Ankyrin-B and ankyrin-G are required for synaptic integrity in the mouse outer retina

Abstract

Retinal circuit assembly relies on the precise timing and positioning of key molecules between neuronal partners to mediate proper synapse formation. In the outer retina, horizontal cells (HCs) are important interneurons that make the first contacts with photoreceptors and begin to segregate visual information into two distinct pathways by selectively forming synapses to the different types of photoreceptors. Dendrites of HCs synapse exclusively to cone photoreceptors, whereas the axon terminal synapses to rod photoreceptors. Failure to properly form these early connections disrupts the downstream connectivity of other postsynaptic neurons and leads to abnormal visual function. Although these early events are critical for proper synapse development, little is known about how this process is coordinated in the outer retina. In the present study, we performed single-cell RNA sequencing and uncovered members of the cytoskeletal scaffolding family of ankyrins to be differentially expressed in HCs during development. Specifically, we found ankyrin-B to be highly expressed in HCs at early time points and ankyrin-G to be expressed at later stages. Genetic deletion of both ankyrin-B and ankyrin-G disrupts outer retinal synaptic integrity and leads to impaired in vivo retinal responses. In summary, our findings uncovered a new requirement for ankyrin-B and ankyrin-G in maintaining synaptic connectivity in the outer retina essential for normal visual function.

Significance Statement.

In the mammalian retina, the first synapse between photoreceptors and their downstream targets begins to separate visual information into two distinct pathways. During retinal development, photoreceptors first make contacts with horizontal cells (HCs) in a temporal- and spatial-specific manner. Although this initial contact is critical for synaptogenesis, little is known about the key molecules responsible for selective wiring of HCs to photoreceptors. In this study, we performed single-cell RNA sequencing and identified the family of cytoskeletal scaffolding proteins, ankyrins to be differentially expressed in HCs. Loss of ankyrins impairs synaptic connectivity between HCs and their photoreceptor partners leading to abnormal visual responses. Taken together, our work uncovered a new function of ankyrins at photoreceptor synapses.

Introduction

The first synapse in the mammalian outer retina begins to process visual information into two distinct pathways: cone- and rod-driven vision. This is largely due to the distinct connectivity patterns between cone and rod photoreceptors to downstream synaptic targets, such as horizontal cells (HCs) and bipolar cells (BCs). Cone photoreceptors synapse selectively to the dendrites of HCs and cone bipolars to mediate daylight vision and color perception (1). Rod photoreceptors synapse to the axon terminal of HCs and the dendrites of rod bipolars to facilitate night vision or low-light illumination (1) (Fig. 1A). During retinal development, photoreceptors first make contacts with HCs before making connections to BCs. HCs first extend neuronal processes around postnatal day 3 (P3) and these are in close contact with the terminals of cone photoreceptors (2, 3) (Fig. 1B). By P5, HCs extend their axon (4), and from P6 to P8, the axon terminal begins to make contacts with rod photoreceptors (5–7) as shown in Fig. 1B. From P9 to P13, dendrites of bipolars begin to innervate the terminals of photoreceptors and form synapses with their appropriate targets (3) (Fig. 1B). By P21, synapse formation in the outer retina is largely complete.

Figure 1.

HCs undergo morphological changes during key developmental time points (A–E). A, B) Schematic drawing of the adult mouse retina. The retina consists of three nuclear layers (ONL, INL, and GCL) and two synaptic layers (OPL and IPL). A) Cone photoreceptors synapse with the dendrites of HCs and the dendrites of cone bipolars. Rod photoreceptors synapse with the axons of HCs and the dendrites of rod bipolar cells (A). B) HCs first contact cone photoreceptors via their dendrites followed by axon elongation from P3 to P5. By P6–P8, the axon terminal of HCs begins to contact rod photoreceptors. This is followed by innervation of the bipolar dendrites around P9–P13 (B). C) HCs (anti-V5) are individually labeled using the Tg(Cx57icre;MORF) mouse line at distinct developmental time points: P3, P5, and P10. DAPI was used as a nuclear stain in (C). D) Cone photoreceptors are labeled with anti-Sopsin at P3 and with anti-CAR at P5 and P10. Dendrite development of HCs begins with contacts with cone photoreceptors starting at P3. These contacts mature from P5 to P10. Insets are zoomed images of dotted boxed region in (D). E) The axon terminal of HCs matures or extends processes from P5 to P10. Scale bar = 10 μm. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer; OPL, outer plexiform layer; IPL, inner plexiform layer.

This initial contact between photoreceptors and HCs has been shown to be required for synapse formation and maintenance in the mouse outer retina. Studies of Lim1 conditional knockout animals where HCs fail to migrate (8) do not form contacts with photoreceptors and lead to connectivity defects between rods and rod bipolars (9). These findings show HCs are required for the dendrites of rod bipolars to properly innervate and form synapses with rod photoreceptors. Similarly, ablating HCs at later time points after migration also disrupts connectivity between cone photoreceptors and cone bipolars, which ultimately leads to photoreceptor degeneration (10). All together, these data demonstrate HCs are not only required at early stages to form proper synaptic connections but also at later stages to maintain photoreceptor connectivity and survival.

Although HCs play a critical role in circuit assembly, little is known about the molecular mechanisms that mediate selective wiring of HCs to their respective photoreceptor target. Previous studies have shown cell adhesion molecules play a vital role in synapse formation by serving as recognition molecules to mediate interactions between pre- and postsynaptic neurons (11). Over the last decade, several cell adhesion molecules have been identified to mediate photoreceptor connectivity in the outer retina (reviewed in (1, 12)). The expression of these cell adhesion molecules is often restricted to particular cell types or specific subcellular domains to facilitate selective interactions between synaptic partners. For instance, the cell adhesion molecule Elfn1 is highly expressed in rod photoreceptors and mediates connectivity to rod bipolars (13), whereas Elfn2 is found in cone photoreceptors and mediates connectivity in the cone pathway (14). Similar restricted expression of cell adhesion molecules has also been observed within HCs. The cell adhesion molecule Ngl-2 is localized to the axon terminal of HCs, and loss of Ngl-2 leads to connectivity defects between rod photoreceptors and HCs (15). However, to date, Ngl-2 has been the only molecule implicated in selective wiring of HCs to their appropriate photoreceptor partner.

In the present study, we set out to identify new molecules responsible for mediating synaptic connectivity between HCs and photoreceptors. First, we carried out a developmental analysis using a stochastic labeling approach to identify the different subcellular compartments of HCs, such as the dendrites and axon terminal, and assess when they make contacts with the different types of photoreceptors. Second, we performed single-cell RNA sequencing (scRNA-seq) of HCs at early developmental time points when they are actively making contacts with their respective photoreceptor partners. From our sequencing data, we found different members belonging to the cytoskeletal scaffolding ankyrin family to be temporally expressed within HCs. We found ankyrin-B (AnkB) to be highly expressed at early developmental stages in HCs, whereas ankyrin-G (AnkG) is expressed at later stages. Moreover, loss of both AnkB and AnkG disrupts synaptic connectivity between HCs and photoreceptors, leading to impaired in vivo retinal responses in both the rod and cone pathways. In summary, our findings demonstrate a new role for AnkB and AnkG in HC connectivity required for normal visual function.

Materials and methods

All mouse experiments were approved by the Institutional Animal Care and Use Committee at BCM. Eyes were collected at various developmental stages and processed for antibody staining or in situ hybridization, as previously described (16). See Table S4 for the list of reagents. Confocal images were acquired using a Zeiss LSM 800 microscope. Technical replicates from different retinal sections are shown as small circles and biological replicates from different animals as big circles in all graphs. Quantification of presynaptic and postsynaptic marker expression was performed using the Imaris software, as previously described (16, 17). Since CtBP2 protein does not form a puncta-like structure, we measured the maximum fluorescence intensity within the outer plexiform layer (OPL) using ImageJ. Values were normalized to account for surface area variability of the OPL between animals. Statistical tests were performed using the average of the three retinal sections per animal for each group. Eye cups were prepared for serial block-face scanning electron microscopy (SBFSEM) from P30 animals, as described previously (18). Photoreceptor terminals and associated ribbon structures were identified according to known ultrastructural features, as done previously (19, 20). Scotopic and photopic electroretinograms (ERGs) were recorded, as described previously (16, 17). Photopic ERG responses were elicited by using a paired-flash protocol as described (16, 21). All statistical analyses were performed using GraphPad Prism version 9 with more details on the statistical tests and P-values given in the text and figure legends.

Results

Developmental analysis of HC specificity to photoreceptors

To investigate the developmental mechanisms underlying HC connectivity to the different photoreceptors, we sparsely labeled HCs by crossing Tg(Cx57icre) mice (22) with the Tg(MORF) reporter line (4). This strategy allowed us to visualize individual HCs, including their dendrites and axon terminals, across key developmental time points (Fig. 1C–E). Consistent with prior studies (2, 3), we found dendrites of HCs are initially vertical at P3 but then transition to lateral processes from P5 to P10, as shown in Fig. 1C. At P3, dendrites of HCs begin contacting cone photoreceptors, and these interactions continue to P10 (Fig. 1D). Notably, these changes in HC morphology are accompanied by distinct changes in photoreceptor terminal morphology, including a flattening and enlargement at the base of the cone terminal by P10, suggesting the onset of mature synaptic contacts. Our whole-mount retinal images of Tg(Cx57icre;MORF) animals revealed the axon terminal of HCs begins to extend processes starting at P5 and continuing to P10 (Fig. 1E), which is consistent with previous work showing HCs make contacts with rod photoreceptors during these developmental time points. The temporally coordinated morphological changes of HCs led us to hypothesize that dynamic transcriptional changes accompany their structural remodeling and connectivity. To explore this further, we performed scRNA-seq at defined developmental stages to capture the molecular programs driving HC connectivity to photoreceptors.

Transcriptomic analysis of HCs during photoreceptor connectivity

To better understand the molecular mechanisms underlying HC connectivity to photoreceptors, we performed scRNA-seq at postnatal day 6 (P6), P8, and P11. These time points correspond to early (P6), mid (P8), and late (P11) stages of HC-photoreceptor connectivity. We purified HCs via fluorescence-activated cell sorting using Tg(Cx57icre;Ai14) animals. Next, we performed scRNA-seq using the 10× Genomics platform (23) (Fig. 2A). After quality control and removal of non-HCs based on the absence of HC marker expression (i.e. Lhx1, Onecut1, Slc32a, and Pax6), we obtained a total of 11,034 high-quality HCs across all developmental time points (Figs. 2B and S1). Although all cells were grouped into a single cluster using unsupervised-based clustering methods, cells from each age were slightly segregated in Uniform Manifold Approximation and Projection (UMAP) space (Fig. 2C). We then performed differential expression (DE) analysis between the different ages: P6 vs. P8, P8 vs. P11, and P6 vs. P11 (Tables S1–S3). Significant DE genes were identified based on an adjusted P-value threshold of 0.05 and a log₂-fold change ≥1.5. This resulted in 408 genes at P6, 114 genes at P8, and 61 genes at P11 (Fig. 2D). Gene ontology (GO) enrichment analysis was then performed on significant DE genes, which showed differences in the most enriched terms across the different time points (Fig. 2D). The GO term “synapse assembly” was significant at P6 and “positive regulation of neuron projection development” at P8, whereas “negative regulation of locomotion” was found at P11. These GO terms are consistent with our developmental analysis showing changes in HC morphology at early stages (P3–P10).

Figure 2.

Transcriptomic data analysis and developmental trajectories of HCs across different time points (A–G). A) Schematic drawing created with BioRender.com illustrates the process of acquiring scRNA-seq data from HCs. B) Dot expression profile plot of known retinal cell markers: Lhx1, Onecut1, Slc32a1, and Pax6 for HCs, Vsx2 for bipolars, Rbpms for ganglion cells, Pde6a for rods, Arr3 for cones, Slc1a3 for Muller glia, Cd74 for microglia, and Pecam1 for endothelial cells. C) A UMAP visualization of 11,034 HCs with cells colored by age. D) DE and GO enrichment analyses were performed between adjacent ages. The number of significant DE genes between ages was determined using the following criteria: adjusted P-value <0.05 and avg_log₂-fold change >1.5. The number of DE genes at each time point along with the associated GO terms is shown in (D). PCA was performed on 500 randomly sampled cells from each age group (1,500 cells total). E) The top left panel shows cells colored by age, while the top right panel shows the same cells colored by pseudotime, as determined by Slingshot trajectory analysis. The bottom panel shows the distribution of cells from each age group across pseudotime. F) Expression of candidate genes across experimental time points in violin plots. G) A heat map displays the scaled expression of DE genes identified.

We then performed a pseudotime analysis to characterize the developmental trajectory of HCs at these different ages (Fig. 2E). This involved performing a principal component analysis (PCA) of 500 randomly sampled HCs from each time point (1,500 cells total; Fig. 2E). To visualize pseudotime progression, cells were plotted in PCA space and colored by their pseudotime values. The results revealed distinct temporal expression patterns corresponding to different phases of HC development. To analyze gene expression changes along the developmental trajectory, we binned cells based on their pseudotime values. The scaled expression values were visualized as a heat map with genes clustered into the three distinct age groups (Fig. 2G). We found the cell adhesion molecule, Ngl-2, in our dataset, which is known to mediate HC-rod connectivity along with other cell adhesion molecules, such as Cntnap1 and Nfasc (Fig. 2F). Moreover, we also identified different members of the cytoskeletal scaffolding ankyrin family known to scaffold a variety of membrane proteins, including cell adhesion molecules, receptors, and ion channels to mediate synapse formation in other regions of the central nervous system (reviewed in (24)). Our sequencing data uncovered promising candidate genes that may be responsible for the selective wiring of HCs to their respective photoreceptor target.

Distinct spatial and temporal expression of HC-specific candidate genes

We validated our HC sequencing data by analyzing the expression of our candidate genes across the three key developmental stages: P3, P5, and P10. We found Contactin-associated protein-like 1 (Cntnap1) to be localized to the somatodendritic region of HCs starting at P3 and continuing to P10 (Fig. 3A). These findings were consistent with published data showing Cntnap1 is expressed in HCs (25). However, Neurofascin (Nfasc), which is known to mediate rod connectivity (17), is expressed starting at P5 and appears to be confined to the axonal region of HCs (Fig. 3B). Previous research has shown localized expression of cell adhesion molecules mediates selective wiring of HCs to photoreceptors (15). Moreover, ankyrins are adaptor proteins that scaffold a variety of membrane proteins, including cell adhesion molecules to different cellular compartments to mediate synapse formation in other regions of the central nervous system (24). Thus, we set out to investigate the role of ankyrins in mediating HC connectivity during retinal development.

Figure 3.

Validation of candidate genes at key developmental time points (A–F). A, B) Protein expression of the cell adhesion molecules Cntnap1 and Nfasc are shown at P3, P5, and P10. HCs are labeled with anti-calbindin. C, D) AnkB and AnkG mRNA expression was detected via in situ hybridization, and HCs were labeled with anti-calbindin at P3, P5, and P10. Yellow arrows denote AnkB or AnkG mRNA expression within HCs. E) Sparse labeling of HCs at P15 using Tg(Ptf1a-cre;MORF) animals revealed AnkB is found in the dendrites and axon terminal of HCs. F) Similarly, AnkG is also highly expressed in the OPL and colocalizes with the dendrites and axon terminal of HCs. Scale bar = 10 μm.

First, we examined AnkB and AnkG expression in HCs at early developmental stages. By performing in situ hybridization followed by antibody staining, we confirmed AnkB mRNA is expressed in HCs from P3 to P10, as shown in Fig. 3C. However, AnkG mRNA is not detected in HCs until P5–P10 and continues to be expressed at lower levels compared with AnkB (Fig. 3D). At P15, AnkB and AnkG proteins are localized to the synaptic layer of the outer retina or OPL and colocalize with the dendrites and axon terminal of HCs (Fig. 3E and F). By P30, ankyrins continue to be highly expressed in the OPL and showed extensive overlap with processes from HCs and their respective synaptic partners (Fig. S2A–D). These findings indicate ankyrins are not only expressed during key stages of HC connectivity but are also localized to the synaptic layer where HCs form synapses with their respective photoreceptor partners.

Loss of ankyrins leads to HC sprouts and OPL defects

We then performed functional studies by crossing published floxed alleles of AnkB (AnkB^flox/flox) (26) and AnkG (AnkG^flox/flox) (27) to the early retinal progenitor cell driver, Chx10-cre (28), to conditionally delete ankyrins, as germline knockouts are known to be embryonically lethal (29). This resulted in Chx10-cre;AnkB^flox/flox and Chx10-cre;AnkG^flox/flox transgenic animals, which we refer to as AnkB CKO and AnkG CKO, respectively. We confirmed loss of ankyrin protein expression in CKO animals by performing antibody staining as shown in Fig. S2E. Next, we examined AnkB CKO and AnkG CKO for morphological and synaptic defects using available antibodies. By P30, processes from HCs seen with calbindin staining are normally confined to the OPL in wild-type retinas as shown in controls (Fig. 4A), and misprojections or sprouts from HCs into the outer nuclear layer (ONL) are correlated with synaptic defects in diseased and aged mouse retinas (15, 16, 30, 31). We found single knockouts of AnkB or AnkG led to minimal HC sprouts compared with controls (Fig. 4A). Moreover, expression of a known presynaptic marker, anti-CtBP2 localized to both rod and cone terminals, appears unaffected in AnkB CKO and AnkG CKO compared with controls (Fig. 4A). As AnkB and AnkG are both expressed in HCs based on our sequencing data (Fig. 2F), and they are known to be functionally redundant with one another (32), we generated Chx10-cre;AnkB^flox/flox;AnkG^flox/flox double knockout animals, which we refer to as AnkB; AnkG DKO. We confirmed AnkB is still present in AnkG CKO and AnkG in AnkB CKO; however, only both are significantly reduced in the OPL of AnkB; AnkG DKO animals (Fig. S2E). We found disruption of both AnkB and AnkG results in numerous misprojections or sprouts from HCs extending beyond the OPL into the nuclear layer or ONL as depicted by arrows in Fig. 4A. Consistent with misprojections of HCs associated with synaptic defects, we observed a significant decrease in the presynaptic marker, CtBP2 within the OPL of AnkB; AnkG DKO compared with controls (Fig. 4A). These observations were quantified by measuring the total volume of calbindin staining in the ONL and CtBP2 immunofluorescence signal in the OPL (Fig. 4D). Controls and single knockout animals showed minimal calbindin staining in the ONL; however, AnkB; AnkG DKO displayed significant sprouts from HCs with reduced CtBP2 expression in the OPL (Fig. 4A and D). Next, we further examined synaptic changes due to loss of ankyrins. Similar to our CtBP2 findings in Fig. 4A, we found reduced immunofluorescence signal of the presynaptic marker, Bassoon (Bsn) in the OPL of AnkB; AnkG DKO animals compared with controls (Fig. 4B). Bsn is normally expressed in both rod and cone terminals and forms puncta-like structure that is ∼0.6 μm in diameter (16, 17). Using the Imaris confocal software, we quantified the total number of Bsn puncta in the OPL and found an 18% reduction in the OPL of AnkB; AnkG DKO compared with controls (Fig. 4D). In addition, we also examined changes of postsynaptic receptors such as the metabotropic glutamate receptor 6 (mGluR6) that is localized to the dendrites of ON bipolars (33, 34) and found a 28% reduction of mGluR6 in AnkB; AnkG DKO compared with controls (Fig. 4C). These results show reduced presynaptic and postsynaptic immunofluorescence signal in the OPL due to loss of ankyrins.

Figure 4.

Loss of AnkB and AnkG results in HC sprouts and OPL defects (A–C). A) Processes from HCs labeled with anti-calbindin (Calb) are normally confined to the OPL by P30 as shown in controls. Single conditional knockouts of Ankryin-B (AnkB CKO) or ankyrin-G (AnkG CKO) leads to minimal misprojections or sprouts of HCs. However, disruption of both ankyrin-B and ankyrin-G or AnkB; AnkG DKO results in multiple HC sprouts in the ONL as depicted by arrows. Expression of the known presynaptic marker, CtBP2 showed a significant reduction in the OPL in AnkB; AnkG DKO but not in AnkB CKO or AnkG CKO compared with controls. B, C) Protein expression of another presynaptic marker, Bassoon (Bsn) and the postsynaptic marker, mGluR6 is also reduced in the OPL in AnkB; AnkG DKO compared with controls. HCs are labeled with anti-calbindin (Calb). D) HC sprouts were quantified by measuring the amount of calbindin staining in the ONL. CtBP2 immunofluorescence signal in the OPL was measured across all experimental groups. Quantification of Bsn and mGluR6 immunofluorescence signal in the OPL was measured using the Imaris software to automatically detect puncta of about 0.6 μm in size. The averages from three different mice were used for statistical analysis. Data from each animal are represented as mean values ± SEM. Statistical significance determined by an unpaired two-tailed Student's t test. *P < 0.05, **P < 0.01, ***P < 0.001. Scale bar = 10 μm.

To determine if these changes were due to retinal degeneration, we measured the different retinal layers in both control and AnkB; AnkG DKO animals at P30 (Fig. S3A and B). We found a slight reduction in the ONL but not the inner nuclear layer (INL) or ganglion cell layer (GCL) of AnkB; AnkG DKO animals compared with controls. Next, we assessed the role of ankyrins on photoreceptors by using antibodies that label the outer segments and axon terminals of rod and cone photoreceptors. We used anti-CNGA1 and PNA staining to label the outer segments of rods and cones, respectively (Fig. S4A–D). We found no significant difference between the rod outer segments of controls and AnkB; AnkG DKO animals (Fig. S4B). Similarly, there was also no significant difference in the total number of cone outer segments or their volume between controls and AnkB; AnkG DKO animals (Fig. S4D). We also examined the axon terminals of rod photoreceptors (anti-PSD-95) and cone photoreceptors (anti-CAR) as well as the dendrites of rod bipolars (anti-PKC) and cone bipolars (anti-Scgn; Fig. S4E). Our results show there is no retraction of axon terminals from photoreceptors even in regions where there is sprouting from HCs as depicted by white arrows in AnkB; AnkG DKO retinas (Fig. S4E). Similarly, we do not observe sprouting or mispositioning of dendrites from rod bipolars nor cone bipolars due to loss of ankyrins (Fig. S4E). These findings suggest that although there is some loss of photoreceptors, ankyrins are mainly required to maintain processes from HCs into the OPL but not for other neuronal types.

Disruption of synaptic innervation between HCs and photoreceptors

To further examine the synaptic changes associated with loss of ankyrins, we performed SBFSEM on AnkB; AnkG DKO retinas at P30. We found processes from HCs, which are usually confined to the cone synapse or pedicle in wild-type retinas, were extending beyond the cone terminal in AnkB; AnkG DKO, as depicted by white arrows in Fig. 5A. These findings were similar to the HC sprouting phenotype seen with calbindin labeling in AnkB; AnkG DKO retinas in Fig. 4A. Interestingly, dendrites of BCs (cyan) appear disorganized but remain confined to the cone pedicle as depicted by white arrowheads in Fig. 5A. These results were also consistent with our confocal data showing no sprouts from cone bipolars as seen in Fig. S4E. Moreover, we found the ribbons of cone pedicles in AnkB; AnkG DKO were not anchored to the base as in wild type, but rather floating as shown by asterisk in Fig. 5A. Similar floating ribbons have been reported previously in Bassoon mutant mice, which also display synaptic defects in the outer retina (35). We quantified the percentage of ribbons associated with an HC, BC, glial partner, or unidentified (UI) neuronal process (Fig. 5C). We found there is a trend toward reduced contact between cone pedicles and HCs, as well as BCs in the AnkB; AnkG DKO. These data were consistent with the decrease in pre- and postsynaptic immunofluorescence signal in the OPL, as shown in Fig. 4B and C.

Figure 5.

AnkB; AnkG DKO mice display aberrant synaptic innervation as seen with SBFSEM (A–C). A) A 3D reconstruction of a wild type and AnkB; AnkG DKO cone synapse or pedicle with included ribbons, HC partners, and BC partners. B) A 3D reconstruction of a wild type and AnkB; AnkG DKO rod synapse or spherule innervated by HC partners and BC partners. C) Quantification of the number of ribbons per cone pedicle, the ribbon length, the percentage of anchored ribbon and the percentage of ribbons within a pedicle associated with HC partners, BC partners, glial partners, and UI neuronal processes. A total of five cone pedicles were reconstructed from two different AnkB; AnkG DKO animals, and four wild-type/control cone pedicles were used for the analysis. A total of six rod spherules were reconstructed from AnkB; AnkG DKO mice, and six reconstructed wild-type rod spherules were used for the analysis. The percentage of anchored ribbons and those associated with BC partners in cone pedicles was statistically significant between wild type and AnkB; AnkG DKO. In addition, the number of BC partners is significantly reduced for AnkB; AnkG DKO rod spherules compared with controls. Statistical significance was determined by an unpaired two-tailed Student's t test. *P < 0.05, **P < 0.01, ***P < 0.001.

Next, we performed a similar 3D reconstruction analysis of the rod synapse or spherule in animals with loss of ankyrins. Rod terminals typically have one large horseshoe-shaped ribbon or synaptic site (19, 36) as shown by the yellow arrow in Fig. 5B. Our data show that one of the rod terminals from AnkB; AnkG DKO contains multiple small, fragmented ribbons (yellow arrowhead) with one of the HC processes (shown in brown) extending beyond the synaptic layer into the ONL, as depicted by white arrows. In contrast, the other HC process (yellow) appears to properly innervate the rod spherule in Fig. 5B. Similarly, dendrites of BCs (shown in cyan) seem to be localized to the rod spherule and do not extend beyond the synaptic layer in Fig. 5B, which is consistent with our confocal findings shown in Fig. S4E. Quantifications of these observations are shown in Fig. 5C. Moreover, we also measured the total number of synaptic vesicles located at the ribbon of cone and rod synapses from wild type and AnkB; AnkG DKO animals (Fig. S5A and B). Our data revealed a significant decrease in the number of vesicles of both cone and rod synapses (Fig. S5C). These data further support the role of ankyrins in mediating proper synaptic innervation in the outer retina.

Impaired rod- and cone-driven retinal responses in animals with loss of ankyrins

We then addressed whether the synaptic defects seen in AnkB; AnkG DKO result in abnormal retinal responses. To answer this question, we performed in vivo full-field ERG recordings in both dark-adapted and light-adapted mice to measure rod-driven and cone-driven responses, respectively. Anesthetized mice were exposed to different flashes of light to measure rod-driven responses (scotopic ERG) or cone-driven responses (photopic ERG) as described in (16, 17, 21, 37). ERG responses are comprised of an initial a-wave response elicited by photoreceptors, followed by a b-wave response generated by postsynaptic neurons, such as bipolars (37). For scotopic responses, we plotted individual ERG traces elicited from various stimulus intensities from controls and AnkB; AnkG DKO animals (Fig. 6A). We quantified these observations by plotting the a-wave and b-wave amplitudes of different animals from both controls and AnkB; AnkG DKO animals as shown in Fig. 6B. We calculated the maximum rod-driven b-wave responses or B_max and found controls to have a B_max value of 355 μV, whereas AnkB; AnkG DKO have a B_max value of 217 μV. Our data confirm that the rod component of the b-wave, but not the a-wave, is reduced at low-light intensities in AnkB; AnkG DKO animals, and only at higher flash intensities, where there is a mixture of rod and cone responses, are a-wave responses affected due to loss of ankyrins (Fig. 6B). We also measured cone-driven or photopic responses in controls and AnkB; AnkG DKO animals using a paired-flash protocol (38). Individual traces of photopic ERG responses showed a significant reduction in b-wave responses (bipolars) in AnkB; AnkG DKO animals compared with controls (Fig. 6C). Moreover, quantification of the a-wave and b-wave amplitudes of cone-driven or photopic ERG responses showed a significant reduction in AnkB; AnkG DKO animals compared with controls (Fig. 6D). The significant reduction in b-wave responses in both the rod and cone pathways is consistent with synaptic defects between photoreceptors and their respective downstream partners.

Figure 6.

Reduced in vivo retinal responses in mice with loss of ankyrin function (A–D). A) Individual ERG traces from controls and AnkB; AnkG DKO are shown at different scotopic or rod-driven stimulus intensities. B) Scotopic b-wave amplitudes are significantly reduced in AnkB; AnkG DKO (n = 8 from four mice) compared with controls (n = 8 from four mice) at both low-light and high-light stimulus. However, scotopic a-wave amplitudes are only significantly reduced in AnkB; AnkG DKO at high-light stimulus. C) Individual ERG traces from controls and AnkB; AnkG DKO are shown at photopic or cone-driven stimulus. D) Reduced photopic ERG responses in both a-wave and b-wave are observed in AnkB; AnkG DKO animals compared with controls using a paired-flash method (n = 8 from four mice per group). Data represented as mean values ± SEM. Statistical significance was determined using Student's t tests with Holm–Sidak correction for multiple comparisons for scotopic ERGs, and an unpaired two-tailed Student's t test for photopic ERGs. *P < 0.05, **P < 0.01, ****P < 0.0001.

Discussion

Our work uncovered a new role for ankyrins at photoreceptor synapses in the mouse outer retina. We found AnkB and AnkG are expressed in HCs at early developmental stages, and protein expression is localized to the synaptic layer of the outer retina in adult mice. Moreover, loss of AnkB and AnkG leads to (i) mispositioned processes or sprouts from HCs into the ONL, (ii) reduced expression of pre- and postsynaptic proteins in the OPL, (iii) defects in synaptic connectivity between HCs and photoreceptors, and (iv) impaired in vivo retinal responses. Taken together, our findings reveal ankyrins are essential at the synapse in the outer retina and are required for normal visual function.

HCs in photoreceptor circuit assembly

In the outer retina, neurons are born shortly after birth, but synaptogenesis is delayed and occurs in a stereotyped and sequential manner postnatally (3). One of the earliest events in synapse development is the formation of connections between photoreceptors and HCs. These initial contacts are known to be critical for the dendrites of bipolars to properly position themselves and form synaptic connections to photoreceptors in the OPL (9). Our developmental analysis uncovered HC specificity to photoreceptors arises at early developmental stages (P3–P10) prior to connectivity with BCs (P9–P13). After HCs have made their initial contacts with photoreceptors, bipolars begin to make contacts with the rod and cone terminals. The cell adhesion molecule Ngl-2 has been shown to mediate synaptic connectivity between HCs and rod photoreceptors. Loss of Ngl-2 results in sprouting of HCs but not of dendrites from rod bipolars (15), similar to AnkB; AnkG DKO animals. However, removal or ablation of HCs leads to sprouting of bipolar dendrites at later stages (∼3–7 months) but not at early time points (9, 10). Although there is no sprouting of BCs at early developmental stages, there is a significant reduction of pre- and postsynaptic protein expression in the OPL (9, 10), suggesting impaired synapse formation due to loss of HCs. These findings imply that the initial extension and positioning of bipolar dendrites to the OPL is not dependent on HCs but is still required for proper synapse development. Dendrites of BCs must use HCs as a recognition signal and/or scaffold to then form proper synapses with photoreceptors, and sprouting of bipolar dendrites observed at later stages may be a consequence of defective retinal connectivity.

Function of ankyrins in the developing nervous system

In this study, we found different members of the ankyrin family to be temporally expressed in HCs. AnkB is expressed at early stages (P3), whereas AnkG is expressed at later stages from P5 to P10. Ankyrins are cytoskeletal scaffolding proteins that are widely expressed throughout the developing nervous system (39, 40). They are known to play a pivotal role in forming specialized structures within the nervous system, such as the axon initial segment and nodes of Ranvier, which are essential for normal neuronal function (26, 41–43). Ankyrins form and maintain these specialized structures within neurons by recruiting different cell adhesion proteins, channels, and receptors to the membrane and linking them to the cytoskeleton (44–48). In mammals, the ankyrin family consists of three members: AnkR, AnkB, and AnkG. Previous studies have shown AnkB and AnkG are highly expressed in the mouse retina (49–51), with AnkB localized to the synaptic layers at adult stages (52). Although ankyrin expression in the retina had been previously reported, its function at the synapse had been largely unknown.

Role of ankyrins in the mouse outer retina

The first synapse of the outer retina is comprised of multiple components, including scaffolding proteins, receptors, and channels (reviewed in (1)). Loss of presynaptic proteins such as Bassoon (35) or the auxiliary subunit of the calcium channel (53) disrupts synaptic connectivity in both the rod and cone pathways. This is often seen as retraction from presynaptic neurons and/or sprouts from postsynaptic neurons. However, these molecules are shared among rod and cone synapses, which raise the question of what drives their distinct connectivity. Recent advancements in RNA sequencing have begun unveiling new molecules whose expression is restricted to either rods or cones (3, 54–56). These include several cell adhesion molecules such as Elfn1 (13), Elfn2 (14), Nfasc (17), and Lrfn2 (57). Functional studies suggest restricted expression of these molecules may mediate selective wiring of photoreceptors to their respective targets, as loss of these cell adhesion molecules disrupts connectivity of one pathway but not both. Similarly, ankyrins in HCs may be responsible for binding and recruiting various synaptic components, including cell adhesion molecules to mediate neural circuit assembly. In our scRNA-seq data, we uncovered cell adhesion molecules such as Nfasc and Cntnap1 to be highly expressed in HCs during retinal development. Moreover, protein expression analysis revealed Nfasc and Cntnap1 to be restricted to distinct subcellular compartments within HCs. Cntnap1 appears to be localized to the somatodendritic region of HCs, whereas Nfasc appears to be concentrated in the axon. The restricted expression of these cell adhesion molecules may be responsible for the distinct wiring of HCs to their respective photoreceptor partner. Interestingly, Nfasc is a known binding partner of AnkG (47). However, loss of Nfasc in the retina does not recapitulate the synaptic defects seen in the retinas of AnkB; AnkG DKO animals (17). Specifically, loss of Nfasc disrupts synaptic connectivity of the rod pathway but not the cone pathway (17). Since we observe connectivity defects and abnormal retinal responses in both the rod and cone pathways due to loss of ankyrins, there must be other cell adhesion molecules that work alongside Nfasc to mediate cone connectivity, possibly through Cntnap1-mediated interactions. Therefore, additional studies are needed to identify the different binding partners of ankyrins at the rod and cone synapse in the outer retina.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Competing Interest

The authors declare no competing interests.

Funding

This work was supported by the National Eye Institute (F31EY035931 to R.M.P., R01EY033037 to E.Z.-S., R01EY035324 to Y.-R.P., T32EY007026 to M.C., P30EY002520 to BCM Department of Ophthalmology, and P30EY00331 to UCLA Jules Stein Eye Institute); Research to Prevent Blindness to E.Z.-S., Y.-R.P., unrestricted department grants to BCM, UCLA, UW Madison; National Institute of General Medical Sciences (T32GM008042l to M.C.); McPherson Eye Research Institute's Rebecca Meyer Brown/Retina Research Foundation Professorship to M.H. and the UW 2020 WARF Discovery award.

Author Contributions

Ross M. Perez (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing—original draft, Writing—review & editing), Jay Campbell (Data curation, Formal analysis, Investigation, Methodology), Martina Cavallini (Formal analysis, Investigation, Methodology), Collin Todora (Data curation, Formal analysis, Investigation, Methodology), Daniela Becerril (Data curation, Formal analysis, Investigation, Methodology), Debalina Goswami-Sewell (Data curation, Formal analysis, Investigation, Methodology), Rajashree Venkatraman (Data curation, Formal analysis, Investigation, Methodology), Cesiah C. Gomez (Data curation, Formal analysis, Investigation, Methodology), Caitlin Bagnetto (Data curation, Formal analysis, Investigation, Methodology), Audrey Lee (Data curation, Formal analysis, Investigation, Methodology), Marlon F. Mattos (Data curation, Formal analysis, Investigation, Methodology), Yi-Rong Peng (Conceptualization, Formal analysis, Investigation, Methodology), Mrinalini Hoon (Conceptualization, Formal analysis, Investigation, Methodology), and Elizabeth Zuniga-Sanchez (Conceptualization, Funding acquisition, Project administration, Supervision, Writing—review & editing)

Preprints

This manuscript was posted on a preprint at https://www.biorxiv.org/content/10.1101/2025.02.11.637690v1.

Data Availability

The scRNA-Seq data generated in this study have been deposited to the Gene Expression Omnibus (GEO) under accession ID GSE336188 and Zenodo (10.5281/zenodo.20707795).