Structure, Dynamics and Implied Gating Mechanism of a Human Cyclic Nucleotide-Gated Channel

Cyclic nucleotide-gated (CNG) ion channels are nonselective cation channels, essential for visual and olfactory sensory transduction. Although the channels include voltage-sensor domains (VSDs), their conductance is thought to be independent of the membrane potential, and their gating regulated by cytosolic cyclic nucleotide–binding domains. Mutations in these channels result in severe, degenerative retinal diseases, which remain untreatable. The lack of structural information on CNG channels has prevented mechanistic understanding of disease-causing mutations, precluded structure-based drug design, and hampered in silico investigation of the gating mechanism. To address this, we built a 3D model of the cone tetrameric CNG channel, based on homology to two distinct templates with known structures: the transmembrane (TM) domain of a bacterial channel, and the cyclic nucleotide-binding domain of the mouse HCN2 channel. Since the TM-domain template had low sequence-similarity to the TM domains of the CNG channels, and to reconcile conflicts between the two templates, we developed a novel, hybrid approach, combining homology modeling with evolutionary coupling constraints. Next, we used elastic network analysis of the model structure to investigate global motions of the channel and to elucidate its gating mechanism. We found the following: (i) In the main mode of motion, the TM and cytosolic domains counter-rotated around the membrane normal. We related this motion to gating, a proposition that is supported by previous experimental data, and by comparison to the known gating mechanism of the bacterial KirBac channel. (ii) The VSDs could facilitate gating (supplementing the pore gate), explaining their presence in such ‘voltage-insensitive’ channels. (iii) Our elastic network model analysis of the CNGA3 channel supports a modular model of allosteric gating, according to which protein domains are quasi-independent: they can move independently, but are coupled to each other allosterically.

interface of the two CNGA3 subunits, the α-carbons of residues R436, E467 and D507 are 11 Å apart from each other ( Figure S3). The model suggests that R436 could form a salt bridge with either E467 or D507 or both and stabilize inter-subunit and/or intra-subunit interaction. Similarly, at the interface of the CNGA3 and CNGB3 subunits, CNGB3 R478 could form a salt bridge with CNGA3 E467 or D549 ( Figure S3). The proximity of these charged residues supports the model structure and suggests the residues' possible functional role in stabilization of the closed state.
This role would be similar to that of the charged triad (R431, E462, D502) in the bovine channel [5,6].

Mapping disease-causing mutations on the model structure
We mapped all known disease-causing mutations of CNGA3 and CNGB3 onto the model structure (Tables S1 and S2; Figure S4). As anticipated, almost all known mutations, i.e., all 10 mutations in CNGB3 and 46 of 57 mutations in CNGA3, are found in evolutionarily conserved residues (ConSurf grades of 5 or larger), many of which are in the pore region and near the cGMP binding site. Of the 11 CNGA3 mutations found in more variable positions and shown in red bold in Table S1, four (E228K, T245M, S341P, E376K) have been found to affect trafficking and have a partial effect on channel function, two (D485V, E593K) have a partially deleterious effect on channel function, two (E194K, G267D) have a deleterious effect on channel function, and three (Y263D, N276S, G329C) have not been investigated experimentally (Table S1).
We also examined a subset of nine disease-associated mutations for which strong experimental data have been obtained in three or more independent studies. Two of these mutations are in the C-linker, and three are in the CNBD, including two that reside close to the cGMP binding site. The remaining four residues are in the VSD, including the conserved R277 and R283 (Table S1). This observation suggests an important evolutionarily selected role for the voltage-sensing motif, although the channel is voltage-insensitive.
Below we present three examples of possible molecular interpretations for the damage due to mutations associated with achromatopsia.
CNGA3 R427C/CNGB3 Y469D. The disease-causing mutation R427C in CNGA3 is of special interest because a mutation in the corresponding position in subunit CNGB3, i.e., Y469, is also disease-causing [7,8]. CNGA3 R427 is located in the A'-helix of the C-linker and can interact electrostatically with E453, located in the C'-helix of the adjacent CNGA3 subunit ( Figure S4A).
R427 can also interact with CNGB3 D488, located in the loop connecting helices B' and C' of the C-linker of the adjacent CNGB3 subunit, and with E495, located in helix C' of the adjacent CNGB3 subunit ( Figure S4D). We suggest that interactions between these charged residues stabilize inter-subunit interfaces, as in the R431-E462-D502 triad in the bovine channel ( Figure   S3) [5,6]. The mutation of the positively charged R427 to Cys would abolish the electrostatic attraction to the negatively charged residues (CNGA3 E453 or CNGB3 D488 and E495; Figures   S4A and S4D), disrupting the compact architecture of the C-linkers. The mutation of Y469 to the negatively charged Asp in CNGB3 (corresponds to CNGA3 R427) might repel D488 and E495 of the adjacent CNGB3 subunit ( Figure S4B), as well as E453 of the adjacent CNGA3 subunit ( Figure S4C). In this case, the architecture of the C-linkers would be disrupted due to the electrostatic repulsion between the negatively charged residues. CNGA3 L186F. CNGA3 L186 is located in the S1 helix, at the interface with the S5 helix and the P-loop ( Figure S4E). Clearly, the replacement of L186 with a bulky phenylalanine might disrupt this interface. Moreover, in voltage-gated channels this interface is functionally important [9].
Although CNG channels are not voltage-dependent, the role of the VSD in these channels is unclear, and the interface between S1 and the pore might be still essential. Regardless of the functional role of the interface in CNG channels, its structure can be significantly disrupted by the L186F mutation.
CNGB3 S435F. CNGB3 S435 faces the central pore ( Figure S4F), and its replacement with any large residue, e.g., phenylalanine, may disrupt the helix bundle, or simply block the pore. Indeed, CNGB3 homologs feature only small amino acids (S, A or G) in the position corresponding to S435, supporting the suggested interpretation of the effect of the S435F mutation.

Evolutionary couplings between amino acids that are not in contact in the model structure
In our analysis, most of the evolutionary couplings that did not correspond to actual amino acid contacts-more specifically, the majority of such couplings in the TM domain, and about half of such couplings in the cytosolic domain-had at least one residue in a loop region (data not shown). We hypothesized that these discrepancies emerged from differences in the conformation between CNGA3 and the bacterial template, and/or loop flexibility. To examine these possibilities, we calculated the evolutionary couplings of the templates directly ( Figure S5). The overlay between the calculated evolutionary couplings and the contacts derived from the crystal structures (MlotiK1 channel and the cytosolic domain of mouse HCN2 channel) was similar to that of our model structure (Figures 3 and S5), and the majority of the observed false positive predictions had at least one residue in a loop region. Taken together, these observations indicate that many of the evolutionary couplings between pairs of amino acids that, according to the model structure, are not in direct contact with each other are likely to represent flexibility in the loop regions.

Equilibrium dynamics of the CNGA3 channel
A Gaussian network model (GNM) detected the dynamic domains of the channel and their cooperative motions. The six slowest GNM modes of motion emerged considerably above the rest in the eigenvalue spectrum ( Figure S11A). These modes, therefore, represented the most important contributions to the overall motion of the channel, and we investigated them. We analyzed the shapes of the residues' mean-square displacement plots in the six slowest GNM modes. GNM modes 2 and 3, as well as GNM modes 5 and 6, shared the same eigenvalues ( Figure S11A) and were degenerate. The average shape of the residues' fluctuations in modes 2 and 3 was very similar to mode 4, indicating that GNM modes 2-4 correspond to the same motion ( Figure S11C). Thus, GNM modes 1-6 represented three types of motion. Next, we carried out anisotropic network model (ANM) analysis, which provided information on the directions of the motions in 3D-space. We associated the fluctuations (and correlations between fluctuations), detected by the GNM, to their directions, obtained from the ANM analysis, comparing the distributions of residue fluctuations in each ANM mode to the GNM modes ( Figure S7). We further validated the modes' associations by mapping the GNM-derived crosscorrelations between residues on the conformations obtained via ANM (Figure 4). Motion I was described by GNM mode 1 and ANM mode 3; motion II was described by GNM modes 2-4 and ANM mode 4; In motion III, described by GNM modes 5-6 and ANM mode 12 (Figures 4 and   S7).
The CNG channel features VSDs but is, in essence, insensitive to the membrane potential.
The role of the VSD is still unclear [10], although a recent study associated it with trafficking [11]. In order to understand the effect of the VSDs on channel dynamics, we performed GNM and ANM analysis of the CNGA3 channel (in holo-conformation), removing the VSDs. Three GNM modes of motion emerged as the slowest in the eigenvalue spectrum ( Figure S11B). GNM mode 1 was associated with ANM mode 3 ( Figures S8A and S8B); this motion was very similar to motion I of the intact CNGA3 channel, in direction, mobility and cooperative dynamics ( Figures 4A, 4B, S8A and S8B). GNM modes 2 and 3 were degenerate, and we related them to ANM mode 4 ( Figures S8C and S8D). This motion resembled motion III of the intact channel, with some alterations in cooperativity dynamics ( Figures 4E, 4F, S8C and S8D). Overall, the removal of the VSDs did not affect the slowest modes of motion of the CNGA3 channel, aside from the motions that directly involve the VSDs (motion II of the intact channel).