Energetics of Cyclic AMP Binding to HCN Channel C Terminus Reveal Negative Cooperativity

Abstract

Cyclic AMP binds to the HCN channel C terminus and variably stabilizes its open state. Using isothermal titration calorimetry, we show that cAMP binds to one subunit of tetrameric HCN2 and HCN4 C termini with high affinity (∼0.12 μm) and subsequently with low affinity (∼1 μm) to the remaining three subunits. Changes induced by high affinity binding already exist in both a constrained HCN2 tetramer and the unconstrained HCN1 tetramer. Natural “preactivation” of HCN1 may explain both the smaller effect of cAMP on stabilizing its open state and the opening of unliganded HCN1, which occurs as though already disinhibited.

Introduction

Binding of cAMP to the cytoplasmic side of the HCN channel makes it easier to open (1); in the heart, this speeds rate and enhances conduction (2). In the sinoatrial node, opening is facilitated with a half-maximal cAMP concentration of 0.2 μm (1) and is marked by a depolarizing shift in the voltage range over which the channel opens (3). Each of four mammalian HCN isoforms (HCN1–4) (4–7) contains a cyclic nucleotide-binding domain (CNBD)³ in the cytosolic C terminus and has significant sequence homology to voltage-gated potassium channel subunits. Thus, HCNs are predicted to have six transmembrane helices (S1–S6) with a cytosolic N and C terminus and to combine as tetramers to form the ion-conducting channel containing four equivalent cAMP-binding sites.

The structure of a fragment of the HCN2 C terminus was solved by x-ray crystallography, which showed that this structure forms a symmetrical tetramer (8). The structure, hanging below the pore, is derived from sequences that begin just after the end of the S6 transmembrane segment and includes both the CNBD and the C-linker, a region that connects the CNBD to the inner side of the channel pore. Cyclic AMP is thought to bind and induce a conformational change that propagates through the CNBD as well as to the C-linker of an adjacent subunit (9, 10), by interactions of select α helices, to form a “gating ring” that disinhibits channel opening (8, 11, 12). Evidence for the formation of a gating ring comes from studies of the purified HCN2 C terminus using analytical ultracentrifugation (8). In the absence of cAMP, monomers are predominantly seen, whereas the proportion of multimers, especially tetramers, increases significantly when cAMP is present.

Curiously, facilitation of opening by cAMP differs among the mammalian HCNs. HCN2 and HCN4 opening is strongly facilitated by cAMP, whereas that of HCN1 and HCN3 is only weakly affected or not affected at all, respectively (12, 13). This variation among the isoforms is thought to arise, at least in part, from subtle differences in primary structure of the CNBD and the C-linker (12, 14, 15). Notably, opening is facilitated even in chimeric HCN2 and HCN4 channels containing the CNBDs from HCN1 and HCN3 (14, 15), suggesting that all HCNs bind cAMP. Nevertheless, the molecular basis for the differences in cAMP effect among the isoforms remains unknown. Here, we set out to directly quantify and compare the energetics of cAMP binding among HCN1, HCN2, and HCN4 using the C terminus in solution.

EXPERIMENTAL PROCEDURES

HCN Protein Purification and Mutagenesis

HCN constructs were cloned in a modified pET28 vector (Novagen) containing an His₆ tag, maltose-binding protein, and a cleavage site for the tobacco etch virus protease denoted as an HMT tag. The constructs utilized were mouse HCN1J (residues 390–645), ΔAB-HCN1J (432–645), mouse HCN2J (443–645), ΔAB-HCN2J (458–645), rabbit HCN4J (522–724), and ΔAB-HCN4J (532–724). The R591E mutation in HCN2J was obtained using the QuikChange protocol (Stratagene). GCN4LI, a leucine zipper protein that forms an artificial parallel tetramer when amino acids in the “a” and “d” positions of the coiled-coil are mutated to a leucine and an isoleucine, respectively (16), and HCN2J were amplified by PCR with a BamHI site and an additional 5 glycines at the 3′ and 5′ end, respectively, making a 10-glycine linker between the two pieces. After ligation, the fusion construct was amplified and cloned into the HMT vector.

All proteins were expressed in Escherichia coli Rosetta (DE3) pLacI grown in 2× YT media at 37 °C, except for HCN1J and ΔAB-HCN1J, which were grown at 25 °C, and induced with 0.4 mm isopropyl-1-thio-β-d-galactopyranoside. Cells were lysed by sonication in 250 mm KCl + 10 mm HEPES (pH 7.4) (buffer A), except in the case of HCN1J and GCN4LI-HCN2J, where 500 mm NaCl plus 30 mm HEPES (pH 7.4) was used, supplemented with 100 mm PMSF, 25 mg/ml lysozyme, 25 mg/ml DNase I, 10% glycerol, and 0.1 m EDTA. Cell debris was removed by centrifugation for 30 min at 35,000 × g, and the filtered supernatant was loaded onto a Talon (Clontech) column, washed with buffer A, and eluted with buffer A + 500 mm imidazole. His-tagged tobacco etch virus protease was used to cleave the tag for 2 h at room temperature and dialyzed against buffer A + 10 mm β-mercaptoethanol. To remove the tag and tobacco etch virus protease, the protein solution was rerun on talon resin. For GCN4LI-HCN2J, the tags were not removed and run on an amylose column instead. The flow-through was collected, dialyzed against 10 mm KCl, 20 mm MES (pH 6.0) + 10 mm β-mercaptoethanol (buffer C), and further purified on a Resource S column (GE Healthcare) in buffer C with a linear gradient to 40% buffer D (1 m KCl, 10 mm MES (pH 6.0) + 10 mm β-mercaptoethanol) over 40 column volumes. In the case of the GCN4LI-HCN2J, a Resource Q column (GE Healthcare) was used with 10 mm KCl, 10 mm Tris (pH 8.0) with a linear gradient to 40% buffer 1 m KCl, 10 mm Tris (pH 8.0) over 40 column volumes. Protein purity was confirmed by SDS-PAGE and MALDI-TOF mass spectroscopy. Homogeneity of the protein was confirmed on HiLoad Superdex 200 (GE Healthcare) (see supplemental Fig. 1).

Direct Measurements of Cyclic Nucleotide Binding by Isothermal Titration Calorimetry (ITC)

Samples were concentrated and dialyzed against 150 mm KCl, 10 mm HEPES (pH 7.4), 10 mm β-mercaptoethanol overnight at 4 °C. Titrations were performed in an ITC200 (GE Healthcare) at 25 or 10 °C, using a concentration of cAMP that was 10× that for the HCN protein and 1-μl injections. Protein concentrations were determined by absorption at 280 nm under denaturing conditions, and extinction coefficients were calculated. The binding isotherms were analyzed using sequential models and the two independent binding site models implemented in MicroCal Origin 7.0.

Binding isotherms are described by the following equation c = K_a[M]N, where K_a is the association constant, [M] is the total protein concentration in the cell, and N is the stoichiometry of interaction. The value for c was between 1 and 1000 for all of the data used for this study, which is generally considered to yield data that is accurate for enthalpy changes that are >10 kcal/mol (17).

Dynamic Light Scattering (DLS)

Samples were centrifuged for 10 min to remove dust particles. DLS measurements were performed on a DynaPro plate reader (Wyatt Technology). Acquisition was performed at 25 °C with Dynamics 7.0 with five scans of 5 s each. Analysis was performed using Dynamic 7.0 and Origin 7.0. Values for molecular weight, along with associated values for polydispersity, were obtained by DLS. Molecular weights of oligomeric forms of protein obtained by DLS were easily separated from aggregated protein, which was identified by its comparatively large value for molecular weight.

RESULTS

cAMP Binds to Cyclic Nucleotide-binding Domain of HCN2 C Terminus with Apparent High and Low Affinity

We examined the binding of cAMP to the isolated C terminus of the HCN2 channel, a region corresponding to the first six α helices of the C-linker and the three α helices and eight β-barrel strands that constitute the CNBD. Examination of this region alone, which we refer to as HCN2J from the initial study of this region (8), allows us to isolate the actual binding event, which is not possible in full-length channels. Although inherently limited because of the absence of the transmembrane and N-terminal regions of the channel, use of the C-terminal fragment also avoids reciprocal interactions between the C terminus and transmembrane portions of the full-length channel that could influence binding. Ample quantities of HCN2J were obtained, and the binding of cAMP to HCN2J was assessed by ITC (18).

Adding cAMP to purified HCN2J at 100 or 200 μm produced two clear transitions, which could readily be fit with a two independent binding site model (Fig. 1A) and yielded binding affinities of ∼1 and 0.1 μm (Fig. 1B). This model gave a better fit than the sequential model for four binding sites, also implemented in MicroCal Origin 7.0, in which the initial binding of the ligand to one site affects the affinity of the others. These two models, as applied in Origin, are described in detail in the supplemental material, along with examples of how each model fit the data (supplemental Fig. 2) and a state diagram illustrating a possible molecular interpretation for the negative cooperativity (supplemental Fig. 3).

FIGURE 1.

Cyclic AMP binds to HCN2J with an apparent high and low affinity. A, plots of heat produced by ITC upon progressive injections of cAMP into 200 μm HCN2J. In the upper plot, the broadening of peaks causes an initial increase in the overall heat per injection. The solid line through the values of the lower plot represents a two independent binding site model, which yielded values for affinity and energetics (ΔG, ΔH, and ΔS) (supplemental table). B, bar graph comparing high and low cAMP binding affinity as determined from the fit in A. Error bars indicate S.D. C, bar graph comparing energetics between the low and high affinity binding events, determined from the fit in A. Error bars indicate S.D. D, plots of heat produced by ITC upon progressive injections of cAMP into 200 μm HCN2J HCN2J-R591E.

The energetics of binding were different for each binding event (Fig. 1C); the apparent low affinity binding event was driven by favorable enthalpy and unfavorable entropy, whereas both were favorable for the apparent high affinity binding event. Thus, the biphasic pattern is due to release of heat specific to each of two binding events, both of which are exothermic but proceed and saturate over different and well separated ranges of cAMP concentration. The lower affinity binding event also occurs with a more negative enthalpy, resulting in a falling phase. The rising phase that follows occurs as both sets of binding sites saturate and the amount of heat released decreases.

The stoichiometry of binding, which is the ratio of cAMP to individual HCN2J subunits, was different between binding events, with values of ∼0.25 and ∼0.75 (supplemental table) for the high and low affinity events, respectively. Because these values for stoichiometry sum up to an overall stoichiometry of ∼1, both binding events likely occur to the same site with altered geometries.

For the low affinity binding event, the values for c are ∼114, and the values for ΔH are >15 kcal/mol, within the optimal experimental guidelines suggested by Wiseman et al. (17). However, for the high affinity binding event, the values for c are high (>400), and the values for ΔH are <6 kcal/mol. To examine the profile of heat changes due only to the high affinity binding event, the μcal/mol versus molar ratio curves were simulated with the experimental data obtained for only this event. These simulated curves show that the decrease in heat over time was comparatively steep (curves not shown). Thus, although the data are within the acceptable range of c-values and the fit to the two independent binding site model is very good, it is probable that the value obtained from the fitting for the high affinity binding event is not as accurate as that for the low affinity binding event (19).

Lowering temperature (from 25 to 10 °C) still resulted in two transitions and affinities much like those at the higher temperature, although the total amount of heat released was considerably smaller and the double transition was less pronounced (supplemental Fig. 4; supplemental table). To determine whether both events reflected binding to the CNBD, we introduced the R591E mutation, which is located in the highly conserved phosphate-binding cassette and coordinates the cAMP phosphate group (8, 14, 20–22). This mutation eliminated both binding events, consistent with two transitions resulting from binding of cAMP to the known CNBD (Fig. 1D).

High Affinity Binding of cAMP Occurs in Concert with Tetrameric Forms of HCN2J

One of the two binding events could have been due to promotion of self-association of HCN2J by cAMP binding to form dimers and tetramers, which has been shown previously by analytical ultracentrifugation (8, 11). To minimize the extent of HCN2J self-association, ITC measurements were repeated using lower concentrations of protein. Because self-association requires the C-linker, ITC measurements were also carried out using HCN2J lacking the first two α helices of this region (A′, B′) (ΔAB-HCN2J). Indeed, cAMP addition to lower concentrations of HCN2J and to ΔAB-HCN2J resulted in only the lower affinity binding event (Fig. 2, A and B), both with a stoichiometry of 1 (supplemental table), suggesting that self-association was necessary for the appearance of a high affinity binding event. Although the ΔH values were <6 kcal/mol for the single binding event of 10 and 25 μm HCN2J, the c-values obtained were ∼8 and ∼19, respectively, which are within the optimal range.

FIGURE 2.

Low affinity binding of cAMP remains at low HCN2J concentrations and when self-association is eliminated. A, plots of heat produced by ITC upon progressive injections of cAMP into indicated concentrations of HCN2J and 200 μm ΔAB-HCN2J, as indicated. The solid line through the values of the lower plots represents a single binding model for all of the above except for 100 μm HCN2J, which was fit with a two independent binding site model. The fits yielded values for affinity and energetics (ΔG, ΔH, and ΔS) (supplemental table). Note that, along with an experiment using 10 μm HCN2J on the far left, data are plotted from an experiment in which cAMP was titrated into buffer lacking HCN protein (upper plot, gray trace; lower plot, gray points); no discernible heat is released when cAMP is added to buffer alone. B, bar graph comparing affinities among the single transitions determined from the fits in A and the lower affinity determined from the two independent binding site model fit for 100 μm HCN2J in A. Error bars indicate S.D. C, plot of apparent molecular mass versus concentration of HCN2J with or without a saturating concentration of cAMP (2 mm), as indicated, determined by DLS. At 200 μm, open (− cAMP) and solid (+ cAMP) circles overlap. The molecular mass of 200 μm ΔAB-HCN2J, with or without cAMP, is also indicated in the plot. Error bars indicate S.D.

To approximate the extent of self-association and assess its contribution to the high affinity binding event, DLS of HCN2J, at a range of concentrations, and ΔAB-HCN2J (200 μm) was performed with or without a saturating concentration of cAMP (Fig. 2C). The apparent molecular mass obtained for ΔAB-HCN2J was ∼26 kDa, close to the calculated mass of the ΔAB-HCN2J monomer (22.4 kDa); this mass was unaffected by cAMP, consistent with the lack of cAMP-induced self-association. At the lowest concentrations of protein used (10 μm), the apparent molecular mass of HCN2J was also ∼26 kDa, suggesting that the protein was monomeric.

However, the apparent weight of HCN2J increased with increasing concentrations of protein, consistent with the appearance of larger forms. This interpretation is supported by values for polydispersity, the standard deviation of the distribution of molecular weights obtained by DLS. Values for polydispersity of greater than 30% are generally considered to indicate populations of particles that are of significantly different sizes, or to be polydisperse, with a molecular weight corresponding to the peak of the distribution (23, 24). At the lowest concentration of HCN2J and for ΔAB-HCN2J, polydispersity was <15%, which, along with the apparent molecular weight, is consistent with very small contributions from oligomeric forms of the protein. For intermediate protein concentrations, polydispersity values were ∼27.5%, suggesting small but more significant contributions from multiple oligomeric forms. This is reflected in the gradual increase in apparent molecular masses, ranging from ∼42 to 59 kDa for 25, 50, and 100 μm HCN2J. At the highest concentration of HCN2J (200 μm), the apparent mass is about 83 kDa, close to the 89.6 kDa as predicted for the tetramer; this, along with a polydispersity value of ∼24%, supports the predominance of the tetrameric form at this highest concentration. The higher HCN2J concentrations are likely a better reflection of its effective concentration when attached to the intact tetrameric channel at the plasma membrane.

Notably, the apparent weight was increased with cAMP at all concentrations of HCN2J except for the highest concentration; its effect was largest at the lowest concentration of HCN2J and progressively decreased as the protein concentration increased. In cAMP, values for polydispersity remained below ∼30% at all concentrations of HCN2J. At 200 μm, the absence of an effect of cAMP on the molecular weight of HCN2J is consistent with the predominance of tetramers. Thus, high affinity binding is apparent mainly when HCN2J is already a tetramer and is not due to self-association per se. Together with the stoichiometry of cAMP binding to HCN2J at higher protein concentrations, these data suggest an inherent negative cooperativity whereby cAMP binds initially to one subunit of the tetramer with high affinity and subsequently to the remaining three subunits with lower affinity. The stoichiometries of the two binding events are calculated in relation to the concentration of individual HCN subunits; thus, their summed value of 1 is consistent with negative cooperativity, in which one cAMP binds with high affinity and the remaining three cAMP molecules bind with low affinity.

Low Affinity Binding of cAMP Remains when C Terminus Is Constrained as Tetramer

To further examine binding of cAMP to the HCN2J tetramer, we fused it to the C terminus of GCN4-LI, a protein known to form a tetrameric coiled-coil (16), and confirmed that it was a tetramer by gel filtration chromatography (Fig. 3D). As pretetramerization reduces the amount of tag-free protein after cleavage with tobacco etch virus protease, and hence the yield, we kept the HMT tag attached for both constrained and unconstrained forms of the HCN2J tetramer.

FIGURE 3.

Only the low affinity binding transition is observed in a constrained HCN2J tetramer. A, plots of heat produced by ITC upon progressive injections of cAMP into 200 μm HMT-HCN2J or HMT-GCN4LI-HCN2J, as indicated. The solid line through the values of the lower plots represents a two independent binding site model and a single binding site model for HMT-HCN2J or HMT-GCN4LI-HCN2J, respectively. This fit yielded values for affinity and energetics (ΔG, ΔH, and ΔS). For the high affinity binding event in HMT-HCN2J, the c-values were ∼200. B, bar graph comparing affinities between the single binding transition of HMT-GCN4LI-HCN2J and the lower affinity transition determined from the two independent binding site model fit for HMT-HCN2J, in A. Error bars indicate S.D. C, bar graph comparing the energetics determined from the single binding transition of HMT-GCN4LI-HCN2J and of ΔAB-HCN2J and from the lower affinity transition of HMT-HCN2J. Error bars indicate S.D. D, size exclusion column profile of HMT-GCN4LI-HCN2J (black) and HMT-HCN2J (gray). mAU, milliabsorbance units.

Unlike the unconstrained HCN2J tetramer, titration of cAMP with the constrained HCN2J tetramer revealed only a single binding transition (Fig. 3A). The possibility thus exists that the HCN2J fragments were unable to form a tetramer. However, the binding is characterized by a low binding affinity and an energetic signature that were identical to those of the low affinity binding event in the unconstrained HCN2J tetramer (Fig. 3, B and C). The binding affinities of both the constrained and the unconstrained tetramer were similar to the single binding event determined for the monomeric ΔAB-HCN2J, but the energetics were different; the absolute values of entropy and enthalpy were much smaller after deletion (Fig. 3C). Thus, the CNBDs are able to make tetrameric contacts, and the changes induced by high affinity binding of cAMP in the unconstrained HCN2J tetramer have probably already taken place in the constrained tetramer.

Cyclic AMP Elicits Only Single Low Affinity Binding Transition in HCN1 Isoform

HCN2 and HCN4 channel opening is strongly facilitated by cAMP, whereas that of HCN1 and HCN3 is only weakly affected or not affected at all, respectively (12, 13). We therefore wondered whether the conformational changes in HCN2J, as seen in the double transitions of cAMP binding, might reflect the stronger effect of cAMP. Therefore, cAMP binding to purified versions HCN1 and HCN4 C termini was next examined. For HCN4J and ΔAB-HCN4J, the biphasic heat signature, apparent binding affinities of cAMP, and associated energetics were similar to those for HCN2 (supplemental Fig. 5).

In contrast, for HCN1J, only a single transition was observed upon cAMP binding in both low and high concentrations (Fig. 4A), with values of binding affinity and energetics (Fig. 4, B and C) that were similar to those for the single binding event in the constrained HCN2J and for the low affinity binding event of the unconstrained HCN2J tetramer. Notably, the energetics of binding for HCN1J were different from those of ΔAB-HCN1J. Although both exhibit a single binding transition of low affinity and similar free energy of binding, the absolute values for both enthalpy and entropy were larger, paralleling the differences in energetics noted between tetrameric HCN2J and ΔAB-HCN2J.

FIGURE 4.

Cyclic AMP binds to HCN1J similar to the constrained HCN2J tetramer. A, plots of heat produced by ITC upon progressive injections of cAMP into the indicated concentrations of HCN1J and 200 μm ΔAB-HCN1J, as indicated. The solid line through the values of the lower plots represents a single binding model for all of the plots above, which yielded values for affinity and energetics (ΔG, ΔH, and ΔS). B, bar graph comparing affinities among single transitions determined from the fits in A. Error bars indicate S.D. C, bar graph comparing energetics among the low affinity binding events for the indicated concentrations of HCN1 and ΔAB-HCN1J, determined from the fit in A. Error bars indicate S.D. D, plot of apparent molecular mass, determined by DLS, versus concentration of HCN1J, with or without a saturating concentration of cAMP (2 mm), as indicated. Values obtained under the identical conditions for HCN2J, from Fig. 2C, are replotted in open (− cAMP) and solid (+ cAMP) squares. Error bars indicate S.D.

Dynamic light scattering across a range of HCN1J concentrations (Fig. 4D), under the same conditions as those utilized for ITC, showed that, as for HCN2J, the apparent molecular weight increased with increasing concentrations, consistent with the appearance of larger forms. However, in the absence of cAMP, the apparent weights for HCN1J were larger than those for HCN2J, at intermediate and high concentrations of protein, suggesting that self-association and tetramer formation occurred to a greater extent under these conditions. Furthermore, the molecular weight of HCN1J, at all concentrations tested, was unaffected by saturating levels of cAMP, suggesting that the cAMP is unable to induce significant conformational changes.

The value of stoichiometry for the single binding event of HCN1J is ∼0.37 and ∼0.68, at 200 and 25 μm, respectively; that for ΔAB-HCN1J is ∼0.88 (supplemental table). Although a stoichiometry near 1 would be expected for a single binding event, the lower values probably reflect an underestimation of concentration that results from nonspecific aggregation that renders the cAMP-binding site inaccessible. Nevertheless, the molecular weights for HCN1 constructs are reliable as they were obtained from DLS experiments in which polydispersity values for the non-aggregated portions were ≤31%. Together, the data suggest that the changes induced by high affinity binding of cAMP in the unconstrained HCN2J tetramer have already taken place in the HCN1 C terminus.

DISCUSSION

Here, we show that cAMP binding to HCN channels can be assessed in detail using the isolated C terminus and interpreted in the absence of downstream changes in the transmembrane domains associated with channel gating. Cyclic AMP bound to one subunit of tetrameric HCN2 with high affinity and, subsequently, to the remaining three subunits with lower affinity. The greater similarity of energetics of the single binding event of the constrained HCN2J tetramer to those of the low affinity binding event of the unconstrained HCN2J tetramer than to those of the HCN2J monomer (ΔAB-HCN2J) suggests that the changes induced by high affinity binding of cAMP in the unconstrained HCN2J tetramer have already taken place in the constrained tetramer.

Cyclic AMP binds to HCN1, HCN2, and HCN4 with similar low affinity, but we found no evidence for a high affinity binding transition in HCN1. Thus, as for the constrained HCN2J tetramer, our data suggest that changes induced by the high affinity binding of cAMP have already occurred in the HCN1 C terminus. This “preactivation” of HCN1 explains both the small effect of cAMP on facilitation of its opening and the opening of the unliganded channel, which occurs as though it has already been disinhibited by the C terminus (12, 25, 26).

Two recent studies found that binding of cAMP to the HCN C terminus produced only a single transition when examined by ITC, with values very close to the low affinity binding values obtained in our study. In one study (27), the amount of HCN4 protein utilized was 75 μm, suggesting that the proportion of tetramers was low; this would explain the absence of a high affinity binding transition. In another study (28), the amount of HCN2 protein used was not reported. Temperature was also not reported in those studies. We found that lowering temperature from 25 to 10 °C reduced the amount of heat released considerably, although two transitions were still resolved provided that high concentrations were used. Nevertheless, reduced heat at lower temperatures could make it more difficult to resolve the transitions, particularly at lower concentrations at which tetramer formation is already low.

The biphasic pattern of heat released when cAMP was added to HCN2J was fit very well using a two independent binding site model, with three parameters per binding each of two sites (K_a, ΔH, and N), and not very well using a sequential model, with two free parameters (K_a and ΔH) for each of four binding sites. The comparatively better fit for the two independent binding site model was probably also due to the availability of the parameter N for each site of the two sites, which was free to adjust independently during the fit. The stoichiometry obtained from fitting the data to the two independent binding site model was close to 1:3, high affinity relative to low affinity.

A possible interpretation for the negative cooperativity inferred by the two independent binding site model is offered in supplemental Fig. 3. Binding of cAMP to the unoccupied HCN2J tetramer changes the conformation of one subunit, resulting in subunit interfaces with increased stability. This hampers the conformational changes of the other subunits upon binding cAMP, resulting in an observed negative cooperativity (a sequential model (29)). An alternative mechanism would involve one high affinity site and three low affinity sites that “pre-exist” within a tetramer. However, such asymmetry is not supported by the structures of the holo and apo HCN2 C terminus obtained by x-ray diffraction, which are 4-fold symmetrical tetramers (8, 30).

Our data support functional assays (28, 31) and modeling (9, 32–34) suggesting that HCN2 can switch between states of low (∼1–2 μm) and high (submicromolar) cAMP affinity, but those studies suggest that this switch occurs when the channel goes from closed to open. We show that a transition from high to low affinity can also be induced solely by cAMP binding to the C terminus without influence from the transmembrane domains. A similar switch has been described in the catabolic activating peptide dimer, a structurally related cAMP-binding protein in which ITC measurements have shown that binding of the second cAMP molecule occurs with lower affinity but more favorable enthalpy than binding of the first molecule (35); thus, for both the catabolic activating peptide and the HCN2 C terminus, the thermodynamic basis for negative cooperativity is completely entropic. It remains to be seen how this entropic C-terminal switch and resulting changes in conformation and/or dynamics interact with the transmembrane elements to facilitate HCN opening.