Separable Gating Mechanisms in a Mammalian Pacemaker Channel*

Despite permeability to both K (cid:1) and Na (cid:1) , hyperpolar-ization-activated cyclic nucleotide-gated (HCN) pacemaker channels contain the K (cid:1) channel signature sequence, GYG, within the selectivity filter of the pore. Here, we show that this region is involved in regulating gating in a mouse isoform of the pacemaker channel (mHCN2). A mutation in the GYG sequence of the selectivity filter (G404S) had different effects on the two components of the wild-type current; it eliminated the slowly activating current (I f ) but, surprisingly, did not affect the instantaneous current (I inst ). Confocal imaging and immunocytochemistry showed G404S protein on the periphery of the cells, consistent with the presence of channels on the plasma membrane. Experiments with the wild-type channel showed that the rate of I f deacti- vation and I f amplitude had a parallel dependence on the ratio of K (cid:1) /Na (cid:1) driving forces. In addition, the amplitude of fully activated I f , unlike I inst , was not well predicted by equal and independent flow of K (cid:1) and Na (cid:1) . The data are consistent with two separable gating mechanisms associated with pacemaker channels: one (I f ) that is sensitive to voltage, to a mutation in the selectivity filter, and to driving forces for permeating cations and another (I inst ) that is insensitive to these influences. Pacemaker channels (also known as

important in thalamocortical neurons where its deactivation produces a slowly decaying after-depolarization that determines the length of refractory periods separating episodes of synchronized oscillations (8,10).
We have recently shown that HCN2 channels produce an instantaneous and Cs ϩ -insensitive current component (I inst ) in addition to the hyperpolarization-activated and Cs ϩ -sensitive I f component (11). I inst was not affected by a mutation in the S4 transmembrane segment (S306Q) whereas I f was greatly reduced. Thus, our data support a role for the S4 segment in voltage-dependent gating (I f ), as suggested previously for HCN channels (12,13) but not in voltage-independent gating (I inst ).
Despite permeability to both K ϩ and Na ϩ , the HCN pore contains a conserved GYG sequence that is found in many potassium-selective channels (14,15). Studies examining the effects of mutations in the selectivity filter of K ϩ channels have shown that conformational changes in this region contribute directly to channel gating (16,17). Permeant and blocking ions such as Rb ϩ , K ϩ , and Cs ϩ also affect the closing of K ϩ channels, implicating the selectivity filter in voltage-dependent gating (18 -21). Finally, the KcsA K ϩ channel selectivity filter and activation gate have recently been shown to have different conformations in conditions of low K ϩ and high K ϩ , which could explain the effects of permeant ions on gating (22). Based on the similarity of the pore structure in HCN and K ϩ channels, it seems likely that the selectivity filter and the voltagedependent gate of HCN channels are linked structurally such that changes in one influence the function of the other. However, this has not been directly demonstrated.
In this study, we carried out experiments to determine whether the selectivity filter is coupled to gating in HCN channels using mouse HCN2 subunits expressed in Chinese hamster ovary (CHO) cells. We found that I f was sensitive to a mutation in the selectivity filter and to different concentrations of permeating cations, whereas I inst was insensitive to these influences. The results may be explained by changes in pore conformation upon hyperpolarization, and/or by the presence of a second pore that is: 1) found within the same channel, 2) formed by a second population of the same channel subunits, or 3) associated with HCN channels in the form of up-regulated endogenous channels.

EXPERIMENTAL PROCEDURES
Mutagenesis and Expression-The G404S mutant was constructed by overlapping PCR mutagenesis from a mouse HCN2 template as previously described (11). The amplified mutagenic product and wildtype mHCN2, in the mammalian expression vector pcDNA3 (6), were subsequently digested using NheI and BlpI. The fragment digested out of wild-type mHCN2 was replaced by the complementary fragment carrying the mutation. The mutation was confirmed by restriction analysis and automated sequencing (Biotechnology Laboratory, University of British Columbia, Vancouver Canada).
Tissue Culture-CHO-K1 cells were obtained from ATCC (Manassas, VA), maintained in Hams F-12 media supplemented with antibiotics and 10% fetal bovine serum (Invitrogen) and incubated at 37°C with * This work was supported by grants from the Canadian Institutes for Health Research and the American Health Assistance Foundation (to E. A. A.), the Heart and Stroke Foundation of Canada, and the Heart and Stroke Foundation of British Columbia & the Yukon (to E. A. A. and C. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Present address: Dept. of Neurobiology, Harvard Medical School, 220 Longwood Ave., Boston, MA 02115.
Electrophysiology-Cells expressing GFP were chosen for whole-cell recordings 24 -48 h after transfection or co-transfection with mHCN2. The pipette solution contained (in mM): 130, KCl; 10, NaCl; 0.5, MgCl 2 ; 1, EGTA; 5, HEPES; pH adjusted to 7.4 with KOH. The extracellular solution contained (in mM) 140, XCl (X ϭ sum of K ϩ and Na ϩ ); 1.8, CaCl 2 ; 0.5, MgCl 2 ; 5, HEPES; pH adjusted to 7.4 with NaOH. For solution changes, a 200-l bath was completely exchanged and perfused (0.5-1 ml/min) for at least 1 min prior to collecting data. Whole-cell patch clamp currents were recorded using an Axopatch 200B amplifier and Clampex software (Axon Instruments, Union City, CA) at room temperature (20 -22°C). Currents were not leak-subtracted and capacitance compensation was not used. Patch clamp pipettes were pulled from borosilicate glass and were fire polished before use (pipette resistance is 2.5-4.5 M⍀). Data were filtered at 2 kHz and were analyzed using Clampfit (Axon Instruments), Origin (Microcal, Northampton, MA) and Excel (Microsoft, Seattle, WA) software. The current densities plotted in Fig. 2, A and B were determined by dividing measured currents by the capacitance, which was estimated by the ClampEx software from the time constant of the current elicited by a 2 mV test pulse at the beginning of whole-cell recording.
Time constants to assess rates of activation and deactivation were generated using a single exponential fitting procedure. An initial delay occurred prior to I f deactivation that was not well described by a single exponential function, as described previously (5,25) and therefore was not used in our fits. Fig. 4 uses values of I inst corrected for the presence of endogenous instantaneous currents. Previously, we found that the amplitude of I inst was 41 and 48% larger than the instantaneous current determined in cells transfected with only GFP or with the pHOOK membrane protein (a single chain antibody, fused to the C terminus of the transmembrane domain from the platelet-derived growth factor receptor, that allows for the antibody to be anchored and displayed on the extracellular side of the CHO cell plasma membrane), at Ϫ140 mV and Ϫ65 mV, respectively (11). These proportions were used to estimate the amplitudes of instantaneous current due only to HCN2, plotted in Fig. 4, C and D (I inst(corrected) ).
Immunocytochemistry and Confocal Microscopy-At 24 -48 h posttransfection, cells on coverslips were washed with PBS and fixed in 2% paraformaldehyde in PBS for 5 min. The cells were then washed with PBS, permeabilized using 0.2% Triton X-100, and blocked with 10% normal goat serum (NGS). After one wash with PBS containing 1% NGS, the cells were incubated with a rabbit polyclonal antibody to HCN2 (Alomone Labs, Jerusalem, Israel) at a dilution of 1:400 in PBS with 1% NGS for 48 h at 4°C. The antibody was removed, and cells were again washed with PBS. Cells were then incubated with a donkey anti-rabbit antibody tagged with cyanine 3 (Cy3) (Jackson Laboratories, West Grove, Pennsylvania) at a dilution of 1:200 in PBS with 1% NGS for 1 h at room temperature in the dark. The antibody was removed, cells were washed in PBS, and the coverslips with cells were mounted on slides using Permount (Fisher Scientific). Cells were examined using wide-field and confocal microscopy with an Olympus BX 40 inverted microscope, equipped with epifluorescence and appropriate filters, and an inverted Zeiss TurboPascal confocal microscope, respectively. For confocal microscopy, serial sections were taken in 0.8 -1.0 M steps using a ϫ63 oil immersion objective lens and an excitation wavelength of 543 nm.

RESULTS
A Pore Mutation (G404S) Reduces I f but Does Not Affect I inst -Our initial strategy to examine the involvement of the selectivity filter in gating was to introduce a mutation into the GYG sequence of mHCN2 (G404S) and determine its effects on expressed currents. Fig. 1 shows that the region of the pore encompassing the selectivity filter has significant homology among different classes of K ϩ channels and cyclic nucleotidegated channels. We mutated the glycine at position 404 to a serine ( Fig. 1, asterisk). Serine was chosen because the corresponding mutation (G629S) in the human ether-a-go-go-related channel (HERG), which is related to HCN channels, eliminated slowly activating current but not trafficking to the plasma membrane (23). Cells expressing G404S produced little I f , even in a high K ϩ (135 mM) extracellular solution ( Fig. 2A). We determined I f and I inst densities (see "Experimental Procedures") in cells expressing wild-type HCN2 and G404S in order to control for variability in cell surface area. Cells expressing G404S produced significantly less I f than did cells expressing wild-type HCN2 (Fig. 2B). The instantaneous current associated with G404S (I inst(G404S) ) and wild-type HCN2 (I inst(HCN2) ) were significantly larger than the instantaneous current observed in cells expressing only GFP (I inst(GFP) ) but were not significantly different from each other (Fig. 2, A and C). The reversal potentials of I inst(G404S) and I inst(HCN2) were significantly more positive than the reversal potential of I inst(GFP) but not significantly different from each other (Fig. 2D). This indicates that the expression of either wild-type HCN2 or G404S added a conductance with a reversal potential positive to that of the endogenous channels found in CHO cells expressing only GFP. The similar reversal potentials for I inst(G404S) and I inst(HCN2) suggests similar selectivities for the wild-type HCN2 and G404S channels. Like I inst(HCN2) , I inst(G404S) was significantly (p Ͻ 0.05) and reversibly reduced by cAMP elevation with perfusion of 100 M IBMX (3-isobutyl-1-methylxanthine), an inhibitor of phosphodiesterase that inhibits the breakdown of cAMP, and 100 M forskolin, an activator of adenylyl cyclase that increases cAMP formation (Ϫ931.2 Ϯ 139.8 pA, n ϭ 17; Ϫ858.8 Ϯ 127.5 pA, n ϭ 17; Ϫ1048.0 Ϯ 174.1 pA, n ϭ 15; before IBMX and forskolin, after 1 min in IBMX and forskolin, and after a 1-min wash, respectively). Finally, we found that I inst(G404S) was unaffected by 2 mM Cs ϩ , like I inst(HCN2) (11) but unlike I f , which is blocked by 2 mM Cs ϩ (26).
For confirmation of the presence of G404S on the plasma membrane, we examined subcellular localization in CHO cells of wild-type or mutant channels and labeled with an anti-HCN2 primary antibody and a Cy3-tagged secondary antibody. As shown by optical sections of cells expressing either wild-type HCN2 or G404S, there was a very strong pattern of fluorescence along the periphery of the cells suggesting the presence of HCN2 protein on the plasma membrane (Fig. 2E). Bright patches of fluorescence were observed in the interior of the cells, which probably represented channel protein present in discrete intracellular compartments such as the endoplasmic FIG. 1. A comparison of the selectivity filter from channels structurally related to HCN channels. Black shaded columns represent amino acid identity and gray shaded columns represent conserved amino acids. Asterisk represents the residue mutated in this study. Mouse HCN-gated 1, 2 channels (mHCN1 (NP_034538) and mHCN2 (NP_032252)) and human HCN4 (NP_005468) channel; sea urchin sperm channel (spHCN, CAA76493), a member of the HCN channel family; Arabidopsis, KAT1 (BAB11079), hyperpolarization-activated potassium channel found in plants; Drosophila melanogaster Shaker potassium channel (S00479); human ether-a-go-go (HERG, I38465) potassium channel; mouse inward rectifier potassium channel (Kir2.1, P35561); Streptomyces lividans, KcsA potassium channel (S60172); rat olfactory cyclic nucleotide-gated subtype a channel (olf-CNG␣, AAD41473); human voltage-gated potassium channel (Kv 2.1, Q14721). reticulum, Golgi apparatus, vesicles mediating transport to or from the plasma membrane, or degradatory compartments. There was relatively little fluorescence in non-transfected cells present in the image field of Fig. 2E, or in mock-transfected cells (not shown). We have shown previously that the wild-type channel and the S306Q mutant demonstrated a similar pattern of localization, whereas a mutant lacking the cyclic-nucleotide binding domain (CNBD) and distal C terminus did not express currents and demonstrated a very different pattern of localization that did not include fluorescence on the periphery of CHO cells (11). The imaging data and the significant increase in instantaneous current support the presence of G404S on the plasma membrane and the selective disruption of hyperpolarization-activated gating (I f ) by the G404S mutation.
Increasing the Ratio of K ϩ /Na ϩ Driving Forces Increases Fully Activated I f and the Rate of I f Deactivation-To further examine the role of the selectivity filter in voltage-dependent gating of HCN channels, we determined the effects of permeating cations on I f kinetics. Whole-cell current traces from cells expressing HCN2 are shown in Fig. 3A. We compared rates of deactivation and amplitudes of fully activated I f at the same potential but at different ratios of extracellular K ϩ /Na ϩ because both cations permeate HCN channels (24). We used a protocol consisting of a prepulse to Ϫ140 mV followed by a test pulse to Ϫ65 mV. These voltages were chosen for several reasons. First, the channels are close to fully activated at Ϫ140 mV and fully closed at Ϫ65 mV in CHO cells (11). Second, fully activated I f amplitude and the rate of I f deactivation were easily determined at Ϫ65 mV over the range of extracellular K ϩ and Na ϩ concentrations studied here. Finally, Ϫ65 mV is The pore mutant G404S eliminates I f and does not affect I inst . A, representative current traces elicited from a holding potential of Ϫ35 mV to Ϫ150 mV and followed by a pulse to ϩ5 mV to close the channels. Traces are shown for cells transfected with GFP alone, G404S and HCN2. Cells were perfused with a solution containing 135 mM K ϩ and 5.4 mM Na ϩ . I f density (B), I inst density (C), and I inst reversal potential (E r (I inst ), D) determined in cells expressing GFP, HCN2, and G404S. Values in B-D represent means Ϯ S.E. Single asterisks represent a significant difference from values determined in cells expressing only GFP. Double asterisks represent a significant difference from values determined in cells expressing HCN2. Numbers in parentheses represent the total number of cells in each group. Reversal potentials in D were determined using current traces elicited by 200-ms pulses to voltages ranging from ϩ5 mV to Ϫ150 mV, from a holding potential of Ϫ35 mV in the high K ϩ /low Na ϩ solution. E, confocal images of cells expressing G404S. A rabbit polyclonal antibody to HCN2 was used at a dilution of 1:600. This was visualized with a secondary anti-rabbit antibody tagged with Cy3. Note the intensity of fluorescence along the periphery of the cells (white arrows).
within the ranges of potentials observed normally in many excitable cells expressing I f (8,9).
Increasing the K ϩ /Na ϩ ratio produced reversible increases in the amplitude of I f at Ϫ140 mV and of the tail current at Ϫ65 mV (Fig. 3, A-C). Current activation for 2-s pulses at Ϫ140 mV and deactivation at Ϫ65 mV were both well fit with a single exponential function (Fig. 3A, inset) from which time constants were determined. Increasing the K ϩ /Na ϩ ratio did not significantly increase the rate of I f activation at Ϫ140 mV but sub-stantially increased the rate of I f deactivation at Ϫ65 mV (Fig.  3, A, D, and E). Deactivation also included a delay (Fig. 3A) that has been attributed to complex changes in conformation during channel closing in both native and cloned HCN channels (25,26). This delay is clearly smaller in the higher K ϩ /Na ϩ ratio as compared with the medium K ϩ /Na ϩ ratio and cannot be distinguished in the lowest K ϩ /Na ϩ ratio. Together, the increased deactivation rate and reduction in delay suggest that permeant cations affect I f deactivation preferentially.  a, b, c). Note the faster deactivation and shorter delay in deactivation in the high K ϩ /Na ϩ solution. Inset, tail currents (dotted lines) fitted with a single exponential function (solid lines), and tau values at each K ϩ /Na ϩ ratio; residuals (i.e. observed, fitted) of each fit are shown as solid lines above each tail current trace while the dashed lines represent the zero current level. Current amplitudes elicited by voltage steps to Ϫ140 mV (B) or Ϫ65 mV (C) at the three different ratios of K ϩ /Na ϩ . Bar graphs of time constants () of activation at Ϫ140 mV (D) and deactivation at Ϫ65 mV (E) at three different ratios of K ϩ /Na ϩ concentrations. Values represent absolute means Ϯ S.E.M., and the numbers in parentheses represent the total number of cells at each ratio.
The Dependence of I f and I inst Amplitudes on K ϩ and Na ϩ Driving Forces-We next compared how I f and I inst amplitudes depended on the changes in the K ϩ and Na ϩ driving forces, and how these compared with theoretical current values generated using the total current calculated in Equation 1, I tot ϭ I Na ϩ I K ϭ g Na,max (E test Ϫ E Na ) ϩ g K,max (E test Ϫ E K ) (Eq. 1) where I tot , I Na , and I K are the predicted current values for total I f , the Na ϩ component, and the K ϩ component; (E test Ϫ E K ), and (E test Ϫ E Na ) are the driving forces for Na ϩ and K ϩ ; and g Na,max and g K,max are the maximum values of conductance for Na ϩ and K ϩ . We assumed that: (i) each ion moved through the pore independently and (ii) g Na,max ϭ 1 ns, a value that was obtained using the above equation and the measured I f value at a ratio of 0 (when only Na ϩ flows through the channel). Fig. 4A (filled boxes) shows the fully activated I f as a function of the ratio of K ϩ and Na ϩ driving forces [(E test Ϫ E K )/(E test Ϫ E Na )] at Ϫ65 mV. An additional set of data at 10 mM K ϩ and 130 mM Na ϩ was added to determine more accurately the range over which the changes occurred. The greatest change in measured I f occurred between 5.4 mM K ϩ /135 mM Na ϩ and 10 mM K ϩ /130 mM Na ϩ where a slope of 677 pA/unit ratio of driving forces was determined (solid line, negative ratios). A smaller change occurred between 10 mM K ϩ /130 mM Na ϩ and 135 mM K ϩ /5.4 mM Na ϩ where a slope of 124 pA/unit ratio of driving forces was determined (solid line, positive ratios). The change in slope coincides with the change in direction of K ϩ flow (dotted vertical line represents a ratio of zero). Using g Na,max equal to g K,max , we found that the measured I f lay on I tot only at ratios where the K ϩ driving force was relatively small (Fig. 4A, lower dashed curve). Extrapolation of the measured I f to 0 pA (x-intercept) yielded a ratio of  Fig. 4. The solid black line represents a single exponential fit of values to determine the region of greatest change (y ϭ y o ϩ A 1 [1 Ϫ e Ϫx/k ], k ϭ 0.12). The theoretical current values for I tot were generated using the total current Equation 1 where g Na,max ϭ g K,max (lower dashed line) and g Na,max ϭ 0.25 g K,max (upper dashed line) and E test ϭ Ϫ140 mV. For A and B, values for fully activated I f are the same as in Fig. 3 but an additional group of cells was added for each of the variables at a K ϩ /Na ϩ ratio of 10 mM/130 mM to determine more precisely the range over which changes occurred.  Fig. 4. The theoretical current values for I tot were generated using the total current Equation 1 as described in the text and Fig. 4, where g Na,max ϭ g K,max and E test ϭ Ϫ140 mV (dashed line). For both C and D, I inst(corrected) represents instantaneous current that has been corrected for endogenous instantaneous current, as described under "Experimental Procedures." Filled squares in C and D are re-plotted from A and B. approximately Ϫ0.24. Thus, the K ϩ driving force in the outward direction needs to be only 0.24 of the Na ϩ driving force in the inward direction to eliminate I f . Therefore, a second theoretical relation for I tot was generated assuming that g Na,max was equal to 0.24 g K,max (Fig. 4A, upper dashed  curve). Using these unequal maximum conductances yielded a better fit to the data, suggesting that different permeabilities of each cation could account for the differences between the measured I f and theoretical current amplitudes.
Similarly, the fully activated I f at Ϫ140 mV was plotted as a function of the ratio of driving forces (Fig. 4B, filled boxes). Fitting with a single exponential function (solid black line) showed that the greatest change in the measured I f occurred between 5.4 mM K ϩ /135 mM Na ϩ and 10 mM K ϩ /130 mM Na ϩ, as it did at Ϫ65 mV. The theoretical values, generated assuming g Na,max ϭ g K,max (lower dashed curve) or g Na,max ϭ 0.24 g K,max (upper dashed curve), fit the experimental values even more poorly at Ϫ140 mV than at Ϫ65 mV. Extrapolation of the measured I f values yielded an x-intercept of 0.21 indicating that I f was eliminated despite the fact that driving forces were inward for both Na ϩ and K ϩ . This behavior has also been reported in studies of native I f (8,(27)(28)(29). The corresponding concentrations determined at the x-intercept were 3.3 mM K ϩ and 136.7 mM Na ϩ , which compares to 3.8 mM K ϩ and 136.3 mM Na ϩ determined at Ϫ65 mV (Fig. 4A). These concentrations of K ϩ and Na ϩ are in a range found in vivo, indicating that the control of external cation levels is a potent mechanism for regulating I f in different tissues.
To determine whether I inst could be predicted by independent flow of Na ϩ and K ϩ through the pore, we plotted the corrected I inst (corrected for removal of endogenous leak currents as described under "Experimental Procedures") versus the ratio of driving forces. This was compared with theoretical current values where g Na,max ϭ g K ,max at Ϫ65 mV (Fig. 4C) and Ϫ140 mV (Fig. 4D), as was carried out for I f . A value for g Na,max was determined using the corrected value for I inst where the ratio of driving forces ϭ 0 (when only Na ϩ flows through the channel) for I tot . The measured values for I inst were much closer to the theoretical current values than were the values of fully activated I f (filled squares replotted from Fig. 4, A and B). These data suggest that hyperpolarization produces conformational changes that enhance the flow of K ϩ versus Na ϩ through the HCN pore in the fully activated open state, leading to greater than predicted values for the measured I f amplitudes. Alternatively, these data could be explained by a model where the two currents (I f and I inst ) flow through two different pores, both permeable to Na ϩ and K ϩ but where each pore has a distinct gating mechanism.
Parallel Dependence of Fully Activated I f and I f Deactivation on Driving Forces for K ϩ /Na ϩ in the Physiological Range-We next re-examined I f deactivation at Ϫ65 mV to determine whether this was modified by permeating cations in the same range of driving forces that modify I f amplitude. Here, we included the set of data points at 10 mM K ϩ /130 mM Na ϩ for both I f amplitude and I f deactivation (Fig. 5). We found that the increase in rate of I f deactivation (filled squares) paralleled the increase in fully activated I f (filled diamonds) at increasing K ϩ /Na ϩ driving forces. Both sets of values changed most dramatically in the range of driving forces normally found in cells and where K ϩ moves in the outward direction. As for the fully activated I f at Ϫ65 mV, deactivation time constants were joined with two straight lines on either side of the K ϩ reversal concentration. The most dramatic change occurred between 5.4 mM K ϩ /135 mM Na ϩ and 10 mM K ϩ /130 mM Na ϩ where a straight line of slope ϭ Ϫ4647 ms/unit ratio of driving forces joined these values. A less dramatic change occurred between 10 mM K ϩ /130 mM Na ϩ and 135 mM K ϩ /5.4 mM Na ϩ , where a straight line of slope ϭ Ϫ182 ms/unit ratio of driving forces joined the values. When both sets of values were normalized to the maximum values obtained, the slopes for the steep component were 2.2 and Ϫ4.0 and for the shallow component were 0.4 and Ϫ0.16, for the fully activated I f and deactivation time constants, respectively. The parallel changes in both sets of values support the idea that permeating cations affect I f deactivation as well as I f amplitude. They also suggest that the conformational changes induced by hyperpolarization occur relatively slowly because activation, unlike deactivation, was not greatly affected by changes in K ϩ /Na ϩ driving forces. Thus, after the hyperpolarizing pulse, changes in pore conformation then lead to changes in the rate of I f deactivation, in the relative conductance of K ϩ versus Na ϩ , and in the fully activated I f amplitude.

Separable Gating Mechanisms in Pacemaker Channels-Our
present findings show that in the HCN2 G404S mutant, a reduction in external K ϩ or application of Cs ϩ reduce or eliminate I f without blocking I inst . Together with our previous work demonstrating the co-existence of a slow Csϩ-sensitive current (I f ) and an instantaneous Csϩ-insensitive current (I inst ) (11), these data suggest that two gating mechanisms, which can be separated functionally and structurally, are associated with HCN channels.
Is it possible for two currents with different properties and structural requirements to travel through the same pore? The inhibition of I f by Cs ϩ and G404S may be analogous to the production of distinct subconductance states in single channel currents of K ir 2.1 by the permeant cation Tl ϩ and by mutations of the glycine corresponding to G404 in HCN2 (16,18). Certain closed-time states in this inward rectifier were inde-  Fig.  3 but an additional group of cells was added for each of the variables at a K ϩ /Na ϩ ratio of 10 mM/130 mM to determine more precisely the range over which changes occurred. Statistically significant differences were found for fully activated current amplitudes and time constants of deactivation at ratios of K ϩ /Na ϩ of 5.4/135 versus 10/130, using a Student's t test (p Ͻ 0.05). pendent of membrane voltage and were not produced by openpore blockade by external or internal cations, suggesting the co-existence of a separate gating mechanism (18,30). Voltageindependent closed to open fluctuations and subconductance levels have also been observed in Shaker K ϩ channels (31,32). By analogy, I inst may thus represent voltage-independent subconductance states and/or closed to open transitions, which are not inhibited by Cs ϩ , G404S, or a decrease in external potassium. In this model, hyperpolarization would then produce additional subconductance states that are sensitive to Cs ϩ and to driving forces for permeating cations. Measurement of single channel currents will be required to examine this hypothesis in HCN channels, although this may be challenging given the small (ϳ1 pS) single channel conductance determined in studies of native HCN channels (33).
Other voltage-gated K ϩ channels, such as the cloned plant K ϩ channels, AKT2 and AKT3 (34,35), exhibit combinations of instantaneous leak and slowly-activating inward rectification. Single channel and whole-cell current analysis suggested that the AKT2 channels exist in two modes whose relative amounts may change over time (36). The co-existence of instantaneous, voltage-independent leak with voltage-dependent, slowly activating rectification has also been described in the KCNK2 K ϩ channel (37). The SLO-2 calcium-activated K ϩ channel has both a time-dependent outward current component and a rapidly activating current component (38). Finally, Liu and Joho (39) have proposed the co-existence of slow voltage-dependent and rapid voltage-independent gating mechanisms in Kv2.1. These authors suggested that there are two gates within the pore of Kv2.1 that are close, but physically separated. Given the conserved nature of the pore structure, voltage-independent leak through a separate gating mechanism may be an inherent property of channels in these related families. However, it is also possible that the co-existence of instantaneous, slowly activating currents is due to a second pore that is found within the same channel or formed by a second population of the same channel subunits or associated with the channels in the form of up-regulated endogenous channels.
Coupling between the Selectivity Filter and I f Deactivation-Permeation and gating were originally thought to be independent in ion channels but more recent evidence indicates that permeant ions affect gating and hence that these properties are coupled. Previous studies of I f in sensory neurons, apical dendrites of hippocampal CA1 pyramidal neurons, and myocytes from the sinoatrial node have shown that deactivation was affected by relatively large changes in external cations (28,40,41). Despite concurrent changes in I f amplitude that reflected interactions of Na ϩ and K ϩ in the pore, a separate mechanism was proposed for the effects of cations on I f deactivation that involve an external binding site (28). The primary argument in favor of an external binding site was that only K ϩ could be present in the pore during measurements of outward tail currents because Na ϩ was not present in the intracellular solution and thus would have left the pore almost instantly upon depolarization. Our data suggest the effects of permeating cations do involve interactions in the pore because the rate of I f deactivation and I f amplitude depended on K ϩ /Na ϩ driving forces in a parallel manner. Our findings also suggest that the changes in I f deactivation and I f amplitude follow from hyperpolarization-induced changes in pore conformation that occur slowly during I f activation. These changes are related to the ability of K ϩ to permeate the channel more readily than Na ϩ after hyperpolarization as well as to the direction of K ϩ flow. Thus, we suggest that the selectivity filter is an important structural element of voltage-dependent gating and is coupled to cation selectivity in HCN channels, as it is in related potassium channels (42).
Physiological Relevance of Two Gating Mechanisms and Modulation by Permeating Ions in Pacemaker Channels-A physiologically important aspect of our experiments is that the changes in the concentration of K ϩ and Na ϩ that influence both I f deactivation and I f amplitude are in a range that exist under certain physiological and pathophysiological conditions and thus could provide a mechanism for controlling cellular excitability in vivo. In the heart, changes in external K ϩ between 3 and 10 mM may occur in situations such as exercise and can produce bradycardia mediated through I f (43)(44)(45). In the central nervous system, extracellular K ϩ may vary from 3 to 12 mM, and extracellular Na ϩ may decrease by up to 7 mM during repetitive activity (46,47). These changes in cation concentration could significantly affect the spontaneous activity of neurons that contain HCN channels, such as thalamic relay neurons, in which I f deactivation and amplitude play critical roles in determining the period of time separating periods of sustained activity (10).
The ability to balance I inst versus I f would greatly expand the role of HCN channels in modulating the firing characteristics of neurons and cardiomyocytes. A dramatic reduction of I f compared with I inst may explain why some neurons express HCN mRNA but do not express I f (4). In the sino-atrial node of the heart, I inst may contribute to the Na ϩ -sensitive background current (I b,Na ) with which it shares several properties (48). The relative effects of I b,Na and I f are important for regulating the pacemaking activity of the sino-atrial node (49), and a similar balance of these current components may be important in neurons.