cAMP control of HCN2 channel Mg2+ block reveals loose coupling between the cyclic nucleotide-gating ring and the pore

Abstract

Hyperpolarization-activated cyclic nucleotide-regulated HCN channels underlie the Na+-K+ permeable IH pacemaker current. As with other voltage-gated members of the 6-transmembrane KV channel superfamily, opening of HCN channels involves dilation of a helical bundle formed by the intracellular ends of S6 albeit this is promoted by inward, not outward, displacement of S4. Direct agonist binding to a ring of cyclic nucleotide-binding sites, one of which lies immediately distal to each S6 helix, imparts cAMP sensitivity to HCN channel opening. At depolarized potentials, HCN channels are further modulated by intracellular Mg2+ which blocks the open channel pore and blunts the inhibitory effect of outward K+ flux. Here, we show that cAMP binding to the gating ring enhances not only channel opening but also the kinetics of Mg2+ block. A combination of experimental and simulation studies demonstrates that agonist acceleration of block is mediated via acceleration of the blocking reaction itself rather than as a secondary consequence of the cAMP enhancement of channel opening. These results suggest that the activation status of the gating ring and the open state of the pore are not coupled in an obligate manner (as required by the often invoked Monod-Wyman-Changeux allosteric model) but couple more loosely (as envisioned in a modular model of protein activation). Importantly, the emergence of second messenger sensitivity of open channel rectification suggests that loose coupling may have an unexpected consequence: it may endow these erstwhile "slow" channels with an ability to exert voltage and ligand-modulated control over cellular excitability on the fastest of physiologically relevant time scales.

Figure 1. Schema and models describing HCN channel gating and Mg²⁺ block.

A. Mg²⁺ block of HCN channels may occur via a simple bi-molecular process (Scheme 1) or via more complex processes (e.g., Scheme 2). For further details see Methods, Results and Discussion. B. Schematic representation of an allosteric gating reaction wherein Mg²⁺ can bind to and block the open channel (reactions going back into the plane of the page) but does so without altering the energetics of either activation (horizontal steps in the plane of the page) or opening (vertical steps). Further details of the model and the methods used to optimize the rate constants associated with gating are given in the methods section. C. Schematic representation of the modular model of gating. Here, as in the basic concerted model shown in panel B, voltage sensors can activate irrespective of the status of the pore and the pore can open whether the voltage sensors are activated or not but voltage sensor activation and pore opening results in a reciprocal stabilization when the allosteric coupling factor, E is >1. Furthermore, tighter binding of agonist when the gating ring is activated leads to a reciprocal stabilization of the ring and bound agonist if the allosteric factor W is >1. The critical divergence between the concerted and modular models is that in the latter case elements of the gating ring can be either activated or deactivated when the pore is open. As shown, pore opening is coupled to the status of the C-linker such that the open pore and the activated C-linker are reciprocally stabilized when the coupling factor Q is >1. Coupling between other modules is not excluded but is not required for, nor included in, our simulations.

Figure 2. cAMP accelerates [Mg²⁺]_in block of HCN2 channels.

A. HCN2 channels activated at −155 mV and deactivated at +100 mV in the presence of 2 mM intracellular Mg²⁺ and the absence (Pre), presence (Plus) and following washout (Post) of 30 µM cAMP. Arrows indicate the instantaneous tail current amplitudes in the absence and presence of cAMP (determined by zero time extrapolation of fits of a single exponential function – e.g,. as shown in B and C). Records are active sweeps before subtraction of flanking leak sweeps acquired using the deactivation protocol (see Methods). B. Expanded views of the initial 2 ms of the +100 mV tails from A following subtraction of the averaged interlaced leak sweeps (shown in blue). Solid red lines represent fits of a single exponential function. The residuals from the fits are shown vertically offset for clarity. In this and all other figures, dashed red lines represent the zero current level. C. Current records (and exponential fits thereunto) normalized to the observed peak amplitude of the plus 30 µM cAMP tail current. Inset: the time constants of decay of the initial phase of the HCN2 tail currents in the presence of 2 mM Mg²⁺ and the absence (open symbols) or presence (closed symbols) of 30 µM cAMP (10 to 27 determinations per point) are significantly different at each potential (Student's t-tests). Data acquired from deactivation and sequential IV protocols (see Methods) were pooled in this plot.

Figure 3. cAMP acceleration of [Mg²⁺]_in block is mediated via ligand occupancy of the cyclic nucleotide-gating ring.

A. Average of 8 consecutive active sweeps acquired from a patch expressing HCN2-R591E channels in response to the sequential IV voltage paradigm. Intracellular Mg²⁺ was 1 mM. B. Expanded view of activation (Left panel) and deactivation (at the holding potential of −40 mV; Right Panel) of HCN2-R591E obtained in the absence (Pre), presence (Plus) and following washout (Post) of 30 µM cAMP. Records are from same patch as A and are each averages of 8 sweeps acquired in response to the active paradigm before subtraction of the averaged interlaced leak records. C. Expanded views of the leak subtracted currents recorded at +50 and +200 mV (as indicated) in the absence, presence and following washout of 30 µM cAMP (traces and legend as in B). Red lines are fits of a single exponential function. Residuals are shown vertically offset for clarity. D,E. Time constant of block by 1 mM intracellular Mg²⁺ of HCN2-R591E (D) and HCN2 (E) in the absence or presence of 30 or 300 µM cAMP. For HCN2 but not the cAMP-disabled construct, HCN2-R591E, block kinetics in the presence of cAMP were significantly different from block in the absence of cAMP while the speed of block of HCN2-R591E in the absence or presence of cAMP was not different from that of block of HCN2 in the absence of the nucleotide (one-way ANOVA at each voltage with 11–20 determinations per point).

Figure 4. Bi-exponential behavior of [Mg²⁺]_in block in the absence of cAMP.

A,B. Expanded views of leak subtracted currents recorded at +100 (A) and +200 mV (B) in the presence of 3 mM Mg²⁺ and absence (gray) or presence (black) of 30 µM cAMP before (upper) and after (lower) normalization to the observed peak tail current. Red lines are fits of single or double exponential functions (30 and 0 µM cAMP, respectively). Blue lines represent the slow component of the double exponential fits. Residuals are shown offset below the current records in the upper panels. Data acquired with the sequential IV protocol (see Methods). C–E. Plot of 1/τ_BLOCK versus voltage at the indicated Mg²⁺ concentrations in the presence (C) and absence (D,E) of cAMP. D and E plot the data for the fast and slow phases of block in the absence of cAMP, respectively. The dashed lines in D are the fit lines from C. r² values for fits to 50, 100, 150 and 200 mV data are C: 0.9974, 0.9997, 0.9991 and 0.9997; D: 0.9985, 0.9676, 0.9782 and 0.9854; E: 0.1004, 0.8165, 0.7023 and 0.9966. Data are from 7–26 and 7–27 separate patches for plus and minus cAMP, respectively.

Figure 5. cAMP abolishes a slow blocking population of channels.

A. k₁ determined from the slopes of the regression lines in Figure 4C–E plotted against the depolarizing step potential in the presence or absence of cAMP. Dashed lines represent fits of equation 1 (see text for details). B,C. Compound rate constants determined in the presence (B) and absence (C) of cAMP. Black symbols: k′′′ (equal to k₂[Mg²⁺]_out+k₋₁+k₋₂ at 1 mM [Mg²⁺]_out) as obtained from the y-intercepts in Figure 4C–E. Teal and blue symbols: k′′ (equal to k₋₁+k₋₂ obtained according to equation 6) at 2 mM and 3 mM [Mg²⁺]_in, respectively. The red symbol at −135 mV is set to 10⁵ s⁻¹ in keeping with the observation that recovery of current is faster than the time constant of the clamp at that voltage . The long and short dashed lines (B) represent the optimized behavior of k₋₁ and k₋₂, respectively obtained from fits to the black and red circles. This fit reported s⁻¹, δ-₁  = 0.306, s⁻¹ and δ₋₂  = 0.303. D. Black and grey symbols show the fractional unblocked current. The ratios at 0 mV are omitted as this potential is close to the reversal potential and, therefore, poorly defined. Teal and blue lines: the probability channels are unblocked (equation 4) using the scheme II parameters determined in A–C. The black line is a fit of equation 4 wherein both k₂ and k₋₂ are zero; it represents the predicted exponential behavior if Mg²⁺ block were to accord to Scheme I. E,F. Plots of the fractional unblocked current and the relative amplitude of the fast component of block (zero time extrapolation of the fast component with respect to the sum of zero time amplitudes of the fast and slow components, A_f and A_s respectively – right hand aspect of F). Open red symbols (F) represent the estimates obtained in the presence of 2 mM Mg²⁺ and absence of cAMP when a block window of 10 ms was employed in place of the normal 2 ms window. The dashed line (F) is the mean of the fractional fast amplitude determined in the presence of 0.3, 1, 2 and 3 mM Mg²⁺ at 200 mV.

Figure 6. [Mg²⁺]_in does not modify closing kinetics and closing does not intrude into the block time domain.

A,B. Leak sweep subtracted tail currents in the absence or presence of 2²⁺ and absence of cAMP normalized to the peak amplitude of each recording then averaged (A: 10 and 16 separate recordings) or same records after scaling of the 0 Mg²⁺ record (B). The SEM of these averaged records is included as a pixilated halo around the records in A and D. C. Deactivation envelopes determined in the absence (open circles) and presence (filled circles) of 2 mM Mg²⁺ (3–8 determinations per point). The continuous line represents a mean +100 mV tail current (14 separate recordings each normalized to the peak amplitude before averaging). At no time were the envelope amplitudes in the absence and presence of Mg²⁺ significantly different (Student's t-tests). D. The initial 2 ms of +100 mV tail currents collected in the absence of internal Mg²⁺ and the absence or presence of cAMP (normalized to the peak amplitude during the 2 ms window then averaged).

Figure 7. Slow block is controlled by cAMP occupancy of the gating ring and not open probability.

A. +100 mV leak sweep subtracted pre-deactivation HCN2 tail currents observed immediately following activation (see inset for opening trajectories) at −155 mV for 2 s (black traces) or 14 s (blue) or for 14 s at either −100 mV (red) or −95 mV (pink) in the presence of 2 mM intracellular Mg²⁺ and 30 µM cAMP. Superimposed smooth lines are fits of each trace with a single exponential function. The V_1/2 and slope factor determined from a fit of the Boltzmann function to an activation curve constructed from 10 s sweeps were −105.7 and 4.5 mV, respectively (data not shown). B. Sweeps and fits from the block records shown in A each normalized to the instantaneous amplitude determined from the cognate exponential fit. C. Single exponential time constants of block from five patches such as that shown in A–B (gray shaded symbols). These are compared to the mean (± SEM) time constants of block by 2 mM Mg²⁺ following activation at −155 mV for 2 s in the presence (filled circle, n = 22) or absence (open symbols, n = 27) of cAMP (open circle and square: fast and slow components of a two exponential fit). Submaximal activation voltages varied between −95 and −115 mV (in 5 mV increments) while times varied between 5, 8 or 14 s at the submaximal voltages and between 2 or 14 s at −155 mV. For clarity, and because varying the durations and activation voltages had no effect on block kinetics (other than altering the open probability at the onset of the block epoch - see, for example, A and B), we do not differentiate between short and long activation pulses or the various activation potentials in this plot.

Figure 8. cAMP control of [Mg²⁺]_in block and channel opening are kinetically decoupled processes.

A,B. Simulated HCN2 currents at −155 mV (Left) and +100 mV (Right) in the absence (Gray) and presence (Black) of cAMP and the absence of intracellular Mg²⁺. The current records were simulated using the rate constants shown in Table 1 wherein cAMP did (B) or did not (A) alter activation transitions. C,D. Probability of occupancy of sum of open and open blocked states with the indicated number of activated voltage sensors (upper panels) and open unblocked probability (lower panels) when cAMP alters only the opening isomerization (C) or both activation and opening reactions (D). In all panels, the probabilities were normalized to the initial maximal open probability under the specified conditions to simplify comparison of simulations generated in the presence and absence of cAMP. Note that the plus and minus cAMP traces in the lower panels superimpose.

Figure 9. Inter-pulse sag arises from cAMP-dependent anomalous closure at hyperpolarized potentials, not closure at +100 mV.

A,B. Averaged active records obtained in response to the depolarized conditioning envelope paradigm (wherein the +100 mV sojourn was 10 ms) from a cell expressing HCN2 (A) or an un-injected cell from the same donor frog (B) before (Black) and after (Gray) inclusion of 300 µM ZD7288 in the bath. In each case, the inset shows the tail currents at +100 mV obtained after the second −155 mV epoch. C–E. Records obtained before (C), during (D) and after (E) the conditioning 10 ms step to +100 mV from the HCN2 (Upper traces) and un-injected (Lower traces) recordings shown in A and B, respectively. Yellow and blue traces are the averaged active and leak records. Where included, the black and gray traces are the difference currents obtained before (Black) and after (Gray) inclusion of 300 µM ZD7288. F,G. Leak sweep subtracted continuous +100 mV tail currents (normalized to the peak amplitude of each recording then averaged; 14 and 13 separate recordings in the absence and presence of cAMP, respectively) and the normalized amplitudes of the instantaneous (Inst) and delayed (delay) envelope currents upon return to −155 mV following steps of varying duration to +100 mV (3–11 determinations per point) are each plotted with respect to time at +100 mV. The instantaneous and delayed amplitudes were determined from fits of a single exponential (e.g., red line in A and D) and plotted as a function of the current amplitude at −155 mV immediately prior to the +100 mV conditioning step. H. Time constant of the −155 mV closing phase as a function of preceding +100 mV conditioning interval.

Figure 10. A modular model describes the cAMP enhancement of HCN2 activation and acceleration of [Mg²⁺]_in block.

A. Observed (obs) and model generated (model) values of the V_1/2 and P_MAX of channel activation and P_CL (the probability that the linker is in the resting configuration which we assume is reported as the slow component of block) each in the absence and presence of cAMP. The observed apparent affinities (K_1/2) were either determined by fits of the Hill equation to model-generated concentration response curves or, for the observed K_1/2 of cAMP modulation of gating, taken from published values . B. Predicted behavior of τ_FAST (thick line) and τ_SLOW (thin line) as a function of the membrane potential. Curves were generated using equation 3 with k₂ set to zero (see text for details).