Fig 1.
CNGA3-mediated channelopathy in the canine model: clinical manifestation and genetic basis.
(A) Scotopic and photopic ERG responses obtained from a 5-month-old CNGA3-R424W-affected dog showing normal rod flash response, and absence of cone-mediated flicker responses compared to the age-matched wild-type control in ERG responses evoked under dark-adapted (rod flash, 0.3Hz) and light-adapted (cone flicker, 29 Hz) conditions. See also S1 Fig for behavioral vision testing under bright- and dim-light conditions. (B) DNA sequencing chromatograms of the wild-type and CNGA3 spontaneous mutants showing the C1270T transition in exon 7, and the 1931_1933delTGG deletion. Both residues are highly conserved among vertebrate genomes. Canis familiaris—dog; Homo sapiens—human; Ovis aries—sheep; Mus musculus—mouse; Rattus norvegicus—rat; Pan troglodytes—chimpanzee; Ailuropoda melanoleuca—panda; Tursiops truncatus—dolphin; Gallus gallus—chicken; Bos taurus—cow. (C) Immunohistochemical representation of CNGA3 protein in the wild-type canine retina. CNGA3 is exclusively expressed in the outer segment of cone photoreceptor cells (labeled in red), shown also in a high-magnification confocal photomicrograph (right) representing a Z-stack of maximum projection images taken at 0.5 mm intervals created with Leica LAS-AF software (Leica Microsystems, Inc.). Cell nuclei were stained with DAPI. RPE: retinal pigment epithelium; IS: inner segment; ONL: outer nuclear layer; INL: inner nuclear layer; Scale bars: 10μm.
Fig 2.
R424W mutation disrupts salt bridge interaction and destabilizes the open state of pore in a homotetrameric CNGA3 model.
(A) Schematic representation of CNGA3 subunit consisting of six transmembrane (TM) spanning segments (S1-S6) and a pore domain between S5 and S6. The highlighted last residue of S6 (blue) is the site of canine CNGA3-R424W mutation; its predicted partner, glutamic acid E306, is the first residue of S4-S5 linker. (B) Amino acid sequence alignment of the S4-S5 linker and S6 segment of selected shaker K+ channel superfamily members. The TM regions of the CNG channel family were assigned using the crystal structure of the chimeric voltage-gated potassium channel Kv1.2/2.1 (PDB ID: 2R9R). Sequence alignments of S5 domain and pore region were omitted for clarity. The R424 residue is shown in blue and its interacting partner, E306 in red. The conserved salt bridges in the Kv channels show opposite charges at these positions. c = canine, b = bovine, h = human, r = rat, m = mouse. (C) Side view of the wild-type CNGA3 homotetramer model and the CNGA3-R424W mutant channel equilibrated in its environment. The voltage-sensing domain (S1-S4) is presented in green, the S4-S5 linker in purple and the pore-forming region (S5-S6) in grey. The residues E306 and R424 are shown as red and blue rods, respectively. The E306:R424 interaction (wild-type) or its loss (R424W mutant) is demonstrated on the higher magnification images. Carbon atoms are labeled in cyan, nitrogens in blue and oxygens in red. Other side chains were omitted for clarity. Note that R424 forms a salt bridge with the E306 molecule in three subunits out of four. (D) Bottom views of the wild-type CNGA3 and CNGA3-R424W mutant channels. S6 is represented as a grey solid surface highlighting the partial closure of the pore in the R424W mutant model.
Fig 3.
CNGA3-R424W mutant channel is non-functional, but CNGA3-E306R/R424E double charge-reversal mutant rescues the phenotype.
(A) Cyclic nucleotide-activated currents of the wild-type CNGA3 and a set of R424- and E306-mutant channels recorded from excised membrane patches at -60mV and +60mV in a presence of saturating concentrations of cGMP (200μM) and cAMP (5000μM). No nucleotide-activated currents were recorded for R424W-, R424E- or E306R-mutant channels. E306R-R424E double mutant channel restored cGMP activation, producing large currents similar to the wild-type CNGA3. Detailed results from multiple patches are summarized in S2 Table. (B) cGMP dose-response relationship of the wild-type and CNGA3-E306R-R424E double charge-reversal mutant channels. The plot shows the ligand concentration-dependent activation of the wild-type (black) and CNGA3 double mutant (blue) homomeric channels at -60 mV. Each data set was taken from a representative single patch. The cGMP K0.5 is 14.2 μM for the wild-type and 118.4 μM for the mutant channel; Hill coefficient [Nh] value is 2.1 and 2.0, respectively. (C) Cellular localization of YFP-tagged wild-type canine CNGA3 and CNGA3-R424W mutant in HEK tsA201 cells. Cells transfected with the wild-type construct showed specific fluorescent signals in the plasma membrane and Golgi-like organelles (arrow); cells expressing the mutant protein exhibited augmented intracellular signals consistent with aggregate formation in addition to membrane and Golgi-like fluorescence. Scale bar: 10μm. (D) Histograms of averaged cellular localization patterns for R424- and E306-mutant constructs versus wild-type CNGA3. A significant increase in intracellular aggregates was found in R424W-transfected cells (red bars) vs CNGA3-WT (black bars), and an apparent reduction in aggregate formation in the E306R-R424E double mutant channels (blue bars). An unpaired t-test was used to compare individual mutants vs WT (mean% ± SD for n>300 cells). * p-value of <0.05; ** p-values <0.0001.
Fig 4.
The V644 deletion alters CLZ core structure indicating unfolding of coiled-coil complex.
(A) Sequence alignment of the CLZ (C-terminal leucine zipper) domains in CNGA-type subunits and conservation of the V644 residue. Note that canine V644 corresponds to V630 of human CNGA3. c = canine; h = human. (B) Helical wheel diagrams looking down the superhelical axis of the CLZ coiled-coil from the N- to C- terminus. Heptad a positions are marked in pink, d positions are denoted in magenta, and V630 is shown in red. The V630del shifts residues from b and e heptad positions to the core a and d positions and causes the predominantly hydrophobic residues of the coiled-coil core in the wild-type structure (left) to shift destabilizing charged and small residues in the mutant complex (right). Sequence diagram of the wild-type and V630del mutant heptad is shown below. (C) Molecular dynamics trajectories and models of the wild-type and mutant CLZ trimeric structures. Snapshots before (t = 0 ns) and after (t = 325 ns) simulation for the wild-type CNGA3-CLZ (left). The CLZ core clearly remains intact with small perturbations near the solvent exposed N- and C- termini over long timescales. Snapshots before (t = 0 ns) and after (t = 116 ns) simulation for the V630del mutant model (right). The V630 deletion dramatically alters the previously stabilizing residue-residue interactions, leading to disruption of the CLZ coiled-coil and helical structure. Small red and blue spheres correspond to oxygen and nitrogen atoms, respectively. See also S4 and S5 Figs for the full-length molecular dynamics simulations.
Fig 5.
Complex cellular phenotype of V644del mutant channel.
(A) Cellular localization of YFP-tagged wild-type canine CNGA3 and CNGA3-V644del mutant in HEK tsA201 cells. Cells transfected with the wild-type construct showed specific fluorescence pattern of expression limited to the plasma membrane and Golgi-like organelles (arrow); an evident increase in intracellular aggregates was observed in cells transfected with V644del mutant construct consistent with abnormal trafficking and potential ER retention. Scale bar: 10μm. (B) cGMP- and cAMP-activated currents recorded from CNGA3-WT and a responsive patch expressing V644del mutant channels. Approximately 40% of V644del mutant patches had no cGMP-activated currents, in contrast to 100% responsive patches from the CNGA3-WT-transfected cells. This partial loss of channel activity might reflect incomplete subunit assembly associated with disruption of the coiled-coil structure as depicted in the simulation studies. The responsive patches showed cyclic nucleotide-activated currents with similar characteristics to WT channels. (C) Histograms of subcellular localization patterns monitored in HEK tsA201 cells co-transfected with V644del and CNGA3-WT cDNA constructs. Cells were transfected with either CNGA3-WT or CNGA3-V644del or both constructs at the indicated ratios. Each cell count represents >300 cells from at least 2 transfections (mean% ± SD).
Fig 6.
Long timescale molecular dynamics simulations of the native CLZ, V630del mutant structure, and 2:1 and 1:2 WT to V630 hybrid deletion models.
(A) Molecular dynamics trajectories of the wild-type, V630del mutant and V630 hybrid CLZ deletion models at the selected time points. The wild-type CLZ structure exhibited long timescale stability. The core structure of the coiled-coil remained intact during the entire trajectory, leading to an impressive long timescale stability and average backbone root-mean-squared deviation (RMSD) of 1.54 Å. The 2:1 WT:V630del CLZ structure showed some unfolding. The core structure of the coiled-coil retained two thirds of its wild-type structure, leading to a greater overall stability compared to the all V630del structure. However, the model did exhibit some unfolding with the V630del helix (cyan) beginning to detach from the complex. The 1:2 WT:V630del CLZ complex showed more unfolding. The core structure of the coiled-coil contains charged amino acids at opposing helical positions: 630Ea 637Da 658Kd 665Ed 669Ka causing significant disruption at these positions. The V630del CLZ structure unfolded quickly. The core structure of the coiled-coil, now composed of destabilizing small and charged amino acids, begun to unfold shortly after initialization of the simulation. After 100 ns, most of the tertiary structure was lost, leaving an amalgam of helices and an average backbone RMSD of 9.09Å. Space filling surfaces: core a (pale pink) and d (magenta) positions applies to all panels. (B) Root-mean-squared deviations of the backbone structure at time t = x ns compared to the starting structure. The wild-type structure exhibits impressive stability with an average RMSD of 1.54 Å; the hybrid structures have intermediate stability with average RMSDs of 2.99 Å and 4.04 Å for the 2:1 (blue) and 1:2 (green) models, respectively. The 3-fold V630 deletion model is clearly unstable (purple) having an average backbone RMSD of 9.06 Å.