Ion behavior in the selectivity filter of HCN1 channels

Abstract

Hyperpolarization-activated cyclic-nucleotide gated channels (HCNs) are responsible for the generation of pacemaker currents (I_f or I_h) in cardiac and neuronal cells. Despite the overall structural similarity to voltage-gated potassium (Kv) channels, HCNs show much lower selectivity for K⁺ over Na⁺ ions. This increased permeability to Na⁺ is critical to their role in membrane depolarization. HCNs can also select between Na⁺ and Li⁺ ions. Here, we investigate the unique ion selectivity properties of HCNs using molecular-dynamics simulations. Our simulations suggest that the HCN1 pore is flexible and dilated compared with Kv channels with only one stable ion binding site within the selectivity filter. We also observe that ion coordination and hydration differ within the HCN1 selectivity filter compared with those in Kv and cyclic-nucleotide gated channels. Additionally, the C358T mutation further stabilizes the symmetry of the binding site and provides a more fit space for ion coordination, particularly for Li⁺.

Significance

Hyperpolarization-activated cyclic-nucleotide gated channels (HCNs) represent the molecular correlate of the currents I_f or I_h in cardiomyocytes and neurons. Here we study the unique low conductance and semi-selective properties of HCNs. The conductance and selectivity mechanisms of ion channels are tightly associated with their physiological role and contribute to the specific properties of the excitable cells in which they are expressed.

Introduction

Hyperpolarization-activated cyclic nucleotide-gated channels (HCNs) belong to the voltage-gated cation channel superfamily. The four HCN isoforms (HCN1–4) are responsible for the generation of I_h in cardiac and neuronal cells, where they play key roles in several cellular functions including setting the resting membrane potential, pacemaking, dendritic integration, and establishing action-potential threshold (1). HCNs are important for learning and memory (2,3), pain sensation (4), sour-taste sensation (5), and vision (6). I_h plays important roles in the mechanisms of epilepsy, pain, schizophrenia, addiction, and other neurological disorders (7,8).

HCNs are composed of four subunits, each consisting of six transmembrane alpha helices and a C-terminal cyclic-nucleotide-binding domain that is attached to the S6 transmembrane domains via an 80 amino-acid linker (9). The S1–S4 transmembrane helices of each subunit generate a non-domain-swapped voltage-sensor domain, which is arranged next to the pore-forming S5–S6 helices of the same subunit. Recent atomic resolution structures have also identified a novel N-terminal HCN domain (10,11), which couples cyclic-nucleotide binding to voltage gating (12). Despite the overall structural similarity to voltage-gated potassium (Kv) channels, HCNs demonstrate 20-fold lower selectivity for K⁺ over Na⁺ ions (P_Na/P_K = 0.2–0.3 in HCNs (9,13, 14, 15) compared with P_Na/P_K < 0.03 in Kv channels (16, 17, 18)). This increased permeability to Na⁺ ions results in a net influx of Na⁺ ions and is critical to the role of HCNs in depolarizing cellular membranes. Intriguingly, HCNs are capable of selecting between Na⁺ and Li⁺ ions, hence they are semi-selective unlike the closely related cyclic-nucleotide-gated (CNG) channels that are non-selective. Differences in selectivity between selective channels like Kv channels, and non-selective channels such as CNGs, have been partially attributed to multi-ion versus single-ion binding in the selectivity filter (19). In non-selective channels, the selectivity filter residues orient differently than those in the selective Kv channels, rendering the top of the pore wide open, eliminating the first (and possibly second) ion-binding sites (11) (Fig. 1 A). In addition to the three-dimensional arrangement of pore-forming residues, the dynamic behavior of the selectivity filter in the presence of various cations could have a major effect on ion selectivity. Structures of HCN1 indicate a wide-open top of its selectivity filter, similar to CNG channels. However, the details of how many ions stably bind within the HCN pore and their coordination remain poorly understood. Here, we investigate the unique ion-binding properties of HCNs and the effect of C358T mutation to ion selectivity and permeation in the selectivity filter of HCNs using unbiased and biased molecular dynamics (MD) simulations. The C358T mutation is particularly of interest as it converts the selectivity filter sequence of the HCN (CIGYG) to that of the Kv channels (TIGYG) (Fig. 1 B). Hence, the investigation of this mutation aims to shed light on why this mutation fails to confer K⁺ selectivity to HCNs but increases their unitary conductance (15,16).

Figure 1.

The arrangement of backbone atoms of the selectivity filter of KcsA, NAK, CNG, and HCNs. (Top) The four-letter codes represent the PDB ID of each protein. The ions in the selectivity filter are represented in spheres, with the black lines indicating the coordination by carbonyl and sidechain oxygens. (Bottom) Sequence alignment of the pore helix and selectivity filters (highlighted in orange) of HCNs, potassium, and CNG channels.

Materials and methods

System preparation

Atomic models of HCN1 channels were constructed based on the cryoelectron microscopy (cryo-EM) crystal structure of the protein (PDB: 5U6O (11)) in a closed conformation with a tightly packed inner helical bundle that constricts the pore to a radius of about 1 Å. Since HCN1 retains hyperpolarization-activated gating in the truncated form which is devoid of the C-linker and cyclic-nucleotide-binding domains (HCN1ΔCNBD) (20), these domains were omitted from the structure of the HCN1 channel used in our simulations to reduce the number of particles in the simulation systems. Therefore, our structural model of the HCN1 channel included residues 94–402 (UniProt: O60741) with the backbone of N- and C-terminal residues neutralized. The first set of simulation systems was prepared to determine the favorable binding sites of K⁺, Na⁺, and Li⁺ ions in the selectivity filter. Three models for each system were built; in each model, four identical ions were placed at S1, S3, S4, and the cavity beneath the selectivity filter (Fig. 2). Both ions located at S1 and the cavity site were fully hydrated, while the ions at S3 and S4 were directly exposed to carbonyl groups of the selectivity filter residues. In the second set of simulations, a single K⁺, Na⁺, and Li⁺ ion was placed in the S3 site of the selectivity filter. In the third set of simulations, a C358T mutation was introduced using Chimera UCSF software (21), and the systems were rebuilt for each ion condition. The pKa value of each residue was calculated with the PROPKA server (22), and all residues were assigned their standard protonation state at pH 7 accordingly. Consequently, His355, located on the pore helix, was protonated. The protein was then oriented with respect to the membrane using the Orientation of Proteins in Membranes webserver (23).

Figure 2.

Identification of a single stable binding site for K⁺ (POT), Na⁺ (SOD), and Li⁺ (LIT) ions in the HCN1 selectivity filter. (A) Time-evolution plots of the positioning of K⁺, Na⁺, and Li⁺ ions along the selectivity filter. The data for each ion are color coded: 1_S1 (black) represents ion number 1, which was initially positioned in the S1 site; 2_S3 (blue) represents ion number 2, which was initially positioned in the S3 site; 3_S4 (green) represents ion number 3, which was initially positioned in S4 site; and 4_C (red) represents ion number 4, which was initially positioned in the bottom cavity site. Horizontal lines represent the average z coordinates of the carbonyl groups in the selectivity filter residues (from bottom to top: C358, I359, G360, Y361, and G362); the mean average and standard deviation for these coordinates are reported on the right side of each graph. (B) Snapshots of the HCN1 selectivity filter with ions and their coordinating waters were taken at the end of 20, 100, and 200 ns, respectively.

MD simulations

The prepared HCN1 channel was embedded in a POPC bilayer and solvated with 150 mM of KCl (in the case of K⁺/Li⁺ systems) or NaCl (in the case of Na⁺ systems) using the CHARMM-GUI web server (24). The total number of atoms in the resulting simulation system is on the order of 170,000 atoms. The CHARMM36 force field (25) for protein, lipids, and ions was used. Explicit water was described with the TIP3P model (26). Standard CHARMM parameters for K⁺, Na⁺, and Li⁺ ions were used.

The prepared systems were refined using energy minimization for at least 2000 steps, and the ions and protein backbone atoms were kept fixed throughout the minimization procedure. For simulation series 2, 3, 4, and 5 (Table S1), the ions located in the selectivity filter continued to be restrained after energy minimization for an additional 20 ns to relax any unfavorable contacts destabilizing the selectivity filter. All simulations were performed under constant isothermal-isoberic (NPT) conditions at 310 K and 1 atm. Periodic boundary conditions were applied. The electrostatic interactions were treated by the particle-mesh Ewald algorithm (27) with grid spacing less than 1 Å. A 12 Å smoothed cutoff (10–12 Å) with a switching distance cutoff of 10 Å was applied for van der Waals interactions. The pressure was maintained at 1 atm using a Langevin piston control (28), with a period of 50 fs and an oscillation decay time of 25 fs. To maintain the temperature, the system was coupled to the Langevin thermostat with a damping coefficient of 1 ps^-1. Equations of motion were integrated with a time step of 2 fs. Bond lengths involving hydrogen atoms were constrained using the SHAKE algorithm as implemented in NAMD. After minimization and equilibration with harmonic positional restraints on all alpha-carbon atoms, MD simulations were performed for 200 ns for wild-type and C358T channels using NAMD 2.12 (29) on the supercomputers of Compute Canada. Simulations were performed in two steps: an initial 20 ns during which the positions of ions in the selectivity filter were restrained for 20 ns to relax any unfavorable contacts destabilizing the selectivity filter with a harmonic potential applied along all Cartesian directions with an exponent of 2 and scaling factor of 3. This initial phase was followed by a 180 ns simulation after the restraints on the ions were removed. All molecular graphics work and figures were generated using VMD (30) and UCSF chimera (21).

Potential of mean-force calculations

To study the energetics of ion permeation through the selectivity filters of wild-type (WT) and C358T HCNs, one-dimensional potential mean force (PMF) calculations were computed as a function of the reaction coordinate, the distance between the permeant ion and the center of the selectivity filter along the pore axis, using umbrella-sampling simulations. The geometric center of the selectivity filter was calculated using the alpha-carbon atoms of the residues 358–362. The initial configurations of the simulation systems were taken from the last frame generated at the end of 200-ns-long, unbiased MD simulations described in the previous section. The region of interest for the reaction coordinate was between -10 and +16 Å and was covered by 27 equally spaced umbrella-sampling windows. In each window, the reaction coordinate was centered at a specific distance within this range using a harmonic bias potential with a force constant of 2.5 kcal/mol/Ų. Each window comprised a relaxation stage of 1 ps and an equilibration phase of 6 ns, followed by a production phase of 6 ns, performed with NAMD 2.14 (29). The initial configurations for this series of windows were extracted from the configuration obtained at the relaxation stage of the preceding window along the reaction coordinate. In order to avoid the permeant ion from escaping from the pore, a flat-bottom harmonic restraint with a force constant of 100 kcal/mol/Ų was used to keep the permeant ion radially 8 Å within the pore using the collective variable distanceXY (31). The rest of the simulation parameters were described in the previous section. The umbrella-sampling windows provided the biased probability distribution of the reaction coordinate. The unbiased probability distribution of this reaction coordinate and the PMF as a function of the reaction coordinate were estimated using the weighted histogram analysis method (32). The PMF calculations for the permeation of K⁺ and Na⁺ through the selectivity filter were performed for both WT and C358T HCN1 channels. The first half of the trajectory for each umbrella-sampling window was discarded for equilibration. The standard error values for PMF profiles were computed from the weighted histogram analysis method estimation of two halves of the last 6-ns-long trajectory.

Results

Dynamics of ions’ motions within the selectivity filter of WT HCN1

To identify the stable binding sites for K⁺, Na⁺, and Li⁺ ions in the HCN1 pore, we performed all-atom MD simulations. In separate simulations for each ion type, ions were simultaneously placed in S1, S3, S4, and cavity with the ion in the cavity hydrated. No ion could be placed in the S2 site because the carbonyl groups are flipped away from the conduction pathway, and the tyrosine sidechain makes the site unavailable for ion binding. In un-restrained simulations, K⁺ and Li⁺ ions initially located in S3 and S4 rapidly leave the pore, and the ion from the cavity stably replaces the S4 ion (Fig. S2). Na⁺ ions move similarly, though the Na⁺ ion that moves from S4 to S3 resided there for the remainder of the 100 ns simulations. We ran a second, longer set of simulations with the ions constrained during the first 20 ns of production simulations so that the carbonyl groups of the pore could adjust their arrangement around each ion, and the ions did not leave the pore simply due to an unfavorable starting arrangement. The trajectory of ion movement within the selectivity filter over the course of the simulation time for each ion is presented in Fig. 2. After 20 ns, all constraints were removed, and ions that were initially located at S1, S3, and S4 once again pop out of the pore and irreversibly join the solvent. The ion, which was initially localized within the cavity, migrates to the S4 site while being coordinated by the carbonyl groups of Cys358 along with water molecules. These results were confirmed in a second independent set of simulations for each ion, performed using the same initial structure but different initial velocities (Fig. S3). Notably, in those simulations, both K⁺ and Na⁺ move from S4 to S3 (largely at the plane of the carbonyl groups) and reside there for 150 ns before ultimately leaving the pore. Therefore, the S3 site appears to be a short-lived (low-affinity) binding site, which may contribute to a knock-on type permeation in an open channel. Since no structure of the open HCN1 pore was available at the time of these studies, our simulations were performed with the closed-pore conformation, which prevents the cavity from being replenished with another ion. Therefore, to evaluate whether the ion that moves from the cavity into the S4 site gets trapped there because there is no additional ion to repel it over the energy barrier into the S3 site (through either a hard or soft knock-on mechanism), we performed an additional set of simulations for each ion type (Fig. S6). In these simulations, only one ion (either K⁺, Na⁺, or Li⁺) was placed in the selectivity filter at the S3 site. The placement of the ion at the S3 position also provides further insight on the ability of this site to coordinate the ion and serve as a stable ion-binding site. In these systems, shortly after initiating the simulations, the ion migrates from the S3 to the S4 site (Fig. S6) and is coordinated by the carbonyl backbone of C358 and waters, similar to what we observed in the previous set of simulations (Figs. 2, S2, and S3). Again, this behavior was consistent for K⁺, Na⁺, and Li⁺ ions. This indicates that regardless of how the simulations are performed, the S4 site is the primary stable binding site within the selectivity filter of HCN1 channels.

A careful examination of ion movement within the selectivity filter indicates that ions fluctuate along the z axis between the plane of C358 carbonyl groups and in the S4 site (Figs. 2 and S3). The ions frequently sample the plane of the C358 carbonyl groups (black lines in Fig. 2). Additionally, ions are partially hydrated while being coordinated by only two of the four selectivity filter subunits (Fig. 3). This differs from observations in Kv and CNG channels (19,33,34), where ions are either dehydrated in the selectivity filter (in Kv channels) or synchronously coordinated by carbonyl groups from all four subunits (in both CNG and Kv channels). To analyze the localization of ions, we calculated the distance between the ion and the carbonyl groups of C358 from each subunit throughout the simulations. For K⁺ and Na⁺ ions, the average distances between the ion and C358 carbonyl groups of subunits C and D are 2.8 and 2.5 Å respectively, while the distances of these ions to the C358 carbonyl groups of subunits A and B are greater than 5 Å (Figs. 3 A and B). However, Li⁺ ions move more freely in the xy plane than K⁺ and Na⁺ ions; therefore, the distribution of distances is more scattered (Fig. 3). Finally, the peak at 2.2 Å indicates that Li⁺ ions also interact with the C358 carbonyl groups. In fact, the stability of ions within the S4 binding site (Fig. 3) follows the same selectivity sequences measured electrophysiologically (14), with K⁺ showing the most stable behavior in the pore, followed by Na⁺ and then Li⁺.

Figure 3.

Ion coordination by C358 carbonyl oxygens and water. (A) Histogram of the distance between each ion and its coordinating oxygen atoms (from CO groups of C358). In the x axes, the distance distributions calculated for each protein chain are shown separately. (B) Histogram of the number of oxygens (from water molecules or backbone of selectivity filter residues) that coordinate each ion during the simulation. The captions under each graph represent the ion number and its location in the selectivity filter: 1_{S1_W} represents water molecules that coordinate ion number 1 initially localized at S1 site. 2_{S3_W} represents water molecules that coordinate ion number 2 initially localized at S3 site. 3_{S4_W} represents water molecules that coordinate ion number 3 initially localized at S4 site. 4_{S4_W} represents water molecules that coordinate ion number 4 following its translocation to S4 site. 4_{S4_CO} represents carbonyl groups that coordinate ion number 4 following its translocation to S4 site. 3_{S4_CO} represents carbonyl groups that coordinate ion number 3 while localized at S4 site.

Taken together, these results suggest that there is only one primary binding site located in the HCN1 selectivity filter. This site is at the S4 position, formed between the carbonyl oxygen of C358 and the sulfhydryl group of its side chain. The off-center location of ions in the pore, high mobility of these ions between HCN subunits, and overall lack of a second stable binding site in the pore may all contribute to the low (∼1 pS) conductance observed for HCNs (35).

Coordination of partially hydrated ions within WT HCN1 selectivity filter

It is evident that ions permeating through Kv-selective pores are mostly dehydrated (33,34). However, our simulations indicate that cations permeating through HCNs are partially hydrated. We examined the number of water molecules coordinating each ion at the S4 position throughout the trajectory using ion-oxygen cut-off distances of 3.6 Å for K⁺, 3.2 Å for Na⁺, and 2.8 Å for Li⁺ (36). Our results show that each ion spends most of the time coordinated by four water molecules in addition to the 2 carbonyl groups of the C358 residue (Fig. 3, right panel). In the case of K⁺ and Na⁺, two of these waters are arranged to fill the gap between the ion and the carbonyl groups from the opposing non-coordinating subunits. The other two water molecules interact with the ions such that they do not interfere with direct interaction between ion and the carbonyl groups (Fig. 2). This arrangement of coordinating water molecules and carbonyl groups stabilizes the off-center localization of K⁺ and Na⁺ ions. In the case of the Li⁺, the ion is more tightly coordinated by four water molecules, which generally prohibits direct interaction between Li⁺ and the C358 carbonyl groups. As a result, the Li⁺ ion sits slightly lower than the C358 carbonyl plane when compared with K⁺ and Na⁺. It should be noted that during the simulation time, the sulfur atoms of C358 are very far apart (around 7–9 Å away from each other) (Fig. S4). In this regard, the S4 site resembles a small cavity that has merged with the bottom cavity and cannot form a caged-like structure to accommodate the ion. Therefore, while partially hydrated K⁺ and Na⁺ ions move along the z axis between the carbonyl plane of C358 and into the wide S4 site, the Li⁺ ion becomes fully hydrated rapidly and primarily wanders in this wide opening of the S4 site.

Limited hydrogen-bond network behind the HCN1 selectivity filter helps to stabilize S4

Ion binding, permeation, conductance, and selectivity in Kv channels are all partially determined by the rigidity of the selectivity filter (37, 38, 39). A hydrogen-bond network between residues on the pore helix and the selectivity filter residues in Kv channels helps to keep the carbonyl groups facing the conduction pathway and forming multiple ion-binding sites. As a result of this hydrogen-bond network behind the pore, the region between the carbonyl groups remains narrow enough to favor the coordination of dehydrated ions in the center of cage-like binding sites formed by the eight carbonyl groups of the selectivity filter residues (Fig. 1). Our MD simulations indicate that this hydrogen-bond network is limited to the lower part of the selectivity filter in HCN1 (Fig. 4 B; Table S2), with hydrogen-bond interactions stabilizing the pore-facing orientation of C358 and, to a lower extent, I359. Specifically, the C358 residue is stabilized in this orientation through hydrogen bonds formed with H355 via their backbone. This C358-H355 hydrogen bond is uniformly present in all four subunits in all the simulated systems (Tables 1 and S2). Additional hydrogen bonds are formed between G360-S354 and I359-S345; however, they are not uniformly observed between HCN1 subunits.

Figure 4.

Hydrogen-bond and electrostatic interactions behind the HCN1 selectivity filter. (A) The frequency of hydrogen-bond interactions between the side chains of D312 and H355 in the WT HCN1 systems as a percentage of the last 120 ns of the trajectory where interacting atoms satisfy 3.8 Å and 120° cut-off conditions. The pore is represented from the top view, and the selectivity filter is shown in the ribbon representation. Protein chains are labeled in red. The positions of D312 and H355 relative to selectivity filter are shown in the diagram confined in the box at the top of panel. (B) The frequency of hydrogen-bond interactions between the main-chain residues of selectivity filter and the residues on the pore helix in the WT HCN1 channel were calculated from the last 120 ns of the simulations. The snapshots show the average positioning of the ion within the selectivity filter.

Table 1.

Frequency of hydrogen-bond interactions in the HCN1 selectivity filter

H-bond pair	H-bond percentage in WT HCN1
	POT				SOD				LIT
	A	B	C	D	A	B	C	D	A	B	C	D
S354-G360	–	63	–	77	–	55	–	59	24	62	40	31
S354-I359	53	20	21	–	48	22	50	45	29	–	36	39
H355-C358	27	35	30	25	33	27	34	22	31	32	28	26
K351-Y361	–	–	–	–	40	–	–	–	–	–	–	–
H-bond percentage in C358T HCN1
S354-G360	28	–	65	42	–	51	–	32	–	52	65	29
S354-I359	57	–	–	20	52	32	118	29	80	–	61	33
H355-T358	26	–	28	33	27	25	34	31	31	21	33	20
T358-L357	78	72	76	76	73	68	43	76	62	76	77	73
K351-Y361	60	–	–	–	–	35	–	–	–	–	40	–

In addition, the protonated imidazole side chain of H355 also forms a stable electrostatic interaction with D312 (Fig. 4 A). This interaction appears to further stabilize the H355 backbone orientation to enable formation of the C358-H355 hydrogen bond (Fig. 4 A). Analysis of the carbonyl group flipping away from the central pore axis over time shows that C358 continuously faces the pore axis, while the remaining selectivity filter residues rapidly sample different orientations (Fig. S3). This provides further support that hydrogen bonds and electrostatic interactions behind the selectivity filter are important for stabilizing the S4 site but insufficient to stabilize residues G362, Y361, G360, and I359 that would establish a multi-ion pore. Intriguingly, the propensity to form the H355-D312 salt bridge again follows the pattern of ion selectivity (Fig. 4 A), with this salt bridge formed more frequently in the trajectory of simulations with K⁺ than Na⁺ and least frequently for Li⁺. This suggests that while the selectivity filter may stabilize ion binding at S4 in the HCN1 pore, the ions at the S4 binding site may also contribute to the stabilization of the HCN1 selectivity filter via formation of additional interactions between residue pairs.

Ion dynamics and coordination within the selectivity filter of HCN1 C358T

The role of the pore-lining Cys residue in ion permeation and selectivity in HCNs has been examined previously by electrophysiological experiments (14,15). Mutations of the CIGYG selectivity filter sequence in HCNs to the typical TIGYG selectivity filter sequence of Kv channels fail to confer K⁺ selectivity to HCNs (14,15). In fact, the equivalent C358T mutation increases ion conductance by ∼30%, increases the permeation of Na⁺ and Li⁺ ions, enables the permeation of large quaternary ammonium ions such as tetramethylammonium, and abolishes the channels’ ability to select between Na⁺ and Li⁺ ions (14). To understand why the C358T mutation enhances ion permeation, including enabling improved passage of Na⁺ and Li⁺, we performed additional sets of MD simulations on HCN1 C358T channels. Similar to what we observed in simulations of the WT channel after 20 ns when all constraints on the ions were removed, ions that were initially located at S1, S3, and S4 move rapidly out of the pore and irreversibly join the bulk solvent (Fig. S7). The ion that was initially localized within the cavity migrates to the S4 site and remains partially hydrated. However, in the simulations of the C358T channel, the ions are notably localized below the carbonyl groups of T358 rather than in same plane. Moreover, Li⁺ ions appear to be more stable at this lower position on the z axis, as the ions in the S4 site move less frequently to the plane of the T358 carbonyl group compared with those in the WT channel. This is also reflected in the distributions of the distance between the side-chain or backbone heavy atoms of opposing subunits: the distributions of these distances are much narrower in the mutated system bearing the Li⁺ ion (Fig. S4). We also observe that in simulations of C358T, the K⁺ ions move toward the central pore axis, as is evident by the more uniform distribution of distances between K⁺ and the carbonyl-oxygen groups (∼2.8 Å for all four subunits), the increase in the number of coordinating carbonyl groups, and the reduced number of coordinating waters (Fig. S8). To a lesser degree, we see a similar effect with Na⁺ ions (Fig. S6); however, the Na⁺ ion cannot fully approach the central pore axis as it is still coordinated by more water molecules compared with the K⁺ ion. This shift toward the central axis, and the lower position of the ion in the S4 site, may contribute to the small (∼30%) increase in ion conductance observed in HCN2 and HCN4 channels with the equivalent C358T mutations (14,15). This shift toward the central axis would reduce the strength of the ion-carbonyl interaction, and an incoming ion would be more likely to knock the ion at S4 through.

The effect of C358T mutation on the geometry and flexibility of the HCN1 selectivity filter

The selectivity filters of both WT and C358T HCN1 channels are wide enough to accommodate partially hydrated Na⁺ and fully hydrated Li⁺ ions. However, by examining the distribution of distance between the oxygen atoms of the T358 side chain, we see that the pore diameter varies from 4.2 to 9.2 Å (depending on the ion) (Fig. S4). Here, the pore is narrower and more rigid compared with the corresponding distance between the sulfur atoms of the C358 in the WT HCN, which varies between 5.2 and 10.2 Å (depending on the ion) (Fig. S4). In the case of K⁺ ions, the narrower region, formed by carbonyl groups and side-chain oxygen atoms of T358, now enables the dehydrated ion to be directly coordinated by 4 carbonyl groups and 4 side-chain hydroxyl groups. Water molecules are forced to engage directly above and below the K⁺ ion (Figs. S7 and S8), similar to what has been observed for Kv channels. The C358T pore at this position is narrower compared with that of WT HCN1 channels in the presence of Na⁺ and Li⁺ ions, tightly restricting the hydration shells of these ions. In this regard, the C358T mutation seems to tighten the S4 site and provide a more uniform space for ion coordination (Figs. S7 and S8). These results provide further evidence for a highly dynamic pore in WT HCNs.

We examined how often the carbonyl groups of the selectivity filter residues flip away from the pore by analyzing the time evolution of the angle between the carbonyl groups and the pore axis (Fig. S9). This analysis shows that only T358 carbonyl groups constantly face the pore axis, while the rest of the selectivity filter residues dynamically sample different orientations. Additionally, a similar pattern of hydrogen bonding is observed between the backbone atoms of the selectivity filter and the residues behind the pore in both WT and C358T HCN1 (Tables 1 and S2). The hydrogen bond between the backbone of T358 and H355 residues has the highest frequency of formation among the different HCN1 subunits and, as a result, the pore has a more synchronous behavior at the T358 site. Intriguingly, we observed that the side chain of T358 is also stabilized by a persistent hydrogen bond to the backbone of the neighboring L357 residue (Table 1; Fig. S11). This stable hydrogen bonding helps to fix the side chain of T358 toward the pore axis, which enables the coordination of dehydrated K⁺ and at least partially hydrated Na⁺ and Li⁺ ions deeper into the S4 site (Figs. S7 and S10). Lastly, the electrostatic interactions between the sidechains of D312 and H355 persist in these sets of simulations as well as those of the WT HCN1 channels.

Energetics of ion permeation through the WT and C358T HCN1 channels

To investigate single-ion binding for Na⁺ and K⁺ ions along the WT and C358T HCN1 selectivity filters, we computed PMF profiles of ion translocation as a function of the distance between the permeant ion and geometric center of the selectivity filter residues along the pore axis (Figs. 5 A and D). The simulations of the WT HCN1 reveal that the permeant K⁺ ion experiences a mild increase in free energy as it moves into the selectivity filter from the extracellular solution. The ion is always partially hydrated within the selectivity filter, which is consistent with our observations from unbiased simulations (Figs. 2 B and 3 B). Throughout the first half of the selectivity filter (0–16 Å), the number of water molecules coordinating the K⁺ ion ranges from 2–8, with an average of 5.3 water molecules. The pore is not wide enough to provide space for symmetrical water coordination for the permeant ion. In addition, protein atoms are often found within 3.5 Å of the K⁺, disrupting the symmetry of the ion’s hydration shell. The major free-energy barrier against K⁺ permeation is around 4 kcal/mol and is located above G360, whereas the most energetically favorable position for the permeant K⁺ is located at -7 Å (Fig. 5 B). Here, the K⁺ ion is coordinated by one C358 residue and ∼5.4 water oxygens. The other three protein subunits do not contribute to the ion permeation. The examination of K⁺ permeation through the C358T HCN1 selectivity filter showed that the behavior of the permeant ion is comparable through the first half of the selectivity filter. The location of the highest energy barrier (∼5.6 kcal/mol) is shifted by 1 Å compared with that of the WT system, and the K⁺ ion at this position is coordinated by ∼4.1 water molecules and three G360 backbones. Two G360 residues coordinate the ion transiently. Interestingly, the K⁺ coordination around the free-energy well is different in the C358T HCN1’s selectivity filter: the ion is hydrated only by two water molecules on average and is coordinated by T358 residues from all four subunits. Both backbone and side-chain oxygen atoms of T358 residues face toward the pore center, creating an energetically favorable space for the permeant K⁺. The distribution of the position of permeant K⁺ ion shows that this ion stays at the pore center (Fig. 5 C).

Figure 5.

Ion permeation through the WT and C358T HCN1 selectivity filters. (A and D) The potential of mean force (PMF) profiles for translocating single K⁺ (A) and Na⁺ (D) ions through the selectivity filters of WT and C358T HCN1 channels are computed as a function of the permeant ion position with respect to the geometric center of selectivity filter residues 358–362. In (A) and (D), PMF profiles are colored in black for the WT and in gray for the C358T HCN1 channels. In (A), a representative snapshot (insert) shows the permeant K⁺ ion coordinated by four T358 residues and two water molecules at -7 Å. The protein and water molecules are shown in licorice and the ion as a green sphere. (B and C) The selectivity filter residues and permeant K⁺ ion positions are plotted for the WT (B) and C358T (C) HCN1 selectivity filters. (E and F) The selectivity filter residues and permeant Na⁺ ion positions are plotted for the WT (E) and C358T (F) HCN1 selectivity filters. In (B), (C), (E), and (F), x and y axes show the radial distance from the pore center and the position along the pore axis, respectively. Data points are color coded: the data for Cys (or Thr) 358, Ile 359, Gly 360, Tyr 361, Gly 362, and the permeant ion are shown in orange, green, dark green, cyan, blue, and navy, respectively. These points are extracted from the last 6 ns of the 12-ns-long trajectories of the umbrella-sampling windows corresponding to the free-energy minimum (left) and maximum (right) for each given system. The permeant ion’s location is labeled on the top left corner of each graph.

Next, we inspected the Na⁺ permeation along the WT and C358T HCN1 selectivity filters. From the extracellular solution toward the WT selectivity filter, the permeant Na⁺ ion experiences a free-energy increase of ∼4 kcal/mol (Fig. 5 D). Between 0 and 16 Å, the Na⁺ is always hydrated by 3 or more water molecules; on average, 5 water molecules coordinate the ion. The most energetically favorable region for Na⁺ is located at -6 Å below C358 (Fig. 5 E). Here, both the most favorable position and ion coordination differ from those found for K⁺. The Na⁺ is only coordinated by one C358 residue and ∼4.1 water oxygens at -6 Å and hence is hydrated by fewer water molecules compared with the K⁺ at its energy minimum. The simulations of the C358T HCN1 channel with a permeating Na⁺ ion revealed that the maximum energy barrier, around 4.4 kcal/mol, is located at 4 Å (Fig. 5 F). At this position, an average of 5.4 water molecules coordinates the ion, and Y361 residues from two subunits are rarely found within 3.5 Å of the permeant Na⁺. A free-energy minimum for Na⁺ is located at -8 Å; in this region, the ion is coordinated by 6 water molecules (Fig. 5 F). Two opposing T358 residues rarely interact with the ion, leaving around 3–4 water molecules. Differing significantly from the uniform and stable coordination of K⁺ by T358 residues, the Na⁺ is almost completely coordinated by water molecules at its most energetically favorable position within the C358T HCN1’s selectivity filter.

Discussion

Kv channels have the backbone carbonyl groups of selectivity filter residues arranged in a tetragonal prism, which gives rise to four equally spaced ion-binding sites known as S1–S4 sites from top to bottom (33,40). In this symmetrical arrangement, a K⁺ ion is coordinated by eight oxygen atoms from carbonyl groups of the selectivity filter as if sitting in the middle of a cubic cage. This synchronous coordination of K⁺ ions in turn helps the carbonyl groups remain oriented toward the conduction pathway and enables multiple K⁺ ions to be coordinated in the pore, allowing for rapid ion conductance via a “hard” or “soft” electrostatic repulsion (knock-on). According to our results, the selectivity filter in the HCN1 channel is more flexible and mobile than what has been reported for a Kv pore, with only one primary binding site (between carbonyl groups of C358) stable during the simulation time. Additionally, we observe that the dilated pore, with a diameter of around 5 Å, accommodates the partially hydrated K⁺ and Na⁺ ions as well as the fully hydrated Li⁺ ion either in-plane with the C358 carbonyl groups or slightly deeper into the S4 binding site. Our simulations also indicate that ions can be coordinated at the S3 position, again in the carbonyl plane of I357; however, binding at this site is shorter lived, suggesting lower affinity. This is in line with recent simulations performed on the open HCN4 apo/LC structure, which similarly showed that K⁺ and Na⁺ ions spend the majority of time with a single ion bound in the pore in the plane of the C358 carbonyl groups with the S3 only transiently occupied (<1 ns) by K⁺ ions during the transition into a two-ion pore (41). In fact, in mixed K⁺/Na⁺ ion simulations of HCN4 conductance, K⁺ can replace Na⁺ from the S4 position without a two-ion configuration (41). These similarities are particularly notable since our simulations were performed using a CHARMM36 force field, while HCN4 simulations were performed with an Amber99sb∗-ILDN force field, indicating that favorable binding to the S4 at the carbonyl plane of C358 is independent of the force field used. Moreover, the conductive selectivity filter of the Y-in structure of the open HCN4 apo/LC structure used for simulations (41) is similar to that of the closed WT HCN1 used here. However, unlike what is observed for Kv and CNG channels (19,33), in HCNs, ions do not reside along the central pore axis, but rather, partially hydrated ions are coordinated in an offset manner more reminiscent of what is observed in Nav and Cav channels. While it is possible that the backbone carbonyl groups are largely polarized and that there is a sizable charge transfer from the backbone to the cations as in K⁺ channels (42), this analysis requires quantum-mechanical calculations to address the HCN selectivity filter. However, the effects may also be limited by the high flexibility the HCN1 selectivity filter, which is considerably less rigid than a K⁺ channel pore due to the lack of a hydrogen-bond network from the residues behind the selectivity filter.

Our data provide some insights on ion selectivity in HCNs; however, given the short timescale of our simulations, and the closed conformation of the pore (preventing us from fully observing conduction events), we are somewhat limited. Conventional views on the mechanism of selectivity in K⁺-permeating channels proposed a “snug-fit” model, which accounts for the rigidity of the filter and fits the ion in the pore according to its Pauling radius (43). This model assumes the ion to be stripped off from its water shell and that this dehydration energy is compensated by synchronous coordination of ion by selectivity filter oxygen atoms. More contemporary models consider the selectivity filter to be more dynamic with liquid-like properties, and selectivity arises a result of several factors such as the effect of strain energy (which accounts for structural perturbation of the host [e.g., selectivity filter] as the result of guest [e.g., ion] presence in the context of host-guest chemistry) and chaperone-like effect of the ions (which accounts for the necessity of ions being present to avoid selectivity filter collapse), in addition to ion hydration-dehydration energy compensation (34,44,45).

Kv channels have a 100- to 1000-fold higher selectivity for K⁺ over Na⁺ and Li⁺ ions (46). It has been argued that the multi-ion arrangement in the Kv-channel pore contributes to both rapid ion conduction and a high degree of ion selectivity. Studies using NaK channels (47, 48, 49) suggest that multi-K⁺-ion occupancy increases the probability that a Na⁺ will exit from the same side it entered (because a K⁺ ion would be blocking the other side due to its higher affinity) and thus permit kinetic selectivity. Additionally, smaller ions such as Na⁺ and Li⁺ have different dehydration energies, which may exclude them from the entering the pore (a molecular sieve). However, even once they can be dehydrated enough to enter the pore, Na⁺ ions interact asynchronously with the selectivity filter carbonyl backbones and contribute to rapid rearrangement of the carbonyl backbone away from the conduction path (flipping) and, in some cases, pinch or collapse the conduction pathway. On the other hand, our simulations of HCN1 in the closed conformation indicated that the HCN selectivity filter does not pinch or collapse the conduction pathway but rather that the selectivity filter remains primed for ion conduction. It is possible that this may contribute to the I_Inst component of HCN currents that is observed upon membrane hyperpolarization.

On the other end of the spectrum are CNG channels, which are effectively non-selective between monovalent cations. Recent atomic structures indicate that the lack of the external tyrosine and glycine residues in their selectivity filter sequence of CNG channels results in the elimination of the outer (S1) ion-binding site (19). The three remaining potential ion-binding sites are again formed by a combination of backbone carbonyl oxygens and hydroxyl oxygen of the threonine residue sidechains, and a continuous elongated density can be observed across all three sites in the atomic resolution structure. This suggests multi-ion binding within the CNG pore, despite their inability to effectively discriminate between K⁺, Na⁺, and Li⁺ ions (19,50). In fact, MD simulations using the engineered NaK2CNG-E channels suggest that the dilated selectivity filter in non-selective CNG-like channels may bind up to two partially hydrated ions with the inner (or lower) ion residing largely in the S3 site along the central pore axis (50). However, unlike the delicate hydrogen-bonding network surrounding the selectivity filter of K⁺ channels (37), the selectivity filter of the CNG channel is mainly reinforced by hydrophobic interactions (19). This results in a pore diameter of the CNG selectivity filter that ranges between 4.7 and 10.1 Å (19) compared with the typical pore diameter of 1.7–5.4 Å in Kv channels (51). This may provide more flexibility for the selectivity filter residues to properly arrange around different cations and may not force the CNG channel to sieve the ions based on their sizes (19).

It is clear from our studies here that ion selectivity cannot arise simply from the presence or absence of multi-ion binding within the pore or by the pore acting purely as a molecular sieve since HCNs can select between K⁺, Na⁺, and Li⁺ ions (14,15). Furthermore, the sulfhydryl group of C358 does not act like the rate-limiting barrier, as previously proposed (14), since the atomic structure, and our MD simulations, do not indicate this side chain to ever act as the narrowest restriction point in the pore. However, it was demonstrated previously that HCN selectivity is reduced when the pore-lining cysteine is mutated to threonine (C358T equivalent) (14,15). Free-energy calculations for K⁺ and Na⁺ permeation also reveal stable ion binding at the S4 site, with less-tight Na⁺ interactions at this position compared with K⁺ (free-energy change is ∼4.8 kcal/mol compared with ∼13.0 kcal/mol obtained for K⁺), which could enable a persistent Na⁺ conductance. In addition, a few small barriers seem to emerge in the free-energy change upon the entry of Na⁺ into the selectivity filter from the extracellular side (Fig. 5 D). Small barriers between relatively favorable positions (between S1 and S2 sites and at the S4 site) could also play a role in improving the Na⁺ flow through the selectivity filter, as suggested in different ion-channel studies (52). The free-energy increase upon entry to the selectivity filter is almost linearly maintained until the permeant K⁺ passes the S2 binding site of the WT selectivity filter, possibly slowing down the ion passage. At the S4 site, K⁺ is partially hydrated and sometimes is also coordinated by a single C358 residue, whereas Na⁺ is coordinated by two opposing C358 residues along with water molecules. Neither of these ions is uniformly coordinated by all four protein subunits, clearly demonstrating the low K⁺ selectivity of the WT HCN1 selectivity filter.

Interestingly, in the C358T HCN1 channels, the energetic cost of entering the pore is comparable for both ions we studied, but their behavior at the S4 binding site differs. The S4 site of the mutant selectivity filter creates a tight conformation for K⁺ ion binding and leaves no gap for the water molecules to sneak in and mediate the ion interaction with CO groups. In other words, the ion is stripped off from shielding water molecules (except for one water molecule at the top and one at the bottom of ion) and directly binds to CO groups of T358. It seems that the C358T mutation increases the chance of K⁺ ion gathering the oxygen atoms of T358 around itself at the S4 site. Establishment of an ordered S4 site in the mutated structure seems to be highly dependent on the precise localization of the T358 sidechain which in turn depends on a stable hydrogen bond between the side chain of this residue and the backbone carbonyl group of L357 (Fig. S8). This improved ion coordination may explain the increase in K⁺ conductance observed in mutant HCNs (14,15). In comparison, Na⁺ permeates by being mostly coordinated by waters. Hence, its location is radially shifted by up to 4 Å off the pore center (Fig. 5 F, left panel). The ring of T358’s carbonyl and hydroxyl oxygens from four subunits forms the proper cage-like solvation shell for K⁺, which cannot solvate Na⁺ as uniformly. However, Na⁺ ions sit deeper in the S4 site and are slightly more toward the central pore axis in C358T HCN1 than in WT HCN1 and, therefore, may be more readily knocked on by an incoming ion, enhancing Na⁺ conductance in the mutant channel.

The importance of the hydrogen-bond network involving selectivity filter residues and those of the pore helix to coordinate the proper arrangement of selectivity filter CO groups has been shown in several studies (37, 38, 39). Interestingly, L478T/C479T as well as S475E/C479T mutations in the HCN4 channel (equivalent to L357T/C358T and S354E/C358T mutations in HCN1, respectively) failed to produce a measurable current despite high levels of N-glycosylated protein expression (14). As shown in Fig. S11, these mutations seem to interfere with the network of hydrogen bonds behind the selectivity filter and specifically those that keep T358 and I350 CO groups in place. In the case of L357T/C358T mutations, the side chain of T357 could be localized in a proper location to establish a hydrogen bond with the CO group of S354. This interaction may perturb the orientation of S354, which in turn may disrupt its important role in reinforcing the hydrogen-bond network behind the pore. Upon S354E mutation, the side chain of E354 could rest at very close vicinity to the I359 amine (NH) group and could interfere with the proper arrangement of the T358 carbonyl group around K⁺ ion by dragging the backbone of this residue toward itself. These effects would lead to disorientation of the CO groups at the only binding site of the pore and, hence, compromise ion permeation along the pore.

While not directly examined in these studies, our data may provide some conceptual insights into the low conductance (∼1 pS) of HCNs. The four symmetrical and equally spaced ion-binding sites in Kv channels enable the coordination of a K⁺ ion by eight oxygen atoms from carbonyl groups of the selectivity filter as if sitting in the middle of a cubic cage. This synchronous coordination of K⁺ ions in turn helps the carbonyl groups remain oriented toward the conduction pathway and enables multiple K⁺ ions to be coordinated in the pore, allowing for rapid ion conductance via a hard or soft electrostatic repulsion (knock-on). However, in HCNs, the high mobility of a partially hydrated ion within the pore, the quantity of coordinating carbonyl groups, and their distance to the permeating ion, as well as the highly flexible and destabilized outer-pore region (which reduces the frequency of observing stable S1, S2, and S3 binding sites), may all contribute to the low (∼1 pS) conductance of HCN. Our data in HCN1 channels, as well as simulations in HCN4 channels (41), indicates that occupation of a second ion-binding site occurs infrequently, and when it does, ions may permeate by electrostatic repulsion (i.e., soft or hard knock-on). However, since this is rarely stable, the equilibrium highly favors a single-ion occupied pore. This model is supported by our simulations where a single ion was placed in the S3 site, and it readily moved back into the S4 site. The shift toward the central axis, and the lower position of the ion in the S4 site, may contribute to the small (∼30%) increase in ion conductance observed in HCN2 and HCN4 channels with the equivalent C358T mutations (14,15), since the strength of the ion-carbonyl interaction would be reduced and an incoming ion would be more likely to knock the ion through. This mechanism also explains how Cs⁺ can act as a pore blocker at low concentrations, yet permeate the channel at high concentrations, and how Cs⁺ permeation is further enhanced with Thr instead of Cys in the pore (14).

Author contributions

S.A., T.N.O., and N.D. devised computational experiments. S.A. and T.N.O. performed and analyzed computational simulations. The manuscript was prepared and edited by S.A., T.N.O., and N.D.

Declaration of interests

The authors declare no competing interests.

Editor: Philip C. Biggin.