Foxn3 is required to suppress aberrant ciliogenesis in nonphotoreceptor retinal neurons

Abstract

The retinal photoreceptors possess specialized sensory cilia critical for phototransduction while the nonphotoreceptor cells typically exhibit simpler primary cilia or lack them altogether. This dichotomy in ciliary architecture underpins the functional specialization of retinal cell types, but how this dichotomy arises and is maintained remains elusive. This study explores the role of the transcription factor Foxn3 in establishing and maintaining this divergence. We generated retina-specific Foxn3 conditional knockout (Foxn3CKO) mice, which show that Foxn3 is essential for repressing ciliary gene expression in nonphotoreceptor cells, such as bipolar and amacrine cells. Foxn3CKO mice exhibit significant reductions in electroretinogram b-wave amplitudes and oscillatory potentials, indicating functional impairments in inner retinal neurons. Loss of Foxn3 leads to ectopic ciliary gene expression and abnormal ciliogenesis in nonphotoreceptor neurons, without affecting retinal cell specification and differentiation. Single-Cell RNA Sequencing, chromatin profiling, and transcription assays reveal that Foxn3 directly binds to and represses the promoters of ciliary genes and their transactivators, including Foxj1 and Rfx family members. Our data together highlight Foxn3 as a key transcriptional repressor that may function to ensure the proper ciliary architecture of retinal neurons by preventing nonphotoreceptor neurons from adopting photoreceptor-like ciliary features and provide insights into the molecular mechanisms governing retinal development and ciliopathies.

Keywords: Foxn3; ciliogenesis; primary cilia; retinal ciliopathy; sensory cilia.

Fig. 1.

Decreased scotopic ERG responses of Foxn3CKO mice. (A) Schematic diagram illustrating retina-specific conditional ablation of Foxn3 (Foxn3CKO) in mice. The floxed (fl) allele was generated by homologous recombination, from which the CKO allele was derived using the Six3-Cre mice to delete exon 2. The arrows underneath exon 2 indicate the positions of oligonucleotide primers used for the qRT-PCR assay. (B) PCR analysis of genomic DNA to identify mice containing the wild-type (WT) allele, fl allele, and Cre. The wild-type and fl alleles yield a product of 246 bp and 305 bp, respectively. (C) qRT-PCR assay of Foxn3 exon 2 expression levels in control and Foxn3CKO retinas using a pair of exon 2-specific primers. Data are presented as mean ± SD (n = 9). ****P < 0.0001. (D) Western blot analysis of Foxn3 in adult control and Foxn3CKO retinas. Glyceraldehyde-3-phosphate dehydrogenase (Gapdh) served as the internal protein control. (E) Comparison of adult (2-mo, 2 M) control and Foxn3CKO eyeballs and optic nerves. (F) The size of the optic nerve (ON), optic chiasm (OC), and optic tract (OT) in adult (4-mo, 4 M) control and Foxn3CKO mice exhibits no obvious difference. (G) Retinal sections from adult (3-mo, 3 M) control and Foxn3CKO animals were stained with HE (hematoxylin-eosin). (H) Representative ERG waveforms from dark-adapted control and Foxn3CKO mice aged 8 mo. The flash intensity used to elicit the responses is given to the left of each pair of responses. (I and J) Intensity–response functions of scotopic ERG a waves (I) and b waves (J) for control and Foxn3CKO mice aged 8 mo. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. (K) Representative ERG waveforms from light-adapted control and Foxn3CKO mice aged 8 mo. The flash intensity used to elicit the responses is given to the left of each pair of responses. (L and M) Intensity–response functions of photopic ERG a waves (L) and b waves (M) for control and Foxn3CKO mice aged 8 mo. (N) Oscillatory potentials (OPs) in scotopic ERG at 3 cd.s/m². OP amplitudes were quantified using rms for two-frequency bands in eight horizontal and seven vertical locations. Data are presented as mean ± SEM (n = 6 or 8). **P < 0.01. (O) OPs in photopic ERG at 3 cd.s/m². OP amplitudes were quantified using rms for two-frequency bands in eight horizontal and seven vertical locations. Data are presented as mean ± SEM (n = 10 or 12). Abbreviations: GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; IS, inner segment; ONL, outer nuclear layer; OPL, outer plexiform layer; OS, outer segment. (Scale bar: G, 20 μm.)

Fig. 2.

Effect of conditional Foxn3 ablation on the differentiation of different retinal cell types. (A) Retinal sections from 8-mo-old control and Fonx3CKO mice were immunostained with antibodies against the indicated cell type–specific protein markers and counterstained with nuclear DAPI. Between retinas of the two genotypes, there was no obvious difference in the number of bipolar cells immunoreactive for Chx10 and PKCα, amacrine cells immunoreactive for Tfap2a and Calbindin, HCs immunoreactive for Calbindin, RGCs immunoreactive for Rbpms, Müller cells immunoreactive for Sox9, rod cells immunoreactive for Recoverin, and cone cells immunoreactive for cone Arrestin. (B) Quantitation of cells that are immunoreactive for several cell type–specific markers. Each histogram represents the mean ± SEM for 3 to 6 retinas. Abbreviations: GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; ONL, outer nuclear layer; OPL, outer plexiform layer. (Scale bar: A, 20 μm.)

Fig. 3.

Changes in single-cell transcriptome profiles between 3-mo control and Foxn3CKO retinas. (A) UMAP plot showing different cell-type clusters in a merged dataset of single cells from 3-mo control and Foxn3CKO retinas. (B) UMAP plots comparing the cell types in control and Foxn3CKO retinas. (C) Percentage of cells in all cell-type clusters of control and Foxn3CKO retinas. (D) Scatter plot analysis of the global gene expression profiles in control and Foxn3CKO retinas. Average gene expression levels are depicted in log10 scale. The diagonal line represents equal expression in the two genotypes. (E) Volcano plot [significance versus fold change (FC)] of differentially expressed genes (DEGs) (FC ≥ 2 and P-value < 0.05) between Foxn3CKO and control retinas, with top 10 down- and up-regulated genes labeled. The dashed horizontal line indicates P-value = 0.05 and the two vertical dashed lines indicate log₂FC = 1 or −1. (F) Heatmap of the z-transformed expression values of the DEGs between Foxn3CKO and control retinas. (G) Dot plot showing expression patterns of the top 50 upregulated genes in the bipolar, amacrine, cone, and rod single-cell clusters of the control and Foxn3CKO retinas. (H) Top 20 enriched GO terms for the DEGs between Foxn3CKO and control retinas. (I) Network plot of the top four enriched GO terms or gene sets (nodes) and their associated DEGs. Node size represents the gene-set size. (J) Stacked violin plots showing expression patterns of representative ciliary genes in the bipolar, amacrine, cone, and rod single-cell clusters of the control and Foxn3CKO retinas.

Fig. 4.

Upregulation of ciliary gene expression in 3-mo Foxn3CKO retinas. (A) Split UMAP feature plots showing the expression patterns of the indicated ciliary genes in control and Foxn3CKO retinas. (B) qRT-PCR assay of the RNA expression levels of the indicated ciliary genes in control and Foxn3CKO retinas. Data are presented as mean ± SEM (n = 3). **P < 0.01; ***P < 0.001; ****P < 0.0001. (C) In control and Foxn3CKO retinas, expression levels of the indicated ciliary genes were examined by section RNA in situ hybridization analysis. All 10 genes display increased expression within the INL of Foxn3CKO retinas compared to that of the control. (D) Retinal sections from control and Foxn3CKO mice were immunostained with the GT335 antibody, or double-immunostained with GT335 and antibodies against Chx10 or Tfap2a, with counterlabeling by nuclear DAPI. The outlined areas 1 and 2 are shown below at a higher magnification as two panels. (E) Quantification of GT335-immunoreactive cilia in the inner and outer halves of the INL in control and Foxn3CKO retinas. Data are presented as mean ± SEM (n = 3). **P < 0.01; ****P < 0.0001. Abbreviations: CC, connecting cilium; GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; IS, inner segment; ONL, outer nuclear layer; OPL, outer plexiform layer. (Scale bar: C, 20 μm; D, 5 μm.)

Fig. 5.

Upregulation in expression of ciliary regulatory TF genes in 3-mo Foxn3CKO retinas. (A) Split UMAP feature plots showing the expression patterns of the indicated ciliary TF genes in control and Foxn3CKO retinas. (B) Violin plots showing expression patterns of the indicated ciliary TF genes in the bipolar and amacrine cell clusters of control and Foxn3CKO retinas. Each dot represents a single cell. (C) qRT-PCR assay of the RNA expression levels of the indicated ciliary TF genes in control and Foxn3CKO retinas. Data are presented as mean ± SEM (n = 3 or 4). **P < 0.01; ***P < 0.001; ****P < 0.0001. (D) Schematic of the Foxn3 gene structure. The numbered boxes represent the six exons. Indicated also are the positions of the three pairs of oligonucleotide primers (F, forward; R, reverse) used for the qRT-PCR assay shown in (E). (E) qRT-PCR assay of the Foxn3 RNA expression levels in control and Foxn3CKO retinas using the three pairs of oligonucleotide primers flanking DNA fragments that include part of exon 2 (F1R1 and F2R2) or not (F3R3). Data are presented as mean ± SEM (n = 3). **P < 0.01; ****P < 0.0001.

Fig. 6.

CUT&Tag analysis mapping genomic sites bound by Foxn3 in adult mouse retinas. (A) Heatmaps of the Foxn3 and H3K27me3 CUT&Tag signals around the Foxn3 CUT&Tag peak region. Each row represents a 1-kb region centered on the Foxn3 peak summit, sorted by Foxn3 signal enrichment. IgG serves as the negative control. (B) CUT&Tag fragment coverage of Foxn3, H3K27me3, and IgG centered around the Foxn3 peak summit. (C) Distribution of the 5′ and 3′ tags (reads) from the H3K27me3-marked nucleosomes near the Foxn3 peaks. (D) Frequency of the Foxn3 peaks located in the promoter, intron, intergenic, exon, and TTS (transcription termination site) regions. (E) A top-ranked Foxn3-binding motif (P = 1e−15) identified by de novo motif search in a 300-bp window centered at the peak summit. (F) Representative ciliary gene-related GO terms enriched for the Foxn3 peak-associated genes. (G) Genome browser view of Foxn3, H3K27me3, and IgG CUT&Tag signals at the Cfap54, Ccdc88a, Nphp3, Rfx1, Rfx2, and Foxj1 loci. The y axis represents the number of normalized reads.

Fig. 7.

ChIP and luciferase assays showing occupation and repression of ciliary gene promoters by Foxn3. (A) Schematics of the promoter regions of Drc1, Fam183b, Tekt1, and Foxj1, which harbor one or more FHL motifs. The TSS and translation start site ATG are indicated. The horizontal arrows indicate the positions of PCR primers used to amplify the precipitated DNA fragments. The negative control (NC) fragment is located in exon 1 of Foxj1. (B) Chromatin DNA was prepared from adult mouse retinas, immunoprecipitated by an anti-Foxn3 antibody, and quantified by qRT-PCR. Data are presented as mean ± SEM. ****P < 0.0001; ns, no significance. (C) Schematic of the luciferase assay. A 2-kb promoter (P) region from Cfap52, Drc1, and Foxj1 was inserted upstream of Luc (luciferase) in the pGL3-Basic vector. The open reading frames of the Foxj1, Rfx3, and Foxn3 TF genes were inserted into the pCI expression vector. (D) Relative luciferase activities after cotransfection of the indicated reporter plasmids with control (pCI), Foxj1, Rfx3, and/or Foxn3 expression plasmids in 293T cells. Histograms represent the mean ± SEM of triplicate assays in a single experiment. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, no significance. (E) Relative luciferase activities after cotransfection of the Cfap52 reporter plasmid with control, Foxj1, Rfx3, and/or Foxn3 expression plasmids in 293T cells. Histograms represent the mean ± SEM of triplicate assays in a single experiment. ***P < 0.001; ****P < 0.0001. (F) Relative luciferase activities after cotransfection of the Cfap52 reporter plasmid with control, Foxj1, or Foxj1 with the indicated increasing amount of Foxn3 expression plasmids in 293T cells. Histograms represent the mean ± SEM of triplicate assays in a single experiment. ****P < 0.0001. (G) Relative luciferase activities after cotransfection of the Axin1 or Creb1 reporter plasmids with control, Maf1, or Foxn3 expression plasmids in 293T cells. Histograms represent the mean ± SEM of triplicate assays in a single experiment. **P < 0.01; ****P < 0.0001; ns, no significance.

Fig. 8.

Working model of how Foxn3 may regulate retinal ciliogenesis. In the mammalian retina, the rod and cone photoreceptors sport a large specialized sensory cilium whereas the nonphotoreceptor neurons possess only a simple primary cilium (e.g., amacrine cells and RGCs) or no cilium (e.g., bipolar and HCs). Foxn3 may play a crucial role in the differentiation and maintenance of this ciliary dichotomy by acting as a transcriptional repressor of ciliary genes. The Rfx family of ciliary gene transactivators, in particular Rfx3 and Rfx7, may promote the differentiation and formation of the photoreceptor sensory cilia. In late postnatal and adult retinas, Foxn3 displays strong expression in nonphotoreceptor cells but no expression in photoreceptors. By directly repressing ciliary gene expression and indirectly repressing ciliary gene expression through inhibiting the expression of Rfx genes and Foxj1, Foxn3 appears to prevent nonphotoreceptor neurons from adopting a photoreceptor-like sensory cilium or even forming a simple primary cilium. Both Foxn3 and Rfx TFs may autoregulate their own expression to fine-tune the expression of ciliary genes, ensuring that they are maintained at proper levels conducive for the formation of either the complex sensory cilia in photoreceptors or simple primary cilia in nonphotoreceptor neurons. In addition, Foxn3 may repress the expression of other ciliary genes (both motile and immotile) that are not normally associated with photoreceptors.