This Article Abstract Full Text (PDF) Supplemental File All Versions of this Article: biophysj.104.045286v1 89/2/932    most recent Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Giorgetti, A. Articles by Torre, V. Search for Related Content PubMed PubMed Citation Articles by Giorgetti, A. Articles by Torre, V.

A Homology Model of the Pore Region of HCN Channels

Istituto Nazionale per la Fisica della Materia (INFM-DEMOCRITOS Modeling Center for Research in Atomistic Simulation) and International School for Advanced Studies (SISSA), 34014 Trieste, Italy

Correspondence: Address reprint requests to V. Torre, E-mail: torre{at}sissa.it.

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
HCN channels are activated by membrane hyperpolarization and regulated by cyclic nucleotides, such as cyclic adenosine-mono-phosphate (cAMP). Here we present structural models of the pore region of these channels obtained by using homology modeling and validated against spatial constraints derived from electrophysiological experiments. For the construction of the models we make two major assumptions, justified by electrophysiological observations: i), in the closed state, the topology of the inner pore of HCN channels is similar to that of K⁺ channels. In particular, the orientation of the S5 and S6 helices of HCN channels is very similar to that of the corresponding helices of the K⁺ KcsA and K⁺ KirBac1.1 channels. Thus, we use as templates the x-ray structure of these K⁺ channels. ii), In the open state, the S6 helix is bent further than it is in the closed state, as suggested (but not proven) by experimental data. For this reason, the template of the open conformation is the x-ray structure of the MthK channel. The structural models of the closed state turn out to be consistent with all the available electrophysiological data. The model of the open state turned out to be consistent with all the available electrophysiological data in the filter region, including additional experimental data performed in this work. However, it required the introduction of an appropriate, experimentally derived constraint for the S6 helix. Our modeling provides a structural framework for understanding several functional properties of HCN channels: i), the cysteine ring at the inner mouth of the pore may act as a sensor of the intracellular oxidizing/reducing conditions; ii), the bending amplitude of the S6 helix upon gating appears to be significantly smaller than that found in MthK channels; iii), the reduced ionic selectivity of HCN channels, relative to that of K⁺ channels, may be caused, at least in part, by the larger flexibility of the inner pore of HCN channels.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
HCN channels are members of the voltage-gated superfamily of ionic channels, regulated by cyclic nucleotides, such as cyclic adenosine-mono-phosphate (cAMP). In contrast to most Na⁺ and K⁺ ionic channels, which open when membrane potential is depolarized (1–7), they are opened when the membrane potential hyperpolarizes below –50 mV.

HCN channels control fundamental biological events such as heart beat and neuronal rhythmic activity, providing the biophysical mechanism for the pacemaker (8–11) in these cells. These channels are composed of a cytoplasmic cAMP binding domain and a transmembrane domain. Although the x-ray crystal structure of the cytoplasmatic domain from the mouse channel has been determined recently (12) (mHCN2; Protein Data Bank accession number, 1Q3E), structural information on the pore domain is still lacking.

Indirect structural information has been inferred from experimental data on cloned channels, which have properties very similar to those of native HCN channels (13). These channels have been cloned from sea urchin sperm, referred to as spHCN (14,15) and from mammalian tissues, referred to as HCN1-4 (16–23). In particular, the HCN2 from mouse (mHCN2) has been experimentally widely investigated (1,24,25). Based on these data, it has been suggested that the pore regions of HCN share several important features with those of K⁺ channels: i), in both families the pore domain includes four identical subunits (16), each featuring a loop forming the inner pore region together with an additional small helix (P-helix), not spanning the lipid membrane. The inner pore and the P-helix form the so-called P-helix-loop. Each subunit of voltage-gated K⁺ channels and HCN channels features six transmembrane helices (referred to as S1-S6), whereas subunits of the KcsA and MthK K⁺ channels have only two transmembrane helices (referred to as TM1-TM2), structurally similar to S5 and S6 in voltage-gated K⁺ channels. S6 and TM2 are involved in gating, whereas the P-helix-loop does not change conformation upon gating (2,3,26). ii), The sequence of K⁺ and HCN channels in the inner pore region is highly conserved and contains the GYG fingerprint, called the selectivity filter (27,28). Recently, this has been supported by an investigation of the pore topology of the spHCN channel (1). iii), Some of the structural features of the closed and open states are similar: In the closed state, the orientation of the S5 and S6 helices of HCN channels is very similar to that of the TM1 and TM2 helices of the KcsA channel (2). This suggestion is based primarily on the assumption that the great majority, if not all, of the ionic channels belonging to the superfamily of voltage-gated channels share the same molecular architecture and in particular the orientation in space of S5 and S6 helices (7). The open state of K⁺ channels differs from the closed state only for the S6 helix (or TM2 helix in the case of KcsA), which is bent toward the lipid phase, around a Gly residue that acts as gating hinge (29). This bending was indeed observed in the x-ray structure of the MthK (3) channel in the open state, and proposed to be a general feature. S6 in HCN channels is believed to exhibit a similar conformational change in the open state (6,30). This proposal is supported by the presence in this family of the Gly residue, which provides the gating hinge in K⁺ channels, as well as of an Ala/Gly residue, five residues downstream in the so-called "activation gate region" (highlighted in dark gray in Fig. 1, A and B). This is further substantiated by electrophysiological experiments with wild-type (wt) and mutant spHCN channels (6,30).

View larger version (60K): [in this window] [in a new window]   FIGURE 1  (A) Sequence alignment of pore regions of KcsA, KirBac1.1, MthK, mHCN2, and spHCN channels. The filter GYG fully conserved signature (16) is shown as underlined residues. The activation gate (29), highlighted in dark gray, begins with the fully conserved glycine residue acting as gating hinge and the Gly/Ala five residues downstream (29). In the KcsA channel, E71 and D80 (bold residues) form a very strong hydrogen bond. Such interaction is represented here with a curved arrow. The long extracellular loop present in spHCN and mHCN2 is highlighted in light gray. Double-headed arrows indicate location of TM1/S5 and TM2/S6. (B) Sequence alignment of pore regions of spHCN, mHCN1, mHCN2, mHCN3, and mHCN4 channels. The conserved residues in the inner pore region are underlined, whereas the residues present in the extracellular end of the inner pore are shown in bold. The activation gate in the S6 transmembrane helix and the large extracellular loop were highlighted in dark and light gray, respectively. The cysteine in the extracellular loop conserved in HCN1-4 channels, responsible for MTSES and MTSEA channel block (58), is marked with an arrow. A thick black line indicates the P-helix-loop, which comprises residues from Tyr-387 to Pro-408 (mHCN2). In panels A and B, HCN inner pore conserved Cys residue is represented with bold underlined fonts.

Here we construct structural models of the HCN channels' pore region by homology modeling based on the available structural information on K⁺ channels, exploiting the structural similarities between HCN and K⁺ channels. Spatial constraints inferred by experimental data are then used to validate our models. We focus on the structures of spHCN (both in the closed and open states) and mHCN2 (in the closed state), as the pore region of these channels has been experimentally investigated (1,24,25). We explore a wide range of possible configurations, analyzing physical quantities that do not depend critically on the exact three-dimensional (3D) structure, but are likely to capture basic differences between K⁺ and HCN channels, and, for the latter, between the closed and open states.

For the closed state, the templates could be in principle the available x-ray structures of the K⁺ channels, namely the KcsA K⁺ channel from Streptomyces lividans (2,26), the voltage gated K⁺ channel from Aeropyrum pernix (KvAP) (4), and the inward rectifier K⁺ channel from Burkholderia pseudomallei KirBac1.1 (5). However, as KvAP exhibits a wide-open conformation, thought to arise from crystal packing effects (4), only the other two structures were used.

Our model of the open state uses the MthK channel as a template (3,6).

The models are compared with structural constraints derived by electrophysiological experiments with wt and mutant channels reported previously (30) as well as with new experiments on Cd²⁺ blockage on spHCN and mHCN2 channels reported here. Models for the closed state derived from crystal structures are fully consistent with this data, however, those for the open state require the introduction of an appropriate constraint based on experimental data (30).

Based on our modeling, we provide a possible explanation for the different ionic selectivity between K⁺ and HCN channels and suggest that the bending of the S6 helix upon gating is significantly smaller than that occurring in MthK channels.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
Bioinformatics
Sequence alignment
The S5/S6 regions in the amino acid sequence of mHCN2 and spHCN channels were identified on the basis of the prediction of transmembrane helices, using the programs: DAS (http://www.biokemi.su.se/server/DAS (35)); tmap, (http://www.embl-heidelberg.de/tmap/tmap_info.html (36,37)); PHDhtm (http://www.embl-heidelberg.de/predictprotein (38,39)); and TopPred2 (http://www.biokemi.su.se/server/toppred2 (40)). The sequences including S5 and S6 turned out to be 368–477 in spHCN and 340–449 in mHCN2. The sequences of spHCN and mHCN2 channels were aligned with residues 26–124 of KcsA, 62–151 of KirBac1.1, and 19–98 of MthK, which correspond to the truncated sequence visible in the x-ray structures (2,3). The alignment was performed using the program ClustalW (41) (http://www2.ebi.ac.uk/clustalw). In the alignment, the following criteria were used: i), similarity of the P region (P-helix-loop) (1); ii), fully conservation of the G-Y-G motif in the filters (16,29); iii), alignment of S5 (spHCN) residues with TM1 in KcsA (2), KirBac1.1 (5), and MthK (3). iv), Alignment of S6 with TM2 in KcsA (2), KirBac1.1 (5), and MthK (3). Notice that a long gap (>11 residues) was inserted between the S5 and the pore-helix in the potassium channels to align the predicted transmembrane helices of the HCN channels with those of the templates.

Building of three-dimensional models
Three-dimensional models of the closed spHCN and mHCN2 and open spHCN were produced using Modeller 6v2 (42). The S5, S6, and the P-helix were restrained to adopt a helical conformation, and the noncrystallographic fourfold symmetry present in the KcsA, KirBac1.1, and MthK structures was enforced. The large extracellular loops, present in the two channels, have a very high variability and cannot be compared to existing regions in K⁺ channels of known 3D structure. The best-scoring models were selected for each channel. These models provided also the best stereochemistry, as monitored by ProcheckV3.5 (43) and WhatifV4.99 (44) programs. Models of the channel in the closed state were consistent with all the available experimental data (see Results section). In contrast, the model of spHCN in the open state was not in agreement with the estimates of distances of opposing C along the S6 helix (30). In the open state, the analysis of Cd²⁺ blockage of mutant channels and of Cd²⁺ containing proteins (45,46) suggests that the distance between opposing C's for Thr-464 must be 11 Å and 14 Å for Glu-461, Asn-465, and Gln-468. These distances were not in agreement with those obtained using the MthK channel as a template (Table 1). Therefore, the MthK channel was used as a template, but with the constraint that the distance between opposing C's of Thr-464 should be 11 Å. The experimental data on residues in positions 461, 465, and 468 were used for model validation.

View this table: [in this window] [in a new window]   TABLE 1  Distances of selected (opposite) C's along the S6 helix from spHCN structural models in the open and closed conformations

Number of H-bonds
To estimate the rigidity/flexibility of HCN channels, relative to the KcsA channel, the number of H-bonds was counted. Counting H-bonds in KcsA is straightforward, but some care must be used when homology models are investigated. Therefore, several models for the spHCN, mHCN2, and mHCN3 channels were generated using Modeller6v2 (42) with different initial conditions, so to explore a variety of possible side-chain configurations. The different initial conditions were achieved by using different random numbers. Once models were generated, an H-bond count was carried out on the P-helix-loop from Tyr-415 to Pro-436 in spHCN, Tyr-387 to Pro-408 in mHCN2, and Tyr-297 to Pro-318 in mHCN3. All N-O and O-O atoms separated by a distance between 2.195 and 3.350 Å (following Swiss-Pdb Viewer protocols; http://www.expasy.org/spdbv) were considered as putative H donors and acceptors, respectively. The configuration space of possible models was sampled in the following way: for each channel 10 new models were generated using modeller6v2 and H-bonds were counted and listed. The procedure of model generation was stopped when for a new set of 10 different models no new H-bond distribution appeared. The procedure stopped after the generation of six groups of 10 different models.

Electrostatic potential along the channel axis
To estimate the electrostatic profile in the absence of a reliable 3D structure, a variety (see later) of 3D models of mHCN1, mHCN2, mHCN3, and spHCN channels were generated and the electrostatic potential profile was computed by solving the Coulomb equation along the channels' z axis. To take into account the presence of water molecules (different from "bulk water") in the cavity and in the inner pore region, a dielectric constant of 4 was used. Because far from the filter, water becomes "bulk", i.e., the dielectric constant increases, electrostatic profiles only in the neighborhoods of the inner pore were considered. Partial charges derived from the AMBER force field were used (51). The different 3D models for each of the HCN channels were generated by considering all possible rotamers of Arg-405 (for mHCN2), of Lys-433 (for spHCN), of Ala-352 (for mHCN1), and Gln-315 (for mHCN3). Rotamers were modeled by using Swiss-Pdb Viewer tools (http://www.expasy.org/spdbv) and the distribution was as follow: 18, 16, 6, and 2 different rotamer configurations for Arg, Lys, Gln, and Ala, respectively.

Experimental section
Molecular biology and expression in Xenopus oocytes
The spHCN channel from sea urchin sperm (14) and the HCN2 channel from mouse (mHCN2) (16) were subcloned in the pGEM-HE expression vector. cRNA (50 ng) was injected into Xenopus laevis oocytes. Stage IV-VI oocytes were surgically removed from female frogs anesthetized by immersion in 0.3% 3-aminobenzoic acid ethyl ester, followed by digestion with 1 mg/ml collagenase solution containing 88 mM NaCl, 1 mM KCl, 0.82 mM MgSO₄, 2.4 mM NaHCO₃, 5 mM TRIS (pH 7.4) for 60 min. Injected eggs were maintained at 18°C in a standard Barth solution supplemented with 50 µg/ml gentamycin sulfate and containing (in mM): 88 NaCl, 1 KCl, 0.82 MgSO₄, 0.33 Ca(NO₃)₂, 0.41 CaCl₂, 2.4 NaHCO₃, 5 TRIS, pH 7.4 (buffered with tetramethylammonium (TMAOH)). During the experiments, oocytes were kept in a Ringer solution containing (in mM): 150 NaCl, 2.5 KCl, 1 CaCl₂, 1.6 MgCl₂, 10 HEPES-TMAOH, pH 7.4 (buffered with TMAOH). All reagents were from Sigma Chemicals (St. Louis, MO).

Electrophysiological recordings
Cell-free inside-out patches were obtained 2–4 days after cRNA injection, and data were acquired using a patch-clamp amplifier (Axopatch 200B, Axon Instruments, Foster City, CA) at room temperature (20–22°C). Borosilicate glass pipettes had resistances of 3–5 M in symmetrical standard solution. The solution in the patch pipette contained (in mM) 150 KCl, 10 HEPES, and 10 EGTA, buffered with TMAOH to appropriate pH 7.4. The solution bathing the intracellular side of the patch contained (in mM) 150 KCl, 10 HEPES, and appropriate amounts of CdCl₂. The intracellular bath solution was supplemented with 1 cAMP. Data were low-pass filtered at 1 kHz and acquired online at 5 kHz. Analysis was done using pClamp (Axon Instruments) and Origin software (Microcal Software, Northampton, MA). I_h currents were elicited by stepping the membrane voltage from 0 to –100 mV for 600 or 1400 ms for spHCN and mHCN2, respectively.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
Here we first describe the structural features of the closed and open state of HCN channels, as obtained by homology modeling. Comparison is then made with experimental data reported in the literature and performed here.

Closed state
The amino acid sequence in the pore region (TM1/S5, P-helix-loop, and TM2/S6) of the KcsA, KirBac1.1, MthK, spHCN, and mHCN2 channels are aligned in Fig. 1 A. The overall sequence identity (SI) between targets and templates is low (18% for all the templates). However, in the P-helix-loop (residues Tyr-415 to Pro-436 in spHCN and Tyr-387 to Pro-408 in mHCN2) SI increases significantly, and becomes 30% between targets and KcsA channel. The sequence alignment between K⁺ and HCN channels requires the inclusion of a gap of >11 residues in KcsA, KirBac1.1, and MthK channel, between residues Gly-53 and Ala-54 (KcsA). In HCN channels, this region corresponds to a long extracellular loop (highlighted in light gray in Fig. 1 A) absent in K⁺ channels.

Fig. 1 B illustrates the sequence alignment in the same region of the mHCN1, mHCN2, mHCN3, mHCN4, and spHCN channels. The overall SI is very high (of the order of 90%), but some significant differences are evident. In the inner pore region there is a cysteine conserved in all HCN channels (bold underlined cysteine residues in Fig. 1, A and B). According to the sequence alignment in Fig. 1 A and the 3D structure of the KcsA channel, these cysteines may form a ring at the intracellular mouth of the channel inner pore. The presence at the inner pore of a ring of cysteines able to oxidize and form disulfide bonds is a remarkable feature and some properties of these cysteines will be studied in spHCN and mHCN2 channels in the following sections. The inner pore region in HCN channels is composed of the residues C-I-G-Y-G-X, (in Fig. 1 B, CIGYG underlined residues; X residues shown in bold) where X can be positively charged, polar, or a neutral residue. X is a lysine in spHCN, an arginine in mHCN2 and mHCN4, but an alanine in mHCN1, and a glutamine in mHCN3. The S5 and S6 transmembrane helices are well conserved among the HCN channel family. In the activation gate region the motif GATCYA (highlighted in dark gray in Fig. 1 B) is fully conserved in HCN1-4 channels, but in the spHCN channel a tyrosine residue is replaced by a phenylalanine.

The large extracellular loop (highlighted in light gray in Fig. 1) has a similar sequence among HCN1-4 channels, but there are some differences between those and spHCN channel. For instance, in the middle of the loop there is a cysteine (indicated by a gray arrow in Fig. 1 B) conserved in all HCN channels, which is replaced by a serine in spHCN channels. This cysteine mediates the blockage of the I_h current by extracellular MTSEA and MTSES (58), indicating that the extracellular loop could approach the pore at a distance comparable to MTSEA length, i.e., 6 Å. As expected from a hydrophilic domain, residues forming the extracellular loops are polar and have a high probability for H-bond formation.

Models were constructed using homology modeling based on the sequence alignment shown in Fig.1 A (models are available at http://www.sissa.it/giorget/pdb_files). Final models were refined by a 1-ns MD simulation (see Methods). As expected, the 3D architecture of the K⁺ channels is maintained. The long loop regions are not shown because of the high variability and uncertainty of their location after using our modeling approach. The stereochemistry and quality of the final models, investigated with ProcheckV3.5 and WhatifV4.99 programs, validated the proposed models (Table 1 in Supplementary Material).

The inner pore
Homology models of the inner pore region of mHCN2 and spHCN channels are illustrated in Fig. 2, A and B, respectively. The C of residues (shown as balls) from His-397 to Arg-405 for the mHCN2 and from His-425 to Lys-433 for the spHCN channels are shown; in Fig. 2 A, a gray stick with a darker tip indicates the carbonyl groups. A marker of all known HCN channels is the presence of a cysteine residue two residues upstream of the filter motif GYG (Cys-428 for spHCN and Cys-400 for mHCN2; see bold underlined cysteine in Fig. 1, A and B). Given the tetrameric structure of these channels, these cysteines form an intracellular ring (Fig. 2, A and B). In mHCN2 and spHCN channels at the outer mouth of the pore there is an arginine and a lysine, respectively, forming an extracellular ring of positive charges. The intracellular ring of cysteines and the extracellular ring of positive residues are key features of the inner pore of mHCN2 and spHCN channels.

View larger version (30K): [in this window] [in a new window]   FIGURE 2  Homology model of the inner pore region of mHCN2 (A) and spHCN (B) channels in the closed state. Residues from His-397 to Arg-405, and from His-425 to Lys-433 are shown in panels A and B, respectively. The side chains of all residues except Arg-405 and Lys-433 are not shown. The rotamer space of Arg-405 and Lys-433 is shown as a shaded area: R1, R2 and K1, K2 indicate the extreme rotamers for Arg-405 and Lys-433, respectively. (C) Cysteine rotamer configurations for Cys-400 and Cys-428: S1, S2, and S3 indicate three different rotamer configurations. The homology model of the intracellular vestibule for mHCN2 and spHCN channels is the same. The solid line indicates the distance between C of opposing cysteines and the dotted lines indicate examples of distances between S atoms of opposing cysteines. Just two opposite subunits are shown.

We now compare our structural models of the pore with electrophysiological experiments of Roncaglia and co-workers (1) and experiments performed here. These experiments are summarized in the next section.

Stoichiometry of intracellular Cd²⁺ blockage
Cd²⁺ ions usually bind in cysteine-containing proteins to two or more (up to four) S atoms. In addition, copper phenantroline (CuP) enhances the formation of disulfide bonds between two S atoms by oxidization. Because both agents experimentally block the current, at least two Cys residues must be able to bind Cd²⁺ ions and form S-S bridges.

Roncaglia et al. (1) have shown that Cys-428 of the spHCN channel mediates the blockage by intracellular Cd²⁺ of the wt spHCN channel. Roncaglia et al. have also shown that the mild oxidizing agent CuP added to the intracellular medium caused an irreversible blockage of the wt channel. Both Cd²⁺ and CuP blockage were significantly reduced in the mutant channel C428S where cysteine was replaced by serine. Cd²⁺ blockage of the I_h current flowing through wt spHCN channels described by Roncaglia et al. and by Rothberg et al. is different: Rothberg and co-workers (6) reported that 20 µM Cd²⁺ did not produce any blockage of the I_h current in wt spHCN channels, whereas Roncaglia and co-workers (1) described a significant blockage of the same current by 50 µM or more Cd²⁺ ions. To resolve this apparent discrepancy, the dose dependence of intracellular Cd²⁺ blockage, of spHCN and mHCN2 channels, was determined. Fig. 3, A and B, reproduce I_h currents recorded in the presence of different amounts of intracellular Cd²⁺ ions obtained from spHCN and mHCN2 channels expressed in Xenous laevis oocytes. Fig. 3 C shows the time course of Cd²⁺ blockage of the I_h current in the spHCN channel, when larger amounts of Cd²⁺ were added to the intracellular medium. The stoichiometry of intracellular Cd²⁺ blockage in the spHCN (solid circles) and mHCN2 (open circles) channels is shown in Fig. 3 D. The experimental results indicated that the residual I_h current flowing through spHCN channels in the presence of 20 µM intracellular Cd²⁺ was 57 ± 22%; as a consequence there is a partial blockage of the I_h current flowing through spHCN channels at the concentration used by Rothberg et al. (6). When a cysteine replaced Glu-461, Thr-464, Asn-465, and Gln-468, Rothberg et al. (6) reported a complete blockage of the I_h current by micromolar amounts of intracellular Cd²⁺. In a more recent work, Rothberg and co-workers (59) probed intracellular Cd²⁺ effects on spHCN mutant channels L466C and Q468C at nanomolar concentrations. At these concentrations we never observed a blockage of the I_h in the wt channels.

View larger version (26K): [in this window] [in a new window]   FIGURE 3  (A and B) The Ih current traces obtained in the presence of indicated micromolar concentrations of intracellular Cd2+ from the spHCN and mHCN2 channels, respectively. The Ih current was elucidated by hyperpolarizing the membrane patch in inside-out configuration from the holding potential of from 0 to –100 mV for 600 or 1400 ms for spHCN and mHCN2 channels, respectively. The traces are leak corrected. (C) Time course of the intracellular Cd2+ effect on the normalized amplitude of the Ih current from spHCN channel. The amplitude of current at a given Cd2+ concentration was determined from current traces like those in panel A by their normalization with respect to the amplitude in the absence of Cd2+. (D) The dose dependence of the intracellular Cd2+ block of the spHCN and mHCN2 channels (•, , respectively). Data were fitted with the Hill function I(c) = 1/(1 + (c/IC50)N), where IC50 is the concentration necessary for the 50% block and N is the Hill coefficient. The best fitting was obtained with the values of N = 1 ± 0.2 and IC50 = 13 ± 1 µM for spHCN channel and N = 1.7 ± 0.1 and IC50 = 41.4 ± 0.7 µM for mHCN2 channel. Each experimental point represents mean ± SE from at least three experiments.

(1)

where c is the concentration of intracellular Cd²⁺. A least-mean-square fitting of the experimental data provides values and N = 1.7 ± 0.1 and IC₅₀ = 41.4 ± 0.7 µM for the mHCN2 channel. The fitting procedure provides instead a value of N = 1 ± 0.2 and IC₅₀ = 13 ± 1 µM for the spHCN channel. This value of N, suggests that one Cd²⁺ ion is necessary to block the spHCN channel. Instead, the data for mHCN2 show that for larger concentrations of Cd²⁺, the current traces seem not to be in equilibrium. Therefore, it may be appropriate to assume a noncooperative Cd-block in the pore.

In this respect, it is interesting to notice that Rothberg and co-workers (59) have fitted the experimental data of Cd²⁺ blockage on mutant channel Q468C with Eq. 1 using an N-value of 4 and IC₅₀ of 72 nM. This result indicates that four Cd²⁺ ions block one mutant channel Q468C with a much higher affinity than the wt.

Comparison with experimental data
We now compare our models with the structural properties of the cysteine ring inferred by the electrophysiological experiments illustrated in Fig. 3 and in Roncaglia et al., 2002. Analysis of Cd²⁺ and Cys-containing protein structures performed by Krovetz et al. (45) and Giorgetti et al. (46), suggest that the range of most probable distances between the C for a pair of cysteines binding to Cd²⁺ falls between 4 and 9 Å (46,60,61) and between 4 and 7 Å for pairs of cysteines forming S-S bridges (46,62). Because the approximate diameter of Cd²⁺ is 1.8 Å and considering thermal fluctuations (63) the maximum distance d(C@Cys-C@Cys) is expected to be of 11 Å or less (45,46,60,61,64,65). In our models, d(C@Cys-C@Cys) is 9 Å and 6.5 Å for opposite and adjacent cysteines, respectively, showing fully consistency with the experiments (Fig. 2 A). We have also investigated the configuration space of all possible rotamers of Cys-400 and Cys-428; three limiting rotamers were illustrated in Fig. 2 C. The S atom can point in several different directions: for some rotamers the distance of opposite S is 7 Å and for other rotamers 10 Å. In our models the expected distance for a Cd²⁺-S interaction is 2.5 Å. This distance is in agreement with that determined by x-ray diffraction between S atoms of cysteines forming the binding pocket for divalent cations (66,67).

Flexibility of the inner pore region
The 3D structure of the P-helix-loop of the proposed models of HCN channels in the close state is naturally very similar to that of the KcsA K⁺ channel (residues from Tyr-62 to Pro-83; see Fig. 4, A and B). However, the number of H-bonds with residues in the outer mouth of the channel decreases significantly on passing from the KcsA (26 H-bonds) to an average of 21.3 ± 1.5, 21.1 ± 1.4, and 21.1 ± 1.6 H-bonds for spHCN, mHCN2, and mHCN3 channels, respectively (shown as histograms in Fig. 4 C). We have generated 60 models for each of the pores of spHCN, mHCN2 (Fig. 4 B), and mHCN3 channels so as to include a wide range of side-chain configurations of the P-helix-loop. Residues from Tyr-415 to Pro-436 for spHCN and from Tyr-387 to Pro-408 in mHCN2 (indicated with a black thick line in Fig. 1 B) were considered. As shown in Fig. 4, A and B, the KcsA channel has more H-bonds (dotted lines) in the upper part of the inner pore. As a consequence, the inner pore of spHCN and mHCN2 channels is expected to be less rigid than that of K⁺ channels. We provide further evidence of the different rigidity of the inner pore of these channels by calculating the ß-factors based on MD simulations of KcsA, spHCN (closed conformation), and mHCN2 channels. To account for the presence of K⁺ ions in the inner pore, three K⁺ ions were located in the positions: S1, S3, and S_cav (68) (see Fig. 1 in Supplementary Material). Although positions S1 and S3 are not stable in MD simulations longer than 1 ns (68), it was known that substitution of Thr-75 (KcsA) with a cysteine increases the instability of S2-S4 configurations (69) and, as HCN channels have an endogenous cysteine in position corresponding to Thr-75 of the KcsA channel, it was decided to position K⁺ ions only in S1 and S3. Fig. 5 shows that the fluctuations in the filter region are much larger in HCN channels than in the KcsA channel, confirming the different flexibility of the two families of ionic channels.

View larger version (27K): [in this window] [in a new window]   FIGURE 4  H-bond distribution in KcsA and mHCN2 channels in the P-helix-loop region (inner pore and P-helix), in the closed conformation. (A) Three-dimensional crystallographic structure of one subunit of the KcsA channel (2); (B) 3D model of one configuration of the mHCN2 channel. In panels A and B only side chains capable of H-bond formation are shown. H-bonds are indicated by dotted lines. The stabilizing H-bond network formed by Glu-71 and Asp-80 in the KcsA channel is absent in the HCN channels. (C) H-bond distribution for mHCN2 (black), for spHCN (dotted-dashed), and for mHCN3 (dotted) channels. The average number of H-bonds was 21.3 ± 1.5, 21.1 ± 1.4, and 21.1 ± 1.6 in the P-helix-loop region for spHCN, mHCN2, and mHCN3 channels, respectively. The configurations were obtained as described in the Methods section.

View larger version (21K): [in this window] [in a new window]   FIGURE 5  ß-Factor of residues in the P-helix, the pore, and S6. Data obtained from 1-ns MD simulation for KcsA (pink), mHCN2 (green), spHCN (black), and KcsA E71V (red). Residue numbering corresponds to that of the spHCN channel (see the alignment in Fig. 1). The ß-factor of the open and closed configuration of the spHCN channel superimpose and only data for the closed state are shown.

The electrostatic potential profile within the pore
As shown in Fig.1 B the inner pore of HCN channels is conserved (underlined residues) with the exception (bold residues) of the Lys-433 in spHCN, which becomes an arginine in mHCN2 and mHCN4, an alanine in mHCN1, and a glutamine in mHCN3. The outer ring of positively charged residues in spHCN, mHCN2, and mHCN4 is expected to affect electrostatics in the pore and in the outer vestibule.

To estimate the electrostatic potential profile in the absence of a reliable 3D structure, a variety of 3D models of mHCN1, mHCN2, mHCN3, and spHCN channels were generated and for each of them the electrostatic profile along the pore axis was computed. All models had the same backbone, illustrated in Fig. 2, but differed in the direction of side chains of the outer ring of arginine, lysine, alanine, and glutamine. In Fig. 2, A and B, the allowed rotamers for Arg-405 and for Lys-433 are shown as shaded regions. The two extreme rotamers are marked R1, R2 and K1, K2, respectively. As discussed in the Methods section for each HCN channel a family of electrostatic potential profiles was obtained.

Fig. 6 illustrates the electrostatic potential profile along the pore axis for different rotamers of the mHCN1, mHCN2, mHCN3, and spHCN channels. For HCN channels with an alanine or a glutamine the electrostatic potential profiles for the different rotamers are very similar, as expected from residues not bearing a net charge. As a consequence, the electrostatic potential profile for mHCN1 and mHCN3 channels is more negative near the extracellular vestibule than that of HCN channels with a positively charged ring, such as spHCN and mHCN2. The electrostatic potential profile for the mHCN2 and spHCN channels varies from –60 to –40 kcal/mol, according to the orientation of the side chain of the arginine or lysine. When side chains point toward the lipid phase (configurations R2 and K2 of Fig. 2 and corresponding potential profile in Fig. 6) the electrostatic potential profile is more negative than when side chains point upward and are almost parallel to the pore axis (configurations R1 and K1).

View larger version (24K): [in this window] [in a new window]   FIGURE 6  Electrostatic potential profile along the channel z axis for different rotamer configurations of mHCN2 (red), mHCN1 (violet), mHCN3 (green), and spHCN (blue) channels. The plot indicates the position of the different carbonyl oxygen from inner pore. The electrostatic profiles corresponding to the limiting rotamers—shown in Fig. 3—are labeled as R1, R2 and K1, K2. Outside of the filter region the potential tends to zero (results not shown for clarity).

View larger version (22K): [in this window] [in a new window]   FIGURE 7  The open (red) and closed (blue) channel conformations. (A) Templates used for homology modeling for the open (MthK channel in red) and closed (KcsA channel in blue) channel. (B) Final models of the spHCN channels in the open (red) and closed state (blue). (C) Superimposition of the initial (red) and final (gray) MD simulation snapshots (open configuration of the spHCN channel). The bending of the S6 transmembrane helix associated with gating in K+ channels is 30° whereas in our model of spHCN channels the bending is 12°. Only two subunits are shown. Selected residues are shown as spheres. Distances between opposite C–C atoms obtained from the modeling are compared with estimates from experimental data in Table 1.

To test the stability of the proposed model, 1-ns MD simulations were performed for spHCN model in the open pore configuration (see Methods). Fig. 7 C shows the initial (red) and the final (gray) structures superimposed; the overall root-mean-square deviation was 1.8 Å, indicating that the model was stable. The displacements between the initial and the final configurations for selected atoms are shown in Table 1. A comment on the structure of the pore filter in the open state is necessary. The functional properties of HCN and spHCN channels are intermediate between those of K⁺ and CNG channels. However, the filters of these channels exhibit very different properties: the filter of K⁺ channels does not have major conformational changes during gating (with the exception of molecular events related to C-type inactivation (29)), whereas that of CNG channels does rearrange upon gating (71–73). Thus, we can draw no conclusions on the structure of the pore obtained by MD simulation, albeit this structure is stable in the timescale investigated.

Fig. 8 reproduces the proposed molecular structure around the gate located between Thr-464 and Gln-468 in the closed (A) and open state (B). The Connolly surface, in gray, obtained using the VMD1.8.2 program (http://www.ks.uiuc.edu/research/vmd/) with a probe radius of 0.90 Å, indicates the accessible free space. It is evident that the gate is closed in panel A whereas it is open in panel B.

View larger version (65K): [in this window] [in a new window]   FIGURE 8  (A and B) The gate of the spHCN channel in the open and closed configuration, respectively. In gray is the Connolly surface obtained using the VMD1.8.2 program (http://www.ks.uiuc.edu/research/vmd/) with a probe radius of 0.90 Å. (C) Cd2+ block in mutant channel T464C in the open state. Distances between C of opposite cysteines in the open state were 11 Å. (D) Cd2+ block in mutant channel Q468C in the closed state. Distances between C of opposite cysteines in the closed state were 15 Å.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
We have constructed the structural models of spHCN and mHCN2 channels' transmembrane domains in the closed state, and that of spHCN in the open state. These models were obtained by homology modeling using the crystal structure of the KcsA channel (2) and KirBac1.1 (5) as templates for the closed conformation and the MhtK channel (3) for the open conformation, using the assumptions outlined in the Introduction. In this section, we discuss insights on structure/function relationships in HCN channels based on the proposed models.

The closed state
The closed state models of spHCN and mHCN2 were validated against structural constraints inferred by experimental data available in the literature (1). The models were found to be consistent with all of the data. This supports the use of the models for investigation of structure features of the filter, the P-helix, and S6.

Flexibility of the inner pore and ion selectivity
The great majority of HCN and K⁺ channels share the GYG motif in the pore. The ionic selectivity of these channels, however, is different: K⁺ channels are primarily permeable for K⁺ and Rb⁺ ions and the permeability ratio P_Na/P_K is below 0.1. On the contrary, a distinct feature of all HCN channels is a significant permeability also to Na⁺ ions, with P_Na/P_K 0.3.

Here we argue that the lower selectivity of HCN channels relative to that of K⁺ channels can be ascribed, at least in part, to the different flexibility of the selectivity filter, which is one of the primary determinants of ionic selectivity (31). In this view, K⁺ channels have a radius of 1.5 Å, large enough for the permeation of alternating, single-file K⁺ ions and water molecules, as confirmed by high-resolution x-ray crystallography (2). In HCN, the inner pore is relatively flexible, as suggested by a calculation of ß-factors based on MD simulations on these channels (Fig. 5). Thus, partly hydrated Na⁺ ions will be able to permeate through the selectivity filter. If, instead, the inner pore is less flexible, as is the case for the KcsA channel (Fig. 5), Na⁺ ions, with several water molecules attached (74) will not be able to pass and the channel will be highly selective for K⁺ ions. In turn, the larger flexibility of HCN channels could be ascribed, in part, to the smaller number of H-bonds in the P-helix and the inner pore region in the HCN channels relative to the K⁺ channel (Fig. 4). However, this cannot be the only key factor for flexibility as shown by an additional simulation on the E71V K⁺ channel mutant, which exhibits a lower number of H-bonds relative to wt and yet similar flexibility (Fig. 5) and permeability ratio (70) as wt. Thus, results from "static" crystal structures (H-bond) do not necessarily allow predictions about origin and determinants of protein flexibility or rigidity, as has been pointed out very recently by Allen et al. (75).

The cysteine rings
All HCN channels bear a cysteine in the inner pore region, as shown in Fig. 1 B (bold underlined Cys). Cd²⁺ ions from the cytoplasmic side of the membrane block spHCN and mHCN2 channels with an IC₅₀ of 13 ± 1 and 41.4 ± 0.7 µM, for spHCN and mHCN2 channels, respectively (see Fig. 3). These results show that these cysteines are located at the intracellular side of the inner pore and form the cytoplasmic vestibule (1).

In CNGA1 channels from bovine rods, Cys-428 of spHCN channels is replaced by a threonine. When this threonine is replaced by a cysteine, the resulting mutant channel, T360C, is blocked by intracellular Cd²⁺ ions in the open state with an IC₅₀ of 25 µM and a value of N = 2 (M. Mazzolini, A. V. Nair, A. Giorgetti, and V. Torre, unpublished data). Therefore, the cysteine ring (Fig. 2 B) in the inner pore region of HCN channels and of mutant channels T360C of CNGA1 from bovine rods is likely to have a similar molecular arrangement.

One Cd²⁺ ion is bound at least to three cysteines in the pore of the mutant channel T464C in the open state (6). As a consequence one Cd²⁺ is likely to coordinate to the four exogenous cysteines of mutant T464C, as shown in Fig.8 C. Cd²⁺ blockage of the mutant channel Q468C of the spHCN (59) is different: indeed, these authors have fitted the experimental data with Eq. 1, but with an N-value of 4 and an IC₅₀ value of 72 nM. This result indicates that four Cd²⁺ ions block one mutant channel with a much higher affinity. The physical arrangement of Cd²⁺ ions and of the cysteines is illustrated in Fig. 8 D.

It is useful to stress the different blockage by Cd²⁺ ions in the pore of spHCN and CNGA1 channels and in the channel mutant Q468C; Cd²⁺ blockage in mutant T360C of the CNGA1 channel and in the pore of the spHCN channel (see Fig. 3) occurs presumably because the Cd²⁺ ions physically occlude the pore, whereas blockage of the mutant Q468C of the spHCN channel is of the lock-closed type in which the channel becomes locked in the closed configuration.

Under normal conditions HCN channels are open and therefore these cysteines do not form S-S bonds. The distance between opposite Cs of these cysteines is 9 Å, suggesting that S-S bridges can be readily formed in the presence of an oxidizing medium. As a consequence, the inner pore of HCN channels is likely to be a sensor of the intracellular oxidizing/reducing conditions. Given the limitation of homology modeling it is very difficult and almost impossible to provide an explanation why Cd²⁺ ions bind with different affinities the ring of the native Cys-428 and of exogenous cysteines introduced in position 464 or 468.

The pore electrostatics
As shown in Fig. 6 the electrostatic potential profile obtained from our modeling is slightly different for the various HCN channels. In agreement with the K⁺ pore structure the carbonyl groups of the signature GYG point toward the pore axis and provide an attracting binding site for cations. The computed electrostatic potential profile in the outer part of the pore is more negative for mHCN1 and mHCN3 channels. Given the positive charge of Arg and Lys present in the outer mouth of mHCN2, mHCN4, and spHCN channels the electrostatic potential profile for all considered rotamers was consistently less negative than for mHCN1 and mHCN3. Therefore, differences of single-channel properties conductance among HCN channels could be expected. Given the small single channel conductance of HCN channels(1,6,30) it may be difficult to experimentally establish these differences.

The open state
As the homology model-based structure of the spHCN channel in the open state is not consistent with the available experimental data in the S6 helix, it has been refined by introducing experimentally derived constraints, which involve residues present in the S6 helix. Notice that the constraints between the C's belonging to opposite T464 are not consistent with the MhtK-based homology model of the channel, even including the fact that the channel may fluctuate between the open state (modeled based on the MthK channel) and the closed state (modeled based on the KcsA channel). Indeed, to satisfy such constraints one should suppose that the channel spends as much as 90% of its time in the closed state in the presence of cyclic nucleotides. Here we discuss structural features of S6 helices, for which comparison is made with the correspondent structural elements in K⁺ channels.

The transition between the closed and open state in bacterial K⁺ channels is associated with a significant bending of the S6 transmembrane helix around a hinge located around a conserved glycine (see Fig. 1 A). This bending in bacterial K⁺ channels is associated with an angular rotation of 30° around the hinge (3,29). The gating in spHCN channels seems to occur by similar mechanisms (6), but with a smaller rotation of 12°. As shown in Fig. 8, the gate in spHCN channels seem to be located near Gln-468: indeed the mutant channel T464C is blocked by Cd²⁺ ions only in the open state, suggesting that Thr-464 is accessible from the intracellular side only in the open state (6). On the contrary the mutant channel Q468C is powerfully blocked by Cd²⁺ ions also in the closed state and therefore Gln-468 is accessible from the intracellular side of the membrane also in the closed state (59). At the speculative level, we anticipate that the gating in the large superfamily of voltage-gated ionic channels is likely to be caused by a movement of the S6 transmembrane helix, but its extent and properties are likely to depend on the specific ionic channel.

We close this section by using our model to provide a rationale for current measurements on adducts between spHCN mutant channels and Cd²⁺, which shows that: i), submicromolar quantities of Cd²⁺ ions greatly slow down closure of L466C, once opened at negative voltages (59). ii), In the presence of Cd²⁺ ions, the double mutant channel H462C-L466C once opened remains locked in the open configuration.

In our model, residues His-462 and Leu-466 do not point toward the channel axis and instead face toward other channel domains, such as S5. The spHCN channel has two native cysteines, Cys-473 and Cys-469 in S5, which according to our homology model are approximately at the same height of His-462 and Leu-466 in S6. Because S6 is relatively far in the closed state, the distance between His-462 and Cys-473 is as large as 13 Å and that between Leu-466 and Cys-469 14 Å. Therefore, in the closed state there is no binding site for Cd²⁺ ions in mutant channel H462C-L466C, as shown in Fig. 9 A. In the open state, these distances become shorter and approach values between 10 and 11 Å so that one Cd²⁺ ion can tightly bind to cysteines in position 462, 466, 473, and 469, stabilizing, therefore, the open channel configuration, as shown in Fig. 9 B.

View larger version (19K): [in this window] [in a new window]   FIGURE 9  Proposed mechanism of Cd2+ binding to the double mutant H462C-L466C. (A) In the closed state, the distance between C @Cys-462 and C@Cys-473, located on S5, is 13 Å, whereas and the distance between C@Cys-466 and C@Cys-469 in S5 is 14 Å and, therefore, too large for Cd2+ ion binding. (B) In the open state, these two distances are reduced to 10 and 11 Å, respectively, so that one Cd2+ ion can bind and lock the channel.

	SUPPLEMENTARY MATERIAL

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
An online supplement to this article can be found by visiting BJ Online at http://www.biophysj.org.

	ACKNOWLEDGEMENTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
We thank Anna Tramontano for her comments on the article and Iain McLay for very useful discussions.

We thank GlaxoSmithKline for financial support. This work was also supported by a Human Frontier Science Program grant (RGP0054/2002) to V.T., and a grant from the Regione Friuli Venezia Giulia and CIPE (GRAND).

	FOOTNOTES

 
P. Mistrik's present address is University of Munich, Dept. of Pharmacy, Center for Drug Research, Munich 81377, Germany.

A. Giorgetti's present address is Biocomputing, Dept. of Biochemical Sciences, University of Rome ‘La Sapienza’, Rome, Italy.

Submitted on May 3, 2004; accepted for publication May 16, 2005.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION SUPPLEMENTARY MATERIAL ACKNOWLEDGEMENTS REFERENCES

 
1. Roncaglia, P., P. Mistrik, and V. Torre. 2002. Pore topology of the hyperpolarization-activated cyclic nucleotide-gated channel from sea urchin sperm. Biophys. J. 83:1953–1964.[Abstract/Free Full Text]

2. Zhou, Y., J. H. Morais-Cabral, A. Kaufman, and R. MacKinnon. 2001. Chemistry of ion coordination and hydration revealed by a K+ channel-Fab complex at 2.0 A resolution. Nature. 414:43–48.[CrossRef][Medline]

3. Jiang, Y., A. Lee, J. Chen, M. Cadene, B. T. Chait, and R. MacKinnon. 2002. Crystal structure and mechanism of a calcium-gated potassium channel. Nature. 417:515–522.[CrossRef][Medline]

4. Jiang, Y., A. Lee, J. Chen, V. Ruta, M. Cadene, B. T. Chait, and R. MacKinnon. 2003. X-ray structure of a voltage-dependent K+ channel. Nature. 423:33–41.[CrossRef][Medline]

5. Kuo, A., J. M. Gulbis, J. Antcliff, T. Rahman, E. Lowe, J. Zimmer, J. Cuthbertson, F. M. Ashcroft, T. Ezaki, and D. A. Doyle. 2003. Crystal structure of the potassium channel KirBac1.1 in the closed state. Science. 3000:1922–