Ankyrin-G Promotes Cyclic Nucleotide–Gated Channel Transport to Rod Photoreceptor Sensory Cilia

See allHide authors and affiliations

Science  20 Mar 2009:
Vol. 323, Issue 5921, pp. 1614-1617
DOI: 10.1126/science.1169789


Cyclic nucleotide–gated (CNG) channels localize exclusively to the plasma membrane of photosensitive outer segments of rod photoreceptors where they generate the electrical response to light. Here, we report the finding that targeting of CNG channels to the rod outer segment required their interaction with ankyrin-G. Ankyrin-G localized exclusively to rod outer segments, coimmunoprecipitated with the CNG channel, and bound to the C-terminal domain of the channel β1 subunit. Ankyrin-G depletion in neonatal mouse retinas markedly reduced CNG channel expression. Transgenic expression of CNG channel β-subunit mutants in Xenopus rods showed that ankyrin-G binding was necessary and sufficient for targeting of the β1 subunit to outer segments. Thus, ankyrin-G is required for transport of CNG channels to the plasma membrane of rod outer segments.

Cyclic nucleotide–gated (CNG) channels initiate the electrical responses to light in photoreceptors and to chemical stimuli in olfactory neurons (1). CNG channels are segregated to sensory cilia, where visual and olfactory signal transduction takes place. This precise intracellular localization is dependent on the channel's β subunit (CNG-β1, CNGB1) in both classes of neurons (24). However, the molecular mechanism(s) of CNG channel targeting to the plasma membrane of sensory cilia, where this channel normally functions, are unclear.

Ankyrin-G is a versatile membrane adaptor involved in the formation and maintenance of diverse specialized membrane domains (59). Ankyrin-G is localized exclusively to rod outer segments (ROSs), where it was found along with CNG channels that have been localized to the ROS plasma membrane (10, 11) (Fig. 1, A and B). In contrast, the plasma membrane of the inner segment was lined with ankyrin-B, which is required for the coordinated expression of the Na+- and K+-dependent adenosine triphosphatase (ATPase), Na+ and Ca+ exchanger, and β2-spectrin (12) (Fig. 1A). Localization of ankyrin-G to the plasma membrane was evident in isolated mouse ROSs, but was better demonstrated in frog ROSs, which are three to four times as large in diameter (Fig. 1B). Ankyrin-G also localized in the olfactory sensory cilia and the principal piece of sperm flagella, together with CNG-β1 (4, 13) (fig. S1). We treated isolated bovine ROSs with a cleavable cross-linker, solubilized them in 0.1% SDS, and used ankyrin- or CNG-β1–specific antibodies for immunoprecipitation. We observed the reciprocal coimmunoprecipitation of CNG-β1 and ankyrin-G (Fig. 1C). The interaction with ankyrin-G was specific, because CNG-β1 was not precipitated by nonimmune or ankyrin-B–specific antibodies, and the major ROS-specific protein, rhodopsin, was not precipitated in either case (Fig. 1C).

Fig. 1.

Ankyrin-G is restricted to photoreceptor outer segments and binds the rod CNG channel. (A) Colocalization of ankyrin-G (AnkG, red, left two columns) with CNG channel (CNG-β1, green) in ROSs and colocalization of ankyrin-B (AnkB, red, right two columns) with Na+- and K+-dependent ATPase (NKA) (green) in inner segments (IS). A schematic of a rod cell is shown to the right. (B) Ankyrin-G (red) localizes to the plasma membrane of isolated mouse and frog ROSs labeled with CNG-β1–specific antibody (green). ROS tangential sections are shown (top) and longitudinal sections (bottom). (C) Coimmunoprecipitation of ankyrin-G with CNG-β1 channels from bovine ROS extracts. Antibodies used for precipitations are indicated on the top (immunoglobulin G, IgG), antibodies used for protein detection are indicated on the left. (D) CNG-α1 (white) alone does not recruit ankyrin-G–GFP (green) to the plasma membrane of HEK 293 cells. (E) Ankyrin-G–GFP (green) is recruited to the plasma membrane of HEK 293 cells coexpressing CNG-β1 (red) and CNG-α1 (white). Scalebars: 5 μm in (A), 10 μm in (D) and (E).

The CNG channel binding to ankyrin-G was further evaluated using a HEK 293 cell–based assay for detecting ankyrin–membrane protein interactions (14). In this assay, overexpressed exogenous ankyrin-G fused to the C terminus of green fluorescent protein (GFP) (ankyrin-G–GFP), which is normally localized to the cytoplasm, is recruited to the plasma membrane when ankyrin binding partners, such as neurofascin, are coexpressed. CNG-α1 expressed in HEK 293 cells localized to the plasma membrane but did not recruit ankyrin-G–GFP (Fig. 1D), whereas CNG-β1 failed to localize to the plasma membrane of these cells when expressed by itself (fig. S2). Coexpression of CNG-α and CNG-β does yield functional heterotetrameric channels in the plasma membrane of HEK 293 cells (15). Such coexpression also resulted in efficient plasma membrane recruitment of ankyrin-G–GFP (Fig. 1E); complete recruitment was observed in 65% of cells coexpressing the three proteins (n = 50). Ankyrin-B–GFP was not recruited to the plasma membrane under similar conditions (Fig. 1E), which indicated that CNG-β1 interacts with ankyrin-G.

We next sought to evaluate whether ankyrin-G is required for localization of CNG channels to ROSs in vivo, using short hairpin RNA (shRNA) to deplete (knockdown) ankyrin-G expression in neonatal mouse retinas. We injected a mixture of a plasmid encoding shRNA targeting mouse ankyrin-G in 10-fold excess over a plasmid encoding GFP into the eyes of newborn pups followed by electroporation (16). Under these conditions, rods expressing GFP are typically co-transfected with the shRNA plasmid. Two weeks post injection, photoreceptors transfected with control shRNA expressed GFP, displayed normal morphology, and were robustly immunostained with ankyrin-G (Fig. 2). In contrast, photoreceptors transfected with ankyrin-G shRNA (GFP+) displayed a major reduction in the ankyrin-G immunofluorescence in ROSs [20 to 30% of control, based on immunofluorescence intensity of samples on the same slide (fig. S3)], and their ROSs were significantly shortened [average length of 4.7 μm versus 15.5 μm in control rods (n = 25); compare rhodopsin-labeled sections]. The immunofluorescence levels of both CNG-β1 and CNG-α1 were also markedly reduced, to a degree comparable with the ankyrin-G reduction (Fig. 2 and fig. S3). In contrast, rhodopsin levels estimated by fluorescence intensities of samples on the same slide were similar for control ROSs and ROSs expressing ankyrin-G shRNA (Fig. 2 and fig. S3). The shortened ROS phenotype in ankyrin-G–depleted retina was more severe than reported for mice lacking the CNG-β1 subunit (2). Thus, ankyrin-G plays role(s) in either assembly and/or maintenance of ROSs in addition to localization of the CNG channel. This result is similar to the requirement of ankyrin-G for biogenesis of the lateral membrane in cultured columnar epithelial cells (17).

Fig. 2.

Ankyrin-G is required for ROS morphogenesis. Retinas of newborn mice were electroporated with either ankyrin-G shRNA or control pFIV (3 μg/μl) plasmid, each mixed with pCAGGS-GFP (0.3 μg/μl) (11). Ankyrin-G (AnkG) staining is shown in green. The staining of rhodopsin (left), CNG-β1 (center), and CNG-α1 (right) is shown in red. GFP staining is shown in white. Scale bars, 10 μm.

The reduction in CNG channel expression in ankyrin-G–depleted rods could be explained by a requirement for ankyrin-G in targeting the channel to ROS plasma membrane from the endoplasmic reticulum or Golgi located in the inner segment. Indeed, ankyrin-G is required for both post-Golgi transport and immobilization of its binding partner E-cadherin in epithelial cells (6). To test this hypothesis, we needed to identify CNG-β1 mutants lacking ankyrin-G binding. We first determined whether ankyrin-G bound to either the N- or C-terminal cytoplasmic domain of CNG-β1 (Fig. 3A). Ankyrin-G–GFP was coexpressed with protein constructs in which the ankyrin-binding domain of neurofascin was replaced with either the entire cytoplasmic N or C terminus of human CNG-β1 [NF-CNG-βN (amino acids 1 to 654) and NF-CNG-βC (amino acids 1041 to 1251), respectively (Fig. 3A)]. Ankyrin-G interacted only with the C-terminal domain of CNG-β1, both in the HEK 293–based plasma membrane recruitment assay (fig. S4) and coimmunoprecipitation experiments (Fig. 3B). Immunoprecipitation experiments in HEK cells were performed in the absence of a cross-linking reagent.

Fig. 3.

The ankyrin-G–binding site resides in a C-terminal motif of CNG-β1. (A) Schematic diagrams of rod CNG-β1 (top) and HA-tagged neurofascin (HA-NF) chimeras with CNG-β1 (bottom). Numbers within parentheses indicate the amino acid ranges of the CNG-β1 polypeptide fused to neurofascin. The abilities of chimeras to recruit ankyrin-G–GFP to plasma membrane of HEK 293 cells is indicated by + or – (AnkBD, ankyrin-binding domain). (B) Ankyrin-G–GFP was coexpressed in HEK 293 cells with the chimeras shown in (A), cells were lysed, and proteins were immunoprecipitated by HA-specific antibodies. Immunoblots of samples from the starting material (left) and precipitated proteins (right) were probed with HA- or GFP-specific antibodies. Lane numbers correspond to the numbered chimeras in (A). (C) Sequence of the 28 C-terminal amino acids of human CNG-β1 and homologous regions from other vertebrates (19). Arrows indicate sites of C-terminal deletions hCNG-βC1243 and hCNG-βC1236 used to identify residues critical for ankyrin-G binding. Colored residues were mutated to alanine; those in red were critical for ankyrin-G binding and those in green were neutral. The human CNG-β1–β-dystroglycan chimera (hCNGβ-DAG) is shown at the bottom with the dystroglycan sequences marked in blue. (D) CNG-β1 (lane 1) CNG-βΔ 28 (lane 2) and CNG-β1 IL1237AA (lane 3) were coexpressed with CNG-α1 in HEK 293 cells and immunoprecipitated using the CNG-β1–specific antibodies. Each CNG-β1 mutant normally coprecipitated with CNG-α1, but failed to bind endogenous ankyrin-G. Starting material is shown on the left and immunoprecipitates on the right.

A truncation of the C-terminal 28 residues of CNG-β1 is associated with retinitis pigmentosa (RP; hCNG-βΔ28) (Fig. 3A) (18). Indeed, neurofascin fused to the CNG-β1 C-terminal domain bearing this deletion failed to recruit or to coimmunoprecipitate ankyrin-G–GFP (fig. S2A and Fig. 3B). Additional deletion mutagenesis (hCNG-βC1243 and hCNG-βC1236, Fig. 3C) (19) narrowed the interaction site to a seven–amino acid stretch in this region (Fig. 3C, underlined, and fig. S4B), and alanine-scanning mutagenesis revealed that the highly conserved residues Ile1237 and Leu1238 were essential for ankyrin-G binding (fig. S4B and Fig. 3C). Mutant CNG-β1 (RP deletion or IL1237AA mutation, in which alanine (A) replaced isoleucine (I) and leucine (L) at positions 1237 and 1238, respectively) coexpressed with CNG-α1 in HEK 293 cells failed to bind ankyrin-G without affecting the normal CNG-α1 and CNG-β1 association (Fig. 3D).

To test whether ankyrin-G binding was required for delivery of the channel to outer segments, human CNG-β1, either wild type or mutants unable to interact with ankyrin-G, were expressed in the rods of transgenic Xenopus laevis (20). We used a specific antibody against human CNG-β1 (21) to distinguish it from the endogenous Xenopus CNG-β1 (Fig. 4). Wild-type human CNG-β1 (WT) was found in ROSs in a distinctive pattern consistent with its plasma membrane localization (Fig. 4). In marked contrast, both the RP (Δ28) and IL1237AA mutants were confined to perinuclear sites within rod cell bodies and were completely absent from ROSs (Fig. 4).

Fig. 4.

Ankyrin-G binding is necessary and sufficient for CNG-β1 transport to ROSs of transgenic Xenopus. Retina sections in each panel are stained with CNG-β1–specific antibody (red) and TOTO-3 to label the nuclei. Nontransgenic tadpole control on the left demonstrates that this antibody does not recognize the endogenous channel. Other panels depict the localization of wild-type (WT) human CNG-β1 or its mutants (indicated above each panel; see results for abbreviations). Scale bar, 5 μm.

We next tested whether an ankyrin-G–binding site from an unrelated protein was sufficient for targeting CNG-β1 to ROSs. The native site in CNG-β1 required for interaction with ankyrin-G was replaced with 14 amino acids from β-dystroglycan, which binds ankyrin-G directly (5) and has little sequence similarity with the CNGβ1 motif (Fig. 3C). The CNG-β1–dystroglycan (CNG-β–DAG) chimera associated with ankyrin-G when coexpressed with CNG-α in HEK 293 cells (fig. S5). When expressed in transgenic Xenopus, this chimera was targeted to the ROSs plasma membrane; however, the mutant CNG-β–DAG IIF/AAA chimera lacking the ankyrin-binding site (fig. S5) (5) was retained in the photoreceptor cell body (Fig. 4). Because there is no retrograde movement of membrane proteins from ROSs back into the cell body (22), we conclude that ankyrin-G binding is both necessary and sufficient for trafficking CNG-β1 to the outer segment. The ankyrin-G pathway could intersect with the microtubule motor Kif17/osm3, which is found in ROSs (23) and is required for ciliary transport of olfactory CNG channels when expressed in MDCK cells (3). Another question relates to the specific β-spectrin partner of ankyrin-G in ROSs. Ankyrins partner with β-spectrins in performing their scaffolding roles in the membrane cytoskeleton and in mediating post-Golgi transport through interactions with phospholipids and motor proteins (2427). β2-Spectrin found in inner segments (28) is reduced in the retina with or without ankyrin-B (12) and thus is a likely partner for ankyrin-B there. β4-Spectrin is associated with ankyrin-G in axon initial segments and is present in rod inner and outer segments (fig. S4), where its presence makes it a plausible ankyrin-G partner in ROSs.

Ankyrin-G accomplishes two critical functions in photoreceptors: It is required for transport of CNG-β1 from its site of synthesis and the assembly and/or maintenance of ROSs. This resembles the role of ankyrin-G in axon initial segments, where it binds to and coordinates the localization of three proteins required for the initiation and regulation of action potentials (Nav1.6, KCNQ2 and 3 channels, and 186-kD neurofascin) (29, 30). Without ankyrin-G, axon initial segments lose these proteins and express dendritic markers (31). In epithelial cells, ankyrin-G is required both for targeting E-cadherin to the plasma membrane and for biogenesis of the lateral membrane (6, 17). We hypothesize that, in addition to targeting the CNG channel, ankyrin-G can interact with other ROS membrane proteins, as well as proteins required for their ROS trafficking, and these interactions are essential for ROS morphogenesis. A conserved ankyrin-G–based mechanism may thus be shared by photoreceptors, neurons, and epithelial cells that accomplishes both the targeting of membrane-spanning proteins to specialized plasma membrane domains as well as assembly and/or maintenance of these domains.

Supporting Online Material

Materials and Methods

Figs. S1 to S6


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article