|Home | About | Journals | Submit | Contact Us | Français|
Margaret J. Ochocinska, Section on Neuroendocrinology, Program in Developmental Endocrinology and Genetics, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Estela M. Muñoz, Institute of Histology and Embryology, School of Medicine, National University of Cuyo, National Council of Research, Science and Technology (CONICET), ANPCyT, Mendoza, Argentina.
Shobi Veleri, Neurobiology-Neurodegeneration and Repair Laboratory, National Eye Institute, National Institutes of Health, Bethesda, MD, USA.
Joan L. Weller, Section on Neuroendocrinology, Program in Developmental Endocrinology and Genetics, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Steven L. Coon, Section on Neuroendocrinology, Program in Developmental Endocrinology and Genetics, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
Nikita Pozdeyev, Deptartment of Ophthalmology and Pharmacology, Emory University School of Medicine, Atlanta, GA, USA.
P. Michael Iuvone, Deptartment of Ophthalmology and Pharmacology, Emory University School of Medicine, Atlanta, GA, USA.
Sandra Goebbels, Deptartment of Neurogenetics, Max-Planck-Institute of Experimental Medicine, Göttingen, Germany.
Takahisa Furukawa, 4th Department, Osaka Bioscience Institute, Osaka, Japan.
David Charles Klein, Section on Neuroendocrinology, Program in Developmental Endocrinology and Genetics, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD, USA.
NeuroD1 encodes a basic helix-loop-helix (bHLH) transcription factor involved in the development of neural and endocrine structures, including the retina and pineal gland. To determine the effect of NeuroD1 knockout in these tissues, a Cre/loxP recombination strategy was used to target a NeuroD1 floxed gene and generate NeuroD1 conditional knockout (cKO) mice. Tissue specificity was conferred using Cre recombinase expressed under the control of the promoter of Crx, which is selectively expressed in the pineal gland and retina. At two months of age NeuroD1 cKO retinas have a dramatic reduction in rod- and cone-driven electroretinograms and contain shortened and disorganized outer segments; by four months NeuroD1 cKO retinas are devoid of photoreceptors. In contrast, the NeuroD1 cKO pineal gland appears histologically normal. Microarray analysis of two-month-old NeuroD1 cKO retina and pineal gland identified a subset of genes that were affected 2- to 100-fold; in addition, a small group of genes exhibit altered differential night/day expression. Included in the down-regulated genes are Aipl1, which is necessary to prevent retinal degeneration, and Ankrd33, which is selectively expressed in the outer segments. These findings suggest that NeuroD1 may act through Aipl1 and other genes to maintain photoreceptor homeostasis.
NeuroD1 is a basic helix-loop-helix (bHLH) transcription factor implicated in cell cycle regulation, retinal cell genesis, and neuronal development (Miyata et al. 1999, Schwab et al. 2000, Lee et al. 2000, Cai et al. 2000, Liu et al. 2000, Cherry et al. 2011, Ochocinska & Hitchcock, 2009). In the retina, NeuroD1 plays a role in terminal photoreceptor differentiation and survival of rod photoreceptors; loss of NeuroD1 results in progressive photoreceptor degeneration (Morrow et al. 1999, Pennesi et al. 2003). NeuroD1 also functions in cone photoreceptor patterning and mediates cone-specific expression through the regulation of thyroid hormone receptor β2 expression during development (Liu et al. 2008).
NeuroD1 mRNA is also highly abundant in the pineal gland and exhibits a developmental expression pattern similar to that of the retina (Bailey et al. 2009, Muñoz et al. 2007). This is consistent with the common evolutionary origin of pinealocytes and retinal photoreceptors (Klein 2006, Bailey et al. 2009, Donoso et al. 1985, Korf et al. 1985, Korf et al. 1992, Rodrigues et al. 1986, Reig et al. 1990, Schaad et al. 1991, Babila et al. 1992, Lolley et al. 1992). In contrast to the retina, the pineal gland is a relatively homogeneous structure, composed 95% of pinealocytes, which share some genetic features with photoreceptors. This makes the pineal gland a useful model for understanding elements of cell biology shared by both tissues.
Previous studies directed at understanding the role of NeuroD1 in the mouse retina have used a global knockout strategy that eliminates NeuroD1 expression in all tissues, including the pancreas and cerebellum (Morrow et al. 1999, Liu et al. 2008). The majority of these animals die shortly after birth, which limits efforts to study NeuroD1 function in the adult. Moreover, global knock out of NeuroD1 allows the possibility that the effects of this manipulation on the retina are mediated by non-retinal and non-pineal tissues.
To obviate this possibility in the current study, a Cre/loxP recombination strategy was used to limit gene deletion to the pineal gland and retina. Cre recombinase was selectively expressed in pinealocytes and retinal cells, including photoreceptors, by placing it under control of the Crx promoter, which is selectively expressed in both tissues late in gestation (Omori et al. 2012, Rath et al. 2006, Furukawa et al. 2002). Crx-cre mice were crossed with NeuroD1flox/flox mice (Goebbels et al. 2005) to generate NeuroD1flox/flox/Crx-cre+ (cKO) and NeuroD1flox/flox/Crx-cre− control mice. The effects of NeuroD1 cKO on the retina and pineal gland were monitored histologically by immunofluorescence and transmission electron microscopy, and by microarray-based gene expression profiling. Retinal function was monitored using electroretinography. The findings of these studies are presented below.
All animal experiments and treatments were performed in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals and the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines. Mice were housed in a 12:12 Light-Dark Cycle with lights on at Zeitgeber time (ZT) 0 and food and water ad libitum. Both male and female mice were used for all experiments.
NeuroD1 cKO mice: A colony of mice was established on a C57BL/6J background using a male NeuroD1flox/flox animal in which exon 2 of the NeuroD1 gene, which includes the initiating codon, is flanked by two loxP sites (Goebbels et al. 2005). These animals were crossed with Crx-cre mice (Omori et al. 2012), which express Cre recombinase under the control of the 2-kb Crx promoter (Furukawa et al. 2002). See Fig. 1A and Supporting Information for further details.
Agr2−/− mice: A colony of Agr2 −/− mice was established by crossing a male Agr2+/− animal (Dr. David J. Erle, University of California, San Francisco) with a C57BL/6J female (Taconic Farms) to generate heterozygous progeny (Lewandoski et al. 1997, Park et al. 2009). These heterozygous progeny were used to generate Agr2 −/− (Agr2 KO) and age-matched wild type littermate control animals. See Supporting Information for further details.
Pineal glands and retinas from two-month-old NeuroD1 cKO and control mice were removed during the day at ZT6 and at night at ZT20; tissue was placed on solid CO2. Tail samples were collected in parallel for genotype confirmation. See Supporting Information for further details.
Total RNA (~200 ng/pineal gland and 1000 ng/retina) was prepared from pools of six pineal glands or retinas; three pools per organ and per genotype were analyzed at each sampling time. Published methods were used for total RNA isolation, quality control, and cDNA amplification (Rovsing et al, 2011). See Supporting Information for further details. Amplified cDNA was biotinylated using the Encore Biotin Module (NuGEN Technologies, Inc.). Biotinylated cDNA (20 ug) was fragmented and hybridized for 18 h at 45°C to the GeneChip Mouse Genome Array 430 2.0 (Affymetrix, Santa Clara, CA). This GeneChip contained 45,000 probe sets corresponding to 39,000 transcripts and variants from over 34,000 annotated mouse genes. The arrays were stained and washed as described in the Affymetrix protocols.
Affymetrix arrays were scanned using a GeneChip Scanner 3000 (Affymetrix). For each array, the raw signal intensity (.CEL) files were generated and analyzed with Genomatix ChipInspector V2.0 software (Munich, Germany). The following filters were used for all analyses: exhaustive matching, a false discovery rate (FDR) = 0, cut off = 1, region size = 300 bp and a minimum of 4 significant probes. Differences of p < 0.05 were considered to be statistically significant. See Supporting Information for further details. The microarray data are available at the NCBI Gene Expression Omnibus (www.ncbi.nlm.nih.gov/projects/geo/GSE35396).
Promoter and transcription factor analyses were done using the following Genomatix Software Suite modules (Munich, Germany): Genes2Promoter, RegionMiner, and MatInspector. Briefly, the results from ChipInspector were loaded into the Genes2Promoter program to identify networks and pathways. RegionMiner was used to identify transcription factor module enrichment from the list of differentially expressed genes and MatInspector was used to inspect the relevant promoter regions.
Twenty differentially expressed transcripts (p<0.05) were validated by qRT-PCR using the LightCyclerR 480 Rapid Thermal Cycler System (Roche, Indianapolis, IN). The SPIA-amplified cDNA generated for the Microarrays was used as template for all qRT-PCR experiments. PCR reactions were performed in a 25 μL volume consisting of 0.5 μmol/L primers (Table 1), Real-Time SYBR® Green PCR Master Mix (SuperArray Bioscience, Frederick, MD), and cDNA as per the manufacturer’s instructions. Published methods were used for all assays (Muñoz et al., 2007). See Supporting Information for further details.
All immunohistochemistry protocols were performed as previously described (Ochocinska & Hitchcock, 2009). Sections were stained with the following antisera: mouse monoclonal anti-Rho (Rho 4D2; Dr. Robert Molday, University of British Columbia, Vancouver, Canada) diluted 1:1000; rabbit polyclonal anti-M-opsin and antiS-opsin (Dr. Cheryl Craft, University of Southern California, Los Angeles), diluted 1:1000. Secondary antisera included goat anti-rabbit Alexa Fluor® 488 and goat anti-mouse Alexa Fluor® 555 (Invitrogen, Eugene, OR), diluted 1:200. The stained tissue sections were visualized under a LSM 510 META Laser Scanning Confocal Microscope (Carl Zeiss, Maple Grove, MN). See Supporting Information for further details.
BrdU (Sigma, St. Louis, MO) was injected intraperitoneally (0.1mg BrdU/g body wt) at postnatal day 0 (P0) and at two months of age. The animals were sacrificed 4 hours post-injection. Three animals per genotype were analyzed (cKO, N=3; CON, N=3) at both ages.
Retina tissue was homogenized and 100 μg sample of retinal protein was separated by SDS-polyacrylamide gel electrophoresis (PAGE). Blots were incubated with a 1:5000 dilution of affinity-purified Rho 4D2 antiserum overnight at 4°C, followed by a 1- to 2-hour incubation with secondary Alexa Fluor® 680 goat anti-mouse IgG, diluted 1:15,000. Immunoreactive proteins were detected using the Odyssey Infrared Imaging System (Li-Cor; Lincoln, NE) and analyzed by video densitometry (Lynx software; Applied Imaging, Newcastle-upon-Tyne, UK). For semiquantitative analysis, blots were stripped and reprobed with a monoclonal antibody against beta-actin (Sigma, St. Louis, MO), diluted 1:1000, and Alexa Fluor® 800 goat anti-mouse IgG secondary antibody, diluted 1:5000. See Supporting Information for further details.
Pineal glands were dissected and fixed (1 h at 4°C in 2.5% glutaraldehyde, made in 0.1M sodium cacodylate buffer, pH 7.4). Eyes were dissected and a slit was made in the superior cornea prior to fixation (2 h at 4°C in fixation mix). After the 2 h primary incubation, the superior cornea was cut away, the lens was removed and fixation was continued at 4°C overnight. Tissues were embedded and polymerized in 100% resin for 18 hours in a lab oven set at 70°C. Thin sections (50nm thick) were cut on a Reichert-Jung Ultracut-E ultramicrotome (Leica, Wetzlar, Germany) and collected on LuxFilm grids (Ted Pella, Inc., Redding, CA, Inc.; 30nm film thickness). The grids were post-stained with uranyl acetate and lead citrate and examined in a Tecnai G2 transmission electron microscope operating at 80kV (FEI, Hillsboro, OR). See Supporting Information for further details.
Log2 transformed microarray data from control and NeuroD1 cKO groups, at ZT6 and ZT20, were compared using two-way ANOVA on GraphPad Prism V4 (GraphPad software). The statistical significance of the microarray, qRT-PCR, and Western blot results was determined by a Student’s two-tailed t-test with Welch’s correction (Armitage 2002). ERG results were compared using two-way ANOVA on GraphPad Prism V4 (GraphPad software) with p-values of < 0.05 considered statistically significant.
To characterize the retinal morphology of NeuroD1 cKO mice, retinal sections from three animals per genotype were analyzed (cKO, N=3; CON, N=3). Histological analysis revealed that the outer nuclear layer (ONL) in the NeuroD1 cKO retina is present at two months of age, although the observed thickness varies (Fig. 1B, middle panels). However, the ONL is absent at four months of age (Fig. 1B, right panels). The inner nuclear layer (INL), inner plexiform layer (IPL), and ganglion cell layer (GCL) are not affected in the NeuroD1 cKO retina based on histological evaluation via light microscopy (Fig. 1B) and transmission electron microscopy, comparing the cellular structures in all three retinal layers, including the ganglion cell layer, inner nuclear layer, and outer nuclear layer (data not shown). Specifically, we examined the morphology of the cells within the respective layers, the organelles within the cells including mitochondria and endoplasmic reticulum and the relationship between respective cells. Due to the absence of the ONL by four months of age, subsequent evaluation of cone and rod photoreceptors was limited to two-month-old retinas.
Cone receptors were examined using antisera against cell type-specific markers M-opsin (Opn1mw) and S-opsin (Opn1sw). M-opsin antisera labeled the outer segments of medium wavelength cone photoreceptors in control retinas. In contrast, in the cKO retina M-opsin immunoreactivity was nearly undetectable (compare Fig. 2A and Fig. 2B). The expression of M-opsin mRNA, was reduced by 80% in the cKO retina compared to controls (Fig. 2C). S-opsin antisera labeled cone outer segments in two-month-old control and NeuroD1 cKO retinas. However, the outer segments labeled with S-opsin were disorganized; in addition, the S-opsin protein was also mislocalized to the cell membrane and synapse (compare Fig. 2D and Fig. 2E). The expression level of S-opsin transcript, in the cKO and control retinas was similar (Fig. 2F).
Rods were examined at two months of age using antisera against the rod-specific marker Rho. Retinal sections from three animals per genotype were analyzed (cKO, N=3; CON, N=3). The control retina showed a strong Rho signal (Fig. 2G). In contrast, the NeuroD1 cKO retina showed a weak Rho signal and the protein appeared to be partially mislocalized to the inner segments (Fig. 2H). The Rho-positive outer segments of the NeuroD1 cKO retina were also shortened and disorganized compared to the controls (Fig. 2H). The red label in the GCL of control and cKO retinas is consistent with the staining pattern of Müller glia end feet.
The immunohistological findings were confirmed by Western blot, which revealed that Rho protein in the two-month-old NeuroD1 cKO retina was significantly reduced relative to control (estimated by dilution analysis to be ~95%), (Fig. 2I). In contrast, Rho mRNA was reduced only 40% (Fig. 2J).
Ultrastructural evaluation of retinas from conditional NeuroD1 cKO and control mice was performed using transmission electron microscopy (TEM). Retinal sections from three animals per genotype were analyzed (cKO, N=3; CON, N=3). No change in the inner nuclear layer and ganglion cell layer was apparent at the ultrastructural level (data not shown). In the NeuroD1 cKO retina, the outer segments were disorganized, shortened in length, and, in some regions, completely absent; the controls, in contrast, showed normal outer segments (compare Fig. 3A and C with Fig. 3B and D). The structural integrity of the outer segment discs appears altered and atypical whorls of membranous material (arrows in Fig. 3D) were present between the outer segments and the retinal pigment epithelium (RPE). This expansion of amorphous membranous material is also illustrated in Fig. 3E and at a higher magnification in Fig. 3F.
To determine if there were any changes in pineal gland morphology between genotypes, pineal gland sections from three animals per genotype were analyzed (cKO, N=3; CON, N=3). In contrast to the striking photoreceptor degeneration phenotype observed in the NeuroD1 cKO retina, there was no obvious change in the morphology of the pineal gland body in NeuroD1 cKO compared to control mice at the light microscopy level (Fig. 4A and Fig. 4B). The pineal stalk was shorter and wider in the NeuroD1 cKO animals compared to controls (compare Fig. 4A with Fig. 4B). However, this observation could not be statistically quantified due to the variability in the pineal stalk morphology, and associated technical limitations.
BrdU was used to label mitotically active cells (Wojtowicz & Kee 2006) with the intention of detecting differences between control and NeuroD1 cKO animals. Injections were performed at postnatal day zero and at two months of age. Brain sections from three animals per genotype and age group were analyzed (cKO, N=3; CON, N=3). BrdU labeling of the control and cKO pineal stalk at both developmental stages was not different (data not shown). This suggests that the pineal stalk phenotype observed in the NeuroD1 cKO animals is not associated with a difference in proliferation.
Ultrastructural evaluation of pineal glands from two-month-old NeuroD1 cKO and control mice was performed using TEM. Pineal gland sections from three animals per genotype were analyzed (cKO, N=3; CON, N=3). The morphology of pinealocytes and blood vessels in both groups was normal (Møller et al. 1978), as judged by examination of the relationship of pinealocytes to each other and to vascular elements and the appearance of cellular components including the nucleus, mitochondria, ER, membranes and vesicles(Fig. 4C and 4D).
The influence of NeuroD1 on the retina and pineal gland transcriptomes was assessed by Affymetrix microarray analysis in two-month-old control and NeuroD1 cKO mice. This revealed that NeuroD1 transcript abundance was reduced by >20-fold in both the retina and pineal gland, confirming the effectiveness of the KO strategy.
The microarray results identified several potential downstream target genes. In the retina, the expression of 55 genes was reduced >2-fold, including genes linked to transcription, phototransduction, and protein folding (Table 2). The most dramatically down-regulated genes included Ankrd33/Panky (7.8-fold), and Agr2 (7.9-fold), which are associated with retinal transcription (Sanuki et al. 2010) and protein folding (Persson et al. 2005), respectively. The following phototransduction-linked genes did not decrease more than 2-fold, including Rho, Gnat1, Rcvrn, Sag, Arr3, Pde6, and Grk1.
In addition to down-regulated genes, it was observed that some genes were upregulated in the retina. The most highly up-regulated gene in the retina was the T cell marker Tal2 (11.8-fold); others include Gfap (5.4-fold), Prtg (5.7-fold), Edn2 (7.4-fold) and A2m (7.6-fold), all of which are associated with the immune response (Bucher et al. 2000, Kumar & Shamsuddin 2012, Takahashi et al. 2010, Rattner & Nathans 2005, Fan et al. 2010).
In the NeuroD1 cKO pineal gland, the expression of 16 genes was reduced > 2-fold (Table 2), including genes linked to transcription and calcium signaling. The most dramatically down-regulated genes included Agr2 (2.2-fold), S100a8 (3.3-fold) and S100a9 (3.1-fold). The most up-regulated gene in the pineal gland was Igl-V1 (9.1-fold), a member of the immunoglobulin family linked to the immune response (Das et al. 2011).
A set of 32 genes exhibited a night/day difference in expression in the control retina; 10 of these genes were also rhythmic in the NeuroD1 cKO retina. Eleven genes that were rhythmic in the NeuroD1 cKO retina did not show daily variations in the control retina at the time points studied (Table 3). In the control pineal gland, a set of 38 genes exhibited differential night/day expression; 28 of these genes also exhibited daily rhythms in the NeuroD1 cKO pineal gland. In addition, 14 genes were only rhythmic in the NeuroD1 cKO pineal gland (Table 3).
Microarray gene results were validated by qRT-PCR (Table 2). The transcripts examined were those that were most strongly impacted by the NeuroD1 deletion. In all cases, the results from qRT-PCR analysis qualitatively confirmed the changes observed. For some transcripts, the differences in expression detected by qRT-PCR were >4-fold greater than those revealed by microarray analysis.
The list of differentially regulated genes in the NeuroD1 cKO retina and pineal gland, shown in Table 2, was analyzed using the Genomatix Pathway System (GePS). Transcription factor binding sites were determined using RegionMiner, in Genomatix, which compares the promoter regions of the affected group of genes to promoters from the entire genome (Z-score >2, p<0.05). NeuroD1 binding sites were present in the promoters of several photoreceptor transduction genes including Opn1mw, Aipl1, Agr2, and Ankrd33 in the retina (Fig. 5A) and S100a8, S100a9, Per3, and Agr2 in the pineal gland (Fig. 5B). NeuroD1 dimerizes with its binding partner E47, a bHLH transcription factor, and forms a heterodimer NeuroD1/E47 to translocate to the nucleus and regulate transcription of target genes (Mehmood et al. 2011, Mehmood et al. 2009, Longo et al. 2008). Promoter analysis of the most affected genes using the MatInspector platform revealed that all of the down-regulated genes in the retina contain potential binding sites for the heterodimer NeuroD1/E47 (Fig. 5A). Only two down-regulated genes in the pineal gland, i.e. Agr2 and S100a9, contain putative binding sites for the NeuroD1/E47 complex (Fig. 5B).
The list of genes with altered rhythmic expression on a night/day basis in the retina and pineal gland (Table 3) were analyzed using Genomatix software. To determine whether there was enrichment in specific transcription factor modules which included the NeuroD1/E47 binding site we used the transcription factor overrepresentation search in Genomatix. This analysis revealed that several transcription factor modules containing NeuroD1 were overrepresented in the genes with altered rhythms in both the retina and pineal gland with Z-scores > 4.0 (Table S1). The most overrepresented modules included the Krüppel-like transcription factor family in the retina (Z-score = 8.1) and the GATA binding factor family in the pineal gland (Z-score = 5.0).
The role of NeuroD1 in visual physiology was analyzed in NeuroD1 cKO mice by electoretinography. Rod- and cone-mediated ERGs were recorded to evaluate retinal function in NeuroD1 cKO mice. Four animals per genotype were analyzed (cKO, N=4; CON, N=4) at two and at four months of age. At two months of age, cKO mice showed significant reductions in rod-driven ERG responses compared to controls (Fig. 6A and Fig. 6C; *p<0.001). Cone ERG responses in cKO mice were reduced to undetectable levels at two months of age (Fig. 6B and Fig. 6D; *p<0.001). Rod and cone ERG responses in the cKO retina were undetectable at four months of age (data not shown).
The finding that selective knockout of NeuroD1 decreases the abundance of Agr2 transcripts raised the possibility that Agr2 deletion would cause retinal damage. To test this hypothesis, we examined rod and cone electroretinogram responses in Agr2−/− (KO) and control (CON) mice. Four animals per genotype were analyzed (KO, N=4; CON, N=4) at two and four months of age. Agr2−/− mice displayed normal rod and cone ERGs at both ages (Fig. 7A and Fig. 7B). Significant differences between the average rod and cone-driven b-wave amplitudes were not detected at these ages (Fig. 7C and Fig. 7D) or at six months of age (data not shown).
This is the first report on the effect of tissue-specific NeuroD1 knockout in the adult mouse retina and pineal gland. Analysis of NeuroD1 cKO retina and pineal gland revealed a marked reduction in NeuroD1 transcript and severely compromised photoreceptor function. This confirmed that the targeted knockout strategy was effective and establishes a useful model for studies of NeuroD1. The results of this study are discussed below from retinal and pineal perspectives, sequentially.
The present study extends our understanding of the sequence of events involved in retinal degeneration induced by NeuroD1 deletion, which we view as a two stage process (Table 4). The first stage is characterized by marked reduction of photodetection as indicated by electroretinography, by changes in outer segments, and by a profound decrease in Rho. However, at this stage, the photoreceptor cell bodies appear to be histologically intact. In agreement, photodetector marker genes remain strongly expressed including transduction genes (e.g., Rho, Gnat1, Rcvrn, Sag, Arr3, Pde6, and Grk1) and transcription factor genes (e.g., Crx, Nrl, and Otx2). Moreover, the expression of thousands of other genes is also normal. This supports the interpretation that at this stage the cell bodies are not globally impacted by the loss of NeuroD1. However, a notable outstanding change observed is a 95% reduction in Rho. The second stage of photoreceptor deterioration is characterized by absence of photoreceptors, as observed by four months of age.
The contrast between the marked change in ERG and the minor change in gene expression at two months suggests that one or all of the small number of genes that are most affected at this time play a critical role in the loss of photodetection and in disruption of outer segments without marked destruction of cell bodies. This group includes Aipl1, Ankyrd33/Panky, and Agr2. We suspect that the reduction in expression of one or more of these genes and their encoded proteins is critical for the initial deterioration of outer segments. Furthermore the presence of putative binding sites in the promoters of these affected genes suggests that NeuroD1 acts directly on these genes.
Aipl1 has been linked to Leber’s congenital amarousis, one of the earliest onset and the most severe forms of inherited retinopathy in humans (Ramamurthy et al. 2004). Experimental deletion of Aipl1 causes a retinal phenotype remarkably similar to the one described in the present study: compromised ERGs, altered structural integrity of the outer segment and the RPE boundary, and photoreceptor-specific degeneration (Ramamurthy et al. 2004, Dyer et al. 2004). Hence, the NeuroD1-Aipl1 link established in this study is of special interest because it raises the possibility that retinal pathologies could reflect a reduction in NeuroD1-dependent gene expression, leading to a reduction in Aipl1 expression, which leads to loss of phototransduction.
A second factor that might contribute to the observed effects of NeuroD1 cKO is Ankyrd33/Panky, which encodes a transcriptional cofactor that suppresses Crx-dependent photoreceptor genes (Sanuki et al. 2010). Accordingly, down-regulation of Ankrd33/Panky may remove suppression of a gene that has a negative influence on photoreceptor biology. An Ankyrd33/Panky knockout mouse model is not available but our expression data, as well as other gene profiling data, suggest that Ankyrd33 is a good retinal degeneration candidate gene (Geisert et al. 2009, Sanuki et al. 2010).
Decreased expression of Agr2 may also contribute to the loss of photoreceptor function through an effect on protein folding. This gene encodes a protein disulfide isomerase (Park et al. 2009, Persson et al. 2005); members of this family insure proper disulfide bonds are formed in newly synthesized proteins. Proper formation of disulfides is required for normal protein function; an example from the retina literature is Rho (McKibbin et al. 2007), misfolding of which prevents membrane insertion and proper function. Accordingly, it is possible that the decrease in Agr2 transcripts contributes to the observed effects because substrates of this enzyme, perhaps Rho, are incorrectly folded. However, the finding that Agr2−/− animals have normal ERG responses and relatively normal photoreceptor morphology indicates that loss of Agr2 alone does not explain the observed NeuroD1 cKO phenotype. Accordingly, it seems valuable to consider that Agr2 might contribute to photoreceptor physiology and that photoreceptor loss in the NeuroD1 cKO mouse reflects the impact of the partially reduced expression of it together with that of Aipl1 and Ankyrd33/Panky.
The observations in this study are consistent with the interpretation that the gradual loss of photoreceptor cell bodies may involve an immune response in which the initial deterioration of outer segments leads to destruction of photoreceptor cell bodies (Whitcup et al. 1998). This is consistent with the increased expression of genes associated with the immune response including Gfap, Prtg, Edn2, A2m, and Tal2. The most highly up-regulated (>11-fold) gene is T-cell acute lymphocytic leukemia 2 (Tal2), that encodes a bHLH transcription factor which functions as an oncogene whose activation can lead to T-cell leukemia, a cancer of the immune system (Bucher et al. 2000, Xia et al. 1994, Wadman et al. 1994).
NeuroD1 also plays a role in cone photoreceptor specification via regulation of thyroid hormone receptor beta2 (TRbeta2) expression (Liu et al. 2008). We found similar effects on cone photoreceptor specification but did not observe a change in expression of TRbeta2 at two months of age. However, the study by Liu et al., (2008) involved embryonic total NeuroD1 knockout and evaluated early postnatal retinal explants; in contrast, in the present study NeuroD1 was deleted late in gestation and adult tissue was evaluated. Therefore, it is possible that the differing results in the two studies reflect differences in the timing of the knockout and in the developmental stages examined.
In contrast to the retina, deletion of NeuroD1 did not cause obvious pineal gland morphological effects, although shorter and wider pineal stalks were consistently observed. Deletion of NeuroD1 did not affect expression of phototransduction- or melatonin synthesis-related genes in the pineal gland, consistent with previous findings in neonatal mice (Muñoz et al. 2007). This suggests that NeuroD1 is not an essential regulator of the melatonin synthesis pathway, but does not eliminate the possibility that it is involved in modulating other aspects of pineal biology, including transcription and calcium signaling, processes which are linked to S100a8 and S100a9 (Ghavami et al. 2009); expression of which was suppressed in the NeuroD1 KO pineal gland.
Comparison of the results of the current study with the previous study on NeuroD1 deletion in neonatal mice reveals notable differences. Total knockout of NeuroD1 results in increased expression of Kif5c and Gad1 (Muñoz et al. 2007) in the neonatal pineal gland; this was not observed in the current study of the adult pineal gland in which the gene deletion occurs late in gestation, when Crx is first expressed. Accordingly the differences may reflect differences in the timing of NeuroD1 deletion and in the ages studied. It is likewise reasonable to suspect that NeuroD1 may play different roles during pineal development because NeuroD1 may control transcription in concert with other factors (Cherry et al. 2011); different combinations of which in the neonatal versus the adult pineal gland could have differential effects on gene expression.
Day-night analyses of the mouse pineal transcriptome have been previously published (Rovsing et al. 2011). Those analyses revealed significant differences in the abundance of rhythmic genes between species: In contrast to >600 genes found to be rhythmic in the rat pineal gland (Bailey et al. 2009), only 51 genes showed a day/night rhythm in 129sv mice (Rovsing et al. 2011). We compared the list of 51 genes that were rhythmically expressed in 129sv mice (Rovsing et al., 2011) with the list of 38 genes that were rhythmic in the NeuroD1 control pineal glands, which are on a C57BL/6J background. Of the 51 genes that were rhythmically expressed on a day-night basis in the 129sv control pineal glands only 15 were also rhythmic in the C57BL/6J; these included melatonin and signal transduction genes Aanat, Gng4, E2f8, Cpm, Nr1d1, and Nrxn3 (Rovsing et al., 2011). This indicates that there are notable strain differences in rhythmic gene expression in this tissue and perhaps others.
The findings of this report confirm that NeuroD1 is required for photoreceptor function and indicate that it may act through a small set of genes to promote retinal survival, specifically through the maintenance of photoreceptor outer segments, loss of which results in photoreceptor degeneration. NeuroD1 may also play a role in pinealocyte homeostasis. The advances from our study are summarized in Fig. 8, which describes the expression of NeuroD1 and its downstream targets in the photoreceptor and pinealocyte lineages. The absence of NeuroD1 and its downstream targets results in deterioration of outer segments and leads to the progressive degeneration of photoreceptor cell bodies (Table 5). The NeuroD1-Aipl1 link established in this study raises the possibility that retinal pathologies could in part reflect a reduction in NeuroD1-dependent gene expression, which in turn could suppress expression of Aipl1 and other genes required for photoreceptor homeostasis.
This research was supported by the Intramural Research Program of the National Institute of Child Health, NIH grants R01EY004864 and P30EY006360, and an unrestricted departmental grant from Research to Prevent Blindness, Inc. (RPB). PMI is a recipient of a Senior Scientific Investigator Award from RPB. We would like to thank members of the Klein laboratory, Dr. Morten Møller and Dr. James Russell for many helpful discussions, Dr. Cheryl Craft for the gift of S-opsin and M-opsin antisera, Dr. Robert Molday for the gift of the Rho (Rho 4D2) antiserum, Dr. David J. Erle for the gift of Agr2+/− mice, Chip Dye (NICHD Microscopy and Imaging Core) for his help with transmission electron microscopy, Dr. Anand Swaroop and members of the Neurobiology-Neurodegeneration and Repair Laboratory and Dr. Haohua Qian in the NEI Visual Function Core for assistance with electroretinography, and Daniel Abebe for his veterinary support.
Any Conflict of Interest: None