. (O) Schematic representation of Fgf15 and Dkk3 enhancer/promoter regions showing amplicons containing conserved potential binding sites for MITF (▪) and PAX6 (consensus , non-consensus ), with positions indicated relative to translation start sites. (P) ChIP assays performed with indicated antibodies on RPE fractions dissected from E11.5 wild-type embryonic eyes. Amplicons as indicated in (O). (Q) ChIP assays performed with the indicated antibodies on optic vesicle tissue dissected from E10.5 embryonic eyes of the indicated genotypes. Anti-acetyl H3 signal represents active chromatin domains and anti-dimethyl H3K9 signal inactive chromatin domains. For details see text.

Figure 7. FGF and DKK3 induce RPE transdifferentiation.

(A–P) Cultures of developing eyes explanted from E10.0 wild-type embryos. (A–D) RPE pigmentation develops within 48 hours in the absence of a bead (A) or in the presence of beads soaked in 0.65 µg/ml of recombinant DKK3 (B) or 0.35 µg/ml of recombinant FGF2 (C) but is segmentally missing in the vicinity of a bead soaked in a combination of 0.65 µg/ml of DKK3 and 0.35 µg/ml of FGF2 (D). The number of cultures with the represented results per total cultures tested is shown in the upper right corner. (E–L) Representative cultures were fixed, cryosectioned and subjected to in situ hybridization with the indicated probes. Note induction of the retinal factors Vsx2 and Six6 in the RPEs only after implantation of a DKK3/FGF2 double-coated bead (H,L). (M–P) Cultures were established from TCF-LacZ transgenic embryos and implanted with beads coated with bovine serum albumin (BSA, M) or the indicated growth factors. They were fixed, cryosectioned, and stained with antibodies to ßGAL. Note absence of ßGAL staining only in cultures implanted with double-coated beads (inset shows higher magnification of the RPE region) (P). The number of cultures with the shown results per total cultures established is shown in the right upper corner. (Q–R) VSX2/ßGAL double-labeled cryosections of E10.5 embryos of the indicated genotypes. Note absence of ßGAL labeling in the transdifferentiating portions of the RPEs in Mitf^mi-vga9 heterozygous or homozygous embryos when they carry a Pax6^Sey-Neu allele (S,T, arrow in T). Numbers represent % ßGAL positive cells in the dorsal RPE. P values based on Student's t test. Scale bar (A–D, M–P, Q–T): 90 µm; (E–L): 115 µm.

Figure 8. Model of the regulatory circuit involving Pax6, Mitf, and Tcfec during mouse RPE development.