|Home | About | Journals | Submit | Contact Us | Français|
Investigation of the mechanistic bases and physiological importance of cAMP regulation of HCN channels has exploited an arginine to glutamate mutation in the nucleotide-binding fold, an approach critically dependent on the mutation selectively lowering the channel’s nucleotide affinity. In apparent conflict with this, in intact Xenopus oocytes, HCN and HCN-RE channels exhibit qualitatively and quantitatively distinct responses to the tyrosine kinase inhibitor, genistein – the estrogenic isoflavonoid strongly depolarizes the activation midpoint of HCN1-R538E, but not HCN1 channels (+9.8 mV ± 0.9 versus +2.2 mV ± 0.6) and hyperpolarizes gating of HCN2 (−4.8 mV ± 1.0) but depolarizes gating of HCN2-R591E (+13.2 mV ± 2.1). However, excised patch recording, X-ray crystallography and modeling reveal this is not due to either a fundamental effect of the mutation on channel gating per se or of genistein acting as a mutation-sensitive partial agonist at the cAMP site. Rather, we find that genistein equivalently moves both HCN and HCN-RE channels closer to the open state (rendering the channels inherently easier to open but at a cost of decreasing the coupling energy of cAMP) and that the anomaly reflects a balance of these energetic effects with the isoform specific inhibition of activation by the nucleotide gating ring and relief of this by endogenous cAMP. These findings have specific implications with regard to findings based on HCN-RE channels and kinase antagonists and general implications with respect to interpretation of drug effects in mutant channel backgrounds.
cAMP facilitates opening of cyclic nucleotide gated (CNG) and hyperpolarization-activated cyclic nucleotide regulated (HCN) pacemaker ion channels by binding to, and regulating the conformation of, a cyclic nucleotide gating ring formed from a binding domain (the CNBD) and a short linker (the C-linker) that physically links the CNBD to the S6 activation gate [1, 2]. Recently, considerable insight has been obtained into the mechanism and stoichiometry by which the gating ring regulates HCN voltage-dependent activation [3–15]. While the nature of the transitions undergone by the gating ring (such as whether it involves a switch from a four-fold to a two fold symmetric “dimer of dimers” arrangement [6–9]) and the extent of coupling of motions of the CNBD and C-linker have yet to be fully resolved [2, 8, 11, 16], the findings are in general agreement with a model wherein the unoccupied, deactivated gating ring stabilizes the closed channel [2–4, 8, 11, 17]. In this framework, tighter agonist binding to the activated gating ring promotes channel opening because it overcomes the basal inhibitory effect of the deactivated gating ring [2–4, 8, 11, 16]. A further refinement of this model indicates that, in addition to stabilizing the open channel relative to the closed channel, cAMP occupancy of the CNBD decreases the occupancy of a “voltage-desensitized” conformation .
Importantly, aspects of this model have been obtained through analysis of mutants that render the gating ring insensitive to physiological levels of cAMP by mutation of a critical arginine (See Fig. 6A) that interacts with the cyclized phosphate of the ligand [5, 6, 9, 14, 18]. Moreover, using a “knock-in” approach, this mutation has been introduced into mice to probe the physiological role of cAMP regulation of HCN channels . A critical assumption in such experiments is that the mutation selectively disables cAMP binding without perturbing the fundamental energetics of the voltage sensors or of the activation gate. Previous studies have appeared to validate this assumption and, additionally, suggested that the arginine-nucleotide interaction contributes only to the initial binding energy that serves to dock cAMP in the pocket and not to coupling energy [2, 7, 18, 20].
Several studies considering a potential role for tyrosine kinases as regulators of HCN channels have shown channel function is altered by the estrogenic isoflavinoid tyrosine kinase inhibitor, genistein, although there has been considerable disagreement as to whether the ligand acts via disruption of tyrosine phosphorylation [21–23] or through tyrosine phosphorylation independent mechanisms [24, 25]. While analyzing the sensitivity of HCN channel gating to such steroid like molecules we observed that, in two electrode voltage clamp (TEVC) recordings from intact Xenopus oocytes, HCN-RE channels responded very differently to genistein when compared to their cognate wild type counterparts. These observations raised the possibility that genistein was serving as a detector of abnormal aspects of gating within these mutants, a finding that could call into question aspects of the cAMP-gating model that are dependent on interpretation of the responses of the HCN-RE constructs and would raise questions with regard to the interpretation of HCN-RE “knock-in” mice. Moreover, the finding suggested the confusion among studies using kinase inhibitors might have an intriguing origin.
By use of excised inside-out patch clamp recording (IOPC) and X-ray crystallography, combined with interpretation within a simple cyclic gating scheme, we find that the apparent RE-sensitivity of genistein’s actions arise from a synthesis of three influences: 1. The strength of the HCN isoform-dependent basal inhibition of activation by the cAMP-vacant gating ring; 2. The efficacy with which this is overcome upon cAMP occupancy of the CNBD at physiological concentrations of the agonist and 3. A phytoestrogen-mediated movement of the channels towards their open state, a change that elicits a corresponding decrease in the coupling energy available from bound cAMP. Thus, as HCN-RE channels are uncoupled from physiological levels of cAMP (and so are fully inhibited by their gating ring) their activation is enhanced by genistein in intact oocytes. In contrast, for wild type channels, the loss of favorable coupling to basal cAMP in intact oocytes offsets the genistein-mediated enhancement of opening such that genistein elicits a weaker facilitation (HCN1) or an overall inhibition of gating (HCN2).
With respect to understanding the function and physiology of HCN channels, our analysis provides a novel demonstration that RE mutant HCN channels behave faithfully as cAMP-decoupled channels, shed light on the conflicting studies addressing the role of tyrosine kinases as regulators of HCN function and suggest that endogenous steroidal compounds may act as novel, cAMP sensitive, regulators of HCN channel function. In a broader context, our findings serve to illuminate how mechanistic interpretation of a residues role in a particular function can be subject to gross misinterpretation.
For functional studies, murine HCN1 and HCN2 channels (and mutants thereof) were subcloned into pGH19 and pGHE expression vectors, respectively. HCN-RE channels and CNBD truncation constructs were made by site-directed PCR mutagenesis as described previously [3, 18]. The sequences of all clones were confirmed by sequencing PCR amplified fragments on both strands. Residues are numbered according to the full-length channel sequences. Constructs were amplified in STBL2 cells (Invitrogen Corporation, Carlsbad, CA). cRNA was transcribed from NheI (HCN1 constructs) or SphI (HCN2 constructs) linearized DNA using T7 RNA polymerase (Message Machine; Ambion, Houston, Tx). 1–50 ng RNA was injected into each Xenopus oocyte.
For crystallization studies, we followed the protocol as previously described by Zagotta and co workers . Briefly, a DNA fragment encoding residues 443–640 of the murine HCN2 channel was expressed with the pETGQ vector in Escherichia Coli (construct HCN2I). Cells expressing HCN2I were harvested by centrifugation and lysed in a HEPES based buffer at 4 °C. The construct was isolated by Ni-NTA chromatography and the octahistidine tag was removed by thrombin cleavage. The tag-less construct was purified using cation exchange followed by size exclusion chromatography. Pooled fractions were dialyzed against a modified HEPES buffer and concentrated to 5–7mg/mL prior to crystallization.
Recordings were made in either TEVC or IOPC configurations from Xenopus oocytes 1–5 days after cRNA injection. Cells were maintained in L-15 media without ficoll (Specialty Media, Phillipsburg, NJ) at 17 °C until use. In both configurations, data were digitized using an ITC-18 interface (Instrutech Corporation, Port Washington, NY) and recorded with Pulse software (HEKA Elektronik, Lambrecht/Pfalz, Germany). Ag-AgCl ground wire(s) were connected to the bath solution by 3M KCl-2% agar salt bridges placed downstream of, but close to, the oocyte. Recordings were obtained at room temperature (22–25 °C). In all cases voltages are reported as the command potentials. In TEVC the measured deviation was less than 1%.
TEVC data were acquired using a Warner Instruments (Hamden, CT) OC-725C amplifier, filtered at 1 kHz then digitized at 2 kHz. Oocytes were bathed in a recording solution of (in mM) 107 NaCl, 5 KCl, 2 MgCl2, 10 HEPES-free acid pH 7.4 (NaOH). Microelectrodes were fabricated from 1B120-F4 borosilicate glass (World Precision Instruments, Sarasota, FL) and had resistances of 0.1–0.5 MΩ (I passing) and 1–4 MΩ (V sensing) when filled with 3 M KCl. An active virtual ground was used to clamp the bath.
IOPC data were acquired using an Axon Instruments 200B patch clamp amplifier (Axon Instruments, Foster City, CA) in resistive mode and then digitized at 5 kHz following filtering at 2.5 kHz (Warner LPF-8 8-pole Bessel filter). The extracellular solution was (in mM) 107 KCl, 5 NaCl, 1 MgCl2, 1 CaCl2, 10 HEPES-free acid pH 7.4 (KOH). The intracellular solution was (in mM) 107 KCl, 5 NaCl, 1 MgCl2, 1 EGTA-free acid, 10 HEPES-free acid pH 7.4 (KOH). cAMP was applied as the Na salt. In experiments where we applied 50 mM cAMP, the bath solution was adjusted so that the NaCl and KCl concentrations in the absence of cAMP were 50 and 62 mM and the Na.cAMP substituted for NaCl. Electrodes with resistances between 1–2 MΩ were fabricated from Kimax-51 borosilicate glass (Kimble Glass, Vineland, NJ) and coated with sylgard (Dow Corning Corporation, Midland, MI). In patches where the maximal current exceeded 500 pA, we applied analogue series resistance compensation (90% correction, 20 us lag with the resistance set to the initial resistance of the electrode). HCN channel gating spontaneously shifts hyperpolarized following patch excision. Accordingly, data acquisition was not begun until 3 minutes after patch excision into the Mg containing solution, a time sufficient to permit activation to largely equilibrate to the cell free level.
Steady-state activation curves were determined from tail currents (measured at 0 mV in TEVC and −40 mV in IOPC) following hyperpolarizing voltage steps applied in −10 mV increments. The holding potential was −40 mV in IOPC and −30 mV in TEVC. Tail current amplitudes were measured by averaging the current during the plateau of the tail (after allowing the voltage-clamp to settle and the uncompensated linear capacitance and Mg block [26, 27] to decay) and subtracting from this the baseline current recorded after the channels had fully deactivated. Tail current amplitudes were plotted versus the hyperpolarizing step voltage and fitted with the Boltzmann equation (Equation 1).
Here A1 is the current offset, A2 the maximal current amplitude, V is the step voltage, V1/2 is the activation mid-point voltage and s is the slope factor defined as RT/zF where z is a measure of the gating charge associated with channel activation and opening and R, T and F have their usual meaning. The tail amplitudes and Boltzmann fit were then normalized to the maximal tail amplitude, [I(v)-A1] / A2.
To follow changes in activation with greater temporal resolution than can be achieved when constructing full activation curves (channel gating is extremely slow), we adopted a protocol that would determine an approximate measure of the change in the V1/2. Each 30 or 60 s, a cell or patch was stepped to an intermediate voltage (VINT) that would achieve ~50% activation (determined from a full activation curve constructed at the beginning of the recording). Following determination of the intermediate current (IINT), the corresponding tail current (IINT TAIL) and a sojourn at the holding potential, the cell/patch was stepped to a voltage that saturated channel opening (VMAX), and the maximal current (IMAX) and corresponding tail current (IMAX TAIL) were determined. From these data we constructed conductance ratios: GINT TAIL/GMAX TAIL and GINT/GMAX by correcting current ratios for driving force using determined values of the reversal potential. These non-linear gating parameters (GR in equation 2) were then converted into linear approximations of the shift in the V1/2, the “ΔV1/2 apparent” (ΔV1/2 APP) using equation 2.
Here, V1/2 and s are activation mid-point and slope determined from a fit of the Boltzmann equation (1) to an initial steady state activation curve (see  for further details). Analysis of GINT TAIL/GMAX TAIL or GINT/GMAX yielded equivalent estimations of ΔV1/2 APP.
Crystals of HCN2I in the presence of cAMP were generated by the hanging drop method as previously described . Crystals were bi-pyramidal in shape and belong to the space group I4. Attempts to incorporate genistein into the crystals were conducted by soaking harvested crystals in solutions containing genistein. Working from a 100 mM stock of genistein in DMSO, three solutions of 4 mM, 8 mM, and 20 mM genistein were made with protein crystallization buffer (20 mM HEPES, 150 mM NaCl, 1mM dithiothreitol, 5mM cAMP, pH 7). Harvested crystals were sequentially soaked for one minute in solutions of increasing genistine concentration made by mixing 5 μL of diluted genistine with 5 μL of mother liquor. Crystals were cryoprotected in the final soak solution supplemented with 20 % glycerol and flash frozen in liquid nitrogen.
Diffraction data from one genistein soaked crystal was collected at the Advanced Photon Source beamline 19-ID, diffracting to 2.0 Å resolution. Integration, scaling and merging of the diffraction data were accomplished with the HKL suite of programs . The structure was solved with the program MOLREP  using the HCN2I-cAMP crystal structure as the starting model (PDB: 1Q43), water and ligand excluded. Refinement was carried out using REFMAC  and a composite omit-map generated using CNS  was used for modeling protein and cAMP. Two molecules of cAMP were modeled in the asymmetric unit and subsequently refined with REFMAC. Rwork and Rfree of the cAMP-included model were 25.8% and 27.4%, respectively. This model was used calculate a difference fourier map (Fo – Fc) and the resulting map used to search for positive peaks indicative of genistein. Solvent building was subsequently accomplished using ARP-wARP . The resulting model (Rwork = 21.6%, Rfree = 24.7%) and Fo – Fc map were further analyzed for evidence of positive peaks. Model building and interpretation of maps were carried out in the programs O  and COOT .
Here V is the transmembrane voltage, α = [cAMP]/Ko, c=Ko/Kc, Ko and Kc are the dissociation constants for cAMP binding to the open and closed states respectively, Lo is the opening equilibrium constant (C/O) in the absence of an applied field and z the gating charge associated with the L(v) equilibrium (assumed to be 2.9 and 4.3 for HCN1 and HCN2, respectively, based on slope values from fits of the Boltzmann function to steady state activation curves – see above). Ko and Kc for HCN2 were set to 15 and 1200 nM based on the estimates from Wang and colleagues . Kc for HCN1 was adjusted to 40 nM to reflect the weaker coupling of ligand to channel opening. Ko and Kc for RE channels were set to 1000x higher values. Lo was set so that the model V1/2 determined by equation 4 was equal to the observed V1/2 of RE channels. [cAMP] was then adjusted to 210 nM so that the model V1/2 of HCN2 was equal to the observed V1/2. The effect of genistein was introduced by a 3.1-fold decrease in Lo combined with loss of cAMP coupling (Kc* = Ko) and determined by solution of equation 4.
Genistein was obtained from Alexis corporation (Carlsbad, CA) and cAMP was obtained from Sigma. Stock solutions of these reagents were prepared in DMSO and H2O, respectively. However, when cAMP was used at 50 mM, the reagent was dissolved directly into the recording solution during its preparation. Drug stock solutions were frozen in aliquots and stored at − 20°C. All other reagents used in electrophysiology were obtained from Sigma and were of the highest available purity. The final DMSO concentration was adjusted to be 0.045% (V/V) in both drug and DMSO control recordings.
Data analysis was performed in PulseFit (HEKA Elektronik) or using custom analysis routines written in IgorPro (Wavemetrics Corporation, Lake Oswego, OR). SigmaStat V3.1 was used to perform statistical analysis (Systat Software, Point Richmond, CA). A T-test was used to determine if differences between populations are significant. A paired T-test was used to determine if differences before and after treatment are significant. All data are presented as mean ± SEM.
Figure 1 shows TEVC recordings obtained from Xenopus oocytes expressing HCN1 (Fig. 1A) or HCN2 (Fig. 1B). In each case current families obtained in response to hyperpolarizing voltage steps (LEFT) and tail currents (RIGHT) were obtained before (TOP) and after (BOTTOM) exposure of the cells to 90 μM genistein. In close accord with a previous study , we observe that genistein appears to have several effects on the channels. 1. The drug elicits a modest decrease in both the inward and outward current carried by both HCN1 and 2 (note the change in the scale bars in Fig. 1A and C). 2. The kinetics of gating are altered, an effect that is most obviously revealed by the change in the deactivation of HCN channels where a general slowing and a perturbation in the shape of the plateau phase is apparent. 3. Genistein inhibits the voltage-dependent opening of HCN2 such that more hyperpolarized potentials are required to open the channels (note that there is no detectable activation at − 65 mV in the presence of the drug whereas there was significant activation at this voltage in the absence of drug) but the presence of the drug only weakly alters voltage-dependent opening of HCN1.
To quantify the effect of genistein on the steady-state properties, we constructed current-voltage (IV) plots (Fig. 1B, D UPPER PANELS) and normalized steady-state activation curves (Fig. 1B, D LOWER PANELS). Inspection of the IV relationships confirms that genistein decreased the maximal current of both channels (IMAX/IMAX INITIAL for HCN1 and HCN2 was 0.69 ± 0.04, n= 9 and 0.71 ± 0.02, n= 11, respectively). Similarly, inspection of the steady-state activation curves reveals that opening of HCN2 was indeed shifted hyperpolarized by genistein while the drug weakly depolarized opening of HCN1. Mean changes in V1/2 of −4.8 mV ± 1.0 for HCN2 and +2.2 mV ± 0.6 for HCN1 were both statistically significant (P < 0.005; see also Fig. 3).
A notable difference between HCN1 and HCN2 lies in the strength of coupling between cAMP binding and channel opening. Thus, cAMP strongly facilitates opening of HCN2 but has only relatively modest effects on HCN1. In light of the observation that HCN2 was more responsive to genistein, we asked if channel regulation by the isoflavonoid was dependent upon the activation status of the gating ring. To address this question we recorded from cells expressing either HCN1-R538E or HCN2-R591E.
Figure 2 shows representative TEVC recordings obtained from cells expressing either HCN1-R538E (Fig. 1A,B) or HCN2-R591E (Fig. 1C,D). Qualitatively, the gating of these two constructs appears to be indistinguishable from their wild type counterparts (compare records to those in Fig. 1). However, quantitatively, it is clear that under control conditions the mutant channels activate somewhat hyperpolarized relative to the cognate wild type channels. This shift is most apparent in the more cAMP-responsive HCN2 background. Thus, inspection of the current records shows there is marked activation of HCN2 below ~ − 60mV (Fig. 1C) while HCN2-R591E does not activate until the voltage is stepped below ~ − 80 mV (Fig. 2C). The hyperpolarizing shift in gating observed in intact Xenopus oocytes upon mutation of the arginine to a glutamate (V1/2 of HCN1 and HCN1-R538E is − 65.7 mV ± 0.3, n=113 versus − 75.0 mV ± 0.3 n=212 while V1/2 of HCN2 and HCN2-R591E is − 77.4 mV ± 0.4, n=131 versus − 93.0 mV ± 0.6 n=74) is considered to represent the loss of coupling of HCN-RE channels to basal levels of cAMP [5, 18].
What of the action of genistein on the cAMP-unresponsive HCN-RE channels? Inspection of the current records in Figure 2 reveals that exposure to genistein results in a similar change in gating kinetics with a slowing of deactivation and a decrease in IMAX as seen when the drug is applied to wild type channels (Fig. 1; IMAX/IMAX INITIAL for HCN1-R538E and HCN2-R591E was 0.58 ± 0.04, n= 8 and 0.75 ± 0.05, n= 9, respectively). Surprisingly, the action of genistein on the voltage-dependent opening of both HCN1 and HCN2 channels is dramatically altered by the introduction of the RE mutations.
Thus, inspection of the current records shows that HCN1-R538E channels open at a markedly more depolarized potential in the presence, as compared to the absence, of genistein. This apparent facilitation of channel gating is clearly revealed by the right shift in the steady-state activation curve (Fig. 2B). Even more dramatically, the opening of HCN2-R591E channels is not inhibited by genistein (as observed for HCN2), but, rather, is robustly facilitated by the drug (Fig. 2C,D). Mean changes in V1/2 of +9.8 mV ± 0.9 for HCN1-R538E and +13.2 mV ± 2.1 for HCN2-R591E were both statistically significant (P < 0.001; see also Fig. 3).
We were concerned that the extended time required to obtain complete activation curves before and after application of genistein may have occluded an accurate assessment of the effects of the drugs by convolving it with an unanticipated rundown of gating during prolonged recordings. To circumvent this possibility, we employed a simple two-step protocol that allowed us to follow changes in the V1/2 of channel activation with a high temporal fidelity during drug onset (See Fig. 3A and Methods). Figure 3B shows plots of the derived parameter ΔV1/2 APP for HCN1 and HCN1-R538E (UPPER PANEL) and HCN2 and HCN2-R591E (LOWER PANEL) exposed to either 90 μM genistein or DMSO vehicle control normalized to the control value of the initial V1/2 determined from full steady-state activation curves for each channel. These data confirm that introduction of the RE mutation results in an enhanced facilitation of HCN1 and the inversion of the response of HCN2 from inhibition to facilitation in response to genistein in accord with the data shown in Figures 1 and and22.
We considered three possible hypotheses to account for the ability of these “cAMP-uncoupling” mutations to alter channel responses to genistein: 1. Mutation of the critical arginine to a glutamate altered the channel architecture and gating in a previously undescribed manner; 2. Genistein can act as an RE sensitive partial agonist at the cAMP site and 3. cAMP occupancy and/or the activation status of the CNBD influences the channel response to genistein interacting elsewhere. The data presented in the following sections suggest that the latter hypothesis is correct.
Inspection of the endpoint for the shift in activation gating reported by either ΔV1/2 APP (Fig. 3B) or the fully determined V1/2 (Fig. 3C) shows that the midpoint of activation of each channel pair (HCN1 versus HCN1-R538E and HCN2 versus HCN2-R591E) converges. One interpretation of this could be that this is coincidental with the altered basal energetics of wild type and mutant channels offset by corresponding changes in the coupling energy of genistein. A simpler, and therefore more attractive, interpretation of these data is that genistein uncouples the channels from the gating ring such that both the ring’s inhibitory effect, and the cAMP-mediated relief thereof, is lost. In such a scenario, wild type channels (which start out gating in a cAMP-modified state wherein inhibition by the gating ring is relieved) and RE channels (which start out gating in a cAMP-unbound state wherein inhibition by the gating ring is intact) end up gating in a genistein modified state independent of where they started with respect to the basal inhibition by the gating ring or its relief by cAMP.
To further investigate whether cAMP-occupancy of the CNBD influences the response of HCN channels to genistein, we recorded HCN2 and HCN2-R591E channels in excised inside-out patches where we could control the presence or absence of cAMP as well as genistein. Upon patch excision, HCN gating shifts hyperpolarized by 30–40 mV due, at least in part, to loss of coupling to basal cAMP but also as a consequence of loss of coupling to intracellular factors including acidic lipids [28, 36–38]. Although this shift is largely completed in the first 3 minutes following excision, a slow drift in gating can obscure drug-mediated changes in gating if activation is determined with time intervals that are too far apart. To circumvent this problem, we again utilized the two-step determination of the parameter ΔV1/2 APP. Figure 4A shows the general protocol and the voltage paradigm used in these recordings. Figure 4B shows the superimposed current records obtained before, during and after perfusion of patches containing HCN2 channels with 90 μM genistein. One patch was recorded in the presence of cAMP (Fig. 4B LOWER PANEL) and the other in the absence of the nucleotide (Fig. 4B UPPER PANEL). In each case the superimposed sweeps have been normalized to the maximum current response to show how channel opening at the intermediate voltage was altered by genistein after compensating for suppression of the maximal current. It is apparent that in the presence of a saturating concentration of cAMP the current through HCN2 at the intermediate voltage is suppressed by genistein. This finding is analogous to the action of genistein on HCN2 in intact oocytes where the basal cAMP level is high enough to facilitate the channels. However, when we determined the effect of genistein in the absence of cAMP, gating of wild type HCN2 was facilitated - the current at the intermediate voltage is enhanced relative to control and washout. Plots of the parameter ΔV1/2 APP obtained from these two recordings reveals the reversible facilitation of gating in the absence of cAMP and inhibition of gating in the presence of cAMP. Consistent with the results obtained in TEVC (Figs. 1 and and2),2), genistein suppressed the maximum current in both the presence and absence of cAMP in IOPC (Fig. 4D).
Previous studies in both HCN and CNG channels have shown that mutation of the critical arginine to glutamate does not prevent cAMP from binding to or activating the CNBD, it simply shifts the apparent affinity for cAMP into the high mM range [6, 18, 20]. Accordingly, we predicted that the mutant HCN2-R591E channel should behave similarly to HCN2 if cAMP were to be driven into its disabled CNBD. In accord with this prediction, in the presence of 50 mM cAMP, gating of HCN2-R591E was inhibited by genistein (Fig. 5A LOWER RECORDS) while channel gating was enhanced if the isoflavonoid was applied in the absence of cAMP (Fig. 5A UPPER RECORDS). The response of HCN2-R591E to genistein in the presence of 10 μM cAMP (a concentration that inverted the behavior of HCN2 to genistein but which is too low to elicit significant binding to the RE CNBD) was indistinguishable from the action of genistein in the absence of cAMP (data not shown). Thus, the response of HCN2-R591E to genistein is qualitatively indistinguishable from that of wild type HCN2.
Figure 5B presents a summary of the effects of genistein on the mid-point of activation gating of HCN2 and HCN2-R591E channels recorded in IOPC. As suggested from the representative data shown in Figures 4 and and5A,5A, the qualitative response of wild type and HCN2-R591E channels to the convergent effects of genistein and cAMP are identical. Moreover, while there is a small discrepancy in the amplitudes of the genistein-mediated shifts in gating in the two constructs, the overall net effect of genistein on channel gating (comparing between cAMP-bound and unbound states) is similar (~ 7 to 10 mV). Furthermore, this effect of genistein is close to the total genistein mediated difference calculated for cAMP-bound (HCN2) and cAMP-unbound (HCN2-R591E) channels in intact cells (~ 15–18 mV, See Fig. 3B,C) supporting the notion that the effects of genistein on wild type versus HCN-RE channels in intact oocytes arises from the difference in the cAMP occupancy and activation of the CNBD. These findings argue against genistein acting as a detector of a heretofore unobserved effect of the RE mutation on HCN channel basal gating and against genistein acting as an RE sensitive cAMP site partial agonist.
How can genistein enhance activation of HCN channel gating in the absence of cAMP but have little effect (on the gating of the weakly cAMP-responsive HCN1 channels) or an inhibitory effect (on the strongly cAMP-responsive HCN2 channels) in the presence of cAMP? We can consider two simple hypotheses to account for this behavior: 1. Genistein acts as a competitive partial agonist at the cAMP binding site but its binding is insensitive to the presence of a positive or negative charge on the amino acid that is important in docking cAMP. 2. Genistein acts as an allosteric modifier that moves the channel closer to the open state thereby facilitating opening but at the expense of weakening the coupling energy imparted by cAMP (see discussion for more details). We addressed this question in two ways. First, we asked whether genistein could occupy the HCN2 CNBD. To do this we soaked crystals of the gating ring in genistein and determined the structure by X-ray crystallography (see materials and methods for details). We observed no density due to genistein indicating that the molecule was not stably located in the structure (see supplemental data for the structure factor file, crystallographic and refinement statistics and the resultant pdb file) be it in the cAMP pocket or elsewhere. Second, we asked if deletion of the CNBD would eliminate the action of genistein as predicted if it is a partial agonist acting within the CNBD. Figure 6 shows that genistein facilitated activation of both HCN1-ΔCNBD and HCN2-ΔCNBD. These findings indicate genistein neither acts as a competitive partial agonist nor does it modify behavior of HCN channels by binding to a discrete site in the gating ring itself. Intriguingly, deletion of the Clinker (which acts not only as the tetramerization interface within the gating ring but also the structural interface between the gating ring and the membrane embedded pore and voltage sensing module) appears to largely, but not completely, abolish the sensitivity of HCN1 to genistein (in the lowest panel of Fig. 6B the application of genistein, but not vehicle, resulted in a small, albeit statistically insignificant, deviation in ΔV1/2 APP). HCN2-ΔCterm does not form functional channels precluding analysis of this construct.
The mechanism by which cAMP binding to the C-terminal gating ring enhances activation of HCN channels has been investigated using a number of approaches, but functional evidence for some aspects of the resultant gating models [2, 5, 6, 9, 14, 18] as well as of the physiological function of cAMP binding in vivo  has depended on utilization of cAMP-insensitive channels generated by mutation of an arginine that is critically involved in initial binding of the nucleotide [2, 7, 18, 20]. Here we investigated an unusual phenotype of these HCN-RE channels that suggested their gating might not well reflect the cAMP-independent gating of wild-type HCN subunits. Specifically, we studied the response of wild type and HCN-RE channels to the estrogenic isoflavonoid tyrosine kinase inhibitor, genistein. Our findings reveal that the divergent responses of HCN and HCN-RE channels actually arise as a direct result of the uncoupling of the CNBD from cAMP and provide support for the hypothesis that the gating mechanism of the cAMP-disabled HCN channels faithfully reflects the cAMP-independent activation of wild type HCN channels. Our results suggest that this action of genistein is exerted through an allosteric uncoupling of the CNBD from channel opening rather than the drug acting as a partial agonist at the CNBD. Below, we consider these ideas in more detail and consider the specific and general implications of our findings with regard to HCN channel regulation and the use of mutants in helping unravel pathways of regulation.
To provide a framework within which to consider how genistein may give rise to the different responses in HCN and HCN-RE channels, we consider a simple cyclic allosteric model (Fig. 7A). In this model, the channel is stabilized in the closed state in the absence of cAMP and genistein (indicated by lines between the subunits). Activation of the channel involves a concerted change that includes opening of the pore and conversion of the CNBD from a low affinity site to a site with high affinity for cAMP. The tighter binding of cAMP to the activated CNBD (Ko < Kc) imparts energy to stabilize the open state as shown by the enhancement of the voltage-dependent opening equilibrium, L(v), by the factor c, which is equal to Ko/Kc. Within this framework, we can interpret the action of genistein (Fig. 7B) as eliciting a weakening of the closed state such that the channel moves closer to the open state (hence its ability to facilitate the channels in the absence of cAMP). However, a cost of this would be to also move the CNBD closer to the open state, thereby reducing the difference between Ko and Kc and, hence, weakening the coupling energy that cAMP can impart to enhance channel opening. As shown in Figure 7C, a modest adjustment of Lo (3.1-fold decrease) combined with collapse of the Ko/Kc ratio are adequate to generate changes in the overall V1/2 that mirror the results we observe upon exposure of HCN channels to genistein. Thus, the model can describe the depolarization of both HCN1 and HCN1-R538E (ΔV1/2 = +2 mV and +10 mV as observed in the presence of a physiological concentration of cAMP in TEVC). Similarly, the same genistein-mediated changes in Lo and Ko/Kc would yield a depolarization of HCN2-R591E (ΔV1/2 = +7 mV) but a hyperpolarizing shift in the gating of HCN2 (ΔV1/2 = −8 mV). It should be noted that while aspects of the model are set according to observed data the adjustments to simulate the effects of genistein were arbitrarily chosen and do not necessarily represent a unique solution. Moreover, it is important to note that the significance of the model is that, despite its simplicity (in that it does not separate activation and opening [10, 13], account for the presence of multiple voltage sensors [39–42] or incorporate either a modal shift in the behavior of the sensors [41, 43, 44] or the presence of inactivated states ), it does, nonetheless, show how the observed divergent responses of HCN and HCN-RE channels to genistein do not require that the channels have fundamentally different gating.
What of the decrease in IMAX and the changes in gating kinetics upon exposure to genistein? The effect of genistein on channel kinetics will likely have at least two components. Thus, the observable rates of activation and deactivation will change in concert with the shift in the V1/2 irrespective of whether the drug is acting to alter the voltage-sensing reactions or the opening transitions but these effects may be convolved with a slowing of gating that occurs independently of a shift in the V1/2 as a result of altered phosphorylation/coupling of a kinase to the C-linker (see below). Similarly, the decrease in IMAX could arise in a number of ways including a decrease in POPEN (if the drug alters the opening reaction directly or binds preferentially to closed but not open states ) or a simple block of the channel. As our primary goal here was to determine the basis of the divergent effects of genistein on voltage gating of HCN and HCN-RE channels and changes in both IMAX and kinetics can have other confounding origins, we have not considered these other factors further.
Why does genistein only weakly facilitate (HCN1) or inhibit (HCN2) opening of full-length HCN channels in the presence of a physiological concentration of cAMP but facilitate gating of HCN channels devoid of the CNBD when each of these interventions are able to relieve the CNBD-mediated contribution to closed state stability ? In wild type channels, the net inhibitory effect of genistein presumably reflects a balance between an enhancement of the ease of channel opening and the loss of the coupling energy of bound nucleotide. However, in the background of HCN-ΔCNBD channels, there is no coupling energy to give up so the response to drug will be dominated by the remaining component – the enhancement of opening.
Where does genistein interact with the channels to elicit the change in opening and cAMP coupling? In light of the observations that genistein does not bind within the C-linker per se but deletion of this motif greatly weakens the drug’s functional effect on the channel, it is tempting to speculate that the isoflavinoid is associating at the interface of the C-linker and membrane and/or membrane embedded pore and voltage-sensing elements of the channel. However, in the absence of definitive evidence for binding to such a site we have no way to exclude the possibility that the loss of the C-terminus weakens the drug’s actions by modifying a site elsewhere in the protein.
In several recent reports the effects of genistein and PP2 on HCN channel gating have been used to support the idea that HCN channels are regulated by tyrosine phosphorylation. An apparent ineffectiveness of daidzein and PP3 (tyrosine kinase inactive analogues of genistein and PP2, respectively) was used to strengthen this hypothesis [21–23]. However, others have reported, as we show here, that genistein can have non-catalytic effects on HCN channels [24, 25] – though such a finding alone does not preclude the possibility that such drugs alter HCN channels via channel association with a genistein:tyrosine kinase complex as has been suggested for CNG channels [46–49]. Moreover, as previously reported by others [21, 22], we find that daidzein is relatively ineffective against cAMP bound HCN channels (data not shown). However, we find that daidzein can mimic the effects of genistein when HCN channels are cAMP unbound (albeit generally more slowly and often with a reduced efficacy - data not shown). Such effects of daidzein can be readily explained within the context of our model but are hard to reconcile with the hypotheses that the effects of genistein (and maybe PP2) are due solely to inhibition of catalytic effects of tyrosine kinases or an allosteric effect of interaction of the channels with a drug:kinase complex. A second line of evidence used to support a role for tyrosine phosphorylation of HCN (and CNG) channels is to probe the effect of site-directed elimination of select tyrosine residues [21, 46, 50–52]. While this is a valid and important line of investigation, our findings suggest that loss of sensitivity to drugs upon mutation of these residues (which all fall within the gating ring and may, therefore, alter cAMP coupling) may have a very different origin than elimination of kinase substrate site. This is not to say that tyrosine kinases do not also exert effects on HCN (or CNG) channels. For example, this hypothesis has received support from studies using over expression of, and interaction with, kinases, phosphatases and dominant negative constructs thereof [21, 50, 53]. Our findings simply serve notice that the complex energetic inputs to gating can confound interpretation of even the simplest experiment and this needs to be taken into account.
In summary, our results suggest that estrogenic type compounds such as genistein and, perhaps, also endogenous steroidal compounds, can act as allosteric modifiers of HCN channel function in a manner such that the phenotype is determined by the activation status of the cyclic nucleotide-gating ring. This idea is in keeping with the finding that HCN channels are functionally regulated by association with lipid rafts  and bears an appealing similarity to the manner in which cortisone alters the function of a Kv1-Kvβ complex . Importantly, we show that the divergent responses of wild type and cAMP-disabled HCN channels to isoflavinoids do not reflect an underlying disruption of channel activation or opening in the mutants, rather these phenotypes directly emerge from the designed loss of coupling to cAMP.
We thank Alex Lyashchenko for helpful discussions and Margaret Wood and faculty in the Department of Anesthesiology for their continuing support. We thank members of Steven Siegelbaum’s laboratory for generously providing us with Xenopus oocytes. This work was supported by grants to GRT (Whitehall Foundation S98-23 and 2003-05-02-REN) and KJF (NIH F31 MH070202-02).