Abstract
The bHLH transcription factor Hes1 is a key downstream effector for the Notch signaling pathway. During embryogenesis neural progenitors express low levels of Hes1 in an oscillating pattern, whereas glial brain boundary regions (e.g., isthmus) have high, sustained Hes1 levels that suppress neuronal fates. Here, we show that in the embryonic mouse retina, the optic nerve head and stalk express high Hes1, with the ONH constituting a boundary between the neural retina and glial cells that ultimately line the optic stalk. Using two Cre drivers with distinct spatiotemporal expression we conditionally inactivated Hes1, to delineate the requirements for this transcriptional repressor during retinal neurogenesis versus patterning of the optic cup and stalk. Throughout retinal neurogenesis, Hes1 maintains proliferation and blocks retinal ganglion cell formation, but surprisingly we found it also promotes cone photoreceptor genesis. In the postnatal eye, Hes1 inactivation with Rax-Cre resulted in increased bipolar neurons and a mispositioning of Müller glia. Our results indicate that Notch pathway regulation of cone genesis is more complex than previously assumed, and reveal a novel role for Hes1 in maintaining the optic cup–stalk boundary.
SIGNIFICANCE STATEMENT The bHLH repressor Hes1 regulates the timing of neurogenesis, rate of progenitor cell division, gliogenesis, and maintains tissue compartment boundaries. This study expands current eye development models by showing Notch-independent roles for Hes1 in the developing optic nerve head (ONH). Defects in ONH formation result in optic nerve coloboma; our work now inserts Hes1 into the genetic hierarchy regulating optic fissure closure. Given that Hes1 acts analogously in the ONH as the brain isthmus, it prompts future investigation of the ONH as a signaling factor center, or local organizer. Embryonic development of the ONH region has been poorly studied, which is surprising given it is where the pan-ocular disease glaucoma is widely believed to inflict damage on RGC axons.
Introduction
The mammalian retina contains six neuronal and one glial cell class. They are generated in a conserved, overlapping birth order, with retinal ganglion cells (RGCs), amacrines, horizontals and cone photoreceptors specified first, while rods, bipolars and Müller glia generated mostly after birth. This diversity arises from a pool of multipotent progenitor cells (RPCs) with dynamic transcriptional profiles, during retinal development (Livesey and Cepko, 2001; Clark et al., 2019). The interplay among signaling pathways and transcription factor hierarchies produces a laminated retina that receives, interprets and relays visual input to the brain.
The Notch signaling pathway is highly conserved, and regulates both the timing of neurogenesis and progenitor cell population dynamics during mouse retinal development. At the molecular level, Notch ligand-receptor binding induces sequential proteolytic cleavages of the receptor, releasing the intracellular domain (N-ICD), which forms a nuclear protein complex with Rbpj, MAML, p300 and other proteins. This complex transcriptionally activates downstream genes, with Hes1 a frequent target (for review, see Kovall et al., 2017). Numerous studies demonstrated that too little or too much Notch signaling profoundly disrupts retinal development. For example, mouse Notch1 or Rbpj conditionally mutant eyes prematurely produce RGCs and cones, progressively lose RPCs, and display retinal rosettes composed of excess photoreceptors (Jadhav et al., 2006a,b; Yaron et al., 2006; Riesenberg et al., 2009; Zheng et al., 2009; Riesenberg and Brown, 2016).
Vertebrate HES/Hes genes are orthologues of Drosophila hairy and the Enhancer of Split complex, and encode basic helix-loop-helix (bHLH) transcriptional repressors (for review, see Kageyama et al., 2008; Harima et al., 2014). Hes1 is arguably the best understood, due to its involvement in a wide array of developmental processes, and association of human HES1 mutations with cancer (for review, see Liu et al., 2015; Dhanesh et al., 2016). During eye development, all cells initially express Hes1, regardless of origin from the neural ectoderm (optic cup, stalk and RPE), surface ectoderm (lens, cornea) or surrounding mesenchyme (muscles and embryonic vasculature). During embryonic growth, Hes1 mRNA and protein oscillate, controlled cyclically by Hes1 autorepression (Hirata et al., 2002; Jensen et al., 2003), to maintain CNS progenitor proliferation. Hes1 is downregulated at cell cycle exit and differentiation. Importantly, Hes1 also exhibits high, sustained expression in domains that are physical boundaries between CNS progenitor cell compartments (for review, see Kageyama et al., 2007; Imayoshi et al., 2013; Imayoshi and Kageyama, 2014). One region is the brain isthmus, where cells expressing high Hes1 proliferate at a slower rate, and differentiate into glia (Baek et al., 2006). In the embryonic eye, at the juncture of the optic cup and optic stalk, is a ring of cells termed the optic nerve head (ONH), which is a boundary between neural retina and optic stalk glia (Tao and Zhang, 2014). We found that ONH cells express Hes1 at high, sustained levels, whereas adjacent RPCs exhibit the “salt and pepper” pattern characteristic of oscillatory behavior.
Previous explorations of Hes1 used a germline mutant mouse, ex vivo retinal culture and viral misexpression (Tomita et al., 1996; Furukawa et al., 2000; Takatsuka et al., 2004; Lee et al., 2005). These studies highlighted Notch-dependent Hes1 activities, including regulating the ratio of proliferating to differentiating RGCs, the embryonic apical outer limiting membrane and retinal lamination. Without Hes1, apical N-Cadherin in RPCs is discontinuous, and RGCs occupy the subretinal space, leading to cellular rosettes, a manifestation of abnormal lamination (Takatsuka et al., 2004). Hes1 is not solely regulated by the Notch pathway, since it also depends on Hedgehog (Hh) signaling (Solecki et al., 2001; Wall et al., 2009). It remains unknown how Hes1 has two simultaneous expression modes, and how it responds to particular signaling inputs. To unravel these questions, we removed Hes1 during retinal development, using two Cre transgenic mouse lines, Chx10-Cre and Rax-Cre, and compared these mutant phenotypes across eye development. The Chx10-Cre driver deletes genes from E10.5 exclusively within RPCs. Although this driver has mosaic activity, it is relatively low before E13.5. Rax-Cre initiates recombination at least one day earlier, without mosaicism, and acts in the ventral thalamus/hypothalamus, optic vesicle, cup and stalk, RPE, ONH and RPCs. (Klimova et al., 2013; Klimova and Kozmik, 2014; Pak et al., 2014).
The onset of neurogenic phenotypes in both conditional mutants phenocopy each other, and germline mutants (Tomita et al., 1996; Takatsuka et al., 2004; Lee et al., 2005). We discovered that both conditional mutants produce significantly fewer cone photoreceptors that persists into adulthood. This is in stark contrast to an overproduction of cones in Notch1, Rbpj, Dll4 or Dll1;Dll4 retinal mutants, suggesting Hes1 does not directly repress cone-specific gene expression, but acts in a more complex manner. Both adult Hes1 Cre mutants have excess bipolar neurons; but only Rax-Cre;Hes1 mutants have abnormally positioned Müller glial cells. Only Rax-Cre mutants revealed a novel role for Hes1 in maintaining the optic cup–stalk boundary, where Pax6+ neural retina expanded at the expense of Pax2+ optic stalk, resulting in a failure to close the optic fissure. Our data significantly extend the known roles for Hes1 during vertebrate eye development, and implicate human HES1 mutations as candidates for optic nerve coloboma.
Materials and Methods
Animals.
Mouse strains used in this study were a Hes1CKO allele (Hes1tm1Kag) maintained on both CD-1 and C57BL/6J backgrounds (Imayoshi et al., 2008); Hes1KO/+ (Hes1tm1Fgu) or Hes1KO/+;Atoh7 Tg GFP/+ mice (Ishibashi et al., 1995; Hutcheson et al., 2005) maintained on a CD-1 background; Chx10-Cre BAC transgenic line [Tg(Chx10-EGFP/cre;-ALPP)2Clc; JAX stock number 005105] maintained on a CD-1 background (Rowan and Cepko, 2004); Ai9 Cre lineage tracer [(Gt(ROSA)26Sortm9(CAG-tdTomato)Hze] maintained on a C57BL/6 background (Madisen et al., 2010); and Rax-Cre BAC transgenic line Tg(Rax-cre) NL44Gsat/Mmucd created by the GENSAT project, cryobanked at MMRRC U.C. Davis (stock no. 034748-UCD) and maintained on a CD-1 background. PCR genotyping was performed as described previously (Ishibashi et al., 1995; Rowan and Cepko, 2004; Hutcheson et al., 2005; Imayoshi et al., 2008; Madisen et al., 2010). Conditional mutant breeding schemes mated one Cre Tg/+ mouse to another homozygous for the Hes1 conditional allele to create Cre;Hes1CKO/+ mice, used in timed matings to Hes1CKO/CKO mice. The date of a vaginal plug was assigned the age of E0.5. All mice were housed and cared for in accordance with guidelines provided by the National Institutes of Health, Bethesda, MD, and the Association for Research in Vision and Ophthalmology, and conducted with approval and oversight from the U.C. Davis Institutional Animal Care and Use Committees.
Histology and immunohistochemistry.
For histology, P0 and P21 eyes were dissected and fixed in 4% paraformaldehyde overnight at 4°C and underwent standard processing for paraffin-embedding, then 4 μm sections were generated and H&E stained. For immunohistochemistry, embryonic tissue was fixed in 4% paraformaldehyde/PBS for 1 h on ice, processed by stepwise sucrose/PBS incubation ranging from 5 to 15%, and embedded in Tissue-Tek OCT, then 10 μm frozen sections were generated and antibody labeled as in (Mastick and Andrews, 2001). Details about the primary and secondary antibodies used are provided in Tables 1 and 2. Nuclei were counterstained with DAPI.
Authentication of primary antibodies used in this study
Secondary antibodies used in this study
EdU incorporation.
EdU pulse labeling was performed by injecting EdU in pregnant dams carrying E13.5 litters (12 μg/g body weight of 1 mg/ml EdU in 0.15 m NaCl) and harvesting embryos after 1.5 h. EdU was detected using Click-iT EdU Cell Proliferation Kit for Imaging (Thermo Fisher Scientific, catalog #C10339).
Microscopy and statistical analysis.
Digital TIF images of living embryos with tdTomato expression were captured on a Leica MZ12 dissecting microscope, equipped with a UV light source, Spot B&W and color cameras and Spot software (v5.2). TIF images of adult mouse heads were imaged using a Leica MZ8 dissecting microscope, DFC290 HD camera and Leica LAS V4.2 software. Histologic sections were imaged with a Zeiss Axio Imager M.2 microscope, color camera and Zen software (v2.6). Antibody labeled cryosections were imaged using a Leica DM5500 microscope, equipped with a SPEII solid state confocal and processed using Leica LASX (v.3), ImageJ software, and Adobe Photoshop (CS5) software programs. All images were equivalently adjusted for brightness, contrast, and pseudo-coloring. For embryonic sections, ≥3 individuals were analyzed, using at least 2 sections per individual. For cell counting, three or more animals were analyzed per genotype and age, with ≥2 sections per control or mutant littermate. Sections were judged to be of equivalent depth by presence of or proximity to the optic nerve. Marker+ cells in tissue sections were counted using the count tool in Adobe Photoshop CS5 and statistical analyses performed using Prism software (GraphPad v7) and Excel (v16.16.11), with p-values determined with one-way ANOVA with Dunnett's test.
Results
At the end of gastrulation, the eye field is specified, bisected laterally and each optic vesicle evaginates from the ventral diencephalon. Multiple signaling pathways regionalize the growing optic vesicles, delineating the neural retina (NR), retinal pigment epithelium (RPE) and optic stalk (OS). This process also produces a unique cuff of cells, called the optic nerve head (ONH), which serves as a physical and molecular boundary between the neural retina and glial optic stalk (for review, see Tao and Zhang, 2014). Retinal progenitor cells (RPCs) exhibit nonuniform Hes1 expression, consistent with its established role in promoting growth and suppressing neurogenesis. Using a validated antibody (Lee et al., 2005), we found that beginning around E12.5, ONH cells express high levels of Hes1 protein, compared with the adjacent retina, RPE and lens epithelial domains. These expression differences are easily seen at E13.5 and E16.5 (Fig. 1A,B). In cross section, the ONH is triangular, defined morphologically by the opening that allows RGC axons to exit into the optic stalk lumen, but molecularly where it is contiguous with the NR (Fig. 1B). The high Hes1 domain corresponds to that of Pax2, previously shown to be essential for ONH and OS development (see Figs. 1A, F, 3A,D) (Torres et al., 1996; Otteson et al., 1998; Schwarz et al., 2000).
Hes1 protein and Cre driver spatiotemporal expression domains. A, Anti-Hes1 labeling delineates low, nonuniform retinal and high, sustained ONH Hes1 expression. B, Hes1 and Tubb3 colabeling clearly shows high Hes1 in ONH and OS cell nuclei, adjacent to RGC axons of the optic nerve. C, D, G, H, Lateral view of live mouse embryos showing Cre-activated Ai9 reporter activity. Rax-Cre is expressed earlier than Chx10-Cre (C, G, arrows). Both drivers express robustly in E11.5 optic cup (arrows in D, H), with Rax-Cre also seen in the developing forebrain. E, F, I, J, E13.5 horizontal sections containing the ONH labeled with anti-DsRed to better show spatial differences in the two Cre transgenes. Chx10-Cre mediated activity is restricted to the neural retina, but the long half-life of tdTomato expression also lineage traces nascent RGC axons (E). Antibody colabeling for Cre-GFP fusion and Pax2 proteins show absent Chx10-Cre expression in the OS or ONH (arrows in F–F′′). I, Rax-Cre activity in the neural retina, RPE, ONH, OS and lineage-traced RGCs axons. J, Comparison of tdTomato (anti-DsRed) and Pax2 highlights Rax-Cre activity within ONH cells. L, lens; NR, neural retina; ONH, optic nerve head; OS, optic stalk; RPE, retinal pigmented epithelium Scale bars: 50 μm in A, B, E, and F; 500 μm in C and D.
Chx10-Cre and Rax-Cre driver activities overlap in RPCs but not in the ONH
To tease apart Hes1 functions in the developing retina versus ONH/optic stalk, we used two conditional deletion strategies. First, a Chx10-Cre BAC transgenic driver was used for consistency with previous retinal analyses of Notch1 and Rbpj (Jadhav et al., 2006a,b; Riesenberg et al., 2009). Because Chx10/Vsx2 expression is quickly confined to RPCs once the optic cup forms, we also searched for another Cre transgenic line that would remove Hes1 at earlier stages, and within the forming ONH/OS. The Rax/Rx transcription factor is one of the earliest genes to demarcate the eye field (Furukawa et al., 1997; Mathers et al., 1997). Previously the GENSAT/MMRRC project created a Rax-Cre BAC transgenic mouse by recombineering a Cre expression cassette into the identical BAC clone used by Sasai and colleagues for a widely used Rax-GFP transgenic mouse cell line (Wataya et al., 2008). So, we acquired this MMRRC Rax-Cre line, and compared its activity to Chx10-Cre (Fig. 1). Although an ONH-specific Cre driver would be ideal for our experiments, an exhaustive search identified only one possible tool, a BAC transgenic Pax2-Cre mouse containing 120kb of the mouse Pax2 locus (Ohyama and Groves, 2004) that unfortunately lacks embryonic eye enhancer(s).
First, we compared the spatiotemporal expression of a Cre-induced tdTomato reporter (Ai9-Rosa26-CAG-loxP-STOP-loxP-tdTomato) in timed matings with each Cre driver (Fig. 1C–J). For Chx10-Cre, we did not detect any tdTomato expression in the E9.5 optic vesicles (Fig. 1C, arrow), but readily saw expression from E10.5 onwards (Fig. 1D, arrow), in agreement with previous studies (Rowan and Cepko, 2004). At E13.5, anti-DsRed labeling of horizontal sections highlighted tdTomato expression within the NR, and lineage-traced nascent RGC axons within the OS (Fig. 1E,F). To confirm retinal-specificity, we colabeled E13.5 Chx10-Cre cryosections with anti-GFP and anti-Pax2, further demonstrating that the Cre-GFP fusion protein is almost entirely absent from Pax2+ ONH and OS cells (Fig. 1F–F′′).
By contrast, Rax-Cre;Ai9 embryos have tdTomato expression in the E9.5 optic vesicle and developing forebrain (Fig. 1G) that is highly visible at E11.5 (Fig. 1H, arrow). The reporter expression within the telencephalon and diencephalon (data not shown), reflects endogenous Rax expression domains (Furukawa et al., 1997; Mathers et al., 1997; Asbreuk et al., 2002). At E13.5 anti-DsRed labeling demonstrated the expected NR, RPE, ONH and OS domains, since these tissues derive from the optic vesicle (Fig. 1I). In E13.5 DsRed/Pax2 colabeled sections all Pax2+ ONH cells are present in the Rax-Cre lineage (Fig. 1J–J′′). Where Chx10-Cre and Rax-Cre activities overlap in the NR, nonuniform Hes1 activity will be removed from RPCs, thereby inducing neurogenic phenotypes akin to those previously described for the Notch pathway (Jadhav et al., 2006a,b; Riesenberg et al., 2009). However, Rax-Cre is predicted to also remove high, sustained Hes1 expression, allowing us to test the requirements for Hes1 in the ONH and/or OS.
Hes1 conditional mutants have retinal lamination defects and persistent coloboma
We, and others, previously showed that Hes1 germline mutant embryos are microphthalmic and die before birth, from multiple organ system defects (Tomita et al., 1996; Lee et al., 2005; Kameda et al., 2011). To demonstrate that these conditional mutants are adult viable, we bred mice homozygous for the Hes1 floxed allele (Hes1CKO/CKO) to Chx10-Cre;Hes1CKO/+ or Rax-Cre;Hes1CKO/+ mice (Fig. 2). Conditionally mutant adults were produced at an expected 25% Mendelian ratio, with completely penetrant ocular deformities (Fig. 2) (n = 7 mutants Chx10-Cre; 3 mutants Rax-Cre). Throughout this study, controls are Hes1CKO/CKO littermates, and Cre; Hes1 heterozygotes (Hes1CKO/+) displayed no obvious phenotypes.
Hes1 conditionally mutant ocular defects. A, E, I, Lateral views of E13.5 live embryos. Chx10-Cre;Hes1CKO/CKO eyes (E) are indistinguishable from control littermates (A). However Rax-Cre;Hes1CKO/CKO eyes have fully penetrant colobomas (arrow in I). B, F, J, Anti-Hes1 labeling of E13.5 cryosections indicates the identical loss of low, nonuniform expression from the retina of both Cre-mutants. However, Chx10-Cre;Hes1CKO/CKO eyes retain high, sustained Hes1 expression in the ONH (arrows in each panel). Rax-Cre;Hes1CKO/CKO uniquely exhibited an abnormal elongated shape to the retina (J). C, G, K, Gross examination of P21 eyes suggests microphthalmia in Chx10-Cre;Hes1CKO/CKO (G) and Rax-Cre;Hes1CKO/CKO (K) mutants. D, H, L, However, H&E-stained P21 sections at the level of the optic nerve support smaller eyes only for Rax-Cre;Hes1CKO/CKO. Both mutants have lamination defects and retinal rosettes (H), along with severe thinning of retinal layers (L). D′, H′, and L′ are higher magnifications of boxed areas in D, H, and L, respectively. L, lens; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bars: 50 μm in B, D, and D′ and 500 μm in A and C. n = 3 biologic replicates per age and genotype were evaluated.
Then, we analyzed the morphology of control and mutant eyes during embryogenesis (Fig. 2A,E,I). E13.5 Chx10-Cre;Hes1CKO/CKO eyes showed no apparent differences compared with control retinas (Fig. 2E). However, E13.5 Rax-Cre;Hes1CKO/CKO eyes exhibited coloboma, a failure of optic fissure closure (Fig. 2I, arrow). The optic fissure is a transient structure that normally closes by E12.5. However, at all ages examined, Rax-Cre;Hes1CKO/CKO animals had incomplete fissure closure (100%, n > 22), thus we conclude this defect is not simply a developmental delay. Next, we confirmed the loss of Hes1 nuclear protein by immunohistochemistry (Fig. 2B,F,J). In Chx10-Cre;Hes1CKO/CKO retinas, low, nonuniform expression in RPCs was absent, although high, sustained expression persisted in the adjacent ONH and OS (Fig. 2F, arrow). In Rax-Cre;Hes1CKO/CKO retinas nearly all expression domains were missing (Fig. 2J), and the optic cup had a fully penetrant, elongated shape.
Next we surveyed the adult eyes from Hes1 conditional mutants. Both Cre mutants grossly appear to have bilateral microphthalmia (Fig. 2G,K), however smaller eye size was only evident in P21 Rax-Cre;Hes1CKO/CKO ocular sections (Fig. 2L). Both Chx10-Cre; Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO retinas had lamination defects that included retinal rosettes (Fig. 2D,D′,H,H′,L,L′). Our data corroborate previous descriptions of Hes1 germline mutants, and allow us examination of Hes1 functions at every stage of retinogenesis. Importantly, Rax-Cre;Hes1CKO/CKO phenotypes most closely matched those of Hes1 germline mutants, and uncovered two significant defects: persistent coloboma and early changes in optic cup shape.
Hes1 maintenance of the optic cup–stalk boundary
The cells situated between the optic cup and stalk, represent a neural-glial cell tissue boundary that is defined molecularly by the mutually suppressive activities of Pax2 and Pax6 (Schwarz et al., 2000). Although these factors are initially coexpressed in the E9.5 optic vesicle, they are simultaneously downregulated in adjacent tissues: Pax2 in optic cup, and Pax6 in the OS. This establishes a boundary via mutually exclusive expression domains. The clearest demonstration that the ONH is a boundary came from Pax2 germline mutants, in which Pax6 NR expression expands proximally, along the optic stalk (Torres et al., 1996; Schwarz et al., 2000). At E11.5, we saw that the Pax6 and Pax2 domains were actively separating, but were still coexpressed in the forming ONH (Fig. 3A, arrows). E11.5 Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO optic cups showed essentially normal Pax6 and Pax2 domains, although the area of coexpression appeared broader in some Rax-Cre mutants (Fig. 3B,C, arrows). At E13.5, in both wild-type and Chx10-Cre;Hes1CKO/CKO optic cups there was complete separation of Pax6 NR and Pax2 ONH/OS domains (Fig. 3D,E). However, in Rax-Cre;Hes1CKO/CKO eyes the Pax6 NR domain was clearly expanded, but the Pax2+ ONH cells were also misplaced proximally (Fig. 3F, arrow). This defect is identical to Hes1 germline mutants (Fig. 3M–P), where all Hes1 expression is lacking. Because Pax6 and Pax2 domains can separate without Hes1, we conclude that its role is to maintain the ONH boundary.
Specific alteration of the optic cup–stalk boundary in Rax-Cre;Hes1CKO/CKO and Hes1−/− embryos. Immunolabeling of E11.5 and E13.5 horizontal sections containing retina, ONH and optic stalk. A–F, Pax6-Pax2 colabeling. These transcription factors are initially coexpressed, but by E13.5, abut each another in nonoverlapping domains, thereby highlighting the NR/OC-ONH boundary (arrow in D). Although the abutting expression domains arise in both Chx10-Cre;Hes1CKO/CKO (B, E) and Rax-Cre;Hes1CKO/CKO (C, F) eyes, the position of the boundary is more proximal and highlights an expansion of the retinal territory (F). This defect is phenocopied in Hes1 germline mutants (N, P). G–L, Rax-Mitf colabeling for the retinal-RPE boundary in nearby sections. In both conditional mutants, these tissues are essentially normal, although Rax-Cre;Hes1CKO/CKO eyes always displayed complete suppression of Mitf in the E11.5 ONH/OS (arrows in G–I and insets of Mitf expression). At E13.5, proximal expansion of RPE matched that of the neural retina. M, O, Normal Pax6, Pax2 (arrow in O) and Atoh7-GFP transgene expression domains. N, P, In Hes1 germline mutants, inappropriate proximal expansion of Pax6+ optic cup cells, accompanied by a proximal shift of the Pax2 ONH domain, which is also misshapen (arrow in P). Atoh7-GFP expression highlights a previously described RGC neurogenic phenotype. L, lens. Scale bars in A, M, O, 50 μm. n > 3 biologic replicates per age and genotype were analyzed.
Next, we investigated whether loss of Hes1 also affected NR and RPE separation, particularly at the ciliary margin, by colabeling with specific antibodies for Rax (NR) and Mitf (RPE). At both ages, control and Chx10-Cre;Hes1CKO/CKO optic cups had indistinguishable expression of these markers (Fig. 3G,H,J,K). Although at E11.5 Mitf is normally present in both the RPE and OS, later it becomes restricted to the RPE (Fig. 3G,H, arrows and insets). In contrast, we never saw Mitf expression within the E11.5 Rax-Cre;Hes1CKO/CKO OS (Fig. 3I, inset), suggesting these cells may be changing their fate to retinal. At E13.5, Rax-Cre;Hes1CKO/CKO eyes displayed an inappropriate expansion of Rax and Mitf into the OS (cf. Fig. 3L and J,K), as shown for Pax6 (Fig. 3F). We conclude that the NR, RPE and ciliary margin territory formation is unaffected in both conditional mutants. But, in Rax-Cre;Hes1CKO/CKO eyes, lacking both low, nonuniform and high, sustained Hes1 expression, a profound requirement for Hes1 in ONH cells to sustain this boundary was revealed.
Hes1 in embryonic retinal progenitor cells
Multiple analyses of Hes1 germline mutants describe precocious and rapid neurogenesis of Tubb3+ neurons that impact the remaining progenitor cell pool (Ishibashi et al., 1995; Nakamura et al., 2000; Ohtsuka et al., 2001; Takatsuka et al., 2004). These mutants also display patchy deformity to the outer limiting membrane, and contain excess retinal neurons in the subretinal space (Takatsuka et al., 2004). Although the excess Tubb3+ neurons are highly likely to be RGCs, this was not formally shown. First, we labeled E11.5 and E13.5 control, Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO sections with Tubb3 to verify that both conditional mutants contained ectopic Tubb3+ optic cup cells (Fig. 4A–F, arrows). At the older age, ectopic Tubb3 expression was quite expanded apically, which was more prominent in Rax-Cre;Hes1CKO/CKO eyes. Some ectopic Tubb3+ neurons and their processes could be seen in the subretinal space (Fig. 4E,F, and see Fig. 5, arrows). The proneural bHLH factor Atoh7 normally initiates expression in the E11.5 central optic cup (Brown et al., 1998) (Miesfeld et al., 2018). Previously we reported precocious Atoh7 mRNA expression at E9.5, but without an increase in expression levels in Hes1 germline mutants (Lee et al., 2005). In the central optic cup of both conditional mutants, we observed ectopic Atoh7+ cells (Fig. 4H,I, arrows), whereas control littermates had not yet initiated expression (Fig. 4G).
Hes1 conditional mutants have expected neurogenic and RPC pool depletion phenotypes. A–F, TUBB3 labeling of nascent retinal neurons at E11.5 and E13.5. Premature neurogenesis is in progress at E11, and by E13.5, there is a dramatic increase of Tubb3+ cells in both Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO eyes. Some ectopic neurons are improperly positioned in the outer retina and/or have aberrant trajectories (arrows in E and F). G–I, Precocious Atoh7 expression at E11 (arrows in H, I), also phenocopies Hes1 germline mutants (Takatsuka et al., 2004; Lee et al., 2005). J–L, P, E13.5 sections containing the ON from embryos pulse-labeled with EdU 90 min before dissection. M–O, Q, Anti-cPARP labeling of nearby sections to assess apoptosis. P, Statistically significant loss of EdU+ RPCs in both mutants. Q, cPARP+ cells (arrows in N, O) were increased, with statistical significance for Rax-Cre;Hes1CKO/CKO eyes. Both graphs display individual data points, the mean and SEM; n.s., not significant; *p < 0.05; **p < 0.01. L, lens. Scale bars in A and M, 50 μm. n = 3 biologic replicates (2 or more sections/replicate) per age and genotype were quantified.
RGC patterning defects in Hes1 retinal mutants. A–C, Colabeling for Tubb3 and Mitf highlights patches of abnormal neuronal processes inappropriately on the apical side of the retina (inset boxes). D–F, Pou4f nuclear expression in differentiated RGCs better highlights RGC cell body mispatterning in the patches (arrows in E, F). Previous examination of Hes1 germline mutants showed an early loss of the OLM, a boundary that also confers apical-basal positional information to RPCs (Takatsuka et al., 2004). G–I, P0 H&E sections with retinal rosettes and breaks in the outer retina (arrows in H, I). L, lens. Scale bars: A and D, 50 μm with insets 400× magnified; G, 500 μm. n = 3 biologic replicates per age and genotype analyzed.
Hes1 has well described functions in influencing cell cycle dynamics (Uriu, 2016). All Hes1 mutants contain excess neurons, which negatively feedbacks on growth. Therefore, we quantified the pool of proliferating RPCs by EdU pulse-labeling timed pregnant mice. We observed significant reduction in the number of EdU+ RPCs in both Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO retinas compared with control littermates (Fig. 4J–L,P). To determine which defects contribute to microphthalmia in Hes1 mutants, we also quantified apoptotic cells at E16.5 by labeling with anti-cPARP (Fig. 4M–O,Q). There was an increase in the number of cPARP+ cells in both Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO retinas, compared with controls (Fig. 4Q). Overall we conclude that both conditional mutants recapitulate Hes1 germline mutant phenotypes regarding loss of proliferating RPCs and excess neurogenesis. However, unlike the previous studies, here, we observed a significant increase in apoptosis, consistent with conditional deletion of Notch1, Rbpj and Dll4 (Yaron et al., 2006; Riesenberg et al., 2009; Zheng et al., 2009; Luo et al., 2012).
To explore the abnormal retinal neuron patterning in Hes1 conditional mutants further, we labeled E16.5 sections with Tubb3, and Mitf, or the nuclear RGC marker Pou4f (Fig. 5). At E16.5, there were many Tubb3+ neurons inappropriately positioned apically, or in the subretinal space, between the retina and RPE cells expressing Mitf (Fig. 5A–C, inset boxes). This phenotype was patchy, with no particular directionality in the nasal-temporal, proximal-distal or dorsal-ventral dimensions. We attribute a missing outer limiting membrane (OLM) previously reported in E14-E16 Hes1 germline mutants, for allowing access to the subretinal space (Takatsuka et al., 2004). To definitively identify the ectopic neurons as RGCs, we labeled E16.5 sections with a pan-Pou4f antibody and saw groups of RGC nuclei mispositioned apically (arrow in 5E,F), and in the subretinal space, in both conditional mutants. At P0, H&E staining highlighted the persistence of outer retinal “gaps” (arrows in Fig. 5H,I). Further analysis at this age was challenging, since lamination defects became difficult to assign as a missing OLM versus formation of a retinal rosette.
Loss of Hes1 causes a reduction in cone photoreceptor genesis
Previous work has consistently shown that blocking Notch signaling increases both RGC and cone photoreceptor differentiation (Jadhav et al., 2006b; Yaron et al., 2006; Riesenberg et al., 2009). However, E13.5 Hes1 germline mutant retinas contained fewer Crx+ cells (Riesenberg et al., 2009). To address this paradox, we quantified developing cone photoreceptors at three ages: E13.5, E16.5, and P21. First, the Otx2 transcription factor is one of the earliest markers for RPC commitment as a tripotential cell (cone, rod or bipolar) (Nishida et al., 2003; Fossat et al., 2007). Although Otx2 is also present in adjacent RPE cells, they are easily identifiable by tissue position and shape. So, we quantified Otx2+ optic cup cells in control, Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO eyes (Fig. 6A–C,J). We found a significant reduction in the percentage of Otx2+ cells in both conditional mutants (Fig. 6J). To specifically examine cone genesis however, we quantified Thrb2+ cells at E16.5, which is the peak of cone neurogenesis (Fig. 6D–F,K). Here, too, we observed a significant reduction in Thrb2+ cones in both Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO retinas (Fig. 6K). To address the possibility that early loss of Hes1 only induced a developmental delay, we also quantified Cone Arrestin+ (Carr3+) cells in P21 control, Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO eyes (Fig. 6G–I,L). Both conditional mutants contained significantly fewer Carr3+ cells, which was more dramatic in Rax-Cre;Hes1CKO/CKO retinas, where the Carr3+ cones had abnormal outer segments (Fig. 6I, arrow, L). We conclude that canonical Notch signaling does not normally suppress cone photoreceptor genesis, via Hes1 transcriptional repression of cone-promoting genes.
Loss of Hes1 results in failed cone photoreceptor formation. A–C, J, RPCs competent to develop into photoreceptor or bipolar neurons express the Otx2 homeobox protein, as well as early-stage E9–E13 RPE cells (Nishida et al., 2003; Fossat et al., 2007). The percentage of Otx2+ retinal cells was quantified (J), using total DAPI+ nuclei to normalize for the elongated retinal shape of Hes1 mutants (arrows in B, C). D–F, K, Anti-Thrb2 labeling (a cone-specific factor) of E16.5 sections also showed a statistically significant loss of cones in the absence of Hes1 (K). G–I, L, Anti-Arr3 (cone Arrestin) labeling of P21 further highlighted this reduction of cones (L), plus shorter outer segments in the disorganized Arr3+ cones for Rax-Cre; Hes1CKO/CKO eyes (I). L, lens; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 50 μm, with boxed insets at 400× magnification. All graphs display individual data points, mean, and SEM; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. n = 3 biologic replicates (2 or more sections/replicate quantified) per age and genotype.
Hes1 and late-forming retinal cell types
There has been no evaluation of the requirements for Hes1 during postnatal retinogenesis, using loss of function mutations. Here we focused on potential P21 phenotypes for INL cell types namely amacrines, bipolars and Müller glia. Given the small size of both conditional mutant retinas (Fig. 2) and irregular ONL architecture, examination of Hes1 functionality during rod neurogenesis should use different Cre drivers that activate at older ages, with restricted expression to those RPCs giving rise to rods and cones. To estimate the percentage of INL amacrine neurons, we labeled P21 sections with Pax6. However, we found no change upon removal of Hes1 for either Cre driver (Fig. 7A–C,M). For bipolar neurons, we quantified Vsx2+ nuclei (Fig. 7D–F,N), where we found a significant increase in both Chx10-Cre;Hes1CKO/CKO and Rax-Cre;Hes1CKO/CKO retinas compared with controls (Fig. 7N). This was somewhat unanticipated given that in vitro culture of Hes1−/− retinas suggested that fewer bipolar neurons would be produced (Tomita et al., 1996). Finally, we examined Müller glia, which are specialized radial glia that stretch across the retina from apical to basal, with their nuclei in the INL. Numerous studies have shown that loss- or gain-of function mutations in the Notch pathway affect gliogenesis, and that Hes1 is weakly expressed by this cell class (Furukawa et al., 2000; de Melo et al., 2016). We used two different Müller glial markers, Sox9 which is confined to the nucleus and allows for quantification (Fig. 7G–I); and Glutamine Synthetase, a membrane-localized protein that highlights radial glial patterning in the mature retina (Fig. 7J–L′). For Sox9, there was no statistically significant difference in glial cell numbers, among the three genotypes (Fig. 7O). Morphologically, although Sox9+ Müller glia nuclei were aligned normally in the INL (Fig. 7H) of Chx10-Cre;Hes1CKO/CKO eyes, they were randomly arranged in Rax-Cre;Hes1CKO/CKO eyes, including displacement to the ONL (Fig. 7I, arrows).
Evaluation of adult inner retinal cell types in Hes1 retinal mutants. A–C, M, The percentage of Pax6+ INL amacrine cells was equivalent among genotypes. D–F, N, For bipolar neurons, Vsx2+ nuclei were quantified, although some Müller glial cells also express this protein. The Vsx2+ cells were significantly increased in both Hes1 mutants (N). G–L′, O, Two Müller glial markers were evaluated, Sox9 in the nucleus (G–I) and glutamine synthetase in the cell membrane (J–L′). Although the percentage of Sox9+ nuclei was not significantly different among the three genotypes (M), in Rax-Cre;Hes1CKO/CKO retinas, these cells were frequently misplaced in the ONL (arrows in I) and had abnormal radial processes (L′). ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar in A, 50 μm. Graphs display individual data points, mean, and SEM. n.s., not significant; **p < 0.01; ***p < 0.001. n = 3 biologic replicates (2 or more sections/replicate quantified) per age and genotype.
Next, we examined Müller glial radial morphology. Normally, the apical processes of Müller glia MG photoreceptors form tight junctions at the outer limiting membrane, which is integral with retinal integrity. Given that all Hes1 mutant embryos have a patchy OLM (Takatsuka et al., 2004), we wondered how this affected mature Müller glial architecture. Glutamine Synthetase expression was normal in Chx10-Cre;Hes1CKO/CKO eyes (cf. Fig. 7J,J′ and K,K′). However, in Rax-Cre;Hes1CKO/CKO retinas Müller glia processes often did not stretch across the apical-basal extent of the retinal layers and appeared to be sparser (Fig. 7L,L′). Overall, we conclude that Hes1 activity is necessary for normal bipolar neuron and Müller glia formation.
Discussion
The bHLH repressor Hes1 has diverse functions throughout vertebrate development, particularly in the CNS, where it regulates the timing of neurogenesis, the rate of progenitor cell division, gliogenesis, and maintenance of tissue compartment boundaries. These activities correlate with two modes of Hes1 gene expression, low, nonuniform (oscillating) or high, sustained; for example, oscillating Hes1 in neuronal progenitor cells, keeps the pace of cell division, and acts in determining the exit point for differentiation (Shimojo et al., 2008). Here, we extended the known roles of Hes1 during retinal development, by uncovering an unpredicted requirement in promoting cone photoreceptor and bipolar neuron formation. By comparing and contrasting the spatiotemporal phenotypes induced in two Hes1 conditional mutants, we also show the embryonic optic nerve head is another CNS location where high, sustained Hes1 expression maintains a neural-glial boundary. Loss of Hes1 in this domain resulted in optic nerve coloboma, thereby inserting this gene into the genetic hierarchy regulating optic fissure closure.
Hes1 and embryonic retinogenesis
It is generally accepted that retinal mutations in the Notch pathway deplete the RPC pool and produce ectopic early-born cell types, namely RGCs and cones. Indeed, we saw that Hes1 retinal mutants have a significant, early loss of cycling RPCs, and a dramatic increase in RGCs. However, we discovered a substantial loss of cone photoreceptors, which sharply contrasts cone overproduction in other experimental paradigms where Notch signaling is blocked (Jadhav et al., 2006b; Yaron et al., 2006; Riesenberg et al., 2009). We believe that the Hes1 mutant cone phenotype reveals an additional level of regulation for cone genesis. Although it is possible that the substantial loss of cones reflects a change in cell fate, to the benefit of RGCs, this seems unlikely, given the simultaneous increase of RGCs and cones in other Notch retinal mutants. Alternatively, developing cones could be eliminated via apoptosis, in the absence of Hes1. We do not favor this possibility because the percentages of Otx2- and Crx-expressing retinal cells are already abnormally low at E13.5, when photoreceptor production is largely cones (Riesenberg et al., 2009; this paper). Although at E16.5, we noted an increase in apoptosis, it is much milder than the loss of Thrb2+ cones. Finally, Hes genes are normally not present in differentiated cone photoreceptor cells, and Hes5 by itself, has no role in embryonic retinal neurogenesis (Hojo et al., 2000). Previously, we reported an exclusion of Hes5-GFP BAC transgenic expression in nascent embryonic cones (Riesenberg et al., 2018). Moreover, Hes5 expression is upregulated in Hes1 mutants (Hatakeyama et al., 2004) which suggests that the removal of Hes1 releases sufficient Hes5 activity to ectopically squelch cone differentiation. To address this possibility, future experiments will evaluate the early retinal neurogenic phenotypes of mutants lacking multiple Hes genes.
We observed multiple rosette-like cell clusters in both Chx10-Cre;Hes1CKO/CKO eyes and Rax-Cre;Hes1CKO/CKO retinas. Rosettes are mainly composed of photoreceptors, but the exact mechanisms underlying their formation are currently unknown. However, it is possible that the precocious overproduction of RGCs and their incorrect positions in the retina can impact the arrangement of photoreceptors into rosette-like structures, or another aspect of retinal lamination. Indeed, RGC apical to basal migration to create the ganglion cell layer is a critical first step in the progression of retinal lamination. It is plausible that after their final cell division, the nascent RGC nuclei are unable to migrate to their final position, thus become “stuck” at the apical surface of Hes1 mutant retinas. Another interrelated possibility is that the rapid appearance of excess RGCs simply blocks the available space for RPCs to rise to the apical surface, disrupting cell–cell signaling dynamics that include Delta–Notch binding (Norden et al., 2009; Hiscock et al., 2018). In other regions of the CNS, Notch signaling is essential for cortical neuron migration (Hashimoto-Torii et al., 2008), and Hes1 mutants in particular affect the movement of hypothalamic neurons (Aujla et al., 2011). Another consequence of premature RGC neurogenesis appears to also affect the outer limiting membrane (OLM), where extra developing RGCs put pressure on the OLM, inducing localized breaks that allow Pou4f+, Tubb3+ retinal cells to spill into the subretinal space. Subsequently, differentiating photoreceptors may be physically blocked from coalescing apically to initiate the ONL, resulting in their clustering. To understand the cause and effect of RPC, RGC and early photoreceptor cell dynamics requires time-lapse imaging in various control and mutant retinas.
Hes1 and postnatal retinogenesis
In addition to major changes in cone numbers, we also report a significant increase in bipolar numbers and a trend towards reduced Müller glia number. In the mouse retina, the generation of different retinal cell types follows a temporal sequence, in which bipolar interneurons and Müller glia are generated last, almost entirely during postnatal development. The earlier premature differentiation of RGCs in Hes1 mutants should negatively impact later born cell types, via progressive depletion of the progenitor pool. However, we found a 1.5-fold overproduction of bipolars in both Hes1 retinal mutants. One explanation could be that in Hes1 mutants, cells that normally develop into Müller glia adopt a bipolar cell fate. Although there is precedent for bipolar-Müller glial cell fate switches, for example in Ascl1/Math3(Atoh3) double mutants (Hatakeyama et al., 2001), we did not observe an equivalent, statistically significant loss of Müller glia in either Hes1 conditional mutants. To provide better evidence for this idea, birthdating studies and possibly Hes1 deletion with a postnatal-specific Cre driver would be necessary.
Hes1 and the optic nerve head
In the vertebrate retina, multiple studies have shown low, nonuniform Hes1 expression, sometimes termed a “salt and pepper” pattern, within RPCs. There is already precedent that such gene expression generates diversity in an otherwise genetically identical progenitor cell population. In other CNS contexts low, nonuniform Hes1+ cells oscillate, which correlates the promotion of neural stem cell or progenitor cell growth and suppression of differentiation, phenotypes first associated with Hes1 mutant retinas >20 years ago (Tomita et al., 1996). Here, we demonstrate a previously unknown role for Hes1 in eye formation: that of maintaining the neural-glial boundary at the ONH via high, sustained Hes1 protein expression. Interestingly, high-level Hes1 expression is strikingly similar to the stripe of Hairy gene expression separating anterior ocular progenitor cells, from those within the morphogenetic furrow, where photoreceptor differentiation begins in Drosophila eye imaginal discs (Brown et al., 1991). Thus, we are curious to understand whether mammalian high Hes1 expression domains might represent components of an evolutionarily conserved module that could even relate to the regulation of Hairy striped expression in the fly embryo, wing, or leg imaginal discs.
In the developing optic cup and stalk high, sustained Hes1 initiates at the border between these two compartments, which is a molecular boundary for OC Pax6 and ONH/OS Pax2 expression. Because both genes initially coexpress and then resolve apart in Rax-Cre;Hes1CKO/CKO eyes, we conclude that Hes1 maintains this boundary. Indeed, the initial report of sustained Hes1 expression was at the midbrain–hindbrain boundary (MHB) or isthmus (Baek et al., 2006), which regulates the development and patterning of neighboring brains compartments by secreting Fgf, Wnt and Shh signals (Gibbs et al., 2017). Interestingly these signaling pathways are also important during eye morphogenesis and patterning (Fuhrmann, 2008, 2010; Cavodeassi et al., 2019). Our discovery that Hes1 is an integral component of the ONH provokes several questions. First, does its removal result in defective morphogen signaling, and if so, which signaling molecules are affected, and are they sufficient to account for the shift in Pax6-Pax2 border and retinal tissue expansion? Moreover, does the ONH represent an ocular signaling center, and how does simultaneous signaling become integrated? What seems clear at this point is that Hes1 regulation of the neural-retinal boundary is independent of canonical Notch signaling, since Rax-Cre deletion of Notch1 or Rbpj removes low Hes1 in RPCs, but not the high Hes1 ONH domain (data not shown).
Interestingly, other mouse mutants exhibit strikingly similar phenotypes to Rax-Cre;Hes1 mutants. Most notably, Pax2 mutant eyes lack both ONH and OS oligodendrocytes, with these tissues erroneously adopting a neural retinal fate (Schwarz et al., 2000). Vax1/Vax2 double mutants also exhibit retinal expansion tissue into the optic stalk (Take-uchi et al., 2003). In both mutants, the Pax6+ RPE was also expanded, at the expense of the optic stalk, which manifested as a hyperpigmentation that extended proximally toward the midline. Given the similarities of ONH phenotypes and directly overlapping expression between Hes1 and Pax2, we envision a direct link between these two factors. Because Pax2 expression initiates normally and is sustained in Hes1 mutants, we conclude that Pax2 is upstream of Hes1. Future genetic epistasis testing and molecular profiling experiments will reveal the relationship between these two genes, and identify other factors that act during ONH formation. We predict that cell adhesion and extracellular matrix genes will be identified through these studies, since these types of proteins are already known to be important during optic fissure closure in vertebrates (Chen et al., 2012; Noh et al., 2016; Gestri et al., 2018; Gordon et al., 2018; Hardy et al., 2019). It will be extremely interesting to integrate Hes1 into the networks of optic fissure closure and coloboma-associated genes.
Footnotes
This work was supported by the National Eye Institute–National Institutes of Health (Grant EY13612 to N.L.B. and Grant P30 EY012576 to U.C. Davis), a Grant-in-Aid for Scientific Research on Innovative Areas (Grant 16H06480 to R.K.) from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan, and AMED-CREST (Grant JP18gm1110002 to R.K.) from the Japan Agency for Medical Research and Development. We thank Doug Forrest for Thrb2 antibody, Brad Shibata and Paul FitzGerald for assistance with histology, Julissa-Suarez Navarro and Amy Riesenberg for technical support, the U.C. Davis Eye Development group for insightful discussions, and Anna La Torre and Joel Miesfeld for critical reading of this manuscript.
The authors declare no competing financial interests.
- Correspondence should be addressed to Nadean L. Brown at nlbrown{at}ucdavis.edu
