Cone-rod homeobox CRX controls presynaptic active zone formation in photoreceptors of mammalian retina

Abstract

In the mammalian retina, rod and cone photoreceptors transmit the visual information to bipolar neurons through highly specialized ribbon synapses. We have limited understanding of regulatory pathways that guide morphogenesis and organization of photoreceptor presynaptic architecture in the developing retina. While neural retina leucine zipper (NRL) transcription factor determines rod cell fate and function, cone-rod homeobox (CRX) controls the expression of both rod- and cone-specific genes and is critical for terminal differentiation of photoreceptors. A comprehensive immunohistochemical evaluation of Crx^−/− (null), Crx^Rip/+ and Crx^Rip/Rip (models of dominant congenital blindness) mouse retinas revealed abnormal photoreceptor synapses, with atypical ribbon shape, number and length. Integrated analysis of retinal transcriptomes of Crx-mutants with CRX- and NRL-ChIP-Seq data identified a subset of differentially expressed CRX target genes that encode presynaptic proteins associated with the cytomatrix active zone (CAZ) and synaptic vesicles. Immunohistochemistry of Crx-mutant retina validated aberrant expression of REEP6, PSD95, MPP4, UNC119, UNC13, RGS7 and RGS11, with some reduction in Ribeye and no significant change in immunostaining of RIMS1, RIMS2, Bassoon and Pikachurin. Our studies demonstrate that CRX controls the establishment of CAZ and anchoring of ribbons, but not the formation of ribbon itself, in photoreceptor presynaptic terminals.

Introduction

Evolution of the mammalian nervous system necessitates diverse and highly specialized neurons, which form intricate yet precise synaptic connections to establish neural circuitry that is critical for complex higher-order functions (1). Fate specification of discrete neuronal subtypes and differentiation of their morphology and function are governed by intrinsic transcriptional regulatory programs in concert with extrinsic cues (2,3). Combinatorial and coordinated actions of a small set of regulatory factors can direct the generation of multiple distinct neurons and glia from neural stem cells or from multipotent progenitors. Dynamic yet stringently controlled patterns of gene expression within each cell type can then produce requisite morphology and highly precise synapse organization that is necessary for carrying out assigned functions. We have limited understanding of regulatory networks that control synaptic morphology of specific neurons.

The retina is a highly stratified and easily accessible sensory tissue with five major classes of neurons that can be categorized in as many as 100 distinct cell types of unique morphologies and function (4,5), making it an attractive model to dissect molecular control of neuronal diversity and connectivity (6). The unique organization of the neural retina in three cellular strata, interspersed by synaptic (plexiform) layers, is designed to maximize the capture, integration, processing and transmission of visual information. Rod and cone photoreceptors in the outer nuclear layer (ONL) initiate phototransduction by capturing photons. Rods contain the visual pigment rhodopsin and allow vision in low-light conditions, whereas cones mediate daylight and color vision via distinct cone opsin visual pigments (7). Functional distinctions between the rods and the cones are reflected in their distinctive morphology, both in the outer segments and synapses and gene expression patterns.

The presynaptic terminals of neurons possess unique membrane and cytoskeletal architecture for neurotransmitter release and synapse formation for specific functions (8); the specialized cytomatrix associated with the active zone (CAZ) includes proteins that participate in vesicle transport and membrane fusion (9). The rod and cone photoreceptors form synapses with bipolar neurons and horizontal cells in the outer plexiform layer (OPL) of the retina. A hallmark of these photoreceptor presynaptic terminals is the presence of synaptic ribbons, which project from the arciform density at the active zone as dense plate-like scaffolds that provide large surface area for attachment of hundreds of vesicles (10–12). Unlike the conventional synapses in most neurons that transmit signals in an all-or-none manner via action potentials, the synaptic ribbon in sensory neurons enables tonic (continuous) and graded release of the neurotransmitter glutamate over a wide dynamic range in membrane potential (13). In mammals, most rod photoreceptors have a single long ribbon whereas cones possess multiple shorter ribbons. Conventional and ribbon synapses contain many common proteins, such as Piccolo, Bassoon and RIMS; however, the synaptic ribbon comprises of additional components including the core structural protein Ribeye, which associates with numerous vesicles and active zone proteins (12).

Rod and cone photoreceptor fate is specified by complex interplay of several key transcription factors (14–16). OTX2 is expressed in retinal progenitors and activates Crx and Nrl, two key transcriptional regulators of photoreceptor development (17–19). NRL is essential for rod differentiation (20); however, CRX acquires the role of the primary regulator by functioning synergistically with multiple transcription factors including NRL to control gene expression in both rod and cone photoreceptors (21–23). Mutations in CRX cause multiple distinct retinal degeneration phenotypes in humans (24,25). In mice, the loss of Crx (Crx^−/−) severely compromises photoreceptor development without altering cell fate (26). Additional mouse models have provided insights into CRX-associated human retinopathies (27,28). Crx^Rip mutant shows retina with immature photoreceptors (19), revealing mechanistic similarities to dominant negative CRX mutations that cause Leber congenital amaurosis (29). Interestingly, photoreceptors are preserved for a longer period in the Crx^Rip/+ retina compared to the Crx^Rip/Rip or Crx^−/− mice, providing an opportunity for examination of developmental defects (19).

To elucidate the role of CRX in morphogenesis of photoreceptor presynaptic structure and delineate underlying regulatory mechanisms, we performed detailed characterization of developing retina in Crx^Rip/+, Crx^Rip/Rip and Crx^−/− mouse mutants and compared their relevant gene expression profiles with that of Nrl^−/− retina. Our studies unravel a Crx-dependent gene regulatory network involving presynaptic genes that encode cytomatrix-associated proteins governing the organization of the photoreceptor ribbon synapse but not the formation of ribbon structure.

Results

Aberrant photoreceptor ribbon synapses in Crx mutant retina

We performed retinal histology analysis of Crx mutants to examine synaptic layers. The retinas of postnatal day (P)21 Crx^Rip/+, Crx^Rip/Rip and Crx^−/− mice were considerably thinner with almost complete absence of outer segments in photoreceptors (Fig. 1A). Quantitative analysis revealed a decrease in ONL thickness in Crx^Rip/Rip and Crx^−/− retina, with relative preservation in the Crx^Rip/+ retina (Fig. 1B). The long-term preservation of the ONL in the heterozygous Crx^Rip mutant has been documented in even 18-month-old animals (19). OPL was somewhat reduced in Crx^Rip/Rip retina, whereas inner plexiform layer (IPL) was largely unaffected in all Crx mutants despite the reported CRX expression, though at low levels, in bipolar cells (30).

Figure 1.

Histological examination of Crx mutant retina. (A) H&E staining of methacrylate sections at P21 of Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− mouse retinas. Scale bar: 50 μm for all images. (B) The thickness comparison of individual retinal layers from Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− mice measured at 1 mm away from the optic nerve head. The measurements were quantified using three retinal sections from three independent mice of each genotype. * P-value ≤ 0.05, ** P-value≤ 0.01, *** P-value ≤ 0.001 and **** P-value ≤ 0.0001. Abbreviations: gcl, ganglion cell layer; is/os, inner segment/outer segment of the photoreceptor; rpe, retinal pigmented epithelium.

We then focused on synapses between photoreceptors and inner retina neurons. Anti-Ribeye antibody detected synaptic ribbons in the OPL of wild-type Crx^+/+ retina as early as P5, and the ribbons were present close to dendritic processes of rod bipolar cells by P8 (Fig. 2), as reported earlier (31). The horseshoe shape of ribbons was clearly evident at P14 within the OPL, and rod and cone ribbons were separate into distinct sublamina in P21 wild-type retina. However, the OPL of P21 Crx^Rip/Rip and Crx^−/− retina exhibited no distinct separation of rod and cone synaptic levels, and Ribeye-positive synapses were reduced near the PKCα (a marker of rod bipolar cells)-positive dendritic tips (Fig. 2). Crx^Rip/+ retina was less aberrant with somewhat less reduction in Ribeye staining. In the wild-type P21 retina, immunostaining of calbindin (shown in green) and ribeye (shown in red) suggested the establishment of connections between horizontal cell dendrites and photoreceptor presynaptic terminals (Fig. 3). However, all Crx mutant retinas revealed reduced number of ribbons (Fig. 3). Notably, the second-order neurons did not show a widespread sprouting in the mutant Crx retina (except in P21 Crx^−/− retina shown in Fig. 2), as evident in retinal remodeling during aging or degeneration (32,33).

Figure 2.

Spatial and temporal development of photoreceptor synapses in Crx mutant retina. Photoreceptor ribbons (identified by Ribeye antibody, shown in red) and rod bipolar cells (marked by PKCα, in green) were stained at different ages ranging from P5 to P21. Abbreviation: inl, inner nuclear layer. Scale bar: 10 μm.

Figure 3.

Synaptic connection between photoreceptors and horizontal cells in Crx mutant retina. Photoreceptor ribbons (identified by Ribeye antibody, shown in red) and horizontal cells (marked by Calbindin, in green) in P21 Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinas. Scale bar: 10 μm.

Ribbon structure defects in Crx mutant photoreceptors

To get further insights into the ribbon structure in Crx mutants, we examined Ribeye-immunostaining in P21 retina. The presynaptic ribbons were observed as fragmented or punctuated in Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinas (Fig. 4A upper panel) compared to the typical horseshoe-shaped ribbons in Crx^+/+ controls. The average ribbon length was also reduced in all three Crx mutants, as revealed by whole mount immunostaining of P21 retina (Fig. 4A upper panel and Fig. 4B). Transmission Electron Microscopy (TEM) revealed fewer ribbons in the Crx^Rip/Rip retina compared to the control (Fig. 4A lower panel and Fig. 4C). It should be noted that the appearance of wider ribbons in whole mount retina (Fig. 4A upper panel) is likely because multiple shorter ribbons are so close to one another that they cannot be clearly discriminated in ribeye immune-preparations. The photoreceptor ribbon is anchored to the presynaptic membrane, adjacent to the postsynaptic horizontal and/or bipolar cell processes. Interestingly, all Crx mutants showed dramatically less ribbons anchored at the active zone (Fig. 4D). A majority of photoreceptor presynaptic terminals displayed multiple ribbons in all Crx mutants (Fig. 4E). Thus, the photoreceptors in the three Crx mutants studied here revealed major defects in synaptic terminals, including altered number, shape, length and anchoring at the CAZ.

Figure 4.

Synaptic ribbons in the OPL of Crx mutant retina. (A) Ribeye (shown in red) immunostaining in the OPL of P21 Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− whole mount retina (upper panel), observed by confocal microscopy. The lower panel shows transmission electron micrographs (TEM) of P21 Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− photoreceptor synaptic terminals (yellow lined). Red arrowheads indicate anchored ribbons, and blue open arrowheads point to floating ribbons. (B) Ribbon length (μm) in the OPL of Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinas, measured from 50 ribbons of each genotype in three independent mutant mice. (C) Ribbon number comparison in Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinal OPL. Total OPL ribbon numbers were counted from 100 μm of TEM. (D) The proportion of anchored ribbons in the total number of observed ribbons of Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− OPL. (E) Comparison of the number of terminals per 100 μm OPL length with one, two and more than two ribbons, measured from of Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinas. * P-value ≤ 0.05, ** P-value ≤ 0.01, *** P-value ≤ 0.001 and **** P-value ≤ 0.0001. Scale bar: 500 nm (A, upper panel) and 5 μm (A, lower panel). For C to E: 6, 7, 10 and 10 fields were used for counting from retina of three independent Crx^+/+, Crx^Rip/+, Crx^Rip/Rip and Crx^−/− mice, respectively.

Regulation of synapse-associated transcripts by CRX

In order to dissect molecular mechanism(s) underlying photoreceptor presynaptic defects, we analyzed previously published transcriptome profiles of P21 wild-type (Crx^+/+) and Crx mutant (Crx^Rip/+, Crx^Rip/Rip and Crx^−/−) whole retina (19), together with expression profile of the Nrl^−/− retina (34) and RNA-seq data from purified developing rod photoreceptors (35,36). We then integrated the transcriptome data with CRX- and NRL-ChIP-seq (23,37) and performed Gene Ontology analysis to identify genes that are largely regulated by CRX and encode synapse-associated proteins (Fig. 5A, Table 1). As predicted, a majority of genes encoding synaptic vesicle and active zone proteins showed a surge in expression during postnatal retinal development. This dramatic enhancement in synapse-associated gene expression after P6 was also observed in developing rod photoreceptors, thereby reflecting the period of synapse morphogenesis in rods (38) that constitute >75% of all cells in the mouse retina. Interestingly, the expression of these genes was high in Nrl^−/− retina as well, indicating their participation in both rod and cone photoreceptor synapse formation. Consistent with these observations, genes encoding postsynaptic proteins demonstrated higher expression only in the whole retina expression data, but not in developing rod photoreceptors.

Figure 5.

RNA-seq and ChIP-seq analysis of whole retina or flow-sorted rod photoreceptors. (A) Transcriptome analysis of synapse-related genes, in order from left to right: developing Crx^+/+ retina from E11 to P21, P21 Crx^Rip mutants, Crx^−/− and Nrl^−/− retinas; followed by CRX and NRL ChIP-Seq analysis from P21 Crx^+/+ retina; and transcriptome analysis of developing rod photoreceptors from P2 to P28. Gene names shown in red, green and blue correspond to genes showing reduced, unchanged and increased expression, respectively, in Crx mutants. (B) Transcriptome analysis of synapse-related genes showing significant change in expression in Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinas compared to the control Crx^+/+ retina. (C). Track view of CRX- and NRL-ChIP-Seq density profiles. The ChIP-Seq peaks on the genes were visualized using the UCSC genome browser. Representative ChIP-Seq density distribution is shown for synaptic related genes Reep6, Cabp4 and Unc119.

Table 1.

Expression changes in photoreceptor synapse-associated proteins of mutant retina revealed by immunostaining

Fourteen synapse-associated genes showed significantly less expression and four revealed higher expression in Crx^Rip/+, Crx^Rip/Rip and Crx^−/− retinas compared to the wild type (indicated in red and blue, respectively, in Fig. 5A). These genes, with the exception of Cast, Unc13b (downregulated) and Sv2a and Trpm1 (upregulated), possess CRX-binding site(s) in their promoter and/or enhancer region (Fig. 5A, third column). Several of the CRX-target genes also included NRL binding sites (Fig. 5C). Most of the genes showing decreased expression in Crx mutants encode proteins associated with synaptic vesicles and the CAZ (Fig. 5A); of these, Reep6, Cabp4 and Rs1 are significantly reduced in all Crx mutants and Mpp4, Rims2, Ano2 and Unc119 in either Crx^Rip/Rip or Crx^−/− retina (Fig. 5B). Rod-specific gene Reep6 (39) was almost undetectable in all three Crx mutants as well as in the Nrl^−/− retina. Taken together, despite extensive functional synergy with NRL, CRX regulates several distinct set of genes during photoreceptor synapse development.

Immunohistochemical analysis of photoreceptor synapse proteins in Crx mutants

We performed immunohistochemical analysis of P21 retina from Crx mutants, and of Nrl^−/− retina for comparison, to validate gene expression changes and examine localization of synapse-related proteins. The antibodies against the following proteins were used for immunostaining: vesicles—REEP6, SNAP25 and SV2; ribbon—Ribeye (Ctbp2), Bassoon, RIMS1; cytomatrix active zone—MPP4, RIMS2, UNC119 and UNC13; synaptic membrane—PSD95; synaptic cleft—RS1 and Pikachurin; postsynaptic—RGS7 and RGS11 (Figs. 6 and 7). Consistent with the transcriptome data, REEP6 expression was reduced or undetectable in all mutant retina, and SNAP25 and SV2 labeling was increased in Crx^Rip/Rip and Crx^−/− retinas (Fig. 6). Expression of RIMS2 was reduced in Crx mutants but not in the Nrl^−/− retina where Crx is still expressed, suggesting its regulation by CRX. PSD95 expression was reduced in all Crx mutants as well as in the Nrl^−/− retina. The CAZ proteins—MPP4, UNC119 and UNC13—were also reduced in all mutants. The ribbon proteins (Ribeye, Bassoon and RIMS1) and the cleft protein Pikachurin showed a disorganized immunolabeling pattern in the Crx mutant retina. Ribeye and Bassoon immunostaining rarely revealed the typical horseshoe ribbon shape, and Pikachurin labeling revealed tiny dots at the synaptic cleft in all Crx mutants but the staining of Ribeye, Bassoon and Pikachurin in synapse terminals was consistently maintained (Figs. 6 and 7). Additionally, the expression of RGS7 and RGS11, postsynaptic proteins in bipolar neurons, was decreased, probably due to the improper connections with presynaptic partners. Finally, RS1 protein, localized in the photoreceptor synaptic cleft and bipolar dendrites (40), displayed no significant change in expression and localization in Crx mutants (Fig. 7). Therefore, CRX appears to largely regulate the expression of genes associated with vesicle and active zone components, but not those involved in the formation of the ribbon structure per se.

Figure 6.

Expression of photoreceptor presynaptic proteins in Crx mutants. MPP4, Ribeye (Ctbp2), Pikachurin (Egflam), UNC119, UNC13, RIMS2, REEP6, PSD95 (Dlg4), SNAP25 and SV2 were immunolabeled and detected in the OPL of Crx^+/+, Crx^Rip, Crx^−/− and Nrl^−/− retinas. REEP6, PSD95, SV2 and SNAP25 belong to group of vesicle protein, and others to CAZ protein group. Scale bars: 10 μm in Ribeye/MPP4 and Ribeye/PSD95, 5 μm in all others.

Figure 7.

Expression of photoreceptor presynaptic and postsynaptic proteins. Pikachurin, Ribeye, Bassoon, RIMS1, RGS7, RGS11 and RS1 were immunolabeled and detected in the OPL of Crx^+/+, Crx^Rip, Crx^−/− and Nrl^−/− retinas. RGS7 and RGS11 are expressed in post-synaptic buttons. RS1 is detected in presynaptic cleft membrane and bipolar neurons. Others belong to CAZ protein group. Scale bars: 5 μm in all images, except last panel of RS1/Ribeye (10 μm).

Discussion

Almost 90 billion neurons in humans generate 150 trillion synaptic connections (41). Connectomics and the dissection of synaptic wiring are among the most appealing current topics of neurobiology. Retinal photoreceptors are part of the complex neuronal network with unique features including ribbon synapses (42). Next generation sequencing technologies have begun to unravel the intricacies of gene regulatory networks underlying retinal development and disease (43); however, the molecular mechanism(s) underlying the genesis of photoreceptor morphology are poorly delineated. In this context, our study demonstrates a direct and critical role of CRX in the regulation of genes associated with synapse morphogenesis in photoreceptors (Fig. 8).

Figure 8.

Schematic summary of synaptic proteins regulated by CRX. (A) Vertical view of synaptic terminal of the rod photoreceptor with horizontal (H) and bipolar (B) processes. AD; archiform density, V; vesicle. (B) Relationship among synaptic proteins in the terminal active zone. Red and blue background labels on proteins indicate their reduced or increased expression in the OPL presynaptic terminals of the Crx mutants.

Given its crucial role in both rod and cone photoreceptor development, we hypothesized that CRX controls the expression of genes associated with ribbon synapse morphogenesis and function. We took advantage of the available Crx mutant mouse strains and performed a detailed immunohistochemical analysis, focusing initially on the synaptic ribbon within the OPL. The loss of Crx profoundly impaired the development of photoreceptor synapses, with Crx^Rip/Rip and Crx^−/− mutants displaying reduced OPL thickness and less total ribbon number. Floating or multiple ribbons in synaptic terminals of many mutant photore ceptors suggested a defect at the CAZ where the ribbons are anchored. An integrated analysis of gene expression profiles from mutant retina with genome-wide transcription factor binding data validated these observations. We observed that a number of CAZ and calcium channel-linked proteins are encoded by CRX-target genes and demonstrate a dramatically reduced expression in Crx mutants.

Formation of photoreceptor presynaptic terminals requires the assembly of the ribbon with attached synaptic vesicles and an arciform density at the active zone (44). Immunohisto-chemistry as well as RNA-seq data show that most of the genes encoding ribbon proteins (such as Piccolo, RIMS1, KIF3a, ERC2 and Bassoon) demonstrate little or no significant change in expression in Crx mutants. One possible explanation is that synaptic ribbons are present in non-photoreceptor sensory neurons as well and that photoreceptor-specific transcription factors may not be a major regulator of their expression. However, ribbon protein distribution, such as Bassoon and Ribeye, was altered in Crx mutants with the absence of a horseshoe-shaped structure aligned with one another, which explains the observed phenotype of floating ribbons. Notably, floating ribbons are also observed in photoreceptors of Bassoon-knockout retina (45,46). In addition, CAST (an inhibitor of the calcium-dependent cysteine protease, Calpain functioning in membrane fusion) which interacts directly with Bassoon (47) showed somewhat lower expression in Crx mutants (Fig. 5A). Furthermore, the presence of multiple ribbons in Crx mutant photoreceptors and their possibly reduced synaptic connectivity with second-order neurons can be explained by decreased expression of a majority of genes encoding the active zone proteins (e.g. MPP4, CABP4, DLG4, UNC13b, UNC119, RIMS2, CACNA1F and ANO2) as well as synaptic vesicle-proteins (REEP6, SLC17A7 and CPLX4). Interestingly, CABP4, MPP4, RIMS2 and UNC119 are specifically expressed in photoreceptor synapses, but not in the bipolar ribbon synapse (48–51). RIMS2 is a scaffold protein that regulates vesicle priming and docking (52). UNC119 is associated with synaptic vesicles (53), and mouse Unc119^−/− retina shows slow progressive retinal degeneration (54). CABP4 (51) and MPP4 are involved in synaptic calcium release and maintenance (48). Loss of Cabp4 in mice (Cabp4^−/−) results in retinal phenotype with certain similarities with the Crx mutants, including shorter outer segments and no synaptic function, and in humans CABP4 mutations lead to incomplete congenital stationary night blindness (CSNB2B) (51,55,56). CABP4 associates with the C-terminal domain of Ca(V)4.1 alpha-1 (CACNA1F), and mutations in CACNA1F cause CSNB, Åland eye disease and have been reported as the cause of one case of X-linked cone-rod dystrophy (57–62). On the other hand, MPP4 assembles with VELI3-PSD95 (DLG4) and modulates calcium release from the cells (48). Mouse Mpp4^−/− retina exhibits enlarged synaptic terminals and ribbons, which result from intracellular Ca²⁺ accumulation (48,49). Thus, CRX-regulated genes play critical role in photoreceptor synapse formation.

Regenerative medicine and stem cell-based therapies, including photoreceptor cell replacement, are attractive and feasible therapeutic strategies for many retinal and macular degenerative diseases. Formation of accurate and appropriate synaptic connections is crucial for the success of such paradigms. Toward this goal, our studies showing a critical role of CRX in presynaptic terminal morphogenesis unravel additional avenues for examining synaptic plasticity and exploring photoreceptor integration in the degenerating retina.

Materials and Methods

Animals and tissue processing

All experiments with mice followed ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Care and Use Committee of the National Eye Institute (ASP NEI#650). The Crx^Rip/+, Crx^Rip/Rip, Crx^−/−and Nrl^−/− mice on C57BL/6 J background have been described earlier (19,20,26). Crx^−/− mice were generous gift from Dr Connie Cepko. The eyes were enucleated and immediately fixed in 4% paraformaldehyde-phosphate buffered saline (PBS) for 1 hr. The anterior part and lens were removed and embedded into low melting agarose gel (Type XI, Sigma-Aldrich). Vibratome (Leica VT1000S) was used to obtain 100 μm-thick retina sections. For cryosections, both fixed and unfixed eyes were embedded in Tissue-Tek Cryo-OCT compound (Thermo Fisher Scientific), frozen in cold isopentane and kept at −80°C until use. The fixed eyes before OCT compound embedding were incubated in an increasing gradient of PBS-sucrose at 10, 20 and 30%. Retina cryosections were freshly obtained at 14 μm–thickness, using a cryostat Microm HM 550 (Thermo Fisher Scientific) and used up within 1 month.

For hematoxylin and eosin (H&E) staining, methylacrylate sections were used. Briefly, the eyeballs were fixed for 1 hr in 2% glutaraldehyde-PBS overnight before fixation in 4% paraformaldehyde-PBS. Six sections (5 μm thickness) were cut from the eyes in each group through pupillary-optic nerve head plane and stained with H&E.

Immunohistochemistry

Immunostaining of retinal sections and whole mounts was performed as described (36), with minor modifications. Briefly, cryosections were incubated with primary antibodies overnight at 4°C, followed with secondary antibo-dies for 1 hr at room temperature. Retinal vibratome sections and whole mounts were incubated with primary antibodies for 3–5 days and secondary antibodies for 2–3 days after overnight blocking and permeabilization. PBS with 3% bovine serum albumin or 5% goat serum and 0.3% Triton X-100 was used for blocking and antibody treatment. The primary antibodies used in the study are listed in Table 2. Secondary antibodies conjugated with Alexa Fluor 488, 594 or 633 (Life Technologies) were used at the dilution of 1:1,000 with nuclear counterstaining solution, 4,6-diamidino-2-phenylindole (Molecular Probes). Images were acquired using either a Zeiss 700 or Zeiss 780 confocal microscopes (Carl Zeiss Meditec).

Table 2.

Primary antibodies used in this study

Antibody	Source	Dilution
PKC	Thermo Scientific	1:500
	Ribeye or CtBP2	BD Biosciences	1:500
		Calbindin D-28K	Calbiochem	1:1000
MPP4			T. Li (48)	1:1000
	PSD95		T. Li (48)	1:1000
		REEP6	A. Swaroop (70)	1:1000
UNC119			F. Haeseleer (51)	1:200
	Pikachurin		Wako	1:1000
		pan MUNC 13	BD Biosciences	1:1000
RIMS 2			Synaptic Systems	1:200
	SV2		DSHB	1:1000
		Bassoon (D63B6)	Cell Signaling	1:500
RIMS1			Synaptic Systems	1:500
	SNAP25 (SP12)		Abcam	1:1000
		RGS7	Wei Li (71)	1:1000
RGS11			Wei Li (71)	1:1000
	RS1		P. A. Sieving (72)	1:1000

Image analysis

Thickness of retinal layers was measured at the level of 1 mm away from the optic nerve head from more than three sections of at least three different eyes for each phenotype using H&E sections. Imaging J software was used for measurements. Statistical analyses were performed by ANOVA with Post-hoc Dunnett’s test.

Transmission electron microscopy

The eyeballs were fixed in 2.5% glutaraldehyde-PBS and 0.5% osmium tetroxide and then dehydrated and embedded in Spurr’s epoxy resin. Thin sections (90 nm) were double-stained with uranyl acetate and lead citrate and imaged in JEM 1010 transmission electron microscope (JEOL).

Transcription factor ChIP-seq analysis

Previously published NRL and CRX ChIP-seq (Replicate 1) data sets (23,37) were downloaded from the public databases (https://www.nei.nih.gov/intramural/nnrldataresource and http://www.ncbi.nlm.nih.gov/geo/ (GSE20012)) for re-analysis. Sequence reads were aligned to Ensembl reference genome release 89 (63) using Genomatix Software Suite, and the genomic regions of significant enrichment of sequence reads over the background were identified by MACS2 v2.0.10 (64). ChIP score for each gene was then determined as a sum of read counts at the individual MACS2 peaks that lay within the gene body and 10 kb upstream of the gene start site by using bedTools (65). The read counts were normalized to 20 million total mapped reads.

RNA-seq analysis

We re-analyzed the previously published Crx mutant whole retina (GEO accession # GSE52006), flow-sorted photoreceptor transcriptome (GEO accession # GSE74660) data sets (19,35,36) and in vivo mouse retina development data (GEO accession # GSE101986). Transcript quantitation was performed with kallisto v0.43.0 (66) for alignment to Ensembl release 89 transcriptome annotation (63). After alignment, the quantification was summarized at the gene level by using R package tximport v1.4.0 (67). EdgeR was used to apply TMM normalization to the whole data set (68). Heat maps have been generated by custom usage of ggplot2 package in R. The package cowplot (https://CRAN.R-project.org/package=cowplot) was used for aligning heat maps and by plotting them together.