NeuroD factors regulate cell fate and neurite stratification in the developing retina

Abstract

Members of the basic helix-loop-helix (bHLH) family of transcription factors have been shown to control critical aspects of development in many tissues. To identify bHLH genes that might regulate specific aspects of retinal cell development, we investigated the expression of bHLH genes in single, developing mouse retinal cells, with particular emphasis on the NeuroD family. Two of these factors, NeuroD2 and NeuroD6/NEX, had not been previously reported as expressed in the retina. A series of loss- and gain-of-function experiments was performed, which suggested that NeuroD genes have both similarities and differences in their activities. Notably, misexpression of NeuroD genes can direct amacrine cell processes to two to three specific sublaminae in the inner plexiform layer. This effect is specific to cell type and NeuroD gene, as the AII amacrine cell type is refractory to the effects of NeuroD1 and NeuroD6, but uniquely sensitive to the effect of NeuroD2 on neurite targeting. Additionally, NeuroD2 is endogenously expressed in AII amacrine cells, among others, and loss of NeuroD2 function results in a partial loss of AII amacrine cells. The effects of misexpressing NeuroD genes on retinal cell fate determination also suggested shared and divergent functions. Remarkably, NeuroD2 misexpression induced ganglion cell production even after the normal developmental window of ganglion cell genesis. Together, these data suggest that members of the NeuroD family are important for neuronal cell type identity and may be involved in several cell type-specific aspects of retinal development, including fate determination, differentiation, morphological development, and circuit formation.

Figure 1.

Distinct atonal-like basic helix-loop-helix transcription factors are expressed in specific amacrine cell types during development. A heat map generated using Treeview software of atonal-like bHLH expression in 32 individual amacrine cells profiled between E16 and P8. Single profiled amacrine cells belonging to GABAergic (green), glycinergic (red), and other (yellow) amacrine cell groups express distinct bHLH family members. Ngn2 is expressed in a subset of GABAergic amacrine cell types, but not glycinergic or other amacrine cell types. NeuroD1 is expressed in cholinergic and AII amacrine cell types, among others. NeuroD2 expression is specific to AII amacrine cells and one other glycinergic amacrine cell. NeuroD4/Math3 is expressed in AII amacrine cells and is biased toward amacrine cells of the glycinergic group, but is expressed in other amacrine cells, including at least one type of GABAergic cell. NeuroD6 was expressed in one amacrine cell that did not express markers of GABAergic or glycinergic amacrine cells. Molecular classification of amacrine cell type was made based on expression of known amacrine cell type markers (Gad1⁺, GABAergic; GlyT1⁺, glycinergic; VAChT, cholinergic; Gtf2h4, AII amacrine; Gad1⁻/GlyT1⁻, other) and according to Pearson's correlation and Ward's clustering method of single-cell transcriptional profiles (Cherry et al., 2009).

Figure 2.

NeuroD2 is expressed in amacrine, ganglion, and bipolar cells of the mature mouse retina. The expression of NeuroD2 was examined by immunofluorescent detection of βgal in mice heterozygous for the NeuroD2 knock-out/lacZ knock-in allele. A, At P22, NeuroD2 is expressed in cells of the upper (b) and lower (c) INL and GCL (d) of the retina, as seen in cross section. B–D correspond to three horizontal planes of the retina at the level of b–d, respectively. B, βgal⁺ cells in the horizontal plane of the upper INL. C, βgal⁺ cells in the horizontal plane of the lower INL. D, βgal⁺ cells in the horizontal plane of the GCL. E, NeuroD2 (green) is expressed in a subset of Chx10⁺ (red) bipolar cells. F, NeuroD2 (green) is not expressed in p27⁺ (red) Müller glial cells. G, NeuroD2 is expressed in a subset of Pax6⁺ (red) amacrine cells in the lower INL. F, NeuroD2 is expressed in a subset of Brn3⁺ (red) ganglion cells in the GCL. DAPI, Blue. In E, G, and H, the arrowheads point to double-positive cells. Scale bars, 40 μm.

Figure 3.

NeuroD6:cre fate maps horizontal, amacrine, and ganglion cells of the mature mouse retina. The fate of NeuroD6-expressing cells was examined by immunofluorescent detection of βgal in mice heterozygous for the NeuroD6 knock-out/cre knock-in allele and the Rosa26 lacZ reporter of cre activity (R26R). A, At P22, NeuroD6 fate-mapped cells are found in the upper (b) and lower (c) INL, as seen in cross section. B and C correspond to two horizontal planes of the retina at the level of b and c, respectively. B, βgal⁺ cells in the horizontal plane of the upper INL directly below the OPL. C, βgal⁺ cells in the horizontal plane of the lower INL directly above the IPL. D, A subset of fate-mapped cells in the lower INL continue to express Cre-recombinase from the NeuroD6 allele at P22, as seen using immunofluorescence for Cre (red). E, NeuroD6 fate-mapped cells expressing βgal (green) do not correspond to Chx10⁺ (red) bipolar cells. F, NeuroD6 does not fate map (green) p27⁺ (red) Müller glial cells. G, NeuroD6 does fate map a subset of Pax6⁺ (red) horizontal cells in the upper INL and a subset of Pax6⁺ amacrine cells in the lower INL. H, NeuroD6 fate maps a small subset of Brn3⁺ (red) ganglion cells in the GCL. DAPI, Blue. In D, G, and H, the arrowheads point to double-positive cells. In D, the chevron points to a βgal⁺ (green) cell that no longer expresses Cre (red). Scale bars, 40 μm.

Figure 4.

NeuroD1, NeuroD2, and NeuroD6 expression is specific to distinct neuron cell types of the mature retina. The expression of NeuroD1 (A–D) or NeuroD2 (E–H) was examined using immunofluorescent detection of βgal in mice heterozygous for the NeuroD1 or NeuroD2 knock-out/lacZ knock-in allele. The fate map of NeuroD6-expressing cells or constitutive expression of NeuroD6 was examined using immunofluorescent detection of βgal in mice heterozygous for the NeuroD6:cre allele and hemizygous for the R26R lacZ cre-responsive allele (I, J) or by immunofluorescent detection of cre, respectively (K). Specific cell types were defined according to immunoreactivity to cell type-specific markers. A, At P22, NeuroD1 (green) is expressed in PKCα⁺ (red) rod bipolar cells (RBPs) (white arrowheads). B, NeuroD1 is expressed in Dab1⁺ (red) AII amacrine cells (white arrowheads). C, NeuroD1 is also expressed in ChAT⁺ (red) starburst (SB) amacrine cells in the INL (white arrowheads). NeuroD1-expressing cells (green) in the GCL are ChAT⁺ (red) displaced starburst amacrine cells (dSB) (white arrowheads). D, A schematic of retinal cell types expressing NeuroD1. Three of these cells, rod photoreceptors (PR), RBPs, and AII amacrine cells comprise the first part of the rod circuit. NeuroD1 is also expressed in at least one unidentified type of PKCα⁻ cone bipolar cell (CBP) and SB and dSB amacrine cells. E, NeuroD2 is not expressed in PKCα⁺ (red) RBP cells (white chevrons), although in the lower INL it is expressed in a subset of PKCα⁺ (red) amacrine cells (white arrowheads). F, Dab1⁺ AII amacrine cells (red) express NeuroD2 (green) (white arrowheads). G, ChAT⁺ amacrine cell types do not express NeuroD2. H, A schematic of retinal cell types that express NeuroD2. NeuroD2 is expressed in at least one type of PKCα⁻ cone bipolar cell, AII amacrine cells, at least one type of PKCα⁺, non-AII amacrine cell, and at least one ganglion cell type (GC) (see Fig. 2H for GC staining). I, A subset of NeuroD6:cre fate-mapped cells (green) express GlyT1 (red) (white arrowhead). Another subset of NeuroD6 fate-mapped cells (green) in the inner INL do not (white chevron). J, NeuroD6:cre fate-mapped cells (green) do not express Gad65/67 (red). K, Cells actively expressing cre from the NeuroD6 locus (red) (white chevrons) [the asterisk (*) denotes background in vasculature] do not express GlyT1 (green) or Gad65/67 (blue). L, A schematic of retinal cells fate mapped by NeuroD6:cre (green) and cells actively expressing NeuroD6:cre in the mature retina. Horizontal cells, at least one type of glycinergic amacrine cell, at least one type GlyT1⁻/Gad⁻ amacrine cell, and at least one type of ganglion cell are fate mapped by NeuroD6:cre (green and red) (see Fig. 3H for GC staining). At least one type GlyT1⁻/Gad⁻ amacrine cell (red) is actively expressing NeuroD6 in the adult retinal. DAPI, Blue. Scale bars, 40 μm.

Figure 5.

Cell type-specific bHLH expression in retinal progenitor cells. A, A heat map of bHLH transcription factor expression in single profiled retinal progenitor cells. Atonal-like bHLH factors that were expressed in specific subsets of developing amacrine cells are expressed in individually profiled retinal progenitor cells between E12 and P5. Ngn2, NeuroD1, NeuroD4/Math3, and Bhlhb5 were widely expressed in retinal progenitor cells throughout development. NeuroD2 was only expressed in one progenitor cell, at P0. NeuroD6 was not expressed in these cells. These cells were previously identified as progenitor cells according to known marker expression and a post hoc classification based on clusters of coexpressed genes centered around known markers of retinal progenitor cells (Trimarchi et al., 2008). B, Retinae were labeled in vivo at P0 with a pulse of EdU (green) and a 2 h chase, and then subjected to dissociation and RNA in situ hybridization for NeuroD1 or NeuroD2 (red). NeuroD1 and NeuroD2 were expressed in a subset of EdU⁺ P0 retinal progenitor cells in, or recently in, S-phase of the cell cycle; however, most EdU⁺ cells did not express NeuroD1 or NeuroD2. There were also many NeuroD1- or NeuroD2-expressing cells that were not labeled by EdU. These cells are either postmitotic cells or progenitor cells that were not in S-phase during the time of EdU labeling. The green and yellow pie charts show the percentage of EdU-labeled cells that express a NeuroD factor (yellow) or do not express NeuroD factors (green). The multicolored pie charts show the percentage of total cells that were labeled with EdU and express NeuroD factors (yellow), that were labeled with EdU, but do not express NeuroD factors (green), that express NeuroD factors, but were not EdU⁺ (red), or cells that were not labeled by EdU or NeuroD⁺ (blue). NeuroD1: n = 3 retinae, >380 cells were counted per retina. NeuroD2: n = 3 retinae, >1370 cells were counted per condition. Scale bar, 10 μm.

Figure 6.

NeuroD2 inactivation results in a reduction in AII amacrine cells. The effects of losing one or both alleles of NeuroD2 on retinal cells were analyzed by immunofluorescent detection of βgal from NeuroD2 knock-out/lacZ knock-in alleles, or by cell type-specific markers in wild type (wt), heterozygous (het), and null littermates. A, βgal-expressing cells (green), including Dab1⁺ AII amacrine cells (red), in a NeuroD2 heterozygous retina. B, βgal-expressing cells (green) including Dab1⁺ AII amacrines (red) in a NeuroD2 null retina. C, The number of βgal-expressing cells per 375 × 375 μm field was quantified in NeuroD2 heterozygous and null retinae and compared using Student's t test, two-tailed, homoscedastic, p = 0.27. D–F, Dab1⁺ AII amacrine cells (red) in wt, het, and NeuroD2 null retinae. G, Numbers of AII amacrine cells were compared between wt, het, and NeuroD2 null retinae, *p = 0.008. C, G, n ≥ 3 retinae for each genotype and ≥2 fields per retina. DAPI, Blue. Scale bar, 40 μm. Error bars indicate SD.

Figure 7.

Forced expression of individual NeuroD family members in the newborn mouse retina. Plasmids encoding GFP, NeuroD1-ires-GFP, NeuroD2-ires-GFP, or NeuroD6-ires-GFP were individually electroporated in vivo into P0 mouse retinae. The fates of electroporated cells were then analyzed in the mature retina at P21. A–D, In vivo electroporation of GFP at P0 labeled photoreceptor cells in the ONL of mature retinae, as well as Chx10⁺ (red; A) bipolar cells, Sox9⁺ (red; B) Müller glial cells, and Pax6⁺ (red; C) amacrine cells in the INL. E–H, Electroporation of NeuroD1 gave rise to photoreceptors, Chx10⁺ bipolar cells, and Pax6⁺ amacrine cells, but suppressed the Sox9⁺ Müller glial cell fate. I–L, Cells electroporated with NeuroD2 became photoreceptors and Pax6⁺ amacrine cells, but did not adopt Chx10⁺ bipolar or Sox9⁺ Müller glial cell fates. K, Cells electroporated with NeuroD2 were also observed in the GCL. M–P, Electroporation of NeuroD6 gave rise to photoreceptors and amacrine cells, but suppressed bipolar and Müller glial cells. Scale bar, 40 μm. The asterisk (*) denotes cells with AII amacrine cell type-like morphology.

Figure 8.

Quantitative analysis of cell fate in the mature retina after forced expression of individual NeuroD family members at P0. A–E, In vivo electroporation of GFP at P0 labeled photoreceptors, bipolar cells, Müller glia, and amacrine cells. A, C, Electroporation of NeuroD1 gave rise to a small, but significant increase in photoreceptors and a significant decrease in Müller glial cells. C–E, Electroporation of NeuroD2 gave rise to significant increases in the percentage of amacrine cells and cells in the GCL, and a significant suppression of Müller glial cells. Electroporation of NeuroD6 gave rise to a significant increase in amacrine cells, compared with GFP alone, and a significant decrease in Müller glial cells. n ≥ 3 retinae per condition; >950 cells were counted per retina. *p < 0.05, **p < 0.005, ***p < 0.0005. Error bars indicate SEM.

Figure 9.

Forced expression of NeuroD2 in the newborn retina induces ganglion cell genesis. Plasmids encoding GFP or NeuroD2-ires-GFP were individually electroporated in vivo into P0 mouse retinae. Electroporated retinae were then analyzed at P12. A, Electroporation of GFP alone at P0 labels cells in the ONL and INL at P12. B, Cells electroporated with GFP at P0 do not extend axons into ONH and GFP-expressing ganglion cell axons are not present in the fiber layer (FL) of the retina. C, D, Electroporation of NeuroD2 at P0 results in cells in the ONL, INL, and GCL at P12. Axons from ganglion cells induced by forced expression of NeuroD2 are present in the fiber layer of the retina and in the ONH. Scale bar, 40 μm.

Figure 10.

Forced expression of individual NeuroD family members in the mouse retina at P3. Plasmids encoding GFP, NeuroD1-ires-GFP, NeuroD2-ires-GFP, or NeuroD6-ires-GFP were individually electroporated in vivo into P3 mouse retinae. The fates of electroporated cells were then analyzed in the mature retina at P21. A–D, In vivo electroporation of GFP at P3 labeled photoreceptor cells in the ONL, as well as Chx10⁺ bipolar cells (red; A), Sox9⁺ Müller glial cells (red; B), and Pax6⁺ amacrine cells (red; C), in the INL. E–P, Electroporation of NeuroD1 (E–H), NeuroD2 (I–L), or NeuroD6 (M–P) gave rise to photoreceptors, Chx10⁺ bipolar cells, and Pax6⁺ amacrine cells, but suppressed the Sox9⁺ Müller glial cell fate. Scale bar, 40 μm.

Figure 11.

Quantitative analysis of cell fate in the mature retina after forced expression of individual NeuroD family members at P3. A–D, The greatest percentage of cells electroporated in vivo at P3 with GFP became photoreceptors, followed by bipolar cells, Müller glia, and amacrine cells. C, Electroporation of NeuroD1 did not significantly change the percentages of cells becoming photoreceptor, bipolar, or amacrine cells, but resulted in a significant decrease in Müller glial cells. C, D, Electroporation of NeuroD2 gave rise to significant increases in the percentage of amacrine cell and a significant suppression of Müller glial cells. Electroporation of NeuroD6 gave rise to a significant increase in amacrine cells and significant decreases in photoreceptors and Müller glial cells. n ≥ 3 retinae per condition, except NeuroD6, n = 2; >950 cells were counted per retina. *p < 0.05, **p < 0.005, ***p < 0.0005. Error bars indicate SEM.

Figure 12.

Stratification of neuronal processes in the IPL at P21 after introduction of individual NeuroD factors in vivo at P0. A, Electroporation of the newborn mouse retina with GFP (green)-labeled amacrine cells in the lower INL that displayed diffusely stratified neuritic arbors in the IPL, including amacrine cells with stereotyped AII amacrine cell morphology (white arrowhead, white brackets). B–D, Cells with stereotyped AII amacrine cell morphologies were also found in NeuroD1 and NeuroD6 electroporated retinae, but not in retinae electroporated with NeuroD2. A–H, Compared with GFP alone, individual NeuroD factors bias the localization of certain neurites in the IPL to three distinct sublaminae (white chevrons), which intercalate between ChAT⁺ amacrine cell processes (red) (E–H). I, J, L, M, N, P, Cells electroporated with GFP, NeuroD1, or NeuroD6 have characteristic AII amacrine cell morphology and colocalize with Dab1, a molecular marker for AII amacrine cells (white arrowheads). K, O, Cells electroporated with NeuroD2 colocalize with Dab1, but do not possess typical AII amacrine cell morphology and instead costratify with other processes. Scale bars, 40 μm.

Figure 13.

Summary of NeuroD family member expression and gain-of-function analyses in the murine retina. A, NeuroD1 is expressed in a subset of retinal progenitor cells (rpc) in the developing retina. In the mature retina, NeuroD1 is expressed in rods and cones, on-bipolar cells (bp^on), and in cholinergic and AII amacrine cells (ac^chat,AII). NeuroD2 is expressed in a smaller subset of retinal progenitor cells during development and in subsets of cone-bipolar (bp^cone) and ganglion (gc) cells. NeuroD2 is also expressed in AII amacrine cells and another subset of amacrine cells (ac^AII+). NeuroD6 is expressed in non-GABAergic, non-glycinergic amacrine cells (ac^subset) in the adult retina. Horizontal cells (hc) and a subset of ganglion cells (gc^subset) have a history of NeuroD6 expression, but this expression is not maintained in the adult. It is unknown whether rpcs expressing one or more NeuroD family members give rise to the cell types that express NeuroD genes in the mature retina. B, The effects of NeuroD family member misexpression on retinal cell fate at P0. The relative thickness of arrows reflects the production of each cell class compared with electroporation of GFP alone. Electroporation of the retina with GFP at P0 labels rpc and postmitotic cells that give rise to rod photoreceptors (rod), bipolar cells (bp), Müller glia (mg), and amacrine cells. Misexpression of NeuroD1 by electroporation increases the percentage of cells that adopt the rod photoreceptor cell fate, reduces the percentage of bipolar cells produced, and suppresses the Müller glial cell fate, but does not affect amacrine cell production. Misexpression of NeuroD2 suppresses the Müller glial and bipolar cell fates and induces amacrine cell and ganglion cell production, but does not influence the rod photoreceptor cell fate. NeuroD6 expression induces amacrine cells most strongly, while suppressing bipolar and Müller glial cell production. The photoreceptor fate is unchanged by NeuroD6 expression at P0. C, The effect of NeuroD family member expression on amacrine cell neurite targeting. Electroporation of GFP at P0 mostly labels diffusely stratified narrow-field amacrine cells. NeuroD1 and NeuroD6 do not affect AII amacrine cell morphology; however, some bias is seen in other amacrine cell processes toward two to three specific strata in the IPL. NeuroD2 causes all amacrine cells, including AII amacrine cells (*), to stratify to two to three specific strata within the IPL.